ADVANCES IN CATALYSIS VOLUME 7, Volume 7

ADVANCES IN CATALYSIS AND RELATED SUBJECTS

VOLUME VII

This Page Intentionally Left Blank

ADVANCES IN CATALYSIS AND RELATED SUBJECTS VOLUME VII EDITED BY

W. G. FRANKENBURG

V. I. KOMAREWSKY

Lancaster, Pa.

Chicago, Ill.

E. K. RIDEAL London, England

ADVISORY BOARD

PETERJ. DEBYE Zthaca, N.Y.

D. D. ELEY

P. H. EMMETT

Nottingham, England

M. G. EVANS Manchester, England

W. E. GARNER Bristol, England

P. S. SELWOOD Evanston, Ill.

Pittsburgh, Pa.

W. JOST Goettingen, Germany

H. S. TAYLOR Princeton, N.J.

1955

ACADEMIC PRESS INC., PUBLISHERS NEW YORK, N.Y.

COPYRIQHT 1955, BY ACADEMIC PRESS INC.

125 East 23rd Street, New York 10, N.Y. All Rights Reserved No part of this book may be reproduced in any form, by photostat, microfilm, or any other means without written permission from the publishers. Library of Congress Catalog Card Number: 49-7755

PRINTED IN THE UNITED STATES OF AMERICA

CONTRIBUTORS TO VOLUME VII

M. McD. BAKER,Department of Physical Chemistry, King's College, London, England

JOSEPHA. BECKER,Bell Telephone Laboratories, Inc., Murray Hill, New Jersey

M. BOUDART, Forrestal Research Center, Princeton University, Princeton, New Jersey

E. CREMER, Physik.-Chem. Institut der Universitat, Innsbruck, Austria ROBERTCOMER,Institute for the Study of Metals, The University of Chicago, Chicago, Illinois

K. HAUFFE,Department of Solid State Physics, Central Institute for Industrial Research, Blindern, Oslo, Norway

G. I. JENKINS,Department of Physical Chemistry, King's College, London, England *S. ROYMORRISON, Department of Electrical Engineering, University of Illinois

G. PARRAVANO, Forrestal Research Center, Princeton University, Princeton, New Jersey

R. SUHRMANN, Institut fiir Physikalische Chemie und Elektrochemie der Technischen Hochschule Braunschweig, Germany

* Present address: Sylvania Electric Products, Physics Laboratories, Bayside, New York.

V

This Page Intentionally Left Blank

NICOLAIDMITRIEVICH ZELINSKY 1861-1953

It was my very good fortune to study and work with two great chemists. I n succession it is now my honorable but sad duty to write their obituaries. One of them was well known to us; the Willard Gibbs medalist, Vladimir Nicholaevich Ipatieff, who passed away in November, 1952. The other is Nicolai Dmitrievich Zelinsky, whose death at the age of ninety-three was announced on July 31st, 1953 in Moscow. The first scientific publication of Zelinsky’s was presented in 1884, the last one in 1953, which makes it nearly seventy years of productive research work. This certainly must represent some kind of a record. During his long and successful scientific career, Zelinsky published more than five hundred papers mostly devoted to catalysis, and raised to scientific maturity scores of successful pupils who became famous on their own accord. Among the numerous collaborators and pupils of Zelinsky such names as Chugaev, Schilov, Reformatsky, Nametkin, Chelinzev, Stadnikoff, Lebedev and Balandin, t o mention only a few, speak for themselves. I n his famous book (‘Grosse Manner,” Ostwald separates famous scientists in two classes. One is “Romanticists”; the representatives of this type are brilliant researchers who study a certain problem with vigor and energy, and who very quickly make valuable discoveries and immediately pass to another problem, sometimes of an entirely different nature, leaving the details of the first to other investigators. The “classicists” after making an important discovery, continue t o investigate the problem systematically and carefully. It is interesting t o note that both of these great men belonged to the two different classes. Ipatieff was a typical “Romanticist,” Zelinsky a. typical “Classicist.” It might also be of interest that their major contributions were in the study of two opposite reactions; Ipatieff-the reaction of hydrogenation. Zelinsky-the reaction of dehydrogenation. After graduation from the Novorossiisk University in Odessa, Zelinsky was sent by the faculty t o Germany where be worked and studied under Victor Meyer (Gottingen) and Wilhelm Ostwald (Leipzig). While in the laboratory of Victor Meyer, Zelinsky was the first to prepare the mustard gas (/3-/3’-dichlorodiethylsulfide) and was the first victim of its toxic properties. vii

viii

NICOLAI DMITRIEVICH ZELINSKY

The first period of scientific work upon returning to Odessa was devoted to study of stereoisomerism of dibasic carboxylic acids, resulting in the discovery of the known Hell-Volgard-Zelinsky method of a-bromination of fatty acids. The major work of Zelinsky was devoted t o petroleum and catalysis. This work was started in the laboratories of the University of Moscow where he moved in the year 1893 as associate professor and where, with the exception of a six year interval, carried out all his research until his death. The classical studies of Zelinsky in dehydrogenation catalysis started with the discovery of platinum (and later of palladium and nickelalumina) catalyst in the dehydrogenation of cyclohexane, its homologues and derivatives, to aromatics. This work was the foundation of a systematic study of hydrocarbon reactions in the presence of metal catalysts, belonging to the 8th group of the periodic system, which resulted in the discovery of decyclohydrogenation of cyclopentane and its homologues t o paraffins; hydrogen disproportionation of cyloolefins and cyclodiolefins t o aromatics and cycloparaffins, and cyclization of paraffins. All these reactions served as excellent tools in the study of the nature of petroleum hydrocarbons and became of tremendous importance in aromatization of petroleum. This work laid the foundation for the modern processes of catalytic reforming of napthas.. One cannot forget the leading role of Zelinsky in development of the gas mask, which saved many thousands of lives of Russian and Allied soldiers during the first World War. As a matter of fact, the Zelinsky mask (Activated charcoal) was so good that it was accepted by the Germans and masks with the sign “Nach Zelinsky” could be found on the German soldiers. It will also be important t o mention the successful work of Zelinsky in the field of proteins. This method of hydrolysis of proteins with dilute acids under pressure resulted in the production of clean hydrolyzates and helped the development of the diketopiperasine and polypeptide theory of protein structure. Professor Zelinsky’s devotion to his beloved science was unique. His work was his life and hobby and his home was in his laboratory. He had the satisfaction t o see the results of much of his research work be transferred to industrial practice. He received numerous decorations and prizes for the Russian Government and was elected an honorary member of the Chemical Society (London). With the passing away of Professor Zelinsky, we are losing another pioneer and great Master of catalysis. V. I. Komarewsky

CONTENTS CONTRIBUTORS TO VOLUME VII . . . . . . . . . . . . . . . . . . . . . . v OBITUARY OF NICOLAI DMITRIEVICH ZELINSKY. . . . . . . . . . . . . . . vii

The Electronic Factor in Heterogeneous CataIysis

BY M . McD . BAKERA N D G. I . JENKINS, Department of Physical Chemistry. King’s College. London. England I . Introduction . . . . . . . . . . . . . . . . . I1. Theories of the Solid State . . . . . . . . . . . I11. The Nature of the Chemisorption Bond . . . . IV . Catalytic Activity of Transition Metals . . . .

. . . . . . . . . .

1

. . . . . . . . . .

2

. . . . . . . . . . . 12 . . . . . . . . . . . 20 V . Catalytic Activity of Alloys . . . . . . . . . . . . . . . . . . . . . 24 VI . Semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . 30 VII . Insulators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 41 VIII . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Chemisorption and Catalysis on Oxide Semiconductors

BY G . PARRAVANO A N D M . BOUDART, Forrestal Research Center, Princeton University, Princeton, New Jersey I . Chemisorption of Hydrogen on Zinc Oxide . . . . . . . . . . . . . . 50 I1. Electronic Properties of Zinc Oxide . . . . . . . . . . . . . . . . . 52 I11. The Hydrogen-Deuterium Exchange on Defect Zinc Oxide . . . . . . . 56

IV . V VI . VII .

Adsorption of Oxygen and Oxidation Catalysis on Nickel Oxide . . . . .

. Electronic Properties of Nickel Oxide

60

. . . . . . . . . . . . . . . . 66 Carbon Monoxide Oxidation on Modified Nickel Oxide Catalysts . . . . 68 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .

72 72

The Compensation Effect in Heterogeneous Catalysis BY E . CREMER,Physik.-Chern. Znstitut der Universitat, Znnsbruck, Austria

I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Experimental Data . . . . . . . . . . . . . . . . . . . . . . . I11. The Interpretation of the Compensation Effect . . . . . . . . . . .

75 76

. 80 IV . Apparent Compensation Effects . . . . . . . . . . . . . . . . . . . 88 V . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . 90 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

Field Emission Microscopy and Some Applications to Catalysis and Chemisorption

BY ROBERT GOMER,Institute for the S t u d y of Metals, T h e University of Chicago, Chicago, Illinois

I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 1I.Theoi-y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

93 94

X

CONTENTS

111. Selected Applications of Field and Ion Microscopy. . . . . . . . . . . 111 Appendix: Experimental Methods and Techniques. . . . . . . . . . . 129 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Adsorption on Metal Surfaces and Its Bearing on Catalysis

BY J O S E P H A. BECKER,Bell Telephone LaboTatOTieS, Inc., Murray Hill, New Jersey

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. The Importance of Adsorption for the Theory of Surface Catalysis . . . 111. The Adsorption of Cesium on Tungsten Using Thermionic Emission. . . IV. The Adsorption of Nitrogen on Tungsten as Deduced from Ion Gauge and Flash Filament Techniques. . . . . . . . . , , . . . , . . . . . . V. The Adsorption of Oxygen on Tungsten as Observed in the Field Emission Microscope. . . . . . . . . . . . . . . , . . . , . . . , , . . . VI. Other Adsorption Experiments with the Field Emission Microscope . . . VII. Discussion of Previous Work on Adsorption, . . . . . . . . . . . . , References . . . . . . . . . . . . . . . . . . . . . , , , , . . .

136 138 141 159 175 192 200 210

The Application of the Theory of Semiconductors to Problems of Heterogeneous Catalysis

BY K. HAUFFE,Department of Solid Stale Physics, Central Institute for Industrial Research, Blindern, Oslo, Norway

I. Introduction . . . . . . . . . . . . . . . . . . , . . . . . . . . 11. The Mechanism of the Reaction of Gases with Semiconducting Crystals , 111. The Mechanisms of Simple Reactions on the Surface of Semiconducting Solids,. . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .

213 216 236 255

Surface Barrier Effects in Adsorption, Illustrated by Zinc Oxide

By S. ROY MORRISON,Department of Electrical Engineering, University of Illinois

I. Introduction . . . . . . . . . . . , , , . . , . . . . . . . . . . 259 11. Electronic Structure and Electron Transfer at Surfaces. . . . . . . . . 261 111. Basic Properties of Zinc Oxide . , . . . . . . . . . . . . . . . . . 266 IV. Irreversible Adsorption and the Electronic State of the Surface . . . . . 272 V. The Relationship of the Photoconductivity to the State of the Surface . . 294 VI. The Relationship of the Fluorescence to the State of the Surface . . . . 298 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300 Electronic Interaction between Metallic Catalysts and Chemisorbed Molecules

BY R. SUHRMANN, Institut f i i r Physikalische Chemie und Elektrochemie der Technischen Hochschule Braunschweig, German9

I. Introduction , . . . . . . . . . . . 11. The Nature of Electronic Interaction , 111. Experimental Methods, . . . , . . . IV. Results of Experiments. . . . . . . . V. Theoretical Considerations . . . . . .

. *. . . . . . . . . . . . . . . . . . .

.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.

. . . . . . . . . . . .

303 305 306 . . . . 320 . . . . 347

CONTENTS

xi

VI . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .

349 350

AUTHORINDEX. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

353

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . SUBJECT INDEX

359

ERRATUM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

363

This Page Intentionally Left Blank

The Electronic Factor in Heterogeneous Catalysis M. McD. BAKER

AND

G. I. J E NK IN S

Department of Physical Chemistry, King’s College, London, England Page

I. Introduction. . . . . . . . . . ....... ........ 11. Theories of the Solid St ..................................... 1. Quantum Mechanical Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a. Alloys., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b. Transition Metals.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . c. Transition Metal Alloys. . 2. Resonating Valence Bond Theory.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. The Nature of the Chemisorption Bond. IV. Catalytic Activity of Transi V. Catalytic Activity of Alloys VI. Semiconductors. . . . . . . . . . . 1. Theory . . . . . . . . . . . . . . . . a. Intrinsic Semiconductors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b. Extrinsic Semiconductors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII. VIII.

..............................................

2 6 8

30 30

41

I. INTRODUCTION The theoretical approach t o the subject of surface catalysis was first considered in a series of classical papers by Langmuir (l), who suggested that the adsorbed particles are held to the surface by chemical forces, and applied the theory t o interaction of adsorbed species a t adjacent adsorption sites on the surface. Langmuir pointed out that steric hindrance effects between molecules might play a prominent part, and the role of the geometric factor in catalysis was greatly emphasized by Balandin and others. The importance of this factor has already been reviewed in this series by Trapnell (2) and Griffiths (3). As early as 1928, Roginskii and Schul’tz (4) stressed the importance of electronic considerations, and Rideal and Wansbrough-Jones (5) related the work function of metals t o the activation energy for their oxidation. Brewer, 1928 (6), Schmidt, 1933 (7), and Nyrop, 1935 (8) proposed that the surface must be capable of effecting ionization of the adsorbed species in some catalytic processes. Lennard-Jones (9) in his 1

2

M. MCD. B A K E R AND G. I. J E N K I N S

theory of chemisorption on metal surfaces considered the possibility of covalent bond formation due t o the pairing of a valence electron of the adsorbed atom and an electron from the conduction band of the metal. Further developments in the fundamental approach t o the electronic structure of catalysts were made possible by the development of the quantum mechanical treatments of solids which followed the work of Sommerfeld, Bloch and others. Similarly, Pauling’s resonating valence bond treatment has lent further impetus t o consideration of metallic catalysts. 11. THEORIES OF THE SOLIDSTATE 1. Quantum Mechanical Theory

The molecular orbital treatment of a crystalline solid considers the outer electrons as belonging to the crystal as a whole (10,ll). Sommerfeld’s early free electron theory of metals neglected the field resulting

FIG.1. Momentum diagram.

from the array of metallic nuclei. A great advance was made by Bloch and others who assumed that the electrons move in a field whose periodicity is that of the crystal lattice. The possible energy states for the electrons can be obtained from the solutions of the Schrodinger equation 2m v2++ p (E -

U)+ = 0

(1)

I n the consideration of the momentum of a large number of particles restricted t o a volume V , it is often convenient t o describe the system by an assembly of points in a momentum diagram (Fig. 1). T h e length OA represents the magnitude of momentum of the particle A , and its direction --t

is O A . The application of Heisenberg’s uncertainty principle leads to the

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

3

restriction that the momentum of any particle cannot be represented by a point, but must be associated with a volume h3/V (h is Planck’s constant). This means effectively that momentum space must be divided into cells of volume h3/V,each cell corresponding to an energy state, and from. Pauli’s Principle each of these states can contain two electrons. Furthermore, the energy of the assembly will be a minimum when the occupied states are clustered about the origin. The wavelength of a free electron is given by 1/X = mu/h. Since the wave number k = 2?r/X, it is obvious that a k-diagram may be constructed in a manner similar to that ‘kY

t

FIG.2. Brillouin zone for a two dimensional crystal.

used for the momentum diagram. The boundary between occupied and unoccupied states in k-space is referred to as the Fermi surface. When the periodic field of the lattice is taken into account, the wave number is no longer simply related to the momentum (mu). The wave mechanical examination of the equation for the motion of an electron in a periodic field shows that the energy does not vary continuously with the wave number. This means that in any one direction in k-space, there will arise energy discontinuities as the states are filled. These regions can be represented in the k-diagram by surfaces; the regions bounded by such surfaces are usually termed Brillouin zones. Figure 2 represents a Brillouin zone for a two-dimensional crystal; in the figure, the Fermi surface is circular and the states are not completely occupied by electrons up to the boundary of the Brillouin zone. A typical energy distribution curve is represented by Fig. 3. An elec-

4

M. MCD. BAKER A N D G . I . JENKINS

tron state, value lc, in k-space, will have an energy E,. The number of states N ( E ) in the energy band dE, is rzlated to the volume of the spherical shell, and it can he seen that N ( E ) in a range dE, will increase with E,. The maximum a t E zoccurs a t the point where the Fermi surface first touches the Brillouin zone. A rapid decrease of N ( E ) then occurs because there are no energy states available for some directions of the wave number k . If the first and second Brillouin zones do not overlap, on completion of the first band, the next available electron state will involve a n increase in energy of (Eq - Eo). If the first and second zones overlap, the addition of electrons will fill the first zone t o the point where the

Energy,E

-

FIG.3. Typical electron bands.

highest filled states of the first zone have the same energy as the lowest filled states of the second zone. More electrons will then enter the states of both zones. The resulting N ( E ) curve has the form shown in Fig. 4. The effect of an external electric field is to produce a n acceleration of the electrons in the direction of the field, and this causes a shift of the Fermi surface. It is a necessary condition for the movement of electrons in the k-space that there are allowed empty states a t the Fermi surface; hence electrical conductivity is dependent on partially filled bands. An insulating crystal is one in which the electron bands are either completely full or completely empty. If the energy gap between a completely filled band and a n empty band is small, it is possible that thermal excitation of electrons from the filled t o the empty band will result in a conducting crystal. Such substances are usually referred t o as intrinsic semiconductors. A much larger class of semiconductors arises from impurities

ELECTRONIC FACTOR IN HETEROGENEOUS CATALYSIS

5

in the crystal. The introduction of foreign atoms generally creates additional energy levels, and these impurity levels may lie within the energy gap, as shown in Fig. 3. At temperatures above absolute zero there are two possible processes which may occur: either electrons may be thermally excited from the impurity levels to the empty band, or electrons from the filled band may be excited into the impurity levels. I n both cases partly occupied bands are produced, and the crystal becomes conducting. (a) Alloys. I n the formation of simple alloys, the progressive substitution of one type of metal atom by another type with different valency

E+

FIG.4. Overlapping electron bands. changes the ratio of valence electrons to metal atoms of the crystal. This ratio has been termed the electron concentration. The Fermi level of the solvent metal will be either raised or lowered depending on whether the solute metal has a higher or lower valency. Many properties of alloys can be related to the electron concentration. Hume-Rothery (12,13) has pointed out that in some alloys the structure of the intermetallic phases are determined by the electron concentration (E.C.). The work of Hume-Rothery and others has shown that the series of changes (i.e. CY phase + 0 phase + y phase -+ e phase), which occurs as the composition of an alloy is varied continuously, takes place at electron-atom ratios of 3/2, 21/13, and 7/4, respectively. The interpretation of these changes in terms of the Brillouin zone theory has been made by H. Jones (14) and can be understood from the N(E)-curves for typical face centered cubic (a) and body centered cubic (b) structures as

6

M. MCD. B A K E R AND G. I. J E N K I N S

shown in Fig. 5. The alloying of univalent metals, such as copper, silver, and gold, with higher valence elements is considered t o be a process in which the solute elements merely serve to increase the electron concentration. The higher the N(E)-curve, the greater is the number of electrons which may be accommodated in states within the energy range concerned, and hence the lower will be the energy necessary to accommodate a given number of electrons. I n Fig. 5 it can be seen th a t up to the point A the face centered cubic structure has the lower energy and will be the stable phase of the alloy. Beyond point A the number of states decreases rapidly, and addition of more electrons into the band causes a rapid increase in energy. At some point, the body centered cubic structure

E-

FIG.5. Interpretation of Hume-Rothery alloys in terms of Brillouin zone theory.

becomes more stable than the face centered cubic structure, since the electron band in the former can accommodate the same number of electrons with a lower energy. H. Jones has calculated the electron-atom ratios corresponding to the maximum N(E)-peaks for each of the four phases, and has obtained the values a = 1.362, p = 1.480, y = 1.538. These values for the a-,p-, and y-phases agree quite well with the HumeRothery values for the corresponding phase changes. (b) Transition Metals. The transition metals are characterized by ferromagnetism or strong paramagnetism and by their comparatively low electrical conductivity. According t o Mott and Jones (lo), the outermost electrons are considered to occupy two bands. I n the first long period of the periodic table these bands arise from the 3d and 4s atomic states. The broadening of the 3d-band is much less marked than that of the

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

7

s-band, since the smaller 3d-orbitals overlap to a lesser extent. The broadening process can be seen from Fig. 6(a). The density of states is five per atom in the d-band and is one per atom in the s-band. The actual N(E)-curves for some transitional elements may be complex, but it is more convenient in this discussion to represent the s- and d-bands in their general form (Fig. 6(b)). Considering nickel as a typical example, the ten outermost electrons per atom must be shared between the two bands. There is strong evidence that there are 9.4 electrons per atom in the 3d-band and 0.6 id the 4s-band. The electron bands of copper will not

A

t”.

E

fB I

I

FIG.6. Electron bands of transition metals.

differ appreciably in energy from those of nickel, but the addition of another electron will completely fill the d-band and half fill the s-band, in agreement with its monovalent nature. From this point of view, the ferromagnetism of iron, cobalt, and nickel is almost entirely due to the spins of the d-electrons and the saturation magnetization intensity expressed in Bohr magnetons per atom is approximately equal to the number of “holes” in the d-band (Fe, 2.2; Co, 1.7; Nil 0.6). The noninteger number of electrons in the d-band can be understood if they are considered as average values for a crystal containing the configurations ( 3 4 lo, (3d)9, and (3d)8. Mott and Jones consider that only the (3d)’O and ( 3 ~ i con)~ figurations are important, the two holes in the shell of the (3d)8-ion having the same spin. The strong paramagnetism of platinum and palladium, which is probably also due to electron spin, and the physical

8

M. MCD. BAKER AND G . I . JENKINS

properties of their alloys can be explained by this theory if there are about 0.55-0.6 holes in the d-band. (c) Transition Metal Alloys. The theoretical consideration outlined above readily explains the changes in properties accompanying alloy formation of transition metals with other metals. If we take as an example the Cu-Ni alloys, the replacement of nickel atoms by copper atoms will add extra electrons to the lattice. Since the density of states in the d-band is much greater than in the s-band, the added electrons will enter the d-band until it is filled, As the atomic percentage of copper in the alloy increases, the saturation magnetic moment decreases; extrapolation to zero magnetic moment shows that the d-band is just full at 60%. The addition of an atom of palladium, zinc, or aluminium will increase the number of electrons in the d-band by 0, +2, or +3 respectively. Similarly, for the palladium and platinum alloys with copper, silver, and gold, the atomic susceptibility drops as the percentage of solute increases, ‘ the susceptibility becoming negligible a t a concentration of about 55% of the monovalent metal (10,ll). This confirms the assumed number of &“holes” in these two metals. The solution of hydrogen in palladium must be considered as a different type of alloy system, because the hydrogen cannot replace palladium in the lattice structure. The fall in susceptibility to zero at a hydrogen-palladium ratio of 0.55 suggests that all the electrons of the hydrogen atoms enter the 4d palladium levels (11).

9. Resonating Valence Bond Theory Pauling (15-17), in an approach more easily described in language familiar to chemists, considers that metal bonds resemble ordinary covalent bonds, all or most of the outer electrons of the metal atoms taking part in bond formation. He suggests that when the number of possible positions at which bonding can occur is greater than the number of bonds, resonance occurs about the available positions. This approach explains satisfactorily the decrease in bond length with increasing valency in the sequences of elements K -+ Cr, Rb -+ Mo, and Co + W. The number of electrons available for bond formation in these metals increases with valency, and theref ore, according to the resonating valence bond theory, the number of bonds resonating about the available positions increases, resulting in smaller bond lengths. The magnetic properties of these elements also fit into this concept. I n metallic vanadium for example (atomic configuration 3d34s2),if the 4s2 electrons only were involved in bond formation, the cores of the atoms would have a large dipole moment with the result that ferromagnetism or strong paramagnetism decreasing with increasing temperature should

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

9

be exhibited. Contrary to this, vanadium is only weakly paramagnetic, the paramagnetic moment being almost temperature independent. This indicates that all its ou$er electrons are involved in bond formation. Similarly, all elements in the sequences mentioned above are either diamagnetic or weakly paramagnetic. Lithium may serve as an example for illustrating the general picture of resonance structures in metals. Pauling points out that it is unlikely that synchronized resonance, i.e. resonance of two bonds simultaneously, of the type Li-Li Li Li

I

Li-Li

Li

I

Li

could give the stabilization necessary to explain the strength of metallic bonding. He shows that, a much greater resonance stabilization can be achieved by resonance of one bond involving electron transfer. Li Li@

I

eLi-Li According to this concept, the metallic properties are based on the possession by some or all of the atoms in a given metal of a free orbital (the “metallic orbital”), in addition to the orbitals required for bonding and nonbonding electrons, thus permitting uninhibited resonance of valence bonds. For the case of tin, the following scheme illustrates these relationships for three electronic structures ( A , B,and C) of this metal: Sn ( A ) v = 4 Kr

atomic electrons valence electrons i.e. electrons occupying bonding orbitals 0 metallic orbitals 0

( B )v

=

2 Core

(C) v

=

0

where v is the valency of the metal atom in each electronic structure. I n Sn ( A ) there is no metallic orbital and therefore no metallic phase; the four electrons are hybridized to give spa tetrahedral bond orbitals. The nonmetallic form of tin, gray tin, does in fact have the diamond tetrahedral structure. Pauling suggests that the metallic form of tin (valency 2.44) consists of Sn ( B ) resonating with Sn ( A ) ,the Sn ( B ) configuration having the larger statistical weight. The loss in energy due to a decreasing number of bonds is largely compensated by the resonance energy gained. Although the number of electrons theoretically available for bond formation increases with increasing atomic number in the series, Cr, Mn,

10

M. MCD. BAKER AND G. I. JENKINS

Fe, Co, Nil the bond length in these metals remains practically the same, indicating about six bonds per atom for all five elements. Pauling suggests, as an additional hypothesis to account for this experimental data, that while some 3d-orbitals may hybridize with 4s- and 4p-orbitals to . give bonding orbitals, other 3d-orbitals may be unsuitable for bond formation (atomic orbitals). The ferromagnetic saturation moment of iron, cobalt, and nickel can be attributed to unpaired, nonbonding, electrons in these atomic orbitals. By accepting this hypothesis, the following pictorial representation of the electronic structure of transition metals in the first long period of the periodic table is derived. Iron has a magnetic saturation moment of 2.22 Bohr magnetons. Therefore, of the eight electrons outside the argon core, 5.78 are bonding electrons and 2.22 are unshared electrons in nonbonding orbitals. Iron: Eight electrons outside the argon core. 3d Fe ( A ) v

=

4s

=

Magnetic saturation moment 2 Bohr magnetons M.S.M. 3 Bohr magnetons

6 ‘

Fe ( B ) v

4p

I

5

Both possible valence states contribute in the ratio 78 :22, giving iron the necessary magnetic saturation moment of 2.22 magnetons. This concept leads to a valency of 5.78. Cobalt: Nine electrons outside the argon core. 3d ,-

CO ( A ) v = 6 : A TT Co(B)v=6lcoreiTJr

T

-

4s i

j.*

*

I

;

4p *

, A

The saturation moment of 1.71 magnetons indicates resonance between the two forms of electronic states in the ratio 35:65. Niclcel: Ten electrons outside the argon core. 3d -.Lvv4

Ni(A)v=6; A /TJT Ni(B)v=6jcoreiTJ’TL

-4s-,-.4 p

T-.! . - #

,

~

I

j

,

*

j

.

a

0

The saturation moment of 0.6 magnetons corresponds to resonance in the ratio 30:70. Developing this concept for elements following the transition metals, Pauling employs a system based on two observations in his paper (16): (1) A linear relationship holds between the “single bond’’ radius and

11

ELECTRONIC FACTOR IN HETEROGENEOUS CATALYSIS

atomic number for bonds of constant hybrid character. ( 2 ) For an element, the single bond radius is approximately linearly dependent on the d-character of the d-s-p hybrid bond orbitals. Applying these observations, he is able to postulate a possible valence scheme and electronic structure for the metals of the first ascending branch, copper, zinc, etc. Copper: Eleven electrons outside the argon core. 3d

inn

CU ( A )v = 7 ; A Cu ( B ) v = 5 j core i

4s

-,-,-

a

TJ fJ TJ

* *

.

I I

I 7

*

4p I j

S

* *

'

j

*

*

*

*

0

Metallic copper, on Pauling's hypothesis, involves resonance between structures A and B in the ratio 2 5 : 7 5 , giving a valency of 5.5. Magnetic evidence indicates that platinum and palladium, occupying identical positions to nickel in the second and third long periods, also contain on the average 0.55 to 0.6 unpaired electrons in atomic orbitals. These elements, as well as other metals of the second and third long periods, are treated by Pauling in an identical fashion to the first long period discussed above. Percentage d-character: Considering the electronic structure of metals thus derived, Pauling then calculates the "percentage d-character" of the metallic bonds, the percentage d-character being an indication of bond strength. As examples, we have chosen cobalt, nickel and copper (Table I). TABLE I Percentage d-Character of Cobalt, Nickel, and Copper (Pauling Theory) (Brackets Indicate Bonding Orbitals) Outer electrons

Resonance ratio

Percentage d-character

"ioo

x 34

3!%~0 X

+

6%00

x 96

N f 'ROO X 8#

2Noo X 94

+

7 % ~ 0X

%

= 39.5%

= 40%

= 35.7%

12

M. MCD. BAKER A N D G . I. JENKINS

A list of the percentage &characters of the elements in the first, second and third long period is given in Table 11. TABLE I1 Percentage d-Character and Valency of the Transition Metals, According to Pauling (Proc. Roy. Sac. Al96, 343 (1949)) Element Sc Valency 3 Percentage d-character 20

Ti

Element Y Valency 3 Percentage d-character 19 Element Lu Valency 3 Percentage d-character 19

4

V 5

Cr (6,3)

Mn (6,4)

Fe 5.78

co 6

Ni 6

Cu 5.5

27

35

39

40.1

39.7

39.5

40.0

36

Zr 4

Cb 5

Mo 6

Tc 6

Ru 6

Rb 6

Pd 6

31

39

43

46

50

50

46

Hf

Ta 5

W

4

6

Re 6

0s 6

Ir 6

Pt 6

29

39

43

46

49

49

44

Whereas Pauling considers that all outer electrons take part in bond formation, Mott and Jones express the view that only the outer s-electrons are involved. Pauling was mainly concerned with explaining the properties of the transition metals and the preceding groups (bond lengths and ferromagnetism). The extension of the theory to latter groups is far less satisfactory. For example, on his hypothesis a valency of 5.44 for copper must be assumed. This has no experimental justification. On the other hand the Mott and Jones interpretation is more satisfactory for the elements copper, silver, and gold (one outer s-electron, therefore monovalent), but underestimates the part played by the d-electrons in the bonding of the transition metals. The theoretical interpretation of the reactions concerning the earlier transition metals are more easily understood on Pauling’s picture. Both the Pauling and the Mott and Jones interpretations can be usefully applied in reactions concerning the later transition elements.

111. THENATURE OF THE CHEMISORPTION BOND It is now generally accepted that chemisorption of one or more of the reactants is essential in heterogeneous catalytic reactions, and a consideration of the nature of the bond between surface and substrate is therefore of decisive importance. To explain the phenomena of “activated ” adsorption, Lennard-Jones (9) in 1935 proposed a scheme for the adsorption of hydrogen on metal

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

13

surfaces as represented pictorially by Fig. 7. The transition from van der Waals adsorption to the chemisorbed states requires an energy of activation E. As recent work (18,19) has shown, there may be cases where adsorption of gases on “clean” surfaces proceeds rapidly with little or no activation energy. The suggestion of ionic bond formation (6-8) has been strongly criticized by Emmett and Teller (20), and later by Couper and Eley (21), who calculated theoretically that all possible ionization processes are -2Mt2H

.2M +H,

’y

/

Distance from metal surface

FIG.7. Potential energy diagram for van der Waals (aa) and chemisorbed hydrogen (bb) [J. E. Lennard Jones, Trans. Faraday SOC.28, 333 (1932)l. A& and Aq represent heats of chemisorption and van der Waals adsorption. AE represents activation energy for chemisorption.

prohibitively endothermic on the surfaces of transition metals. In support of this criticism Eley (22) has pointed out that the experimental values of surface potentials, determined by Bosworth and Rideal (23), Oatley (24), and Mignolet (25), correspond to only a fraction of an electronic charge on a hydrogen atom adsorbed on tungsten and platinum respectively. However his calculation, that the small dipole of 0.4Debyes for W6f - H6- corresponds to the transfer of approximately a tenth of an electronic charge, is only applicable if it can be assumed that the W + H dipole is perpendicular to the surface. Boudart (26) suggests that the presence of the electrical double layer produced by the surface dipoles can account for the observed fall in the heat of adsorption and change in work function as the surface coverage is increased. Furthermore, assuming that the dipole interaction is negligible, as will be the case for small surface coverages, the heat of adsorption and work function changes should be related by the equation

14

M. MCD. BAKER AND G. I. JENKINS

where n is the number of valence electrons taking part in bond formation, Aq the change in heat of adsorption, and A$ is the work function change. Unfortunately only in a few cases is there sufficient data to test this relationship. In Table I11 the observed decreases in heats of adsorption are listed against the values calculated from the observed work function changes using Equation (2). The agreement is good for the adsorption of hydrogen and nitrogen, but is less satisfactory in the case of oxygen adsorption. It is probable that the large polarizability of oxygen increases the dipole interaction, and this. may account for the deviation between the observed and calculated values. TABLE I11 Comparison between Calculated and Observed Heats of Adsorption (Boudart, J . Am. Chem. SOC.74, 3559 (1952))

Aq (kcal.) Aq (kcal.)

Calculated Observed

Hz on W

HZon Ni

02 on W

Nz on W

Cs on W

12 13.5

4

24 30

15.9 15

5 5.4

5

The drop in heat of adsorption, and therefore in strength of the bond, will cause an “induced ’’ heterogeneity. Boudart suggests that the presence of “active centers’’ is not necessarily due to an a priori heterogeneity of the surface, but may arise from this induction effect; only a small fraction of the surface will be active at any one time although the whole surface will be involved in catalysis. Dowden (27) has considered the factors favoring (a) positive ion, (b) negative ion, and (c) covalent bond formation on metal surfaces. (a) The ratio of the ionized to unionized species in the surface phase is given by (see Figure 8)

c,+/c, where and

=

fm exp [ f(A)

I‘ = I

+ AU+

(11

+ pEM)/kT]

(3) (4) (5)

f ( A + ) ,f ( A ) = partition functions for ionized and neutral particles, p e M = thermodynamic potential per metal electron per unit volume, I = ionization potential of the neutral particle, AU+ = adsorption energy of the ion, C$ = electron exit work function.

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

15

It can be seen that positive ion formation will be favored by (i) a large exit work function 4, (ii) a large positive value of

("

at

the Fermi surface, and (iii) controlled quantities of electronegative additives which will increase 4.

Eng. Chem. 44,

(b) For negative ion formation

E, = electron affinity of the isolated substrate, AU- = adsorption energy of the anion. Negative ion formation will be favored by (i) a low exit work funca t the Fermi surface, and (" (iii) controlled quantities of electropositive additives.

tion, (ii) a large negative value of

(c) Covalent bond formation: The surface is asymmetric and bonding can occur with unused metal-metal bonds and atomic d-orbitals. A rehybridization of the surface orbitals is suggested, and as the d-s-p hybrid bond strength increases with d-content, strongest bonds are expected when atomic d-orbitals are available to contribute towards the metal-substrate bond. Covalent bonding will be favored by (i) large values of the exit work function, (ii) large positive values of and (iii) the presence of unfilled atomic orbitals.

16

M. MCD. B A K E R A N D G . I. J E N K I N S

Thus the complete removal of an electron will undoubtedly require a prohibitively endothermic energy, I - (b, as pointed out by Emmett and Teller (20). Such a view uses the concept of the removal of the electron t o an infinite distance. If, however, the electron is moved to a finite distance in the solid (that is, partial ionization), the energy required, I’, is less than I , and the small dipole moment of the chemisorption bond can be explained. Dowden takes care of this partial ionization by introducing the term AU+ without actually specifying the physical mechanism of the electron transfer. Data on the relationship between the work function (b of metals and the activation energy E for their oxidation has been tabulated by Ward and Bharucha (28). The equation proposed by Rideal and WansbroughJones ( 5 ) (b-E=K (8) where K is a constant, is shown to be of wide validity. Equation (8) indicates that the rate-determining step in the oxidation process involves the transfer of electrons from the metal to the oxygen forming the corresponding anion. This suggestion is in agreement with the work of Pilling and Bedworth (29) on the oxidation of copper-nickel alloys. The decrease in reaction velocity where the d-band begins to empty has been interpreted by Dowden (27), assuming transference of an electron from the alloy to the oxygen to be the rate-determining step. The strength of the bond between catalyst and substrate will be directly related to the heat of adsorption, and Beeck, Cole, and Wheeler (30) , using an adiabatic calorimeter, have determined the heats of adsorption for hydrogen and ethylene on a number of transition metals. The bond strength will obviously depend on the nature of the metal; using the Pauling Treatment, Beeck (31) relates percentage d-character to the experimentally determined heats of adsorption (on sparsely covered surfaces) (Fig. 9). He suggests that bonding occurs with “atomic” orbitals of the metal and considers that the greater the percentage d-character of the metallic bond the less orbitals are available for bonding with the adsorbate. It must be noted however, from the Pauling treatment outlined above, that the percentage d-character is not inversely related to the number of unpaired atomic orbitals in the metal (Table IV). Theoretical calculations of heats of chemisorption on a sparsely covered surface have been made by Eley (32) assuming the bond between catalyst and substrate to be covalent. The heat of adsorption Qo for the reaction 2M Hz -+ 2M-H is given by the expression

+

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

17

calculating E ( M - Husing ) Pauling’s rule for covalent bond strength (33), i.e.

+

+

E(M-H, = + ~ [ E ( M - ME(H-H)] ) 2 3 . 0 6 ( X ~- XH)’ where (XM- X H ) is the electronegativity difference. I n this treatment the (M-H) bond is assumed t o be of similar type to the (M-M) bond,

1

I

I

35

I

I

40 45 Per cent d character of metallic bond

50

FIG.9. Relationship between heats of adsorption and per cent d-character (Beeck).

i.e. essentially d3sp3 orbitals. Reasonable agreement is obtained between calculated and observed values for the heats of adsorption of hydrogen on transition metals. Eley extends his discussion to the calculation of heats of chemisorption of ethylene using “associative” and “dissociative” models for the adsorption. The two models are, respectively CHz-CHz

/

+ 2Ni Ni + (5 - z)Ni--,

CzHr CzH4

\

---f

Ni

Ni - C2H,

+ (4 - z)Ni - H

TABLE IV Comparison of the Number of Unpaired Atomic Orbitals and Percenlage d-Character on Pauling’s Theory Element

Fe

co

Ni

Percentage d-character Unpaired atomic orbitals

39.7

39.5 1.7

40 0.6

2.2

18

M. MCD. BAKER AND G. I. JENKINS

He concludes that the first (associative) mechanism gives values nearest the observed heat of adsorption determined by Beeck (30), and is therefore accepted “as nearest the truth” (34) (Qo (calculated) = 42 kcal./ mole; Qo (observed) = 58 kcal./mole). Experiments on tungsten and nickel films (Beeck (35), Trapnell (36), and more recent work in Rideal’s laboratory) have shown that when ethylene is added to a clean metal surface ethane appears in the gas phase. A self hydrogenation mechanism must be operative and at least in these cases dissociation of ethylene must occur on the catalyst. It is suggested that the calculations might be complicated by the energy of bond strain in the adsorption of an ethylene molecule to the fixed lattice distances of the metal. Trapnell (18) has used the standard high vacuum technique to study gas adsorption 011 a large number of evaporated metal films incorporating data of Allen and Mitchell (37), Beeck, Smith, and Wheeler (38), and Beeck (19,31). The metals were placed in an order of activity, the activity being defined by the number of the following gases chemisorbed : oxygen, hydrogen, acetylene, ethylene, carbon monoxide, and nitrogen. The gas was considered to be chemisorbed if it covered at least 50% of the geometric area of the metal surface. The following order of activity was obtained: W, Ta, Mo, Ti, Zr, Fe, Ca, Ba > Ni, Pd, Rh, Pt > Cu, A1 > K > Zn, Cd, In, Sn, Pb, Ag. It was impossible to correlate these activities with either the work function of the metal or the lattice spacing of the metal surface. The transition metals were the most active, and this is attributed to holes in the d-band of these metals, though copper, gold, and aluminium, which have complete d-bands, chemisorb carbon monoxide, ethylene and acetylene, and potassium chemisorbs acetylene. Trapnell suggests that promotion of electrons from the full d-band to the s-band might take place in copper and gold, where the d-s optical promotion energy is known to be small, allowing chemisorption of carbon monoxide, ethylene, and acetylene. The smaller affinity of hydrogen and nitrogen for metal surfaces might be insufficient to allow the promotion of electrons, thus preventing chemisorption of these gases. The activities of aluminium and potassium are tentatively explained assuming possible bonding with s- and p-bands. The high activities of calcium and barium have been discussed theoretically by Manning and Krutter (39). These authors suggested small overlap in the s-, p-, and d-bands, some of the electrons being in a second Brillouin Zone, where they have d-band character. Oxygen was chemisorbed on all the metals studied by Trapnell (18), with the exception of gold, which indicates that oxygen adsorption does not require d-band vacancies. The 0 2 - ion was suggested as the adsorbed state, the donation of electrons taking place from the s- or p-bands rather than the d-band.

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

19

The effectiveness of atoms or molecules as catalyst poisons will depend on their size and strength of bonding to the catalyst, and will therefore depend on geometric and electronic considerations. The geometric approach has been investigated by Herington and Rideal (40) ; and long and intensive research by Maxted and co-workers (41) has provided a method of studying the nature of the chemisorbed bond between catalyst and poisons, using as a criterion of effectiveness of the poison its effect on the rate of hydrogenation of unsaturated organic compounds in solution with metals of Group VIII and Group Ib as catalysts, in particular with platinum at temperatures below 100°C. Molecules containing elements of the periodic Groups Vb and VIb were found to act as catalyst poisons if the potentially poisonous element had free electron pairs, e.g. : Toxic

H H : P.. : H

Nontoxic

0 b : i : O ]

3-

.. “Thus it appears that the chemical bond by means of which the poison is linked to the metallic surface resembles the ordinary dative bond in which the poison is donor.” Work by Dilke, Eley, and Maxted (42) on the changes of the magnetic susceptibility of palladium on adsorption of dimethyl sulphide, indicates that the electrons were donated t o the d-band of the catalyst, thus filling up the fractional deficiencies or holes in the d-band. An interesting relationship is found to hold when metallic ions are used as poisons, the ions being adsorbed from solution; the results are summarized in Table V. The ions apparently act as poisons only if the d-shells immediately preceding the s or p valency orbitals are singly or doubly occupied by electrons. These d-shells presumably play a significant part in the bonding of the poison to the catalyst, though the true significance is still not understood since “d-orbitals do not appear to lend themselves easily, without some form of hybridization, to the formation of ordinary chemical bonds.” Whereas in the bulk phase each metal atom is completely surrounded by its neighboring atoms, in the surface phase it is only partially surrounded. I n some cases it has been suggested that the chemisorption bond

20

M. MCD. BAKER AND G .

I.

JENKINS

makes use of modified valence orbitals of the surface atoms. Another view is that a sharing of an electron from the adsorbate occurs with an electron from the "atomic" orbitals of the metal. Both points of view have some experimental support. We suggest that chemisorption in general involves a bonding orbital of the metal which is intermediate between a valence and an atomic orbital. TABLE V Tozicity of Metallic Ions (Mazted) [ J . Chem. SOC.p . 1987 (1949)] I

Toxicity towards platinum

Electronic occupation of external orbitals

Metal ions tested

-l Li+ Na+ K+ Rb+ Ce+

Beg+ Mg2+

Ca%+ Sr*+ Ba*+

Zr4+

Las+ Ces+

Zna+ CdZ+

In*+ Sn2+

Au+

Cr*+ CrS+

Hg9+ Hg+

4d 5d

0 0

0 0

0 0

0 0

6d

0

0

0

3d 3d

0 Q

0

0 0

0

4d 4d 5d 5d 5d

0 0 Q 0

0

0 0 Q 0 Q 0

3d 3d

0 0

Als+

Th4+ Cu+ cu2+ Ag+

3d

No d-shell No internal d-shell 0 0 0 0 04s

TI+

Pb2+

BiS+

Nontoxic 11

'I

05s 06s

0 0 0

0

07s

0

11

0

0 0 0

Toxic

"

0 0

Q

0 4s 04s 05s 05s 06s 06s

Q

Q 6s 0

G

0

0 4s

0 Nontoxic

0

0

0 0

04s

0

Q 0

0

0 0 0

' I

'I

0

"

0

"

Q

"

---

11

-I___

Mn2+ Fez+ coz+ NP+

3d 3d 3d 3d

0 0 0 0

0 0

0 0 0

0

0

0

0

Q

0 0

0 48 (34s 04s 04s

0 0 0 0

Toxic " " "

21

ELECTRONIC FACTOR IN HETEROGENEOUS CATALYSIS

He suggests that generally the activity rises to a maximum for nickel, palladium and platinum in the first, second and third periods, respectively. He further suggests that at least in the second and third periods, when there is less complication due to ferromagnetism, the paramagnetic susceptibility also increases to a maximum for palladium and platinum (Table TI). The paramagnetic susceptibility is determined in part by the structure of the d-band, indicating a relationship between catalytic activity and d-holes (molecular orbital theory) or atomic d-orbitals (resonating valence bond theory). TABLE V I Paramagnetic Susceptibility (gm. atom) of Transition Metals (Second and Third Long Periods). Listed bg Mazted ( J . Chem. SOC.p. 1987 (1949)) Ru

Rb

Pd

Ag

0s

Ir

Pt

AU

+50.8

+114.2

+476.2

-21.6

+9.5

3-29.0

+24.7

-29.6

From various sources Dowden (27) has accumulated data referring to the density of electron levels in the transition metals and finds an increase from chromium to iron. The density is approximately the same from a-iron to p-cobalt; there is a sharp rise between the solid solution iron-nickel (15 :85) and nickel, and a rapid fall between nickel-copper (40 :60) and nickel-copper (20 :80). From Equation (2), the rates of reaction can be expected to follow these trends of electron densities if positive ion formation controls the rates. On the other hand, both trends will be inversely related if the rates are controlled by negative ion formation. Where the rate is controlled by covalent bond formation, singly occupied atomic orbitals are deemed necessary a t the surface to form strong bonds. In the transition metals where atomic orbitals are available, the activity dependence will be similar to that given for positive ion formation. In copper-rich alloys of the transition elements the activity will be greatly reduced, since there are no unpaired atomic d-orbitals, and for covalent bond formation only a fraction of the metallic bonding orbitals are available. Using a constant-volume high-vacuum system, Beeck (31,35) has made a systematic study of ethylene hydrogenation on evaporated films of numerous transition metals. He related the activities of the catalysts to their crystal parameters (Fig. 10). It can be noticed that whereas most of the metals fall on a smooth curve, with a maximum activity for rhodium, the crystal parameters of copper, silver, and gold suggest high activities when, in fact, there is relatively little, as in copper, or practically none, as in silver and gold. Boudart (44), using Beeck’s data, found the loga-

22

M. MCD. R A K E R AND G. I. J E N K I N S

rithm of activity to be a steadily increasing function of the percentage d-character, with the exception of tungsten (Fig. 11). An even better correlation has been suggested by Schuit (45), who takes into account the valency of the metal on the Pauling theory (Fig. 12). Beeck (31) 0.0

-

-2

2 3.0-

M

0-0,

-

6.0

W Ta

NI

I

I

Rh

Pd Fe Pt Cr

I I I ,I1

W

Ta

I

I

FIG.10. Catalytic activity as a function of crystal parameter (Beeck).

states, however, that it would be unwise to ignore crystal parameter considerations for reactions involving adsorption of reactants on two or more sites. Similar work on acetylene hydrogenation (Sheridan and Reid (46)) has shown no clear relationship between percentage d-character and

;I p -

52 52-

m

RI

a_ U

0 c

n

* m

E

x 440 W

‘D .c

Ta

2c

4 4

cr

0 0

Random

Y

2

36-

I

I

-4 0

I

-2.0 Loglo klhydrogenation of CZH41

0

FIG.11. Catalytic activity as a function of per cent d-character (Beeck).

catalyst activity. The catalysts used, in order of decreasing activity, are Pd > Pt > Nil Rh > Fe, Cu, Co, I r > Rh, 0s. It can be seen that rubidium and osmium which have the highest percentage d-character and should exhibit high catalytic activity, have been observed to have a very low activity. There is, however, in this case a parallelism between catalyst

ELECTRONIC FACTOR IN HETEROGENEOUS CATALYSIS

23

activity and magnetic susceptibility as emphasized by Maxted (Table VI), which reflects an undoubted influence of some parameters of electronic structure of the metals on their catalytic activities. Sheridan and Reid suggest that it would be interesting to study the hydrogenation of ethylene over an iridium catalyst, which has a large percentage d-character but low activity for acetylene hydrogenation.

0-

-

s

0

. I -

0

c 0 I m

3

-2-

D

h

f

-x

0

3

-1

-4

0

1 2 3 Valencyx per cent a‘-character

FIG.12. Activity plotted against valency times per cent &character (Beeck).

In comparing the results of Beeck and of Sheridan, it must be pointed out that, whereas Beeck used hydrogen adsorption as a measure of the surface area of the various catalysts, Sheridan inappropriately used the catalyst weights. Kemball (47) has shown, for evaporated metal films, that the energy of activation for the ammonia-deuterium exchange reaction increases in the order Pt < Rh < Pd < Ni < W < Fe < Cu < Ag, whereas the frequency factor remains fairly constant for most of these metals. The criterion of surface area applied was the amount of deuterium adsorbed at 0.1 mm. pressure and 0°C. No obvious relationship was found in this work between the energies of activation and the crystal structures, but a slight tendency was observed for the activation energies t o decrease with increasing work functions. Fitting Kernball’s results to the percentage

24

M. MCD. BAKER AND G. I. J E N K I N S

d-character of the metals (Pauling theory), however, we find a more reasonable correlation (Fig. 13).

V. CATALYTIC ACTIVITYOF ALLOYS A very convenient method of investigating electronic effects in catalysis is to study the activities within a series of metal alloys. Changes in alloy composition will alter the energy density of electron levels at the Fermi surface. Moreover, in the particular case of transition metals, it has been seen that alloying with a metal of Group I b decreases the number of holes in the d-band. These effects should profoundly alter the catalytic activities of the alloys. It is important to keep in mind that within such

Xi?

32

34

36

38

40 42 44 Per cent d character

46

FIG. 13. Relation between activation energy (NH3 (Kernball) and percentage “ d ” character (Pauling).

48

50

+ Dz) exchange

reaction

a series, changes in lattice spacing will be expected and it is desirable that these changes should be small in the alloy system studied. An elegant method of studying the activity of palladium-gold alloys has been developed by Couper and Eley (21) using the ortho-para hydrogen conversion. The energy of activation of the reaction was used as a measure of the catalytic activity and compared with the magnetic susceptibility (a property related to the electronic structure). The palladium-gold group of alloys was chosen be:ause the increase in lattice spacing from palladium to gold is small (0.19A) and almost linear, all the alloys being homogeneous face-centered cubic systems. The catalyst was used in the form of a wire sealed down the axis of a cylindrical reaction vessel and care was taken to prevent its contamination, by baking it out and using a liquid air trap. Instead of the usual

25

ELECTRONIC FACTOR I N HETEROQENEOUS CATALYSIS

degassing of the wire in vacuo a t high temperatures, the reaction vessel was surrounded by a high frequency oscillator and filled with hydrogen a t a pressure of 0.1 mm, so that the hydrogen atoms in the luminous discharge would clean the filament. After this treatment, the wires were heated in vacuo to remove any hydrogen atoms left on their surface. The effect of temperature on the first-order rate constant of the ortho-para hydrogen conversion, given by the Arrhenius equation k = Be-E/RT,was determined for the series of Pd-Au alloys from 0 t o 100% ' gold concentration (Fig. 14). The paramagnetic susceptibility falls

I

FIQ.14. Activation energy as a function of alloy composition (Couper and Eley). Curve X denotes the paramagnetic susceptibility.

t o zero at sixty atom percent of gold, indicating, according t o Mo tt and Jones, t ha t there are 0.6 holes in the d-band of palladium inducing paramagnetism, the holes being filled by the s-electrons of the added gold. According t o Pauling (16), 5.78 electrons in palladium are used in bonding orbitals [(4d)2.66(5~) ( 5 p ) 2 . 2 2 ]The . remaining 4.22 electrons are contained in 2.44 atomic orbitals, leaving 2 X 2.44 - 4.22 = 0.66 unpaired electrons per atom in the atomic orbitals inducing paramagnetism; the unpaired electrons are paired by surplus electrons of the added gold. It was shown t ha t the energy of activation for the ortho-para hydrogen conversion is markedly dependent on d-holes (or free atomic electrons). A low activation energy was observed on all alloys containing d-vacancies. A sharp increase in the activation energy did not occur until all the d-vacancies were filled (at alloy composition 60 % Au :40 % Pd.) . This can

26

M. MCD. BAKER AND G . I. JENKINS

only be explained by assuming that d-orbitals from the metal interior are made available for the surface reaction. Dissolved hydrogen in palladium was found to increase the energy of activation for the ortho-para hydrogen reaction, the dissolved hydrogen contributing its 1s electron, thus filling the d-vacancies of palladium in a similar manner to gold. Here, however, a small atom percent (8%) concentration of hydrogen atoms increases the energy of activation from three to eleven kilocalories per mole. Couper and Eley suggest a skin effect, the hydrogen being concentrated near the surface; thus an 8% concentration in the wire is equivalent to 64% in the first 30,000 layers, this being sufficient to completely fill the d-vacancies of palladium. “The only certain conclusion is that d-vacancies are essential in bonding the activated complex of the conversion reaction.” Eley (34) suggests that the electron of the adsorbed hydrogen is paired to the metallic electrons of the metal. The reaction would occur through a complex contained on a single site H ,~ H H ,

\

M the extra bond being formed by combination with the atomic d-electrons of the metal. Dowden (48) has discussed the experimental results of Reinacker, of Long, Frazer, and Ott, and of Emmett and Skau in the light of the modern electronic theory of metals. Reinacker has studied the ethylene hydrogenation on nickel-copper, palladium-copper and platinum-copper alloys as catalysts. The hydrogenation of benzene has been studied by Long, Frazer, and Ott and by Emmett and Skau. Dowden points out that in all these cases the catalyst activity drops appreciably and the energy of activation rises at the point where the d-band is full, though the change of activation energy does not always occur a t the critical alloy composition. Dowden and Reynolds (49,50) in further experimental work on the hydrogenation of benzene and styrene with nickel-copper alloys as catalysts, found a similar dependence. The specific activities of the nickelcopper alloy catalysts decreased with increasing copper content to a negligible value at 60% copper and 30-40% copper for benzene and styrene, respectively. Low-temperature specific heat data indicated a sharp fall (1) in the energy density of electron levels N ( E ) at the Fermi surface, where the d-band of nickel becomes filled at 60% copper, and (2) from nickel to the binary alloy 80 nickel 20 iron. Further work by these authors (50) on styrene hydrogenation with nickel-iron alloy

+

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

27

catalysts showed that the specific activity decreased with decreasing values of N ( E ) . This is a result to be expected on the Dowden theory, for a reaction proceeding a t a velocity depending on the rate of formation of, or the concentration of, chemisorbed positive ions (or radicals) on the surface (Fig. 15). The rise in activity from 80 nickel-20 iron to 100% nickel is attributed to the large increase in N ( E ) , and the fall from nickel to nickel-copper is attributed to both the decrease in the number of d-holes and t o the large fall in N ( E ) .

I

I

Ni (At. Fr.1

FIG.15. Effect of nickel content on the rate of hydrogenation of styrene by alloy catalysis. Curve A : Hydrogen uptake. Curve B : Number of holes per atom in the 3d-band. Curve C : Coefficient of the electronic specific heat term ID. A. Dowden and P. W. Reynolds, Disc. Faraday SOC.8, 187 (1950)l.

Haber and Weiss postulated an electron transfer mechanism for the ferrous ion catalyzed decomposition of peroxides, the metal donating an electron HzOz-+ OH OHMetal e (10)

+

+

Dowden and Reynolds observed that the rates of decomposition of hydrogen peroxide decreased from pure copper t o copper-nickel alloys, thus suggesting that negative ion formation takes place in the heterogeneous catalytic reaction, in agreement with the Haber and Weiss mechanism based on catalysis in solution. A systematic investigation of dehydrogenation activity over a variety of metal alloy catalysts has been made by Schwab and co-workers (51-53), who have successfully interpreted their results in terms of the bulk metal theory. The test reaction used, namely the dehydrogenation of formic

28

M. MCD. BAKER AND G . I . JENKINS

acid, has zero-order kinetics, which permits the evaluation of the true energies of activation from the temperature coefficient of the Arrhenius equation k = koe-E/RTwithout correcting for heats of adsorption (51). The alloys were used in the form of foils or, in the case of brittle alloys, as small pieces, the catalytic surface area being assumed to coincide with the geometric surface area. The effect of increasing the electron concentration (E.C.) of the alloy catalyst within a phase domain on the rate of the test reaction was studied. Using silver and gold as solid solvents and adding the multivalent metals of Groups 11-V, the electron concentration can be increased up to an E.C. of 1.33 in the period Vb (Cd, In, Sn, Sb) and to an E.C. of 1.1 in the period VIb (Hg, Pb, Bi). The observed rate changes cannot be explained purely by considerations of the atomic surface geometry. Although the Group 11-V metals with greater atomic radii than silver (1.44A) increase the energy of activation og the catalyzed dehydrogenation, the same was found for copper (1.28A), zinc (1.37&), and gallium (1.38A). There is, however, a straightforward correlation between .the activation energies and the electron concentrations of the alloy catalysts: increases in the atomic fraction of the solute metal (z) cause a corresponding increase of the activation energy ( E ) .For small additions, with silver as solvent (i.e. for a-phase alloys), the equation

E

=

EAg

+ A z ( ~- 1)'

(11)

is obeyed, where n is the valency of the solute and A a constant. For Group V metals, A has a value of 11 and for Group VI metals, a value of 100. Similar results were obtained for gold alloys. Equation (11) shows that the activation energy E increases with the square of the E.C. within the a-phase. With heterogeneous alloys of the Hume-Rothery type, other intermetallic phases are formed. Using these alloys, Schwab (51-53) found that similar increases in activation energy occur as the E.C. was increased within a phase domain. At the phase boundaries, the activation energy reached a maximum for a given phase. The saturated a-, p-, e-, and q-phases roughly show the same activation energy, whereas the 7-phase exhibits a sharp maximum (52,53) (Fig. 16). Hence it can be seen that the catalytic activity of both homogeneous and heterogeneous alloys are dependent on the degree of completion of the Brillouin Zones. Within a phase domain, addition of solute metal has the effect of raising the Fermi level, which causes a corresponding rise in the activation energy. The change from a saturated phase to an unsaturated phase, on the Mott and Jones theory, is accompanied by a lowering of the Fermi surface, and this produces a decrease of the activation energy. The unusually high

ELECTRONIC FACTOR I N H E T E R O G E N E O U S CATALYSIS

29

peak of the energy of activation for the ?-phase is obviously the result of the shape of the Brillouin zone. The y-phase of these alloys, due t o their lattice complexity, have nearly spherical Brillouin zones, and a t the saturation point the Fermi surface can nearly fill the zone. The large degree of completion of the zone causes the maximum of the activation energy. The mechanism proposed by Schwab for the formic acid decomposition is t ha t two protons of the hydrogen atoms of this compound intrude into interlattice spaces of the catalyst, that the electrons dissolve in the

cu

I

Composition FIG.16. Dehydrogenation of formic acid on the Cu/Zn alloys [G. M. Schwab and S. Pesmatjoglou, J. Phys. Chem. 62, 1046 (1948)l.

electron gas of the metal, and that the process requires a n activation energy which depends on the electron saturation within the solid. The catalytic efficiencies of iron, cobalt, and their mutual alloys for the hydrogenation of carbon dioxide have been studied by Stowe and Russell (54), who measured the catalytic activities by determining the ratio of methane t o carbon monoxide in the hydrogenation products. It is assumed that the water gas shift-reaction first occurs a t the surface; the carbon monoxide formed is subsequently hydrogenated and the main product of this secondary step is methane. The changes in activity, when observed as a function of varied composition of the alloys, cannot be simply related to lattice changes. Furthermore, the values for the d-character and the coefficients of the electronic specific heat terms of cobalt and iron are almost identical. It was noticed, however, that the activity of a given alloy for the hydrogenation of carbon monoxide thus producing methane decreased as the number of holes in the d-band of the alloys was increased.

30

M. MCD. BAKER AND G. I . J E N K I N S

There occurs a minimum in the methane-carbon monoxide ratio for a n alloy composition of 35 cobalt to 65 iron, which is characterized by a maximum of the saturation magnetic moment, and therefore by a maximal number of d-holes. It is suggested by Stowe and Russell that this relationship is due to an increased strength in the bonding of the chemisorbed hydrogen as the number of d-band holes is increased. As proposed by Beeck (31) for ethylene hydrogenation, the hydrogenation activity varies inversely as the strength of hydrogen chemisorption on a given catalyst. VI. SEMICONDTJCTORS 1. Theory

We have seen, in the case of metal catalysts, how the Mott and Jones and Pauling theories have been applied t o correlating the catalytic activities and electronic structure of metals. For semiconductors, the modern electronic theory is based on work by Wilson (55) using the (‘band” theory to interpret the electrical conductivity. Many authors have attempted to relate catalytic activity of semiconducting solids to such theories. Miilson supposes th at all semiconductors are insulators at O O K , when the allowed bands are either completely full or empty (Fig. 17a). At higher temperatures, the energy difference between full and empty bands in semiconductors is sufficiently small for the promotion of electrons t o take place, this liberation and movement of electrons within the solid giving rise to conductivity. I n insulators, however, the energy difference is large, and no liberation and movement of electrons takes place. It is convenient to distinguish between intrinsic and extrinsic semiconductors. The former contain stoichiometric amounts of the lattice constituents, whereas the latter depend on impurities present in solid solution. These impurities can be due either to foreign elements or t o a stoichiometric excess of one lattice constituent (Figs. 17b and 17c). ( a ) Intrinsic Semiconductors. Excitation of electrons from the filled band will introduce electrons into the empty conduction band, leaving positive “holes” in the previously filled band, and both electrons and positive holes will contribute to the electrical conductivity. ( b ) Extrinsic Semiconductors. Impurity levels can be either donor levels near the empty zone (normal or n-type), or acceptor levels near the filled band (abnormal or p-type). Conductivity in n-type conductors will be due to electrons in the empty band donated by the impurity levels, and in p-type conductors, t o positive holes in the previously filled band, arising from the transition of electrons to the impurity acceptor levels.

ELECTRONIC FACTOK. I N H E T E R O G E N E O U S CATALYSIS

31

Nonstoichiometric composition producing impurity levels can arise in two ways, either (1) excess atoms in interstitial positions or ( 2 ) holes in the lattice. Both methods [( 1) and ( 2 ) ]are theoretically possible in n- and p-type semiconductors. I n n-type semiconductors, method 1 is characterized by a n electropositive interstitial atom, for example Zn in ZnO. This can be considered as an interstitial ion and its trapped electron, whose orbit may well extend over many atomic diameters; a t higher temperatures the electron will be free t o move throughout the lattice, giving rise t o electronic conductivity. I

(a) (b) (C) FIQ.17. Diagrammatic illustrations of intrinsic ( a ) and extrinsic (b,c) semiconductors.

Method 2 is characterized by the loss of a negative ion from its lattice position. To preserve electroneutrality, an electron is trapped a t the point where the negative ion is missing. I n p-type semiconductors, method 1 is characterized by a n electropositive interstitial atom, for example Zn in ZnO. The large work required for bringing an electronegative atom into an interstitial position might be the reason that no cases of this type are known. Again method 2 is characterized by the loss of an ion from the lattice, in this case a positive ion, e.g. copper from cuprous oxide. T o preserve electroneutrality in this case, a metal atom in the vicinity of the hole must carry a n excess positive charge. These positive charges are not fixed, but can be considered as moving in an orbit around the defect. At higher temperatures the positive charge is no longer localized, and can move through the lattice giving rise to positive hole conductivity. Thus, the electrical conductivity will be a measure of the number of free charge carriers of the catalysts. Adsorption processes which produce or destroy defects, or trap free electrons or holes, will alter the conductivity. Magnetic susceptibility, which will usually be changed by

32

M. MCD. B A K E 8 .4ND G. I. JENKINS

alteration in the number of unpaired electrons, has also been studied in connection with adsorption effects in semiconductors (56,57). 2. Adsorption Studies Dowden (27) in a theoretical approach similar to that used for metals, has examined the probability of positive ion formation on intrinsic and extrinsic semiconductors. The energy of activation of this process in intrinsic semiconductors is considered t o be proportional to

+ @), where I* is the ionization potential of the activated 2 ( I + A U * ) ; the activation energy decreases a s rp (exit work

(I -* rp complex

defining the Fermi rp ( level. I n the case of n- and p-type semiconductors, the Fermi level will be defined by ( - + and ( - - respectively, and the

function) increases and as A E decreases,

-

)A :

A:") 2 activation energy for positive ion formation in these cases will be proI$J

I*

AE

- I$J

"I> ~

+ AE -

I* - rp

AE" + -). 2

Adsorption on oxides has been considered by Garner (58), who pointed out that adsorption of hydrogen, for example, can take place on the cations, a process similar t o that on pure metals (M-H), or by bonding t o the oxygen ion (MO-13). I n evidence, Garner and Kingman (59) found reversible adsorption of hydrogen (M-H) a t low temperatures on both zinc and chromic oxide, and irreversible adsorption of hydrogen a t higher temperatures (M-OH). On raising the temperature still further, water was desorbed. The work has been extended to the adsorption of carbon monoxide on cuprous oxide (p-type conductor) by Garner, Stone, and Tiley (60,61). This gas was reversibly adsorbed a t room temperature. Above 80°C the gas desorbed was carbon dioxide and the formation of a surface carbonate (COs-) is postulated, the carbon monoxide reacting with oxygen ions of the lattice. According to the authors, carbon monoxide is held on Cu+-ions at room temperature. Garner, Gray, and Stone (62) found that the room temperature adsorption of carbon monoxide decreased the conductivity of cuprous oxide. On evacuation, the conductivity rapidly returned t o its initial value. To account for this conductivity change Garner et al. conclude that a COZor CO, complex is most likely on the surface. On heating the adsorbed carbon monoxide t o above lOO"C, carbon dioxide was desorbed, resulting in a permament decrease in conductivity (excess oxygen ions removed from the lattice). Whereas Garner, Stone, and Tiley

33

ELECTRONIC FACTOR I N HE T E R OGE NEO U S CATALYSIS

found that only part of the adsorbed carbon monoxide was rapidly removed by evacuation, Garner, Gray, and Stone observed a rapid change in conductivity to the initial value on pumping. In our opinion these results indicate that the adsorption of carbon monoxide at room temperature occurs in two different ways, with only the rapidly desorbed fraction influencing the conductivity. Chemisorption of oxygen would be expected t o increase the number of positive holes in the cuprous oxide lattice, and in agreement with this, Garner, Gray, and Stone (62) noted an increase in the conductivity. The mechanism suggested by Stone and Tiley (61) for the adsorption is that 02cu+

cu+

02- c u +

02- Cu+

cu+

cu+

02-

cu+

02-

Cu+

02-

CU2+ 02-

cu+ cu+

Cu+ 02-

cu2+

cu+

cu+

02-

cu+

PI

I4

c u + 02- c u +

-

4

Cu+ 02-

cu+

/cu+

a

02- Cu+

cu+\

m e 02- c u + 0 2 FIG. 18. Pictorial representation of the adsorption of oxygen on cuprous oxide [W. G. Garner, F. S. Stone, and P. F. Tiley, Proc. Roy. Sac. A211, 472 (1952)l.

oxygen dissociates to 0- or O= on the surface, electrons being donated by the process c u + + Cuff E

+

To calculate the number of surface ions Cu+, the adsorption of krypton was determined and, assuming that the (001, 011, 111) planes of the oxide are exposed in equal amounts, the average number of Cu+ sites can be estimated. The adsorption of oxygen exceeded the calculated number of these sites and indicated a penetration of the surface by oxygen, a suggestion which is compatible with the observation that carbon monoxide does not react with all the oxygen previously adsorbed producing carbon dioxide. The mechanism is illustrated pictorially in Fig. 18. Oxygen adsorption has also been studied on zinc oxide, an n-type oxide, by Wagner (63) and later by Bevan and Anderson (64). The conductivity decreases with increase in oxygen pressure, a result interpreted by Wagner by assuming the equilibrium

+

Zn2+lattloe 0'-

?402

+ Znointeratitiai

(12)

34

M. MCD. BAKER AND G. I. J E N K I N S

An increase in oxygen pressure will obviously decrease the amount of interstitial zinc atoms which can also be conceived as zinc ions combined with conduction electrons. The work of Stone, and Tiley (61) on cupric oxide surfaces is also interesting from an electronic view point. They found th a t the amount of oxygen adsorbed on this oxide, after its previous saturation with carbon monoxide, corresponded to the ratio 460,: CO, in agreement with earlier work by Garner (65) on ZnO and Cr203. The explanation proposed in these cases was that carbon monoxide forms a surface complex with two oxygen ions of the lattice, thus leaving a n anion deficiency which can be filled by an oxygen atom (Fig. 19). Me++ Me++ Me++

o=

o=

Me++ Me++ Me++

o=

Me++ Me++ Me++

0”

Me++ ’Me++ Me++

0‘ CO ___c

Me++ Me++ Me++

o=

2€ 102

Me++ Me++ Me++ ---+ 0’

co3=

Me++ Me++ Me++

0’

Me++ Me++ Me++

o=

coa=

Me++ Me++ Me++

FIG.19. Illustration of carbon monoxide adsorption followed by oxygen adsorption on metal oxides [W. G. Garner, J . Chem. SOC.1239 (1947)l.

3. Reaction Studies

Garner et al. (60-62,66) have made an extensive study of the catalytic oxidation of carbon monoxide on cuprous oxide, following the effects taking place on the surface by measuring the semiconductivity and the heats of adsorption. The electrical conductivity of a cuprous oxide film during the oxidation was found to be approximately the same as of a film of the same material that has been saturated with carbon monoxide. It differed from that of an oxygenated surface of Cu20. Furthermore, it was demonstrated by Garner, and his coworkers that preadsorption of oxygen caused a much larger time lag preceding the establishment of a stationary state in electrical conductivity than would be the case for a bare surface. This suggests that during reaction the surface concentration of carbon monoxide is high and the oxygen surface concentration is low. It is suggested that oxygen and carbon monoxide react on the surface giving a carbonate complex which subsequently reacts with more adsorbed carbon monoxide t o yield carbon dioxide. and Wagner (67) has put forward the following tentative scheme for th e

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

35

decomposition of nitrous oxide on oxide catalysts, in which the exchange of electrons between the oxide catalyst and reactants plays an important part.

+ +

+

Cat. 2 ~ - NzO = N Z 02- NzO = N z O2 2e-

+

+ +

02-ads

+ Cat.

(15) (16)

This proposed mechanism accounts for the retardation of the rate of decomposition of nitrous oxide by oxygen, observed with the oxides of zinc and cadmium. If the reaction step (15) is rate-determining, the reaction will be favored by the quasi-free electrons in n-type semiconductors such as zinc and cadmium oxides. Increasing the oxygen pressure over these catalysts will decrease the number of free electrons due to the adsorption of oxygen as negative ions and will result in a decreased rate of the N2O decomposition. Wagner has shown that the conductivity changes observed for zinc oxide in oxygen and nitrous oxide are of the direction expected from Equation (15). Using the results of Schwab and co-workers (68), of Schmid and Keller (69), and of their own work on cuprous oxide, Dell, Stone, and Tiley (70) have shown that the solid oxides active for the decomposition of nitrous oxide can be divided into three groups depending on the temperature required to produce measurable decomposition. Group 1. Below 400°C Group 2. 400-550°C Group 3. Above 550°C

Cu20, NiO, COO CuO, MgO, CeOz, CaO A1203,ZnO, Fe203,TiOz, Cr203.

It is of interest t o consider the semiconducting properties of these oxides. The first group are all p-type semiconductors. In the second group MgO, CuO, and CeOz are known t o be insulators below 550°C, whereas CuO is an intrinsic semiconductor. The third group consists of n-type semiconductors with the exception of Cr2O3, which very nearly has the properties of an intrinsic semiconductor, though it does show p-type conductivity. This existence of catalyst classes operative at diff went temperatures points to the importance of electrons as reactants as indicated by Wagner's mechanism for the reaction. In the case of n-type oxides, the reaction step (15) is favored by the quasi-free electrons already existing in the catalyst, due to ionization from impurity levels t o the conducting band. I n agreement with the Wagner mechanism, the reaction decreases the conductivity of the n-type oxides because the free conducting electrons are trapped by the chemisorbed oxygen. Dell, Stone, and Tiley (70) suggested that the high temperature required in the case of n-type oxides is probably due to the difficulty of these oxides to chemisorb oxygen, the adsorption process requiring a deep electron trap and thus taking place at relatively few sites

36

M. MCD. B A K E R A N D G. I. JENKINS

in the surface. With p-type oxides the observed increase in conductivity due t o the nitrous oxide decomposition reaction must arise from a n increase in the number of positive holes. I n this case the chemisorbed oxygen must be acting as an acceptor impurity level near t o the top of the filled electron band. Removal of electrons from this band (reaction step (15)) t o these additional acceptor levels will increase the number of positive holes. The catalytic activity of the insulator oxides for the nitrous oxide decomposition reaction is more difficult t o explain. Dell et al. (70) suggest that surface anionic vacancies may occur at temperatures above 4OO0C, these vacancies facilitating chemisorption of oxygen in the reaction step (15).

Mn. per cent

FIQ.20. Catalytic activity of low ignition manganese oxides on alumina as a function of manganese concentration [J. Mooi and P. W. Selwood, J . Am. Chem. SOC.74, 1750 (1952)J.

The effect of catalyst supports (e.g. manganese dioxide on alumina) on the catalysis of the hydrogen peroxide decomposition has been studied by Mooi and Selwood (56). Manganese dioxide appeared as patches on the support. On lowering its percentage, dispersion of the MnO2 occurred, thus presenting a larger available surface area of this substance. The catalytic activity per unit of weight of manganese dioxide was determined (Fig. 20) and, as expected from the enlarged surface area, increased with decreasing percentage of manganese dioxide. Less easy to understand, however, was the fall in rate a t still lower manganese dioxide percentages. The mechanism suggested is a reduction-oxidation process of the catalyst, involving Mn3+ and Mn4+ oxidation states. Mnz03 2Mn02

+ HzO2 2Mn02 + H2O + HzO2-t + Mn203 + HzO --+

(17)

(18) Mooi and Selwood suggest that a t the lower concentrations the cata0 2

ELECTRONIC FACTOR I N H E T E R O G E N E O U S CATALYSIS

37

lyst is strongly influenced by the support, valence induction occurring, which stabilizes one of the catalyst forms. I n agreement with this concept an increased energy of activation was observed for catalysts containing lower percentages of manganese dioxide on alumina. Also reduction of the MnOz catalyst first produced an increase in activity, followed by a decrease a t higher percentages of Mn3+. A similar oxidation-reduction mechanism in the carbon monoxide oxidation reaction on oxide catalysts has been proposed by Benton (71), Bray (72), Frazer (73), and Schwab (74). I n this reaction also, Mooi and Selwood (57) found th at a decrease in the percentage of iron oxide, manganese oxide or copper oxide on the alumina support first increased the rate, and then a t lower percentages decreased the rate, of carbon monoxide oxidation, indicating that valence stabilization is again operative in these cases. Oxidation and reduction reactions on zinc oxide, a typical n-type oxide, have been investigated by Parravano (75) in a n attempt to find a possible relationship between activity and electron concentration of the catalyst. He did this by modifying the electronic characteristics of the catalyst by the controlled addition of foreign ions. I n all the reactions studied by Parravano, namely the oxidation of carbon monoxide, the hydrogen-deuterium exchange and the reduction of nickel oxide, the energies of activation were increased by monovalent cations, and decreased by ions of a valency greater than two. I n both the oxidation of carbon monoxide and the hydrogen-deuterium exchange, a n initial induction period was noted, a decrease in rate occurring in the former and a slow increase of activity in the latter. Parravano points out th a t zinc oxide prepared in air will have a large concentration of adsorbed oxygen and t ha t carbon monoxide will react with this oxygen, after which more oxygen will be adsorbed on the surface.

On a surface covered with oxygen, step (19) will be faster than step (20), but as the surface concentration of oxygen decreases, a stage will be reached where reaction (19) becomes the rate-determining stage. I n consequence, the overall rate of reaction will decrease until it reaches a steady state. The mechanism proposed by Parravano is also in accord with the fact that the initial rate of the carbon monoxide oxidation can be decreased by baking the nickel oxide in helium, a process which will undoubtedly reduce the oxygen concentration on its surface. The suggested mechanism for the hydrogen-deuterium exchange is adsorption of hydrogen forming anions foreign t o the lattice (OH-), followed by a n

38

M. MCD. BAKER AND G. I. JENKINS

exchange taking place between the gas phase hydrogen or deuterium, and the chemisorbed hydrogen. The hydrogen-deuterium exchange reaction has also been studied on chromia (p-type oxide) by Voltz and Weller (76). This exchange reaction is convenient for the determination of activity-conductivity relationships, since i t can be measured a t temperatures so low that specific surface states achieved by pretreatment of the catalyst are not affected. High temperature hydrogenation would cause an activity-levelling effect due to a reduction of the surface. The pretreatment of the oxide consisted of heating i t t o 5OO0C, either in a hydrogen atmosphere to give a ((reduced” catalyst or in a n oxygen atmosphere to give a n “oxidized” catalyst. It was found that in its oxidized state, chromic oxide has a low activity for the exchange and a high electrical conductivity, whereas in its reduced state, it has a high activity and a low conductivity. Thus the activity rises as the number of defects responsible for electrical conductivity falls. Work by Clark (1952) (77) has shown a similar dependence on the pretreatment of the catalyst. A number of both n- and p-type oxides supported on Si02-A1203 were pretreated b y either heating in air or in hydrogen a t 500°C. The activity in all cases was considerably higher after the hydrogen treatment. Clark suggests a n association of activity with metal ions, reduction resulting in a relative increase in the number of metal ions in the catalyst. We would like to point out that in all these cases the rate of H z - D ~ exchange is increased by an increase in the electron concentration of the oxide catalysts (Table VII). This means th at in all these systems a pretreatment which increases the electron concentration also increases the rate of the exchange reaction, and vice versa. Beeck (31) has observed on metal films an increase in hydrogenation activity with decreasing strength of the hydrogen-metal bond, the heat of hydrogen adsorption being used as an indication of the bond strength. It is suggested th a t if hydrogen adsorption on the metal oxides involves either the formation of positive ions or covalent bonds, an electron excess will presumably decrease th e strength of bonding. Thus, in view of Beeck’s work, the observed increase in activity would be explainable. No direct relationship can be expected between the electrical conductivity and catalytic activity because a n increase in electron concentration of the catalyst will increase the conductivity of n-type semiconductors, but decrease the conductivity of p-type semiconductors. VII. INSULATORS This catalyst class consists mainly of the oxides and halides of the lower elements of Groups 111, IV, and V of the periodic table, distin-

ELECTRONIC FACTOR IN HETEROGENEOUS CATALYSIS

39

TABLE VII Hydrogen-Deuterium Exchange on Oxide Catalysts

Catalyst

ZnO

Author

Parravano

Voltz and Weller

A. Clark

Effect on oxide

Pretreatment

Addition of cations Increases the with valence greater number of electrons than 2 Decreases the Addition of cations number of with valence of 1 electrons Increases the 0 2 at 500°C number of positive holes Decreases the Hz at 500°C number of positive holes 0, a t 500°C Decreases the number of electrons 1 Increases the Hz at 500°C number of electrons

1

Effect on exchange activity Rate increased Rate decreased Rate decreased

Rate increased

Rate decreased Rate increased

guished by their ability to catalyze polymerization, isomerization, cracking, and alkylation reactions, etc. of hydrocarbons. An understanding of the nature of the surface processes occurring on these insulator catalysts is facilitated by the observation that these reactions are also catalyzed by strong acids, for which a mechanism has been postulated that involves carbonium ions as intermediates. It follows that similar intermediates may be expected in the heterogeneous catalytic reactions. In recent years a large amount of evidence, mostly obtained with the very widely used silica-alumina catalysts, has shown that the catalysts are potential acids. The acidic nature of the SiOz (A1203) catalysts over the whole range was explored by Thomas (78) using a titration procedure with potassium hydroxide as neutralizer. A general relationship was observed between the amount of catalytic cracking and acidity. His method of determining the acid nature of the catalyst has been criticized by Miesserov (79) whose work indicated that NaOH solutions reacted with other protons

40

M. MCD. BAKER AND G. I. J E N K I N S

of the catalyst as well as with protons responsible for catalysis. Tamele (80) found that potassium hydroxide gave a n indefinite titration endpoint, and used instead a titration method with a weak base in an anhydrous medium t o determine the number of acid centers on the catalyst. A straight-line relationship was found between this number and the rate of polymerization of propylene. Confirmatory evidence of the acid properties has been presented by Milliken, Mills, and Oblad (81). The tests used were (1) reaction with carbonate solutions, (2) titration with solutions of bases, (3) hydrolysis of sucrose, (4) reaction with volatile bases, (Fj) reaction with Zerewitinoff reagent.

H

Si

OH-+

9" p

Si (a)

6+

0-H I

si

:0: H+ , O : . - A1~ ~ : O : - 0

..-1

Si

:o: Si

(b)

(4

FIG.21. Interpretations of the acid centers in insulator catalysts.

Dowden (27) considers the active centers for carbonium ion formation to be associated with surface cation vacancies. A proton, derived from water contained in the catalyst, is attracted t o the anions surrounding the vacancy. A hydrocarbon molecule is assumed to be held by polarization forces above this lattice defect and the proton will be distributed between the hydrocarbon and the anions, forming a carbonium ion of a definite lifetime. Evidence t hat the presence of water is important in these reactions has been obtained by Hansford (82). Pretreatment of the catalyst with a stream of predried air a t 500°C resulted in a marked decrease in the rate of cracking. Further, if deuterium oxide was substituted for the water in the catalyst a large percentage of the deuterium exchanged with hydrogen atoms contained in the hydrocarbon undergoing reaction. Hansford also pointed out that effective catalysts for cracking reactions are always prepared from one or more hydrous oxides. The way in which the proton is associated with the alumina-silica catalyst is a matter of some doubt. Thomas (78) assumes the aluminium to be tetrahedral when linked with tetrahedral silicon, the extra valence electron being supplied by hydrogen from water contained in the catalyst (Fig. 21a). Both aluminium hydroxide and silicic acid are very weak acids because of the affinity of oxygen for the hydrogen (83), and a coordination of aluminium with the hydroxyl oxygen contained in the catalysts

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

41

SiOz, A1203,xHnO has beensuggested by Hansford (82) (Fig. 21b). Tamele (80), in a development of the Thomas interpretation, pictures a displacement of electron along A1-0-Si bonds, the acidity arising because of the tendency of aluminium to accept an electron pair, thus completely filling the p-orbitals (Fig. 21c). All these concepts are based on the assumption that the acid centers important in these catalytic reactions exist independently on the catalyst. On the other hand, Milliken et al. (81) consider that the acid centers are onIy stable in the presence of the basic substrate. The mechanism postulated by these authors for cracking type reactions, involving carbonium ions held at acid sites on the surfaces, is as follows Me

Me

I

Me

I

HfA-

R-C-CHZ-C=CH*

I

R-C-CH2-C-CHs

I

I

Me

Me

I

+ A-

93

(21)

Me Me

Me

Me

I R-C-CH2-C-CH3 I e3

-+

Me

Me

I + CHz=C-CHs

(22)

+ H+A-

(23)

Me

Me

I R-C@ I

I R-C@ I Me

+ A-+

I

R-C=CHz

Me

H+A- represents the acid centers at the surface of the catalyst. This general scheme can also be applied for other reactions catalyzed by “acidic” catalysts. An increase in the electronegativity of R is expected to result in a decreased rate of carbonium ion formation, a relationship shown to hold for substituted 1: 1-diaryl-ethanes (84).

VIII. CONCLUDING REMARKS We have seen that whereas some authors have used the energy of activation of a reaction as a criterion of catalyst activity, others have felt it more desirable to use the “rate,” k . Often the Arrhenius equation has been found t o be useful k =

BOe-E’RT

and in some cases a relationship has been found between the Bo factor and E. In a number of reactions, neither the energy of activation nor the rate of reaction can be considered as completely dependable yardsticks

42

ill. MCD. B A K E R AND G . I . JENKINS

for measuring the relative activities of catalysts. For example, in Beeck’s experiments on the rate of hydrogenation of ethylene, the amount of surface available as reaction proceeds is diminishing with time due to complex formation, resulting in a change of the Bo factor. If the mechanism proposed by Beeck is correct, the measured speed of reaction depends on the rate of reaction of adsorbed ethylene with adsorbed hydrogen occurring on a decreasing number of sites, the rate of decrease presumably depending on the catalyst used. K e suggest that the results of tlideal and Trapnell (85) should be considered carefully when dealing with the activation energy. Rideal and Trapnell have pointed out th a t adsorption of reactants would be expected to decrease as the temperature is increased, reaction thus occurring on fewer sites. Therefore the “energy of activation” measured will not necessarily be a true measure of the catalyst activity. For convenience, the catalysts considered in this review have been grouped in three classes. The distinction between the metal, semiconductor and insulator catalysts, however, is not a sharp one. It has been pointed out (Dowden (86)) th at the activity of zinc oxide (a semiconductor oxide) for methanol decomposition can be correlated to the relative catalytic activities within the series of catalysts, nickel, nickel-copper alloys, copper-zinc alloys, zinc oxide. The high activity of the nickel-rich alloys can be coordinated to the d-band vacancies in these alloys. I n the copperrich nickel-copper alloys and in the copper-zinc alloys, where the d-band is full, the decreasing catalytic activity is t o be associated with the raising of the Fermi level, i.e. to the filling of the electron conductivity bands, if the reaction is controlled by electron donation to the catalyst. Zinc oxide, as a n excess semiconductor, has a larger electron concentration and will therefore be expected to have a decreased activity. Furthermore, the transition metal oxides can act as dehydrogenation-hydrogenation catalysts similarly t o the transition metals, and can also promote acid-type reactions (compare insulators) (77). An interesting relationship between color of the catalyst and the type of reaction catalyzed is indicated by Roginsky (87). He distinguished two reaction types, electronic transfer playing an important part in the first, while the essential role in the other is the formation of unstable chemical bonds such as hydrogen or hybrid bonds. Whereas highly colored solids (presence of color is evidence for a transfer of electrons) exert the greatest influence in accelerating reactions of the first type, color is of no significance in the second. As many authors have indicated, it is well to be aware of the dangers involved in assuming that bulk and surface properties are the same; differences in bond hydridization are to be expected. However, the bulk

ELECTRONIC FACTOR IN H E T E R O G E N E O U S CATALYSIS

43

phase will undoubtedly influence the properties of the surface and, in the absence of any comprehensive surface theory, it is useful and advisable as a first approximation to use the theories of the solid state, correlating them with the activities of the solid in catalytic reactions.

REFERENCES 1. Langmuir, I., J. Am. Chem. Soc. 38, 2221 (1916); 40, 1361 (1918); Trans. Faraduy SOC.17, 607 (1922). 2. Trapnell, B. M. W., Advances in Catalystk 3, 1 (1951). 3. Griffiths, R. H., Advances i n Catalysis 1, 91 (1948). 4. RoginskiI, S., and Schul’tz, E., 2. physik. Chem. A138, 21 (1928). 5. Rideal, E. K., and Wansbrough-Jones, 0. €I., Proc. Roy. SOC.Al23, 202 (1929). 6. Brewer, A. K., J. Phys. Chem. 32, 1006 (1928). 7. Schmidt, O., Chem. Revs. 12, 363 (1933). 8. Nyrop, J. E., “The Catalytic Action of Surfaces.” Williams & Norgate, London, 1937. 9. Lennard-Jones, J. E., Trans. Faraday SOC.28, 333 (1932). 10. Mott, N. F., and Jones, H., “Properties of Metals and Alloys.” Oxford Univ. Press, London, 1936. 11. Hume-Rothery, W., “Atomic Theory for Students of Metallurgy.” Institute of Metals, London, 1946. 12. Hume-Rothery, W., J. Znsl. Metals 36, 309 (1926). 13. Hume-Rothery, W., “The Metallic State,” p. 328. Oxford, London, 1931. 14. Jones, H., Proc. Roy. Soc. A147, 396 (1934). 15. Pauling, L., Phys. Rev. 64, 899 (1938). 16. Pauling, L., J. Am. Chem. SOC.69, 542 (1947). 17. Pauling, L., Proc. Roy. SOC.A196, 343 (1949). 18. Trapnell, B. M. W., Proc. Roy. SOC.A218, 566 (1953). 19. Beeck, O., Advances in CataZysis. 2, 151 (1950). 20. Emmett, P. H., and Teller, E., 12th Report of Comm. on Contact Catalysis 68, New York, 1940. No. 8, 172 (1950). 21. Couper, A., and Eley, D. D., Discussions Faraday SOC. 22. Eley, D. D., Quart. Revs. 3, 209 (1949). 23. Bosworth, R. C. L., and Rideal, E. K., Physica 4, 925 (1937); Proc. Roy. Soc. A162, 1 (1937); Proc. Cambridge Phil. SOC.33, 394 (1937). 24. Oatley, C. W., Proc. Phys. SOC.61, 318 (1939). 25. Mignolet, J. C. P., Discussions Faraday SOC.8, 105 (1950). 26. Boudart, M., J . Am. Chem. Soc. 74, 3556 (1952). 27. Dowden, D. A., J . Chem. SOC.p. 242 (1950). 28. Ward, A. F. H., and Bharucha, N. R., Rec. Irav. chim. 72, 735 (1953). 29. Pilling, N. B., and Bedworth, R. E., Ind. Eng. Chem. 17, 372 (1925). 30. Beeck, O., Cole, W. A., and Wheeler, A., Discussions Faraday SOC.NO. 8, 314 (1950). 31. Beeck, O., Discussions Paraday SOC.No. 8, 118 (1950). 32. Eley, D. D., Discussions Faraday SOC. No. 8, 34 (1950). 33. Pauling, L., “The Nature of the Chemical Bond.” Cornell, Ithaca, N. Y., 1940. 34. Eley, D. D., “Catalysis and the Chemical Bond.” Reilly Lectures, No. 7 . Univ. of Notre Dame Press, Indiana, 1954.

44

M. MCD. BAKER AND G. I. J E N K I N S

35. Beeck, O., Revs. Mod. Phys. 17, 61 (1945). 36. Trapnell, B. M. W., Trans. Faraday SOC.48, 160 (1952). 37. Allen, J. A,, and Mitchell, J. W., Discussions Faraday SOC.No. 8, 361 (1950). 38. Beeck, O., Smith, A. E., and Wheeler, A., Proc. Roy. SOC.A177, 62 (1940). 39. Manning, 111. F., and Krutter, H. M., Phys. Rev. 61, 761 (1937). 40. Herington, E. F. G., and Rideal, E. K., Trans. Faraday SOC.40, 505 (1944). 41. Maxted, E. B., Advances in Catalysis. 3, 129 (1951). 42. Dilke, M. H., Eley, D. D., and Maxted, E. B., Nature 161, 804 (1948). 43. Maxted, E. B., J . Chem. SOC.p. 1987 (1949). 44. Boudart, M., J. Am. Chem. SOC.72, 1040 (1950). 45. Schuit, G. C. A., Discussions Faraday Sac. No. 8, 205 (1950). 46. Sheridan, J., and Reid, W. D., J . Chem. SOC. p. 2962 (1952). 47. Kemball, C., Proc. Roy. SOC.A214, 413 (1952). 48. Dowden, D. A., I n d . h'ng. Chem. 44, 977 (1952). 49. Dowden, D. A,, and Reynolds, P. W., Discussions Faraday SOC.No. 8, 184 (1950). 50. Reynolds, P. W., J . Chem. SOC.p. 265 (1950). 51. Schwab, G.-M., Trans. Faraday SOC.42,689 (1946). 52. Schwab, G.-M., and Pesmatjoglou, S., J . Phys. & Colloid. Chem. 62, 1046 (1948). 53. Schwab, G.-M., Discussions Faraday SOC.No. 8, 166 (1950). 54. Stowe, R. A., and Russell, W. W., J . Am. Chem. Soc. 76, 319 (1954). 55. Wilson, A. H., Proc. Roy. SOC.A133, 458 (1931); A134, 277 (1931). 56. Mooi, J., and Selwood, P. W., J . Am. Chem. SOC.74, 1750 (1952). 57. Mooi, J., and Selwood, P. W., J . A m . Chem. SOC.74, 2461 (1952). 58. Garner, W. E., Dzscussions Faraday SOC.No. 8, 211 (1950). 59. Garner, W. E., and Kingman, F. E. J., Trans. Faraday SOC.27, 322 (1931). 60. Garner, W. R., Stone, F. S., and Tiley, P. F., Discussions Faraday SOC.No. 8, 254 (1950). 61. Garner, W. E., Stone, F. S., and Tiley, P. F., Proc. Roy. SOC.A211, 472 (1952). 62. Garner, W. E., Gray, T. J., and Stone, F. S., Discussions Faraday SOC.No. 8, 246 (1950). 63. Wagner, C., 2. physik. Chem. B22, 181 (1933). 64. Bevan, D. J. M., and Anderson, J. S., Discussions Faraday SOC.No. 8,238 (1950). 65. Garner, W. I<., J . Chem. SOC.p. 1239 (1947). 66. Garner, W. E., Gray, T. J., and Stone, F. S., Proc. Roy. SOC.A197, 294 (1949). 67. Wagner, C., J. Chem. Phys. 18, 69 (1950). 68. Schwab, G.-M., and Schultes, H., 2. physik. Chem. B9, 265 (1930); Schwab, G.-M., Staegar, R., and Baurnbach, H. H., ibid. B21, 65 (1933); Schwab, G.-M., and Schultes, H., ibid. B26, 411 (1934); Schwab, G.-M. and Staegar, R., ibid. B26, 418 (1934). 69. Schmid, G., and Keller, N., Naturwissenschajten 37, 43 (1950). 70. Dell, R. M., Stone, F. S., and Tiley, P. F., Trans. Faraday SOC.49, 201 (1953). 71. Benton, A. F., J . Am. Chern. SOC.46,887, 900 (1923). 72. Almquist, J. A., and Bray, W. C., J . Am. Chem. SOC.46,2305 (1923). 7 3 . Whitsell, W. A., and Frazer, J. C. W., J . Am. Chem. Sac. 46,2841 (1923); Pitzer, E. C. and Fraeer, J. C. W., J . Phys. Chem. 46, 761 (1941). 74. Schwab, G.-M., and Drikos, G., 2. physik. Chem. A186, 405 (1940). 75. Parravano, G., J. Am. Chem. SOC.74, 1194 (1952); 76, 1448, 1452 (1953). 76. Volte, S. E., and Weller, S., J . Am. Chem. SOC.76, 5227 (1953). 77. Clark, A., I n d . Eng. Chem. 46, 1476 (1953). 78. Thomas, C. L., Znd. Eng. Chem. 41, 2564 (1949).

ELECTRONIC FACTOR I N HETEROGENEOUS CATALYSIS

45

79. Miesserov, K. G., Doklady Akad. Nauk. S.S.S.R. 87, 627 (1952). 80. Tamele, M. W., Discussions Faraday SOC.No. 8, 270 (1950). 81. Milliken, T. H., Mills, G. A., and Oblad, A. G., Discussions Faraday SOC.No. 8, 279 (1950). 82. Hansford, R. C., Znd. Eng. Chem. 39, 849 (1947). 83. Pauling, L., “The Nature of the Chemical Bond.” Cornell, Ithaca, N. Y., 1940. 84. May, D. R., Saunders, K. W., Kropa, E. L., and Dixon, J. K., Discussions Faraday SOC.NO.8, 290 (1950). 85. Rideal, E. K., and Trapnell, B. M. W., Discussions Faraday SOC.No.8,114 (1950). 86. Dowden, D. A., Discussions Faraday SOC.No. 8, 296 (1950). 87. Roginsky, S.Z., Rept. Acad. Sci. U.S.S.R. 47, 439 (1945).

This Page Intentionally Left Blank

Chemisorption and Catalysis on Oxide Semiconductors G. PARRAVANO

AND

M. BOUDART

Forrestal &Search Center, PTinceton University, Princeton, New Jersey Page I. Chemisorption of Hydrogen on Zinc Oxide.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 52 11. Electronic Properties of Zinc Oxide.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. The Hydrogen-Deuterium Exchange on Defect Zinc Oxide.. . . . . . . . . . . . . . 56 IV. Adsorption of Oxygen and Oxidation Catalysis on Nickel Oxide.. . . . . . . . . 60 66 V. Electronic Properties of Nickel Oxide.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Carbon Monoxide Oxidation on Modified Nickel Oxide Catalysts. . . . . . . . . 68 VII. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 References. ....................................................... 72

Since in any catalytic reaction, at least one of the reactants must be chemisorbed, no theory of catalysis can be constructed before the mechanism of chemisorption on the surfaces under investigation is fully understood. I n recent years, as a result of the advances in solid-state physics, it became apparent that an important group of adsorbents and catalysts, namely the oxides of the transition metals, were typical semiconductors. This simple idea stimulated both research and speculation in the catalytic field. Its justification is quite simple, a t least in qualitative terms. Let us imagine a perfect surface of a perfect single crystal of zinc oxide : . . . - 0-- - Z n t t - 0-- - Zn++ - 0-- - . . . Suppose now that we want to chemisorb a molecule of hydrogen on this surface. For geometrical reasons, at least during the first stage (I) of the adsorption process, one hydrogen atom will form a bond with an oxygen ion while the second hydrogen atom will attempt to form a bond with a neighboring zinc ion. As the latter has no valence electron, the bond formed with hydrogen would be a one-electron bond, and presumably a weak one, were it not that the oxygen ion can form a strong bond with a proton t o give the hydroxyl ion. Thus it is reasonable to assume that hydrogen is chemisorbed as a proton on 0-- and as a negative ion on Zn++, the latter being now able to form a more stable two-electron bond:

-0-

- (ZnH)+ - (OH)- - Zn++ - 0--

(1) Conceivably the adsorption will not necessarily stop a t this stage. The (ZnH)+ complex may dissociate, the proton moving t o the next 47

48

G. PARRAVANO AND M. BOUDART

oxygen ion t o form another stable hydroxyl ion: -Zn++ - (OH)- - Zn - (OH)- - Zn++ - 0-- -

(11)

At this second stage there has thus been produced by chemisorption of a hydrogen molecule a neutral zinc atom which possesses a n interesting property. Its ionization energy in the crystal is much lower than for the isolated atom, as a consequence of the high dielectric constant of the crystal k . For a n isolated zinc atom, the ionization energy is 9.4 e.v. I n the crystal it is equal (1) to 9.4/k2. Since k for ZnO is about 10, the ionization energy is only 0.1 e.v. Thus a t moderate temperatures the neutral zinc atom produced a t the surface by the adsorption process looses its valence electron, which can wander about in the crystal quasifreely and, in particular, be accelerated by an external electric field. Thus the perfect zinc oxide crystal has become an impurity semiconductor, th a t is, a solid in which quasifree electrons can be produced by means of thermal energies as a result of impurities present in the crystal. Here the impurities are two hydroxyl ions of lower valency than the original oxygen ions. The link between chemisorption and semiconductivity, as illustrated by this example, was first clearly perceived by Wagner and Hauffe (2) in 1938. Whereas the production of a semiconductor b y chemisorption presents relatively little interest for our purpose, the reverse problem is currently receiving a great deal of attention. How is a given semiconductor going t o behave in chemisorption? Is it possible to relate semiconductor characteristics with catalytic properties and, if so, what are the properties of the semiconductor that have to be changed in order to modify and control catalytic activity? I n the preceding example it was deliberately assumed that the zinc oxide crystal used as starting material was perfect and thus, in particular, of stoichiometric composition. This was done solely in order t o show th a t the crystal became an impurity semiconductor following the adsorption process; otherwise the reasoning did not involve in any way the behavior of the bulk of the crystal, and the mechanism of the adsorption process did not depend upon the existence or nonexistence of semiconductivity prior t o the uptake of hydrogen. If this were a general situation, further examination of the effect of semiconductivity on chemisorption and catalysis would hardly seem profitable. However cases where semiconductivity may play a direct role can easily be imagined. Let us consider again the surface of a zinc oxide crystal but suppose now that the underlying crystal is nonstoichiometric and contains excess zinc atoms in interstitial lattice positions. A fraction of these defects is ionized and their valence electrons move quasifreely t,hrough the crystal. A molecule such as nitrous oxide, adsorbed on the surface, can make use

CHEMISORPTION AND CATALYSIS O N OXIDE SEMICONDUCTORS

49

of these “free valences” to reinforce its bond with a zinc ion. If the bond is strong enough the molecule may break into a nitrogen molecule escaping to the gas phase and an oxygen ion bound to the surface and forming a (ZnO)+ complex.

-zntt-

-

Na0 o---zn+i-O---Zn++-

Zn

+’

N*

Zn+

A second free electron of the crystal may eventually be captured by this complex, with formation of a still stronger bond in a neutral (ZnO) complex. This final situation may be also described as follows:

O--

I

Zn++ - 0-- - Zn++ - 0-- - Zn++ - 0-- -

(11)

Although it is clear from this picture that dissociative chemisorption of nitrous oxide may be easier on a zinc oxide semiconductor than on stoichiometric zinc oxide, it is much more difficult to assess the help of semiconductivity in catalyzing the decomposition of nitrous oxide. If the supply of free electrons from the interior is abundant, stage I will be quickly replaced by stage I1 and the oxygen may be so tightly bound in this second stage that its removal from the surface becomes the ratedetermining step. If the supply is limited, some of the adsorbed oxygen will not go beyond state I, and either its removal or the chemisorption of NzO may be rate determining, depending on the specific characteristics of the process. To conclude this brief introduction, it appears that there may be some relation between semiconductivity and chemisorption on oxide semiconductors. In catalysis a much more complex situation is expected and generalizations will be hard to come by. Indeed, it is well known that chemisorption is a necessary condition in heterogeneous catalysis but it is by no means a sufficient one. Furthermore, as was illustrated above, a given molecule can be chemisorbed in different ways on an oxide surface, and it will prove necessary to decide in each particular case which mode of adsorption is important in a catalytic process. The problems that have been sketched in the preceding paragraphs

50

G . PARRAVANO AND M. BOUDART

will be discussed in the light of available experimental evidence. The data have been selected with the idea of examining critically the various speculations or theories pertaining to the subject. Thus the present survey is incomplete in many ways but sufficiently broad, we hope, to serve as a guide to further work in the field.

I. CHEMISORPTION OF HYDROGEN ON ZINC OXIDE Because zinc oxide is a relatively well-understood oxide semiconductor, we shall first review its properties as a hydrogenation catalyst in the catalytic hydrogen-deuterium exchange reaction. Since the latter essentially measures the rate of reversible chemisorption of hydrogen a t equilibrium, data on the hydrogen chemisorption will be included in this survey. Any theory of hydrogen chemisorption on zinc oxide must explain all the following well-established facts. A sample of zinc oxide adsorbs hydrogen only after an activating pretreatment. This activation may be carried out either by heating the sample in oucuo around 400°C. or by treating it in hydrogen at a slightly lower temperature. Water vapor and carbon dioxide are evolved during the pretreatment (3). A suitably activated and evacuated zinc oxide adsorbent takes up considerable amounts of hydrogen. Below room temperature hydrogen is rapidly adsorbed in a reversible manner. At 77"K., a temperature that is some 2.3 times the critical point of hydrogen, and at 1 atm. pressure hydrogen covers 80% of the B.E.T. surface (4). Above room temperature a fraction of the hydrogen taken up is slowly adsorbed. This fraction increases with temperature. A zinc oxide surface saturated in hydrogen at 1 atm. at 300°C. and cooled down to 0°C. in hydrogen is covered with hydrogen t o the extent of 6 to 45% depending on the nature of pretreatment (4). The hydrogen taken up above room temperature is also reversibly adsorbed. The rate of the hydrogen uptake above room temperature cannot be described by any simple rate law. Empirically it was found that a logarithmic expression of the kind q

=

a In (1

+ bt)

where q is the amount of hydrogen taken up at time t, gives an adequate description of the rate process (5). This expression clearly indicates a slow, strongly decelerating velocity of adsorption, as if an increasing activation energy were required. An interesting feature of this slow adsorption was pointed out by Taylor (6) in a discussion of early data obtained on zinc oxide between 132" and 184°C. (7). From the temperature coefficient of the slow adsorp-

CHEMISORPTION AND CATALYSIS ON OXIDE SEMICONDUCTORS

51

tion in this temperature range it is possible to calculate the number of molecules of hydrogen striking 1 sq. cm. of surface/hr. with the required activation energy. This number was found to be 2 X loz1a t 184°C. Since 20 g. of oxide was used, from a preparation that is now known t o have a specific surface area larger than lo6sq. cm./g., and since at 184°C. only 7 cc. (N.T.P.) of gas was taken up by this sample during 1 hr., viz. about 2 X 1020molecules, we are led to believe that there exists a steric factor as small as 5 X for this adsorption process. Taylor concluded that the molecules receiving the necessary activation energy are first adsorbed on the surface and obtain the energy required f r o m the surface. Thus shortly after Taylor’s discovery of “activated adsorption” one of its most interesting features was made apparent. An additional difficulty in studies of slow chemisorption is evidenced by the Taylor-Liang technique of measuring adsorption isobars a t successively higher temperatures without pumping between runs, a technique that has been reviewed in an earlier volume of this series (8). The adsorption-desorption phenomena observed between room temperature and about 150°C. clearly show in this temperature range the existence on the surface of zinc oxide of two types of hydrogen chemisorption (9). Alternatively, these and similar data may be interpreted in terms of two types of surfaces on zinc oxide. This duality is again revealed by the isotopic method of Roginskii. Keier and Roginskii (10) report that when hydrogen is chemisorbed first on zinc oxide at low pressures and deuterium is subsequently admitted to the surface, the gas admitted last (in this case deuterium) is first evolved upon desorption while the species adsorbed first is desorbed last. It would be of interest to repeat this kind of experiment by use of a more refined technique, since it is somewhat surprising that no isotopic exchange took place in at least some of the experiments of Roginskii even though they were performed in a temperature range where exchange is known to proceed easily at higher gas pressures. The rate of slow chemisorption is sharply decreased on a surface of zinc oxide poisoned with water (11).This observation is in agreement with the necessity of activating zinc oxide before it becomes an adsorbent for hydrogen. The complexity of the adsorption process, in particular its duality as illustrated above as well as in more recent data of Wicke (12), also shows in the irregular behavior of zinc oxide as a catalyst for the hydrogendeuterium exchange (13). Thus this reaction proceeds at measurable rates a t temperature as low as 140”K., indicating that at least part of the low-temperature adsorption is of the dissociating type. The apparent activation energy a t low temperatures is low, but in the temperature

52

G. PARRAVANO AND M. BOUDART

range between 180" and 490°K. it changes rather erratically, its value being practically zero between 400" and 430"K., about 7 kcal./mole below this latter interval, and about 12 kcal. above it. The mechanism of the exchange seems to involve both the low- and the high-temperature types of chemisorption. In particular, the rate of the exchange at high temperature exceeds by a factor of 500-fold that of slow chemisorption measured separately. It is important to note that the principal features of hydrogen chemisorption, which are summarized above, apply equally well to other adsorbents than zinc oxide, for instance to chromium oxide. A satisfactory theory therefore must not depend on specific properties of zinc oxide. I n this connection let us recall the important experiments of Pace and Taylor (14) and Kohlschutter (15), who found th a t the slow rates of chemisorption of hydrogen and deuterium on chromium oxide, zinc oxide-chromium oxide, and nickel on kieselguhr, were identical within experimental error. It would be interesting to perform such a n experiment .on zinc oxide because it permits one t o make a decision on the nature of the slow activated step (16). Another observation which has recently been made on chromium oxide (17) emphasizes the difference between the two types of hydrogen chemisorption: it was found that hydrogen chemisorbed a t high temperature on chromium oxide does not exchange with deuterium a t a low temperature where the exchange reaction itself proceeds a t an appreciable rate. Before any further discussion of the implications of these adsorption characteristics, some pertinent properties of zinc oxide as a semiconductor will be reviewed.

11. ELECTRONIC PROPERTIES OF ZINC OXIDE The excellent behavior of zinc oxide as a catalyst in reactions requiring the dissociative chemisorption of hydrogen molecules appears somewhat surprising a t first glance. Indeed, zinc oxide is a diamagnetic crystal and its Zn++ ions have filled d-bands. It is now fairly well established (18) that unfilled d-bands in the electronic structure of transition metals are essential t o low-temperature chemisorption of hydrogen. The complete lack of activity of zinc metal films in the catalysis of the exchange reaction between NH, and DS has recently been recorded by Kemball (19) and agrees with this line of thought as the d-band of zinc metal is full. Onthe other hand a zinc oxide sample, prior t o its activation, is equallyinactive with respect t o hydrogen chemisorption. A zinc metal film however becomes a hydrogenating catalyst according to Kemball if after its preparation i t is purposely contaminated with oxygen. Conversely, in-

CHEMISORPTION AND CATALYSIS O N OXIDE SEMICONDUCTORS

53

active zinc oxide becomes active after a pretreatment in vacuo or in hydrogen. This contrast suggests that the active system is some sort of intermediate between zinc and zinc oxide, possibly nonstoichiometric zinc oxide. The existence of the latter has been recognized for many years, since Wagner (20) applied his thermodynamic theory of defect oxides to the system zinc oxide-oxygen (21). According to this scheme, a t sufficiently high temperature an equilibrium sets in between zinc oxide and oxygen in the gas phase, whereby excess zinc (Zni) can be accommodated in interstitial positions of the lattice: Zn2+

+ 02-

Zn,

+ 4502

If now the interstitial zinc atom loses one electron by thermal ionization of the lattice : Zni Z Zn,f e (2)

+

the conductivity U , assumed to be proportional to the concentration of free electrons due to (a),will depend on the oxygen pressure Pozbecause of (1) in the following manner: u = P02-5i

This relation was verified by von Baumbach and Wagner (21) and checked more recently by Bevan and Anderson ( 2 2 ) in an extended range of temperatures and pressures. The conductivity measurements show that equilibrium (1) sets in rapidly at temperatures as low as 500°C. Since the melting point of zinc oxide is about 2100°C. and accordingly its Tamman temperature about 9OO"C., the process under consideration cannot possibly involve the bulk of the crystal because defects could not diffuse rapidly enough through the lattice a t such low temperatures. Except a t very high temperatures, the defect equilibrium is realized only a t the surface of the crystal, that is, in a layer of a few unit cells' thickness. If a sample of zinc oxide is first heated in a vacuum a t 800°C. and then under low pressures of oxygen mm.) a t 608"C., its resistance increases slowly with time in a manner suggestive of a diff usion-controlled process. This shows, according to Bevan and Anderson, that oxygen is slowly taken up by the solid and traps conduction electrons. Slow changes of conductivity of this nature are especially important a t temperatures lower than 5OO0C., where equilibrium (1) cannot be reached in finite times even in the surface layer. Thus Fritzche (23) observed a slow increase of conductivity of a thin film of ZnO when it is heated in a vacuum a t 460°C. Fritzche's data are represented in Fig. 1 on a semilogarithmic diagram for reasons which will become apparent later. Ac-

54

G. PARRAVANO AND M. BOUDART

cording t o Fritsche, the slow variation of conductivity is due to a n activated diffusion of oxygen from the interior of the film to the surface, From these data a diffusion coefficient of oxygen in the oxide can be calculated in good agreement with values arrived a t by other methods. The conductivity of zinc oxide also increases in a hydrogen atmosphere. Thus Stockman (24) shows that the conductivity of ZnO heated in hydrogen at

log t [minutes]

FIQ.1. Time variation of electrical conductivity of zinc oxide film a t 460°C. in vacuum (data replotted from ref. 23).

All these data demonstrate that the surface layer of a zinc oxide crystal can be modified chemically by addition or subtraction of oxygen in a temperature range where the bulk of the material remains unaffected and by methods (heating in vacuo or in hydrogen) which are precisely those used in the “activating” pretreatment of a zinc oxide adsorbent or catalyst. There exists thus far very little information about the composition, structure, and energetics of such surface layers. From their conductivity data a t low oxygen pressures and in a vacuum, Bevan and Anderson conclude that the conductivity of a zinc oxide surface should be greater than that of the bulk phase. Deoxygenation of the surface in vacuo must give rise to surface atom layers that are richer in zinc than is the interior of the crystal lattice. This may be visualized as a consequence of equilibrium (1) or alternatively of a surface equilibrium of the type ZnO, Znt0) $502 zinc atoms being now chemisorbed on the zinc oxide surface. While such

+

CHEMISORPTION A N D CATALYSIS O N OXIDE SEMICONDUCTORS

55

a picture is rather satisfying insofar as it indicates that a catalytically active zinc oxide surface is some intermediate between zinc and zinc oxide, its consequences are rather unpleasant. Thus, a n active oxide catalyst must be considered as a “Zwischenzustand” supported on bulk oxide and not in equilibrium with the latter, in the sense of Huttig’s school ( 2 5 ) . I n spite of the low temperatures of chemisorption and catalysis, some equilibration between the surface phase and the bulk phase may proceed over long periods of time, the rate of this process being controlled by activated diffusion. Even if a n adsorbent or catalyst were activated above the Tamman temperature, its departure from stoichiometry as a function of depth below the surface may still vary as a function of the history of the cooling process. The inhomogeneity of an oxide semiconductor quenched from high temperatures has recently been demonstrated by Fritsche in the case of cuprous oxide (26). Finally, another consequence of this situation is the danger of attributing t o the surface layer properties which have been established for the bulk material. This applies in particular t o semiconducting characteristics of a defect oxide. Thus, with zinc oxide, the surface layer may be nearly stoichiometric and poorly conducting as a result of oxygen adsorption or conversely may present quasimetallic properties after activation in mcuo, irrespective of the composition of the bulk material. Pending further detailed information on the constitution of the surface layer, it is possible neverthelsss to get useful qualitative information on the relationship between defect structure of a given semiconducting oxide and its behavior as an adsorbent or catalyst. Thus a semiconductor can be modified by addition of controlled amounts of impurities. If it is then assumed that the direction of the modification is the same a t least qualitatively a t the surface and in the bulk, a comparison of oxides respectively unmodified and modified in opposite directions can reveal trends of interest as t o the requirements of a given surface reaction. This method was first tried by Wagner (27) and further amplified by one of us (28,28a). As mentioned earlier, in zinc oxide there exists a certain concentration of interstitial zinc in excess of stoichiometry as a result of equilibrium (1) and a corresponding concentration of free conduction electrons according to equilibrium ( 2 ) . If now a cation of higher valency than 2, such as Ga3+ is dissolved in the zinc oxide lattice and takes the place of a Zn2+ ion in regular lattice position, electrical neutrality can be maintained if a Zn,+ interstitial ion disappears while a free electron remains in the lattice. Thus solution of a trivalent oxide in zinc oxide decreases the concentration of interstitial excess zinc but increases the concentration of free electrons. This was verified by Wagner (27) for solutions of G a z 0 3in ZnO and by

56

G. PARRAVANO AND M. BOUDART

Hauffe and Vierk (29) for dilute solutions of AlzOa and GapOain ZnO. The conductivity, a t a given temperature, under a given oxygen pressure is higher for the modified samples than for the pure oxide. On the contrary, upon substitutional solution of a cation of valency lower than 2, electrical neutrality is maintained a t the price of the free electrons, whercas the concentration of interstitial zinc ions increases. Thus the conductivity of ZnO 1% LizO is lower than for pure ZnO, as also shown by Hauffe and Vierk. Since the rate of oxidation of zinc depends on the migration of interstitial zinc ions, foreign alloying elements oxidized to the trivalent state slow down the rate of oxidation a t a fixed temperature and oxygen pressure since they decrease the Concentration of interstitial zinc in the oxide layer. Conversely, a Zn-Li alloy is oxidized more rapidly than pure zinc in view of the larger concentration of interstitial ions (30). T o sum up, addition of monovalent or trivalent ions changes the defect characteristics of bulk zinc oxide in opposite directions. If it is shown t hat the rate of a catalytic reaction also changes in opposite directions on these modified catalysts, it can be concluded that the ratedetermining step of the reaction is closely related to the defect structure of the oxide, provided that a qualitative correspondence exists between modifications in the bulk and a t the surface.

+

111. THE HYDROGEN-DEUTERIUM EXCHANGE ON DEFECT ZINC OXIDE This method was applied successfully to the hydrogen-deuterium exchange reaction on both modified and unmodified zinc oxide samples. The results and their limitations will now be reviewed. Alumina-, galliaand lithia-containing samples of ZnO were prepared b y impregnation of the pure oxide with definite amounts of standard solutions of the corresponding nitrates so that the finished samples contained 1 mole % of the foreign oxide. These samples as well as the unmodified zinc oxide were sintered at 800°C. in air for 3 hr. Although it is very unlikely th a t the resulting catalysts were of homogeneous composition throughout, this circumstance is not essential to the validity of the demonstration, as has been emphasized above. The most important data of this work can be summarized a s follows by a representative example. On weights of catalysts, having the same B.E.T. surface areas, the following conversions (in percentage from equilibrium) were recorded a t 160OC. when a given Hz-D2mixture was flowed over them at a constant volumetric rate of flow : ZnO ZnO ZnO ZnO

+ Liz0

+ A1203 + GazOa

5%

25 % 40 % 65 %

CHEMISORPTION AND CATALYSIS ON OXIDE SEMICONDUCTORS

57

These figures prove a direct relationship between catalytic activity and semiconductivity for the reaction under study. Indeed, the amount of conversion is smaller on a sample of lower conductivity and larger on samples of higher conductivity. It is important t o note that the demonstration would have remained unconvincing if only the ZnO-LLO or the ZnO-GazOs had been investigated, besides the unmodified sample. Indeed, in the first case a decrease in the rate could have been ascribed t o some kind of "poisoning" by lithia and in the second case the beneficial action of gallia could have been attributed to a promoting effect, say, of a geometric type. Both effects together could however hardly be coincidential or fortuitous. The limitations of these data are numerous. First of all, they do not pretend t o show a proportionality between reaction rate and the concentration of conduction electrons which is not measured by a coilductivity experiment relative to the bulk material. The main obstacle to a deeper analysis of the rate data lies in the fact that the mechanism of the exchange is still obscure, as evidenced by a number of complicating effects. For this reason we refrained from comparing the apparent activation energies of the modified and unmodified catalysts although they lie in a regular sequence, the lowest value corresponding t o the gallia-containing sample (6.3 kcal./mole) and the highest one ( > 2 5 kcal./mole) to the catalyst with lithia. Some of the difficulties encountered in the work of Smith and Taylor (13) appeared again in this investigation. Thus Arrhenius plots of the kinetic data for the first-order rate constants per unit surface area do not yield straight lines for the entire temperature range studied (Fig. 2). For all the sintered catalysts, above a certain temperature in the neighborhood of 115°C. the apparent activation energy becomes zero, exactly as was found by Smith and Taylor on unsintered pure zinc oxide between 100" and 130°C. This break in the Arrhenius plot does not correspond to a transition between rate-determining steps controlled respectively by the surface process and by diffusion. Indeed, above the short temperature range of zero activation energy, the exchange picks up speed again, evidencing a temperature coefficient a t least as large as before. Moreover, even in the lower temperature range, where activation energies can be calculated, there is evidence for a compensation between activation energies and frequency factors (Fig. 3) in the usual manner. This alone would restrict considerably the merit of a comparison of activation energies t o assess catalytic activity. Further observations confirm the impression that in the temperature range of 0" t o 100°C. the mechanism of the exchange is complex and probably involves two types of hydrogen chemisorption. For instance, the rate data in this region are

58

G. PARRAVANO AND M. BOUDART

-2 -

I

l

l

22

23

24

25

26

27

28 29 30 31

+

x 104(0~-1) FIG.2. Arrhenius plots for the hydrogen-deuterium exchange reaction on sintered zinc oxide catalysts; OZnO, oZnO GazOl; VZnO A1203; mZnO LLO-all GapOs; QZnO-both vacuum activated a t hydrogen activated a t 350°C.; OZnO 450°C. (ref. 28a).

+ +

+

+

FIG.3. Relation between frequency factors and heats of activation for the hydrogen-deuterium exchange reaction on zinc oxide catalysts (ref. 28a).

CHEMISORPTION AND CATALYSIS ON OXIDE SEMICONDUCTORS

59

very sensitive t o the mode of pretreatment of the catalysts. On samples which have not been sintered and are activated in hydrogen, it is possible to obtain straight-line Arrhenius plots in an extended temperature range (25' t o 185°C. or more) without any break, in agreement with the results of Holm and Blue (31). Samples which have been sintered in air show the break in the Arrhenius plot even after a similar activation in hydrogen. Activation in vucuo, for both sintered and unsintered samples, leads to some irreproducibility of the rates, which tends to disappear after the samples are heated in the reacting mixture. The difference between activations in

10

0

I

I

0.2

I

I

I

I

I

I

0.4 0.6 0.8 LOG ( t +to) MOMS)

r

FIG.4. Activation of zinc oxide catalysts in hydrogen 0.119 cc./sec.: 0215°C.; A, X 231"C., 0230°C.

r

10 .

l

]

1.2

+ 2 % deuterium, flow rate

vacuo and in hydrogen suggests th at during the latter process, two phenomena take place: a "cleaning" of the surface with production of the highly defect surface phase, and slow chemisorption of hydrogen, which will take a part in the exchange process. It is interesting to note th a t the kinetics of the activating process obey the same empirical law as the change of conductivity with time of a zinc oxide film activated in vacuo (Figs. 4 and 1). This similarity provides additional evidence of the role of the defect surface phase in the exchange reaction. The effect of sintering, although not clear, is apparently similar: presumably, upon sintering, the surface becomes richer in zinc and this is reflected by a larger activity per surface area of the sintered samples.

60

G. PARRAVANO AND M. BOUDART

I n spite of the lack of quantitative evidence, it seems permissible to speculate on the mechanism of hydrogen chemisorption on zinc oxide in view of the proof just presented of a definite relationship between semiconductivity and the rate of the exchange. Because of the features of the activation process evidenced by both chemical and electrical data, it is clear that an active adsorbent or catalyst exposes both zinc and oxygen ions a t its surface. Both are essential to the adsorption of a hydrogen molecule since reoxidation of an active surface b y water or oxygen cuts down the hydrogenating ability of the surface, and also in view of the unreactivity of zinc metal toward hydrogen. Thus the naive picture sketched in our introduction seems t o be essentially correct. The relationship between semiconductivity and the rate of exchange makes it clear tha t conduction electrons (“free valences 1 1 ) take part in the adsorption process. Whether they do it or not will decide whether the adsorption is of the low-temperature kind (small heats of adsorption, low energy of activation) or of the high-temperature variety (large heats of adsorption, slow rate of adsorption). The adsorbing surface is consequently of the type first described by Volkenstein (32). Such a surface would have all the desirable properties. The high-temperature chemisorption would be of the “activated” type; the activation energy originates in the solid adsorbent, and Taylor-Liang effects are easily explained (16). At the same time two adsorption centers identical in all ways except for the presence or absence of a n extravalence electron would simulate surface heterogeneity. The consequences of this picture (16) have been pointed out before. We shall show now that the main features of hydrogenation oxide catalysts, as exemplified by zinc oxide, find their counterpart in the characteristics of oxidation catalysts of the type of nickel oxide. OF OXYGENAND OXIDATIONCATALYSIS IV. ADSORPTION O N NICKELOXIDE

I n the case of zinc oxide-a semiconductor with excess metal-hydrogen chemisorption brings about a surface reduction with consequent increase of the concentration of current carriers. Similarly, in the case of nickel oxide, oxygen chemisorption oxidizes the surface owing to the fact that oxygen molecules upon dissociation draw electrons from the adsorbent. As a result, trivalent nickel ions are produced and current, can flow owing t o charge exchange between neighboring pairs of the type Ni2+ - Ni3+. As before, however, this oxidation process need not stop a t this surface stage. Indeed, nickel oxide, as can many other oxides of transition metals, can accommodate a stoichiometric excess of oxygen in the bulk crystal (33). While the amount of this excess is not particularly large as compared

CHEMISORPTION

AND CATALYSIS

ON OXIDE SEMICONDUCTORS

61

with that in other oxide systems like Ti-0, V-0, Mn-0 (34), it can nevertheless be determined by chemical analysis. By use of a n acid solution of potassium iodide in the absence of atmospheric oxygen, a n accuracy of 1 X lo-& g. at. of oxygen/g. of oxide can be readily obtained. By this method Krauss measured the rate of oxygen uptake by nickel oxide (35) and a number of other oxides (36) a t 300"C., as well as the amount of oxygen taken up a t that temperature after 2 hr., when apparent saturation is reached. For the rate he found a parabolic law: q = kt'l"

q being the amount taken up at time t. The amount after 2 hr. was found to be proportional to the square root of the oxygen pressure. The rate equation indicates, according t o Krauss, a process controlled by oxygen diffusion, whereas the pressure dependence shows dissociation of the oxygen molecules. Krauss interprets his data as representing adsorption of oxygen in the interior of the oxide, and the quantity taken u p after 2 hr. as equilibrium amounts. Since, however, the melting point of nickel oxide is about 26OO0K.,its Tammann temperature should lie around 1000°C., and it is hard t o believe that ionic motion could be so rapid a t 300°C. as to reach equilibrium throughout the crystal in so short a time. A more recent investigation by Engell and Hauffe (37) confirms this impression. Engell and Hauff e measured oxygen adsorption volumetrically between 25" and 700°C. at constant pressures of oxygen between 30 and 200 mm. Hg. At room temperature and below 300°C. they were able to identify two kinds of rate processes for the uptake of oxygen. There is first a relatively rapid uptake lasting approximately 10 min. The quantity of oxygen taken up during that time decreases with increasing temperature and decreasing pressure, as shown in Table I. TABLE I after 10 Min. by 68.9 g. of NiO (Surface Area 61 " t [ . * ) (Ref. 39)

Quantity of Oxygen (cc. N T P ) Taken

U,3

t,T.

25

75

200

mm. Hg 30 80 98 197

0.3 0.6 0.8

0.4

0.2 -

-

-

After this fast process there follows a smaller slow uptake increasing with temperature and insensitive t o pressure. Thus the amount taken

62

G . PARRAVANO AND M. BOUDART

up between 10 and 100 min. a t 25"C., 98 mm. Hg, is about 0.02 cc.; a t 200°C. between 20 and 500 min. (60 mm. Hg) it amounts to about 0.1 cc.; between 300" and 700°C. there seems to be no separation between the two proccsses, the quantity taken up during the first one being presumably too small t o be measurable. Thus at these temperatures the rate process obeys a single equation: q = a In (1 bt)

+

which was already mentioned above in connection with the slow rate of adsorption of hydrogen on zinc oxide. As a matter of fact there is a striking parallelism between both systems, H2-ZnO and 02-Ni0. Thus the adsorption isobar (60 mm. Hg) corresponding t o the amount of oxygen taken u p after 100 min. b y NiO between room temperature and 700°C. has the characteristic shape of hydrogen adsorption lobars: a curve with a flat minimum between 200" and 300°C. (37). The analogy goes further if the activity of NiO as an oxidation catalyst is examined closely. I n the case of the H2-D2 exchange reaction on ZnO, there was an activation period corresponding t o a surface reduction. For NiO, there is a deactivation period also corresponding t o surface reduction during the catalytic oxidation of carbon monoxide. This deactivation was first observed, with NiO, by Roginskii and Tselinskaya (38) and confirmed in a detailed study by one of us (28). I n the latter investigation, which started with a fresh sample of nickel oxide fired in air a t 640" and then slowly cooled in air to room temperature, i t became clear that the catalyst was undergoing a n irreversible change during the course of carbon monoxide oxidation. After a number of runs the activity of the catalyst reached a minimum value which could be considered as constant. Two sharply different stages were recognized: an initial stage (a) characterized by high rates and rapidly diminishing activity, followed by a stage (b) of constant activity. Data for stage (a) were found t o obey a kinetic law of the type q

=

a In (1

+ bt)

where q is the amount of conversion a t time t. These kinetics are of the same general form as those observed for the activation of ZnO and the slow rate of adsorption on ZnO and NiO. It is very significant for the meaning of this deactivation process th at stage (a) could be eliminated by pretreatment of a catalyst with carbon monoxide. After such a pretreatment a t 106°C. rates of the catalytic reaction were measured and found to be very close to those obtained on a n untreated catalyst after prolonged use. Consequently, it appears that a fresh catalyst contains

CHEMISORPTION AND CATALYSIS ON OXIDE SEMICONDUCTORS

63

excess oxygen endowed with a high reactivity and that this excess oxygen is removed during a carbon monoxide preheating or the catalytic process. Confirmation of this view is offered by the fact that a catalyst fired in helium is less active initially than a catalyst fired in air. Just as in the case of the H2-D2exchange on ZnO, two mechanisms are also discernible for the carbon monoxide oxidation [stage (b)] on nickel oxide below 300°C. There is a low-temperature mechanism operative between 100' and 180°C. characterized by a low activation energy of 2 kcal./mole and a high-temperature mechanism, above 180"C., with a higher activation energy of 13 kcal./mole. The kinetics are different and are respectively: r = k p 0 z 0 ~ 6 p c , 0 ~ 5 (106' - 174°C.) r = k p ~ , ~ ~ ~ p , ,(205" - 222") Carbon dioxide was found to have no effect on the rate, r , in the higher temperature range. Similar information is missing for the low-temperature mechanism. Let us indeed represent the over-all oxidation process b y the two steps (X)O co = ( S ) c02 (i> (8) $ 8 0 2 = (S)O (ii)

+

+

+

where S designates a bare surface site. If the removal of surface oxygen is rate determining, the rate will be given by

r = k1p,,8 where 0 is the fraction of surface covered with oxygen. Since (i) is the rate-determining step, 8 will be given by the equilibrium adsorption isotherm corresponding t o (ii) :

or approximately 8 = (bpo2')m

where 0

< m < 1. Thus finally r

=

k'pco(p~z)m/2

This expression agrees with the observed one for the high temperature rate, for m = 0.5. On the other hand, i t must be appreciated from the work of Engell and Hauffe (37) that below 200°C. th e surface of nickel oxide is covered

64

G. PARRAVANO A N D M. BOUDART

with appreciable amounts of loosely bound chemisorbed oxygen. I n particular, it is t o be noted that the change over from one mechanism to the other occurs in a temperature interval th at coincides with the minimum in the adsorption isobar. I n the lower temperature range, carbon monoxide may be oxidized by the loosely bound chemisorbed oxygen. Since the coverage of the surface by the latter is still small, as shown by the adsorption data, one may write simply for the rate expression of the reaction a t the surface following Langmuir:

or approximately r

= lirr2p02M p,,"

This agrees with the experimentally found equation for n = 0.5. These rate laws were formerly derived on the assumption of surface heterogeneity according to the scheme first put forward by Temkin and Pyzhev (39) for the ammonia synthesis. As shown by one of us (40) the assumption of heterogeneity is unnecessary for the derivation of the rate expression even in t ha t case. Presumably, in the low-temperature range the adsorbed oxygen readily migrates over the surface since its heat of adsorption must be rather small. I t ' m a y be called mobile oxygen in contrast to extralattice oxygen, which designates the tightly bound oxygen at higher temperatures. The existence of these two types of adsorbed oxygen was first perceived by Garner, Gray, and Stone (41) in their study of oxygen adsorption on cuprous oxide films and demonstrated subsequently by the Bristol workers by means of magnetic susceptibility measurements (42) and heats of adsorption of oxygen on nickel oxide (43). If it is assumed that the mobile oxygen differs from the extralattice oxygen by the absence of an additional electron supplied by the solid, it is quite likely that modifications of the electronic levels of nickel oxide by impurities will not affect substantially the low-temperature rate of carbon monoxide oxidation. Indeed, the rate depends on surface diffusion with subsequent reaction of the adsorbed partners if our scheme is correct. On the contrary such modifications might affect the rate of the high-temperature process insofar as it depends on the availability and heat of adsorption of the extralattice oxygen. As will be seen later, this prediction is correct. The importance of extralattice oxygen receives considerable support in the work of Krauss (36), who measured the yield of nitrous oxide as a function of time during catalytic flow oxidation of ammonia a t 30OoC. on the oxides of nickel, cobalt, iron, and manganese. By titrating a t the

AND CATALYSIS

CHEMISORPTION

ON OXIDE SEMICONDUCTORS

65

same time the excess oxygen adsorbed by these catalysts, Krauss was able t o find a most remarkable straight proportionality between the excess of oxygen and the percentage of NzO in the exit gases. This relationship is valid for a given oxide as it is “activated” by the reaction, i.e., as the amount of excess oxygen slowly increases with time of reaction toward its steady-state value. It is also valid for all oxides studied irrespective of their chemical constitution (Fig. 5 ) . As was mentioned above, the excess oxygen titrated by Krauss must be considered as belonging t o a surface layer which must be quite thin.

40 -

30-

2

4

6

0 0

MnO

x v

Ni 0 Fe,03 (460Ocl

coo

8

(O/MeO) x IO~gr,atorn/gr] FIG.5. Steady-state concentration of oxygen in different oxides during ammonia oxidation to nitrous oxide (ref. 36). .

At higher temperatures, as in the investigation of Engell and Hauffe, this excess oxygen must be diffusing slowly to the interior. T h a t such a diffusion is possible at not too elevated temperatures is shown by the work of Allen and Lauders (44), who report that a sample of nickel oxide labeled with 01*came to 15% of isotopic equilibrium with ordinary gaseous oxygen in 4 hr. a t 650°C. Further work with the oxygen isotope exchange seems indicated in order to clarify these problems. To sum up, the available data relative t o oxygen adsorption or catalysis on nickel oxide show the existence of two types of chemisorbed oxygen, one of them being related to the ability of this oxide to accommodate excess oxygen. The evidence concerning this latter property will now be reviewed with emphasis on the defect structure of bulk nickel oxide.

66

G. PARRAVANO A N D M. BOUDART

V. ELECTRONIC PROPERTIES OF NICKEL OXIDE The physical and chemical properties of nonstoichiometric nickel oxide have been studied by a large number of investigators. The very existence of stoichiometric nickel oxide has even been questioned by Klemm and Hass (45), who found th at the paramagnetic susceptibility of a n oxide of composition NiOI.026changes slightly when its excess oxygen is progressively removed down t o an amount corresponding to NiOl.Oo6.When this composition is reached, there is a sharp increase in the susceptibility, which becomes dependent on field strength. This seems to indicate phase instability giving way t o metallic nickel and nickel oxide with excess oxygen. Nickel oxide shows antiferromagnetism with a Curie temperature of around 240°C. (46). In this temperature range anomalies are present in the heat capacity and in the thermal expankon, and a change in crystal structure occurs (47). However no effect of this transition on catalytic properties has been detected. Possibly a t the Curie temperature there is only an anomaly in the temperature coefficient of the catalyzed reaction (48). The antiferromagnetism of nickel oxide indicates exchange forces between neighboring nickel ions through an intervening oxygen ion. The ' mechanism of electrical conductivity in nickel oxide may well depend on the nature of these forces. Indeed, nickel oxide is not a semiconductor to which a simple band model can be applied t o account for its lack of conductivity when nearly stoichiometric. If the Ni2+ ( d s ) ions with their unfilled d shell formed a wide band, the latter would also be unfilled and conductivity would be expected. I n nonstoichiometric crystals of nickel oxide there exists a certain concentration, determined by the excess oxygen of Ni+3 ions, and conductivity becomes possible by charge transfer between adjacent Ni+2-Ni+3pairs; however, as shown by Morin (49) this transfer is not a n easy one. Thus the mobility of the carriers is quite small and requires a n activation energy which for large concentrations of defects is about 0.1 e.v. According t o Morin, occupied and vacant d levels of Ni2+ form narrow wavy bands and act as donors and acceptors of electrons respectively. These levels are situated between the full valence sp band of oxygen ions and the higher vacant s p band of the latter. Because mobility is higher for the donor levels than for the acceptor levels, nickel oxide behaves as a p type of semiconductor; th at is, current is carried b y positive holes. Since no Hall effect is observed, owing to the small mobility, the p character of nickel oxide is shown by the positive sign of the Seebeck effect (thermoelectric power &, in volts per degree). For narrow bands or localizedlevels the kinetic energy term in the Seebeck effect does not contribute

CHEMISORPTION AND CATALYSIS ON OXIDE SEMICONDUCTORS

67

to the total effect, and for conduction by holes alone one has simple (49)

QT

=

Ef

(3)

where T is the absolute temperature and E f is the Fermi level. The role of excess oxygen in the conductivity of nickel oxide was first demonstrated by Wagner (50), who studied the conductivity of nickel oxide a t high temperatures (800"-1000°C.) as a function of the pressure of a n oxygen atmosphere coming t o equilibrium with the bulk of the solid. The equilibrium may be written 2Ni2+

+ 3502

2Ni3+

+ 0--

(9

It must be realized that actually for each oxygen ion built into the lattice, according t o (i) a vacant lattice site must be created in the sublattice of nickel ions. This is due to the geometrical impossibility of accommodating excess oxygen in the lattice. Excess oxygen really means nickel deficiency. More complex notations than the notation used here are necessary t o deal with this situation (51) but for our purpose we need not go into this. If now the ionization equilibrium Ni3+ g Ni2+

+ e+

(ii)

is taken into account, it is seen from (i) and (ii) that the conductivity, which is proportional to the concentration of positive holes e+, should depend on oxygen pressure as follows: u CI: P ~ , "

This relation was verified by Wagner et al. (50). The conductivity of nickel oxide may also be changed by dissolved impurities as shown b y Verwey et al. (52). Let us consider what happens when increasing amounts of lithium, the ion of which has the same radius as Ni2+ ion ( 0 . 7 8 k ) , are dissolved into nickel oxide. First, the lithium ions will fill up the vacancies in the nickel ion sublattice. During this process a corresponding number of Ni3+ ions, which have a smaller radius than Ni2+,are reduced to the Ni2+ oxidation state. Consequently, the lattice must expand and the conductivity decrease. When all vacant lattice sites are filled up, further addition of lithium can be achieved only by substitution of NiZ+by Li+. Then for each Li+ introduced, a Ni2+ion must be oxidized to the Ni3+ state, and so the lattice must now contract again and the conductivity increase. This inversion in physical properties has been observed for lattice size (53), conductivity (54), and thermoelectric power (55). Similar but opposite behavior should be expected when trivalent ion such as Cr3+ is introduced into the NiO lattice. Only the second stage of this process has been observed so far (56).

68

G . PARRAVANO AND M. BOUDAliT

The possibility of an inversion of physical properties in nickel oxide is intimately connected with its ability to accommodate excess oxygen. Another consequence of this state of affairs, t o which we will refer later, is that before the inversion the conductivity of NiO containing foreign ions must still depend on the oxygen pressure. After the inversion no dependence of conductivity on oxygen pressure should be observed (56). Let us note t hat substitution of oxygen ions by a n ion of lower negative charge, e.g. C1-, is equivalent, from the conductivity viewpoint, to addition of a cation of positive charge higher than 2. Finally, the self-consistency of this general scheme is well supported by the equivalence between defect concentrations of impurity-containing nickel oxide as determined chemically and electrically (19,55). The method outlined above in the case of zinc oxide will now be applied t o the carbon monoxide oxidation on nickel oxide catalysts modified in both ways. If it is assumed, as before, that semiconductivity trends in the bulk and in the surface layer are qualitatively the same, a correlation between semiconductivity and catalysis will he established if cationic impurities of valences lower and higher than 2 are found to affect the catalytic rate in opposite directions.

VI. CARBONMONOXIDE OXIDATION ON MODIFIED NICKEL OXIDE CATALYSTS Such a proof of the carbon monoxide oxidation was first given by one of us (28). It is very important to remark that the catalysts containing impurities were prepared by firing together in air at 600°C. for 3 hr. a mechanical mixture of the required components in adequate proportions. As pointed out by Fensham (54), this is much too low a temperature to ensure homogeneous solid solutions. Consequently, when a catalyst is described as NiO 0.1% Li20, there is no assurance that this nominal composition is realized at all either in the surface layer or in the bulk of the sample. As will be shown, this reservation is quite essential. The main features of this investigation are the following. First, a “deactivation” process similar to that observed on the pure nickel oxide was found on the modified catalysis as well, with the same logarithmic law t o represent its evolution with time. Second, the kinetic equations which were found t o fit the data on pure nickel oxide also apply t o the modified catalysts. Thus there is a low-temperature mechanism operative between 100” and 180°C. For all the samples assembled in Table 11, the activation energies were practically the same, about 2 kcal./mole and essentially equal to the value for pure nickel oxide. This indicates that, for this particular mechanism of the reaction, the added ions and the semiconductivity changes do not affect directly the catalytic process.

+

CHEMISORPTION

AND CATALYSIS

ON OXIDE SEMICONDUCTORS

69

TABLE I1 Aetioation Energies for Carbon Monoxide Oxidation on Nickel Oxide Catalysts Catalyst NiO NiO NiO NiO NiO NiO NiO NiO

+ 0.01 mole % WOa

+ 0.01 mole % Crz03 + 1 mole % NiClz + 0.01 mole % CezOz + 0.01 mole % AgzO + 1 mole % AgpO + 0.01 mole % I&O

Temp. range, "C.

Activation energy, kcal./mole

180-220 160-220 160-220 160-220 180-250 180-250 225-260 230-280

6.5 7.9 8.2 8.9 13.7 14.9 17.5 18.0

This again confirms the reasoning presented above, according to which loosely adsorbed mobile oxygen which does not exchange electrons with the catalyst, or does so in a restricted sort of way, is the essential partner for the surface reaction. On the contrary, a t about 180°C. the various impurities directly affect the catalytic process by changing its overall activation energy. Moreover, as shown in Table 11, the activation energies are higher than for pure nickel oxide when the added impurity is a cation of change lower than 2. They are lower than for pure nickel oxide for impurities with a positive change higher than 2 or a negative change lower than 2 . It can be concluded that the rate-determining step of the process depends on the changes in the population of electronic levels due to impurities of both kinds. From the kinetic analysis given above, this rate-controlling step is the reaction of adsorbed oxygen with carbon monoxide striking it from the gas phase. The activation energy according t o this scheme ought to depend on the details of the electron transfer process between oxygen and the catalyst. More recently, Schwab and Block (57) have studied again the same reaction on nickel oxide containing both chromium oxide and lithium oxide. Their observations are diametrically opposed t o those just discussed above. They too find that these impurities change the activation energy in opposite directions; however they find that the activation energy is lower (12 kcal./mole) on the lithia-containing samples and higher (19 kcal./mole) on the samples with chromia. For this discrepancy, which appears surprising a t first, the following explanation is proposed. There are four differences between Schwab's work and the previous investigation. First, the samples in Schwab's work were fired at a higher temperature, vie., 850°C. Second, much larger quantities of impurities were used, up t o 5 mole % compared with 0.01% in the original work.

70

G . PARRAVANO AND M. BOUDART

Third, the kinetic law shows no dependence on oxygen partial pressure but is straight first order in carbon monoxide partial pressure. Fourth, the temperatures a t which the catalytic investigation was carried out were substantially higher, between 250" and 450°C. When all these contrasting particularities are taken into account, the opposite results of both investigations can be reconciled rather simply. It appears that the first study concerned itself with catalytic properties of modified nickel oxide before the inversion point of their physical properties. I n Schwab's work, however, the modified catalysts were on i

400

500

600

700

800

90(

T [OK1

FIG.6. Variation of Fermi level E l in nickel oxide as a function of temperature, nature, and concentration of additions to nickel oxide (ref. 55).

the other side of the inversion point. Thus in the mechanism of Schwab the adsorbed oxygen plays no role: the carbon monoxide is presumably oxidized by lattice oxygen, differing in its binding energy from extralattice or adsorbed oxygen. Indeed, the conductivity results recalled above suggest that below the inversion point excess oxygen still plays a role in the structure of the modified oxide. This is not so above the inversion point, where in particular the semiconductivity is independent of oxygen pressure. Confirmation of these views is offered by recent observations by one of us (55) on the Seebeck effect (Fig. 6) in pure nickel oxide and in nickel oxide containing respectively 0.2 (curve l), 1 (curve 2), and 2 (curve 3) atoms of Li per cc. It is seen th a t the Fermi level of these samples, calculated by means of Equation (3), is lowered by addition of lithium except for the sample with the low lithium content where at low temperatures a small but distinct anomaly is measured. For that sample,

CHEMISORPTION A N D CATALYSIS ON O X I D E SEMICONDUCTORS

71

below 450"K., the Fermi level of the modified oxide lies above that of the pure oxide. This conforms with out interpretation th a t considers that particular sample as not having reached the inversion point of physical properties. Presumably, the same consideration might also explain the trend recorded for the other modified oxides of Table 11. The reason why the inversion point has not been reached in those samples containing nominally up t o 1% foreign oxide might be the low temperature a t which they have been fired together (600°C.); whereas Fensham (54) finds that a temperature of 1100°C. has t o be reached before homogeneous solution of a foreign oxide into nickel oxide can be accomplished. Schwab's work also indicated th at a t higher temperatures than those used in the Princeton work lattice oxygen takes part in the oxidation of carbon monoxide. This leads us t o recognize three different types of surface oxygen (mobile, extralattice, and lattice) of increasing binding energy. They might be identified tentatively with 0, 0-, and 0-- respectively. It is also known from Nakata's studies with labeled oxygen (58) that a t high temperatures (300°C.) lattice oxygen takes an important part in the oxidation of carbon monoxide. Similar inversion of catalytic properties on nickel oxide catalysts containing increasing amounts of lithium oxide has been also found by Hauffe, Glang, and Engell (59) with respect t o the decomposition of nitrous oxide. These authors, who fired the oxide mixtures a t 9OO"C., showed that below 500°C. the sample containing nominally 0.1 mole % Liz0 was more active than pure N i 0 and samples containing larger amounts of LizO. A minimum value for the apparent activation energy was also noted. If further work confirms our explanations which connect catalytic inversion with the inversion of physical properties of the modified nickel oxide catalysts, the correlation between semiconductivity and oxidation catalysis found in the Princeton work and in Schwab's studies will appear quite convincing. T o sum up, the activation energy of the carbon monoxide oxidation has been found t o decrease with increasing semiconductivity on both sides of the inversion point of physical properties of nickel oxide catalysts. Let us emphasize again that this correlation does not pretend t o establish a proportionality between semiconductivity and catalytic activity. It shows however that certain changes in electronic structure of a solid which can be brought about by selected impurities are reflected upon both the conducting and catalytic properties of a solid. A deeper analysis is required if the direction of, or the change in, catalytic activity is t o be predicted on a rational basis. I n the last analysis these changes

72

G . PARRAVANO AND M. BOUDART

depend on the strength of the chemisorption bonds, which in its turn depends on the characteristics of electron transfer between adsorbent and adsorbate. Although the reality and catalytic importance of this electron transfer must be considered as definitely proved by the critical experiments reviewed above for oxides of two opposite types, the details of this process are still uncertain, largely owing to our ignorance of surface states. While it is tempting t o speculate further on the direction of electron transfer and the role of the Fermi level in adsorption and catalysis, we shall not pursue this subject any further at this stage for a number of reasons. First, changes in Fermi level are expected t o modify the heats of chemisorption, but no prediction can be made on the activation energy for adsorption from this thermodynamic property of the system unless additional assumptions are made t o correlate heats and activation energies of adsorption. Second, conditions in surface layers may be extremely complex, depending on rates and conditions of quenching and on the formation of boundary layers of the Mott-Schottky type. Third, the discrepancy between results obtained on apparently similar samples by different investigators indicates that further speculations are premature before the differences between the catalytic samples and their precise definition are better understood. VII. CONCLUSION In spite of the shortcomings of the available data, the advances made can be summarized as follows, on the basis of our discussion: 1. It is clearly recognized that on oxide semiconductors various types of chemisorption can and do occur as a result of various types of electron exchange between adsorbent and adsorbate. Slow rates of adsorption may be due t o the conditions of this exchange. The logarithmic rate law, however, seems t o represent a number of different processes (bulk or surface diffusion, “activation ” or “deactivation” of catalytic surfaces, chemisorption). It appears futile to explain this empirical relation in terms of a unique mechanism. 2. Catalytic behavior of semiconducting oxides has been modified in opposite directions by the adfition of impurities which modify their electrical characteristics in opposite directions. Further work is required to put this qualitative correlation on a quantitative basis.

REFERENCES 1. Mott, N. F., and Gurney, R. W., “Electronic Processes in Ionic Crystals,” p. 83. Oxford, New York, 1948. 2. Wagner, C., and Hauffe, K., Z. Elektrochem. 44, 172 (1938). 3. Taylor, H. S., and Kistiakowsky, G. B., J . Am. Chern. SOC.49, 2468 (1927).

CHEMISORPTION AND CATALYSIS ON OXIDE SEMICONDUCTORS

73

Taylor, H. S., and Liang, S. C., J . A m . Chem. Sac. 69, 1306 (1947). Taylor, H. A., and Thon, N., J . Am. Chem. SOC.74, 4169 (1952). Taylor, H. S., Trans. Faraday SOC.28, 131 (1932). Taylor, H. S., and Sickmann, D. V., J . Am. Chem. SOC.64, 602 (1932). 8. Taylor, H. S., Advances in Catalysis 1, 1 (1948). 9. Taylor, H. S., and Strother, C. O., J . Am. Chem. SOC.66, 586 (1934). 10. Keier, N. P., and Roginskii, S. Z., Irvest. Akud. N a u k S.S.R. Otdel. Khim. Nauk 27 (1950). 11. Burwell, R., and Taylor, H. S.,J . Awl. Chem. Sac. 68, 1753 (1936). 12. Wicke, E., 2. Elektrochem. 63, 279 (1949). 13. Smith, E. il., and Taylor, H. S., J . Am. Chem. Sac. 60,362 (1938). 14. Pace, J., and Taylor, H. S., J . Chem. Phys. 2, 578 (1934). 15. Kohlschutter, H. W., 2. physik. Chem. A170, 300 (1934). 16. Boudart, M., Taylor, H. S., and Farkas, L., Memorial Volume, Research Council Isruel Pribl. Mem., Vol. 222 (1952). 17. Voltz, S. E., and Weller, S., J . AWL.Che7n. SOC.76, 5227, 5231 (1953). 18. Dowden, D. A., J . Chem. SOC.1960, 242; Couper, A , , and Eley, D. D., Nature 164, 578 (1949). 19. Kemball, C., Proc. Roy. Sac. (London) A214, 413 (1952). 20. Wagner, C., 2. physik. Chem. B22, 181 (1933). 21. Baumbach, H. H. von, and Wagner, C., 2. physik. Chem. B22, 199 (1933). 22. Bevan, D. J. M., and Anderson, J. S., Discussions Faraday Sac. No. 8, 238 (1950). 23. Fritzche, H., 2. Physik 133, 422 (1952). 24. Stockmann, F., 2. Physik 127, 563 (1950). 25. Huttig, G. F., Discussions Faraday SOC.No. 8, 215 (1950)26. Fritzche, H., 2. Physik 167, (1954). 27. Wagner, C., J . Chem. Phys. 18, 6 (1950). 28. Parravano, G., J . Am. Chem. SOC.76, 1448, 1352 (1953). 28a. Molinari, E., and Parravano, G., J . Am. Chem. SOC.76, 5233 (1953). 29. Hauffe, K., and Vierk, A. L., 2. physik. Chem. 196, 160 (1950). 30. Gensch, C., and Hauffe, K., 2. physik. Chem. 196, 427 (1950). 31. Holm, V. C. F., and Blue, R. W., I n d . Eng. Chem. 44, 107 (1952). 32. Volkenstein, F. F., Zhur. Fiz. K h i m . 27, 159 (1953). 33. Le Blanc, M., and Sachsse, H., 2. Elektrochem. 32, 58, 204 (1926). 34. Klemm, W., Naturwissenschaften 37, 172 (1950). 35. Krauss, W., and Neuhaus, A , , 2. physik. Chem. B60, 323 (1941). 36. Krauss, W., 2. Elektrochem. 63, 320 (1948). 37. Engell, H. J., and Hauffe, K., 2. Elektrochem. 67, 773 (1953). 38. Roginskii, S. Z., and Tselinskaya, T. S., Zhur. Fiz. K h i m . 22, 1350 (1948). 39. Temkin, M. I., and Pyzhev, V., Acta Physicochim. U.R.S.S. 12, 327 (1940). 40. Boudart, M., Ind. chim. Belge 19, 489 (1954). 41. Garner, W. E., Gray, T . J., and Stone, F. S., Proc. Roy. Sac. (London) A177, 314 (1949). 42. Fensham, P., Thesis, Bristol, 1952. 43. Dell, R. M., Stone, F. S., Trans. Faraday SOC.60, 501 (1954). 44. Allen, J. A., and Laudere, I., Nature 164, 142 (1949). 45. Klemm, W., and Hass, K., 2. anorg. allgem. Chem. 219, 82 (1934). 46. Bizette, H., J. phys. radium 12, 161 (1951). 47. Rooksby, K., Acta Cryst. 1, 226 (1948). 48. Parravano, G., J . Am. Chem. Sac. 76, 1497 (1953). 4. 5. 6. 7.

74 49. 50. 51. 52.

G. PARRAVANO AND M. BOUDART

Morin, F. J., Phys. Rev. 93, 1199 (1954). Wagner, C., 2. physik. Chem. B22, 181 (1933). Rees, A. L. G., “Chemistry of the Defect Solid State.” Wiley, New York, 1954. Verwey, E. J. W., Haaynian, P. W., and Romeijn, F. C., Chem. Weekblad 44, 706 (1948).

Brownlce, L. D., and Mitchell, E. W. J., Pror. Phys. SOC.(London)B66,710 (1952). Fensham, P., J . Am. Chem. Soc. 76, 969 (1054). Parrsvano, G., J . Chem. Phys. 22, 5 (1954). Hauffe, K., Advanres in Metal Phys. 4, 71 (1953). Schwab, G. M., and Block, J., Reunion chim. phys. Paris (1954); Z . physik. Chem. [N.F.] 1, 42 (1954). 58. Nakata, S., J . Chem. Soc. Japan. 63, 41 (1942). 59. Hauffe, K., Glang, R., and Engell, H. J., 2. physik. Chem. 201, 223 (1952). 53. 54. 55. 56. 57.

The Compensation Effect in Heterogeneous Catalysis E. CREMER Physik.-Chem. Institut der Universitat, Innsbruck, Austria

Page I. Introduction. . . . . . . ........................................... 75 11. Experimental D a t a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 111. The Interpretation of the Compensation Effect.. . . . . . . . . . . . . . . . . . . . . . . . 80 1. Basic Relations.. . . . . . . . . . . . . . . . . . . . . . . . . a. Relations between Enthalpy and Entropy b. Compensation Effect Due to a Tunnel Effect 2. Simultaneous Occurrence of Reactions on Different Activation Energies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 IV. Apparent Compensation Effects.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 V. Concluding Remarks. . . . .... ... ... . . . . . . . 90 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

I. INTRODUCTION The temperature dependence of the equilibrium of a reaction can be represented by X = C exp ( - A H o / R T ) (1) where X is the equilibrium constant and A H o is the standard enthalpy of the reaction, assumed to be independent of temperature. Likewise the temperature dependence of the rate of a reaction involving the rate constant x may be represented by

x

=

A exp ( - A E I R T )

(2)

where AE is the activation energy and A is the frequency factor. Equation (1) may be applied t o the equilibrium between vapor and liquid of a pure substance ( X = vapor pressure) or to the equilibrium between an ideal dilute solution and the pure phase of a solute ( X = solubility) or t o the equilibrium of a chemical reaction ( X = equilibrium constant). Equation (2) may be used for the rate constant k of a chemical reaction or applied t o the diffusion coefficient in liquid or solid phases or to the fluidity of liquids (reciprocal of dynamic viscosity) or to the specific electrical conductivity of semiconductors. Equation (1) may be also written as

X

=

exp (AX"/R) exp ( - A H " / R T ) 75

(3)

76

E. CREMER

where AS” is the standard entropy change of the reaction. AS” is related t o C of Equation (1) by C = exp (AS”/R) According t o Eyring’s absolute rate theory (I), in which there is postulated the establishment of a quasiequilibrium between the reactants and a n “active complex,” the rate constant may be expressed as

k

=

z = ( k T / h ) exp ( A S S / R )exp ( - A H ’ / R T )

@a)

where ASS and AHS, respectively, are the entropy difference and the enthalpy difference between the activated complex and the reactants, lc is the Boltzmann constant, and h is the Planck constant. AHS = AE if one single reaction is considered. In regard to various equilibria or processes characterized by different values of AH (AH’ or A H S , respectively) the following possibilities may be anticipated: Case 1: A H 1 > A H z > A H e . . . ; AS1 = A S z = AS3 . . . Case 2: AH1 > A H z > A H e . . . ; AS1 < A S z < AS3 . . . Case 3: A H 1 > A H z > A H e . . . ; AS1 > A S , > AXB . . . . Case 3 will be of special interest in this paper. It is encountered in all the examples listed above for Equations ( I ) and (2). Especially in heterogeneous catalysis it shows what has been called the “compensation effect” (C.E.). This term indicates that an increase in the enthalpy of activation AHS frequently has not the cxpected result of a considerable decrease in the rate constant, because there occurs a simultaneous increase in the entropy of activation ASs or of the frequency factor A , which compensates partly or entirely for the change in the exponent ( A H $ / R T , or A E I R T ) . 11. EXPERIMENTAL DATA The first example of a C.E. in heterogeneous catalysis was reported by Constable (2), who found a variation of A E from 20 t o 24 kcal. accompanied by a change in A by a factor of about 10 for the dehydrogenation of ethanol on copper catalysts, prepared by reduction of CuO a t different temperatures (see Figure 1).The effect was only small and even the author himself looked on the phenomenon as occurring only in this special case. But soon more experimental material, extending over a much wider range of A E and of A , led t o the opinion th at the C.E. is a more general phenomenon (Cremer, 4).At first the authors who reported a C.E. varied the catalyst, but not the substrate (2,4,5). Later also examples of a C.E. were found in cases where, with a given catalyst, different substrates had been used (6,7).

THE COMPENSATION E F F E C T I N H E T E R O G E N E O U S CATALYSIS

77

I n most cases of reactions showing a C.E. the relation between A and

AE can be represented by the empirical formula log A = AE/a

+ constant

(4)

where a is a constant (2,4). Figures 1, l a and 2 show examples of heterogeneous reactions typical for case 3. Figure 1 shows a log A versus AE plot. Different values of AE and

I

9.0 L..L

lo9 A

5

:/ 4 50

30

20 0

E

40

in kcal

FIG.1. Decomposition of ethanol on copper according t o Constable (2) and decomposition of nitrogen oxide on cupric oxide according t o Cremer and Marschall (3); log A plotted versus AE.

log A are found for the same reaction (e.g. decomposition of ethanol or of NzO) if a catalyst (Cu or CuO) has been prepared a t different temperatures. For each reaction a straight line is found, according to Equation (4). Figure 2 gives an Arrhenius plot for the decomposition of formic acid, measured on MgCOs/MgO catalysts of different thermal pretreatments. If the experiments obey Equation (4), there must be a temperature T , = a / R where all the lines intersect. Tables I and I1 list examples of the C.E., where for each case the maximum and minimum value of AE, the value of the intersection temperature T, (found mostly by extrapolation of the straight lines in the

TABLE I Examples of the Occurrence of a C . E . if the Composition of the Catalyst or of the Substrate is Changed* No.

Catnlyst

AE,i.

AE,.,

T." K

13

32

920

4 . 3 Cremer (1)

31

37

770

2

XI1 = Mg, Ca, Sr, Ba, Pb 15 S " ' ( 1 V ) = Y,La, Ce, (Cr) 15

28 38

685 640

4 7

1" = Mg, Ca

30

470

6.5

Metal = I't, Os, Xi 9.7 Compounds = CIHlo0, CaH12, C ~ H I I SC, I O H I ~

23.7

460

6 . 4 Bnlandin (6)

m : n from 0 to 0, 5

15.6

19.8

530

1.7

ICrkrll (!I)

X = l'd, Pt, .4U, TI, Cd, In, Sb, 1'11, 1%

13

37

800

8

Schwah (10)

5

21

-

5

9

13

-

4

Test reaction

Change

S I I I

S'CI

CH,CH?CI 4 CHICH?

-1

S"CI? S"'c13 X"Cl4

00

;r;llF?

+ HCI

JIrtals

Dehydrogenation of nronintic compounds

ttFt',Oa

2co

3

+

+

0 2

4

2c02

=

Al, In, Sc, 1,s and

S1 = Li, Sa, Ag, TI,

16

A log A

Author

Grirnni and Schwnniherger (5)

It1 A I : O 3

~

Alloy Ag/S Alloy 6 nl't

+ tttcu

H C O O H 4 CO?

+ H?

CzHd 4- H ? = C2Ha

from n = 100, ?ti = 0

to n = 0, p H ? + OII?

?ti

=

________-

100

Itieniicker (11)

TABLE I (Continued) No. 7

Catalyst Pt, Pt/Ni

Test reaction

Change Content of Ni

p H2 4 OH?

AEmin

11.4

AE,,,

15.2

T.' K A log A 670

2

Author Cremer and Rued1 (12)

HgnCls

C2Hz

+ HC1+

CHsCHCl

n 1 and 2

9.6

15.5

210

6 . 5 Patatand

Weidlich (13) MOO, 9

g

won

H2

+ D2-t

2HD

uon

10 Raney-Co n", Co 11 Pt/Als03

+ XO,

CoS

+ Hz + Co + HZS

N204Nz

+ 36

0 2

* Only examples where A log A > 1 are listed.

n from 1 t o 2.9 n from 1.5 to 2.9 n from 2 to 2.7

7.5 4.0 6.0

25.6 14.7 14.0

353 383 400

10 Molinari (14) 6.5

n from 0.7 to 100

6

34

823

7 . 4 Herglots and Lksner (14a)

26

92

800

X = Mo, Ru, Fr Mg, Ti, Cr, V

5

19

Mikovsky and Waters (31)

80

E. CREMER

TABLE I1 Examples of the Occurrence of a C.E. if the Temperature of Pretreatment of the Catalyst I s Changed ~~

Catalyst (reference) Ba C ~ (7,15) Z

Test reaction

Interval of pretreatment, "C. AE,,,

AE,,,

T.

A log A

C2HbCl = CzH4 350 to 600 HCl HCOOH = COZ 370 to 800 Hz

18

38

890

7

15

(52)

690

9

530 to 830

19

(64.5)

755

13

600 to 920

15.2

27.2

610 10

460to800

8.1

32.2

652

17

6OOto830

33.8

41.9

586

3

+

MgO (MgCO3) from Zillertal Magnesite (8) MgO (MgC03) from Zillertal Magnesite (16,17) MgO (MgCOa) from Radenthein Magnesite (16,17) MgO/CaO (MgC03/ CaC03) from Radenthein Dolomite (16,17) MgO from synthetic MgC03 (17,18) 2Nz0 = 2N2 MgO from Radenthein 0 2 Magnesite (17,18) CaO from CaC03 (Schering) (17,18) SrO from SrCO3 (Mallinckrodt) (17,W CUO (3,7) Fe/Ni alloy (17,19) pHz + OHz

+

+

680 to 1150 30.9

5 4 . 5 -1170

750to 1150 31.8

46 2

750to 1150 27.0

56.4 -1200

4

450 to 680 200to410

42 10.0

9 3

12 4.3

-

640

830 465

4.3 4.7

Arrhenius plot), and the maximum difference in A (A log A ) are given in special columns. Table I11 and Figure 3 show examples where the pretreatment of the catalyst a t different temperatures caused no C.E. I n all these cases AE was found t o be virtually constant, but in A there were differences up t o 103.

111. THE INTERPRETATION OF THE COMPENSATION EFFECT 1. Basic Relations

a. Relations between Enthalpy and Entropy Values. For a heterogeneous catalytic reaction, energy levels for two different catalysts 1

EFFECT IN HETEROGENEOUS CATALYSIS

THE COMPENSATION

81

30

25

t

*O

9 0 0

- I5

10

5 20

30

40

50 60 AE.(kcal/mole)

70

+

80

90

100

FIG. l a . C.E. for decomposition of NgO over Pt-ALOa catalysts (Mikovsky and Waters, 31).

FIG.2. Decomposition of formic acid on magnesite decomposed at different temperatures (370" t o 800'C.) according to Cremer and Kullich ( 8 ) ; Arrhenius plot.

82

E. CREMER

TABLE I11 Examples Where N o C.E. Occurred if the Temperature of Pretreatment of the Catalyst W a s Changed

Catalyst (reference)

Test reaction CzHbCI = CzH4

NiCh (15) MgO (Schering) (16,171 MgO (Schering) (17,181 MgO (by oxidation of pure Mg) (17,18) MgO from synthetic MgCOa (17,18) Ni (plate) (19) Pt (plate) (12)

0-

Interval of pretreatment, "C. AEmin AE,.,

+ HCI

HCOOH = COz + H z 2N20 = 2Nz

pHz

I

+ 02

- OH,

I

I

I

A log A

250 to 400 500 to 600

22.4

22.4

3.1

430t0800

46.7

46.7

0

654 to 1100 4 2 . 3

43.7

0.7

612to 1184 3 9 . 8

41.2

0.6

658to 1150 3 6 . 0 225 to 510 9.3 235to475 15.2

36.0 10.0 15.9

0 2.6 2.3

I

I

I

I

1

I

p-H,/Ni (First Order)

AE=IO.O 9.9 9.6 -I

-

-2

-

9.3 k col

510 405

305

225'C

111111111111

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

2.8

3.0

3.2

I / T . 10'

FIG.3. Para-ortho hydrogen conversion on nickel foil annealed at different temperatures (225" to 500°C.) according to Cremer and Kerber (19); Arrhenius plot.

THE COMPENSATION EFFECT IN HETEROGENEOUS CATALYSIS

83

and 2 are schematically shown in Figure 4. It is assumed th a t the activated complex has the same composition for both the homogeneous and the heterogeneous reaction. It follows from Figure 4 th a t AH: = AHhom- A:,

AP2 = AHhom - A;,

(5) where AHhon,is the activation energy of the homogeneous reaction, AH; and AH', are the activation energies on the catalysts 1 and 2, respectively, and X i and A', are the respective desorption energies of the activated complex. It follows from Equation ( 5 ) that

AH; - AH',

= A', - A;.

(6)

:jA;(iyf

I n the case shown in Figure 4 (AH: > AH:) the adsorption is stronger on catalyst 2 than on catalyst 1 (A: > A:), and therefore vibration quanta GAS PHASE

ON CATALYST

CATALYST 2

-!j----ON

4-L-

A A2

'

I

AH2

\'\ \\:'

PRODUCT

\\

L

FIG.4. Energy levels for two different catalysts.

of the bonds between the activated complex and the catalyst will be more easily excited on catalyst 1 than on catalyst 2 ; in other words, S: > Xi. As ASS = ss - 2s (ZS = the sum of the molar entropies of the reactants in the gas phase, being constant for the two reactions considered) the condition of case 3 exists: AH: > AH; corresponds t o AS: > AX:

Examples where for chemical equilibria such a correspondence of the enthalpy (AH') and entropy (AX") values may be shown are given in Table IV. b. Compensation Efect Due to a Tunnel Effect. I n 1932 Born, Franck, and Weisskopf (20) had suggested a "tunnel effect" for the explanation of catalytic reactions. The width of the energy barrier across which a n atom has t o pass, for example in the case of heterogeneous hydrogenation (21), is about 2 X cm. For this value, however, the theory of th e tunnel effect gives only a very low reaction rate.

84

E. CREMER

TABLE IV Example of a Compensation Eflect at Chemical Equilibria (AH" in kcal., AS" in E.U.)

Fluorine Chlorine Bromine Jodine

AHo 64.2 22.1 12.3 1.2

AS" -1.1 -2.4 -2.5 -2.6

AH"

366.3 348.8 340.7 330.2

AS" 60.25 50.92 47.15 41.67

If an electron rather than an atom tunnels, the transition probability for this particle is much larger in view of its smaller mass; therefore the tunnel effect may account for the observed rates of such chemical reactions (22) in which an electron transition can be considered an essential step. Detailed calculations based on the tunneling theory have been

20

-

I2

-

0

10

20

30

40

A € In k c o l

FIG.5. Relation between log A and A E values for electron exchange reactions in solutions; measurements from Huisenga and Magnusson (24), Gryder, Silvennann, and Dodson (25,26), Mcyer and Kahn (27).

presented by this author for the decomposition of ethylene chloride on metal halides (22), but in this case a C.E. may also be explained with the help of the considerations suggested in section 111, l a and 2. Thus a t present no clearly established example of a C.E. caused b y a tunnel effect may be given in the field of heterogeneous catalysis. B u t

THE COMPENSATION EFFECT IN HETEROGENEOUS CATALYSIS

85

there was pointed out recently by Marcus, Zwolinsky, and Eyring (23) a very interesting example of a homogeneous reaction in the liquid phase: the transition of electrons between ions of different valence states, for example, Men+ + Mem+--, Me(n+l)++ Me(m-1)+ (Me = metal). Figure 5 shows that there exists for this type of reaction approximately a linear relation between log A and AE in accordance with Equathe experimental results may be explained by tion (4). If A > classical assumptions, but if A < 1OI2, a tunnel effect seems to be the only possible explanation. If a tunnel effect takes place, a transition coefficient appears on the right side of Equation (2a), which is given-for heterogeneous as well as for homogeneous reactions-by the expression K~ =

exp { - (8rr/3h)[2m(V - W)lS')

(7)

where V is the height of the triangular energy barrier, W the excess energy of the tunneling electron, m its mass, and r the width of the barrier. The formula shows that the higher the energy W (= A E ) , the larger the probability of tunneling and the larger also the value of A . Therefore, if a process obeys Equation (7), a C.E. will occur (22,22a,23). 2. Simultaneous Occurrence of Reactions on Surface Centers Involving

Diferent Activation Energies In the first place it will be shown that a C.E. is likely to be found when a reaction takes place on two kinds (1 and 2) of active centers involving the activation energies AE, and A E z and the frequency factors A 1 and A z , respectively. In this case, the overall rate constant k becomes k

=

A1 exp ( - A E , / R T )

+ A z exp (-AE2/RT)

(8)

If AEl - AEz >> RT and if the two terms on the right-hand side of Equation (8) are assumed to be of the same order of magnitude, A1 has to be considerably larger than A z . An Arrhenius plot for k expressed by Equation (8) does not yield a straight line but a curve with asymptotes approaching the first term at low temperatures and the second term a t high temperatures. At intermediate temperatures where the two terms on the right-hand side of Equation (8) are of the same order of magnitude, the slope of the log k line varies necessarily with temperature. Thus a conventional calculation of AE and A can yield only values characteristic for certain limited ranges of temperature or rate constants.

86

E. CREMER

Figure 6 shows three plots calculated from Equation (8) for three different values of Az,the values of AE1, A E z , and A , being the same for all the three plots. It is assumed that between the measurements of line a and b, as well as between b and c, a treabment of the catalyst (e.g., a thermal treatment) has changed only the value of A z . This change might be caused by sintering (or forming) of surface centers of kind 2 or also by a change in the concentration of surface impurities. Trying to lay straight lines through the points calculated according to Equation (8) with the values listed in the subscript of Figure 6, one finds a n approximate intersection point at 1/T, = 1.1 lop3and thus gets a result which resembles the plot of experimental data in Figure 2 .

SUPERPOSITION OF 2 CENTERS

-I

-

-2

-

-3

-

-

1 (I)

s

0.9

1.0

1.2

1.4 IIT . l o 3

1.6

1.8

2.0

FIG.6. Arrhenius plot for a reaction on two different kinds of centers according to Equation (8). log A , = 10.3; AE1 = 38 kcal.; AEt = 18 kcal.; curve a : log A2 = < 2 ; curve c: log A2 = 5 . 6 ; curve h : log A l = 3.9.

Equation (8) can be generalized by considering more than two kinds of centers. If there are many different kinds of active centers, their relative proportions may be represented approximately by a continuous distribution function, as has been suggested by Constable ( 2 ) . I n particular, it may be assumed that their relative numbers decrease exponentially with decreasing activation energy AE. Thus the number of active centers d n involving activation energies between AE and A E dAE can be assumed to be dn = C exp ( A E / a ) d A E (9)

+

where C and a are constants. Moreover it may be assumed that there are no centers present involving activation energies lower than AEminand larger than AEmsx.

THE COMPENSATION EFFECT IN HETEROGENEOUS CATALYSIS

87

Hence the over-all rate constant k is obtained as

/E:rx

d n A exp ( - A E / R T ) = n A ( a / R T - l ) - I exp [-(AEmax- AErnin)/~] exp (-AE,nin/RT)

k =

(10) (if AE,,, - AErni,,>> a > R T , and n = total number of centers). The variability of the denominator with temperature is negligible compared with the variability caused by 1 / T in the exponent. Then upon comparison of Equations (2) and (10) it follows that

AE S AE,in and

+

In A Zg AEmin constant Thus the activation energy calculated in the conventional way is nearly equal t o the activation energy AE,,, of the most active centers. Different kinds of pretreatment of the catalyst may form or destroy different types of active centers and thus change the value of AEmin. Furthermore the logarithm of the frequency factor A is related to AE, as has been proposed in view of experimental data, according t o Eq. (4). It might be questioned whether a change in the minimum activation energy might occur without a change in the distribution parameter a. Next is considered the possibility of deriving th e exponential distribution function of active centers as postulated by Equation (9). One may suggest a Boltamann distribution among the surface centers which is frozen in a t a temperature 0 (28,28a). Denoting the energy of formation of 1 mole of active centers by E*, one obtains according to Boltzmann's law the number of active centers having energies between E* and E* dE* as dn = C' exp (-E*/RO)dE* (13)

+

Cremer and Fliigge have suggested for the derivation of the OstwaldFreundlich adsorption isotherm (29) that the energy of desorption from a catalyst possessing many kinds of such surface centers is a linear function of the energy E*. Thus 1' = BE* y (14)

+

where p and y are constants. If the relationship given in Equation (5) is now considered, one may write

AE

=

AH

=

AHhom- As

(15)

The combination of Equations (15), (14),and (13) yields dn = i exp (AE/pRO)dAE

(16)

88

E. CREMER

with i being a constant. Equation (16) is equivalent to Equation (9) with the special value a = pRe (17) In some cases the estimated temperature of preparation of solid catalysts seems t o be close to 0 and therefore in accordance with the foregoing working hypothesis; moreover, adsorption measurements on some catalysts show a dependence on the temperature of catalyst pretreatment in accordance with a Boltzmann distribution of their surface centers. However for catalysts such as chlorides (15) and oxides of the type MezOa(7), which were first considered to be suitable objects for rationalizing the compensation effect, Equations (10) and (17), which interrelate the overall rate constant k with the temperature e of the pretreatment of the catalyst, did not fit the experimental data. The occurrence of a Boltzmann distribution of centers at a catalyst can, therefore, be regarded only as one of the possible explanations of the compensation effect. So far the underlying assumptions have not been rigorously confirmed for any particular catalyst and catalytic reaction. The interpretation of the C.E. by a superimposition of reactions occurring at different active surface centers is compatible with the fact that many multicomponent catalysts exhibit a C.E. but no C.E. is found when very pure substances have been subjected to different thermal pretreatments (17). This implies the possibility that many active centers are due to “impurities” and that their numbers may change with the pretreatment of the catalyst, e.g., by means of aggregation, volatilization, etc. As an illustration, data for the decomposition of NzO on MgO, prepared from synthetic and from natural magnesites, and data for the para-ortho hydrogen conversion on pure metals and on alloys are presented in Tables I1 and 111.

IV. APPARENT COMPENSATION EFFECTS If for a given group of catalysts the Arrhenius equation strictly holds and the values of A E and A are equal from one catalyst to the next, a C.E. may also result from experimental errors. From the Arrhenius equation it follows that

where e is the magnitude of error in the evaluation of A E . Hence the calculated value of log A will be too high if the value of the activation energy derived from experimental observations is larger than its actual value, and vice versa. Likewise, errors in the determination of the catalyst temperature and

THE COMPENSATION

EFFECT IN HETEROGENEOUS

CATALYSIS

89

especially errors in the order of a reaction can lead to apparent compensation effects. The case illustrated by Figure 7 deserves a special discussion. If it is assumed that the state of the surface of a catalyst undergoes a reversible change between the temperatures T Iand T2, then the apparent activation energy of the catalytic reaction calculated in the conventional way by means of the Arrhenius equation will depend (a) on the activation energy of the reaction at a surface that remains unaffected by temperature changes and (b) on the temperature dependence of the surface state. For instance, changes of the effective surface area with temperature may be caused by diffusion limitations, the opening and closing the pores, changes

.m

0

3 0

I-

1.0

1.1

1.14

1.2

I .3

I/T.I O - ~

FIG.7. Arrhenius plot for a reaction a t two different catalysts, 1 and 2, undergoing a reversible change in their effective surface areas with temperature. Dotted lines

refer to the true A H : values. The solid lines 1and 2 interconnect corresponding points of the dotted lines at k = 10 and k = 100.

of grain boundaries, or shifts of the extent to which different solid phases are represented in the catalyst surfaces. Intersections of the straight lines of an Arrhenius plot are in general obtained by extrapolation, This implies that AE is virtually constant between the experimental temperatures and the extrapolated temperature T,. The possibility that AE may vary with the temperature and also with specific surface conditions, such as the density of its coverage with chemisorbed reactants, has been suggested in order t o explain the existence of compensation effects (17). That AE may show very significant variation with temperature if the penetration of the reactants into the pores of the catalyst must be taken into account has been pointed out especially by Wicke and Brotz (30). I n this case two extreme values of AE may be obtained: AE,., if the activation energy of the chemical reaction determines the temperature

90

E. CREMER

dependence (k = ko) and AEmin= >dE,,, if both reaction and diffusion are rate determining ( k = Thus in cases where the relation AE,,,ax:AE,,,,,, = 2 (or <2) is found, diffusion must be considered as a possible explanation. V. CONCLUDING REMARKS

4).

To summarize, a compensation effect may be caused by several reasons. 1. I n some cases the statistical mechanical concept of entropy changes makes it conceivable that relations exist between AH’ and A S ” or between AHS and A S f for a number of reactions. As a special case, reactions involving the tunneling of electrons between ions of different valence states may serve as an example for the possibility that a tunnel effect may account for the occurrence of compensation effects. 2. I n many cases a compensation effect is indicated on comparison of the rates of the same reaction a t the same type of catalyst after the latter has undergone special pretreatments or has been changed in its composition. I n such cases the occurrence of a compensation effect may be expected if different kinds of active surface centers act simultaneously as sites of the catalytic reaction and if the proportions of these different active centers (involving different activation energies) are shifted by means of special pretreatments of the catalyst or by changes in its composition. 3. An apparent compensation effect can result from errors in the experimental data used for an Arrhenium plot. Besides trivial errors, there may also occur errors in the calculation of rate constants, for instance when a homogeneous and a heterogeneous reaction occur simultaneously or when a heterogeneous reaction undergoes a change from a certain reaction order t o another order. A temperature dependence of the activation energy, and the variability of the effective surface of the catalyst with temperature, especially caused by diffusion processes, may also account for apparent compensation effects. ACKNOWLEDGMENT

This work was done during a stay at the Massachusetts Institute of Technology during 1953-54. The author is indebted t o Professors John Chipman and John T. Norton for making available the facilities of the Department of Metallurgy and to Professor Carl Wagner for many valuable and stimulating discussions. Support by the American Association of University Women and the Fulbright Foundation is gratefully acknowledged.

REFERENCES 1. See Glasstone, S., Laidler, K. J., and Eyring, H., “The Theory of Rate Processes.” McGraw-Hill, New York, 1941. 2. Constable, F. H., Proc. Roy. Soc. (London) A108, 355 (1923), and later papers.

T H E COMPENSATION E F F E C T IN HETEROGENEOUS CATALYSIS

91

3. Cremer, E., and Marschall, E., Monatsh. Chem. 82, 840 (1951). 4. Cremer, E., Z.physik. Chem. A144, 231 (1929). 5. Grimm, H. G., and Schwamberger, E., R h n i o n intern. chim. phys., Paris 3, 214 (1928). 6. Balandin, A. A., 2.physik. Chem. B19, 451 (1932). 7. Cremer, E., Z. Elektrochem. 63, 269 (1949); J. chim. phys. 47, 439 (1950). 8. Cremer, E., and Kullich, E., Ruder Rundschau 4, 176 (1950); see also Kullich, E., Thesis, Innsbruck, 1950. 9. Eckell, H., 2. Elektrochem. 39, 855 (1937). 10. Schwab, G.-M., Trans. Faraday SOC.42, 689 (1946). 11. Rienacker, G., and Sarry, B., Z . anorg. Chem. 267, 41 (1948). 12. Ruedl, E., Thesis, Innsbruck, 1953. 13. Patat, F., and Weidlich, P., Helv. Chim. Acta 32, 783 (1949). 14. Molinari, E., private communication. 14a. Herglotz, H., and Lissner, A., 2. anorg. Chem. 260, 161 (1949). 15. Cremer, E., and Baldt, R., Monatsh. Chem. 79, 439 (1949); Z. Naturforsch. 4a, 338 (1949). 16. Conrad, F., Thesis, Innsbruck, 1951. 17. Cremer, E., 2. Elektrochem. 66,439 (1952). 18. Stadlmann, W., Thesis, Innsbruck, 1952. 19. Kerber, R , Thesis, Innsbruck, 1952. 20. Born, M., and Franck, J., Nachr. Ges. Wiss. Gottingen 11, 77 (1930); Born, M., and Weisskopf, V., 2. physik. Chem. B12, 206 (1931). 21. Cremer, E., and Polanyi, M., 2. physik. Chem. Bl9, 443 (1932). 22. Cremer, E., Ezperientia 4, 349 (1948); J. chim. phys. 47, 439 (1950). 2%. Patat, F., Z. Elektrochem. 63, 216 (1949). 23. Marcus, R. J., Zwolinsky, B. J., and Eyring, H., Technical Report X, Inst. for Study of Rate Processes, Univ. Utah, J . Phys. Chem. 68, 432 (1954). 24. Huizenga, J. R., and Magnusson, L. B., J . Am. Chem. SOC.73, 3202 (1951). 25. Gryder, J. W., and Dodson, R. W., J . Am. Chem. SOC.73, 2890 (1951). 26. Silvermann, J., and Dodson, R. W., J . Phys. Chem. 66, 846 (1952). 27. Meyer, E. G., and Kahn, M., J . A m . Chem. SOC.73, 4950 (1951). 28. Cremer, E., and Schwab, G.-M., Z . physik. Chem. A144, 243 (1929). 28a. Assuming this the term “I’hcta Rule” is sometimes used instead of C. E. See Schwab, G.-M., Advances in Catalysis 2, 251 (1950). 29. Cremer, E., and Fliigge, S., Z . physik. Chem. B41, 453 (1938). 30. Wicke, E., and Brotz, W., Chem.-Zng.-Tech. 21, 219 (1949). 31. Mikovsky, R. J., and Waters, R. F., private communication.

This Page Intentionally Left Blank

Field Emission Microscopy and Some Applications to Catalysis and Chemisorption ROBERT GOMER Institute for the Study of Metals The University of Chicago Chicago, Illinois Page I. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 1. Cold Emission of Electrons from Metals.. . . . . . . . . . . . . . 2. Principle of the Field Emission Microscope. . . . . . 3. Resolution in the Field Emission Microscope, . . . 4. The Field Ion Microscope.. . . . . . . . . . . . . . . . . . . . 5. Mechanism of Field Ionization.. . . . . . . . . . . . . . . . 111. Selected Applications of Field and Ion Microscopy. . . . . . 1. Pattern Types.. . . . . . . . . . . . .................... 111 2. Molecular Images . . . . . . . . . . 3. Surface Phases.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Chemisorption and Mobility of Oxygen on Tungsten.. . . . . . . . . . . . . . . . . 115 a. Use of Liquid Helium for Vacuum Technique. . . . . . . . . . . . . . . . . . . . . . 115 5. Mass Spectrometric Analysis of Yields from the Ion Microscope. . . . . . . . 125 .. Appendix: Experimental Methods and Techniques. . . . . References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

I. INTRODUCTION The field emission microscope was invented by E. W. Muller in 1937 and developed by him in the years following (1). For various reasons the device failed to attract general attention until the end of World War 11. Its potentialities are now being recognized; the field and ion microscopes, in conjunction with recently developed ultra high vacuum techniques, are rapidly becoming important tools for the study of physical and chemical surface phenomena. This chapter is devoted to a short theoretical account of field and ion microscopy, followed by selected applications t o problems of catalysis drawn largely from work now in progress at the University of Chicago. No attempt a t completeness is made, and much recent work not directly related t o catalytic problems has been omitted. A very brief summary of experimental technique is appended. 93

94

R OB E R T GOMER

11. THEORY 1 . Cold Emission of Electrons from Metals

The quantum mechanical theory of tunneling was first applied to the cold emission of electrons by Fowler and Nordheim (2). A simplified plausibility argument, rather than a rigorous derivation of their result, will be given here. Figure l a represents a potential energy diagram for electrons in a metal and the adjoining vacuum in the presence and absence of external electric fields. It will be noted th a t the electrons are

I

,4498-13

FIG.la. Schematic potential energy diagrams for electrons in a metal with and without applied field. Clean metal, no image potential assumed. x = work function, p = depth of Fermi sea.

not all in the lowest energy state, but are accommodated pairwise in successive levels until all electrons have been stacked in this manner. This results from the Pauli exclusion principle, which demands Fermi statistics for elementary particles. The energy of the highest filled level, measured from the potential minimum in the metal, is called the Fermi energy M , and is equal t o the chemical potential of electrons in the metal. The energy difference between the Fermi level and the potential energy of electrons in the vacuum is known as the thermionic work function, x. Before discussing field emission, it might be well t o recall a few important characteristics of particles obeying Fermi statistics. For details and proofs of the facts mentioned here the reader is referred to standard texts (3). We note first that the energy distribution of Fermions at 0°K differs

FIELD EMISSION

95

inappreciably from that at room temperature, a direct result of the fact that they may have high energies a t all temperatures. We also note that the number of quantum states near the top of the Fermi sea is much larger than near the bottom, so th at most electrons are accommodated in energy levels near p. We shall also make use of the fact that the maximum energy a Fermion can have a t 0°K (and except for a negligible “Boltzmann tail” a t room temperature) is the Fermi energy p, so that the distribution of momenta among the Cartesian degrees of freedom is not independent, a s in the case of Boltzmann particles. I n the absence of an external field, electrons in the metal are confronted by a semi-infinite potential barrier (upper solid line in Fig. l a ) , so that escape is possible only over the barrier. The process of thermionic emission consists of boiling electrons out of the Fermi sea with kinetic energy 2 x p . The presence of a field F volts/cm. a t and near the surface modifies the barrier as shown. It follows from elementary electrostatics that the potential I‘ will not be noticed by electrons sufficiently far in the interior of the metal. However, electrons approaching the surface are now confronted by a finite potential barrier, so th a t tunneling can occur for sufficiently low and thin barriers. The probability, P, of barrier penetration is approximately given b y

+

P

=

const . exp [-2’4m35/h

jzd m ax]

(1)

where m is the particle’s mass, h Planck’s constant divided b y 21, E the kinetic and V the potential energy, and 1 the width of the barrier. Reference t o Fig. l a shows that A , the area under the square root of the barrier, is nearly triangular in shape and hence approximately equal t o

A

=

+hx$4.x / F

=

>6x9*/F

(2)

if we consider only electrons a t the top of the Fermi sea. The transmission probability then becomes

P

=

const . exp [ ( P m m ” / h ) x ’ 5 / ~ ~

(3)

Multiplication of P by the number of electrons arriving a t unit surface in unit time should then give the field emission current density J . The argument just presented has limited itself to electrons a t the top of the Fermi sea. It is apparent from Equation (1) and the nature of the electron energy distribution th at this assumption is good. We see that the negative logarithm of field emission current varies as the 96 power of the thermionic work function x and inversely as the applied field. A more rigorous derivation leads to the same type of dependence on field and work function and is given by the Fowler-

96

ROBERT GOMER

Nordheim equation (2)

J

=

6.2 X 106(p/x)"(p

+x)-P

exp [-6.84 X 107x'("/F]

(4)

for p and x in e.v. Fig. l b shows a slightly more realistic potential model, taking into account the image forces experienced by a n electron leaving the metal. It is seen that the image potential decreases the effective barrier area. The effect may be thought of as a n increase in field or a decrease in work function. Nordheim (4) has corrected the exponential

FIG.lb. Schematic potential energy diagrams, based on image potential, for electrons in a metal with and without applied field. Barriers are shown for clean metal and metal with a dipole layer of nitrogen; x = work function; p = depth of Fermi sea; PI represents the contribution of the dipole layer to the potential for a n assumed nitrogen-nitrogen spacing, a, of 5A., and is drawn from an origin a t 4.5 volts on the diagram.

part of Equation (4) for a classical image potential. The resultant correction factor a by which the exponent of Equation (4) must be multiplied is given a s a function of y = 3.78

x

10-4~~~/~

(5)

in Table I. I n the range F = 107-108v./cm. a has values of 0.9-0.4. It should be noted that a changes very little over the small range of applied fields usually encountered in a given field emission experiment. Equation (41, corrected for image potential, has recently been verified quantitatively by Dyke and co-workers (5). Nordheim Correction y 1y

0 1

0.2

0.951

TABLE I 3.78 X lO-4FFf$/,, for field F in u./cm. and

(a)as a Funcdion of y =

0.3 0.904

Work Function x in e.v.

0.4 0.849

0.5 0.781

0.6 0.696

0.7 0.603

0.8 0.494

0.9 0.345

1.0 0

FIELD EMISSION

97

2. Principle of the Field Emission Microscope

The field emission microscope utilizes the phenomenon of cold emission as follows: A wire etched to a very sharp point is surrounded by a spherical anode, usually in the form of a fluorescent screen. The system is evacuated to pressures of the order of to 10-11 mm. of Hg and the wire is heatpolished electrically. This outgasses the metal and produces a smoothly A

Fro. 2. Schematic drawing of one form of the field emission microscope. E , glass envelope; S phosphorescent screen; M , metal backing; A , anode lead-in; T,emitter tip; C, tip support structure; V , vacuum lead.

rounded tip, whose radius varies from t o 10-4 cm., depending on the melting point and ease of outgassing of the metal used. Figure 2 shows a schematic diagram of a microscope assembly. If a potential of the order of lo4 volts is applied between tip and screen, cold emission occurs, since the field a t the tip is F = kV/r (6) with k

-

>$* so that F

=

10-7-108 v./cm. Electrons leave the tip with

* The value of k depends on the exact geometry of the tip. If the latter were a free sphere, k would be unity. For a hyperboloid of revolution of radius of curvature r Miiller (1) uses k = 2/ln (4 x / r ) (64 with x the tip-anode distance. For a paraboloid of focal length ro Dyke and his coworkers have used (6) 1 ko = ro In x/ra ~

with k(O)/ko = (7.7-6.7 cos O)-s, where e is the angle between the axis of rotation (apex direction) and the radius vector t o a point on the paraboloid. The radius of curvature r = 0 . 4 ~ 0More . precise determinations of k and its variation with angle, based on electron microscopical examination of the tip, are given by Dyke and w-workers (6).

98

ROBERT GOMER

very low initial kinetic energy and will therefore follow paths parallel t o the lines of force, a t least initially. Since these enter the metal tip perpendicularly, electron paths like those shown in Fig. 3 result. Thus an electron emission map of the tip, magnified by the amount D/6 = X / T , appears on the screen. Linear magnifications of the order of lo6to 106 are possible. Typical emission patterns from nickel tips are shown in Plate I A and B. It will be noted that the patterns display the symmetry characteristic of cubic crystals; that of IA is oriented with the (111)-axis, and th a t of

i

Screen

FIG.3. Schematic diagram showing the optics of the field emission microscope. T , radius of curvature of the tip; 2,tip-to-sereen distance. A region of linear dimension 6 will appear as D on the screen. PLATEI A . Emission pattcrns from clean Ni tips. Principal directions markcd. (111) oriented. B. Emission patterns from clean Ni tips. Principal directions marked. (100) oriented. C . Emission pattern from clean tungsten. Central dark region is (110) face, surrounded by four (211) faces. Two smaller dark spots are (100) faces. Two molecules of zinc phthalocyanine, one flat, the other on edge, are visible in and a t the edge of the (110) face. D. Emission from nickel heated to about 700°K with the field applied. E . (111)-oriented nickel tip, heated to 1300°K after oxygen contamination. F. Si-contaminated (0.3%) Ni tip after heating to 1538°K. G . Tip of Plate IF after 1466°K. H. Tip of IF after 1438°K. I. Tip of IF after 1403°K. J . Tip of IF after 1369°K. Rings around (210) faces developed. Bridge on (110) faces incipient. K . Tip of IFafter 1324°K. Bridge over (110) faces closing. Rings on (111) incipient, L . Tip of IF after 1296'K. Steps in (111) apparent. Rings around (110) enlarged.

FIELD EMISSION

PLATEI

(Legends an page 98)

99

100

ILOBEHT GOMER

IB with the (100)-axis, parallel to the wire direction. It is generally found t h a t the tips obtained by etching are much smaller than the average grain size of the crystallites composing the original wire. Thus the tips are usually part of a single crystal, almost invariably of preferred orientation with respect t o the wire axis. If the work functions of every portion of the crystal constituting t h e tip were identical, one mould simply observe a uniformly bright circular patch on the fluorescent screen. Different crystal faces have slightly different work functions, however, so t h a t one sees a pattern showing different intensities for different faces. I n general, closely packed faces have higher work functions than loosely packed ones.* In fee cubic crystals like Ni the ( l l l ) , (loo), and (110) faces are t h e most closely packed and appear darker than the rest of the pattern (Plate I A and R). I n bcc cubic. metals like tungsten, the (110), (loo), and (211) faces are most closely packed arid appear darkest (Plate IC). Crystallographic indices can he assigned unequivocally from the symmetry of the patterns arid the angular separations of the various faces. It may be pointed out t h a t the patterns will always appear slightly I( compressed” from the ideal shown in Fig. 3, because the tip is not a free sphere but a hemisphere supported by a cylindrical or conical rod. Since lines of force repel each other, the effect of this structure is a coinpressiori by a factor p 1.5 and a derrease in the field (at a free sphere, F = V / T ) .I t is always possible to identify a t least two directions from symmetry alone, so that ,f3 may be computed and the rest of the pattern indexed. It has been fourid possible to “decompress” the pattern and obtain an approximately 1.5-fold increase in magnification by placing a n anode a t screen potential just behind the tip, in the form of a ring through which the latter projects (8). It is doubtful, however, whether such devices or magnetic lenses van produce better resolution (see below). Differeiices in work furictlion orcur not only on different faces of a clean metal crystal, but result also from the presence of adsorbed films, or monolayers of gases. Since the work function appears to the %-power in the exponent of the Fowler-Nordheim equation (Equation 4), changes in x of a few hundredths of a n electron volt produce marked changes in emission. The changes in work fuiictiori produced by adsorbates are generally

=

* Values of the work functions of different farcs are best known for tungsten. A review of thcrmionic measurements and their theoretical interpretation is given by Herring and Nichols (7s). The reader interested in the theory of the phenomenon is also referred to Hrnoluchowski (7b). Work functions of two crystallographic directions in tungsten have recently been deterrnined by field emission technique by Drerhsler and Mullcr (7r).

FIELD EMISSION

101

thought t o arise from the creation of dipole layers. Thus electronegative atoms like oxygen on tungsten tend to attract electrons, resulting in an excess of about electron charges. This induces a n electrical image in the metal, resulting in a dipole. If the dipole moment of individual ad-atom complexes varies only slightly with coverage, the layer may be considered as a condenser or a dipole sheet whose potential is proportional t o the number of individual dipoles per unit area, i.e., proportional to the coverage of adsorbate. This potential, or work function increment, is usually called contact potential, and is given by

pl, = 2 n ~ o e

(7)

where P o is the dipole moment per unit area a t maximum coverage and e is the percentage of occupied sites. In thermionic emission and contact potential work, it is immaterial whether one considers the electric double layer resulting from adsorption as consisting of discrete or uniformly spread out charges. In field emission, however, the discreteness of the dipoles of the ad-layer results in a slight decrease of the effective work function increment A x compared t o th a t predicted by Equation (7). The reason for this is that the potential of the discrete layer does not build up t o P I , a t once, but only a t varying distances from the surface. Therefore the contribution of the layer t o the barrier that electrons must penetrate in field emission will be decreased. The decrease will be most marked a t low coverages (slowest build-up of the layer potential) or high fields (narrowest barrier, hence irregularities near the surface are most important). The effect is illustrated in Fig. lb . Calculated results (9) are summarized in Fig. 4. Changes in emission may also be caused by local enhancement of the electric field a t small protuberances on portions of the tip. I n addition, prot,uberances tend to act as if they were independent tips; that is, there will be additional divergence of electrons emitted in their vicinity, with the result that local magnification may be increased by a factor of 10-30 over that corresponding t o the radius of curvature of the tip as a whole. It must be remembered that ordinary concepts of conduction and insulation break down a t the high fields involved, so that even organic dye molecules transmit electrons easily. These facts make it possible t o observe small, bright images superimposed on the substrate pattern, corresponding t o individual molecules of quite small size (Plate IC). Incipient overgrowths and surface phases may also be observed in this way. 3. Resolution in the Field Emission Microscope Before discussing the ion microscope or describing experimental applications, we must consider the resolution obtainable with the field

102

ROBERT GOMER

emission microscope. It might be thought at first that the limiting factor is the de Broglie wavelength of the image-forming electrons. However, a little thought shows that the wavelength t o be considered is that of electrons arriving a t the screen, which corresponds to the final voltage and is therefore very small. I n most cases (with the possible exception of molecular images) the limiting factor is the statistical distribution of electron velocities transverse to the direction of emission (1,lO). For a

FIG.4. Effective work function increment AX in terms of the contact potential Pi, as a function of coverage 8, for various applied fields P.

transverse kinetic energy E , the resolution d can be shown t o be

since the time of flight t is given with sufficient accuracy by

"(&)

$5

t =

(9)

It has been mentioned that the distribution of kinetic energies among the Cartesian degrees of freedom is not independent for Fermi particles. If we choose the x-axis along the emission direction, it follows that an electron having E , energy in this direction can have a t the most 1.1 - E , transverse energy. Calculation shows that the most probable energy,

FIELD EMISSION

103

independent of its distribution among the y and z coordinates, E,,, for a given E , is E,, = % ( P - E.J (10) Thus we see at once that a barrier permitting only high energy electrons to escape will cut down the statistical distribution of transverse energies; the lower we make the barrier, on the other hand, the deeper we reach into the Fermi sea, with a concomitant spread in transverse momenta. Since the barrier is determined by the applied field, it is not surprising t o find that a linear relation exists between applied field and the average transverse energy E,,

E,, = 4.33 X 10-9x-44(F/~)e.v.

(11)

Combination of Equations (8) and (11) shows that the resolution is given by

(,“:n)” cm.

d = 1.31 X 10-4p __

It is seen that d is independent of applied voltage. This results from the fact t ha t the increase in E,, caused by an increase in F (proportional to V ) is exactly balanced by the decrease in the time of flight of the electrons. The general validity of this reasoning is confirmed by the experimentally found dependence of the forward energy distribution E, on field (11).

4. The Field Ion Microscope E. W. Muller recently developed an ingenious modification of the field emission microscope, based on the field ionization of hydrogen (1 ,12). The device consists of an ordinary field emission tube containing hydrogen a t a pressure of about mm. When the tip is made the anode, adsorbed hydrogen in the form of ions can be pulled off at fields of the order of 2 x 108 v./cm. A pattern, correspoonding t o the electron emission, but with greatly increased resolution (-3A.) results. The obtainable current is limited by the maximum gas pressure th a t can be maintained without resulting in a discharge. Thus the rate-determining step is not the desorption of ions under the influence of the high field, but the diffusion of gas to the tip. The latter is a n order of magnitude greater than would be expected from kinetic theory because gas near the tip is polarized and thus attracted. At room temperature ionization seems to occur from definite sites, so that the image produced on the screen is a hydrogen desorption map of the substrate. The sharpness with which this pattern is produced is limited by the lateral momentum of adsorbed hydrogen, that is, by the

104

ROBERT GOMER

zero-point energy of the hydrogen-bending vibrations (10).Equation (8) for the resolution can be used with E o replacing E. It is seen that a n increase in applied voltage should result in better resolution in the case of ion patterns, since such an increase does not affect Eo. I n practice the applied voltage is limited by the fact that autoionization occurs in the gas phase near the tip a t fields of the order of 3 X los v./cm., which results in unsharpness of the pattern (1,13). The mass of the adsorbate is not contained in (8) explicitly for the ion case. Nevertheless such an effect exists bcrause E , cc 1/m3‘ (12)

so that for identical force constants better resolution should be achieved with the heavier isotope, e.g., deuterium in place of hydrogen. An estimate of the zero-point energy for hydrogen leads t o values of the order of E o = 0.03 e.v., which is of the same magnitude as 1cT a t room temperature. Thus the resolution found by Muller is seen to be possible. It should be pointed out that only weak physical adsorption, caused by the interaction between the field and permanent or induced dipole moments, or no adsorption a t all, is required for ion image formation, if there are local differences over the surface in ease of ionization. Small local differences in field and image or exchange potential may be sufficient for a differentiated pattern with gases like He, Ne, etc. If no adsorption occurs, the resolution is determined by the average velocity of the gas molecules a t the time of ionization. Although the latter is of the order of (lcT/m)$‘,it will in general be somewhat larger and depend on the applied field and the polarizability of the atom or molecule. This is so because molecules traveling near the tip with thermal velocities will be accelerated toward the latter by the field-dipole interaction, which may impart tangential (to the final “emission” direction) velocity components t o molecules before they are ionized. 5 . Mechanism of Field Ionization

Muller first noted field desorption for layers of barium on tungsten (1). He concluded correctly that tunneling could hardly be responsible for the ionization of heavy particles and assumed that the potential curve for the ad-atom substrate complex was deformed to the point where activated desorption over the barrier could take place. This view was supported by the fact that he found that the field necessary for the desorption of thorium was quite temperature dependent, changing from the (remarkably low) value of 6.7 X lo7 v./cm. a t room temperature t o 3.5 x 107 v./cm. a t 1500°K (1). Similar results were found b y him for barium

FIELD EMISSION

105

(1). The dependence of field on coverage showed th a t desorption occurred most readily a t lowest coverages, suggesting a screening effect. I n the course of a recent examination of ion yields from the field ion microscope with a mass spectrometer (13a) it was found that there is only a very slight mass effect when deuterium is substituted for hydrogen. It was also found that both H+ and H2+are formed. The ratio is a function of pressure, applied field, and the state of the tip. At low fields (-lo8) H2+ predominates, whereas H+ exceeds H2+a t high fields. Increase in pressure

'I

20

""

241

28 E .%

FIG.5a. Potential energy diagram for the Is electron of a hydrogen atom a t a distance of 5.5& from a tungsten surfacc in the absence of external fields. I = ionization potential; x = work function; p = depth of Fermi sea; P H = proton-electron potential; PW = image potential.

has the effect of shifting the changeover to higher fields. The results obtained with other gases will be discussed in the last section of this chapter. It suffices here to point out that the absence of a mass effect for hydrogen ions precludes a tunneling in this case also. (It will be recalled from Equation (1) that the square root of the mass appears exponentially in the barrier penetration coefficient.) I n spite of these facts the phenomenon of field ionization is probably not a classical one in the case of gases like hydrogen, oxygen, ethane, and so on. When dissociated, these do not exist on the surface a s anything resembling positive ions. Quite aside from chemical intuition, this is borne out by the fact that their contact potentials, unlike those of alkali and alkaline earth metals, are of a sign suggesting the formation of negative

106

ROBERT GOMER

ions, or a t least of negatively polarized bonds. The existence of H2+, 02+, or C2He+ions on a tungsten surface also seems most unlikely. It is more probable that the phenomenon consists of a tunneling of electrons into the metal.* We shall first consider the case of an unbound neutral atom or molecule near a surface, choosing H for simplicity. Since electronic motion is much faster than that of nuclei, we may consider the

0 4

8

12

I6 20

I I I I I

28

I

E .V. FIG.5b. Potential energy diagram for the

1s electron of a hydrogen atom a t a distance of 5.5& from a tungsten surface with a n applied field of 2 X los v./cm. x = work function; p = depth of Fermi sea; PW = image potential.

position of the latter fixed a t a reasonable distance, say 5 i . from the surface. I n the ground state of the hydrogen atom the expectation value of the proton electron distance is the first Bohr radius, r = 0.5A., the electron’s potential energy is -28 volts, and its kinetic energy is 14 volts, so that the ionization potential is 14 volts. Figure 5a shows this potential and the conduction band of tungsten. Figure 5b shows the same system under the influence of an applied field of 2 X lo8v./cm. The energy level

* This idea has been proposed

independently b y F. Kirchner (13b).

FIELD EMISSION

107

of the hydrogen electron has now been raised t o th a t of the metal, so th a t tunneling can occur. It can readily be calculated that the half-life of the atom in Fig. 5b is of the order of sec. For comparison, the same field applied t o a free H atom is drawn in Fig. 5c. The half-life for ionization is now lO-'4 sec. Since the time spent by a molecule in the region of high field is of the order of 10-'1 sec., it is apparent th a t all H atoms approaching the tip will become ionized a t this field.

E.V..

FIG.5c. Potential energy diagram for the 1s electron of a free hydrogen atom in a field of 2 x 1 0 8 v./cm. I = ionization potential. The calculation is th at of barrier penetration, using Equation (1) and the barriers drawn in Figs. 5b and 5c. The frequency with which the hydrogen electron arrives a t the barrier is estimated from the expectation value of the momentum and the first Bohr radius as lo1* sec.-I. It is interesting t o note th at tunneling occurs more readily near the metal surface than in free space for the same applied field, because the effect of image and exchange forces decreases the potential barrier confronted by the tunneling electron. Unlike field emission from metals, the effect becomes more important with decreasing field as inspection of Figs. 5b and

108

ROBERT GOMER

5c will show. Table I1 lists the calculated mean lives of H atoms near a

tungsten surface and in free space as a function of applied field. The ion current i t o be expected at a given field is given by

i

=

n(1 -

e--t'7)

(13)

where n is the number of molecules per second arriving in the region of high field, t the titne spent in transit in this region (if ionization were not occurring), and T the mean life with respect to ionization at the given LOW

FIELD

I

INTERMEDIATE FIELD

I

HlGH F I E L D

ION

ENERGY

I I

_cC

KEY,,

FIG.6. Typical peak shapes for physically adsorbed gases as functions of field.

field. It will be seen therefore that the postulated mechanism predicts ionization only near the tip at low fields and ionization throughout the region of high fields when the latter is very high. Although the calculations have been made for H atoms, they should apply qualitatively t o other molecules and atoms. I n the case of complex molecules polarization will lower the effective field. Also, except for a electrons, the calculation of electronic frequency becomes more complex. Nevertheless, values of T can be estimated roughly by using Equation (3) with the ionization potential in place of x and assuming an electronic frequency of 10l6see.-'. The variation in origin of the ions as a function of field is borne out

109

FIELD EMISSION

strikingly by mass spectrometric experiments (13a). Figure 6 shows typical peak shapes for ions produced from weakly adsorbed molecules. It is seen that at low fields peaks are sharp within the resolution of the spectrometer. At intermediate fields a low-energy tail develops, whereas a t very high fields the ion energy is spread out considerably and drops to zero a t values below that of the original peak. Since the maximum energy an ion can have is the full potential difference between tip and mass spectrometer, a decrease corresponds t o ion formation in front of the tip.

k

Y

!e

Y

4

a

I 5000

6000

7000

8000

9 K)

VOLTS

FIG.7. Relative ion peak heights for t h e field ionization of hydrogen by a n oxygen contaminated tungsten tip. Figures indicate peak half-widths. Decrease in peak height a t high fields results from wide energy spread of the total current.

Because of the very sharp curvature of tbe latter, i.e. because of the very high field in its vicinity, a distance of 2A. corresponds to approximately 4 volts, which is within the limit of resolution. Thus i,t can be shown th a t a t very high fields ions may be formed as far as 200A. from the tip. The drop off t o zero a t energies below that of the original peak under these conditions corresponds to the fact that the probability of ionization becomes so high that all molecules approaching the tip become ionized before actually reaching it. It can readily be shown that the spread in energies is not primarily due t o the drop off in field with distance from the tip. The latter is closely approximated by

F I F O = (T/z)'

(14)

110

ROBERT GOMER

TABLE I1 Mean Life of H Atoms in High Fields F (volts/A.)

Tla(SeC.)

0.5 1.0 1.5 2.0

1 . 3 x 10-1 1 . 6 X 10-l0 1.6 X 1 . 7 x 10-14 2 x 10-16

m 1% =

6

1 2

=

lifetime of H atoms in free apace. lifetime of H atoms at a distance of -9.5 / F

T2’(SeC.)

4

x

10-4

1 . 2 x 20-12 8X 2 x 10-16 2 x 10-16

11/72

300 130 20 8 1

k. from a W surfare.

where F is the field a t a distance x from the center of curvature of the tip and F o the field a t the surface, where x = r, the radius of curvature. An inspection of Fig. 7 will show that the broadening of Hz+, for instance,

2

0 2 4

6 8

10

FIG.8. Schematic diagram showing the first step in “field corrosion” of a two-electron bond. ‘I’ - x = heat of binding; x = work function; p = Fermi level; a / 2 = expectation value of electron-to-surface distance.

cannot be explained on the basis of Equation (14), being much more marked for the same percentage interval a t 8000 volts than a t 7000 volts. The first step in the “corrosion” of a two-electron bond can be treated similarly. Figure 8 shows the case of hydrogen with a n assumed bond

FIELD EMISSION

111

strength of 0.5 e.v. and a surface-to-proton distance of l i . It will be seen that the critical field is that which raises the level of the bonding elertrons t o that of the Fermi sea of the substrate. Now the distance a t which this occurs is no longer arbitrary, but -45 bondlength. It is readily seen therefore that a field of -1 X lo8 v./cm. is necessary, and that thermal excitation may influence the average bondlength and hence the critical field. Once a bonding electron has tunneled, the atom is held by very weak forces, and if it was vibrationally excited in the bound state the Franck-Condon principle will insure its moving to a distance where the situation depicted in Fig. 5b applies. We see therefore that a simple criterion for the minimum field can be written as

FInin (‘I’ - x ) / a G H / a

(15)

where x is the substrate work function, ‘I’ the ionization potential, and H the heat of binding of a metal adsorbate complex. a is the expectation value of the surface t o bond-electron distance, -36 bondlength. I n the case of chemisorbed atoms or radicals either st,ep may be rate-controling, if the ionization potential of the neutral atom or radical exreeds the heat of binding, these may be field desorbed without being ionized. 111. SELECTED APPLICATIONSOF FIELDAND IONMICROSCOPY 1. Pattern T y p e s

The images of clean metal tips shown in Plate I A, B , and C are observed only after the most painstakiug precautions and under conditions of extremely high vacuum. The patterns commonly observed after moderate initial heating of the tip or in intermediate vacuum are strikingly different, and fall into two general classes. The first consists of patterns in whirh the emission seems concentrated into narrow regions, or even lines, with extreme contrast between the emitting and dark regions. The second consists of patterns which lack both the extreme contrast and narrow lines of the first group, but are considerably more complex than the clean tip patterns. In both types the crystal symmetry is preserved. Patterns of the second type are obtained when rlean tips are exposed to various gases or when volatile substances are sublimed onto them, and are the result of work-function changes due to chemisorption. These will be discussed in the next section. Patterns of the first kind are due to relatively nonvolatile impurities like carbon, silicon, or (under rertain conditions) oxygen. These patterns show their characteristic form only after heating of the tip to fairly high temperatures, 700 to 15OO0K,and can be most easily explained as the result of overgrowths or surface phases of

112

HOBEHT GOMER

definite orientation with respect to the substrate.* The sharply confined emission often noted in these patterns is probably due to the fact, that surface phases occur as proclivities or built up regions leading to field enhancement and hence to increased emission. Very similar patterns can be obtained by heating clean tips in the presence of the applied field. One may then watch the formation of bright line emission patterns (Plate ID). If the heating is discontinued with the field on, the pattern can be ‘(frozenin.” If the field is turned off with the tip hot, however, the pattern reverts t o the smooth form. Patterns of the first kind may often be changed into those of the second by heating, if the adsorbate reacts with the substrate t o form a nonvolatile phase. Thus oxygen chemisorbed on tungsten or nickel is converted to an oriented oxide (Plate IE). This will be discussed in connection with chemisorption. 2. Molecular Images

Perhaps the simplest surface phases observable are the images of individual molecules. These were first noted by Muller (I), who found that the sublimation of small amounts of various organic compounds gave rise t o small patterns superimposed on the substrate (Plate IC). The mechanism of image formation has already been mentioned. The significant fact, is that sufficient field enhancement and extra magnification can be obtained from entit,ies of a few angstroms in size. Early work indicated a one-to-one correspondence between the patterns observed and the molecules under discussion. It appears, however, that very similar patterns can be obtained from a number of quite different molecules. This is undoubtedly due tlo the fact that molecules are distorted by the surface and the field, as well as to insufficient resolution. It is also probable th a t emission from molecules depends more on the form of the highest energy orbitals than the shape of the molecule itself. Nevertheless, there is ample evidence (1,14) that t,he patterns are due to individual adsorbed molecules, even though the effect cannot be used for their identification. One possible application of molecular images is the estimation of * It must be pointed out that the author’s interpretation of the bright line patterns is not the only possible one. Thus J. A. Beeker (see this volume) interprets them as recrystallization of the tip with facet formation, resulting from the changes in relative surface free energies of various crystal faces caused by adsorption (see also reference 7a). The undoubted field enhancement, known reactivity of oxygen with metals, and temperature dependence of the Ni-Si patterns (vide injra) inclines us to prefer the interpretation given above. In the case of oxygen and Si on Ni it can also be seen visually (although not readily on photographs) t h a t the bright line pattern is superimposed on the normal clean Ni pattern. This suggests very strongly that recrystallization is not involved here. I n the case of carbon on W, Mtiller’s ion emission patterns (1) show that the surface phase consists of atom rows, or ridges.

FIELD EMISSION

113

ionization potentials. The method consists of adsorbing a few molecules and measuring the current from a single molecule photometrically as a function of voltage. Analogues to the Fowler-Nordheim equation may then be used for estimating the “work function,” i.e., the ionization potential of the adsorbate (14).

5. Surface Phases Carbon on tungsten was first studied by Muller and later by Klein (15). We shall not describe this work, but use the behavior of silicon in nickel (16) as an example of the unique suitability of field emission technique for discovering and studying overgrowths. Plates IF-L and IIA-D show the temperature dependence of patterns from nickel tips containing 0.3 per cent Si as principal impurity. The photographs were obtained with the tip at room temperature, after heating a t a given temperature produced no further changes. They are remarkable in that they are equilibrium forms. It is possible to approach a given pattern from above or below in a completely reversible manner. The time it takes for a given pattern t o emerge without further changes depends on the temperature and the initial pattern. Moreover, a pattern will approach its final form via the patterns intermediate between the initial and final ones. Thus a tip heated t o 1400°K and then kept a t 1000°K will approach its equilibrium form by going through the 1200°, 1050”K,etc. ones. These facts indicate that one is observing the equilibrium distribution of silicon on nickel. The concentration of silicon in the surface phase increases with decreasing temperature so that more and more silicon is “sweated out” as the temperature is lowered. At high temperatures the regions of best fit with the substrate are occupied first. As more silicon must be accommodated in the surface phase, rings already present become enlarged. This preserves cohesion within the phase, but leads to a slightly poorer fit with the substrate. At still lower temperatures (Plate IK) a point is reached where the misfit for further enlargement of existing rings becomes uneconomical, so that new regions of the substrate are colonized. At this temperature the rings on the (111) regions can be first observed. At still lower temperatures increases in the surface concentration lead t o the build-up of already existing epitaxed layers, resulting in great field enhancement and correspondingly increased emission from the built-up regions (Plate IIB-D). The lack of emission from the substrate in IIA-D is due to the field enhancement on the built-up regions, which emit a t voltages too low to show the substrate. The embryo-like behavior of the growth indicates that the amount of silicon actually on the surface a t a given moment, rather than the temperature per se, controls the form of the pattern. Temperature is effective

114

ROBERT G O M E R

PLATE I1

(Legends on page

FIELD EMISSION

115

only in controlling the equilibrium concentration. This suggests a simple way of determining the activation energy for the bulk diffusion of Si in Xi. The tip is heated t o a high temperature and then heated t o a lower temperature T1, but only for a time t l short enough t o produce a pattern still far from equilibrium. The tip is then reheated t o the original temperature, and now heated a t some temperature T z for a length of time t z just sufficient t o produce the exact state of the pattern resulting from heating a t T I for a time t l . From a knowledge of temperatures and times, the 6 Kcal. This probably (*orreactivation energy was found t o be 48 sponds t o bulk diffusion, which should be the rate-controlling step. It should be mentioned that most authorities (17) consider the solid solubility of silicon in nickel t o be several per cent in the temperature region of this study. The present sample contained only 0.3 per cent Si. This would indicate that a temperature-dependent fraction of the total finds it more economical, from the free energy standpoint, t o occur as a surface phase. It may be that certain types of catalyst poisoning consist of the formation of surface phases of this kind on normally active regions of the catalyst.

4. Chemisorption and Mobility

of Oxygen on l’ungsten

a. Use of Liquid Helium for Vacuum Technique. One of the most fundamental properties of chemisorbed layers from the standpoint of catalysis is their mobility. It is clearly paramount in determining the mechanism of reactions or adsorption phenomena t o have some idea whether adsorbed atoms or molecules remain rigidly fixed on their initial sites or whether they are free t o migrate over the surface, and if so a t what rate. Up t o now conventional methods have been rather unsuccessful in elucidating this phenomenon, since the interpretation of isotherms is dubious, and surfare mixing experiments difficult. PLATEIT A . Tip of Plate I F after 1269°K. B. Tip of IF after 1218°K. Build-up of corners near (110) beginning. C. Tip of IF after 1185°K. Build-up of corners becoming general. D. Tip of IF after 1085°K. Corners built-up to point where emission from these predominates. E. Tungsten, after cleaning, in liquid helium. x = 4.5 e.v. F. Tip of IIE contaminated on one side with oxygen (dark region). x = 4.5 e.v. G. Tip of IIF after spreading a t 40°K. x = 5.9 e.v. H . Tip of IIG after long exposure to oxygen. x = 6.3 e.v. I. Tungsten, oxygen covered, after heating to 620°K. x = 6.2. J. Tip of IIZ after 700°K. x = 6.18. K. Tip of I I J after 775°K. x = 6.05. L . Tip of I I K after exposure to oxygen. x = 6.13.

116

ROBERT GOMER

The field emission microscope offers a very clear-cut and basically simple method of determining the mobility of adsorbates quantitatively. If it were possible to evaporate the gas under study from a suitable source (e.g., a heatable CuO filament for oxygen) in such a way that only a portion of the tip became contaminated, one could determine how, and a t what temperatures of the tip, migration occurred. If one attempted to evaporate from a gas emitter placed on one side of the tip while the microscope tube was a t room temperature, gas rebounding from the walls would instantly contaminate the whole tip and the experiment would fail.

D

FIG.9. Sketch of field emission microscope assembly for mobility studies. I),inner Dewar; S , screen; A , anode; Y', tip, " A , tip assembly; M , platinum foil mortar, filled with copper wires, oxidized in situ; electric heating of M produces a controllable flux of oxygen; M A , gas emitter assembly; V , vacuum lead, sealed off. Outer Dewar and electrical leads are not shown.

If the tube is kept immersed in liquid hydrogen or helium, however (20-4°K) , the vapor pressures of all gases except helium are negligible and their sticking coefficients on surfaces very high, so that one may carry out the experiment just outlined. Under these conditions gas does not return from the walls, and only that portion of the tip directly exposed to the gas emitter is contaminated. It is possible to heat the tip electrically and t o determine its temperature by measuring its resistance while keeping the tube immersed in liquid helium. If unsilvered dewars are used, the patterns may be observed directly and photographed (Fig. 9).

FIELD EMISSION

117

The use of liquid helium as a means of obtaining vacua of the order of lO-3O mm Hg has tremendous advantages over conventional methods. The latter require laborious baking out a t 400-500°C for hours, followed by gettering. Once a high vacuum is obtained, the admission of gas to the system necessitates repetition of the entire procedure, including the separate heating of all metal surfaces. In the case of liquid helium, however, a system may contain several millimeters of selected gas. Nothing beyond immersion is required to produce the high vacuum instantly. Thus it is possible to clean off a field emitter or t o leave it in any given condition of contamination while the tube is immersed in helium. The tube may be warmed to room temperature, resulting in exposure t o the gas; recooling then produces the vacuum necessary for further study. It will be seen that this is important in the case of oxygen on tungsten, where heating to the usual outgassing temperatures seriously alters the chemisorbed layer. The mobility experiment described above has been carried out for oxygen on tungsten. Concomitantly many interesting facts about the nature of the chemisorbed layer have been learned; these will be described simultaneously (18). It is desirable t o know the approximate coverages of various portions of the tip in these experiments. This is accomplished by comparing the slopes of Fowler-Nordheim plots (Equation 4) of the clean and contaminated tip. The average work function of the latter can then be found. If the work function for the most highly covered tip is then taken as that of e = 1, the average coverage for any other tip condition can be calculated a t once. Since most of the emission comes from the regions of lowest work function, regions of higher e will be underestimated in this way. It is possible, however, t o determine the correct coverages by photometric comparison of the emission from different faces. For most of the observations to be discussed here, we shall speak of average coverage. It should be pointed out here that large decreases in the pre-exponential term of the Fowler-Nordheim equation occur on oxygen adsorption, so that the slope of a log I / V 2 vs. l / V plot should be used for determination of x. These decreases were also found by Miiller (I) for oxygen and by the writer for hydrogen on tungsten. They may correspond, a t least in part, to a decrease in effective emitting area, both micro- and macroscopically. Two distinct types of mobility must be recognized. If the amount of oxygen evaporated is so small th at a monolayer or less is formed on the covered portion of the tip, diffusion occurs only at temperatures above 400°K; this will be described in detail below. If more oxygen is evaporated onto a tip kept init,ially at 4"K, totally different behavior results. Under

118

ILORERT GOMElL

these conditions spreading sets in a t about 40°K. It is possible t o watch the film cover the tip like the unrolling of a carpet or the drawing of a blind, with a sharp, uniformly advancing boundary. The dramatic effect of this experiment cannot be conveyed in writing or in still photographs. The layer thus formed is not itself mobile; if the amount of oxygen initially evaporated is dosed t o "unroll the carpet" only part way, the sharp boundary which is formed will not move until the temperature is raised t o above 400°K or until more oxygen is evaporated onto the tip. The spreading shows not only a lower temperature limit a t about 4O"K, but an upper one a t about 120°K. These results indicate that molecular oxygen is mobile on an oxygenated tungsten surface a t -1O"Ii. The low-temperature spreading consists of diffusion on top of the already covered regions of the surface; oxygen becomes bound on the clean tungsten surface a t the edge of the layer, making it, possible for other molecules t o diffuse over the newly covered region. The situation is analogous t o the storming of a walled and moated town by attackers advancing over the bodies of their dead. The formation of a sharp boundary in diffusion with precipitation is well known (19). Careful work function measurements indicate that about 80 per cent of the layer can be formed b y low-temperature spreading (Plate IIE-G). If the tube is warmed up so that the tip is exposed t o the oxygen accumulated in the tube mm.), slight further changes occur in the pattern, and its work function increment goes up about 20 per cent* (Plate IIH). The exart degree of coverage depends on the speed of spreading. Highest coverage is obtained by slowest spreading. Two conclusions are strongly suggested by these facts. First, the major portion of the chemisorbed layer is formed without appreciable activation; a small fraction of the process occurs slowly. At present it is not possible t o say whether the latter process involves more than a monolayer. It is possible that a more careful study of the time and temperature dependence of the slow step will shed more light on this point. The second concblusion is that the apparent attraction of clean tungsten for oxygen is weakened by orders of magnitude when the surface is only 80 per cent covered. This is true for all portions of the tip within fairly narrow limits, hut may simply mean that dissociative adsorption is very slow at these coverages. Before describing the high temperature diffusion experiments, some properties of the chemisorption layer will be discussed. Changes, including evaporation, arc inappreciable below 800°K for fully or partially covered surfaces. At this temperature and above, irreversible changes occur in the * R. C. 1,. Bosworth (20) finds values of the contact potential of 1.82 volts, which is within the experimental error of the present value, uncorrected for field effect (9).

FIELD EMISSION

119

layer. These are shown in Plates III-L, IIIA-L, and IVA and B. The terms " reversible " and "irreversible '' mean here that warming up and exposure t o oxygen (Plates IIL, I I I D and K ) either result in the appearance and work function of the fully chemisorbed layer or do not. This is a simple way of differentiating between desorption and oxidation. As pointed out previously, apparent decreases in work function can be due t o the formation of surface overgrowths and resultant field enhancement. While desorption changes should disappear on recontamination with oxygen, oxidation patterns should not alter, or do so only in part. The series of changes outlined in Plates III-L, IIIA-L, and IVA and B started with a fully covered tip. When smaller initial coverages were used, irreversible oxidation set in a t higher temperatures, depending on the coverage. This phenomenon was also noted by Muller (1). It will be seen from Plate IIIJ that the field enhancement for some oxidation patterns can be so high that apparent work functions below 4.5 e.v. result. Although the detailed analysis of the oxidation patterns, depending as it does on a number of unknown factors, is almost impossible a t the moment, the general behavior ran probably be explained quite satisfactorily. We note first that the areas on which genuine oxide builds u p most noticeably are the (111) regions. Unlike the (100) vicinals, these fail t o take up oxygen once they have assumed the characteristic triangular shape shown in Plate IIIC-L. Since irreversible oxidation involves a reaction with the substrate, it seems reasonable that this should occur most readily on the most loosely parked crystal fares, i.e., (111) for bcc metals. The dependence of this irreversible oxidation on coverage seems also reasonable. Apparently the free energy per adatom of oxygen is lowest when the latter is chemisorbed on the surface, rather than present as oxide. However, only as much oxygen as there are surface sites can be accommodated this way. If there is more oxygen, say, in the gas phase, the system as a whole will find it more economical t o consist of oxide, where a greater number of oxygen atoms can be accommodated, even though less favorably than the few that could be chemisorbed. It is instructive t o ralculate pertinent free energies. If the heat of adsorption of atomic oxygen on tungsten is taken as 160 Kcal. per mol. and if the entropy decrease is ascribed to the loss of three translational degrees of freedom, so that A S = -36 e.u., it is possible t o calculate the enthalpy and entropy changes shown in Table 111. It will be seen that the equilibrium can be shifted t o the oxide side in the presence of oxygen even at low temperatures. This explains the fact previously pointed out, and well known t o inorganic chemists, that tungsten becomes oxidized in oxygen a t temperatures as low as 400°C. It also explains why oxidation does not occur until much higher temperatures in the absence of gas-

120

KOBERT GOMEH

PLATEI11

(Legends on page 121)

121

FIELD EMISSION

phase oxygen. The data of Table I11 show that WOS is the most likely oxide t o be formed in bulk from the chemisorbate. It is probable th a t some type of surface oxide, intermediate in composition and thermodynamic properties between the bulk oxide and the chemisorbed layer, is actually formed. Nevertheless, the trend of numbers indicated in Table I11 must TABLE I11 Standard Enthalpy and Entropy Changes for the Tran.sition from. the Chemisorbed Layer (Assumed to have Composition WOI) to the Three Bulk Oxides Reaction

(WO) (WO) (WO) (WO) (WO) (WO)

+

%W

+O n +

-++ -

+ WO,

3502-

WOa

wo2

t.gw0, + jgw 3402Y2Ww,O' f.$W20, 3tw

+

AH'

AS"

+ 1 3 . 4 Kcal - 120.0 Kcal - 5 4 . 5 Kcal + 1 3 . 0 Kcal -87.0 Kcal + 1 3 . 0 Kcal

2 . 5 e.u. - 4 3 . 5 e.u. -24.0 B . U . - 3.Oe.u. - 4 3 . 5 e.u. - 6.2e.u. -

hold. Experiments in which oxidized tips were heated to low temperatures (700°K) showed that no changes occurred. Although the equilibrium would favor a chemisorbed layer, it is quite reasonable to assume th a t all rates involved are much too slow to show this behavior in a finite time. Examination of Plates IIIH-I, and IVA shows that the changes occurring in the oxide on raising the temperature consist of a reduction in its amount, indicated by a decrease in field enhancement. The pattern seen just before clean tungsten (Plate IVA and B respectively) is obtained resembles the beginnings of a chemisorbed layer. These facts suggest that removal of oxygen from the tungsten surface occurs by evaporation of the oxide (heat of sublimation of WOs = 110 Kcal.). At very high temperatures the decomposition of oxide may compete with the evaporation. Since very little oxide is left a t these temperatures, it is possible PLATEI11 A . Tip of Plate 11.5 after 900°K. x = 5.92. B. Tip of Plate IIIA after 1013°K. x = 5.85. C. Tip of IIIB after 1093°K. x = 5.83. D. Tip of IIIC after exposure to oxygen. x = 5.82. E . Tip of IIID after 1161°K. x = 5.76. F. Tip of IIIE after 1302°K. x = 4.87. G. Tip of IIIF after 1303°K. x = 4.82. H . Tip of IIIG after 1403°K. x = 4.73. I . Tip of IIIH after 1473°K. x = 4.53. 1. Tip of I I I I after 1558°K. x = 4.34. K . Tip of I I I J after exposure to oxygen. x = 4.5. L. Tip of IIIK after 1643°K. x = 4.61.

122

IZOBEIZT GOMER

(Legends an page i23)

F I E L D EMISSION

123

that rapid cooling then results in small amounts of oxygen remaining as chemisorbate rather than oxide. * A phenomenon resembling oxidation, or a t least multilayer formation, can also be observed a t room temperature. If a tip kept at 300°K and exposed t o oxygen a t 0.1 mm. Hg for varying times is reexamined, no changes in pattern or work function beyond the initial rise to 6.3-6.4 volts are found. However, the total emission, i.e. the pre-exponential part of the Fowler-Nordheim function, keeps decreasing with time, even after periods of days. This suggests adsorption of molecular oxygen in the interstices of the monolayer (presumably atomic). Such adsorption would not add t o the work function since no strong dipole moment is associated with it, but would reduce the total effective emitting area, since the multipoles or weak dipoles associated with adsorbed oxygen molecules would prevent emission at the sites where the latter are adsorbed. On heating to 600°K only a small fraction of the molecular oxygen becomes desorbed. Further heating to 700-800°K results in the formation of small oxide crystallites, appearing more or less a t random over the surface. This suggests that a chemisorption of molecular oxygen occurs and facilitates the formation of oxide. The crystallites so formed are undoubtedly very small, resembling molecular images in shape and apparent size.

* The author’s interpretation of patterns resulting from the heating of oxygen covered tips to temperatures above 800°K as partial oxidation forms is not shared by J. A. Becker (see this volume) who interprets them as desorption stages. It is very probable t h a t some desorption accompanies oxidation, While field emission technique cannot decide this point, Becker’s results with the flash filament technique indicate that there is gas evolution on heating oxygen covered tungsten wires. The readsorption experiments cited here, indicate that desorption, if it occurs, is accompanied by oxidation of the substrate. PLATEIV A . Tip of Plate IIIL after 1673°K. x = 4.97. B. Tip of Plate IVA after 1873°K. x = 4.5. Clean tungsten. C. Clean tungsten tip, contaminated with slight amount of oxygen on one side. D. Tip of IVC after 1800 seconds at 453°K. E. T i p of IVD after 1220 seconds a t 498°K. F. Tip of I V E after 5000 seconds a t 498°K. G. Tip of IVF after 100 seconds at 600°K. Equilibration is now complete. H . Tip of IVG after further oxygen contamination and equilibration. x = 4.89. I. Tip of I V H after further contamination and equilibration. x = 5.52. J. Tungsten tip, after partial contamination by exposure to oxygen at 77°K. x = 5.19. K . Tip of I V J after heating to 416°K for a few seconds. x = 5.18. L. Tip of IVK after heating to 509°K. x = 5.09. M. Tip of IVL after heating to 600°K. x = 5.12,

124

ROBERT GOMER

It has been mentioned that diffusion sets in only a t high temperature if the amount of oxygen evaporated on the tip is less than a monolayer. It is possible t o obtain the activation energy of this process by comparing the spreading times of equal doses of oxygen a t different temperatures. This has been done for low coverages. I t is found th a t the activation energy is 30 & 4 Kcal. Spreading sets in a t 450°K and is fairly rapid (complete equilibration in 100 sec.) a t 600°K. If a small dose of oxygen is evaporated and spread slowly, the patterns shown in Plate I V C 4 are obtained. If successive doses are evaporated and spread, the patterns shown in Plate IVH and I are obtained after equilibration. It is seen th a t I V I corresponds to an average e of 0.55. Although these experiments are not complete, there is no indication that the activation energy for surface migration changes up to e = 0.5. It becomes more and more difficult, however, t o cover the tip by high-temperature diffusion as the coverage goes up. The reason for this is probably th at less and less excess oxygen can be accommodated in the monolayer by the exposed portion of the tip. Thus a t 0 = 0 one might cover the receiving section to 0 = 0.8, resulting in 0 = 0.2 after spreading. A new dose of oxygen can only bring 0 to 0.8 in the exposed section, so that after spreading the tip as a whole would be covered t o the extent of e = 0.6 X 0.25 0.2 = 0.35. Successive saturating shots of oxygen would produce, after spreading, coverages of 0.46, 0.55, 0.61, and so on. If a large dose of oxygen is evaporated onto the tip a t B = 0.5, low temperature spreading occurs, with a resultant average e = 0.8, as previously mentioned. Experiments t o measure the sticking coefficient, which are now being planned, will help to decide the reason for this behavior. Although we have talked of average coverages, it is evident from Plate IVC-I t hat rather $evere inequalities in work function exist a t partial coverages. I n particular, it will be noticed th a t the regions connecting the (110) and (211) faces are the first to darken when diffusion sets in (Plate IVC-G). More oxygen produces darkening of the (211) and (111) regions. Further oxygen produces changes in the (100) faces and their vicinals. This behavior suggests strongly that the dark regions are those which are able to take up and hold oxygen preferentially. It seems improbable that the same amount of oxygen on various faces should produce so very different changes in work function, since the emission from fully covered tips is relatively uniform (Plate IIG and H ) . This explanation is supported by the following experiment. A clean tip was contaminated uniformly by warming the microscope bulb to 77°K for a short time (Plate IVJ). On recooling, the pattern of IVK was obtained. This corresponded t o partial coverage (e 0.4) but not to equilibration. On heating to 500°K and 600°K the patterns of IVL and M were obtained. Although

+

-

F I E L D EMISSION

125

there is only a slight change in overall work function, redistribution of the oxygen has occurred; again the diagonals connecting (110) with (211) are darkest. It may be useful to summarize the information on the surface-phase oxygen on tungsten which has been obtained in this work. Oxygen forms a strongly bound, immobile layer on tungsten. Mobility occurs at 600°K with an activation energy of 30 Kcal. Eighty per cent of the layer seems to form with little activation a t temperatures as low as 40°K. Detailed studies of sticking coefficient as a function of temperature and coverage are planned and will elucidate this point. Desorption from the layer does not occur simply, since oxidation sets in below the expected desorption temperature. Despite the more or less uniform maximum coverage, there seem t o be differences in free energy of adsorption on different crystal faces, as shown by the preferential occupation of sites for partially covered layers when these are equilibrated by diffusion. The vicinals of the (211) faces seem to take up oxygen first, followed by the (111) and then the (100) faces and their vicinals. At maximum coverage the (110) vicinals seem t o have the lowest coverage. Oxygen is mobile on a n 80 per cent covered tungsten surface even below the melting point of 0 2 . The study just described should give an indication of the scope and pertinency of field emission technique for problems of adsorption. T h e present experiments are being extended t o other gases. Thus the mechanism of the heterogeneous hydrogen-deuterium exchange hinges on the mobility of hydrogen on surfaces. Results with hydrogen show th a t mobility in the monolayer sets in below 250°K and is complete a t room temperature. Desorption occurs rapidly and completely below 700°K. I n this case the heat of desorption may be obtained from the temperature coefficient of the evaporation rate, even though this will be a minimum heat, since diffusion to the sites of lowest energy and evaporation from the latter occur. While these experiments are still in a n early stage, it is clear already that H-D exchange can occur by surface mixing at room temperature. The two-dimensional melting of physically adsorbed gases is of interest. The formation of mobile fragments from complex molecules is important in catalysis and can perhaps be studied with this technique. The reader will probably be able to think up any number of applications. 5. M a s s Spectrometric Analysis of Yields f r o m the Ion Microscope

The mass spectrometric determination of ions from the field microscope has already been mentioned (13a). Figure 10 shows a schematic diagram of the apparatus. A small fraction of the ion beam is permitted to penetrate through a 30-mil hole in the screen of a field emission tube into a sensitive mass spectrometer. Electron emission in high vacuum

126

ROBERT GOMER

permits the determination of the tip orientation. Gas a t a pressure up to several microns may then be admitted by variable leaks into th e field emission tube. The spectrometer, built by M. G. Inghram, with whom this work is being done, is equipped with a n electron multiplier detector and has a sensitivity of about 1 ion/sec. The analyzer sees a n area of about 400A.z on the tip, in most experiments a portion of the (110) plane. Before admission of gas the vacuum attainable in the spectrometer and emission tube is of the order of mm. so that oxygen contamination TO MASS SPECTROMETER

FIELD MICROSCOPE ION SOURCE

FIG.10. Schematic diagram of apparatus for the mass spectrometric analysis of field ionization.

of the tip takes place. No attempt a t accurate field determinations has been made. T he fields listed in reference (13a) are probably too high, because of field evaporation of tungsten from the tip itself. The occurrence of this phenomenon was pointed out t o us by Professy E. W. Miiller. However, the fields ranged from about 0.5 t o 3 voltslh. It has been mentioned th at most of the peaks observed show a characteristic broadening with increase in field. Occasionally a peak is observed which does not show this behavior but stays sharp, indicating th a t the ion in question is formed only a t the tip, or a t least a t a fixed distance from the latter. Figure 11 illustrates this behavior for methanol. The parent peak, CH,OH+, shows a well developed low energy tail while the only other prominent peak in the spectrum, COH3+, is sharp. These facts suggest strongly that the latter originates on the surface itself and in fact represents a product of the dissociative adsorption of methanol on oxygenated tungsten. This assumption is substantiated by the following observations. Th e onset of the CH30+ peak with field is

127

FIELD EMISSION

much sharper than th at of the parent peak, corresponding to the existence of a critical field for desorption from the bound state, in accordance with Equation (15). The ratio CH30+/CHsOH+increases with decreasing pressure, corresponding to the fact that CH30 is more strongly adsorbed. Even more striking evidence is provided by the following experiment. If the chemisorption of methanol involves the abstraction of a hydrogen atom, so that the chemisorbed species is CH30, its amount on the tip a t pressures of mm. must far exceed the number of methanol molecules in the gas phase near the tip where ionization can occur, since the density

I

I

I

33

32

31 MASS

30

29

NO

FIG.11. Peak shapes in the field ion spectrum of methanol on tungsten.

of an adsorbed film is roughly that of a liquid. If the high field is applied for very short time intervals, so that the diffusion of methanol to the tip becomes negligible, only those ions which originate from the adsorbed state should be formed in appreciable quantity. By using high voltage pulses of microsecond duration and variable frequency it was possible t o show that the CH30H+peak disappears relative to the CH30+peak as the spacing between pulses is increased. At high pulse repetition rates, approaching DC conditions, the CH30H+peak can be restored. Pulse technique also makes it possible t o estimate sticking coefficients from the variation in peak height with repetition rate, since the latter determines whether the tip will have a chance t o regain its coverage of adsorbate. The variation of peak height with pressure makes it possible

128

R O B E R T GOMER

t o determine what amounts to adsorption isotherms, without the perturbing influence of steady state fields. Since the spectrometer sees only a small portion of a single crystal face it is possible t o vary the tip orientation and study the anisotropy. Work of this kind is now in progress. It is surprising a t first glance that no H+ ions are observed in the methanol spectrum. The reason is most probably th a t the CH30H+ and CH30+peaks appear at considerably lower fields than H+ and H2+ since the ionization potential of the latter is nearly 15 volts. However, the heat of binding of H on oxygenated tungsten is probably low enough to permit its desorption by the field as neutral atoms. Although the formula of the adsorbed species is C H 3 0 its structure could conceivably be CH20H. Experiments with deuteromethanol are planned to settle this point. TABLE IV Ion Yields From Various Gases With a Tungsten Field Ionizition Source l'arcnt gas

Ions observeda (primary)

Ions observedb (secondary)

HzO

a For the sake of clarity this table does not list isotopic peaks. Their magnitudes are in accord with accepted natural abundances. Ions listed as secondary result from collisional or vibrational breakup of primary ions as shown by pressure dependence, peak shape, and apparent fractional mass. Relative yields vary with field, pressure, and tip condition (see text).

*

Results with other gases will be discussed briefly. Table IV shows that only parent ions are produced with gases like Oz and N2 although, at least in the former case, adsorption with dissociation undoubtedly occurs. The reason is that the heat of binding of atomic oxygen on tungsten is so high that all molecules approaching the tip would become ionized before reaching it a t fields high enough t o produce O+ ions from the adsorbate.

FIELD EMISSION

129

I n the case of hydrogen, pulse experiments show that there is weak adsorption of Hz, not H, a t least on the oxygen contaminated tungsten surfaces encountered here. The broadening of the H+ peak with field, indicated in Fig. 7, supports this view. The formation of H+ under the conditions of OUT experiments must therefore be regarded as a secondary process, most probably the loss of a n electron by Hz+. The relatively low heat of binding of H on tungsten makes it improbable that the field for its desorption was not reached in these experiments. It is interesting t o note t ha t the field desorption of H from clean W might also give a broad peak if the heat of binding is low enough to permit a n appreciable amount of neutral atoms to be field desorbed. Polar molecules like HzO show apparent polymerization to a n extent quite impossible in the gas phase a t low pressures. The dipole field interaction, which is of the order of 1 ev., results in a n “artificial” multilayer physical adsorption a t pressures and temperatures where ordinarily only a minute fraction of the first layer would exist. Since multilayer adsorption is quite liquid-like, the high degree of polymerization can be explained. It is interesting to note that a t low fields individual peaks show some substructure, which could be due to alignment differences a t the time of ionization or could correspond to ionization from different layers within the adsorbate. It is hoped t o study physical adsorption near the condensation point a t low pressure with nonpolar rare gas atoms to see if layer structure can be elucidated in this way. A look a t Table IV will show the general simplicity of the spectra. This is due t o the fact that ionization occurs by electron tunneling, which does not excite the ions vibrationally, as is the case with impact ionization. Thus the combination of field ionization and mass spectrometry has considerable analytical potentialities, when it is considered that acetone, for example, gives rise t o 19 peaks of comparable intensity by impact ionization and shows no peaks over 0.1% except the parent in field ionization. Since the methods outlined permit determination of the origin of ions encountered i t is clear that the technique should represent a powerful tool for the study of fragmentation processes associated with adsorption. Ry a felicitous accident i t is just those relatively weak chemisorption processes which are most readily studied by field ionization that are of greatest catalytic interest.

APPENDIX Experimental Methods and Techniques The following is a highly condensed summary of experimental informationuseful in field and ion microscopy. This section is intended t o acquaint

130

ROBERT G O M E R

the reader with the problems involved, rather than to provide him with a n exhaustive guide to their solution. Preparation o f Microscope Tubes. Tubes may be conveniently prepared from round-bottom Pyrex vessels of arbitrary size, but flat screens are preferred by some workers. Screens may be deposited by coating the lower half of a bulb with a suspension of millemite in nitrocellulose solution t o which a few drops of butyl phthalate are added as plasticizer.* After thorough drying (evaporation of excess amyl acetate solvent), the flasks are heated slowly in an oven to 400-500°C. This decomposes the nitrocellulose and leaves a fairly stable phosphor deposit. T h e latter is stabilized further arid made electrically conducting by evaporating aluminum or platinum onto it. A less satisfactory, but much simpler backing, can be obtained by wetting the phosphor deposit with Hanovia liquid platinum, then drying and heating to 200-400°C. The backing obtained by this method is electrically satisfactory, but is not as reflecting, so that light intensity is lost. Tungsten through Nonex seals are used for the tip assembly arid the anode lead (Fig. 2). The latter is connected to the screen by liquid platinum or by a small spring touching the conducting coating. For some applications it is desirable to have a minimum of metal in the tube. It is possible to avoid the metallic backing of the screen by making the auodic part of the bulb conducting. A simple way of performing this operation consists of placing the bulb in ail oven a t 400°C. A lorig Pyrex tube of about 8 nim. diameter is inserted into the center of the bulb. The tube must be long enough t o project about G inches out of the oven. A few crystals of SnC12.2H20are inserted into this section of the tube, which is supplied with a rubber hose, The SnClzcrystals are melted with a Bunsen flame and the operator now blows a few puffs of air into the blow hose, carrying the white fumes from the molten salt into the bulb to be coated. This suffices to give surface resistivitics varying from 10-3 to 3 ohm-cm. without impairment of transparency (21). I n tubes t o be used for ion microscopy, a slightly different arrangement is necessary. As pointed out by Muller ( I ) , it is desirable to make the anode cathode spacing as small as possible, so that the chances of secondary ionization in the region of potential gradients are minimized. A ring electrode, connected electrically to the screen, is therefore inserted and placed as close to the tip as possible. The mean free path of ions is then long compared to the region of gradients. The ring-to-screen distance can be quite large, since ions produced there will drift harmlessly. It is

* The author is happy to acknowledge the kindness and cooperation of Dr. Samuel Isenberg and the Sampson Chemical and Figment Corporation, Chicago, for supplying him with phosphors and instructing him in the technique of application.

FIELD EMISSION

131

also desirable to use conducting glass instead of a metal backing for ion tubes. T i p Assemblies and Etching of Tips. Assemblies of the form shown i l l Fig. 2 are convenient for repeated breaking and resealing. It is best to coat the lead-in wires with glass as far as possible t o minimize outgassing problems. Tip structures may be squeezed on or spot-melded t o the leadins. In the latter case these should be beaded or sheathed with platinum to make good melds. Tips are most easily prepared by electrolytic etchitig. The following represents a few recipes which the author has found serviceable. Tungsten-NaOH, 10 g./l. 0.5-30 volts ac. NaOH solution must be carbonate free. Tantalum-Layer of 40% HF on conc. H2SO4 0.5-30 volts dc. Hydrogen embrittlement is produced by ac. Nickel-Saturated solution of KCIOa in 33% HCl 0.5-30 volts ac. P

FIG.12. Tip assembly for accurate temperature control and measurement in liquid helium. P , press seal; W1, tungsten leads (30-40 mil); W?, tungsten heating loop (5-10 mil); N , nichrome thermal barriers (40 mil); W1, tungsten potential lratla (2 mil); T , tip.

I t is easiest to etch tips if a long (1-2 cm.) section of wire is available for immersion into the etch bath. The wire to be etched is examiricd periodically under the light microscope until the tip cannot be resolved a t 500-fold magnification. A carbon rod is satisfactory for the other electrode, or a platinum crucible containing the etch solution may be used. After final etching, tips and assemblies are rinsed in distilled water. 5-15 mil wires are satisfactory, but smaller sizes can be used. Tip assemblies used for work at helium temperatures may be modified as shown in Fig. 12. The inclusion of the nichromc sections is necessary if temperature control is desired. Since both the heat capacity and the electrical resistivity of tungsten at helium temperatures are extremely low and its heat conductivity is very high, a rise in temperature from 4°K to about 1000°K corresponds to changes of a few milliamperes in the heating current if all-tungsten assemblies are used. The nichrome sections act as thermal barriers, since alloys do not lose their high-temperature thermal properties a t 4"K, and permit fine vontrol of temperature. For

132

IWBERT GOlMEll

accurate determinations of the latter by resistance measurements, it is necessary to use potential leads across a very short section of the tip structure, since the nonuniformity in a 4 cm. loop of 5 mil tungsten wire can be as high as 70" a t 700°K. Vacuum Technique. There is a growing literature on ultra high vacuum technique by convent>ionalmethods (22). It will suffice t o say that vacua of the order of lo-$ t o lo-" mm. Hg ('an be obtained readily with the use of either oil or mercury diffusion pumps. The crucial points are : I . System arranged so that all parts up to and iricluding oiie cold trap may be baked out a t 400-450°C for a t least a n hour. This means 1 1 0 stopcock grease beyond the first cold trap (which may be kept cold during the out-baking of the system and other trap). 2. Provision for the heating of metal parts t o a t least 800°C. Very short lengths and screen coatings need not be so heated. 3. Provision for sealing off from the pump, either by collapsing a glass pumping tube under vacuum or by the use of Alpert valves (22). 4. Provision for gettering. A side bulb containing an electrically heatable tantalum filament serves this purpose. After initial heating of the tantalum t o the evaporation temperature while the system is pumping, a film of tantalum is deposited on the walls of the gettering bulb after seal-off.

These methods may be replaced by the use of liquid helium as coolaiit where this is feasible. The sealed off tube is immersed in the cryostat, helium is admitted and the vacuum is attained. T o prevent the diffusion of helium through the glass walls of the tube, the cryostat and (Iontents are cooled t o liquid nitrogen temperature before helium is admitted, or tubes are constructed from Lime Glass 1710 (Corning). The design of cryostats and the handling of liquid helium are beyond the scope of this vhapter. If the tubes do not contain hydrogen, excellent vacua niay be achieved by using liquid H2 as coolant. Elwtrical. Work function determinations require the accurate nieasureinent of currents of 10-9 to 10-6 amperes t o one part in 103, and voltages of several thousand volts to one part in lo4. The reason for using very low currents for the examination-contaminated tips is that one wishes to kecp the applied field as low as possible in order t o minimize changes in iniagcpotential and contact-potential errors. Currents may be measured with rlectromctler type dc amplifiers; vibrating reed electrometers are more aatisfactory. Standard circuits may be used for the voltage supply, but must be d l regulated and free from ripple. The latter is important t m a use a field emission tube represents a highly nonlinear impedance.

FIELD EMISSION

133

Accurate voltage measurement is most simply done by using a high (20-50 megohm) bleeder resistance and using a potentiometer across a small fraction of it. Since the reproducibility of measurements depends on the constancy of the high and low resistances, these are best prepared from standard wire-wound resistors. I n the case of the high resistance, the maximum voltage drop across individual resistors must not be exceeded. General Hints. Despite the minute size and extreme delicacy of fieldemission tips, remarkably long life can often be obtained by observing a few precautions, which have been learned the hard way b y workers already active in this field. High voltage should not be turned on until the tip has been outgassed a t high temperature. The raw tips obtained after etching are very much sharper and are often multiple. If a high field is applied to such a structure, loss by pulling off or vacuum arcing is probable. After the tip has been blunted and annealed by heating, this is much less likely to occur. High voltage should not be applied if the pressure in the tube exceeds 5 X lo-’ mm. Although one may examine the tip a t higher pressures, the chances of loss by ion bombardment are greatly increased. It is unsafe to heat the tip in the presence of fields. Build-up and eventual pulling off of the tip occur. The critical temperature for this depends on the field and on the tip metal. For low melting metals, overheating of the tip (resulting in excessive blunting) must be avoided.

REFERENCES 1. For an excellent summary of work u p to 1953, see “Feldemission” by Muller, E. W., Ergeb. erakt. Naturwiss. 27, 290-360 (1953). 2. Fowler, R. H., and Nordheim, 1).W., Pmc. Roy. Soc. A l l l , 173 (1928). 3. Mayer, J. E., and Mayer, M. G., “Statistical Mechanics.” Wiley,New York, 1940. 4. Nordheim, L. W., Proc. Roy. Soc. A121, 628 (1928). 5. Dyke, W. P., and Trolan, J. K., Phys. Rev. 89, 799 (1953). 6. Dyke, W. P., and co-workers, private communication; Dyke, W. P., Trolan, J. K., Dolan, W. W., and Barnes, G., J . Appl. Phys. 24, 570 (1953). 7a. Herring, C., and Nichols, M. H., Revs. Mod. Phys. 21, 185 (1949). 7b. Smoluchowski, R., Phgs. Rev. 60, 661 (1941). 7c. Drechsler, M., and Muller, E. W., 2. Physik 134, 208 (1953). 8. Gomer, R., unpublished. 9. Gomer, R., J. Chem. Phys. 21, 1869 (1953). 10. The problem of resolution is discussed inter alia in the following papers: Miiller, E. W., 2.Physik 120, 270 (1943); Richter, G., 2. Physik 119, 406 (1942); Gomer, R., J. Chem. Phys. 20, 1772 (1952). 11. Muller, E. W., 2. Physik 120, 261 (1943). 12. Miiller, E. W., 2.Physik 131, 136 (1951). 13a. Inghram, M. G., and Gomer, R., J . Chem. Phys. 22,1279 (1954); Gomer, R., and Inghram, M. G., J. Am. Chem. SOC.77, 500 (1955). 13b. Kirchner, F., Nuturwiss. 6, 136 (1954). 14. Gomer, R., and Speer, D. A,, J : Chem. Phys. 21, 73 (1953).

134

ROBERT GOMER

15. Klein, R., J . Chern. Phys. 21, 1177 (1953). 16. Gomer, R., J . Chern. Phys. 21, 293 (1953). 17. Hansen, M., “Der Aufbau der Zweistofflegierungen,” pp. 947-51. Springer, Berlin, 1936. 18. A very preliminary account of some of this work will be found in Gomer, R., and Hulm, J. K., J . Am. Chenz. Sac. 76, 4114 (1953). A more detailed version will be published soon. 19. Fujita, H., J . Chern. Phys. 21, 700 (1953). 20. Bosworth, R. C. L., Trans. Roy. Sac. (N.S.W.) 79, 53, 166 (1946). 21. Gomer, It., Rev. Sci. Znstr. 24, 993 (1953). 22. Alpert, D., J . A p p l . Phys. 24, 860 (1953); see also Alpcrt, D., Rev. Sci. Znslr. 24, 1004 (1‘358).

Adsorption on Metal Surfaces and Its Bearing on Catalysis JOSEPH A . BECKER Bell Telephone Laboratories. Inc., Murray Hill.New Jersey Page I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 I1. The Importance of Adsorption for the Theory of Surface Catalysis . . . . . . . 138 111. The Adsorption of Cesium on Tungsten Using Thermionic Emission . . . . . . 141 1. Experimental Procedure . . . . . . . ............................... 141 a . Determination of the Amount of Cesium Absorbed . . . . . . . . . . . . . . . . 143 b . Positive Ion Emission . . . . . . . ............................... 145 2 . Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

3 . Interpretation of the Experimental Results in Terms of an Energy-Level Diagram and Activation Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV . The Adsorption of Nitrogen on Tungsten as Deduced from Ion Gauge and Flash Filament Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......................................... 2 . Experimental Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 . Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Interpretation of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 The Bearing of These Experiments on Catalysis . . . . . . . . . . . . . . . . . . . . . V The Adsorption of Oxygen on Tungsten as Observed in the Field Emission Microscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . ....................... 2 . Experimental Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 . Experimental Results and Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . a . Oxygen Adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....................... b . Oxygen Desorption . . . Adsorbed Molecules . . . . . . c . Seeing Individual Ato d . Conditions Necessary to Observe Adatoms and Adrnol e . Changes in Surface Free Energy Due to Adsorption . . 4 . The Bearing of These Experiments on Catalysis . . . . . . . . VI Other Adsorption Experiments with the Field Emission Microscope . . . . . . . 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Adsorption of Tungsten on Tungsten . . . . . . . . . . . . . ............... 3 . The Desorption of Positive Hydrogen Ions by High Fields . . . . . . . . . . . . 4 . The Desorption of Barium on Tungsten by High Fields . . . . . . . . . . . . . . . 5 . Preliminary Catalytic Experiments with the Field Emission Microscope a . The System Hz and O2 on W . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....................... b . The System C and 0 2 on W . . . . . . VII . Discussion of Previous Work on Adsorption., . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Introduction . . . . . . . . . . . . . . . . . ........................ 2. Langmuir and Associates . . . . . . . ........................ 3 . Frankenburg and Associates . . . . . . . . ........................ 135

.

.

. .

.

152

164 168 173 175 175 178 180 180

192 192 192 193 194 197 197 198 200 200 200 202

136

JOSEPH A. BECKER

4. Roberts, Rideal, and Their Associates.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Beeck and Associates.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Page 205 208 210

I. INTRODUCTION Modern chemistry has been enriched and elucidated by the description of chemical reactions in terms of electronic transfers and the energies involved in such transfers. Recently there has been an increasing tendency t o describe catalytic processes in these same terms. It has been realized for a long time that the key to an understanding of heterogeneous or surface catalytic processes is a thorough knowledge of adsorption phenomena. I n these phenomena electronic transfers play a particularly important role. Since small amounts of adsorbed materials have a pronounced effect on the rate of emission of electrons from metal surfaces and since electron currents can be measured easily to a high degree of precision, electron emission can be used as a means of accurately measuring small amounts of adsorbate. From the way in which the emission current varies with the amount of adsorbate it is possible to deduce the kind and degree of electron transfers between the adsorbate and the metal surface. These electron transfers also determine the nature and value of the forces which hold the adsorbate t o the surface. Most catalytic reactions involve a number of species of atoms and molecules. To deduce the mechanism of the reaction and the forces between the various species and between the species and the surface is obviously a complex procedure, but the problem is simplified by a study of the adsorption phenomena of a single species of atom or molecule. Such studies have shown that when some molecules are adsorbed on some adsorbents, the molecular bond is broken and is replaced by two bonds with the adsorbent; the admole is changed to two adatoms. A surface chemical reaction has taken place and the adatoms are said t o be chemisorbed. If a t sufficiently low temperatures this reaction does not take place, the adsorbed molecules are not broken up into two adatoms, and the admoles are said to be physisorbed. Above a certain temperature the rate of the reaction is sufficiently rapid to be appreciable; a t higher temperatures, the rate is very rapid. Such reactions have led to the concept of a n activation energy, that is, the energy which must be given t o a n admole to convert it into adatoms. Even if a n admole is not completely dissociated, i t is to be expected th at the strength of the molecular bond has been weakened as a result of the adsorption; hence it is likely th a t the probability of reaction with other adsorbed species will be quite different from what it is between the two species in free space.

ADSORPTION ON METAL SURFACES

137

It has been customary to express heat or energy involved in such reactions in kilocalories per gram mole. Since in this article electronic transfers are emphasized, energies will be expressed in electron volts (ev.). This unit is the energy which an electron acquires in a potential difference of 1volt. One electron volt per molecule is equivalent to 23.06 kcal./g. mole. A still more complete insight into the nature of adsorbed species can be obtained from experiments: (I) on thermionic emission, ( 2 ) with the field emission microscope, and ( 3 ) with the ion gauge. From some thermionic experiments, particularly with cesium adsorbed on tungsten, it is learned t ha t ( a ) Cs can be adsorbed as positive ions as well as adatoms; ( b ) as the concentration of adsorbed cesium increases, the ratio of adions to adatoms decreases; (c) the forces produced by adions are long-range forces which have appreciable effects over distances of 10 t o 20 atom diameters; (d) adatoms and adions can migrate over the surface a t much lower temperatures than those a t which they evaporate from the surface. From experiments with the field emission microscope it is learned th a t for a system like oxygen on tungsten ( a ) the crystallographic plane of the tungsten has a marked influence on the adsorption properties; ( b ) the heat of adsorption increases with the number of W atoms a particular 0 atom can contact; (c) the heat of adsorption for the first layer, in which 0 atoms make first valence bonds with W atoms, is about 4 ev., for the second layer, in which 0 atoms make second valence bonds with W atoms, only about 2 ev.; (d) a t a constant pressure the rate of adsorption is constant until the first layer is complete, and for the second layer it is slower b y a factor of 100 or more; ( e ) beyond the second layer oxygen is adsorbed as admoles of 02, 0 4 , 0s. From experiments with a modern ionization gauge and for a system like nitrogen on tungsten it is learned that ( a ) only a fraction of the nitrogen molecules that strike the tungsten surface stick to it and become chemisorbed, this fraction being called the “sticking probability”; ( b ) this sticking probability is about 0.4 a t 300°K.; (6) it is constant t o about one layer and then decreases rapidly until at two layers it is only (d) the activation energy increases with the amount adsorbed beyond the first layer. These and other experiments will be discussed in greater detail later. This article will stress, throughout, the importance of starting with a clean surface and working with a vacuum system which is so good that the surface stays clean except for the adsorbate being studied. The chief reason why we choose cases in which tungsten was used as the adsorbent is the fact that tungsten can be heated t o 2400” or 2600°K. for hours and thereby be freed of all impurities. After a series of experiments with a particular adsorbate, the tungsten surface can be cleaned again b y flashing

138

JOSEPH A. BECKER

it a t 2300°K. for a short time. The same surface can thus be used repeatedly without dismantling the apparatus. It is much more difficult t o free metals like nickel or iron from impurities such as carbon or oxides. Besides starting with a clean surface, one must keep the surface free from contaminants for the duration of the experiment. This requires a vacuum system in which the pressure can be reduced t o mm. Hg or lower. The necessity for such low pressures follows from the now well-established fact that a tungsten surface will adsorb one layer of nitrogen in about 1 sec. a t a pressure of mm. Even a t a pressure of lo-* mm. it takes only 100 see. for one layer of nitrogen to be adsorbed. Other experiments make it highly probable that oxygen, hydrogen, or water vapor will contaminate metal surfaces in comparably short times. I n view of these recently established results it is probable that most of the adsorption data in the literature are not characteristic of the adsorbents reported but were obtained for adsorbents covered with unknown amounts of unknown contaminants. Many observers, from Langmuir on, have pointed out t ha t the adsorption characteristics of a surface can be drastically altered by less than one layer of contaminant. I n the following discussion of some earlier work in the field of adsorption in the light of the experimental results and concepts in this article, we shall attempt to evaluate how much of the results can still be relied upon and what results should be reexamined with modern tools and techniques.

11. THE IMPORTANCE OF ADSORPTION FOR OF SURFACE CATALYSIS

THE

THEORY

Catalytic experiments and processes are usually performed with finely divided powders which expose a large variety of crystallographic planes, are full of pores, and contain accidental or intentional impurities. The surface of such powders cannot be freed completely from adsorbates introduced in the production of the catalyst. Numerous experiments have shown that the kind of reaction products obtained and the yield depend on a great number of variables, a few of which are ( a ) the particle size, ( b ) the exact method used in preparing the catalyst, (c) the amount and kind of activator used, ( d ) the atmosphere and maximum temperature used in pretreating the powder, ( e ) the kind and amount of impurities in the reactants, (f) the length of time and conditions of use. These powders are rarely heated above 900°K. They undoubtedly have been exposed t o O2 or HzO before being used as catalysts. If they behave a t all like tungsten, such temperatures will not desorb the first layer of adsorbate. This may still be true if they are heat-treated with hydrogen. These remarks do not mean that all deductions from such experiments are

ADSORPTION ON METAL SURFACES

139

worthless; far from it. What we wish to emphasize is that the system being studied in such experiments is not fully defined and that deductions which do not take into account the effects of such residual films may be erroneous. I n many cases it is probably true that these residual films are beneficial or even necessary t o get the desired result; if so, their presence should be recognized. I n comparison with the complexity and difficulty in defining the system in most catalytic processes, adsorption experiments are relatively simple and easy, particularly if they involve adsorbents which can be cleaned at very high temperatures, are performed in enclosures which can be thoroughly outgassed, and use sensitive detectors. Under such circumstances it is possible t o work with adsorbents th a t have a smooth that consists of only a relatively surface, with a n area of about 1 small number of single crystals. I n the field emission microscope, the area being studied may be as small as cm.2 and consist of a single crystal. I n a comprehensive and fundamental study of adsorption it is best to start with the simplest kind of adsorbate, namely a monatomic vapor such as Cs. Next it is desirable to study a simple molecular gas such as 0 2 , Nz, or Hz. After these systems are understood, one can pass on to molecular compounds such as HzO or CO. After th a t one is in a position t o study a simple catalytic reaction such as 2Hz O2 = 2Hz0. In investigating such a sequence of experiments the following fundamental questions arise. When Cs is adsorbed on W , will it evaporate as an ion or as a n atom? Is it adsorbed as an atom or as a n ion? Can both adatoms and adions exist on the surface? Will the answers depend on the concentration of adsorbed Cs? When oxygen is adsorbed on W , will it exist as adions, adatoms, or admoles? Which species of particle will evaporate most readily? When HzO is adsorbed on W , will it decompose? Will H, Hz, 0, 02,and HzO exist on the surface? Which species will evaporate most readily and a t what temperatures? What will be the influence of one species on the behavior of the others? Lest such questions be considered idealistic and incapable of being answered, let the reader be assured that some of them already have been answered by experiments and analyses to be described later. These experiments will show th at Cs adions do exist and that they produce strong fields which produce large forces on the electrons near the surface; these fields also modify the behavior of other adions or adatoms over distances of 20 to 40 X cm. Because of these long-range forces, the adsorption properties of Cs are greatly dependent on the concentration of adsorbed Cs. I n other cases in which the tendency t o form ions is much smaller or in which the ionic radius is smaller, the electrical fields are smaller and die off more rapidly with dist,ance from an adion. For

+

140

JOSEPH A. BECKER

these the forces are short range and the adsorption properties do not vary so drastically with the adsorbate concentration in any one layer. Experiments with the field emission microscope show in a striking manner that the adsorption properties of a metal vary considerably from one crystallographic plane t o another. The sticking probability, which determines the rate of adsorption, may be about 100 times smaller on one plane than on another; as a result, a t a fixed low pressure one plane may build up t o only one layer while another builds up to two layers; at a higher pressure, both planes may build u p to two layers before a steady state is reached. If the surface is exposed to a high pressure until the whole surface is covered, and if the pressure is then reduced to a very low value, the rate of evaporation a t a particular temperature from one plane may be hundreds of times greater than from a second plane. From such experimental facts one may safely conclude that the catalytic activity too will be found t o vary for different crystallographic planes. From the temperatures a t which the evaporation rate for a particular plane is equal t o about one layer/min., one can calculate approximate values of heats of adsorption and on this basis conclude that the heat of adsorption varies by about 20% for different planes; for a given plane the heat of adsorption may vary by a factor of two depending on whether the oxygen is bonded to tungsten atoms which are unattached t o other oxygen atoms or to atoms which have already made one bond t o a n oxygen atom. The experiments also suggest that the heat of adsorption t o a tungsten atom depends on the number of bonds this tungsten atom makes with other tungsten atoms, also th at in certain cases where the spacing of “sites” on the tungsten plane is just right, so that oxygen atoms may just contact oxygen atoms on neighboring sites, there is a n increase in the heat of adsorption. So far we have discussed the effects of the adsorbent on the adsorbate. We shall now describe one case in which the adsorbate has a n appreciable effect on the adsorbent. Langmuir (1) showed that when tungsten is heated t o temperatures in the range of 1200” to 2000°K. in oxygen a t pressures in the range of to 10 mm., WO3 is formed and evaporates at quite appreciable rates. I n a vacuum the evaporation rate of tungsten atoms in this temperature range is negligible. From this we can conclude tha t when three oxygen atoms are bonded to one surface tungsten atom the bond strength of this atom to other tungsten atoms is reduced b y a factor of two or three. We have performed this kind of experiment in a field emission microscope, in which the tungsten surface which is being observed is the tip of a needlelike point. When this point is heated t o 2500°K. in a vacuum, the tip assumes a shape which closely approximates a hemisphere. When the point is then heated t o 1300°K. in oxygen a t a

ADSORPTION ON METAL SURFACES

141

pressure of about mm. for a few minutes, the shape of the tip is distorted in such a way that one set of planes (the 111 planes) sticks out like mountain peaks on the surface of the earth. Apparently the WO, formation and evaporation are not so rapid on this plane as on other planes. These planes are “eroded” away leaving the one plane sticking out. This experiment again shows the importance of rates of adsorption and rates of reaction on different crystallographic planes. Many cases have been reported showing that the amount of a given adsorbate is decreased by a previously adsorbed different adsorbate (2) ; usually, too, the heat of adsorption is decreased, but not always. The presence of one layer or less of oxygen on tungsten increases the amount of cesium adsorbed (3) and also increases the heat of adsorption of cesium, as determined by evaporation rates. On the other hand, a small amount of adsorbed oxygen decreases the sticking probability of nitrogen on tungsten by as much as a factor of ten. One is tempted to generalize by saying that if a strongly electronegative impurity is adsorbed, it decreases the bond strength of a weaker electroiiegative adsorbate but increases the bond strength of an electropositive adsorbate. The converse should also be true. This kind of rule would be expected from the electron-transfer concept. The cases of adsorption discussed in this section and the rather detailed picture of the forces due t o the interaction of the adsorbate species with the structural arrangement of the adsorbent atoms could hardly have been obtained by conventional catalytic experiments. Yet these same forces which are found in the simpler cases of adsorption must be present in the more complex cases of catalytic reactions. I n arriving at this picture, we have found thermionic emission, the field emission microscope, and the modern ion gauge very useful tools. The next three sections will describe these tools in greater detail.

111. THE ADSORPTION OF CESIUMO N TUNGSTEN USING THERMIONIC EMISSION The study of the adsorption of cesium on tungsten is particularly fruitful because the rate of adsorption and the amount adsorbed can be measured easily and also because it is possible t o measure the rate of evaporation of electrons, atoms, and positive ions from the same surface. From these data also can be deduced the nature of the adsorbed species and the forces between the adsorbed species and the tungsten atoms (3,4). 1. Experimental Procedure The tube used in these experiments consists of a cylindrical glass envelope highly evacuated (see Figure 1). A long tungsten wire passes down the axis of the envelope.

142

YOSEPH A. BECKER

Three cylindrical collectors surround the wire. The central cylinder collects the emission current from the central part of the tungsten wire. Since data are taken a t rather low temperatures, the end cooling effects are very pronounced; hence the central cylinder should be short so as t o collect current from only about 10% of the length of the wire. All three collectors are usually at the same potential, but only the current from the central collector is recorded. A side tube which contains metallic cesium is so placed that its temperature can be controlled separately. The entire glass envelope, evacuated by diffusion pumps, is baked and cooled. All metal parts are then heated until they no longer give off gas. The baking and heating are repeated once or twice.

FIG. 1. Schematic diagram for cesium on tungsten tube. The tube is sealed off while the filament and collectors are hot. The tungsten wire is heated to 2600°K. for a few minutes and a t 2400°K. for many hours. This stabilizes the properties of the wire; in particular it permits the crystals to grow t o such a size that no further growth takes place during subsequent use. The t i p of a glass capsule containing metallic cesium is then broken with a magnetic hammer. The vapor pressure of the cesium in the main tube is varied b y raising or lowering the temperature of the side tube. This vapor pressure can be deduced from kinetic theory and the rate of arrival of Cs atoms per square centimeter per second (3,4). This arrival rate, A, can be calculated from the measured value of the saturation positive ion current. Langmuir first showed that if the tungsten is hot enough every cesium atom t h a t strikes the surface evaporates off as a positive ion of cesium. This saturated positive ion current, i,,, can be measured easily with a medium sensitive galvanometer when the collector is negative. It is related to A by the equation

i,, = A X electronic charge in coulombs X tungsten surface area

(1)

143

ADSORPTION O N METAL S U R F A C E S

a. Detcririination of the Arnount of Cesium Adsorbed. The amount of cesium adsorbed can be determined from the following experiments. The tungsten filament is heated so hot that its surface is clean. The filament is suddenly cooled, a t time zero, to a temperature of about 600"K., a t which the electron emission current, i,, is easily measurable. The collector is made positive with respect to the filament and the current is observed as a function of time. A t first the current is very small, but it increases rapidly and steadily with time until it reaches a maximum or optimum value at time t,,t; then it decreases rapidly until it reaches a final steady value which depends on the temperature and the arrival rate. The optimum current forms a sharp maximum and tOptcan be determined quite precisely. Figure 2, curve a, shows a plot of i. vs. t.

D

FIG.2. Electron current versus time during which cesium is adsorbed; the maximum current is taken as 1.0. The value of tOptis independent of the testing temperature provided t h a t this temperature is low enough that a n optimum is observed. From this fact it is concluded t h a t the sticking probability, s, is unity. The sticking probability is defined as the ratio of the cesium atoms adsorbed per second t o the cesium atoms striking per second. It would seem very unlikely t h a t s could have a value less than unity and still be temperature independent. If we take s = 1.0, we can readily compute the number of cesium particles adsorbed per square centimeter a t the optimum activity. This is equal t o A X tOpt.It is found to be independent of A even though A number, NOpt, is varied over a tenfold range. The value of NOpt is found t o be 3.7 X 1014 cesium particles/cm.a. This is nearly equal to the number of sites available t o cesium atoms on a 110 plane of tungsten. If s had a value less than 1.0 such a simple result would not have been obtained. Later we will dwell a t some length on the question of how the cesium is adsorbed on the surface with respect to the tungsten atoms.

It is convenient t o define a quantity e as the fraction of the surface covered by an adsorbate. The surface will be said to be covered with a monolayer, and 0 will be equal to 1.00 when every available site is occupied. The number of available sites per square centimeter will differ from one crystallographic plane to the next; it will also depend on the size

144

JOSEPH A. BECKER

of the adsorbate particle. On the 110 plane of tungsten there are 1.4 X loL6sites/cm.2; but since the cesium atom is very nearly twice the size of a tungsten atom, only one out of every four sites is available t o a cesium atom. For a n adsorbate atom, which is the same size as a tungsten atom, all sites would be available. On the 100 plane of tungsten there are 1.0 x loL6sites/cm.2; and again because of the size of the cesium atom only one in four is available. From these numbers it follows th a t the number of adsorbed atoms when 0 = 1.00 for the 110 plane is 40% larger than for the 100 plane. It would be logical to define a monolayer as the number adsorbed per square centimeter when e = 1.00. With that definition the amount adsorbed per square centimeter for a monolayer would depend on the crystallographic plane. The utility of the concept of a monolayer is that one or more observable properties of a surface change drastically when a monolayer is reached. It would be highly desirable if the term monolayer were clearly defined whenever it is used in an important treatise. Sometimes a monolayer has been defined as the greatest number of adsorbate atoms which can be packed on 1 em.* regardless of the nature of the adsorbent. Such a definition might be useful for physical adsorption, in which the adsorbate is little affected by the structure of the surface; however, it would be of little use in chemisorption, in which the surface structure of the adsorbent greatly influences the adsorption effects. I n view of this discussion, why compare the experimentally determined value of cesium adsorbed per square centimeter a t the optimum emission with the number of available sites per square centimeter on the 110 plane rather than on any other plane? There are two reasons: (1) the 110 plane is the densest packed plane and is more likely to form on a tungsten surface than any other plane and ( 2 ) Martin ( 5 ) has shown that in the case of cesium on tungsten near the optimum activity, the 110 plane contributes more current than any other plane. Hence in the experiments described above, tOptis observed when the 110 plane reaches its optimum. The fact that the electron emission changes its trend abruptly when all the available sites on the 110 plane are filled fits in with the concept of a monolayer. Next i t may be explained how 8 is determined after a steady state has been reached a t some particular pressure and temperature. The temperature should be high enough SO that 0 is less than 1.0. At time zero the temperature of the filament is suddenly reduced to a testing temperature and the time t l to reach an optimum is observed. During this time the surface is being covered by the fraction (1 - O), which must be equal to tl/topt. Hence e = 1 - tl/toPt. (2)

ADSORPTION ON METAL SURFACES

145

By repeating the experiment at different steady-state temperatures one can determine tl as a function of T. From this a plot can be made of 8 vs. T for a constant A , which means a constant vapor pressure, p . Changing the temperature in the side tube can establish a new value of p or A in the tube, and another 8 vs. T curve can be determined. From such a family of 8 vs. T curves one can plot a family of A vs. 8 curves for constant values of T . Since the e values were determined for a steadystate condition, the arrival rate, A , must be equal to the evaporation rate, E,, expressed in atoms per square centimeter per second. Hence the A vs. 8 curves are also E , vs. 8 curves a t a constant T . Such a family of isothermal curves together with a companion family of isothermal curves of sticking probability, s, vs. 8 would completely describe the adsorption behavior. At a given A and T , 6 will reach th a t value for which AS = E,* (3a)

If a steady state does not prevail, then the rate a t which the amount adsorbed per square centimeter, N , changes with time t is given by

_ d N - A s - E, at

In the case of cesium on tungsten the problem is simplified for two reasons: (1) s = 1.0 for all values of 8 less than unity and for all testing temperatures a t which an optimum is reached and (2) E , is negligibly small compared with A for these same conditions. Hence, dN/dt is simply equal t o A . For 8 > 1.0 it appears that s is still unity; but when 8 gets near t o its steady-state value, E , becomes comparable with A. At the steady state 8, E, = A s = A. b. Positive Ion Emission. A hot tungsten surface partially covered with cesium not only evaporates electrons and cesium atoms, it also evaporates positive cesium ions. At a constant temperature as 0 increases, the positive ion current, i,, increases, comes to a maximum when 8 is about 0.025, then decreases to very small values at 8 of about 0.25. To establish this is somewhat more difficult than it is t o establish the course of electron emission as a function of e a t a fixed temperature. Several methods for investigating the emission of positive ions as a function of 0 and T were given in the original paper ( 3 ) . Only one method will be outlined here. Suppose the vapor pressure is adjusted so that the arrival rate, A , is about 2.6 X 10l2 cesium atoms/cm.2/sec. If the temperature of the tungsten filament is about 550°K., the surface will be covered with more * E , does not include those atoms that strike but do not stick.

146

JOSEPH A. BECKER

than a monolayer: 0 > 1.0. The steady state 0 will be the same whether the collectors are positive or negative. As the temperature is increased in small steps, 8 is decreased because more atoms evaporate than arrive. When the temperature reaches about 790°K.) a small number of positive ions evaporate as well as atoms provided that the collector is at a negative potential. Now the steady-state condition is

A

=

E,

+ E,

(4)

Under these conditions 0 will be about 0.24 and will be slightly smaller if the collector is negative than if it is positive. As T is increased, i, increases approximately exponentially and 8 decreases t o about 0.16.

1-

FIG.3. The variation of the positive ion current as the tungsten temperature is increased and decreased.

Figure 3 shows a plot of i, vs. T in the region a t o b. At a critical temperature T,I, which is a function of A , the ion current increases t o much larger values. For A = 2.6 X 10l2, TC1is about 820°K. At this temperature more cesium ions leave the surface than atoms arrive and ip exceeds the value which corresponds to A . This excess current lasts only a short time and soon i, becomes steady. We are then at point c in Figure 3. 0 is found t o be about 0.01. As T is increased still more, i, remains constant since it is saturated or limited by the arrival rate. This is the region in which every cesium that strikes the hot filament evaporates as a positive ion. I n this region c t o d, 8 slowly decreases toward zero. Beyond d two spurious currents add to the true value of i8,: (1) owing to the power dissipated in the filament, the temperature of the collectors and glass walls increases slowly, increasing A and hence i,; ( 2 ) the light from the filament ejects photoelectrons from the collector, which give an apparent increase in i,. When T is decreased from point d t o point c, i, does not change but 0

ADSORPTION ON METAL SURFACES

147

increases slightly to 0.01. I n fact i, maintains its value of iaPdown to point e in Figure 3, when the temperature reaches a second critical value Tcz.From c t o el 8 increases from 0.01 to 0.025. For A = 2.6 X T,z is about 770°K. At this temperature i, changes with time from point e t o point f. During this time 8 increases from 0.03 to about 0.36. At first i, decreases slowly, then more rapidly, and approaches a constant value at point f. As T is decreased below Tc2, i, follows the curve from f to a,

FIG.4. Electron evaporation rate versus amount of adsorbed cesium; 0 = 1.0 for 3.7 X l O I 4 Cs atoms/cm.Z.

and 8 increases. I n the regions a - b, c - d, and d - el the changes in i, and 8 are reversible with changes in T . At Tcl and Tcz irreversible changes in i, and 8 take place. I n the regions a-f and c-d, there is only one value of 8 for any T. For any T between Tcl and TCzthere are two steadystate values of 8. The significance of this will be clearer after Figure 7 is discussed. I n this and the preceding paragraph, the values of 8 were determined by the procedure given in connection with Equation (2). 2. Experimental Results

This section will give first the direct results of the experimental methods. These are plots of the rate of emission of electrons per square centimeter of surface, E,,as a function of 8 for a series of T , and similar

148

JOSEPH A . BECKER

plots for atoms, E,, and for positive ions E,. Then will be given formulas for converting these rates of emission to work functions pe,pa, and pp as functions of 8. The work function cp is the work th a t must he done t o remove a n electron, atom, or ion respectively from the surface t o an infinite distance or to such a distance th at it will escape from the surface. Figure 4 shows a family of isothermals of E , vs. 8, Figure 5 shows a similar family for En vs. 8, and Figure 6 shows a few isothermals for

FIG.5. Cesium atom evaporation rate versus amount of adsorbed cesium.

E , vs. 0. T o draw such complete families of curves it was necessary t o make rather large interpolations and extrapolations of the original data. However any errors introduced thereby will not alter the conclusions t o be derived from the curves. I n Figure 7 we have assembled all three rates in a single plot for two temperatures, namely the two critical temperatures shown in Figure 3. The dashed curves are for En E,. Any horizontal line on this plot can represent a particular value of arrival rate, A . We have drawn one such atoms per square centimeter per second. Note line for A = 10.0 x that A = En E , at three values of 0, namely, B1 = 0.005, O2 = 0.075, and 63 = 0.27. This means that there exist three values of B at which a steady state is possible. The first and last of these are stable. The second one is unstable; if T should exceed 820"K., En E, for 8 2 will exceed A

+

+

+

ADSORPTION ON METAL S U R F A C E S

149

FIG.7. Evaporation rates for ions, atoms, and electrons versus amount of adsorbed cesium.

150

JOSEPH A. BECKER

+

and hence 0 will decrease; as a rcsult E, E , increases and e will decrease still more; this continues until 0 reaches 0.005. Conversely if a t 02, T should temporarily drop below 820"K., e will increase until it reaches 0.27. For the 820°K. curve if A is reduced, O2 increases and O3 decreases. For A = 2.0 X 10l2,O2 and e3 merge; for A less than this only one steadystate value, namely, el, is possible. Similarly, a t 820°K. if A exceeds 26 X 10l2,only B3 is possible. Since O1 and O3 are stable values, it is possible to have the surface of the tungsten wire in two different states a t the same temperature. T h e central region will be covered to el and will emit predominantly ions; the two end regions will be covered to O3 and will emit predominantly atoms. Where two such states meet on the tungsten surface, there must exist a boundary region in which O has all values between B1 and e3. Special experiments show that this boundary region extends over a distance of about 1 mm. At A = 9 X 10l2 and T = 820°K. the two boundary regions remain fixed in position. If the temperature is raised 1 or 2 degrees, the boundaries move away from the center; if the temperature is lowered slightly, the boundaries move toward the center of the wire. In this way one boundary region can be made to move slowly toward the spare between an end collector and the central collector. The ion current passing through this space can be collected and measured on a n auxiliary collector. At first this slit current has a normal value, which is to be expected from i,, and the slit opening. When the boundary reaches the slit, the slit current increases to a value which is decidedly larger than normal, then reaches a maximum, and declines to a value which is a small fraction of its normal value. Now the ion current to the central collector decreases linearly with time. The whole sequence can then be reversed by raising the temperature a few degrees above 820°K. However, if the two boundaries are allowed to merge, the whole surface is covered to 03. To generate the state el it is then necessary t o raise the temperature about 25" or more. Then there is a rather abrupt cleaning off of the surface, which registers as a ballistic throw of the galvanometer. All of these effects and others substantiate the shapes of the E, and E , curves in Figure 7 . These experiments will be interpreted in the next section. Figures 4 t o 7 are the direct result of experiments. These results can be generalized by deducing the work functions for the three evaporation processes. For electron emission the relation between current density and ( p e is given by the well-established Richard-Dushman equation

i, in which i, = a./cm.2; T

=

=

120T2e-'p.e/kT

OK.;

qe is

in volts; e

(5a) =

electron charge in

ADSORPTION ON METAL SURFACES

151

coulombs; k = Boltzmann's constant in joules per degree. Converting i, t o E, or electron emission per square centimeter per second and using the base 10, we obtain E,

=

8.0

x

1020T210-a5050/T

(5b)

The analogous equations for atoms and for ions are

in which N , and N , are the number of adatoms and adions, respectively,

FIG.8. Concentration of cesium adions and adatoms versus total amount of adsorbed cesium.

per square centimeter; 1013Tn is the number of collisions a surface particle makes per second with its neighbors; the exponential factors give the probability that a collision will give the particle enough energy to escape. pe can be calculated from the E, curve in Figure 7 and Equation (5b). T o calculate pa and pp we must know in addition how N , and N , vary with 0. This relationship, given in Figure 8, will be established in the next section. From these data we can then plot Figure 9, which shows how cp,, pa, and p, vary with 0. Since the surface from which these data were obtained consisted of polycrystalline tungsten the values of cp are some kind of average for all crystallographic planes which were present in the

152

JOSEPH A. BECKER

wire. This is indicated by the symbol ip. A still greater insight into the nature of the adsorption process could be obtained if all the experiments were repeated with a single crystal ribbon of tungsten in which the surface was predominantly a single plane. I n Figure 9 note that as 0 increases, pedecreases, pp increases, and pais nearly constant or decreases slightly.

FIG 9. Variation of the work function for electrons, ions, and atoms versus amount of adsorbed cesium.

3. Interpretation of Results

I n this section certain conclusions will be drawn regarding the nature of the adsorbed cesium particles and the effects that these particles have on the electron emission and on other adsorbed particles. From the fact that both atoms and ions of cesium evaporate from the surface a t the same time, we conclude that both atoms and ions exist on the surface. This conclusion is based on the probable assumption that as the cesium atom approaches the clean surface, the forces which tend t o remove the valence electron from the cesium atom become increasingly larger the nearer the atom is t o the surface. Th at these forces are comparable with the forces that tend t o keep the electron associated with its nucleus follows from the fact that pe for clean tungsten is about 4.6 volts and the ionization potential for cesium is 3.9 volts. Th at both adatoms and adions

153

ADSORPTION ON METAL SURFACES

exist also follows because on this basis a simple explanation can be obtained for the way in which (a, and ( p p change with 9. First let us consider how the presence of adions will affect the electron work function. T o do this quantitatively let us consider a 100 and a 110 plane of tungsten. Figure 10 shows a top view and a section view of the location and sizes of the cesium ion and the tungsten atoms. The Csf is shown in the position in which it contacts the largest number of tungsten S

S

TOP VIEW

s

S I00 PLANE

+& - (

___

110 PLANE

&-f2

f-

--

SECTION

S-S

FIG.10. Top and section view of a cesium ion adsorbed on a 100 plane and on a 110 plane of a tungsten crystal.

atoms. This is probably the lowest energy state for the system. Let us define a n electronic surface plane as the mathematical plane which is tangent t o the radii of the surface tungsten atoms. This is the plane which is the image plane in electrostatic theory since any excess electrons placed on the metal will tend to get to this plane. A theorem in electrostatics states that the electrical fields established outside the metal by the positive ion are the same as would be produced b y a dipole consisting of a and - charge separated by 21, where 1 is the distance between the nucleus and the image plane. This theorem makes it relatively easy to calculate the field due t o an ion or due t o an array of ions and also t o calculate the integral of the force of an electron as it tries to escape. Another theorem, based on the first one, states th a t the decrease in work of a n escaping electron due to an array of ions is 300 X 4aN,el. The factor 300 converts electrostatic units to volts. Hence

+

Aqe = -1200a,el

= -1.8 X 10-6Npl.

(8)

154

JOSEPH A. BECKER

The value of 1 depends on the sizes of the W atom and Cs ion and on the plane. The values of I for the 100 and 110 planes are given in Figure 10. A(a, can be obtained from Figure 9; for 0 = 0.10 it is about 0.55 volt. For small 0 values most of the emission comes from the region surrounding the 100 plane; for this region 1 is smaller than for the 100 plane. Hence we will take 1 = 0.78 X 10-8. From Equation 8 it follows th a t Np

=

1.8 X

0.55 - 3.9 X 0.78 X10-8 -

x

1013

For e = 0.10, the number of Cs particles per square centimeter is 3.7 X 1013. The agreement is remarkably good considering the uncertainties in Figure 9 or in choosing the plane from which to deduce 1. From this we conclude that for small values of 0 nearly every adsorbed Cs particle exists as an adion. By choosing points for various values of 0 on a (a, vs. 0 curve and substituting values of Ape and 1 in Equation 8 we have computed N , as a function of 0. As 0 increases, other planes than the 100 plane become most important or most electron emitting; near 0 = 1.0 the 110 plane contributes most of the electron current. Hence we have used values of I which increased from 0.73 to 1.04 X lo-* as 0 increased from 0 to 0.7. These values of N , are plotted in Figure 8. Since N , N , = 3.7 X 10148, we have computed and plotted N , vs. 0 in Figure 8, in which we have indicated our estimate of the uncertainties by vertical lines. The values of N , and N , in Figure 8 were used to compute the values of ( a p and (aa in Figure 9. While qe decreases with 0, p, increases. This result follows from the fields produced by the adions. These fields are in a direction to help negative electrons to escape; hence they hinder the escape of positive ions. But the rate of increase in g p is only about half the rate of decrease in pel because the positive ions start their escape at a distance 1 above the electronic plane. Hence the integral of the field from this distance t o infinity is smaller than that for the escape of an electron. T o compute the ratio of one half requires a much more involved calculation. As far as these can be made they are not in discord with the experimental facts. The shape of the E, vs. 0 curve in Figure 7 is at first sight surprising since this curve means that the more cesium is adsorbed on the surface the fewer ions evaporate; however, this is just what is demanded by Equation 7 if N , and ‘p, vary as shown in Figures 8 and 9. From the value of e a t which E , reaches a maximum, namely, 0.025, one can conclude tha t the force on one ion due to a neighboring ion is quite appreciable a t distances of about 20 x 10-8, or about 8 tungsten atom diameters. These are long-range forces in comparison with forces between atoms

+

ADSORPTION ON METAL SURFACES

155

which have appreciable effects only a t separations less than 1.5 atom diameters and die off a t about the inverse sixth power of the distance. Another interesting feature in Figure 9 is that for e near 0.07, pa is nearly equal t o pp. Nevertheless nearly all the cesium evaporates as ions. On the other hand, a t 0 = 0.2 about fifty times as many atoms evaporate as do ions even though on the surface there are about eight times as many ions as atoms. This means that a t this 0 an atom has a probability of escape 400 times as large as that for an ion. At larger e, the discrepancy in escape probability is even greater. This should be a warning not to conclude that one species of adsorbate exists predominately on th e surface because that species is practically the only one th a t evaporates.

FIG.11. Potential energy versus distance for a cesium atom (upper curve) and for an ion (lower curve) near a tungsten surface.

I n Figure 9 note also that pa is nearly constant with 0. This means that the fields produced by neighboring ions is not strong enough t o polarize a n adatom appreciably and thus to change its escape probability. T he value of pais 2.4 volts, which is three times as large as 0.8 volt, the heat of evaporation of cesium atoms from bulk cesium. This shows that the Cs-W bond is three times as strong as the Cs-Cs bond. One of the most instructive graphs for an adsorption system is its potential-energy diagram. Figure 11 shows how the potential energy of a system consisting of a cesium atom and a clean tungsten surface changes as the atom approaches the surface. The energy of the system is taken as zero when the atom is a t a large distance from the surface. Since we shall deal with atoms and ions close to the surface, it is necessary to define the term surface. We shall use the electronic surface previously defined and take into account that an atom or a n ion may partly extend below this surface, a s shown in Figure 10. In Figure 11 as the atom approaches the surface it is attracted by it and hence the potential energy decreases.

156

JOSEPH A. BECKER

When it reaches its distance of “closest approach” to the tungsten atoms, strong repulsive forces are exerted between the electrons in the cesium atom and the electrons in the tungsten atoms and the energy of the system rises extremely rapidly. According t o Lennard-Jones (6), the attractive energy varies as the inverse sixth power* of the distance while the repulsive energy varies approximately as the inverse twelfth power.oThe radius of a cesium atom is about 2.7A., and it can sink about 0.5A. below the surface. Hence its nucleus will be 2 . 2 k above the surface. According to Figure 9, (a, = 2.4 volts. This locates the minimum in the atom curve in Figure 11. Suppose we again start with the atom a t infinity and ionize it. The energy of the system will then be +3.9 volts. Now let the electron drop into the tungsten. This reduces the potential energy b y qe, or 4.65 volts. The system now consists of a Cs+ ion a t infinity with a n extra electron in the metal; the system energy is -0.75 volt. Now allow the ion to approach the surface. The system energy will decrease according t o the image Iaw, as the inverse firs: power of the distance. When the ion reaches a distance of about l.OA., it will encounter very strong repulsive forces which will prevent it from coming any closer. According to Figure 9, ( a p is 2.2 volts, and so the energy of the system a t the minimum point will be -2.95 volts. Since a system will try to get t o its lowest energy state, i t is easy to see from this diagram that if a cesium atom were adsorbed as an atom there would be a great tendency for it to be converted t o an adion since this process would lower the system energy by0.55 volt. Conversely if the potential energy of the system is raised by forcing the ion farther from the surface the syttem energy would be the same a s for an atom at a distance of about 2.2A. If the ion receives an electron from the metal while it is a t this distance it can temporarily exist as a n adatom. Define a surface ionization potential, I,, as the energy required t o convert an adatom to an adion. From Boltzmann’s equation it follows that N , / N a = E-l.e/kT = 10-1~6060/T (9) For small 8, I ,

=

-0.55 volt, and for T

=

820°K.

This is why nearly every cesium particle on clean tungsten exists as an adion.

* Professor P. Debye has pointed out to us t h at for a mathematically smooth surface this should be the inverse third power. Recently M. Drechsler (6a) has calculated the force law quantitatively taking into account the atomicity of the surface. From this it follows that the inverse sixth power is more nearly correct. For our purpose the exact law is unimportant.

ADSORPTION ON METAL SURFACES

157

From the experimentally determined values of N , and N , shown in Figure 8, we can calculate N J N , for any 0. By putting in these values into Equation 9 for T = 820°K. we can calculate values for I , as a function of 8. These values are plotted in Figure 12 and are indicated by triangles,

FIG.12. The surface ionization potcnt.ia1 in electron volts versus amount adsorbed for Cs on W.

An independent set of values of I , vs. e can be calculated from the energy balance equation, pa =

I,

+ + ~p

p e

-I

(10)

This follows in an obvious manner from the potential energy diagram in Figure 11; hence I , can be computed for any e from the values of Val pP,and pegiven in Figure 9 and the known value of the ionization potential of cesium, namely I = 3.9 volts. These values are shown in Figure 12 by small circles. The fact that these two sets of values agree with one another shows that the energy values form a self-consistent set." If we now assume th at pe, pa, and pp do not vary appreciably with temperature, we can calculate E , or E , or E , vs. 0 for a range of temperatures above and below 820°K. This is the basis on which the curves in

* This self-consistency was obtained only after carefully taking into acount t h a t the experimental data were taken on a polycrystalline surface for which small crystals of tungsten which contribute most of the electron current are not the same crystals which contribute most of the ion current. This is the reason why the curves in this article differ somewhat from previously published curves.

158

JOSEPH A. BECKER

Figures 4, 5 , and 6 were drawn. The results were checked against the experimental data where they were available. To test whether the work functions are indeed independent of temperature would require th a t the experiments be repeated using single crystal tungsten ribbons which are so treated that the surface exposes only a single plane. We urge more scientists t o undertake this problem. The rewards would seem t o be attractive. It is of obvious interest t o inquire what factors determine the ratio of ions t o atoms which evaporate from the surface. This can be answered easily from Equations (6), (7), (9), and (10).

E,/E,

= (N,/N,) l O - ( ~ - d 5 0 5 0 / T = ~ O - ( ~ ~ - ~ . + ~ , ) ~ O S= O /~TO ( ~ ~ - - I ) S O ~ O / T

(11)

This equation states that the ratio of the rates of ion evaporation to atom evaporation is determined by the temperature and by the difference between the electron work function and the ionization potential; it depends only indirectly on ‘ p p , qa, I,, N,, and N,. This conclusion agrees with thermodynamical reasoning. Before proceeding with the next topic, we should like to emphasize again the significance of this work on the problem of understanding catalysis. The conversion of a cesium atom to a cesium ion is the simplest illustration of a “catalytic” reaction since it involves the transfer of an electron. To convert about 10% of the cesium atoms in free space into cesium ions would require a temperature of 20,OOO”K. This can be calculated from the equations ion concentration in gas phase atom concentration in gas phase

10--3,9X5050/T

=

o.l

(12)

If the cesium is enclosed by walls of tungsten a t about 1000°K. or less, however, it would be possible to convert nearly all the cesium in the space into ions. On the surface of the tungsten an almost complete conversion of atoms t o ions takes place a t as low a temperature as 300°K. and probably even a t much lower temperatures. This simple system permits us t o get a rather detailed insight into the forces and energies involved in the reaction and into the importance of the arrangement and number of nearest neighbors for the tungsten atoms. This makes it easier t o understand under what conditions “active centers” should be important. The reason why minute amounts of foreign materials which are strongly adsorbed act, as powerful poisons is an obvious consequence of the concepts here proposed. Finally this analysis brings out the importance of properly choosing the relations between pressure and temperature in a catalytic reaction. These determine the steady-state surface concentra-

ADSORPTION ON METAL SURFACES

159

tion, which in turn has a tremendous influence on what reactions take place on the surface and which reaction products will easily evaporate from the surface.

IV. THE ADSORPTION OF NITROGEN ON TUNGSTEN AS DEDUCED FROM IONGAUGEAND FLASH FILAMENT TECHNIQUES 1. Introduction

The system nitrogen on tungsten, differs from cesium on tungsten in three respects: ( I ) the nitrogen strikes the surface as a molecule instead of as an atom; ( 2 ) nitrogen accepts electrons and cesium donates electrons; and ( 3 ) the size of the nitrogen which is adsorbed as a n atom is comparable with the size of the tungsten atom but the cesium atom is twice as large. Previous work on this system has shown that near room temperature nitrogen is chemisorbed with a heat of adsorption of about 4 ev. (2). From this it follows th at the incoming molecules, which remain adsorbed or stick, do so as adatoms. The work to be described will show that only about one third of the molecules which strike the surface stick to it and that two thirds reevaporate as molecules in a small fraction of a second. The sticking probability, s, is found to depend on the temperature of the tungsten and very likely varies appreciably from one crystallographic plane t o the next. A startling result is th a t s decreases rapidly and drastically a t about one “layer” when about one quarter of the tungsten sites are occupied by nitrogen; a t two layers, s is only about 10-4. The effect of the second difference, mentioned above, is that nitrogen acts as a negative dipole and increases the electron work function; also, as the nitrogen atom accepts or shares more electrons with the tungsten, its size increases but even so its effective diameter is only slightly larger than that of a tungsten atom. To appreciate the significance of this smaller size, the reader is asked to look a t the top view of the 100 plane for tungsten as shown in Figure 10 and t o imagine this plane extended indefinitely in the x and y directions. Let one nitrogen atom be placed in any site in which it contacts four tungsten atoms; let us say it makes first valence bonds with these four atoms. The four nearest sites in the x and y directions are then no longer equivalent t o the first site since a nitrogen atom in these sites would make second valence bonds with two of the four tungsten atoms. However, if all alternate sites in both the x and y direction are filled, then every nitrogen atom will make first valence bonds with four tungsten atoms and all tungsten atoms will be equivalent. Let us define this situation as “one layer” and say that e = 1.0. There will then be one nitrogen atom for

160

JOSEPH A. BECKER

four tungsten atoms. On other planes this ratio will be different. Thus on the 111 plane, when 8 = 1.0, there will be one nitrogen atom for three tungsten atoms. At every center of an array of four nitrogen adatoms there will still be room for another nitrogen adatom, but these can contact only tungsten atoms each of which has already shared electrons with one adatom. Hence a second layer can be formed but its bond strength will be less than that of the first layer. When the second layer is complete there will be one nitrogen atom for every two tungsten atoms and e = 2.0. After this an incoming nitrogen molecule can no longer find a place where both of its atoms can contact tungsten atoms, and hence it must be satisfied to rest on top of the first two layers of adatoms. The sticking probability and hence the rate of adsorption may again show rather drastic changes at the end of the second layer. I n this section will be described the experimental procedures which measure the rate of adsorption and the sticking probability; the experimental results will be given and will be interpreted in terms of a n energylevel diagram and an activation energy; and finally the bearing of these results on catalysis will be discussed. 2. Experimental Procedure

Since the procedure and results have been published recently (7), their descriptioh here will be rather brief, yet complete enough to clarify the subsequent interpretation and significance. The system, shown in Figure 13, consists of well-evacuated glass containers. At one place nitrogen can be introduced into the system a t a controlled rate through a metal valve. The main bulb contained the tungsten ribbon on which the nitrogen was adsorbed and desorbed. This ribbon was about 15 cm. long, about 1 mm. wide, and had a surface area of 3.2 cm.2. It was thoroughly outgassed by heating a t 2400°K. for hours and a t 2600°K. for shorter periods. This heat-treatment was necessary in order t o “fix” the crystal structure so that it did not change during subsequent experiments. Uriless this is done the experimental results are not reproducible. About one half the total surface area consisted of about ten single crystals; the remainder consisted of many more smaller crystals. This main bulb was connected to a second bulb containing a modern ion gauge which measures or records gas pressures in a fraction of a second. It can measure pressures from about 10-lo to mm. The system was connected to a liquid air trap and mercury diffusion pump through a movable glass plate with a small hole. The effective area of this hole was 0.010 cm.21and the pump speed was 0.12 liter/sec. This pump speed could be increased more than tenfold by removing the glass plate.

161

ADSORPTION ON METAL SURFACES

Pumping through such a small hole during a n experiment is highly desirable because it affords a definite pumping area and a definite volume of the system; both of these must be known and remain constant for the calculation of correct numerical results. The system was baked; all metal parts were thoroughly outgassed; nitrogen was passed through the system and then pumped out. This whole process may be repeated several times in order to reduce the possibility that any gas other than nitrogen TO PUMPS

J CT

PC FF PSC 61 ESS

PRESSUREGUAGE FLASH-FILAMENT P U M P SPEED CONTROL COLD.TRAP COLD-TRAP ELMTRO-STATIC SHIELD

FIG.13. Schematic diagram of experimental system used to study the adsorption of nitrogen on tungsten.

can strike the tungsten ribbon. The volume V of the system was 2.3 liters. A typical experiment is performed as follows: The ion gauge is turned on and kept in operation continuously, the pressures being automatically recorded. The leak valve is opened and adjusted until the pressure p becomes steady a t some value p , , such as 5.0 X 10-7 mm. The tungsten ribbon is cleaned off by heating it to 2300°K. This suddenly releases the adsorbed gas into the volumc V and raises the pressure above 10-6 mm. This pressure is reduced by the pumps until it again reaches the steady-state pressure po. This takes about 2 min. The ribbon is then cooled to room temperature for about 20 see. and heated for 20 see. The heating and cooling cycles are repeated until the system is in a reproducible state and the ribbon is hot. The main experiment is now ready to begin. At time t = 0, the ribbon is cooled. The pressure rapidly drops from p , to a much lower value and remains steady for perhaps 100 sec. (see Figure 14 regions B to C and C to D). Then the pressure rises, slowly at first, then more rapidly, and then more slowly (see regions D to E and E to F). After about 10 or 20 min. the pressure asymptotically rises to p , . A t any time t during this experiment the ribbon can be reheated suddenly or “flashed” to its high temperature. The pressure suddenly rises to a maximum value p,,,, which is noted or recorded. The difference (pm,, - p ) is called A p . The amount of adsorbed gas can be calculated from A p .

162

JOSEPH A. BECKER

The reason for this shape of curve is as follows: From A to B the ribbon is hot and is neither adsorbing nor desorbing gas. The pressure becomes steady a t p , when as many molecules leave the system through the hole as enter it through the leak valve. At B when the ribbon is cooled, additional molecules are removed from the volume by being adsorbed on the ribbon; i.e., there is a n additional pump speed. This rapidly reduces the pressure in the volume V to a new quasisteady state determined by the leak rate into the V and the sum of the two pumping speeds. At D the tungsten surface is covered with so much nitrogen that its pumping action decreases. We shall soon see that at room temperature

FIG.14. The variation of the nitrogen pressure after the tungsten ribbon is cooled.

this happens when 0 A 1.0. The ribbon pumping speed decreases and hence the pressure rises. After quite a long time the ribbon pumping ceases completely and the pressure becomes constant at p,. Before we can calculate the sticking probability s and the fraction of surface covered 19 we must derive quantitative expressions for the processes just described. Let L be the number of nitrogen molecules entering the system per second through the “leak.” The number of molecules leaving the system per second through the pump = 3.8 X 1020pa,.* The number of molecules removed from the volume per second by the filament = 3.8 X 1020pap.Let c1 = the number of molecules per second accumulating in the volume V . Then c1 = 3.2 X 10lgVdp/dt. Except in the

* Definitions for the symbols appearing in this and the next few paragraphs follow: 3.8 X 1020is the number of nitrogen molecules striking 1 cm.*/sec. when the pressure is 1 mm. and T = 300°K. It depends on the molecular weight of the gas and on T. V = volume of system in liters: V = 2.3 t = time in see. p = pressure in mm. of mercury A p = sudden increase in p when ribbon is flashed a, = area of “pump” or hole through which the gases are pumped out of the volume

ADSORPTION ON METAL SURFACES

163

region BC, d p l d t has such low values th at c1 is only a small “correction” term. Let c, be a “correction” term t o take into account the molecules per second which may be taken on or given off by the glass or other cold surfaces. Values of cg have been obtained by subsidiary experiments and found t o be less than 30% of the major terms. I n order t o simplify the presentation of the major factors which determine the course of the experiment, we shall omit the correction terms c1 and c, even though we have evaluated them or allowed for them in our computations. At any time t in the experiment, the rate a t which molecules enter the volume less the rate a t which they leave the volume must equal the rate at which molecules accumulate in the volume. Hence

+ pays) + c,

L - 3.8 X 1OZ0(pa,

=

c1 A 0

(13)

When s is small or zero, either because the filament is hot or because it is saturated, L = 3.8 X 1OZ0p,a, (14) in which p , is the value of p when s puted from

=

0, or t

=

0. Hence s can be com(15)

Let N , = number of atoms per square centimeter adsorbed on the tungsten surface a t any time t. Then

dA= 2 X 3.8 X dt

a

1020ps= 7.6 X 1020 2 (PO- p )

af

(16)

and

N,

=

7.6 X lo2’%

(17)

This equation states that the number of atoms adsorbed per square centimeter of tungsten is proportional to the area between the p , and p V . It is usually 0.010 cm.2. If this hole is not in a plane, a correction must be made for the length of the hole. at = area in cm.2 of the filament or ribbon on which adsorption is being observed. s = sticking probability = number of molecules adsorbed divided by number of molecules striking the surface. 3.2 X 1019 = number of molecules in 1 liter when p = 1 mm. and 2’ = 300°K. It is deduced from Avagodro’s number. el = correction term for molecules accumulating in 1.’ eB = correction term for effects due to glass walls. N o = adatoms/cm.2. N1 = value of N , when 8 = 1.0; N 1 = 2.5 X 1014.

164

JOSEPH A. BECKER

curves from 0 to t. It does not include the additional amount adsorbed in the region BC due to the sudden decrease in p. This amount can be calculated easily since it is equal to 3.2 X 1019V(p,- P ~ , ~I n) .this way we can calculate and plot N , vs. t. Such a curve shows a small rapid rise in the region BC, a linear rise in CD, and then a gradually decreasing rate of rise in DEF. Such a curve can be calculated in an entirely different way by flashing the filament a t 2300°K. a t a series of values of t, observing Ap, and calculating N , from N , = 6.4 X 10’9VAp/~j (18) By this procedure we obtain a curve which shows all the features of the first curve and differs from it by values which are less than 20% of N,. Whatever differences exist can be accounted for by considering values of c1 and c,. This independent check gives us confidence that our calculations of s and N , are reliable t o within about 10%. If we wish t o express the amounts adsorbed in terms of 8, the fraction of the surface covered, we can do so b y defining 8 by

e

=

N,/N~

(19)

where N 1 is the number of atoms adsorbed per square centimeter for one layer on an average plane. For N1 we have tentatively chosen 2.5 X 1014. This corresponds to one adatom for four tungsten atoms on a 100 plane. 3. Experimental Results

The primary results are pressure-vs.-time curves like the one shown in Figure 14. We have recorded hundreds of such curves. I n one family of curves the “cold” temperature a t which adsorption takes place is kept constant while p, is changed by changing the leak rate. Then this whole family of curves is repeated a t other “cold” temperatures such a s 600”, goo”, or 1100°K. One should expect that as long as the tungsten ribbon is not drastically heat-treated such curves are quite reproducible; however, if this ribbon is drastically heated and as a result the crystal structure changes, the p vs. t curves should differ appreciably. The most easily recognizable changes are those in the ratio po/p,,, and in the time a t which p starts t o increase. Such changes cannot be reversed by subsequent heating a t 2400°K. Another treatment which appreciably alters the results is heating the ribbon in nitrogen a t pressures of mm. or higher and temperatures higher than about 1200°K. for times longer than about 10 min. Apparently this treatment changes the arrangement of the tungsten atoms on the surface sufficiently so that the adsorption properties are changed.

These changes can be wiped out by subsequent heating a t 2400°K. at pressures below lo-* mm. for 1 to 30 min. This is another instance in which the presence of an adsorbate changes the surface of an adsorbent. A second experimental result can be shown in plots of A p vs. “time cold.” As before, faiuilies of such curves can be obtained for different leak rates and for different [(cold” temperatures. T o obtain one such curve requires a much longer time than that required to record a p vs. t curve. Since the chief value of such curves is to check the deductions from the p vs. t curves and to evaluate the magnitude of the correction terms, we shall not pursue them further than to point out one of their advantages. 0.6

05 v)

i

3- 0.4 rn

m $0.3 Q

a

5I!0.2 v)

* ‘ * O 0o

425

a50

0.75

LOO

1.25

1.50

1.75

Q

FIG.15. The variation of the sticking probability of N t on W with t h e amount of nitrogen adsorbed.

For long adsorption times, the p vs. t curves and their analysis become uncertain, because p is nearly equal to p , and hence the computation of s and of 0 is subject to large errors. With the flash-off method the errors are smaller. A derived experimental result consists in plots of s vs. e such as the two shown in Figure 15 for 300” and 600°K. At 300°K. the sticking probability is 0.55 and remains constant with the amount adsorbed until 0 = 1.0; between 0 = 1 and 0 = 2 it decreases rapidly; and a t 0 = 2.0 it is about 4 X Between 0 = 1.25 and 2.0, log s decreases linearly with 0. A t higher tungsten temperatures, s is again constant with 0 but has a smaller value and begins to decrease at, a smaller value of 0. We have obtained curves like these for a number of tungsten ribbons. They all have the same shapes and all vary with T in the same may; however, the values of s differ from ribbon to ribbon, probably because of a difference in crystal size and differences in the types of crystal planes

166

JOSEPH A. BECKEIE

exposed by the surfwe. Even for the same ribbon, the s trs. 0 curves ai*(* changed if the ribbon is exposed to pressures of nitrogen of about nim. a t temperatures of about 1200°K. or higher. We believe this t o mean that under these colditions the exposed plane of a single crystal is rearranged into small areas of more stable planes or farets. These facets are small compared with the size of a crystal but are large 011 an atomic scale. Such facets would be expected t o have different adsorption properties from those of the plane which the single crystal forms as a result of high-temperature treatment in a vacuum. Experiments with the field emissioii mic’roscope show direct visual confirmation of this proposed explanation. Another adsorption characteristic which can be deduced from the primary experiments described above is the evaporation rate E . This is the number of molecules of nitrogen which leave the surface per square (aeiitiineter per second and which previously mere chemisorbed as atoms. That nitrogen leaves the surface as molecules rather than as atoms follows from the experimental fact that the flash-off experiments yield the same values as the pressure-time curves. If the nitrogen came off as atoms, it would stick t o glass walls arid would not reach the ion gauge hull) t o be recorded as a sudden increase in pressure. Only at very high flash-& temperatures is the evaporation rate of atoms comparable with the evaporation rate of molecules. Values of the evaporation rate E can be deduced from the arrival ratc and sticking probability when e reaches its steady-state value at a fixed p and 2’. For this steady state

3.8

x

1OZops= E’

In a particular case in which 1’ = 1100°K. and p = 8 X 10-7 mm. e reached 1.8. For this 1’ and 0, s was found to be 4 X lo-‘. Hence

E

5

3.8 X loPo X 8 X

X 4 X 10-‘ = 1.2 X 10”

inolecules/cm.*/sec., or 1.0 X layers/sec. By keeping 1‘ constant but changing the leak rate to other values and determining s, p , and the steady-state el we can calculate an E 17s. e curve for the test 1’.The uncertainties in the calculated values of B are large, and unless 7‘ is 900°K or higher, only lower limits for E can be ohtained.

.I second method for determining E is t o proceed as follows: Fis the leak rate and thus p , . Allow the ribbon to comr to R steady state a t any value of YC,i.e., Yc,,ld. Detcrmine e by the flash-off method. Call this O.,. or steadystate 8. Again let e incrrase to e,, at 7’ = Now reduce thc leak rate to a very low value and wait until the prrssure in the systcni reaches a new low value of po’. This will hr d(1terniined by thc rate of gas cvolution from t h r glass walls and the pump speed. Now sutldenly raise the rihbon temperature to I’l, a t which the evaporation rate E’ is to he determined. From a preliminary test this 7’1 should he so chosen t h a t p mill rise to ahout loop,’. Rerord p vs. time t . B a t any timc t can then bc calculated froin t h r following equation : ‘/IL.

AIX3ORPTION O N METAL SURFACES &~:rc., =

3.8 X 102"aysy(p- p,,')

167

+ 3.8 X 10zOa,(p - po') + 3.2 X 10L9Vdp/dt

(21) In our experiments a, = 3.2 cm.e, a p = 0.01 cm.2, V = 2.3 liters, a, = 1300 cm.?, and sy = 10P. Since avsg = 0.013 om.? while a p = 0.01 crn.l, the glass pumping cannot be neglected. Hence

+

E in molecules/cm.2/sec. = 2.7 X 1018(p - p,' 8dpldt) E in layers/sec. = 2.2 X 104(p - p,' 8dpldt)

+

(Ha) (216)

Then plot E vs. t . For any t compute e from the equation

e

=

e,. - 1.25

x

1014 [ E d 1

Then plot E vs. 0 for 1' = T,.

e

FIG.16. Tentative values for evaporation rate versus amount adsorbed for nitrogen on tungsten. 0 = 1 for 2.5 X 1014 nitrogen atoms/cm.*. When p gets so low that i t is nearly equal to p:, suddenly raise T t o a higher value and again record p vs. t . Analyze this curve in the same way as above and plot E vs. B for 2' = T?.Repeat this until e approaches 0. In such experiments it is very desirable t h a t the glass walls be kept a t a constant temperature. 7'2

Figure 16 shows sample E vs. e curves for 1150°, l l O O o , and 900°K. Because of uncertainties in the data no great significance should be attached to the exact shapes of the curves or to the exact locations of the curves. only approximate values of the heat of evaporation (Oh can be calculated from these data. For 0 near 1.0 we calculate (Oh 5 ev.; for 8 near 1.5 we calculate (Oh 1 ev. These values permit one to conclude that nitrogen is chemisorbed in both the first and second layers and that t,he heat of adsorption in the second layer is smaller than in the first layer. Another safe conclusion is that for 6 less than 1.0 and T less than 1000°K. the rate of evaporation is less than 1O'O molecules/cm.2/sec. This

is small compared with 3.8 x 10*" p s if p is lo-# or larger. Hence ill analyzing dN/dt from the p vs. t curves, E ran be neglected until B approaches close to its steady-state value.

4 . Intcrpretution of the Experimental Results in Terms of an Energy-Levd Diagram and Activation Energy While the experimental results for the variation of sticking probability and evaporation rate with e and T are interesting as such, their real importance is that they tell us more about the mechanism of adsorption. From the fact that the heats of adsorption are in the range of 5 to 1 ev. for rhemisorption, i t has been concluded that in the initial stages of adsorption, the nitrogen molecule is decomposed and is adsorbed as atoms on the tungsten surface. In the later stages of adsorption, particularly a t higher pressures, the heats of adsorption approach about 0.1 ev., and it is generally held that the nitrogen is then physisorbed and is probably held as molecules. From the fact that at low enough temperatures nitrogen is only physisorbed, Taylor (8) and others concluded that the conversion from physi- to chemisorption, i.e., the conversion from admoles to adatoms involves an activation energy. We shall start with these concepts and show how the sticking probability values permit us t o caalculate activation energies for various values of B and T.These values, together with heats of dissociation and heats of adsorption, permit us t o construct potential-energy diagrams. Such diagrams explain why only molecules evaporate even though the nitrogen is adsorbed as atoms. They also lead t o rather detailed pictures of some of the processes which take place on the surface. The potential-energy diagram is the one proposed by Lennarddones (6). Consider a nitrogen molecule at an infinite distance from a single crystal surface of clean tungsten and let it approach the surface. The potential energy of the system will decrease owing to Van der Waals attractive forces. When the molecule is close to the surface, strong repulsive forces arisr between the electrons in the molecule and the electrons in the surface tungsten atoms; as a result the potential energy rises very rapidly. Curve M in Figure 17 is a semiquantitative sketch of how the potential cnergy varies with distance. Zero distance is taken at, the electronic surface plane previously described in connection with Figure 10. The cxact value and cxact location of the minimum will vary with the crystallographic plane and with the orientation of the molecule with respect t o the plane. Again start with a molecule a t an infinite distance. Dissociate it into two separated atoms. This raises the system energy by q,t, the heat of dissociation, which is given as either 7.4 or 9.8 ev. Now let both atoms

ADSORPTION ON METAL SURFACES

169

approach the surface. The system energy will be decreased owing t o polarization forces and interactions between the electrons in the atom and the tungsten surface. At very small distances strong electron repulsive forces will raise the system energy very rapidly (see curve A , Figure 17). Again the exact value and location of the minimum will depend on the crystallographic plane. Curves A and M will intersect a t some point P. According t o Lennard-Jones these curves are very steep and if the distances of closest approach, or the minimum for A and M , differ by about cm., the intersection will occur about as shown. Let us define a number of energy differences by (ad, (a2a, (Oh, cornl Pal and (aa by the symbols shown in Figure 17. Then (aza represents the work to desorb two adatoms into two gas atoms. (ah represents the energy or heat of adsorption for a molecule which is chemisorbed as two adatoms. (pm represents the energy of adsorption for a molecule which is physisorbed as a n admole. pa represents the activation energy, i.e., the energy that must be given t o an admole before it can be converted into two adatoms. (a, = (aa - (an is the energy which determines the sticking probability s. Note that a t P there is no difference in the energy of the system whether the nitrogen is an admole or two adatoms. At this point adatoms and admoles split up or combine readily. Let us now consider what can happen to a molecule which strikes the surface. As it approaches the surface, it loses potential energy and gains kinetic energy. Beyond the minimum the reverse is true. If it loses small amounts of energy by collision with tungsten atoms, it will come to rest or will oscillate about the minimum. Some of these collisions will convert kinetic energy normal to the surface to kinetic energy parallel t o the surface, which means that the molecule can migrate from site to site on the surface. At 300°K. the average kinetic energy will be about 0.03 volt or (35)kT, which is about the energy which might be required to move from one site t o another. Hence we should expect surface movement of admoles t o occur very often-perhaps l O I 3 times/sec. Approximately one in 1000 collisions will give the molecule an energy of 0.1 volts, which is roughly equal t o the estimated value of (am. If this energy is properly directed the molecule will evaporate. Occasionally a collision will result in a n inward movement and may take the molecule t o point P . Here the molecule can split into two atoms both of which are strongly attracted by the surface, and hence they will soon oscillate about the minimum of curve A . The probability t ha t an atom in this “well” can reach point P again is extremely small a t 300°K. and so the atoms will remain permanently adsorbed at t ha t temperature. Quantitative expressions (9) for Em,the number of admoles which will evaporate per square centimeter per second, and for C,, the number of

170

JOSEPH A. BECKER

Where N , is the number of admoles per square centimeter. Hence

Cm / E m

=

=

10--6060(+'a--pm)/T

10-6060pw'T

(25)

I n a quasisteady state the ratio of C, to the total rate at which molecules strike the surface is C,/(E, C,). This by definition is s; hence s = Cm/(Em Cm), or

+

+

l/s

=

1

+ Em/C,

=

1

+ 10+6060~~'T

or

T

CPS

=5050 log

(; - 1)

s,

Thus for s = (P, = 0, which means that curves A and M intersect a t zero energy in Figure 17. For s = 4i0, and T = 300°K. cp. =

If

s=

0.95 X T/5050

= 0.056 ev. cp, = 1 2 0 9 g ~ 5=~ 0.24

ev.

In this way we have computed pS vs. e for several values of T from the family of s vs. 0 curves. The results are shown in Figure 18. Of course, for each 0 and for each T there exists a slightly different set of A and M curves like those shown in Figure 17. Figure 18 permits us t o deduce how point P shifts as 0 and T are changed. At 300°K. P does not change until e reaches 1.0. This means that the curves do not change relative to one another; probably both remain fixed. This would mean that cph does not change between 0 = 0 and 1.0. Between 0 = 1.0 and 2.0, point P moves upward. I n this range we should expect (Ph to decrease by perhaps one quarter of its initial value, The shape of the lower part of the A curve probably does not change very much. The upward shift of curve A would increase cps by about 1 volt; however, the A curve probably also shifts t o the right, which would decrease (P,. From this we conclude that the atoms in the second layer are slightly farther from the surface than those in the first layer. Similarly, we conclude that as T increases, both the A and the M curves shift to the right but curve M shifts slightly more than curve A. This increases cp, and decreases s. As e increases t o about 0.8 a t 600"K., the surface movement of the atoms from one site t o others is probably appreciable, and so perhaps 20% of the atoms are in sites corresponding t o B = 2 and 60% in sites of e = 1. A molecule in a small region for which 0 = 2 will have a much smaller s, and hence the average

ADSORPTION ON METAL SURFACES

171

FIG.17. Potential energy versus distance for two nitrogen atoms or one molecule near a tungsten surface.

FIG.18 tungsten.

on

172

JOSEPH A. BECKER

s would decrease. Some such mechanism probably causes (pl to increase before e reaches 1.0. Thus one can get some clues as t o what goes on as e and T change, from the shape of the s vs. 6 curves. When more precise data for s are obtained on single crystallographic planes, this picture will be clarified. Better data on the heat of adsorption for single planes would also help. We now turn to the interpretation of the evaporation rate vs. e curves. From the potential-energy diagram (Figure 17) one can calculate E,, the number of atoms th at evaporate per square centimeter per second.

E, cpza

= Na1013T~’”10-6050‘p’./2T

(27)

is divided by 2 since it represents the work to remove two atoms.

E , calculated from Equation 27 At T = 1100°K. and N, = 2.5 X is about 1000 atoms/cm.2/sec. Figure 16 shows that under these conditions about 1O1O molecules evaporate/cm.2/sec. Hence lo7 times as many molecules evaporate as do atoms. From Figure 17 it is apparent that the molecules evaporate by a twostep process. The atoms “evaporate” to the plane which passes through P and is parallel to the surfare. Any atoms which collide with one another can form a molecule. Any molecule thus formed has a high probability of evaporating. Hence to get an expression for the evaporation rate E , we calculate the numbers which pass through 1 cm.2 of this plane in 1 sec. and multiply this by the probability that they will collide with a n atom in this plane. The number of atoms per square centimeter in this plane is equal t o N , (the number of atoms per square centimeter a t the minimum) times the Boltzmann factor 1O-5050~h/z2’,where ph/2 is nearly the energy required to remove an atom from the minimum to P . For the collision cross scction for two atoms we take 16 X 10-l6 cm.2. Hence

Em = Na1013T~10-6050‘p”/ZT X 16 X

1O-16Na10-6050ah/2T

Na2 X 16 X 1 0 - 1 6 1 0 1 3 T ~ ’ ” 1 0 - 5 0 5 0 p h / T = 1.6 X 10-2N,2T4*10-5060~”/T -

or ‘Ph

=

-(1.8

5050

+ 2 log N , + 56 log T - log E )

(28b)

By applying this equation to the data in Figure 16, we calculate th a t at 1150°K. for the first layer is 4.0 ev. For the second layer the value of ph averaged for 1100 and 900°K. is 3.5 ev. If further experiments should confirm the shape of the evaporation curves in the second layer, they

ADSORPTION ON METAL SURFACES

173

would show that as the second layer nears completion, (Ph increases. This would be expected if when e = 2, the nitrogen atoms not only contact tungsten atoms but contact and make bonds with nitrogen atoms in neighboring sites. For oxygen on tungsten, the field emission microscope shows direct visual evidence that this is the case on certain planes. This would mean that oxygen and nitrogen atoms have a diameter of 3.1 X lo-* em. when adsorbed on tungsten in the second layer. Thus from the s vs. e curves and the E vs. 0 curves one can draw the following conclusions for nitrogen on tungsten: ( 1 ) At 300°K. the activation energy for the conversion of an adsorbed nitrogen molecule into two chemisorbed atoms is constant until the first layer is complete; then it increases considerably as the second layer is adsorbed. (2) A t 600” t o 1100°K. the activation energy increases; it starts to increase with 0, the amount adsorbed, even before the first layer is complete (see Figure 18). ( 3 ) The heat of adsorption for atoms is constant and equal t o about 6 ev. for the first layer; in the second layer it is about 5 ev. (4) The heat of chemisorption for an incoming molecule is about 4 ev. for the first layer and about 3.0 ev. for the second layer. ( 5 ) I n the first and second layers, the nitrogen is adsorbed predominantly as atoms, and yet it evaporates as molecules. (6) Molecules which are adsorbed directly on tungsten have a very short lifetime even a t 300°K.; during this time the admole moves from one site t o another; it either evaporates as a molecule or is decomposed into two adatoms. (7’) Above 300°K. both the atom and molecule distance from the surface increase with temperature but the increase for a molecule is greater than that for an atom. (8) When the second layer is complete, the nitrogen atoms on certain planes make bonds with their neighbors as well as with the tungsten. (9) I n this condition the diameter of the nitrogen atom is about 3.1 X lo-* em. (20)At 1100°K. and for one layer, about lo7 molecules evaporate for every atom that evaporates. 5 . The Bearing of These Experiments o n Catalysis

It is apparent that the conclusions just summarized will have a bearing on the concepts currently being used to correlate chemisorption and catalysis. There are two more aspects in which these experiments with the ion gauge and flash filament will influence the interpretation and the theories for previous experiments in adsorption and catalysis. I n the literature it has frequently been reported th a t when gases are adsorbed on metals, the first part is adsorbed “instantaneously” and later parts are adsorbed much more slowly. This slow adsorption has been ascribed t o pores or capillaries or to a n approximate balance between adsorption and evaporation. Only rarely has the analysis mentioned small sticking probabilities as the chief cause. The present work shows

174

JOSEPH A. BECKER

tha t on a clean metal the first layer is adsorbed very rapidly but not instantaneously and that the primary cause of the subsequent slow adsorption is due t o a rapid decrease in the sticking probability in the second layer. Any theory which fails to consider sticking probability as a function of amount adsorbed needs drastic revision. Many theories of adsorption, following Langmuir, have assumed that the rate of adsorption is proportional to (1 - e), i.e., to the fraction of the surface which is bare or not yet covered. Langmuir first “proved” the (1 - 0) “law” by measuring experimentally how the thermionic work function cp changed with time as thorium reached the surface of a tungsten filament a t a constant rate (10). He then assumed that cp decreased linearly with 0 and thus deduced that do/& was proportional to (1 - 0). But this assumption has been shown to be incorrect for such cases as Cs on W, Ba on W, SrO on W, and other systems. Hence it follows th a t the (1 - 0) law is not valid. The experiments described above for Nz on W not only show that df?/dt is not proportional t o (1 - O), but they show by a direct experiment that de/dt for a constant arrival rate is independent of e between e = 0 and 1.0. In many theories, nevertheless, the (1 - e) law has been used a s the foundation on which to build elaborate superstructures, which must necessarily fall if the (1 - 0) law is invalid. Even from a theoretical analysis of what should happen when an incoming molecule strikes a surface which is partially covered, one should become suspicious of the (1 - e) law. Should one not expect that an incoming molecule, which collides with a n adsorbed atom (or molecule), is deflected by this atom and strikes a near-by bare spot where it can be adsorbed? The view that the incoming molecule bounces off the surface if i t collides with a n adsorbed atom is in our opinion far too naive. Probably the reason why the (1 - 0) concept has received such widespread credence is that Langmuir was able to derive his famous adsorption isotherm on the basis of this concept. Since the Langmuir isotherm equation has been experimentally verified in many cases, i t was felt th a t the (1 - 0) concept must be essentially correct. This again is fallacious reasoning, since in the derivation two other assumptions are necessarily made which are not in accord with recent experiments. These are (I) the rate of evaporation is proportional to e and ( 2 ) one can treat the experimental data as if the surface were homogeneous. Because of this situation, it is desirable that someone derive the Langmuir isotherm equation on more realistic assumptions. (See ref. 10a.) For the benefit of those readers who might feel that we are basing farreaching conclusions on the results of one system, namely nitrogen on tungsten, we add that exploratory experiments similar t o those described

ADSORPTION ON METAL SURFACES

175

above and those which will be described later have shown that similar conclusions can be drawn for other systems such as 0 2 on W, H2 on W, Nz on Mo, and N2 on Ta. No experiments which we have tried are in disagreement with the conclusion that at room temperature the initial sticking probability is less than 1.0 and greater than 0.03; it is constant up to concentrations such as one adsorbed atom for four metal atoms; for higher concentrations of adsorbate, the sticking probability drops to much lower values such as or I n most cases of adsorption previously described in the literature, the pressures used in the experiments were so high that the first layer was covered instantaneously and the experiments dealt with amounts adsorbed corresponding t o two, three, or four of our layers, for which the sticking probability is very small and may vary rapidly with the amount adsorbed.

V. THE ADSORPTION OF OXYGEN ON TUNGSTEN AS OBSERVED IN THE FIELD EMISSION MICROSCOPE 1. Introduction

I n a field emission microscope (11) one sees on a fluorescent screen a picture that consists of bright and dark regions produced by electrons which are emitted from a very sharp, needlelike, metallic point. These electrons are pulled out of the metal by very high electrical fields surrounding the point. The intensity of the electron beams emitted from the point increases rapidly with field strength and decreases rapidly as the work functions of a small area on the point increases. Thus the picture on the screen is a highly magnified image of the distribution of work functions on the surface of the point. Since the point is very small, it consists of a single metal crystal whose surface contains all crystallographic planes. Since adsorbed gases and vapors change the work function, such changes can be followed on all crystal planes by observing the changes in the pattern on the fluorescent screen. It is thus apparent that this emission microscope can shed a great deal of light on adsorption phenomena. It is particularly valuable in showing the large variations in adsorption effects on different crystallographic planes and the importance of the special arrangement of the metal atoms and of their valences. Under certain special conditions when molecules are adsorbed, there is brought about an additional local magnification which permits one to see individual atoms in some adsorbed molecules. One can observe, for example, when a molecule changes its axis with respect to the surface, when it combines with other molecules to form larger molecules, when it rotates rapidly about an axis, and when it disappears as a molecule. The

176

JOSEPH A. BECKER

effects of temperature and high fields on these processes can be observed, or recorded on motion picture film. The microscope can also provide evidence to answer such questions as which crystallographic planes have the greatest tendency to form or develop on a spherical surface? how does temperature affect this tendency? how does the adsorption of a gas, such as oxygen, change the crystal planes which are most stable? a t what temperature are the surface metal atoms mobile enough to rearrange themselves? how is this temperature related t o the melting point? a t what temperature do admoles and adatoms become mobile? how do these effects depend on the crystallographic plane of the adsorbent? I n order t o understand and interpret the many details that can be observed, i t is necessary t o examine how the metal atoms can arrange themselves on a spherical surface. The best way to do this is t o make a model of a body-centered cubic crystal, such as tungsten, whose surface is as close to a mathematical sphere as the size of its atoms permits. We have constructed such a model in which marbles represent tungsten atoms. The radius of curvature of the model is 25 atom (or marble) diameters. This is 40 to 100 times smaller than the metal points used in the microscope but does not change any of the essential features which we wish to bring out. Figure 19 is a photograph of this model, showing black, gray, and white marbles. A black marble or atom touches four other marbles; i.e., it has four nearest neighbors with which it makes bonds. A gray marble has five nearest neighbors. A white marble has six or more nearest neighbors. An atom in the interior of the crystal has eight nearest neighbors. It is important to portray this since the fewer bonds a tungsten atom makes with its neighbors, the more bonds it can make with foreign adsorbed atoms; also, the more electrons that are still free to be pulled out by the high field, the greater will be the emission current. A second feature brought out by the model is that certain regions form flat planes of marbles. These are the simplest and most important crystallographic planes, In Figure 19, in the middle of each wall of the model, there is a plane of white marbles. This is the 110 plane, the most densely packed plane; it contains 1.41 x 1015 tungsten atoms/cm.Z and has the largest step between successive layers. I n each corner of the model there is a 100 plane, which has 1.00 X l O I 5 tungsten atoms/cm.2; each such atom contacts four tungsten atoms in the plane underneath it but contacts no atom in its own plane. Midway between two 110 planes there is a 211 plane, containing 0.82 X loL6tungsten atoms/cm.2; it consists of rows of gray marbles separated from other rows by a considerable space. It is apparent that adsorbed atoms should be able to migrate more easily

ADSORPTION ON METAL SURFACES

177

over this plane in the direction of the “furrows” than crosswise thereto. I n the center of Figure 19 is a triangular region of black marbles which constitutes the 11 1 plane; it contains 0.58 X l O I 5 tungsten atoms/cm.2; each atom contacts three neighbors in the plane immediately beneath it and one more neighbor directly beneath that.

FIG.19. A model showing the arrangement of tungsten atoms on a sphericalsurface of a single crystal. The “colors” indicate the number of contacts a surface atom makes with its neighbors.

Note t ha t the edge of the 110 plane, for example, consists of “black” and ((gray” atoms, which can make more and stronger bonds with foreign adsorbed species than can the (‘white” atoms in that plane. This leads one t o expect that such edges should be especially active in adsorption. Note also that the regions surrounding any of these simple planes have the steplike structure of plateaus. As one proceeds from any such plane, the distance between plane edges gets smaller. As one proceeds from any one plane t o another type of plane the structure changes from that similar t o the first one t o that similar to the second one. Hence it seems proper t o refer to the region around a plane a s the 110 region or the 21 1 region, etc. Many of the adsorption characteristics observed experimentally confirm the expectations based on this model of the surface structure of metallic tungsten.

178

JOSEPH A . BECKER

2 . Experimental Procedure (12)

A schematic diagram of the apparatus used in the experiments on oxygen on tungsten is shown in Figure 20. The tungsten point, WP, consists of a 3-mil tungsten wire the end of which has been sharpened into a needlelike point by electrolytic polishing. It is welded onto a thicker tungsten loop, WL, which can be heated by passing current through it. The length of the needle is about 0.05 cm. and the radius of FLUORESCENT SCREEN

TO GAS

WL WP MGV

TUNGSTEN LOOP

= TUNGSTEN POINT = MAGNETICALLY OPERATED

GLASS VALVE

FIG.20. Hclicrnatic of evperinicntal system used to study oxygen on tungsten with a firld eniissiori niirroscopr.

curvature of the point is about 2 x cm. A variable potential of from 3 to 15 kv. is applied to the anode, which eonsists of a film of “aquadag,” or colloidal graphite. The field electrons which are pulled out of the point travel in approximately straight lines from the point t o the fluorescent screen. The secondary electrons from the screen are collected by the anode. The field emission current is measured by a sensitive microammeter. Oxygen gas can be fed into the tube through a valve whose leak rate can be controlled. In some experiments this valve consisted of a heated silver tube. The pressure is measured with a modern ion gauge. The gas is pumped off through a magnetically operated glass valve the

ADSORPTION ON METAL SURFACES

179

pump speed of which can be adjusted to any one of three values. I n some cases the system also included a tungsten ribbon, which was used t o measure the rate a t which oxygen was adsorbed on clean tungsten by the methods described in section IV. The system was pumped and baked and all metal parts were outgassed in the usual manner until the residual gas pressure was as low as mm. or even 2 x 10-lo mm. The system was flushed with oxygen and again evacuated. The tungsten loop and point were heated in a good vacuum for several hours a t 2400°K. until the characteristic "normal clean tungsten " pattern was consistently obtained on the screen whenever the loop was flashed at 2200" to 2400°K. and quickly cooled to room temperature. The pattern on the fluorescent screen is observed visually or is photographed a t intervals. Recently motion pictures have been taken of the screen. It is also possible t o measure the intensity of the light on any small portion of the screen by means of a photomultiplier tube and t o follow this intensity as a function of time. The simplest experiment is performed as follows: adjust the leak rate and pump speed until the oxygen pressure attains some predetermined value and remains constant with time. If the initial stages of adsorption are t o be investigated this pressure should be near mm.; for larger or 10W mm. I n the latter case the amounts adsorbed it should be first stage of adsorption occurs so rapidly that it can no longer be followed conveniently. This type of experiment shows on which planes the oxygen is most readily adsorbed and how the work function changes on the various planes. A second experimental procedure is to allow a large amount of oxygen to be adsorbed a t a pressure of about mm. The pressure is then decreased t o a low value such as mm. With the tungsten at room temperature, the anode voltage is increased until the field emission is about 1 pa. The pattern is then bright enough so th a t it may be photographed in about 56 sec. The tungsten loop is heated to a temperature of 500°K. for a fixed time such as 1 min. The tungsten is then cooled t o room temperature, the anode voltage is increased until the current is 1.0 pa., and another photograph is taken. This procedure is repeated for approximately 100°K. steps up t o 2OOO"K., when the tungsten has become clean and the oxygen is completely desorbed. I n order t o interpret the results it is important that the procedure be systematic and that the temperature intervals be sufficiently small. For T less than about 900°K. the heattreatment can be carried out with the field on and the effects of the treatment can be observed as they occur. Above this temperature the high fields rearrange the oxygen and tungsten and thus introduce complications. A series of photographs taken by this procedure shows startling

180

JOSEPH A. BECKER

changes in the emission from different planes. From the analysis of these changes it can be concluded that oxygen is adsorbed in two distinct layers with heats of adsorptioii respectively of about 5.0 and 2.5 ev./molecule. In a third experiment the amount of adsorbed oxygen was increased still morc hy surrounding the tungsten with gas a t a pressure of 10-2 mm. or higher. The tungsten loop was sometimes heated t o about 600°K. a t this high pressure. In other runs the pressure was reduced by shutting off the leak a i d the whole glass system was baked a t 650°K. for about 1 hr. When the system is cool, the anode voltage is increased until a pattern is seen on the screen. This voltage is always less than that necessary to see a pattern for two layers of oxygen on tungsten and frequently IS less than that for clean tungsten. The pattern is of an entirely different nature from the patterns normally observed : it consists of intensely bright spots or groups of spots, any one of which may suddenly change its orientation or its intensity or change into another grouping. These grouph of spots bear no obvious relatioil t o the underlying tungsten planes. From the behavior of these groups we conc*lude that they reveal individual atoms in iridividual molecules which are adsorbed on the undcrlyiiig layers of strongly chemisort)cd oxygen. A fourth experiment requires that the tungsten loop and point be mm. or higher to temperatures of heated i n oxygen a t pressures of 1200” t o 1500°K. The tungsten is cooled and the patterns are observed. These patterns show that the major planes for clean tungsten change their size and shape and that certain new planes develop which rannot be developed by heating the tungsten in a good vacuum. From this we cboncalude that the distribution of the surface free energy for the various planes is materially altered by the adsorption of oxygen.

Rrwlts and Interpretation (18) a. O x y ! i m Adsorption. In the adsorption experiments a t a constant low prebsurc one can observe how the current density from any plane or region cbhanges with time. Figure 21 shows a series of negatives of photographs taken as oxygen was being adsorbed a t a low pressure of 3 x 10-8 mm. At this pressure it should take about 50 sec. for one layer to be adsorbed on the average plane. This is based on subsidiary experiments with a tungsten ribbon and ioii gauge in the same system. The upper three photographs show that the most rapid deactivation of electron emission occurs in the 21 1 region and in the 111 zone. These regions consist of rows of tungsten atoms which are separated from each other (see Figure 19). At the terminals or a row of atoms, an oxygen atom finds a “deep hollow” where it can rontact four or five tungsten atoms. Apparently this structure favors the rapid conversion of admoles t o adatoms, J. Experimcntal

ADSORPTION ON METAL SURFACES

181

which results in a high sticking probability. The next most rapid deactivation occurs in the 111 region. The center of the 111 plane emits electrons for a much longer time than does the surrounding region. The 100 region, which initially is the most emissive, deactivates quite slowly. Apparently the surface structure of this region is not very favorable for molecular decomposition. While it is not possible to deduce values for the sticking probability s it is quite clear that s varies by appreciable factors for the different planes. As oxygen adsorption proceeds, the applied voltage must be increased to maintain the emission a t 1.0 pa.

FIG. 21. Series of field emission patterns as oxygen IS adsorbed on a tungsten crystal at a pressure of 3.0 X 10W mm. On each photograph the upper numbcrs give the time and temperature of treatment; the lower numbers give the applied voltage, the field emission current, and the exposure time. The upper thrrc pictures show the rapid adsorption of the first layer; the lower three show the slower adsorption of the second layer.

The lower three photographs show that after the first layer of adsorbed oxygen has been completed, the electron emission changes in two respects: ( I ) the deactivation proceeds much more slowly and ( 2 ) the pattern of the 100 and 111 regions now consists of numerous small round spots, while for the first layer the emission from these regions was smooth and continuous. In later stages some of these spots suddenly appear or disappear. If the tungsten is heated slightly, these spots become agitated in some regions but not in others. We tentatively propose that these spots are due t o field emission from individual oxygen adatoms in the serond layer. After a long adsorption time the pattern is almost completely reversed

182

JOSEPH A . BECKER

from that of' clean tungsten. The 110 region and 111 zone, which were least emissive for clean tungsten, are now relatively the most emissive. The 100 region, which was the most active region for clean tungsten, is now quite inactive. If a t this stage the tungsten is heated to about 800°K., the 100 region brightens a great deal while some other regions hardly change at all. In this 100 region the pattern is spotty and the spots are agitated, but the spots in the 111 region are steady. If the tungsten is then cooled, the spots in the 100 region stop their agitation and become steady. Then more oxygen is adsorbed slowly and the activity in this region decreases while that in the 111 region remains constant. This heating and cooling cycle can be repeated with the same results. Apparently at these low pressures the 111 plane and region can be covered only with one layer of oxygen because the sticking probability is extremely small for the second layer; the 100 region, however, can be covered with two layers because its sticking probability is larger. All the rhanges in pattern which have been described and many mow \vhic.h have been observed are sigriiticant, but their full significance is still far from clear. The only c.onc1usions we have drawn are that the sticking probability varies considerably among the different planes: for the sccwnd layer s is much smaller than for the first layer, and some planes can he covered with more layers than others. b. Oxygen Desorption. I n contrast with this situation, the second type of experiment, in which the oxygen is desorbed a t successively increasing temperatures, has led to some important semiquantitative conclusions. The first layer of adsorbed oxygen on tungsten cannot be desorbed until the temperature reaches about 1Ci0OoK., which means th a t the heat of adsorptioii is 4 to 5 ev./molecule. The second layer of adsorbed oxygen (bannot be desorbcd until the temperature rearhes about 80O0K., which means that t hr binding energy is 2 t o 255 ev./molecule. The differences of the binding energy for different planes amount to ahout 20%. The reasoning whivh led to these c*onclusionswas as follows. From thc photographs wr deduce whivh fraction of the total current comes from particular areas, sucbh as the I 1 1 plane, the 100 plane, the 100 region, and the 110 region. We then compute current densities for these areas for each photograph. These current densities are caonverted t o electron work functions cp by means of the well-established Fowler-Nordheim equation. * * Dr. El. Gonier and others prefer to deduce a work function from the slope of a plot of log i / F 2 vs. 1/P. This procedure is correct only if (o is independent of P. For clean metal surfaces the two methods yield nearly the same value of 'p. For adsorbed films the values differ by appreciable amounts. For these films the high applied fields probably modify the electronic distriliution near the adatoms, and t h u s cp will vary with F. IIenw our procc&m for caleulating 'p is t o be preferrd.

ADSORPTION ON METAL SURFACES

183

The constants in this equation are evaluated by using the value of 4.4 volts for the work function of the 111 plane for normal clean tungsten. We thus calculate values of cp for the 111P (i.e., the 111 plane) as a function of the heat-treating temperature Th. The results are shown by the 1112' curve in Figure 22. This figure shows also the results for the other planes or regions. T o interpret these curves, let us temporarily focus our attention on the lllP curve. A fairly obvious interpretation is that a t 600" to 700°K. 7.0

6.8 6.6 6.4

3 6.2 9 6.0

u)

8.

5.8

z 5.6 0

F U

5.4

Z IL 3

5.2

rr S 5.0 0

3 4.8 4.6 4.4 4.2 4.0

400 600 800 1000 1200 1400 1600 1800 2000 2200 T E M P E R A T U R E OF HEAT T R E A T M E N T FOR 1 MINUTE IN DEGREES KELVIF

FIG.22. The electron work functions for various crystallographic planes versus thc temperature of heat trratrnrnt: for oxygen 011 tungsten.

a considerable amount of oxygen is desorbed and as a result cp decreases from 7 t o 6 volts. Retween 700" and 1300°K. very little additional oxygen is desorbed and hence cp remains constant. At 1400'K. more oxygen is desorbed rapidly and hence cp decreases. Above 1600°K. the oxygen is completely desorbed and (a remairis constant a t 4.4 volts. Apparently the oxygen is desorbed in two successive stages or "layers" a t about 700" and a t 1400"K., respectively. Let us call the 700" stage the second layer, and the 1400" stage the first layer. From the figure it then follows th a t on the l O O P and lOOR (i.e., the 100 region) the second layer is rapidly removed at 800°K. while the first layer is not removed rapidly until T reaches 1600°K. For the llOP and 11OR the second layer does not produce a large change in cp, but the first layer does, and this layer is rapidly desorbed at 1400'K.

184

JOSEPH A. BECKER

That the changes in work function are indeed due t o the desorption of oxygen is confirmed by subsidiary flash filament experiments. For this test, the tungsten loop and point are again treated as described and covered with a large amount of oxygen at 300°K. The pressure is then reduced t o a low value. If the temperature is then raised to BOO'K., there is only a slight rise in pressure, showing th at no large amount of gas is desorbed; however, when the temperature is raised from 600" t o lOOO"K., there is a large and rather sudden increase in pressure. This rise in pressure is roughly what one would compute if one layer were suddenly desorbed from the surface area of the loop. When the temperature is raiscd from 1000" to 1300°K., there is a smaller sudden rise in pressure. This might reasonably be due to desorption from the cooler ends of the loop, which now reach 1000°K. When the temperature is raised from 1300"to 1700"K., the pressure again rises suddenly by an amount which is compatible with the desorption of one layer. Still higher temperatures produce no sudden comparable rises in pressure. While these results are not very precise, they do show t hat approximately one layer of gas is desorbed a t each of the temperature regions in which the work functions drop drastically in Figure 22. From the temperature a t which a layer is desorbed in about 1 min. from any plane or region, one can compute the heat of adsorption 'ph for that, layer and that plane. As an example, we choose the second layer adsorbed on the 100 region. One layer, or about 1.2 X l O I 4 molecules/ cm.2, evaporate in 60 sec., or 2 X 10l2 molecule~/(cm.~)(sec.). From Equation 2% i t follows that 7' 27 1.4 - 12.3) = 1'/350 pjt = -- (-1.8 (29) 5050

+ +

This equation is a very convenient one for determining approximate values of the heat involved in a process in which the rate on 1 cm.2 is about 1.2 x 1 0 1 4 elementary acts/min., or 2 x 10l2acts/sec. The divisor 350 is strictly not a constant; however i t varies only slightly with the concentration of adsorbate or with the temperature of the surface. By applying this equation to the data in Figure 22 we calculate that for the second layer of oxygen on the 111 plane of tungsten, the heat of adsorption is 2.0 ev. and for the first layer it is 4.0 ev. For the 100 plane or region, p,, = 2.3 ev. for the second layer and 4.6 ev. for the first layer. For the 110 plane and region, (oh = 4.0 ev. for the first layer. c. Sering Indivadual Atoms in Adsorbed Molecules. We now pass on t o the third experiment, in which the pressure and temperature are such that a third layer can be adsorbed. In this case the field emission pattern consists of intensely bright spots or groups of spots which act as a unit.

ADSORPTION ON METAL S U R F A C E S

185

Figure 23 shows schematically the units or groups of spots which have been seen in the course of observations extending over a total of about 100 hr. in perhaps ten different tubes. Such spots have never been observed unless the pressure during the treatment of the point exceeded about mm. After this treatment, the pressure should be reduced t o 10-*mm. or less. At higher pressures the observations are complicated by positive ion bombardment of the surface. The spots can be observed at any temperature below about 600°K. At higher temperatures they disappear. From this fact and Equation (29) we conclude th a t they are bound t o the surface with an energy of about 2 volts and that hence they are chemisorbed. UNITS AND TRANSITIONS OBSERVED FOR

02 O N 0 - W

FIG.23. Patterns of units and transitions observed in field emission microscopes when the tungsten surface has been treated in special ways. It is believed t h a t each bright spot in a unit is due to an individual atom in simple molecules like 0 2 , 0 4 , or 0 s .

If any one spot or any one unit is observed a t 300"K., it may persist unchanged for minutes or even hours; on the other hand, it may change suddenly and drastically in a variety of ways. The first two rows in Figure 23 depict some of these changes. Thus a doublet may suddenly change into a single spot, it may drastically change in intensity, it may disappear completely, it may suddenly change its orientation, it may change into a quadruplet of four round spots a t the corners of a square, or it may change into a bright toroid with a dark centw. A quadruplet may suddenly change into any of the forms shown in the second row. The bottom row shows some units which have been observed on rare occasions. Any kind of unit may suddenly appear a t a location which previously was dark. Sometimes a unit may transform back and forth between two varieties 10 or 30 times at a rate which may vary from once a second t o ten times a second. T o appreciate fully the definiteness, suddenness, and great variety, one must see these transformations on the screen. They can be recorded by motion pictures. If the temperature is increased t o about 500"K., the rate of trans-

186

JOSEPH A. BECKER

formation increases and the spots become somewhat more fuzzy; however even after an hour the total number of spots remains unchanged. During this time thousands of transformations occur and about a hundred equivalent layers of residual oxygen molecules must strike the surface. From this we conclude that the transformations are caused by interactions between “something” in the third layer and oxygen in the first two layers rather than between “something” and the gas phase. Between 600” and 800°K. the number of units on the screen slowly disappear. This we interpret as evaporation of “something.” It may be that this disappearance is related to the evaporation of the second layer, which occurs a t 700” and 800°K. Out of perhaps 100 bright units, two may be seen t o overlap on the screen. The lower right sketch in Figure 23 shows a case in which a singlet and doublet overlap. I n time one of these units will suddenly disappear. When this happens the other unit remains unchanged i n size, intensity, and orientation. This proves that even though the screen images of two units overlap, the source of the electron beams must be far enough apart so that the disappearance of one has no effect on the other. From this we conclude that the “something” which causes the spots must be magnified t o a greater extent than the tungsten and substrate. Let us summarize some conclusions about the spots and their origin. The current density is very much greater than from surrounding atoms in the first two layers. This could mean only one of two things: either the work function is locally greatly reduced or the local field is greatly increased. We believe the second alternative is the more likely. The spots must be caused by (‘something” in the third layer. It is chemisorbed by the substrate, which means that electrons are transferred or shared with the substrate. Physisorbed molecules do not produce bright spots. The sudden transformations of units must be due to interactions or displacements with the substrate rather than with the gas. To account for these conclusions, we propose the following hypothesis: The spots are due to molecules which rest or on protrude above the first two layers and which share electrons with the atoms in these layers. On this basis we can treat these molecules as if they were small metallic spheres or parts of spheres on a metal surface. We can then draw equipotential lines above the surface and above these spheres and deduce differences in field strength. IVe can also sketch probable paths for electrons from those parts of the spheres where the field strength is greatest. Figure 24 is a schematic diagram of equipotential lines and lines of electron flow for a doublet which is clearly resolved. If, as we have postulated, the atoms share or interchange electrons with the substrate, it should be permissible to treat them as conductors and t o draw equi-

ADSORPTION O N METAL SURFACES

187

potential lines as shown. From these it follows (1) that the highest fields should occur near the outer surface of the atoms in about the positions indicated by the highest concentrations of flow lines and (2) th a t the flow lines strike the screen in such a manner that the electrons appear t o come from a region in the neighborhood of the atom whose diameter is several times the atom diameter. The enhanced resolving power, the enhanced magnification, and the extra high field strength should depend upon how far the atoms are separated from each other and how far they protrude

FIG.24. Schematic equipotential and electron flow lines near a n oxygen molecule adsorbed on a substrate of chemisorbed oxygen atoms.

above the substrate of chemisorbed oxygen. At 300” to 500°K. this substrate is likely t o be sufficiently agitated so that occasionally a hole or vacancy is momentarily produced. A molecular complex near this hole can sink into it and become adsorbed in the substrate. The extra high field disappears and with it the bright doublet. If the hole is only large enough to permit the “molecules” t o sink part way into it, there may result a closer approach of the two atoms, a smaller resolution or unresolved doublet, or a single spot; it might also result in less protrusion, less field, and hence a dimmer doublet. Conversely there should occasionally occur an unusually high concentration of atoms in the substrate, which pushes

188

JOSEPH A. BECKER

two atoms from the substrate into the third layer. This would result in the sudden appearance of a doublet. Higher temperatures should cause such changes more frequently. The interpretation that the observed quadruplet unit is due t o a n 0 4 molecular complex is supported by the first and last sketch in the second row of Figure 24. On a few occasions we have observed a square quadruplet suddenly change into an extended quadruplet like that in the last sketch; however, it would persist in this form for only a fraction of a second and would then return to a square unit. During about 3 t o 5 sec. the unit was observed to oscillate back and forth between these two forms about ten or twenty times. It then disappeared completely or became steady as another form of unit. One might ask why the first few molecules or atoms which are adsorbed on clean tungsten should not produce singlets or doublets. Our answer is that in this case the adsorption forces are so great th a t the shared electron spends almost its entire time between the oxygen nucleus and the tungsten atoms. The probability of finding the electron beyond the oxygen nucleus is so small that even the extra high field in this region produces negligibly small field currents. The observed field currents come most likely from the W atoms, which are not covered by oxygen. Quantitative calculations on the magnification and resolving power for a molecule on top of a substrate as discussed above have been made by D. Rose (13) of the Bell Telephone Laboratories. He finds th a t the observed sizes and resolutions of the bright spots are too large if the tungsten point is smooth and has a radius of curvature of about 1 X cm. If on such a point, however, there are protrusions or ridges cm., then any molecules having a radius of curvature of about 2 X on such protrusions or ridges could be magnified and resolved t o give bright spots like those observed. Such sharp ridges or protrusions can conceivably be produced by the interaction of chemical etching due to adsorbed atoms and temperature agitation or by the interaction of temperature agitation and high fields. d . Conditions Necessary to Observe Adatoms and Admoles. We might summarize by restating our views on the conditions necessary t o obtain bright spots or units from adatoms or admoles: (I) the adatoms or admoles must share electrons with the substrate so that the unit acts like a metal and so that emitted electrons may very quickly be replenished from the substrate; ( 2 ) the unit must protrude above the surrounding surface in order that the field above the unit be abnormally high and th a t the emission come predominantly from the unit; ( 3 ) the bond strength of the atoms in the unit with the substrate must not be too great; ( 4 ) the bond strength between atoms in the unit must not be too great, ( 5 ) the

ADSORPTION ON METAL S U R F A C E S

189

atom-atom bond must be weakened and must be comparable with the atom-substrate bond, (6) the number of units per square centimeter of surface must be small enough so that the images from neighboring units overlap only occasionally; i.e., the protruding layer should be only a few per cent covered; (7) the substrate must be reasonably compact or complete; (8) the surface of the point must be distorted in such a way that very small ridges or protrusions are produced. None of these conditions imply that the units must consist of oxygen only. I n preliminary experiments with nitrogen we have found bright units similar t o those described above. We expect that other molecules will show intensely bright spot emission. We also expect that metal atoms which are not adsorbed too strongly will yield bright spots. Cu atoms on W have been reported t o do this (14). However, we should not expect strongly electropositive metal atoms to yield bright spots. Here the valence electron has been transferred to the underlying metal and is not available t o be pulled out by the extra high field above the adatom. I n such cases the field emission electrons come from many underlying W atoms and are not concentrated in one bright spot on the screen. Only after the tungsten has been completely covered with Ba and a second layer is partly formed can one observe intensely bright, highly localized individual spots which might be interpreted as coming from individual Ba adatoms. If in the pretreatment of the point the oxygen pressure is raised t o mm. but the tube is not rebaked, the number of bright units, their intensity, and their persistance are greatly reduced. This raises the question what goes on during the rebake? We have no conclusive answer but suggest the following possibilities. At the baking temperature of 600°K. water vapor reacts with the tungsten surface to enlarge certain planes which meet other planes and produce ridges. Where three planes meet they may form sharp protrusions or “mountain peaks.” This produces a second stage of magnification and resolution for molecules adsorbed on the ridges or peaks. Another possibility is that a small amount of hydrocarbon vapor deposits on the tungsten point during the baking or while the liquid air is removed from the trap. This might result in the adsorption of a very few carbon atoms with the oxygen in the first two layers. These carbon atoms might act as centers a t which molecular complexes in the third layer are held more tightly. It is known th a t carbon is strongly adsorbed on tungsten, and one might expect that adsorbed carbon forms molecular complexes with adsorbed oxygen. Perhaps this is one of the reasons for the great variety of units shown in Figure 23. If this suggestion should be confirmed by further experiments, it would illustrate once again the great versatility of the field emission microscope

190

JOSEPH A. BECKER

in the study of adsorption, catalysis, and the arrangement of atoms in adsorbed molecules. e. Chnngrs in Surface Free Energy Due to Adsorption. The first three types of experiments dealt with the question how docs the arrangement of the adsorbent metal atoms affect the adsorption properties of the adsorbate? l'hc fourth type of experiment deals with the question how does the presence of the adsorbate affect the arrangement of the adsorbent metal atoms? Before considering the last qucstion, let us consider a simpler one: how do tungsten atoms tend to arrange themselves on a clean tungsten surface in the shape of a sphere? This question has been discussed by C. Herring (15). The general answer is th a t they will t r y t o rearrange themselves in such a manner th at the total surface free energy is a minimum. Translated into the language used in this article, this means that as many atoms as possible will try t o make as many bonds as possible, or t ha t those planes in which the atoms make six or five bonds will enlarge while those planes in which the atoms make four bonds will grow smaller. From Figure 19 it follows th at the 110 and 211 planes will grow a t the expense of the 111 and 100 regions. Of course this process will happen only if the temperature is high enough so th a t surface atoms may migrate a t an appreciable rate. For metals this occurs a t about one third the absolute melting temperature. For tungsten this is about 1200°K. Previous work with the field emission microscope (16) shows that this prediction is verified by experiments a t 1200" to 1600°K. At temperature of 2000°K. or higher the surface atoms are so violently agitated that the surface acts somewhat like a liquid in which surface tension forces tend to make the surface approach a sphere. Let us now consider how this situation is altered if oxygen is adsorbed on the surface. The surface will now tend to take a shape such th a t the surface free energy of the system for both tungsten and oxygen will be a minimum or that the tungsten and oxygen atoms will make the greatest number of strong bonds. From experiments described above it follows that oxygen makes strong bonds on the 211 planes or 211 regions. Hence in the presence of oxygen gas a t a pressure a t which a layer of oxygen is adsorbed a t 1400°K. we should expect the 211 planes t o grow. Some of the tungsten which is removed should be deposited onto the 111 plane from the three nearest 211 planes. This should make the 111 regions protrude above the surface. This should increase the field strength and hence increase the emission from this region. Experiments verify this prediction. I n the 100 region similar effects occur when the tungsten is heated a t about 1400°K. in the presence of adsorbed oxygen. This is shown in a striking manner if the emission pattern is observed under these conditions. The emission from the 100 region is observed to fluctuate violently in

ADSORPTION ON METAL SURFACES

191

subdivisions of this region as if the surface were being violently agitated. After a while the 411 plane and 310 plane decrease in emission while the 210 plane increases its emission. This explains the peculiar behavior of the 100R curve in Figure 22 a t 1400°K. At 1600°K. or higher, when the oxygen has been desorbed, the point assumes a shape characteristic of clean tungsten. These observations show that in catalytic experiments in which the temperature approaches bne third of the melting point, the presence of an adsorbate may alter the surface planes of catalyst crystallites and thus alter the adsorption and catalytic activity. If the particles are very small, such effects may occur a t even lower temperatures.

4. The Bearing of These Experiments o n Catalysis The significance of the results described in this chapter for catalytic theory and practice are so obvious that we need do no more than summarize the results. When oxygen is adsorbed on tungsten, the sticking probability averaged over all planes is about 0.10 until one layer is adsorbed. Some planes adsorb a t a faster rate than others. Beyond the first layer the adsorption rate decreases quite materially and again the rate varies from plane t o plane. At a low enough pressure some planes may adsorb only one layer while others adsorb two or more layers. When oxygen is desorbed, the first layer cannot be removed below about 1600°K.; it is held with a bond strength of 5 ev. The second layer cannot be removed below 800°K.; its bond strength is 2.5 ev. The variation of the bond strength for any layer for different planes amounts to about 20%. The first two layers are chemisorbed as atoms but evaporate as molecules. If the pressure exceeds about 10V mm., a third layer is adsorbed. At least part of this is chemisorbed as molecules with a bond strength of about 2.0 ev. Even at room temperature these molecules may change their orientation with respect to the substrate. 0 2 and 0 4 molecules are observed frequently, O3 and 0 6 less frequently. Some of these molecules have vibrational and rotational degrees of freedom. These molecules do not migrate over many atom diameters or lattice sites before they disappear in the substrate or evaporate. There are probably physisorbed molecules in the third layer which are held with smaller energies and which migrate over the surface even at 300°K. If the temperature exceeds about one third the melting point and one or more layers are adsorbed, the metal atoms rearrange themselves; some planes are enlarged while others shrink, and new planes develop. If the temperature exceeds one half the melting point, small metal particles tend t o become spherical in shape.

192

JOSEPH A. BECKER

Probably in most catalytic processes the metal surfaces are covered with three or more layers. The dissociation energy for admoles in these layers is probably only half or one third of th at for molecules in free space, and the binding energy to the substrate are probably small enough so that the evaporation rates are appreciable and migration rates are high. The sticking probabilities are quite small. These conditions are favorable t o catalytic reactions, and for best results, the pressure and temperature must be adjusted to bring about these conditions.

VI. OTHERADSORPTION EXPERIMENTS WITH EMISSION MICROSCOPE

THE

FIELD

1 . Introduction

I n this chapter we shall discuss briefly a few of the many interesting experiments on adsorption which have been performed with the field emission microscope, chiefly by E. Mueller (11). These will include the adsorption of tungsten on tungsten, the desorption of positive hydrogen ions by high fields, the desorption of positive barium ions by high fields, the adsorption of Hz and HzO on W, and the adsorption of C and 0 2 on W. This discussion will show that high electric fields in the proper direction can pull off positive ions; that the field strength required to do this depends on the plane, on the concentration of adions, and on the temperature. From some of these experiments some interesting details of surface catalysis can be deduced. 2. Adsorption of Tungsten on Tungsten

I n a specially designed microscope, Mueller (17) deposited tungsten from a hot wire onto the single crystal point of the microscope and observed the electron emission pattern on the screen. These patterns show very clearly that the sites on which tungsten atoms are adsorbed depend markedly on the temperature of the point while the tungsten is deposited and on the plane on which it strikes. At room temperature the tungsten atoms are deposited on all planes and can move only a small atomic distance from the place where they strike; they thus form very small crystallites, particularly on the 110, 211, and 100 planes. These planes are normally the poorest electron emitters, but when tungsten atoms are deposited on them in a rapid and irregular manner they become the best emitters. There are two reasons for this: because of the irregular shape of the crystallites, the local field strength is abnormally high, and a freshly deposited tungsten atom is not built into its normal lattice position, makes fewer contacts with its neighbors, and hence gives up its electron more easily.

ADSORPTION O N METAL SURFACES

193

When the point is a t 600°K. while the tungsten is deposited, the atoms which strike the 21 1 plane migrate rapidly in the I 11 direction and deposit themselves on the edges of the plane, where they form a crescent-shaped mound; on the 110 plane they can move only short distances t o form fairly large, brightly emitting crystallites; on the 100 plane they can form only very small crystallites. If t he tungsten is deposited on a 750°K. point, it, can migrate and become deposited on the edges of the 211 and 110 planes but it still cannot migrate on the 100 plane. The tungsten which is deposited on a point at 1290°K. can migrate on all three planes and forms broad edges on all of them. If a larger amount of tungsten is deposited a t this temperature, these edge deposits become so broad that two 110 edges meet or a 110 edge meets a 100 edge. If only a small amount of tungsten is deposited while the point is a t 300"K., there are observed on the 110 plane not only bright crystallites but also numerous faint round spots. The diameter of these spots is th a t to be expected from individual tungsten atoms. ,4s the temperature is raised, these spots become agitated and disappear in 200 sec. a t 830°K. or in 3 sec. a t 1160°K. From this Mueller calculates an energy of migration on the 110 plane of about 1.2 ev. Similarly he calculates that the energy of migration on the 100 plane is about 2.0 ev.; tungsten atoms which have been deposited on the 110 edge require about 3.4 ev. for migration. From the model shown in Figure 19 it follows that to make tungsten atoms migrate in these three locations requires that the atom break one, two, and three bonds respectively. Hence we deduce th a t t o break one W-W bond requires about 1.1 ev. 3. The Desorption of Positive Hydrogen Ions by High Fields

Mueller (11) found that when sufficiently large positive fields (i.e., with the point positive) were applied a t a tungsten point on which hydrogen molecules were impinging, a steady positive ion current could be drawn from the point. It is presumed that this current consists of protons, i.e., singly charged hydrogen ions." I n a particular case in which the cm. the pressure of radius of curvature of the point was 860 X hydrogen was 6 X mm., the applied field was 280 million volts/cm., and the ion current was 6 X amperes. When these protons strike the fluorescent screen they produce a faint but clear pattern consisting of many individual spots. Two spots which were separated b y 2 t o 4 X em. on the point can still be resolved on the screen. Such increased *Recent work by Ingrrtm and Corner shows that hoth ITf and €IL+ ions i x w cl~sorhcdby high fields.

194

JOS 15 I’ I I .\

. 1I W I i E 11

resolving power is to be expected because the proton has a much greater mass than the electron. For a clean tungsten point immersed in hydrogen gas, the pattern consists of a dark central ellipse with a bright edge on the 110 plane. This is interpreted to mean that hydrogen molecules which strike the 110 plane are desorbed as positive hydrogen ions a t the edge of this plane. In other regions of the surface the hydrogen molecules need to travel shorter distances to reach the edge of a plane, where they are pulled off as protons. Heiice these regions show dimmer spots or short lines of spots. I n another case, in which the tungsten point was heated after being covered with a small amount of carbon and then immersed in hydrogen at 6 X mm., the pattern shows five to eight bright concentric elliptical rings around each 110 plane. The separation between these rings correspoiids to about 25 X 1 0 P cm., or 10 tungsten atom diameters. This is about the spacing between 110 edges to be expected from a fullscale model like that shown in Figure 19. The pattern also shows a few bright rings around the 211 planes, as might be expected from the model, and a single small circular ring around the 334 planes. This plane does not form on normal clean tungsten but forms in a characteristic manner on a tungsten surface on which carbon is adsorbed. When the tungsten is heated to 400” to 500°K. while the applied field is between 200 t o 300 million volts/cm., the pattern becomes less sharp, the emission spots appear to be agitated, and the 110 rings shrink toward the renter. The proposed interpretation is th a t a t these high fields the tungsten atoms can be pulled off the edges of planes even a t 450°K. These experiments show again that the adsorption properties are quite different on the various crystallographic planes. They show particularly clearly that the edges of the planes have unusual adsorptive properties. Very likely these edges are the active catalytic centers which have been advocated chiefly by Taylor (8). These edges can readily be made inactive by small amounts of “poisons,” which are strongly adsorbed on the edges. The development of the 334 planes by carbon on hot tungsten is a typical case in which an adsorbate modifies the planes exposed by a n adsorbent.

4. ?’he Desorption

o,f

Barium on Tungsten by High F i d d s

When strong positive fields are applied t o a tungsten surface partially (+overedwith barium, the Ba is desorbed even a t 300°K. (1 1). The experiment is performed as follows. Ba is vaporized onto the point, which is then heated to a temperature a t which Ba migrates readily but does not evaporate. The average e is determined from the average electron emissioii

195

ADSORPTION ON METAL SURFACES

and applied voltage as described previously in section 111. A positive field is then applied for a measured time. The effect, if any, is observed by again reversing the field and observing the pattern. A higher positive field is reapplied and again the effect is observed with a negative field. This procedure is repeated at a series of values of 0, and this entire series of experiments is repeated at different tungsten temperatures. Mueller found that for an average 8 of 0.35 no Ba was removed with any field equal to or less than 101 Mv./cm. (i.e., 101 million volts/cm.) no matter how long the field was applied. However, when the field was increased to 102 Mv./cm., in a short time the Ba was completely removed from the edges of the 110 planes, i.e., from the 110 region. At 113 Mv./cm. the Ba is removed everywhere except in the 111 and 100 regions. At 125 Mv./cm. Ba is completely desorbed from the 111 region. At 150 Mv./cm. Ba is desorbed everywhere including the 100 region. It is presumed that the Ba is desorbed as Ba+. The critical field strength at which Ba is removed from the 110 edges is only 77 Mv./cm. when the average e is nearly zero. It increases linearly with the average 0 or until at = 1.0 it is 144 Mv./cm. These values pertain t o 300°K. If the tungsten temperature is 800°K.) the critical field is 61 Mv./cm. for 8 0; it increases linearly to 87 Mv./cm. at = 0.3; it is 95 a t t3 = 0.50; and it increases linearly to 106 at 8 = 1.0. For a tungsten temperature of 1200°K. the critical field for the desorption a t the 110 edges is 35 Mv./cm. at e 0 and increases linearly until it is 72 at = 0.70. If the temperature of the tungsten is high enough so that Ba migrates on all planes, then the Ba is desorbed on the entire surface at the critical field for desorption on the 110 edges. Obviously this means that Ba migrates to the 110 region, where it can be desorbed most readily by the positive field. We believe that these effects are due to the same causes which produce the emission of positive cesium ions at high temperatures, described in section 111. When Ba is adsorbed on tungsten, some of the Ba exists as singly charged positive adions whose radius is about 1.8 X cm. For small concentrations nearly every Ba particle is ionized; as 0 increases, the fraction ionized decreases. When large positive fields are applied to the surface, the Ba ions are pulled farther away from the surface. This decreases the attractive forces between the ion and its image charge. At some critical value of the applied field, the force tending to pull the ion off the surface slightly exceeds the attractive force and the ion is desorbed. As the temperature is increased, the average distance of the ion above the surface is increased and hence the critical field is decreased. Next let us consider how the critical field for desorption can be affected by the presence of additional adions. The field a t any ion due to its

e

e

e

e

196

JOSEPH A. BECKER

neighbors can be computed readily on the basis of the dipoles consisting of these ions and their electrical images. These fields are comparable with the critical fields but tend to keep the ion on the surface, so th a t the more Ba is adsorbed, the harder i t is to pull off any one ion. However, as soon as one ion has been desorbed, the attractive force on near-by ions is reduced and hence the applied field desorbs these more rapidly. This explains the suddenness with which the desorption occurs a t a critical field and why the desorption is complete as soon as the critical field is exceeded. Our proposed mechanism quantitatively explains several experimental constants. First consider the field required t o remove a single adion on the surface of a 110 plane. Mueller’s experiment gives 77 Mv./cm. The field holding a n ion on a metal surface due t o its image is 300e/412 = 360 X 10-10/Z2 if 1 is the distance the Ba nucleus is above the electronic image plane. At em., which is only 20% larger than the the critical field 1 = 2.1 X ionic radius. It is quite reasonable that such high fields can increase the equilibrium distance of the ion by this amount. Next let us compute the field at any one ion due to its neighbors when r3 is 0.10. For a square array of ions, the distance between ions occupying available sites is about 20 X 10-8 em. The ion field

h

3.6 X 1014 X 2 = 7.6 X lo6 volts/cm.

for 1 = 2.1 X cm. This is practically equal to 7.7 X lo6 volts/cm., which Mueller finds as the additional field that must be applied for e = 0.1 over that a t 0 = 0. For higher 8 values the problem becomes more complicated since only a fraction of the Ba is ionized a t any instant. Finally our proposed hypothesis explains the observed dependence of the critical field on the planes on which the Ba is adsorbed. On the basis of Figures 10 and 19, one would expect that 1 would decrease in the sequence for planes 110, 211, 111, and 100. Since the field on a n ion due to its image varies as l/lzl the critical field should increase in this sequence. This is what Mueller found experimentally. The conclusions t o be drawn from these experiments are 1. The fields required to desorb adions yield considerable information on the binding forces involved in adsorption. 2. These forces vary for different planes in a predictable manner. 3. For ionic adsorption the forces between adions are long-range forces and cannot be neglected. 4. An atomic model showing the arrangement, spacing, and valences of the surface atoms is a valuable guide for interpreting adsorption phenomena.

ADSORPTION ON METAL SURFACER

197

5 . Preliminary Catalytic Experiments with the Field Emission Microscope

I n a n exploratory sort of way the F.E.M. has been used to study catalytic reactions on metal surfaces. While such studies are far from complete, they clearly indicate th at this new tool can reveal details of the mechanism of such reactions to a much greater extent than is possible by any other known technique. We shall here discuss only two systems: H2, O2and H 2 0 on W and C and 0 2 on W. a. The System Hz and 0 2 on W . For 132 on W, Mueller (11) reports that i t is adsorbed as adatoms a t 300°K. and forms a saturated film for pressures from to mm. The mean work function of this film is 4.93 ev. At 70°K. part of the hydrogen exists as adatoms and part as admoles. The admoles are desorbed as the temperature is increased and are completely desorbed a t 300°K. if the pressure is less than mm. At 700°K. all the hydrogen is desorbed in a few seconds. I n some preliminary work we have found that some of the hydrogen adsorbed on tungsten a t 300°K. is desorbed a t lower temperatures than 700"K., also that the average work function is increased t o about 5.5 ev. The increase in (a is different for different planes. We found th a t reproducible results can be obtained only if the partial pressure of 0 2 or H2O is mm. or lower. Even for a partial pressure of HzO of mm. a t a n H2 pressure near 10-~mm., the tungsten is contaminated in a few minutes and clean tungsten cannot be regenerated below 2200°K. This emphasizes again that only in a system in which ultrahigh vacuums can be obtained are the adsorption results characteristic of H2 on W. Most of the work on th e adsorption of H2 on metals needs t o be reexamined under more nearly ideal conditions. Oxygen on tungsten has been studied in the F.E.M. by Mueller ( l l ) , Klein (18), and Becker (12), and the effects reported by all three investigators agree with each other. This system is comparatively insensitive to small amounts of other adsorbates. Oxygen increases the work function on all planes, but the rate of adsorption and the amount adsorbed are quite different for different planes. At 300°K. the amount of oxygen once adsorbed is not decreased by reducing the pressure of 0 2 to indefinitely small values. At temperatures from 650" to 800°K. some 0 2 is desorbed from the various planes. Between 800" and 14OO0K. the first layer is stable. One layer corresponds to one oxygen atom for every three or four tungsten atoms in a plane. The adatoms make only first-valence bonds with tungsten atoms; in the second layer oxygen adatoms make secondvalence bonds with the tungsten atoms. Above 1400°K. some planes begin t o desorb 0 2 , and a t about 2200°K. the 0 2 is completely desorbed on all planes. At about 1300" to 1100°K. the adsorbed oxygen reacts with

198

JOSEPH A. BECKER

the surface tungsten t o enlarge certain planes t o a much greater extent than for clean tungsten. At the higher temperatures a t which Oz is desorbed these planes shrink back to their normal size. When hydrogen is adsorbed on a tungsten surface covered with one layer of oxygen, hydrogen and oxygen are not desorbed a t 1150°K. even in 20 min. (11). Evidently H is held more firmly on 0 - W than it is on clean W. However, a t 1200°K. HzO is desorbed from the 100 region in a short time; a t 1300°K. HzO is desorbed on the 111 plane; a t 1500°K. clean tungsten is obtained in 2 min. Water vapor is adsorbed on tungsten in a manner similar t o th a t of oxygen, but the increase in work function is not so pronounced (11). The saturated film of H2O is not affected by temperatures below 700°K. Above 700°K. apparently a second layer is desorbed. Between 1250" and 1300°K. all evidence of adsorbed hydrogen disappears, and the pattern looks and acts like that observed when only oxygen was adsorbed. Apparently the tungsten surface decomposes the H2O and a t 1300°K. hydrogen is desorbed. The simplest explanation for this group of experiments is th a t when a tungsten surface is exposed to either Hz and 0 2 or to HzO, there exist on the surface H, Hz, 0 , 0 2 , and HzO. The amount of each species depends on the gas pressure of Hz, 0 2 , or HzO, on the plane, and on the temperature. The species which are desorbed are Hz, HzO, and Oz; the amounts desorbed depend on the relative amounts adsorbed, on the plane, and on the temperature. Oxygen is bound more firmly to tungsten than is either H z or HzO. H z is held on oxygen on tungsten more firmly than it is held t o clean tungsten. From these experiments it would appear that a tungsten surface which has been exposed t o OZor HzO cannot be completely freed from oxygen by heating it in a pressure of hydrogen of about 0.1 mm. or less a t temperatures below 1500°K. It is questionable whether oxygen can be removed from tungsten a t temperatures near 1000°K. when exposed t o hydrogen a t or near atmospheric pressure. b. The System C and 02 on W . Mueller (11) has reported the following characteristics of carbon on tungsten. Carbon can be deposited either by the decomposition of organic vapors on hot tungsten or by the evaporation from a carbon filament. I n the latter case the C which is deposited on the cold tungsten forms crystallites on all planes; because of the sharp corners, these cause increased field emission. At 1000°K. the C migrates very readily over the surface and produces a characteristic emission pattern. The exact details depend on how much C is deposited. The distinctive features of these patterns are that the 110 and 211 regions increase in emission while the 100 and 111 regions decrease; three new small dark spots appear on the 334 planes, which are near the 111 planes; a t a critical

ADSORPTION ON METAL SURFACES

199

temperature these 334 planes suddenly become bright but they darken again when the temperature is decreased. This critical T increases with the C concentration and varies from 900" to 1500°K. Presumably this behavior is due to the "melting" of a surface crystal of C on a particular plane on which the arrangement of the W atoms and C atoms are just right. All these changes in pattern originate from the changes in work function caused by adsorbed C and by the rearrangement of the W atoms due to the C and high temperatures. At 2800°K. the surface can be freed from C, but subsequent heating between 1000" and 2000°K. brings some C back on the surface. This shows that C not only evaporates from the surface but also diffuses into the interior of the tungsten. Carbon contamination of a W system can be removed only by heating the system for many hours at 2400°K. or higher. Klein (18) has studied the reaction of oxygen with carbon on tungsten. He first deposits C from a filament onto his point and permits it to migrate at 1250°K. for 5 min. He gives no data on how much C was deposited, but from a comparison of his photograph with photographs of Mueller's, we deduce that the amount was small. Klein then exposes this surface to O2 at a pressure of mm. for an unspecified time at 300°K. From our work we would expect that about two layers of 0 were adsorbed by this treatment (about 5 X 1014 oxygen atoms/cm.2). Klein pumps out the 0 2 gas to a pressure of mm. and observes the pattern at 300°K. The average work function p increases from 4.6 to 6.5 volts. The observed pattern shows that p for the 100 and 111 regions increases considerably more than p for the 110 and 211 regions. Klein then heats the surface to 1000°K. for 2 hr. This changes the pattern only slightly but reduces p to 6.2 volts. For 0 on W this treatment would have produced drastic changes in the pattern and in cp. From this we concluded that the presence of the adsorbed C greatly increases the binding energy of the 0 to the surface. Temperatures above 1000°K. produced drastic changes in the pattern. Heating at 1100°K. for only 30 sec. reduced p to 5.7 volts; the cp for the 100 region was most drastically reduced; 45 seconds more a t 1100°K. decreased 9 to 5.6 volts and produced only minor changes in the pattern. However, 11 min. more at 1100°K. increased 9 to 5.8 volts and produced more changes in the pattern: cp for the 100 plane and 111 plane and region was greatly decreased. The dark 334 holes, which characterize adsorbed C, disappeared; however, the tungsten was far from clean. The pattern resembled that for one layer of 0 on W. We believe that this treatment at 1100°K. removed all the adsorbed C in the form of CO. It was first removed rapidly from the 100 region and then gradually from the 100 plane and 111 region. We surmise that the

200

JOSEPH A. BECKER

gradual removal of CO resulted from the migration of the CO from the 100 plane and 111 region to the 100 region, where it evaporated rapidly. Th a t the C was removed as CO we deduce from experiments made a t Bell Telephone Laboratories by Francois and Hannay (19). I n these experiments CO was adsorbed on polycrystalline W, the CO gas was pumped out, and the products which evaporated at various temperatures were determined with a sensitive mass spectrometer. It was found th a t very little evaporation occurred below 1000°K. At 1070°K. large amounts of CO evaporated and little else. This experiment suggests that when CO is adsorbed on W, the C to 0 bond is not broken even though the CO is chemisorbed and must share electrons with W atoms. The abruptness with which the CO evaporation begins with rising temperatures indicates tha t in the 100 region the 0 atoms in the C-0-W complex touch neighboring 0 atoms and form a surface crystal of CO which melts a t a critical temperature and disappears a t the edges of the crystal. A similar effect was described in section V for 0 on W in the 100 region a t a lower temperature. From the paragraphs above it is clear tha t the study of the C-0-W system is incomplete primarily because the adsorption and the desorption of CO on W in the F.E.M. has not yet been investigated. Klein’s study should also be expanded for various amounts of adsorbed C and of 0. VII. DISCUSSION OF PREVIOUS WORKON ADSORPTION

1. Introduction I n this section we shall discuss the relationship between the work described in this chapter and that which preceded it. The chief objective will be t o compare the techniques, the experimental results, the interpretation of these results, and the theoretical concepts. We shall attempt t o point out which results have stood the test of time and which results need t o be discarded or modified. This discussion will be limited t o the work of the following people and their associates: Langmuir, Frankenburg, Roberts and Ilideal, and Beeck. I n making this selection, we unfortunately exclude many important contributions and for this we apologize in advance. To compensate partially we refer the reader t o the excellent and comprehensive books by Adam (20) and Dushman (21). 8. Langmuir and Associates (1911-35)

Starting about 1911 and extending to about 1935, Langmuir and his associates a t the General Electric Research Laboratories published a long series of important articles on adsorption phenomena on metal surfaces a t comparatively low gas pressures (20,21). This work has not only

ADSORPTION ON METAL SURFACES

201

stimulated others but has dominated a good deal of the Theory of the physics and chemistry of surfaces. Important advances in a field of science are often preceded by the discovery and development of new tools or techniques. I n Langmuir’s case these were the thorough baking of glass systems, the outgassing of metal parts, the development of the mercury diffusion pump t o produce better vacuums, the use of thermionic emission to follow adsorption processes, and ingenious methods for measuring small amounts of reaction products. He repeatedly emphasized the importance of a good vacuum and the avoidance of unwanted contamination. His quantitative results showed that relatively small amounts of impurities can drastically affect the adsorption in particular systems. Thus a small amount of oxygen on tungsten materially increases the amount of cesium adsorbed at a given temperature and pressure; the distance atomic hydrogen can travel in a glass system depends on the amount of water vapor adsorbed on the glass. By working with highly refractory metals such as tungsten, molybdenum, tantalum, and platinum, Langmuir could eliminate impurities which are usually present in metals and which can be eliminated only by hightemperature treatments. Langmuir’s studies of hydrogen on tungsten] oxygen on tungsten] water vapor on tungsten, and cesium on tungsten established the concept that adsorption energies were just as great as th e energies involved in typical chemical reactions. He showed that when Hz molecules strike a hot tungsten surface a fraction of the molecules are decomposed and come off as atomic H ; the fraction increases with the temperature. Oxygen molecules which strike tungsten a t temperatures above 1300°K. a t high enough pressures t o maintain an adsorbed layer combine in part with the tungsten and evaporate as WOs; most of the oxygen evaporates as 0 2 . When HzO strikes hot tungsten, it decomposes; some of the products evaporate as WO, and H; whether Hz and 0 2 evaporate was not determined. The WOs deposits on the glass walls, and subsequently some of it is reduced a t room temperature by atomic H, and HzO is given off. When cesium vapor strikes a hot enough tungsten surface, every cesium atom is ionized and evaporates as an ion. If the vapor pressure is high enough or the temperature is low enough so th at the tungsten surface is covered by more than about lo%, only Cs atoms evaporate. For intermediate conditions both atoms and ions evaporate. The cesium which is adsorbed on the tungsten drastically lowers the electron work function. These and other experiments led Langmuir t o develop some fundamental concepts on adsorption. One of these is that the forces involved in adsorption are just as large as those in chemical compounds. These forces are usually short-range forces. If the adatom is removed from the surface

202

JOSEPH A . RECKER

by more than a n atom radius, they are negligible. A related concept is that of a monolayer; when the surface is covered to such a n extent th a t additional adatoms can no longer contact adsorbent atoms, the adsorption forces abruptly decrease to much smaller values. For quantitative purposes, 1,angmuir usually calculates a monolayer as the number of adatoms which can be closely packed on a square centimeter of surface. I n some cases he considers the structure of the underlying surface. I n later work with Cs on W, he enlarged these concepts t o include somewhat longer range forces due to electrical dipoles; these forces die off more slowly and have appreciable values even a t 3 atom diameters. Another concept is that the electronic work function changes linearly with the amount adsorbed or that the dipole moment is independent of the concentration. The (1 - 0) concept states that the rate of adsorption for a constant arrival rate is proportional t o the fraction of the surface which has not, yet been covered. The last two concepts permitted Langmuir t o derive his famous adsorption isotherm, which has been verified by experiment in many systems (see the discussion in section IV). Langmuir’s experimental work for Cs on W convinced him that the (1 - 0) law was not applicable in this system. This work also led to the concept that the energies involved in surface migration were much smaller than the energies involved in evaporation. The techniques and experiments described in this article lead t o the following modifications of Langmuir’s concepts: (1) the (1 - 0) law is not applicable t o most systems; ( 2 ) the work function does not vary linearly with 0; (3) in certain systems, such as Cs on W, it is advantageous t o distinguish between adatoms and adions; (4) in other systems it is advantageous t o distinguish between adatoms and admoles; (6) the concept of a monolayer must be refined to take into account the structure of the substrate and the valence of the bonds with the substrate (the adsorption properties, such as sticking probability and heat of adsorption, change abruptly for layers which are only one half or one third of Langmuir’s monolayer) ; ( 6 ) various crystallographic planes differ greatly from one another with respect to electron work function, sticking probability, heat of adsorption, amount adsorbed, and catalytic activity ; (7) the edges of crystallographic planes have unusual adsorption properties; (8) the presence of an adsorbed layer a t high temperatures changes the crystallographic planes exposed on the surface and thus changes the adsorption properties. 3. Frankenburg and Associates (1944-46)

A great many experiments on adsorption and catalysis have been performed on finely divided metal powders, and much useful information

ADSORPTION ON METAL SURFACES

203

has been deduced from them. Since such powders are prepared by various processes, and since these powders can never be heated high enough t o degas them thoroughly and to free them from impurities introduced in their production, there is always the question of whether the results are characteristic of the metal or merely characterize the particular sample. Of all the work on such powders, th at done by Frankenburg (22) and by his associate Davis (23) is most likely to have approached clean surfaces. Frankenburg investigated the adsorption of hydrogen on tungsten and Davis investigated the activated adsorption of nitrogen on a finely divided tungsten powder. I n both cases the tungsten was prepared with extreme care, was heated in a purified hydrogen stream a t 1020°K. for 24 hr., and was then evacuated a t this temperature for about 50 hr. Pressures were measured by a McLeod gauge sensitive t o 5 X lo+ mm. The system included fourteen stopcocks and two COz snow traps. About half the glass parts could not be baked and hence were potential sources of water vapor. Even after the severe heat-treatment of the powder, the system still evolved enough gas t o be measurable after 16 hr. at room temperature; the pressure was "below mm." The water vapor pressure may have been even higher since a McLeod gauge does not measure it. When the powder was reheated t o 1020"K., th e pressure rose t o about loF3mm. Even if the tungsten was clean a t 750"C., it would soon be contaminated while it cooled to room temperature if the partial pressure of 0 2 or HzO were or even lo-$ mm. There is the further question of whether 0 2 or HzO, which must have been adsorbed in the preparation of the powder, could be completely removed by flowing Hz even a t 1020°K. Judging from modern vacuum practices, one would expect HzO contamination in such a system. I n the experiments on Nz on W and Hz on W with the F.E.M. described in previous sections, it is found th at residual gas pressures must be reduced t o about 2 X 10-lo mm. before Nz or Hz are introduced if the tungsten surface is t o remain clean for several minutes. If the residual gas pressure is lov9mm. and HZis introduced for 10 min. while the tungsten is a t 300' t o 6OO"K., the tungsten not only adsorbs hydrogen but also adsorbs some other gas which acts like oxygen. This is shown by the following experiments: ( 1 ) The HZis pumped out and the tungsten is glowed a t 1000°K. for a short time. This desorbs the hydrogen, but the surface is not clean because the pattern is very different from that for clean W, and the applied field required for 1 Fa. is considerably higher than t ha t for clean W. As the tungsten is glowed a t successively higher temperatures, the whole behavior goes through the same series of distinctive patterns as are observed for Oz on W. I n a glass system part of the residual gas pressure is HzO. This is decomposed on the tungsten into

204

JOSEPH A. BECKER

oxygen and hydrogen adatoms. The adsorbed hydrogen is desorbed a t 1000"K., but the oxygen cannot be completely removed until the tungsten is glowed a t 2200°K. ( 2 ) I n the second type of experiment for the same residual pressure of mm. the H 2 pressure is adjusted to about mm. and the tungsten is glowed a t 2400°K. t o clean it and is then cooled t o 300°K. for say 3 min. The tungsten is then flashed a t 1000°K. and the sudden rise in pressure is recorded. The tungsten is again cooled for 3 min. and reflashed a t 1000°K. The sudden rise in pressure is less than before. Each time this is repeated, the sudden pressure rise is smaller, which means that the rate of hydrogen adsorption decreases continuously as the tungsten surface becomes more contaminated with water vapor. This increasing contamination can be followed by observing the pattern each time the tungsten is flashed a t 1000°K. and cooled t o 300°K. If the tungsten is flashed a t 2400"K., the pattern returns to th a t for clean tungsten and the rate of hydrogen adsorption a t 300°K. returns t o its original value. These experiments show that unless the residual gas pressures in a glass system are lower than mm. a tungsten surface becomes contaminated in a few minutes and cannot be cleaned unless the tungsten is heated t o 2200°K. This is true even if H z is in the system. The contamination decreases the rate of adsorption of Hz or of Nz and changes other adsorption characteristics. That such a contaminant was present in Davis's experiments for N z on tungsten powders can be deduced by comparing his results with our results for N z on a tungsten ribbon. He found that the rate of adsorption at temperatures between 300" and 700°K. was extremely slow and that the rate increased with temperature. We find th at the rate of N2 adsorption for clean tungsten is not extremely slow and that it decreases as the temperature increases. Furthermore Davis finds that a t a pressure of mm. and a temperature of 900°K. the equilibrium amount adsorbed is only about 2 x 1013 N atoms/cm.2; under the same conditions we find about 50 X l O I 3 N atoms/cm.2. Frankenburg finds for H2 on W that the adsorption characteristics, such as the heat of adsorption, change rapidly when the amount adsorbed reaches 2 x 10'3 H a t o m ~ / c m . ~we ; find th at the adsorption characteristics do not change until the amount adsorbed reaches about 25 X 1013 H atoms/cm.z. Frankenburg found th at the heat of adsorption for H on W powders decreased rapidly when the tungsten surface was covered with about 2 x H atoms/cm.2, while Roberts reported that on a W wire the heat of adsorption was more nearly constant and decreased rapidly only after the surface concentration of hydrogen was about twenty times as great. Apparently tungsten powders heat-treated as drastically as those treated by Frankenburg are still not clean; a t least the adsorption

ADSORPTION ON METAL SURFACES

205

properties differ markedly from those of clean tungsten wires and ribbons.

4. Roberts, Rideal, and Their Associates

(1935-39)

The English school made a number of significant advances in the study of adsorption phenomena (24). They developed several ingenious techniques by means of which adsorption effects could be measured on the surfaces of tungsten wires. Because such wires can be heated readily to temperatures near 2600"K., the surface can be completely freed from all adsorbates. The area of these surfaces can be measured easily and fairly accurately. J. K. Roberts (24) and later A. B. Van Cleave (24) developed the technique for measuring the accommodation coefficient of neon on tungsten and showed that this coefficient increased considerably when gases such as hydrogen or oxygen were adsorbed on the tungsten surface. They used this effect to determine the total amount of H2 or 0 2 adsorbed and the temperatures at which hydrogen and oxygen could be desorbed. For oxygen on tungsten they found the three stages of adsorption described in section V. Roberts also used the tungsten wire as a sensitive calorimeter to measure heats of adsorption. I n both these techniques temperature changes of 0.01" t o 0.001"C. must be measured. Consequently after the wire is flashed it must be allowed t o cool nearly to the ambient temperature. This requires from 4 to 10 min. During this time considerable quantities of residual gas can contaminate the surface unless the residual partial pressures of 0 2 , Nz, CO, or HzO are less than or mm. Even a t mm. enough 0 2 strikes the surface in 10 min. to cover it with about ten monolayers if every molecule that strikes it sticks to it. Even if the sticking probability is 0.1, the contamination is serious. Roberts et al. report th a t in their system the residual gases did appreciably increase the accommodation coefficient and did make their values of heats of adsorption vary by about 20% from one run t o the next. T o minimize the effect of residual gases they inserted several liquid nitrogen traps, some of which contained charcoal. They first measured pressures with a McLeod gauge and later used a very sensitive Pirani gauge, which could detect pressures of about 2 X lo-' mm. This should be compared with modern ion gauges, which can measure pressures of 10-lo mm. Their system included three greased stopcocks, two ground-glass joints, a McLeod gauge, two mercury cutoffs, and six cold traps. It is now well established th a t the stopcocks, joints, mercury columns, and any glassware which is not baked give off large quantities of gas for months. It is assumed th at cold traps prevent these gases from passing through the trap. Unfortunately this is only partly true. The ion gauge and field

206

JOSEPH A. BECKER

emission tubes show that a considerable fraction of gases liberated between a diffusion pump and a liquid nitrogen trap pass through the trap and contaminate the glass and metal parts; this can be shown convincingly by torching glass or a condensed mercury droplet in these locations. These tools also show that any system containing greased stopcocks or unbaked glass and cold traps cannot be pumped out to pressures below 10V and that a clean tungsten surface anywhere in the system is seriously contaminated in a fraction of a minute. How seriously this contamination affected the results is difficult to decide. From our experience we estimate that for clean tungsten, the initial heats of adsorption and the equilibrium amounts adsorbed might differ from the reported values by 30 to 50%; the shape of the curve for heat of adsorption vs. amount adsorbed may be quite different from that reported: instead of being smooth it may show sharp drops a t a critical value of 0. As a result we feel that any theory based on these results needs t o be reexamined. Another ingenious technique used by the English school is exemplified by the work of Bosworth and Rideal (25) on 0 2 on W. They determined changes in the average work function of a tungsten filament, on which oxygen was adsorbed, when it was heated t o a temperature between 1270" and 1930°K. The change in average work function is equal t o th e measured change in contact potential between a clean hot tungsten filament, which serves as the source of electrons, and the test filament. The test filament a t room temperature was first exposed t o oxygen gas a t some moderate pressure, which was then reduced. The test filament was heated t o 1270°K. and the contact potential followed for 8 min. The whole procedure was repeated for higher temperatures. From their data they concluded t hat the test filament very quickly reaches a work function 1.8 volts higher than that of clean tungsten; in time the work function decreases rather rapidly and then more and more slowly; the energy of adsorption for oxygen on clean tungsten is 6.5 ev. and decreases to about 3.0 ev. for the coverage a t 1270°K. Very likely they would have found even higher work functions and lower heats of adsorption if their experiments had been performed a t lower test filament temperatures, a t which more oxygen could have been adsorbed. I n all the experiments discussed here, the test surface consisted of many crystals, and hence the results are average values for all the planes that may have been exposed. How significant this averaging can be is brought out by more recent experiments, which show th a t the work function for clean tungsten varies from 4.2 t o 5.5 or even 6.0 volts, while for tungsten covered with one layer of oxygen it varies from 5.2 t o 6.6 volts. Thus the variation between planes is almost as large as the change

ADSORPTION ON METAL SURFACES

207

produced by the adsorbed oxygen. Other recent experiments show th a t the rate of adsorption, sticking probability, heat of adsorption, surface mobility, and equilibrium amounts adsorbed all vary widely between planes. Roberts was, of course, interested in knowing the number of oxygen atoms per square centimeter of surface which were adsorbed in the first layer. This is the layer which starts to desorb a t about 1600” t o 1700°K. and has the highest heat of adsorption. He measured this amount by letting in measurable amounts of oxygen in successive increments, or what we shall call “shots.” After the first, the second, or the third shot the pressure was soon less than 2 X lo-? mm., indicating rapid adsorption. After the fourth shot the pressure was measurable on the Pirani gauge and kept on decreasing slowly with time. This slow decrease in pressure may be due to slow adsorption on the glass walls, particularly in the cold trap or traps. From the known volume of the system and the amount in one shot Roberts could calculate what the pressure should have been if there had been no oxygen adsorbed on any glass surfaces, mercury surfaces, or the tungsten wire. From the residual pressure after any shot he could then calculate how many molecules were adsorbed. For the first three shots he then assumed that the oxygen is adsorbed only on the tungsten whose surface area is about 0.5 cm.2 and th a t no oxygen was adsorbed on the cold-trap surfaces, on the mercury, or on the glass surfaces, whose areas were hundreds of times larger than the tungsten area. On the basis of this assumption, he calculated th a t the tungsten surface when i t was covered with the first film or layer had a surface concentration such th at there was approximately one oxygen atom adsorbed on every tungsten atom, and hence he called this a “monolayer.” I n five successive experiments the saturation concentration differed by a factor of 1.3. On the other hand, we find that when the first oxygen layer is complete the concentration is about 2.5 X l O I 4 atoms of oxygen/cm.2 of surface, which is only about one fourth the concentration th a t Roberts reports. We propose that the assumption which Roberts made is not valid and that the oxygen in the first three shots is partly adsorbed in the cold trap and only partly on the tungsten. I n his system mercury vapor was continuously being deposited on the walls of the cold trap. A fresh layer of mercury was deposited in a matter of minutes. It is known th a t a clean mercury surface adsorbs oxygen or hydrogen. I n his system the shot of oxygen first passed through the cold trap before it could reach the tungsten. Why then should it not in part be adsorbed on the walls? Roberts justifies his assumption by pointing out th a t the eighth or ninth shot does not result in any appreciable adsorption on any surface.

208

JOSEPH A. BECKER

If t he tungsten surface saturates, why should one not expect that the cold trap saturates also? If there is a long interval between experiments the trap adsorption can be rejuvenated by fresh films of mercury. Hence one would expect variations in the trap adsorption between various experiments depending on the time interval between experiments. This may be the reason why Roberts’ calculated concentrations have such a large spread. Owing to the importance of the experiments performed by the English school and the important part that the experimental results have had in shaping concepts on adsorption phenomena, it would seem to be highly desirable t o repeat the experiments on accommodation coefficient, on heats of adsorption, on saturation amounts adsorbed, and on contact potential in improved systems by use of the following suggestions: ( 1 ) eliminate all stopcocks, greased joints, mercury cutoffs, and mercury gauges; ( 2 ) use modern ion gauges to replace Pirani gauges; ( 3 ) bake out all glassware including cold traps up to the diffusion pump; use only valves which can be thoroughly degassed; and ( 4 ) make measurements on single crystals of metal ribbons. Judgment on theories of adsorption should be suspended until such experiments have been made. 6. Beeck and Associates (1937-50)

Another great step in advance in the understanding of adsorption and catalysis was made by the late Otto Beeck and his associates a t the Shell Development Laboratories in California. This work was in part described by Beeck (2) in Advances in Catalysis in 1950 and was reviewed in 1952 by his chief associate, A. Wheeler (26). The unique feature of this work is that adsorption was studied on metal films evaporated onto glass, either in a good vacuum or in an atmosphere of inert gases. A typical sample might weigh 50 mg., have an apparent area of about 30 but have an interior or atomic area of 5000 cm.2;its thickness might be 5000 atomic layers. Metal films deposited in a good vacuum were unoriented, while those deposited in argon were oriented with a simple plane, having a high surface energy, parallel to the glass surface. Pressures were measured either with a McLeod or with a Pirani gauge; no ion gauge was used. Residual gas pressures “were always less than 10-6 mm.” The atomic area of the film was measured by the amount of physisorption a t low temperatures (B.E.T. Method). Their system included stopcocks, mercury columns, cold traps, and portions of glass that could not be baked. Nevertheless, their films were probably free from serious contamination. Any residual gas molecules which entered their testing tube would first strike thinner portions of the deposited metal film, where they could be adsorbed. Furthermore, even

ADSORPTION ON METAL SURFACES

209

if all the residual gas which could enter their testing tube were adsorbed on their main film, it would probably amount to less than layers. The self-consistency of their experiments under varying conditions supports this conclusion. With this system they measured the equilibrium amounts of gas adsorbed on the metal films near 78” and near 300°K.; they also measured the heats of adsorption. They studied CO, H2, 02, N2, ethylene, acetylene, and other gases on such metal films as W, Ta, Ni, Fe, Pd, and Rh. Some of the conclusions of Beeck et al. are 1. Sintering greatly reduces the atomic area of the films; thilower the melting point of the metal, the lower the temperature a t which sintering occurs. This conclusion agrees with our generalization that metal atoms are mobile on their own surface a t about one third of their melting point. mm. essentially a monolayer of Hz, CO, 2. At pressures of ethylene or acetylene was adsorbed in less than 6 sec. They define a monolayer as one adsorbed atom or molecule for each surface tungsten atom. When the pressure was increased from to lo-’ mm., less than 20% more was adsorbed. From their data we have calculated th a t the average sticking probability for their exterior surface must have been about 0.1. Of course, the mechanism of adsorption on spongy films is more complicated than that on simple solid surfaces, and so no exact value can be calculated. I n 6 sec. at mm. about 1000 layers strike the exterior surface. If one in ten molecules is adsorbed, 100 layers stick. Such molecules must rapidly “migrate” or evaporate into the pores and, after many collisions on interior surfaces, become adsorbed. At any moment the concentration of adsorbed particles must be larger near the outer surface of the film than near the interior portions. Any molecule which reaches a spot deep inside a pore must make many surface collisions before i t can again escape. 3. Very rapid chemisorption of Nz on W was observed at room temperature. Nz on Fe was initially very fast; a t 0.1 mm. the adsorption vs. time curve had a long slow tail. On Fe only one fourth as much Nz as H2 was adsorbed. On W more than half as much Nz as H2 was adsorbed. For p > lO--4 and T < 373°K. twice as many molecules of CO were adsorbed as of H2 molecules. “This indicates that H2 molecules are dissociated into two H atoms.” 4. They calculated heats of adsorption from the measured rise in temperature of the glass on which their metal films were deposited. This rise was about 0.1”C. Their procedure was to admit a n increment or shot of gas sufficient t o cover the atomic surface to about one fifth of a monolayer (i.e., one adatom for five metal atoms). They assume th a t each shot covers the metal film uniformly throughout its depth. We believe that

210

JOSEPH A. R E C K E I l

this assumption is not likely t o be correct and th a t the first shot covers part of the surface to a monolayer and other parts to less than a monolayer. Consequently, we believe that their heat value for the first shot is lower than the true initial heat of adsorption, also that the shape of their heat vs. 0 curves is not the correct one. We predict that when true heat vs. 0 curves are obtained on single crystallographic planes, the heat values will decrease abruptly a t those values of concentration a t which th e adsorbates make second- or third-valence bonds with metal atoms. They report initial heats of adsorption in kilocalories per gram mole. When converted to electron volts, some of these are for O2on W, 5.5;for N2 on W, 4.1; for Nz on Fe, 1.7; for H2 on W, 2.0. These are average values for all crystallographic planes and are probably about 20% too low. They also report that the heat of adsorption decreases as the amount adsorbed increases. Near a monolayer the adsorption energy is only 0.7 ev.

REFERENCES 1. Langmuir, I., J. Am. Chenz. SOC.37, 1139 (1915). 2. Beeck, O., Advances in Catalysis 2, 151 (1950). 3. Becker, J. A., Phys. Rev. 28, 341 (1926). 4. Langmuir, I., and Taylor, J. B., Phys. Rev. 44, 423 (1933). 5. Martin, S. T., Phys. Rev. 66, 947 (1939). 6. Lennard-Jones, J. E., Trans. Faraday SOC.28, 333 (1932). 6a. Drechsler, M., 2. Elektrochem. 68, 327 (1954). 7. Becker, J. A., and Hartman, C. D., J. Phys. Chem. 67, 157 (1953). 8. Taylor, H. S., Advances in Catalysis 1, 1 (1948). 9. Becker, J. A., in “Structure and Properties of Solid Surfaces,” p. 459. Univ. of Chicago Press, p. 459, 1952. 10. Langmuir, I., Phys. Rev. 22, 357 (1923); see also Brattain, W. H., and Becker, J. A., Phys. Rev. 43, 428 (1933). 10s. Fowler, R. H., “Statistical Mechanics,” 2nd ed., pp. 829-831. Cambridge Univ. Press, New York, 1955. 11. For a detailed description of the Field Emission Microscope and its applications see the excellent review article by E. W. Miiller on “Feldemission,” Ergeb. exakt. Nalurwiss. 27, 290 (1953); for a brief description of the FEM, see ref. 7. 12. Becker, J. A., and Brandes, R. G., J . Chem. Phys. 23, 1323 (1955). 13. We are indebted t o our colleague D. Rose for this information. 14. Ashworth, F., Advances in Electronics 3, 1 (1951). 15. Herring, C., i n “Structure and Properties of Solid Surfaces,” p. 5 . Univ. of Chicago Press, Chicago, 1952. 16. Becker, J. A., Bell System Tech. J . 30, 907 (1951). 17. Mueller, E. W., 2. Physik 126, 642 (1949). 18. Klein, R., J . Chem. Phys. 21, 1177 (1953). 19. We are indebted to our colleagues for this information. 20. Adam, N. K., “The Physics and Chemistry of Surfaces.” Oxford Univ. Press, London, 1941. 21. Dushman, S., “Vacuum Technique,” Wiley, New York, 1949.

ADSORPTION ON METAL SURFACES

211

22. Frankenburg, W. G., J. Am. Chem. Soe. 66, 1827, 1838 (1944). 23. Davis, R. T., Jr., J. Am. Chem. SOC.68, 1395 (1946). 24. This work is described in the book by Miller, A. R., “Adsorption of Gases on Solids.” Cambridge, Univ. Press, New York, 1949. 25. Bosworth, R. C. L., and Rideal, E. K., Proc. Roy. SOC.(London) A162, 1 (1937). 26. Wheeler, A., in “Structure and Properties of Surfaces,” p. 439. Univ. of Chicago Press, Chicago, 1952.

This Page Intentionally Left Blank

The Application of the Theory of Semiconductors to Problems of Heterogeneous Catalysis K. HAUFFE D e ~ u T t ~ of e ~Solid t State Physics, Central Institute for Industrial Research, Blindern, Oslo, Norway Page I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 11. The Mechanism of the Reaction of Gases with Semiconducting Crystals. . . 216 1. The Electron Defects in Oxides Influenced by Gases.. . . . . . . . . . . . . . . . . 216 219 2. The Boundary Layer Theory of Chemisorption.. . . . . . . . . . . . . . . . . . . . . . 3. The Work Function of Electrons and the Kinetics of Chemisorption. . . . 230 111. The Mechanism of SimpIe Reactions on the Surface of Semiconducting Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 1. The Mechanism of the Decomposition of NzO on Semiconducting Oxides 237 2. The Mechanism of Carbon Monoxide Oxidation on Solid Oxides.. . . . . . . 243 3. Some Further Simple Reactions Catalyzed by Semiconductors. . . . . . . . . 249 References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

I. INTRODUCTION In spite of the great technical success in the use of heterogeneous catalysts, the explanation of their working mechanism is still unsatisfactory because of the complicated reactions at the catalytic surfaces. Without treating the metal catalysts, we shall discuss here the catalyzed gas reactions on semiconducting catalysts, e.g., on oxides and sulfides. This restriction has a logical reason. As far as is known today, in heterogeneous catalyzed reactions the electrons are very important to the catalytic behavior of a catalyst. As we will show later in detail, the electronic structure of a catalyst will be markedly changed by interaction with the reacting gases, especially in the region of its surface. It is important to remember that a catalyzed reaction is believed to be preceded by a chemisorption step which is often accompanied by direct electron interchanges between the chemisorbed gas and the catalyst. Semiconductors are particularly attractive for study as catalysts, for we can treat their behavior and reactions with electrons by the well known methods of classical physical chemistry, e.g., by the Boltzmann statistics. In solid state physics, it is well known that many inorganic solids, e.g., the oxides and sulfides, can dissolve metals and nonmetals in excess, and that by this process electron and ion defects in the lattice will be formed. Wagner and co-workers (1) have developed the basic theory of 213

214

K. HAUFFE

semiconducting crystals possessing deviations from the ideal integral, stoichiometric composition. An excess of metal may be present in the form of metal ions in interstitial positions or in the form of anion vacancies (unoccupied anion places in the lattice) and an equivalent number of quasi-free electrons 8. In distinct contrast to this group of solids there are solids like NiO, where there are, due to an excess of a nonmetal component, cation vacancies and an equivalent number of electron holes $. Under the influence of an electric field or of a gradient of chemical potential of the electrons, the electrons can migrate by jumping from hole to hole in a p-type conductor or as free electrons in an n-type conductor. According to the band theory (2), the energy level of these disordered electrons is quite different from that of the electrons in the valence band. It is therefore evident that the interchange of electrons between the chemisorbed gases during chemisorption and desorption process can be influenced decisively by particularly favorable energy levels of the disordered electrons and by their mobility, Under the assumption of a homogeneous surface, any place in the surface will be equivalent to any other with regard to an exchange of electrons. In spite of this, it turns out that only a small part of the surface can be occupied by chemisorbed particles and, therefore, effective in the reaction. The amount of the occupied surface depends on the concentration of disordered electrons of the catalyst. The catalytic activity of only a small part of the catalyst surface was postulated by Taylor thirty years ago, and was ascribed to the existence of “active centers” in the surface of the catalyst. From the modern viewpoint we can consider Taylor’s active centers as the centers where the exchange of electrons is energetically favorable. But these centers need not be identical with the locations of vacancies, edges, and points at the surface. Although Taylor’s hypothesis served many useful purposes, later work showed it to be inadequate. The concept of active centers was linked with the concept of special geometric arrangements of the atoms in the surface of the catalysts (“roughness of the surface”). However this effect, as we shall show later, is of secondary importance. Of much greater importance is the nature and concentration of the electron defects in the surface, or more exactly, in the region near the surface of the catalyst. An example of this can be found in the catalytic decomposition of NzO. When a smoothly surfaced NiO catalyst is used, the rate of reaction is much faster than when an extremely rough surfaced catalyst, such as FezOaor SiOz, is used. By considering the chemisorption of oxygen, the role of interchange of electrons between the gas and solid will become increasingly evident.

APPLICATION OF T H E THEORY OF SEMICONDUCTORS

215

At a rough surface of a solid without electron defects, the molecules of oxygen do not dissociate faster into their atoms than in air because no electron interchanges take place. I n contrast to the homogeneous gas phase, the thermodynamic situation a t the surface of a suitable catalyst is quite different. Due t o the electron affinity of oxygen, the electron can be transferred t o the chemisorbing oxygen

and, consequently, there will be no chemisorbed oxygen atoms, but oxygen ions, in the surface. This can be proved by the change of the electrical conductivity of the oxides due t o the change of the concentration of electron defects in the region near the surface. Since charged as well as uncharged particles are involved in the over-all reaction, it is necessary to use electrochemical thermodynamics for the treatment of the mechanism of chemisorption. Besides the chemical potential, we have t o include the electrical potential involved during the chemisorption. Therefore, we need quite different relations between the chemisorption and the physical adsorption. The explanation of the influence of the electron defects of a semiconducting catalyst upon its catalytic activity is, at present, the subject of intensive research in the field of heterogeneous catalysis. Wagner and Hauffe (3), in 1936, began their research on the heterogeneous catalytic decomposition of NzO and on oxidation of CO, with NiO and CuO a s catalysts. Using measurements of electrical conductivity, these authors could demonstrate the role of electron defects in these catalytic reactions. At a later time, Wagner (4) studied the decomposition of NzO using ZnO, and ZnO containing small amounts of Gaz03 as catalysts. Dowden (5) has given a review of previous work and has pointed to a parallelism between the catalytic activities of p - and n-type conducting catalysts and their electron defect structure. Stimulated by the work of Anderson (6) and of Garner and co-workers (7), Dowden proposed mechanisms for the reactions of gases and of electrons a t the surface of catalysts, applying the simplified band theory of semiconductors. This treatment is qualitative and is not yet refined enough for deducing from it a specific mechanism. More work along the same lines led Garner and associates (8) t o some very interesting observations concerning the mechanism of chemisorption and desorption processes. We shall deal with these later. Weyl (9) has also outlined a picture of the mechanism of heterogeneous catalysis, which is similar t o the schemes proposed by the above authors. His suggestions, based on the “quanticule theory’’ of Fajans (lo), also result in a qualitative description. The importance of the nature and the concentration of electron

216

K . HAUFFE

defects in oxidic catalysts was emphasized by Krauss (11) too, who was able t o show that the amount of NzO formed in reacting mixtures of ammonia and oxygen is dependent upon the oxygen excess (identical with the concentration of electron holes) in the oxide catalysts. BBnard (12) also published a qualitative description of the correlation bet ween catalytic activity and the electron defects of oxidic catalysts. Taylor’s lecture “Catalysis in Retrospect and Prospect ” (13) represents a n excellent contribution, in which the present knowledge about the interchange between solids and gases is treated (14). I n the field of nonmetallic catalysts, particularly of oxides, Hauffe and co-workers (14a) used only semiconductors for which information concerning electronic and ion defects was available from measurements of electrical conductivity, thermoelectric properties, and Hall effect. These workers obtain a quantitative correlation between the reaction rate, the amount of chemisorption, and the number of electron defects of the catalysts. Since every catalyzed reaction is initiated b y a chemisorption process involving one or several of the reacting gases, and because the nature of this chemisorption process determines the subsequent steps of the reaction, it seems appropriate t o begin with a discussion of the mechanism of chemisorption.

11. THEMECHANISM OF THE REACTION OF GASES WITH SEMICONDUCTING CRYSTALS 1. The Electron Defects in Oxides Influencedby Gases

At high temperatures the attack of oxygen on a surface of a n oxide is quite different from that a t low temperatures. Due t o the great mobility of ions in solids a t high temperatures, metal ions can diffuse in the presence of oxygen from the interior t o the surface and there combine with chemisorbed oxygen, forming more oxide molecules a t the surface. It is, therefore, evident that the number of ion and electron defects increases or decreases with increasing oxygen pressure, depending on the disorder nature of the oxide. In the case of a p-type conductor (e.g., NiO) we can write $~OZ(Q i= ) NiO Nio ” 2c~ (1)

+

+

I n the presence of oxygen, the concentration of nickel ion vacancies, Nio”, and of electron holes, CB, will be increased as shown in Equation (1). Oxides containing an excess of metal behave in a different way; a representative of this group is zinc oxide. An excess of metal may be present in the form of zinc ions in interstitial positions, Zno’ or Zno”, and

APPLICATION O F T H E THEORY O F SEMICONDUCTORS

217

an equivalent number of quasi-free electrons 8. For high temperatures we can write the reaction of ZnO with oxygen in the following way

j.50,(0)+ Zno”

+28

ZnO (2) In this case, the number of zinc ions in interstitial positions and the number of free electrons will be decreased by an increase in the partial pressure of oxygen. These disorder reactions result in a dependence of the electrical conductivity on the oxygen pressure. This effect is a well known phenomenon in the field of semiconductors (1). Complicated relations, however, will occur a t lower temperatures, a t which no equilibrium can be attained between the gas phase and the lattice defects in the whole

20-

n w )r

e

W C

W

Distance from the Surface FIG.1. Energy scheme of chemisorption and physical adsorption of oxygen vs. distance from the surface according to Lennard-Jones. E ~ fisf the electron affinity of atomic oxygen, E Dthe dissociation energy of oxygen molecules, E c the ~ chemisorption energy, and EAct the activation energy. Position A is that of physically adsorbed 0 2 , and position B is t h a t of chemisorbed 0-.

crystal in a reasonable time because the mobility of the ion defects is too small. Under these conditions the reaction will not proceed much beyond the first step, namely the chemisorption of oxygen. Thus at such lower temperatures, often only chemisorption is found, without noticeable incorporation of oxygen into the lattice proper. I n this case a mere displacement of electrons results, i.e. either a transfer of free electrons in the case of a n n-type conductor, or a transfer of electrons over holes in the valence band from the solid to the oxygen during the chemisorption in the case of a p-type conductor. The “chemical” forces causing the chemisorption are greater than the van der Waals forces involved in physical adsorption, having values of the order of 1 to 3 e.v. Furthermore, the chemisorption is characterized by a n activation energy. Figure 1 shows schematically a n

218

K. HAUFFE

energy diagram for the two kinds of adsorption, as suggested by LennardJones (15). A precise definition of chemisorption for all the observed adsorption systems is a t present difficult, because the concept of chemical forces is by no means clear. Furthermore, it is difficult t o decide how far the influence of strong dipole forces enters into the general phenomenon of chemisorption. Furthermore, the requirement of a n activation energy cannot be considered as a positive criterion for chemisorption processes. Engell and Hauffe (16-18) have limited their theoretical considerations of the chemisorption, on a solid, t o particles in the form of ions (ionosorption), and similarly this review will deal with this type of chemisorption. I n the case of an n-type oxide, e.g. ZnO, CdO, and TiOz, the concentration of quasi-free electrons in the region near the surface of the oxide will be decreased when electrons are transferred from the oxide t o the chemisorbed oxygen. The concentration of free electrons a t a sufficiently great depth within the oxide will remain unchanged. With p-type oxides, e.g. CuzO, NiO, and COO, the concentration of electron holes will be increased in the region near the surface, with increasing chemisorption of the electronegative gas. Inverse effects have t o be expected for gases such as CO or HzO, which transfer a n electron t o the semiconductor during the chemisorption. I n any particular case, the direction of the electron transfer depends upon the relative position of the two electron levels or Fermi potentials in the semiconductor and in the chemisorbing gas, respectively. I n the case of low concentrations of electron defects in the catalyst, one may tentatively apply the laws of ideal solutions and assume the Fermi potential to be equal t o the chemical potential of the electrons. At low concentrations, Boltzmann statistics would be applicable. There will be an exchange of electrons between the solid and the chemisorbed gas as long as the electrochemical potential of electrons in the solid, ~l;(~), differs from the electrochemical potential of electrons of the chemisorbed surface phase, 7 P .Generally, a t equilibrium, .we obtain Ili(H)

pi(H)

+ zicp(H)

=

.qi(o)

pi(C)

+

zicp(U)

(3)

where cp denotes electric potentials, p chemical potentials, and zi the valency of the electron. The index (H) symbolizes the interior of the solid, and (u) the chemisorption layer as surface phase. The flow of electrons from the semiconductor into the chemisorbed layer, and vice versa, without any diffusion of ionic species a t the same time, induces a space charge between the interior of the semiconductor and its surface. This space charge will exist only for a small depth into the solid near the surface. The physical and chemical behavior of this

APPLICATION O F THE T H E O R Y O F SEMICONDUCTORS

219

boundary layer of a chemisorption system is quite similar to the boundary layer of crystal rectifiers, which was developed by Schottky (19), Mott (20), and Davidov (21). Therefore, one is obviously led to applying the crystal-rectifier theory of electronic equilibrium between a metal and a semiconductor to the system of a semiconductor and its adjacent chemisorption layer. Following this point of view, the equilibrium chemisorption process has been treated theoretically by Aigrain and Dugas (22), Hauffe and Engell (16), and Weisz (23). Theoretical consequences conperning the rate processes in chemisorption and catalysis were explored by Engell, Hauffe (17), and Weisz (24). I n the subsequent section we shall consider the mechanism of the formation of a boundary layer caused by chemisorption, following the treatment by Engell and Hauffe (17).

2. The Boundary Layer Theory of Chemisorption It has often been pointed out that the electrical conductivity of sintered samples of ZnO and of other n-conducting oxides is frequently caused by the conductivity of thin layers near the surface, and not by the conductivity of the bulk (25-28). According to our present knowledge, these thin layers near the surface of oxides are caused by electron transfer from the layers to the chemisorbate during the chemisorption, and the amount of chemisorption may be related to the electronic properties of the gas molecules and of the solids. The dependence of the electrical conductivity of some semiconductors on the pressure of CO, COZ, and on the vapor pressure of ethanol, methanol, acetone, and water, as observed by Ljaschenko and Stepko (29), can be explained by the same mechanism. The dependence of conductivity of some mixed oxides at high temperatures can be explained in a similar way (30). The possible formation of boundary layers by very high energy levels at the surface has been discussed too (31). Tamm (32) was the first to point to the possible importance of such special electron levels in the surface, caused by the difference between the binding forces at the surface and in the bulk of solid (((Tammstates”). The experimental evidence for the existence of these surface states, proposed by some authors, is not convincing, since these authors exposed the solids investigated t o air before (33) or during their measurements (34). During all these experiments, chemisorbed oxygen was present in the surfaces, with the result that the concentration of the electrons and holes in the surface was considerably changed by this gas. Experiments carried out with crystal surfaces produced in vacuo have given no evidence for the existence of Tamm states (35). For this reason we shall not consider these hypothetical states in the following discussion. According to our definition of chemisorption as equivalent to (‘iono-

220

K. HAUFFE

sorption,” the accumulation of electrons in the chemisorbed oxygen layer will be denoted in the following text by a minus sign, even though, conceivably, the electrons are not completely transferred to the oxygen atoms. A certain distribution of free electrons or of electron holes in the boundary layer will be produced for the steady state: The diffusion current of electrons or of holes due t o the concentration gradient will be just balanced by a current in the opposite direction due to the gradient of the resulting electric field. This concept of chemisorption is represented schematically in Fig. 2. I n mathematical terms, it is described by the

‘I*

71

+

b FIG.2. Diagram for the chemical potential p, electrochemical potential 7, and electrical potential ‘p, in the interior ( H ) and in the boundary layer ( R ) of a n n- and p-conducting oxide due to chemisorption of oxygen, according t o Engell and Hauffe. This chemisorption consumes electrons.

following formulae valid a t low enough temperatures at which the concentration of bulk defects can be considered frozen. Under these assumptions we can write > $ 0 2 ( u ) + @(a) O-(e) (e.g. on ZnO) (44 $@z‘u’ O-(d + $ ( a ) (e.g. on NiO) (4b)

*

Here e(R) and da)are the concentrations of free electrons and of electron holes in the boundary layer ( R ) ,respectively, and the index (v) marks the chemisorbed surface phase. Whereas a t certain temperatures we may have mainly 0- ions chemisorbed, the process

APPLICATION O F T H E T H E O R Y OF SEMICONDUCTORS

221

may gradually become predominant with increasing temperature. The equilibrium condition of the chemisorption is fixed by Equation (3) and, generally, we can write qe(R)

= qo(H);

qe(R)

=

(5)

,,@(H)

The distribution of positive and negative charges with depth 4 into the boundary layer is given by the gradients of the electrical and chemical potentials :

I n agreement with this and according to Fig. 2, the concentration of free electrons ne in the boundary layer will be decreased by the chemisorption of oxygen on n-type solids. Likewise, in the case of p-type solids, the concentration of electron holes n@ in the boundary layer will be increased. Accordingly, we shall distinguish the boundary layers as “exhaustion boundary layer ” (Verarmungsrandschicht) and ((inundation boundary layer ” (Anreicherungsrandschicht), respectively. If the concentration of electrons or of holes is not too great, we can use the Boltzmann equation, and we obtain for the distribution of electrons and holes ng(R)= ne(H)exp ( - e V D / k T ) n@(R)= exp ( +eVo/kT)

(74 (7b)

Here V Dis the potential difference between the interior of the solid and the surface ( 4 = 0). By V(4)we shall denote the potential difference between the interior of solid and any point 6 in the boundary layer. In the case of consumption in the solid of electrons captured by the chemisorbed oxygen on the surface, we obtain, for the electrochemical equilibrium, from Equations (4a) and (4b) and by using the general formula of electrochemical thermodynamics

zqi = 0 = and zq’I . --

0

=

- 1,2po*(Q) /

+

+ vo

(84

+ p@(R)+

+V D

(8b)

-

-~/.kl~(’)

With regard to pe = - p a , these two formulae are identical, except that in (8a) has been replaced by --p@ in (8b). Both are always valid. If we use the concentrations, instead of the chemical potentials, and write the basic potentials pi(i) ( j = any phase index), as pe

K. HAUFFE

222

and consider t hat the sum of all the p's will equal log K , then from (71, (8), and (9) we obtain ne(H) (104 exp ( e V d k T ) = K I P O , ~ ~ (%-type) and exp ( e V D l w = KZP0,S

1

(P-tYPe)

n@(H)no-(u)

(lob)

_-_--1 I

I

I

I

I I

I

a

b

FIG. 3. Scheme of the distribution of the concentration of free electrons ne, interstitial metal ions TIM,O' (Fig. 3a), electron holes T I @ and metal ion vacancies nuen' (Fig. 3b), in the boundary layer and in the interior of n- and p-conducting solids respectively, according to Engell and Hauffe. Here E is the critical thickness of the boundary layer.

where no-(") is the surface concentration of the 0- ions in the a-phase. A further relation between V D and no-(") can be obtained by the Poisson equation (11) Here & is the electric field strength, e the dielectric constant, and p the density of the space charge. Assuming n@CR) >> n @ ( H )the , correlation between the density of the space charge and the concentration of holes can be expressed in the case of an inundation boundary layer by p([) =

ene(6)

for0


(12)

where I represents the maximum thickness of the boundary layer (see Fig. 3). Combining (12), (11), and (6b) we get the following differential equation for the distribution of electron holes in the boundary layer

APPLICATION O F T H E THEORY OF SEMICONDUCTORS

223

A similar expression is t o be expected for the distribution of electrons in exhaustion boundary layers. The evaluation of these equations is not easy. Therefore, it is desirable to simplify our assumptions concerning the distribution of the electrons and holes in the boundary layer. This simplification is illustrated by Fig. 3. With a suitable choice of the thickness 1 of the boundary layer, the simplification will satisfactorily approximate the real case. The distribution of space charge in a n inundation boundary layer can be similarly calculated, and will be shown below in parentheses. The simplification applies for eVD>> kT,in which case n e ( H> )> ne(R) (or n@CR) >> necH)); i.e. the concentration of free electrons in the boundary layer, nO(R), is negligibly small in the case of a n n-type oxide with a n exhaustion boundary layer (Fig. 3a) (or the concentration of electron holes in the interior of the crystals, n@cH), is small in the case of a p-type oxide with a n inundation boundary layer Fig. 3b). We thus have for (12) p = p =

for 0

< 5: < 1

p = o

for E

>1

(1 3 4 (13b)

and Since, in the first approximation, every oxygen atom captures one electron, the number of surface charges equals the total number of ionized defects in the boundary layer having a density equal t o ne(H) (or the number of additional electron holes in the boundary layer produced b y chemisorption) ; the surface concentration of chemisorbed oxygen atoms (gram atoms per square centimeter) nCu) is therefore

Introducing (13) and (14) into the integrated Poisson equation (111, we obtain for the diffusion potential V Dafter integration between f = 0 and .$=l

I n (15a) V Dmay be calculated immediately. However, it is not possible to solve Equation (15b) directly. We have t o eliminate in (15b) by means of (7b). Equation (15b) then takes the form

224

K. HAUFFE

As a first approximation, let V be a constant, with the dimension of an electrical potential, and with the additional assumption V zz 1. and n e ( H )are known from Hall-effect measurements, the If important diffusion potential V Dcan be calculated from (15a) and (15c), since the amount of chemisorption is easily determined. The quantity of oxygen chemisorbed on the surface of a n n-conducting oxide a t constant temperature and oxygen pressure is obtained from (14a) and (15a)

According t o (IB), the quantity of chemisorbed oxygen is proportional t o the square root of the concentration of the free electron in the interior of the solid. Later we shall see that V Ditself is indeed This logarithmic dependence, however, can be dependent on nG neglected in the first approximation. (Similarly, from (15b) we could obtain an expression equivalent to (16) for the quantity of oxygen chemisorbed a t the surface of a p-conducting oxide.) It is interesting to estimate the maximum number of atoms which may be chemisorbed by an electron transfer process, in terms of the fraction of surface sites covered, 0,.,, and of the relative concentration of free electrons t o the total number of atomic sites, n, on a szmiconductor. Following the treatment of Weisz (24) we obtain, with 4.6 A2 as a typical size of a surface “site”

em,,= 0.15(eVon)’

( W

e = 10, and V oequals approximately unity (e.v.). where n s ne(H)/1023, Even if we assume the absurdly high value of n = 1, this does not allow, according t o (16a), to attain complete covering of the surface with the chemisorbed gas. Thus, equations (16) or (16a) indicate that only a small fraction of the surface is available for chemisorption, for this is the chemical potential of the electrons in the boundary layer. For example, for a concentration of n = a dielectric constant c = 5, and V o = 1 e.v., Weisz obtains em,,= 0.4 per cent. After this preliminary discussion, we shall now consider the adsorption isotherms in the case of chemisorption involving boundary-layer effects. At equilibrium, the number of oxygen atoms being chemisorbed per unit time and area equals the number of oxygen atoms desorbed per unit time and area. According to (4a) we obtain

APPLICATION O F THE THEORY OF SEMICONDUCTORS

225

where k, and kd are rate constants of the chemisorption and desorption, respectively. Combining with (7a), we have

By means of (15a) we eliminate Vn and obtain with lc,/kd of the law of mass action)

=

K (constant

Because of the strong dependence of such an expression on the exponential, Equation (19) can be transformed, by means of an approximation similar t o that used for (15c). Accordingly, we have

and therefore

ncO)= const. { In p o t )jP

for ne(H)and T = constant

(20a)

This form of a n isotherm is characteristic for a chemisorption process on an n-type catalyst forming a boundary layer with a space charge. It is evident that this mechanism is quite different from the mechanism of physical or van der Waals adsorption, which is represented by the Langmuir equation. We must now answer the question of what experimental methods t o use t o investigate cases of chemisorption involving boundary layers. This is possible by means of suitable electric methods. According t o measurements of the electrical conductivity and of the Hall effect of polycrystalline ZnO samples by Anderson (36), Hahn (27), Miller (26), and Volger (37), the electrical conductivity is primarily determined by the boundary layers. Chemisorption of water vapor on single crystals of CuzO at room temperature decreases the electrical conductivity, as observed by Brauer (38). This is readily understood, if one considers that water vapor will be chemisorbed with transfer of electrons to the oxide

HzOw)+

@(W

2 HzO+(")

(21)

Thereby, the concentration of the electron holes in the boundary layer of CuzO will be decreased. The chemisorption of HzO destroys the inundation boundary layer of CuzO, produced by the preceding chemisorption of oxygen originally present after contact with air 02(d+ 0-(d + @ ( R ) (22)

226

K. HAUFFE

The remaining concentration of electron holes and, therefore, the electrical conductivity are functions of the water vapor pressure. This function can be derived by applying the formula (15a) of the exhaustion for ne(*). boundary layer. There we have, however, to substitute n@(*) Using the law of mass action of ( 2 1 ) , we obtain for the diffusion potential

Considering that the total conductivity of the CuzO crystals consists of the constant bulk conductivity U H and the boundary layer conductivity 3.0 M

.= 5h

2.8

5

2.6

i!

a

C .-

.-.-

-

$

2.4

2.2

0

2.0 I

I

.LO

0.5

1.5

FIG.4. Dependence of the boundary-layer conductivity K R of Cut0 on the oxygen pressure, as measured by Brauer and evaluated by Engell.

which is only dependent on the water vapor pressure, we get for the total conductivity utot= U H U R

uR,

Utah= K ( H )

+ B - (D - 1) + L

K(R)

B L

for pHnO and T = const.

) K ( * ) is the specific electrical conductivity of the boundary where K ( ~ and layer and the bulk, respectively, L the length, B the breadth, and D the

thickness of the crystal. Under the conditions of I of exp (-eVu/kT)

<< 1, we obtain with K

N

=

ns)

(&

t6

and

n for utOt:

where A , B, and C contain the constant terms. Figure 4 shows a comparison of Brauer's experimental results with relation (24).

APPLICATION O F THE THEORY O F SEMICONDUCTORS

227

Ljaschenko and Stepko have studied the decrease of the electrical conductivity of very thin Cu20 films after these films had chemisorbed methyl alcohol, ethyl alcohol, acetone, and water vapor. Engell (18) has explained this decrease of conductivity by extending the explanation given above to the chemisorption on thin films whose total thickness is less than the thickness of the boundary layer. If K ( ~ is ) the conductivity before the chemisorption of any of the vapors listed above, K ( ~ the ) mean longitudinal conductivity after the chemisorption has taken place, and K ( ~ )- K ( " ) = AK, then AK is proportional to the number of the electron holes which have been filled, and is therefore also proportional t o the concentration of the chemisorbed particles ncU)

Here R is the resistance of the CuzO film without chemisorption, AR the increase of the resistance, identical t o AK, and n@cH) the concentration of the holes in the pure film of thickness D. << n a ( H )for , the formation of Since our previous assumption, exhaustion boundary layers is not applicable t o this case because A R / R = 1, the evaluation of n as a function of n @ ( Hand ) p,,,, requires solution of the differential equation

This equation, is not easily solved. Therefore, Engell used an approximation method, assuming that the concentration of electron holes n @ ( T ) in the boundary of the semiconductor and glass (the supporting medium) is proportional t o the number of electron holes filled up during chemisorption. We obtain thus for

Then, in analogy to (4a) we obtain

ncU)= Jc2pg,,n@(T)exp ( - eVD/kT) or,

2ae Since V D= - ncU)D,we obtain from Equation (25)

228 Here

K. HAUFFE

2ae2 ~

E

ngj(H)D2is the maximum value of V,, i.e. the value for pgse-+ w ,

or the value for complete annihilation of the electron holes in the boundary layer. Denoting this value by VJiand D / k by p*, we finally obtain

Figure 5 represents the values measured by Ljaschenko and Stepko, and the theoretical values according to Equation ( 2 8 ) . The calculated values

6

4

=a 61a

-c

2

0 0

0.2

0.6 I+&

04

I

0.8

1.0

AR

FIG.5. Dependence of the conductivity of thin CuzO films upon the chemisorption of some vapors, as measured h y Ljaschenko and Stepko, and evaluated by Engell. Here p is the vapor pressure in mni. Hg, R the resistance of the pure film, and AR the increase of the resistance due to chemisorption.

of V M are marked at the curves. The calculation can be refined by using a boundary layer with electron defects reserve (Randschicht mit Storstellen- Reserve). Likewise Antipiria and Frost (38) observed considerable chemisorption of ethyl alcohol, ethyl ether and water between 150" and 310°C on A1203 previously fired at 550" in air. The experimental results could not be represented by the Langmuir isotherm, but followed a relation of the form

l/n(")

-

l/pgas

Noteworthy in this connection are Muscheid's (39) results on the dependence of the electrical conductivity of single crystals of CdS upon their treatment in vacuo or in oxygen a t various pressures. This author observed a strong and sudden decrease of the conductivity when oxygen a t

APPLICATION O F THE THEORY O F SEMICONDUCTORS

229

20°C contacted the CdS crystal (n-conductor), which had been a good conductor in high vacuum. This decrease of the conductivity of CdS crystals is independent of the oxygen pressure within a large range; oxygen of one mm. Hg pressure causes the same effect as oxygen of 600 mm. Hg pressure. Here, the basic effect is apparently also the annihilation of the inundation boundary layer by the chemisorption of oxygen. Furthermore, it has been observed that the chemisorption is reversible up to 350°C. At higher temperatures, there are irreversible changes which can be attributed to a chemical reaction of the oxygen extending into the CdS lattice.

MIn ut08

FIG. 6. Resistance R and contact potential AEK us. time, according t o Morrison At time 0 the oxygen is dry and becoming saturated with water vapor at 12 minutes. Then the ambient is changed back to dry oxygen. ( R is in arbitrary units and A E , is in volts.)

Stimulated by the investigations of surface states on germanium crystals by Brattain and Bardeen (34), Morrison (39a) studied the dependence, on the surrounding gas atmosphere, of the electrical conductivity and the contact potential of p- and n-conducting Ge samples during illumination at room temperature. Oxygen and mixtures of oxygen and water vapour were used as gas atmospheres. As shown in Fig. 6, the resistance of p-conducting germanium in oxygen is lower than in HzO vapor, whereas n-conducting germanium shows an inverse effect. The measurements of contact potentials AEK also confirm these measurements. These results can only be understood by assuming that the chemisorption of oxygen takes place by consuming electrons, whereas the water vapor is chemisorbed by delivering electrons. From the time-resistance curves during chemisorption, we see that the desorption of water vapour on p-conducting germanium occurs slower than on n-conducting germanium or, in other words, that the oxygen can exchange chemisorbed HtO faster on n-conducting Ge than on p-conducting Ge. These results are in agreement with the above explana-

230

K. HAUFFE

tion, and represent a good example of a chemisorption of a mixture of a n electron-consuming and an electron-delivering gas. The formation of boundary layers a t the surface interface between semiconductor and gas influences also the luminescence and the electrooptical qualities of semiconductors. These effects offer interesting possibilities for studying experimentally the mechanism of chemisorption, the stationary state of chemisorption, and electron defects in the catalyst during cataIysis. Experiments along this line have been carried out by some investigators (40,41) who have studied in a qualitative way the factors influencing the oxidation of phenols catalyzed by zinc oxide under the influence of light. Further work on this subject is desirable. 3, The Work Function of Electrons and the Kinetics of Chemisorption

I n the preceding chapters the conditions determining chemisorption and the thermodynamic treatment of surface equilibria have been discussed. We shall now derive a general formula for the dependence of the work function of a semiconductor (e.g. of a n oxide) upon the surface concentration of the chemisorbed oxygen ions. I n the preceding chapter we pointed t o electrical conductivity as one of the physical properties of semiconductors which is changed by a chemisorption process and is accessable t o measurement. A further possibility for investigating the mechanism of chemisorption is the relation between the work function and the external electric field of the semiconductor as influenced by chemisorption. These effects have been used for the interpretation of the mechanism of chemisorption and heterogeneous catalysis by Suhrmann (42), and have been experimentally demonstrated in chemisorption processes by Lj aschenko and Stepko. These effects shall here be correlated with our concept of the boundary layer formed in the presence of oxygen and hydrogen. I n the case of an n-type oxide the work involved in transferring an electron from the surface into the free space can be subdivided into three contributions: 1. One contribution is independent of the concentration of the electron defects, containing merely the “chemical” portion of the work t o transfer one electron t o infinity. We shall call this E L . This term includes also the image-force effect. 2. Another part of the work expresses the dependence of the “chemical” work on the concentration of the electron defects; it is given for an n-conducting oxide by

APPLICATION OF T H E THEORY O F SEMICONDUCTORS

23 1

where neo represents the concentration of free electrons when the necessary chemical work is just EL. 3. Finally, we must consider the contribution of the electrostatic work required t o transfer one electron into free space. After overcoming the short range chemical forces, the electron must be moved a certain distance against the electric field in the surface. Under the assumption th at the lines of force of the electric field are located between the ion defects in the boundary layer and the surface charges represented by the chemisorbed gas atom, we obtain the expression u& for this electrostatic work term. & is the boundary field strength represented in Equation ( l l ) , and u is the distance between the surface of the oxide and the centers of charge of the chemisorbed atoms in the a-phase. Assuming that the boundary field strength will be changed only by the concentration of the chernisorbed particles, and that the structural and electrostatic terms will be practically uninfluenced, we get for the difference between the work functions for a free surface, and for one occupied by chemisorbed particles, #oceu,,ied, the expression

Using (7a), (11), and (15a) we have

The corresponding formula for a p-type oxide can be derived exactly in the same manner, in considering th at the number of electrons removed from the surface of a p-type oxide is identical to the production of electron holes in the surface. Then, applying (7b), (11), and (15b) it follows

with V = Vu as a constant. If we first consider only those chemisorption processes in which a n electron transfer takes place from the semiconductor t o the chemisorbing gas, we can summarize the result of these calculations as follows: The value of the work function must increase if a chemisorption takes place with the consumption of electrons b y the chemisorbed gas. The increase of the work function can be expressed in the case of a n n-conducting adsorbent by a quadratic, and in the case of a p-conducting adsorbent by a combined linear-logarithmic dependence on the surface concentration of the

232

K . HAUFFE

chemisorbed gas. Inverse relations will be obtained for a chemisorption with electron transfer from the gas into the semiconductor, e.g. for H2 chemisorption on ZnO or NiO. At present, no quantitative experimental results are available. Qualitatively, the postulated correlations are confirmed by the investigations of Ljaschenko and Stepko. We can use the relations derived in (30) and (31) unchanged, if A$ is t o represent the Volta potential difference between a free surface and a similar one occupied by a concentration of chemisorbed gas equal t o n("). Equations (30) and (31) could be checked by measurements of the kinetics of the chemisorption of oxygen and hydrogen on oxides. However, we shall first apply these correlations t o the kinetics of the chemisorption of oxygen on NiO, and check the results obtained with the experimental observations made by severaI authors. Assuming t hat the rate-determining step is the electron transfer between the adsorbent and the adsorbate, the work function will be increased with increasing coverage of the surface, since the energy barrier, t o be overcome by the electrons moving from the solids t o the adsorbate, will be increased with increasing coverage of the surface. If Uo is the height of the energy barrier a t the beginning of chemisorption (n'") = 0), we obtain for the rate of chemisorption

The frequency factor of the chemisorption, which is proportional to the fraction of the electrons that are in a state of transition from a lower t o a higher energy level, is included in k , which is approximately a constant, at least for the start of chemisorption. At a sufficient distance from the equilibrium, where the reverse reaction can be neglected, we obtain for the kinetics of chemisorption of oxygen on NiO

Here, again, the exponential term is of primary importance in determining the rate of chemisorption and the amount of the surface concentration of the chemisorbed atoms. Considering the other terms t o be constant, therefore, integration gives

Such a n adsorption law may be experimentally tested. I n the case of

APPLICATION OF T H E T H E O R Y O F SEMICONDUCTORS

233

diatomic gases, (e.g., 0 2 and Hz) the chemisorption takes place with dissociation of the molecules into atoms; then the surface concentration n(") [particles per ~ m . from ~ ] measured chemisorbed volume V of gas [ ~ mon . ~ surface of F ~ m . will ~ ] be obtained by the formula

Thus the change with time of the chemisorbed gas volume is obtained as

( t + t o ) min.

FIG.7a. Rate of chemisorption of oxygen on nickel oxide at 25°C and at various oxygen partial pressures, according to Engell and Hauffe. The chemisorbed volume in t o ) , where t is the time of chemisorption cubic centimeters is plotted against log ( t in minutes, and t o a constant.

+

According t o this equation, the chemisorbed gas volume is a linear function of the logarithm of ( t to).I n Fig. 7a the chemisorption of oxygen on NiO is plotted against time of contact for various oxygen pressures at 25". The constant to was determined as ca. 1 minute. Figure 7b represents the rate of chemisorption, on N O , of oxygen a t 60 mm. pressure and a t various temperatures. Any of these curves consist of two rectilinear branches, of which the first is interpreted as representing the chemisorption process, and the second the formation of new oxide molecules on the surface (chemical reactions, occurring by the same mechanism as the formation of thin oxide films on metals (43,44)). Elovich and Zhabrova (45) have found a relation similar t o (35), without connecting i t with any such specific mechanism. If one evaluates the experimental results of Taylor and Strother (46) on the chemisorption of hydrogen by ZnO by means of Equation (35) (volume of HZagainst

+

234

K. HAUFFE

log time), one also finds two rectilinear curves with different slopes. This anomaly had already been found earlier by Sickman and Taylor (47). More investigations about this phenomenon, occurring during the chemisorption of hydrogen on ZnO, were carried out later by Taylor and Liang (48). The rate of chemisorption was measured by suddenly increasing the temperature and observing the rate of change of pressure. Mixed oxides of ZnO and Crz03 show a similar behavior toward the chemisorption of hydrogen (49). Taylor and co-workers assume that these effects are caused by the heterogeneity of the surface. This, however, is not necessarily the

FIG.7b. Rate of chemisorption of oxygen on nickel oxide a t a pressure of 60 mm. Hg and a t three different temperatures, according to Engell and Hauffe.

only explanation, as has been shown above. On the contrary, Gray (50) and Engell and Hauffe (22) have shown that the incorporation of chemisorbed atoms and the reduction of the lattice by a consumption of lattice ions can also play an important role in the chemisorption process. The rate of the chemisorption of hydrogen on ZnO a t 184" can be presented by means of an equation resembling Equation (35). The chemisorption of hydrogen on the n-conducting zinc oxide

%H2(o)

H+b) +

$(R)

(36)

involves an increase in the number of free electrons in the boundary layer, as was explained for the chemisorption of oxygen on nickel oxide. Therefore, Equation (35) is applicable. Likewise, Taylor and Thon (51) could demonstrate the validity of the logarithmic relation (35) for the chemisorption of hydrogen on Crz03at 184°C (52) and on 2Mn0.Crz03 at 100°C (53) measured by Taylor and co-workers (52,53) (see Figs. 8 and 9). The hydrogen-deuterium exchange reaction between 140" and 491"K, Hz Dz = 2HD, which will be described in more detail in the next chapter, is caused in the higher temperature range by a predominating

+

APPLICATION O F THE THEORY O F SEMICONDUCTORS

1.0

1.4

1.8

235

2.2

log ( t i t , ) at a pressure of 0.5 atm. and FIG.8. Rate of chemisorption of hydrogen on at 184"C, as measured by Burwell and Taylor and evaluated by Taylor and Thon.

log i t * to) FIG.9. Rate of chemisorption of hydrogen on 2Mn0.CrzOsa t a pressure of 1 atm. and at lOO"C, as measured by Williamson and Taylor and evaluated by Taylor and Thon.

chemisorption of hydrogen. At lower temperatures, Smith and Taylor (54) observed a preferred physical adsorption of deuterium. By applying the homogeneous disorder theory without the consideration of boundary layers, as discussed above, Volkenshtein (55) assumes that new "active centers" will be produced by the chemisorbed gas. No

236

K. HAUFFE

conclusive evidence for such an effect seems to be a t hand. Similarly, there appears to be no need for the proposals of Taylor and Thon (51) elucidating the kinetics of chemisorption by means of the “theory of dublons,” further developed by Volkenshtein and Bonch-Bruevich (56), or the particular explanation of the mechanism of the decay of luminescence (57). After having described the mechanism of electron interchange between the catalyst and the reactive gas molecules, we will show in the following chapters how the electron disorder and the space-charge effects are expected to influence the rate of reaction. The electron interchange between the catalyst and the different species of molecules involves a different mechanism of potentially higher reaction rates than is the case in the homogeneous gas phase reactions.

111. THEMECHANISM OF SIMPLE REACTIONS ON THE SURFACE OF SEMICONDUCTING SOLIDS As has been explained in the first chapter, the catalytical properties of a catalyst are frequently based on its ability to interchange electrons with one or several species of the gas molecules participating in the catalytic reaction. The geometrical arrangement of the atoms in the surface of the catalysts (“roughness of the surface”) contributes also to the catalytical activity, because the number of the places at which electron interchanges can take place will be increased. This contribution to the catalytical activity, however, is of secondary importance. This situation is demonstrated by the heterogeneous catalytical decomposition of NzO on various oxides, carried out by Schwab and co-workers (58,59). The catalytical activities of several oxides, used as catalysts of the NzO decomposition decrease in the following sequence CuO >> MgO

> AlzO, > ZnO > TiOz > Crz03 > Fez03

(I)

Cupric oxide shows by far the best catalytical activity of all the oxides with regard t o the decomposition of NzO. The surfaces of Fez03or Crz03 catalysts for the decomposition of NzO can have the highest possible roughness without approaching the outstanding catalytical activity of smooth surfaces of CuO, NiO, or COO. From this point of view Schenck (60) and Huttig (61) have emphasized the important role of the (‘chemical affinity’’ of the gases to the catalyst. The postulated properties of a catalyst are manifold. It has to chemisorb one or several species of the reactive gases without, however, forming too strong a bond with the reactants, because the desorption processes would then be more difficult and the electron interchange in the next step, e.g., during the formation of the end-product molecules, would then become too slow. On the contrary, the surface of a good catalyst has to

APPLICATION O F THE THEORY OF SEMICONDUCTORS

237

allow easy desorption of the reaction products and continually has to restore the ' I electronic state of disorder." Furthermore, it is necessary that large molecules come in good contact with the energetically favorable points of the surface (steric factor) (62). As we have pointed out in the preceding chapters, the chemisorption and desorption processes will be decisively influenced by the nature and extent of the electron disorder and by the space-charge or boundary-layer effects caused by the interchanges of electrons. Therefore, it is desirable t o investigate the connection between the electronic structure of catalysts and their catalytic activity as observed for some relatively simple gas reactions. 1. The Mechanism of the Decomposition of NzO o n Semiconducting Oxides

Let us consider the mechanism of the decomposition of NzO on p - and n-conducting oxides. The first step of the decomposition of NzO is always the chemisorption of N20, or of oxygen derived from NzO. I n the case of a p-type oxide (e.g., NiO), the number of electron holes in the boundary layer will be increased as a result of the chemisorption of NzO as follows

or I n the case of an n-conducting oxide (e.g., ZnO), the number of free electrons in the boundary layer will be decreased as follows

NzO(s) + or

NzOb) +

NzO-(u) e(R)

-.j

O-(uj

+ N2(8)

(384 (38b)

The second step will be a desorption of the chemisorbed oxygen 2@(R)+ 2 0 - ( u , 2 OZ(d or 20-(")

* Oz(0)+ 2 dR)

(p-type conductor)

(39a)

(n-type conductor)

(39b)

or, possibly, a reaction between the chemisorbed oxygen atoms and the NzO molecules striking the surface, according to e ( R ) + O-(u) + N2O(8)--+ N Z ( g )+ 0 2 ( 8 ) (p-type conductor) (404 or O-(U) NzO(g)3 N2(g) O z ( p ) @(R) (n-type conductor) (40b)

+

+

+

By means of measurements of the electrical conductivity, carried out during the catalytic decomposition of gaseous nitrous oxide in the pres-

238

K . HAUFFE

ence of oxygen or in an atmosphere of oxygen and nitrogen of the same partial pressure of oxygen, one observes for a p-conducting oxide (e.g., NiO), an increase of its conductivity (3) and for a n n-conducting oxide (e.g., ZnO), a decrease of its conductivity (4). These results indicate a n increase of the amounts of chemisorbed oxygen in both cases. Other experimental results, such as the change of the rate of decomposition with changing concentration of the electron defects produced by the incorporation of foreign ions (14a), make it likely that one of the desorption processes, (39) or (40), is the rate determining step. It is evident th a t only such oxides which accelerate step (39) or (40) will effectively catalyze the decomposition of N20. This acceleration, however, will only be achieved if a lower energy level is in the catalyst than in the chemisorbate, and is offered t o the electrons produced by the desorption (39) or by the reaction (40), in view of the fact that this low energy level decreases the activation energy of the rate determining step (39) or (40). Solids offering such low energy levels are the p-conducting oxides (e.g., NiO, Cu20, and COO), whose electrical conductivities depend strongly on the oxygen pressure, i.e., p-type solids with a large capacity of chemisorption in the absence of appreciable ion-defect mobility a t sufficiently low temperature. On the other hand, the n-conducting oxides (e.g., ZnO, CdO, and TiO2),with and without added foreign oxides, only moderately catalyze the decomposition of N20, unless the foreign ions invert the electron disorder by changing the solid from an n-type conducting t o a p-type conducting system (chlorine treatment of oxide). This catalytic behavior of the n-type conductors is caused by the position of the Fermi level or of the chemical potential of the free electrons in the catalyst with respect t o the Fermi level in the chemisorbed species. These free electrons accelerate the already fast rate of chemisorption, but retard the slow process of desorption, which is rate determining. This conclusion is supported b y the previously quoted experimental results of Schwab and co-workers (58,59). All the oxides t o the right of CuO in series (I) above, are either n-type conductors or essentially insulators with a slight tendency to form electron defects, except for Cr203 (63),* which will be discussed later. CuO as a n intrinsic semiconductor (1) is the only oxide other than Cr203th a t has electron holes. A further confirmation of the validity of the assumption t ha t oxygen desorption is the rate-determining step in the NzO decomposition is the recent investigation of the N 2 0 decomposition on ZnO and on a catalyst consisting of ZnO plus 1 mole percent of GazO8, carried out by Wagner (4). I n accordance with our previous discussion,

* CrtOs is a n intrinsic semiconductor in which the electron holes are produced by an intrinsic ionic disorder (64).

APPLICATION O F T H E THEORY O F SEMICONDUCTORS

239

the increase of free electrons in ZnO produced by the gallium ions causes no appreciable increase of the rate of decomposition of NzO. The interpretation of Wagner's results, proposed by Boudart (30), seems hardly convincing. In this connection, a calculation of the energy involved in the electron transfer from the semiconductor t o the chemisorbate and vice versa is desirable. By the aid of the simple band model of semiconductors, Dowden ( 5 ) has tried t o calculate this energy and to give a physical interpretation. Although his explanation was not complimented by a refined interpretation in which the space charge effects were considered? as was done especially by Weisz (24) and Hauffe (17), Dowden's viewpoint was valuable in two respects. I n the first place it makes the importance of the

80

g60 C

'40 20 0 400

500

600

T in "C

700

800

FIG.10. Percentage decomposition of NzO on a nickel oxide catalyst, with various additions of LizO and InzOs, as a function of the temperature, according to Hauffe, Glang, and Engell. (Composition of the reacting gas: 15% N 2 0and 86% air; flow rate 1200 cm.s/hour.)

electron interchange during the catalysis easily understandable and, secondly, Dowden's simplified explanation has the merit of having pointed in a new direction? even though it requires modifications. What property changes would effect the catalytic activity of a catalyst for given reaction conditions of temperature and composition of the gas mixture? I n the case of the decomposition of NzO, this question was answered and experimentally verified by Hauff e and co-workers (14a). As shown in Fig. 10, the catalytic activity of NiO is considerably superior to the activity of CuO, in agreement with theory. Furthermore, the experimental results of Schmid and Keller (65) show that the p-type COOis the best catalyst for the decomposition of NzO. By the addition of between 0.1 and 0.3 mole percent of LizO, the concentration of electron holes will be increased (66) as follows

XOz(8)

+ Li20

2Lio'(Ni)

+ 2, + 2Ni0

(41)

240

K. HAUFFE

and is virtually equal t o the Li+-ion concentration in NiO.* I n this way, the catalytical activity of COO for the decomposition of NzO is raised. If, however, one increases the content of LizO in NiO u p to 3 or 5 mole percent, thereby further increasing the concentration of electron holes, the number of the lattice electrons or the chemical potential of the electrons becomes so low that the reaction proceeds by a different mechanism. As a result of the high concentration of electron holes the chemisorption (37) becomes so slow that it becomes rate determining, with a simultaneous decisive decrease of the catalytical activity of NiO (Fig. 10). These conclusions become more understandable, if we write th e chemisorption of NzO in the following way

NzO(8) NzO-(o)

+ $(It)

(step consuming electrons, slow)

(37a)

and

NzO-tc) -.+

+ N2(v)

0-f~)

(step without electron consumption, fast) (42)

According t o the experimental result that the rate of reaction is proportional t o pNSo,the electron consuming step (37a) is the slowest, and is therefore rate determining for the chemisorption. From (37a) we obtain for the rate of chemisorption, in the case of a p-type catalyst

and, in the case of an n-type catalyst

Here nNao-(u)is the surface concentration of the chemisorbed NzO molecules, and Y L @ ( and ~ ~ n e ( R )are the concentrations of the electron holes and quasi-free electrons in the boundary layer, respectively (14a). t Obviously, the second terms of (43) and (44) can be neglected in the first approximation, if in the first case n@(R) is not too great and in the second case n e ( R ) is not too small. The corresponding expressions for the rate-determining

* When lithium oxide is dissolved in nickel oxide, monovalent lithium ions replace nickel ions. We obtain, therefore, a n increase of the electron-hole concentration with increasing concentration of Li20. The prime represents the negative charge of this “impurity center” Li.’(Ni). t In addition to the electron defect concentration of thc catalyst the important role of the diffusion potential on the rate of reaction becomes evident, if one substiexp ( - e V D / k T ) in Equations exp ( e V n / k T ) and n e(X)= n tutes n e ( R ) = n (43) and-(46), respectively.

APPLICATION O F T H E THEORY O F SEMICONDUCTORS

24 1

desorption processes (40) are

and

Therefore, the rate of the desorption process must increase with increasing concentration of electron holes for a p-type catalyst, and must decrease with increasing concentration of free electrons for a n n-type catalyst. This is in accordance with the experiments illustrated by Fig. 10. Due t o an increasing concentration of electron holes in NiO, caused by increasing amounts of added LizO, the second term in (43) becomes continually larger, and the chemisorption continually slower, with the result t h a t finally the rate of decomposition is determined by the slow chemisorption of NzO on the catalyst. Using n-type catalysts (e.g., ZnO), we observe the opposite course. With increasing concentration of free electrons in ZnO, caused by the addition of increasing quantities of Ga203,the chemical potential of free electrons and the number of reaction centers on the surface both increase, Thus, the chemisorption is promoted. As shown in Table I, the most active catalyst has the lowest activation TABLE I Change of the Order of Reaction and of the Activation Energy of the Decomposition of NgO on NiO-LiIO Catalysts, According to Hau$e, Glang, and Engell

NiO NiO NiO NiO NiO

Catalyst

Approximate order of reaction

Activation energy (kcal./mole of NzO)

+ 0.1 mole percent of LizO

44

15.9 19.1 24.6 31.7 61.8

+ 0.5 mole percent of LizO + 1.0 mole percent of LizO + 3.0 mole percent of LizO

1

w

34 2

energy and the lowest order of reaction, a relation whose solution seems t o be desirable. The decrease of the catalytic activity of NiO by the addition of other oxides with cations of a valency higher than two (e.g., h 2 O 3and Cr203) is also understandable, if one considers that by such additions the concentration of the electron holes as well as the speed of the rate-determining desorption will be diminished. I n this connection, the poor catalytical activity of p-type Cr203 is

242

K. HAUFFE

surprising at first glance. Hauffe and Block (64), however, could show that the production of electron holes in Crz03involves a mechanism quite different from that operative in the case of NiO or CUZO.I n Cr203 the production of holes is not caused by an excess of oxygen in the oxide lattice but mainly by an intrinsic disorder mechanism as follows CrZ03undisLorted

=

Zero

Cro”’

+ Cro“ + @*

(47)

Therefore it is evident that Crz03 will catalyze neither step (37) nor steps (39a) or (40a). Likewise, the catalytical activity cannot be changed by additions of foreign oxides with cations of different valencies (see Fig. 11). In the case of CuO, the chemisorption is predominant at lower temperatures with formation of boundary layers, accompanied by polarization effects, especially at the start and during the steady state of chemisorption. At higher temperatures, however, we must expect a noticeable I00

80

re 60 c 0

40

20 0 400

500

600

T in ‘C

700

FIG.11. Percentage decomposition of NzO on ZnO and 0 2 0 3 with and without additions of foreign oxides, respectively, according to Hauffe, Glang and Engell.

incorporation of oxygen in the lattice of the catalyst. Due to cations migrating to the surface, the chemisorbed oxygen atoms can be incorporated into the lattice at the surface, e.g., by the process 0-olierniaorlmd

2

CUZOonthosurfaoe

+ 2Cun’ +

@

Investigations of the decomposition of NzO on an oxidized copper sheet covered with a thin CuzO layer, which were carried out by Dell, Stone, and Tiley (67)) indicate that the chemisorbed oxygen produced by the decomposition of N2O is partly consumed by a progressive oxidation of the copper sheet. Because a Cu2O surface with chemisorbed oxygen increases the rate of decomposition of N2O by about a factor of 10, compared with a free Cu20 surface, it is justifiable to conclude that the desorption (39a) can be neglected, and that only reaction (40a) is important for the mecha-

* The notation “Zero” refers to the perfect crystal with no electron holes, free electrons, or disordered ions.

APPLICATION O F T H E T H E O R Y O F SEMICONDUCTORS

243

nism (Fig. 12). The treatment of the CuzO catalyst with oxygen involves practically a similar increase of the concentration of electron holes in the boundary layer and, therefore, a decrease of the activation energy of the rate-determining step (40a), in a manner analogous to the incorporation of small quantities (<0.3 mole percent) of Liz0 into the NiO lattice. Treatment of the CuzO catalyst with water vapor should decrease the rate of decomposition, because the concentration of electron holes will be diminished, as shown in (21). The temperature of pretreatment is also important for the catalytic activity. Cremer and Marschall (68) showed,

Minutes

FIG.12. Dependence of the rate of decomposition of NaO on the pretreatment of the sur€ace of a CusO catalyst at 60°C, according to Dell, Stone, and Tiley. (I and I1 are on a n evacuated CurO surface, and I11 and IV are on a CuzO surface covered with chemisorbed oxygen.)

for nitrous oxide decomposition on CuO, that the activation energy increases almost linearly from 11 to 42 kcal. per mole with increasing temperature of the pretreatment of the catalyst. For the formaldehyde decomposition on a silver powder, Rienacker et al. (69) observed a similar dependence of the activation energy on the temperature of pretreatment of the catalyst. Recently Schwab (70) also emphasized the important role of electron defects for catalytic activities of solid oxides. Following this concept, he tried t o develop a mechanism for the decomposition of H2O2and for the oxidation of CO using inverse spinels as catalysts. Since the mechanism of the electron disorder and conductivity in spinels is much more complicated than in simple cubic oxides, it is not surprising that a satisfactory interpretation of these catalytic effects is still lacking. 2 . T h e Mechanism of Carbon Monoxide Oxidation on Solid Oxides

Garner and associates (71-73) investigated the oxidation of CO on crystalline CUZOand also measured the chemisorption of oxygen and

244

li. HAUFFE

carbon monoxide on the same Cu20 crystals. As expected, the chemisorpt)ion of oxygen on Cu20, observed by measurements of the electrical conductivity, is faster in the temperature range from 100 to 270°C than the oxidation of Cu20 to CuO. This confirms the assumption that the predominant result of this chemisorption of oxygen is the formation of a boundary layer. Furthermore, it follows from this observation that the chemisorbed oxygen is more reactive for the catalytic oxidation than oxygen incorporated into the Cu20 lattice. During the chemisorption of oxygen, the heat of chemisorption decreases from 61 kcal. per mole a t 5 % surface coverage to 55 kcal. per mole a t 50% of the maximum possible covering of the surface. The interpretation of this result does not necessitat e the assumption of a heterogeneity of the surface, as has often been

FIG.13. Dependence of the activation energy AH^ on the difference of the amount of chemisorbed oxygen, (no-(u))l+m - (no-(@),. Here (no-('))t+, is the surface concentration of chemisorbed oxygen in equilibrium, and (no-(u))tis the corresponding value a t time t.

proposed, including recently by Laidler (74) in his latest discussion of the mechanism of chemisorption. The boundary-layer theory of chemisorption provides that, during the chemisorption, the chemical potential of the electrons decreases, and therefore the activity energy of the electron transfer will necessarily increase (16, 17, 22-24). Applying Equation (16), we obtain for the heat of chemisorption a t the time t , A H , AH,

=

const ( (no-(u))l,, - (no-(u))l)

(48)

Here ( ~ t ~ ~ - ( ~ ) )ist . the , ~ equilibrium concentration of chemisorbed oxygen a t constant temperature and oxygen pressure. In Fig. 13 the decrease of the heat of chemisorption with increasing surface concentration of chemisorbed particles is schematically represented. Frankenburg (75) proposed a similar interpretation for the decrease of heats of chemisorption of hydrogen on tungsten. According t o further investigations of Stone and Tiley (73), the CO molecules react at room temperature with the oxygen atoms chemisorbed

APPLICATION OF T H E THEORY O F SEMICONDUCTORS

245

on a surface of CuzO with the formation of COz molecules. Furthermore, at 20"C, the chemisorption of CO is reversible and faster than the chemisorption of oxygen. I n any case, the chemisorbed CO can be completely desorbed by slow heating to 50°C. As a result of fast heating u p to lOO"C, the chemisorbed CO molecules react with the oxygen ions on the surface of the CuzO lattice forming COz molecules which escape into the gas space. No appreciable chemisorption of COz occurs on a CuzO surface which has been outgassed in a high vacuum. If, however, C o t is admitted t o a CUZOsurface on which oxygen has been chemisorbed, there occurs a n immediate chemisorption of COz. The heat of chemisorption of COZis found t o be reproducible, and amounts to 23 kcal. per mole. I n view of this, Garner et al. believe the following t o be the mechanism of this chemisorption COZ'B' + O-"' e COs-(u) (49) The chemisorption of COz can also be interpreted in a slightly different manner without the assumption of a COa--complex, if we consider the amphoteric behavior of COZ; i.e., chemisorbed COZ can be found as COZas well as in the form of a COz+-complex a t the surface. I n the case of chemisorption of COz on metal surfaces having a small work function (high-lying electronic energy levels), electron transfers in the direction of the chemisorbing carbon dioxide will predominate. On the other hand, in the case of chemisorption of COz on a p-type solid with a sufficiently low-lying chemical potential of the electrons, which will be additionally diminished in the surface regions b y a preceding chemisorption of oxygen, we get electron transfer in the opposite direction. Therefore, we picture the electronic mechanism accompanying chemisorption of COZ on CuzO surfaces partly covered by chemisorbed oxygen in the following way

Finally, the following surface reaction is possible O-(u)

*

+ COZ+(U)

(yJ3X'"'

(49c)

Contrary to the case of Equation (49), the chemisorbed COs-complexes are, in this case, electrically quasineutral. Depending on the position of the Fermi level, the COs surface-complex will have a more negative or positive overcharge. It will become a carbonate ion only when a cation migrates from the interior to the surface. Recent investigations into the chemisorption of COZon spinel, ZnO, and ZnO-CrzOa catalysts by Kwan

246

K. HAUFFE

and co-workers (76,77), confirm the above assumptions as applying t o these substances. I n a somewhat modified manner-but quite similar to Garner's scheme-we suggest a mechanism for the oxidation of CO on a CuzO surface, as follows

*

~ O z ( ~ O+) ) CO(8) ea(R)

+

+ ea(R) CO+(')

- 5 5 kcal./mole -20 kcal./mole

(50%) (50b)

I n contrast t o carbon dioxide, carbon monoxide is always chemisorbed with a n electron transfer from CO to the solid. Following this concept, we get the surface reactions

+ O+)

e COZX(0) (slow)* COZX(O) $ COP) (fast) Since the first of these steps is slow, we can write CO+(S) + 0 - ( u ) $ COZ'd CO+(r)

(50c) (50d) (5Oe)

Further, we must also consider the following reactions CO(U)+ O-(S) + @ ( R ) $ C02(6)) (50f) and >$Oz(8) + CO+(u) COz(o) + $(R) (50g) At room temperature Garner's COs-complex mechanism is probable. The position of the Fermi level in the boundary layer determines whether the mechanism of the chemisorption of COz follows Equation (4%) or (49). The kinetics observed a t 20°C indicate that the rate of oxidation of CO is proportional to the oxygen pressure and independent of the CO pressure. Accordingly, the chemisorption of CO is preferred to, and faster than, the chemisorption of oxygen. Therefore, either step (50f) or (50g) is rate determining. At higher temperatures, e.g. 725"C, however, we obtain another rate law and another mechanism for the oxidation of CO. Here, the rate of oxidation on NiO and CuO is proportional t o the pressure of CO and is independent of the oxygen pressure (2). Furthermore, the electrical conductivity of a NiO film is decreased in a CO-02 gas mixture in exactly the same manner as it is in the presence of a Nz-02 mixture of the same partial pressure of oxygen. These results were confirmed by Roginskii and Tselinskaya (78) and by Parravano (79), Schwab, and Block (80). Both results support the assumption th a t a t higher temperatures the chemisorption of 0 2 is fast, and the chemisorption of CO, or the reaction step (50f) determines the rate of the oxidation. Since the absolute number of electron defects per square centimeter in the surface of the oxide catalyst is contained in the frequency factor of

* A cross denotes the uncharged state.

APPLICATION O F T H E THEORY O F SEMICONDUCTORS

247

the rate equation, it is understandable that the rate of oxidation will be decreased when higher temperatures and longer times are used in the sintering of the catalyst. This pretreatment will result in a catalyst with a decreased surface area and with a smaller number of electron holes. Substantial changes of the activation energies of the CO-oxidation over NiO and ZnO, on addition to the catalysts, of foreign oxides with cations of a higher and lower valency than two (e.g., of LizO, Cr203,and Ga203) were observed by Schwab and Block (80),and are quite impressive. When the concentration of electron holes in NiO is decreased by the incorporation of Cr z 03or GaZO3, corresponding t o

2@

+ Crz03

+ 2Ni0 + 3 5 0 2 ' g )

2Cro.(Ni)

(51)

b NiO + L i Q

I 5

l

4

l

l

3 2 1 0 M l E X G o r O ~ CQO,

l

l

l

1

l

,

2 3 4 mole nliro

l

l

5

FIG.14.Dependence of the activation energy for the oxidation of carbon monoxide on the content of foreign oxides in NiO and ZnO catalysts, according to Schwab and Block. The lowest activation energy is caused by an addition of 1 mole percent of LizO to NiO.

then the activation energy for the oxidation of CO on a NiO catalyst containing 1 mole percent of Crz03increases from the value 15 kcal./mole found for pure NiO t o 19 kcal./mole. On the incorporation of G a z 0 3into an n-type oxide (e.g., ZnO), as expressed by the formula Gaz03

2Gao-(Zn)

+ 2 8 + 22110 +

>$02(g)

(52)

the concentration of free electrons in the solid is increased. The activation energy of the rate of oxidation of CO on a ZnO catalyst containing 1 mole percent of Gaz03 decreases from 28 kcal./mole for pure ZnO t o 19 kcal./mole. Results in opposite directions have been found on the addition of Liz0 (see Fig. 14).

248

K. HAUFFE

The dependence of the activation energy of the CO-oxidation catalyzed by NiO-Liz0 mixtures on the content of LizO, can be described in the following way: Assuming that process (50f) is the rate-determining step in the reaction of CO molecules with the chemisorbed oxygen atoms, the activation energy must decrease with increasing content of Lid3 in NiO and, therefore, the rate of oxidation must increase. If, however, the concentration of Liz0 in NiO and, likewise, the concentration of electron holes become too high, the chemisorption of oxygen becomes energetically unfavorable and-similarly as outlined for the decomposition of NzOthe chemisorption becomes the rate-determining step. This, however, is equivalent t o a n increase of the activation energy in the range of concentrations of L i 20 exceeding 1 mole per cent. As shown in Fig. 14, no such increase of the activation energy is observed. This apparent contradiction, however, disappears, if one considers th at with increasing concentration of electron holes in the boundary layer of the catalyst, the chemisorption of CO according t o (50b) becomes energetically more favorable. From this we can conclude that on NiO-LizO catalysts containing high concentration of LizO, the chemisorption of oxygen will be displaced by a n overwhelming chemisorption of carbon monoxide. Accordingly, a t higher contents of LizO, the reaction of CO molecules with chemisorbed oxygen atoms according t o (5Oe) and (50f) can be neglected. Instead, we obtain as the predominant reaction the combination of oxygen molecules with chemisorbed CO molecules according to (50g). Obviously, the rate-determining step of the oxidation of CO is here the chemisorption of CO, in accordance with the observation th a t the rate of the over-all reaction is proportional to pco. This is a n instructive example of a reaction in which the concentration of electron defects in the catalyst can be changed in various ways, thereby modifying decisively the mechanism of the catalysis. I n Fig. 15, the discussed mechanism of the oxidation of CO is schematically represented. Parravano (79) has also investigated the influence of foreign oxides added t o NiO (e.g., of AgzO, LizO, Cr203,and WO,) on the activation energy of the oxidation of CO in the temperature range from 160' t o 250°C. The activation energies obtained by Parravano show the opposite trend from those of Schwab and Block; i.e., an increase of the activation energy after the addition of AgzO and of LizO is observed, and a decrease of the activation energy after additions of foreign oxides with cations of higher valencies than two (e.g., of Crz03,CeOz, and WO,). This result is understandable, because under these experimental conditions, the chemisorption of CO is the starting reaction and (50g) is the rate determining pol!). step (therefore k Rienacker and co-workers (81) investigated also the dependence of the

-

APPLICATION OF THE THEORY OF SEMICONDUCTORS

249

activation energy of CO-oxidation on the content of foreign ions in Crz03 and Tho2 catalysts. For instance, the activation energy for the oxidation of CO on a Crz03 catalyst containing 2 mole percent of CuO decreases from the value of 28 kcal./mole found for pure Cr203to 19 kcal./mole. By the incorporation of 1 mole percent of CeOz into ThOz catalysts, the activation energy of the rate of oxidation of CO decreases from 17 kcal./mole for pure Tho2 to 11 kcal./mole. These experiments strongly indicate the important role of electron interchanges during the catalysis.

NIO

I I

I I

X$

xg

I

*XLi,0.

x,

FIG.15. The dependence of the activation energy for the oxidation of CO upon increasing amounts of LizO in NiO. Curve 1 represents the decrease of the activation energy with increasing concentration of Li,O in NiO. According to Equation (50f), the decrease of the activation energy results in a n increased desorption rate. After the critical concentration xe*,an increase of the activation energy (dotted curve 1) on further addition of LizO should be expected. According t o Schwab and Block, however, the activation energy remains practically unchanged at higher Liz0 concentrations. This becomes understandable in view of the decrease of the activation energy for the CO chemisorption with increasing concentrations of electron holes ( = I L ~ ~ O ) . At the critical concentration ze*, the activation energy of the CO chemisorption becomes lower than the activation energy for the oxygen chemisorption. Therefore, to the left of xe. we obtain a predominant chemisorption of oxygen (50a), and t o the right of ze* we get a predominant chemisorption of CO (50b) as the first step of the CO oxidation.

5. Some Further Simple Reactions Catalyzed by Semiconductors

By means of magnetic measurements, Selwood (82) observed th a t the concentration of electron holes (here identical with Ni3+ ions) in NiO films can be increased, if these NiO films are prepared on y-alumina supports. Obviously the chemisorption or the incorporation of oxygen into the lattice of NiO is favored by the effect of the epitaxy (oriented growth) of NiO on y-AL03. The n-conducting A1203as support for NiO cause high concentration

250

K . HAUFFE

of electron holes in NiO as an incorporation of Li ions. Besides increasing the surface of NiO, the Fermi level of electrons will be changed by the supporting material, a fact which as yet has not been recognized. As a matter of fact the changing of the Fermi level is the primary factor in increasing the catalytical effect. According to Selwood, the nickel assumes a valency of four on the rutile structure of titania (i.e., two electron holes per nickel), the oxide being bright yellow in color. Investigations of Steiner (83) and Huttig (61) on valence induction in other oxides illustrate the same effect. While Alz03,Mooz (83), ZnO, BeO, and Fe103 (61) as single oxides have a poor catalytical activity for some hydrogenation-dehydrogenation reactions, their catalytic activities are strongly increased in combinations resembling Fez03that of Ni0-r-Al2O3,for instance in the mixtures MOOZ--~-A~ZO~, ZnO, and Fe203-Be0.I n the case of Fe203,a small addition of titania should cause a similar increase of the catalytic activity. Further investigations are desirable concerning the connection between the catalytic activity of a given metal oxide and the nature and concentration of electron defects brought about in this oxide by supporting oxides. The metal sulfides represent an interesting group of catalysts for certain hydrogenations and dehydrogenations. Without treating the great number of technically important investigations we only make reference to the summarizing report of Kirkpatrick (84,85). It was observed that NiS with a deficiency of sulphur or an excess of nickel is much more catalytically active than nickel sulfide of a stoichiometrical composition. At present, however, it is not possible t o decide whether the chemical compound Ni3S2 plays a role. Further investigations are necessary on a series of metal sulfides (e.g., NiS, CdS, MoSz, and FeS), in order to see whether electron defects and boundary layers can contribute to an understanding of the mechanism by which these catalysts operate. Furthermore, it is likely that the mechanism of catalytic dehydrocyclizations, studied by Steiner (83) on Crz03with and without additions of foreign oxides and on MoS2, will be better understood by applying the theory of electron defects and of space-charge layers. Also, it will be fruitful to use isotopes for such studies, as has been done by Winter (86,87). Eucken (88,89) and Wicke (90) have tried to explain the dehydrogenation and dehydration of isopropyl alcohol by an electron interchange between the alcohol and the zinc oxide alumina catalysts used for these conversions. We shall modify the mechanism proposed by Eucken and Wicke, following the theory of chemisorption. Contrary to these authors, we do not believe that the positions of the zinc and oxygen ions on the surface of the zinc oxide catalysts have any appreciable influence upon

APPLICATION O F T H E T H E O R Y O F SEMICONDUCTORS

25 1

the selectivity of the catalysts for the dehydrogenation or dehydration of the alcohol. I n the author's opinion the selective decomposition of the alcohol will be determined only by the chemical potential of the electrons, i.e., the Fermi potential of the semiconducting catalyst and, furthermore, by the height of the potential barrier of the electrical field in the boundary layer which the electrons have to overcome during the exchange process. Following this concept, the following mechanism for the dehydration of isopropyl alcohol on a ZnO catalyst is proposed

The reaction starts with the chemisorption of the alcohol. The hydrogen atom of the alcohol loses an electron to the boundary layer of the catalyst. The next step is the chemisorption of the OH group, whereby a free electron in the boundary layer, Q ( ~ ) ,will be removed from the OH group

H

H H ZC---C--CH I OH

3

+

H+(u)

e(R)

-Z7n14

H 2C-C-CH

3

I I [ p + w OH-M

-A

(53b)

Finally, desorption of the products, water and propylene, takes place

H H,C-C-CHs

I

H+(u)

1

OH-(u)

-5

ZnO

H (CH2=C-C€L)

(8)

+ H20(g)

(53~)

As t o the mechanism of the dehydrogenation, we suggest that the reaction starts by a chemisorption of the hydrogen atom attached t o the second carbon atom of the alcohol as follows

I n a second step, the OH group will be chemisorbed with a displacement of the electron toward the oxygen atom

CH3 I

CH3 I

252

K. HAUFFE

In the next step, hydrogen will be desorbed, consuming a free electron CH3

CH3

I H3C-C-O-(") I \ H+(u)

I + 2 @ ( R ) =.&HZ(Y)+ HaC-C-O-(C)

(54~)

ZnO

Ij+Cu)

As a last step, acetone is desorbed

According t o Wicke (YO) and Kriicke (91) the desorptian of acetone seems t o be the rate-determining step for the dehydrogenation, because a pretreatment of the ZnO catalyst with acetone retards the dehydrogenation, whereas the rate of dehydration remains the same. According to reaction (54) the desorption of acetone, and therefore the rate of dehydrogenation, should be further decreased by increasing the concentration of free electrons, e.g., by adding Gaz0 3to the ZnO catalyst. On the other hand, we may expect a slight increase of the rate of desorption of acetone, and therefore of the dehydrogenation rate, by small additions of LizO to the zinc oxide, resulting in a decreased concentration of free electrons. According t o Kriicke (Yl), the desorption of hydrogen (54c) is a fast process. It would be of interest t o investigate whether p-type oxides and sulfides with small additions of suitable foreign oxides and sulfides, respectively, are more active catalysts than n-type oxides and sulfides. The above assumption have recently been confirmed by the results of Weisz (92), who studied the dehydrogenation of cyclohexane t o benzene and hydrogen, using Crz03as a p-conducting catalyst. During chemisorption of hydrogen, butane, or cyclohexane, the electrical conductivity of Crz03decreased from about 3 X ohm-' cm.-' at 480°C t o about 5 X ohm-' cm.-'. I n contrast t o hydrogen, the last two gases cause a conductivity inversion after a short reaction time. The boundary layer of Crz03receives so many electrons from the chemisorbing gas, t ha t all the electron holes will not only be destroyed, but free electrons will be introduced into the boundary layer, forming an inundation layer. Simultaneously the catalytic rate was measured. I n every case the rate of reaction was higher for greater concentrations of electron holes or smaller concentrations of free electrons in the boundary layer. Therefore the chemisorption of cyclohexane is the slowest and rate-determining step. If this conclusion is right, an addition of Ti02 to Crz03would further decrease the chemisorption of cyclohexane and the rate reaction, whereas

APPLICATION O F T H E THEORY O F SEMICONDVCTORS

253

an addition of NiO, for instance, would increase the chemisorption and thereby the rate of reaction. From a similar viewpoint, Clark (93) attempts to give a n interpretation of the catalytic activity of oxides of the transition metals using the simplified band model of semiconductors. The important mechanism of the electron transfer in a space-charge boundary layer is not discussed in Clark's publication. I n this connection, the dependence of the rate of the hydrogendeuterium exchange reaction catalyzed by zinc oxide, upon the concentration of a foreign oxide in ZnO is interesting. This system was investigated by Molinari and Parravano (94). As shown in Fig. 16, the rate of

T in"C

FIG.16. Rate of Ha-D2exchange] 6 =

{

1x

100, as a func-

tion of the temperature] on sintered ZnO containing foreign ions, according toMolinari and Parravano. The samples were activated in hydrogen at 350°C. The flow rate was 0.238 ~ m . ~ / s e c .

the exchange reaction decreases with decreasing concentration of free electrons caused by the addition of Li20. On the other hand the rate increases with increasing concentration of free electrons caused by additions of A1203 or Gaz03.These results lead us to conclude th a t the rate of electron transfer t o H+ and D+ ions during the desorption is the rate-determining step. By hydrogen pretreatment of zinc oxide samples containing chemisorbed oxygen, the surface of the catalyst will be freed of this oxygen in a short time. Therefore, we propose the following mechanism

254

K. HAUFFE

Because of the presence of oxygen, the formation of chemisorbed OHgroups on the surface is quite possible, as pointed out by Anderson (95) and Parravano (94). Air treatment of activated catalysts, however, reoxidizes the hydrogenated ZnO surface, thus producing again a n inactive surface of a nearly stoichiometrically composed ZiW. Thus, the boundary layer with a high concentration of free electrons disappears almost completely. It should be emphasized that the pretreatment of the surface of n-type oxides seems to be a necessary condition for their catalytic activity. Data of Holm and Blue (96) and of Parravano (94) show th a t several specimens of molybdenum, tungsten, and uranium oxides and tungsten bronzes are not active for &he hydrogen-deuterium exchange unless they are subjected t o pretreatments similar t o those described above for zinc oxide. I n catalytic studies it is important to recognize also that the addition of impurities will affect the sintering processes, and thus the specific surface areas after such treatment. I n the case of zinc oxide, a n addition of Liz0 will facilitate the sintering process and thereby decrease the surface area. On the other hand, the addition of G a z 0 3(94) will make the sintering process more difficult. This result is in agreement with the influence of small additions of lithium and aluminium t o zinc on the rate of oxidation of zinc alloys, as observed by Gensch and Hauffe (97). Russian authors such as Roginski (98) are also of the opinion that the high catalytic activity of oxides and sulfides for hydrogenation-dehydrogenation reactions of aliphatic hydrocarbons is brought about by the mobility of their free electrons. Griffith and co-workers (99) also point out qualitatively the connections between the electron disorder in solid catalysts and the rate of reaction. These authors use n-type MoO2, and show that by increasing additions of Si02 t o MoO2, a maximum of surface and a minimum of conductivity will be attained a t about 4 mole percent of Si02. The strong decrease of the conductivity a t 400°C resembles very closely that described by Johnson and Weyl (loo), in which ZrOz was added t o partly reduced TiOz. Furthermore, Griffith et al. suggest th a t MoOz, and not metallic Mo, was effective in the used Si02-Mo02 catalyst for hydrocarbon decomposition. Also, Herington and Rideal (101) believe th a t the oxide is essential for the production of aromatic hydrocarbons from paraffins, because the low chemical potential of electrons is energetically more favorable than this of the metal.

SUMMARY The preceding survey was not written for the purpose of giving a complete review of all important experimental results on the exchange of electrons between semiconducting catalysts and reacting gases. Its main

APPLICATION O F T H E THEORY O F SEMICONDUCTORS

255

purpose is rather to stimulate the workers in the field of heterogeneous catalysis by presenting some of the more recent theoretical ideas regarding the mechanism by which semiconducting solid catalysts can influence by means of their electronic disorder, the mechanism and the rate of catalytic reactions.

REFERENCES 1. See, e.g., Hauffe, K., Lattice defects and conductivity of ionic and electronic solid conductors. Ergeb. ezakt. Naturw. 26, 193 (1951). 2. See, e.g., Mott, N. F., and Gurney, R. W., “Electronic Processes in Ionic Crystals.” Oxford, New York, 1948; Seitz, F., “The Modern Theory of Solids.” McGraw-Hill, New York, 1940. 3. Wagner, C., and Hauffe, K., Z.Elektrochem. 44, 172 (1938). 4. Wagner, C., J. Chem. Phys. 18, 69 (1950). 5. Dowden, D. A., Chemistry & Industry 1949,320; J . Chem. SOC.1960, 242. 6. Anderson, J. S., Discussions Faraday SOC.No. 4, 163 (1948). 7. Garner, W. E., Gray, T. J., and Stone, F. S., Proc. Roy. SOC.(London) A197,294 (1949). 8. See, e.g., Garner, W. E., Discussions Faraday SOC.No. 8, 211 (1950). 9. Weyl, W. A., Penn. State Coll. Bull. NO.67 (1951). 10. Fajans, K., Chem. Eng. News 27, 900 (1949). 11. Krauss, W., Z . Elektrochem. 63,320 (1948). 12. BQnard, J., Ind. chim. Belge 17, 553 (1952). 13. Taylor, H. S., Discussions Faraday SOC.NO.8, 9 (1950). 14. See also, Boudart, M., Ind. Eng. Chem. 46, 898 (1953); Boudart, M., J. Am. Chem. SOC.74, 1531 (1952). 14a. Hauffe, K., Glang, R., and Engell, H. J., 2.physik. Chem. 201, 223 (1952); Engell, H. J., and Hauffe, K., 2. Elektrochem. 67, 776 (1953). 15. Lennard-Jones, J. E., Trans. Faraday SOC.28, 341 (1932). 16. Hauffe, K., and Engell, H. J., Z . Elektrochem. 66,366 (1952). 17. Engell, H. J., and Hauffe, K., Z. Elektrochem. 67,762, 773 (1953); see also Hauffe, K.,Angew. Chem. 67, 189 (1955). 18. Engell, H. J., in “Halbleiterprobleme” (W. Schottky, ed.), Vol. 1, p. 249. Vieweg, Braunschweig, 1954. 19. Schottky, W., Naturwissenschajten 26, 843 (1938); 2.Physik. 113, 367 (1939); 118, 539 (1942); Schottky, W., and Spenke, E., Wiss. Verijftentl. SiemensWerken 18, 25 (1939). 20. Mott, N. F., Proc. Roy. SOC.(London)A171, 944 (1939). 21. Davidov, B., J. Phys. (U.S.S.R.) 1, Heft. 2 (1939). 22. Aigrain, P., and Dugas, C., 2.Elektrochem. 66, 363 (1952). 23. Weisa, P. B., J . Chem. Phys. 20, 1483 (1952). 24. Weisz, P. B., J . Chem. Phys. 21, 1531 (1953). 25. Bevan, D. J. M., Shelton, J. P., and Anderson, J. S., J . Chem. SOC.1948, 1729. 26. Miller, P. H., Jr., in “Semiconducting Materials” (H. K. Henisch, ed.), p. 172ff. Academic Press, New York, 1951. 27. Hahn, E. E., J . A p p l . Phys. 22, 855 (1951). 28. Henisch, H. K., Z.phys. Chem. 198, 41 (1951). 29. Ljaschenko, W. J., and Stepko, J. J., Doklady Akad. Nauk. S.S.S.R. phys. ser. 16, 274 (1952). 30. Hauffe, K., and Block, J., Z. physik. Chem. 196, 438 (1950).

256

I<. HAUFFE

31. Mott, N. F., in “Semiconducting Materials” (H. K. Henisch, ed.), p. 5 . Academic Press, New York, 1951. 32. Tamm, I., Physik. 2. Sowjetunion 1, 733 (1932). 33. Brattain, W. H., and Bardeen, J., Bell System Il‘ech. J . 32, 1 (1953). 34. Brattain, W. H., and Shockley, W., Phys. Rev. 73, 345 (1947); Bardcen, J., and Brattain, W. H., Phys. Rev. 76, 1208 (1949); Shockley, W., and Pearson, 0. I,., Phys. Rev. 74, 232 (1948). 35. Dubar, L., Compt. rend. 202, 1330 (1936); see also measurements carried orit by J. N. Shive, quoted by Bardeen, J., Phys. Rev. 71, 718 (1947). 36. Bevan, D. J. M., and Anderson, J. S., Discussions Faraday SOC.No. 8,238 (1950). 37. Volger, J., Phys. Rev. 79, 1023 (1950); in “Semiconducting Materials” (H. K. Henisch, ed.), p. 162ff. Academic Press, New York, 1951. 38. Brauer, P., Ann. Phys. 151 26, 609 (1936). 38a. Antipina, T. V., and Frost, A. V., Doklady Akad. Nauk S.S.S.R. 84, 985 (1952). 39. Muscheid, E., Thesis, Humboldt-Universitat, Berlin, 1952; Ann. Phys. 161 13, 305 (1953). 3%. Morrison, S. R., J . Phys. Chem. 67, 860 (1953). 40. Rubin, T. R., Calvert, J. C., Rankin, G. T., and MacNevin, W., J . Am. Chem. Soc. 76, 2850 (1953). 41. Markham, M. C., Hannan, M. C., and Evans, S. W., J . Am. Chem. SOC.76, 820 (1954). 42. See, e.g., Suhrmann, R., 2. Elektrochem. 66, 351 (1952); Suhrmann, R., and Schulz, K., Naturwissenschaften 40, 139 (1953) ; see also Dr. Suhrmann’s article in this volume. 43. Cabrera, N., and Mott, N. F., Repts. Progr. i n Phys. 12, 163 (1949). 44. Engell, H. J., and Hauffe, K., Metal 6, 285 (1952); Hauffe, K., and Ilschner, B., 2. Elektrochem. 68, 467 (1954). 45. Elovich, S. Y., and Zhabrova, G. M., Zhur. Fiz. Khim. 13, 1761, 1775 (1939). 46. Taylor, H. S., and Strother, C. O., J . Am. Chem. SOC.66, 586 (1933). 47. Sickman, D. V., and Taylor, H. S., J . Am. Chem. SOC.64, 602 (1932). 48. Taylor, H. S., and Liang, S. C., J . Am. Chem. SOC.69, 1306 (1947). 49. Sastri, M. V. C., and Ramanathan, K. V., J . Phys. Chem. 66, 220 (1952). 50. Gray, T. J., private communication. 51. Taylor, H. A., and Thon, N., J. Am. Chem. SOC.74, 4169 (1952). 52. Burwell, R. L., and Taylor, H. S., J. Am. Chem. Soc. 68, 697 (1936). 53. Williamson, A. T., and Taylor, H. S., J . Am. Chem. SOC.63, 2168 (1931). 54. Smith, E. A,, and Taylor, H. S., J . Am. Chem. SOC.60, 362 (1938). 55. Volkenshtein, F. F., Zhur. Fiz. Khim. 23, 917 (1949). 56. Volkenshtein, F. F., and Bonch-Bruevich, V. L., Zhur. Ekspll. l’heoret. Pzz. 20, 624 (1950). 57. See, e.g., Gurevich, D. B., Tolstoi, N. A., and Feofilov, P. P., Zhur. Ekuptl. Theoret. Fiz. 20, 769 (1950) ; Krylova, E. S., ibid. 20, 905 (1950). 58. Schwab, G.-M., and Schultes, H., 2. physzk. Chem. B9,265 (1930); 26,411 (1934). 59. Schwab, G.-M., and von Baumbach, H. H., 2. physik. Chem. B2l,26 (1933). 60. Schenck, R., Angew. Chem. 49, 694 (1936); Rabes, I., and Schenck, R., 2. anorg. Chem. 260, 154 (1949). 61. Huttig, G. F., Discussions Faraday SOC.No. 8, 215 (1950); see references therein of former works of the author and his co-workers. 62. See, e.g., Eucken, A., “Lehrbuch der chemischeri Physik,” 2nd ed., Vol. 11,2. Akademische Verlagsges., Leipzig, 1944.

APPLICATION O F THE T H E O R Y O F SEMICONDUCTORS

257

63. 64. 65. 66.

Boudart, M., J. Chem. Phys. 18, 571 (1950). Hauffe, K., and Block, J., 2. physik. Chem. (Schottky-Festband) 198,232 (1951). Schmid, G., and Keller, N., Naturwissenschaften 37, 42 (1950). Verwey, E. J. W., Haijman, P. W., and Romeijn, F. C., Chem. Weekblad 44, 705

67. 68. 69. 70. 71. 72.

Dell, R. M., Stone, F. S., and Tiley, P. F., Trans. Faraday SOC.49, 201 (1953). Cremer, E., and Marshall, E., Monatsh. Chem. 82, 840 (1951). Rienacker, G., Bremer, H., and Unger, S., Naturwissenschaften 39, 57 (1952). Schwab, G. M., Chimia (Switz.) 6, 247 (1952). Gray, T. J., Proc. Roy. SOC.(London) A197, 314 (1949). Garner, W. E., Stone, F. S., and Tiley, P. F., Proc. Roy. SOC.(London) A211,472

(1948).

(1952). 73. Stone, F. S., and Tiley, P. F., Discussions Faraday SOC.No.8, 256 (1950); Nature 167, 654 (1951). 74. K. J. Laidler, in “Catalysis” (P. H. Emmett, ed.), Vol. 1, pp. 85 and 101. Reinhold, New York, 1954. 75. Frankenburg, W. G., J . Am. Chem. SOC.66, 1827, 1838 (1944). 76. Kwan, T., Kinuyama, T., and Fujita, Y., J . Research Znst. Catalysis 3,28 (1953); Kwan, T., and Fujita, Y., ibid. 2, 110 (1953). 77. Kwan, T., and Fujita, Y., Nature 171, 705 (1953). 78. Roginskii, S. Z., and Tselinskaya, T. S., Zhur. Fiz. K h i m . 22, 1350 (1948). 79. Parravano, G., J. Am. Chem. SOC.76, 1448 (1952). 80. Schwab, G. M., and Block, J., 2. physik. Chem. [N. S.] 1, 42 (1954); Block, J., Thesis, Munchen, 1954. 81. Rienacker, G., and Burmann, R., 2. anorg. Chem. 268,280 (1949); Rieniicker, G., and Birckenstaedt, ibid. 262, 81 (1950). 82. Selwood, P. W., Bull. soe. chim. (France) 1949, 489. 83. Steiner, H., Discussions Faraday SOC.No. 8, 264 (1950). 84. Kirkpatrick, W. J., Advances in Catalysis 3, 329 (1951). 85. See also, Adkins, H., Rae, D. S., Davis, J. W., Hager, G. F., and Hoyle, K., J . Am. Chem. SOC.70, 381 (1948). 86. Winter, E. R. S., Discussions Faraday SOC.No. 8, 231 (1950). 87. Houghton, S., and Winter, E. R. S., Nalure 164, 1130 (1949). 88. Eucken, A., Naturwissenschaften 36, 48 (1949). 89. Eucken, A., and Heuer, K., 2. physik. Chem. 196, 40 (1950). 90. Wicke, E., 2. Elektrochem. 62, 86 (1948); 63, 279 (1949). 91. Krucke, E., Diplomarbeit, Gottingen, 1945. 92. Weisz, P. W., Prater, C. D., and Rittenhouse, K. D. J . Chem. Phys. 21, 2236 (1953). 93. Clark, A., Znd. Eng. Chem. 46, 1467 (1953). 94. Molinari, E., and Parravano, G., J . Am. Chem. Soe. 76, 5233 (1953). 95. Roberts, L. E. J., and Anderson, J. S., Revs. Pure and A p p l . Chem. (Australia) 2, 1 (1952). 96. Holm, V. C. F., and Blue, R. W., Znd. Eng. Chem. 44, 107 (1952). 97. Gensch, C., and Hauffe, K., 2. physik. Chem. 196, 427 (1950). 98. Roginskii, S. S., Compt. rend. S.S.S.R. 67, 97 (1949). 99. Griffith, R. H., Chapman, P. R., and Lindars, P. R., Discussions Faraday SOC. No. 8, 258 (1950). 100. Johnson, D., and Weyl, W. A., J. Am. Ceram. SOC.32, 398 (1949). 101. Herington, E. F. G., and Rideal, E. K., Proc. Roy. Soc. (London) A184,434 (1945).

This Page Intentionally Left Blank

Surface Barrier Effects in Adsorption, Illustrated by Zinc Oxide* S. ROY MORRISONt Department of Electrical Engineering, University of Illinois Page I. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 11. Electronic Structure and Electron Transfe . . . . . . . . . . . . . . . . . 261 1. The Band Model.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 2. Surface States and the Surf 3. Electron Transfer in Equilibrium Chemisorption; the Effects o Surface Barrier.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .............................. 266 111. Basic Properties of Zinc Oxide 1. Introduction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 2. Bulk Properties of Zinc Oxide.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. The Role of the Surface and Adsorbed Oxygen in Conductance.. . . . . . . . 268 IV. Irreversible Adsorption and the Electronic State of the Surface.. . . . . . . . . . 272 1. Electron Transfer as the Rate-Limiting Step in 2. The Relation between Adsorption and Conduct a. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . b. Low-Temperature Conductivity (Below 20°C). . . . . . . . . . . . . . . . . . . . . 275 c. High-Temperature Conductivity (Above 500°C). . . . . . . . . . . . . . . . . . . 277 d. Intermediate Temperatures (20°C to 500°C). ...................... 279 e. Theoretical Considerations. . . . . . . . . . . . . . . . . . . . 282 f. Further Comparison with Ex .................... 3. A Suggested Mechanism for the Adsorption of Hydrogen on Zinc a. The Proposed Model.. .................................... .......................... 292 b. Comparison with Experiment V. The Relationship of the Photoconductivity to the State of the Surface VI. The Relationship of the Fluorescence to the State of the Surface.. . . . . . . . . 298 References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300

I. INTRODUCTION The role of electron transfer during chemisorption, and its importance in adsorption, catalysis, and solid state physics, has been demonstrated repeatedly during recent years, examples occurring in every phase of surface study.

* Supported in part by the U.S. Office of Naval Research. Based, in part, on a thesis submitted t o the Physics Department, University of Pennsylvania (1953). t Now at Sylvania Electric Products, Physics Laboratories, Bayside, New York. 259

260

S. ROY MORRISON

This work will attempt to demonstrate the importance of adsorbed oxygen on many properties of zinc oxide, namely the conductance, fluorescence, photoconductance, and the adsorption of hydrogen. The contribution, potential, and current of each of these studies to a more complete understanding of the adsorption process will be discussed. The electron transfer theory of chemisorption will be briefly summarized, indicating the important role played by the potential barrier associated with a double layer between adsorbed ions on the surface and ionized impurities of opposite charge in the semiconductor. It will be shown that the electron transfer theory provides a consistent interpretation of many anomalous time-dependent effects occurring on zinc oxide, Assuming electron transfer is the rate-limiting process, we are able t o interpret the slow reversible conductivity change in the neighborhood of 5OO0C, the slow reversible and the slow irreversible conductivity changes in the neighborhood of 100°C, slow photoconductive rise and decay processes studied mainly between room temperature and 100°C, and the slow adsorption of hydrogen on zinc oxide. Zinc oxide is certainly not unique in exhibiting a strong correlation between adsorption and its electrical and optical properties. Dowden and Reynolds (1) have cited several examples of the parallelism between reaction rates in ratalysis and electron density in other adsorbents. Extensive work by Garner and his co-workers (2,3) has shown clearly the intimate relation between adsorption and the conductivity of copper oxide. Clarke (4) and Morrison ( 5 ) have shown th a t adsorption changes the resistance of germanium. I n metals, there appears a similar effect. Suhrmann and Schulz (6) have demonstrated the dependence of the conductivity of thin films of nickel on the adsorption of various gases. Properties dependent on adsorption are not confined t o conductivity. Luminescence of materials may be affected, as Ewles and Heap (7) have shown for the case of silica, for which the luminescent peak a t 4000 8. was shown to be associated with the adsorption of the OH- radical. Many workers have demonstrated the dependence of the contact potential on the adsorption of gases. For example, Brattain and Bardeen (8) have shown that the contact potential of germanium varies with the adsorption of water vapor. Photoconductivity may be dependent on the adsorption. For example, Bube has shown (9) that the adsorption of water vapor has a marked effect on the photoconductivity of cadmium sulphide. He concluded (10) that the effect was indirect; surface changes affect the lifetime of the excess carriers, thus affecting the photoconductivity. Melnick (1l), however, working with zinc oxide, has produced evidence th a t part of the photoconductivity in this case is directly associated with excitation from adsorption levels.

SURFACE BARRIER EFFECTS I N ADSORPTION

26 1

11. ELECTRONIC STRUCTURE AND ELECTRON TRANSFER AT SURFACES 1. The Band Model

When one is considering electronic processes in solids, the band theory provides the most useful basis for discussion. According to the band theory, the allowed energy values for an electron in a crystal are grouped, as a function of energy, into “bands.” Between these allowed energy regions, there may be one or more bands of energy which are “forbidden”; that is, in an ideal crystal, an electron cannot possess an energy in a forbidden range. If there are sufficient electrons in the crystal to fill the allowed energy levels exactly to the topmost energy level in a given band, and if there is a forbidden band of energy between the highest filled band and the next allowed band, the crystal will be a semiconductor or insulator. I n a semiconductor the topmost normally filled band is termed the “filled band” or “valence band”; the next higher band of allowed energy levels is termed the “conduction band.” The intermediate band of forbidden energies is termed the “forbidden band,” “forbidden gap,” or simply the ccgap.” Electric conduction in the semiconductor takes place by the movement of electrons with energies in the conduction band (N-type conduction), or by ((holes,” which are unoccupied states in the normally filled band (P-type conduction). A hole may be pictured roughly as a n empty valence bond; conduction occurs through the exchange of electrons from neighboring filled valence bonds, and the net effect, is that the hole, acting as a positive charge, may wander through the lattice under the influence of a n electric field. It can be shown by quantum theory that a hole behaves in most respects like a positive charge ( + e ) with an effective mass of the order of the free electron mass. I n most semiconductors, there are, in addition t o the allowed energy levels for electrons in the conduction and filled bands of the ideal crystal, discrete levels with energies in the forbidden gap which correspond t o electrons localized a t impurity atoms or imperfections. I n zinc oxide, such levels arise when there are excess zinc atoms located interstitially in the lattice. At very low temperatures the interstitial zinc is in the form of neutral atoms. However, the ionization energy of the interstitial atoms in the crystal is small and at room temperature most are singly ionized, their electrons being thermally excited into the conduction band. These electrons give rise t o the observed N-type conductivity. The band structure of a semiconductor is often plotted with the total energy of a n electron as the ordinate and the distance through the crystal as the abscissa. A typical example is shown in Fig. 1. Energy levels

262

S. ROY MORRISON

corresponding t o the interstitial zinc atoms (termed “donor ” impurities, since they “donate” electrons t o the conduction band) are shown as small dashes, t o indicate that the electron is localized in space a t the position of a n interstitial zinc atom (the energy levels which make u p the conduction and filled bands do not correspond to electrons localized in space). It is indicated in Fig.ll thatfthe ionization energy for single ionization of a n interstitial zinc atom is 0.04 electron volts; that for double ionization is 2.2 electron volts. Not shown in Fig. 1, although used in subsequent illustrations is the energy p, which is the electrochemical potential, or

0FIG.1. Proposed energy-level diagram for zinc oxide.

Fermi level, of the electrons. If G ( N ) is the free energy of a crystal containing N electrons, then p = aG/aN I n Fermi-Dirac statistics, p is the Fermi energy EF, which is such th a t the probability, that a state of energy EF is occupied, is 1/2. States with energies higher than EF have a smaller probability of being occupied, those with lower energy, a higher probability. The position of the Fermi level in a semiconductor depends markedly on the temperature and on the concentration of impurities. The Fermi levels of two conductors in electrical contact and in thermal equilibrium are the same. 2 . Surface States and the Surface Barrier

Energy levels corresponding t o electrons localized near the surface may also be present. These are termed “surface states.” For example, adsorbed ions are one type of surface state; they may be of the form of donors, such as hydrogen, which yield electrons to the material, or in the form of acceptors, such as oxygen, which “accept,” or ‘(tra p ” electrons from the material. I n Fig. 1, surface traps of the acceptor type are shown, and i t is indicated that there are two possible levels present. Other possibilities are impurity atoms a t the surface, which are introduced’in the preparation of the sample or diffuse from the interior during heat treatment, or nonstoichiometry of the surface layers of the compound. Surface

SURFACE BARRIER EFFECTS IN ADSORPTION

263

levels with energies in the forbidden gap may exist at the surface of an ideal crystal, as indicated by quantum theoretical arguments of Tamm (12), and later by Shockley (13) and Baldock (14). These latter are termed “Tamm levels.” As we shall see later on, the existence of other surface levels may have an important effect on the adsorption characteristics of a given substance. Energies of electrons in the surface levels of a semiconductor may be affected strongly by a “potential barrier” a t the surface, which makes the electrostatic potential at the surface differ from that in the interior. The existence of such a potential barrier, sometimes referred to as a “barrier layer” or “space-charge layer” was first proposed by Bardeen (15) in 1947, to account for observed surface properties of germanium and silicon. The formation of a surface barrier may be illustrated by the model in Fig. 1. Electrons will become trapped on the acceptor levels, indicated to be adsorbed oxygen, because these levels, as drawn, have a lower energy than those of interstitial zinc. Transfer of electrons from the interstitial zinc impurities near the surface t o oxygen adsorbed on the surface leaves a positive space charge of zinc ions compensated by a negative surface charge of adsorbed oxygen ions. The electric field in the space-charge region gives a voltage drop between the interior and the surface traps such that the potential energy of an electron increases as it approaches the surface. Thus the energy of each allowed state near the surface, including the surface states themselves, becomes located a t a higher energy as the surface becomes more negatively charged. I n Fig. 1, the energy bands are curved up near the surface to indicate this condition. The energy rise becomes greater as more oxygen is adsorbed. The Fermi level, of course, remains at the same energy, in equilibrium. When the bands are curved up near the surface, as in Fig. 1, it should be noted that there is a potential barrier to electron transfer between the interior and the surface states. This is what we have called the surface barrier. If the space charge in the semiconductor arises from the ionization of impurities only, as in the model we have used, the surface barrier is termed a “Schottky barrier.” The barrier region near the surface of the crystal is sometimes called the “exhaustion region,” as the mobile electrons have been removed from this region (16). A surface barrier to the transfer of holes will arise if a positive surface charge is established on a P-type semiconductor, i.e., a semiconductor in which holes are the electrical current carriers. With such a surface charge the bands will bend down. Effects occurring on a P-type semiconductor will not be discussed explicitly in this article, although the same general principles apply with minor modifications.

264

S. ROY MORRISON

3 . Electron Transfer in Equilibrium Chemisorption; the Effects of the Surface Barrier

As early as in 1937, Nyrop (17) suggested that electron transfer may occur during chemisorption. Dowden (18) clarified the situation by classifying the possible reactions with respect to the type of bond (ionic, covalent, or mixed) and the type of adsorbent (metal, semiconductor, or insulator). He attempted to indicate some probable criteria to be used in the choice of the best adsorbent for use with a given adsorbate. Volkenshtein (19) and Schwab (20) suggested that the active centers of adsorption were actually lattice defects in the semiconductor. Schwab felt that the concentration of free or quasi-free electrons was an important factor in metal catalysts. An important step in the understanding of adsorption was the incorporation of the surface barrier into the theory. The barrier may affect not only the equilibrium amount adsorbed, but also the rate of adsorption. A surface barrier, may exist as a result of other surface charges, prior to the adsorption of the gas under consideration. We will discuss first the case where no surface barrier exists prior to adsorption, then briefly discuss the case where one does exist. The effect of the surface barrier on equilibrium adsorption has been discussed by Aigrain and Dugas (21) and Weisz (22,23). They point out that the surface barrier is of interest especially in the case of adsorption of electronegative adsorbates on N-type semiconductors, or of electropositive adsorbates on P-type semiconductors (Schottky barrier). In adsorption of electropositive adsorbates on a n N-type semiconductor, the bands will turn down, since the surface is positive. This will form a potential well for electrons, rather than a barrier a t the surface. A high concentration of electrons mill be in the well; the charge of these excess electrons will be equal t o the positive charge of the adsorbed ions. Thus the semiconductor may be expected t o be more “metallic” in its adsorption properties, the charge balance arising from mobile charges (electrons in the N-type semiconductor) rather than fixed impurity ions. Adsorption of oxygen on CUZOis an example. The negative oxygen ions are compensated by mobile positive holes. Aigrain and Dugas discuss features desirable for adsorption of various species. Weisz (22) derives quantitative expressions for the heat of adsorption, the rate of adsorption, and the amount of adsorption, for a simple model. The model used involves a simple surface barrier of the type in Fig. 5, with adsorption traps as the only surface traps, and where, if the system reaches equilibrium, adsorption occurs until the adsorption traps are a t the energy of the Fermi level. m’eisz shows that the surface cannot become

SURFACE BARRIER EFFECTS I N ADSORPTION

265

completely covered during adsorption. He also demonstrates that, for deep-lying adsorption traps, the adsorption is insensitive to temperature. The rate of adsorption will be determined by the height of the dipole barrier, and will decrease as the barrier rises with increasing surface coverage. The model gives a constant rate of desorption. Weisz’ assumption that, when equilibrium is attained, the energy of the adsorption traps is at the Fermi level, may not be valid in all cases. This assumption has the effect of removing the temperature dependence from the equilibrium adsorption. It is equivalent t o the assumption that the number of empty adsorption traps is about equal to the number of ionized adsorption traps) and is invalid if empty adsorption traps are physically adsorbed atoms or molecules. For the latter case, AE’ in Fig. 5 will in general be greater than zero. We shall present here a calculation of the equilibrium adsorption when it is assumed that physically adsorbed molecules are the empty adsorption traps. This gives a temperature dependence of the high-t>emperature equilibrium adsorption. Assume again a simple surface barrier, produced by ionization of donor impurities of concentration Nd, a constant. Poisson’s equation, using MKS units, becomes in the exhaustion region

d2V - -eNd dx2 K Integrating this twice, and using the boundary conditions th a t V = 0 a t x = 0, dV/dx = 0 a t x = xo, Ti = E P / e (as in Fig. 5 ) a t x = 20, where xo is the thickness of the exhaustion region, we find that the solution of the above equation yields EP = e2Ndxo2/2K (2)

If it is assumed that each donor yields one electron, and that the adsorbed ions accept one electron each, the surface concentration of chemisorbed ions will be Nu = N ~ x oand , from ( 2 )

E P = e2Nu2/2KNd

(31

Now, neglecting entropy changes due to the adsorption and desorption of physically adsorbed particles, and assuming that the empty adsorption traps are physically adsorbed particles present in surface concentration No (which is assumed independent of Ez), the number of these will be, by the Boltzmann distribution N o = N , exp [- ( E 3 - E P - p ) / k T ]

(4)

It is assumed th at Nu>> No, so the Boltzmann distribution can be applied. The symbols are those of Fig. 5 . Combining (3) and (4)) one

266

S. ROY MORRISON

arrives a t the relation

It is seen that Equation ( 5 ) is essentially that obtained by Weisz for the case of a large trap depth, with the addition of a term which gives the temperature dependence. A modification of this simple adsorption theory must be made if there are other surface levels present. If the surface concentration of these levels is very large compared t o the concentration of ions to be adsorbed, one would expect the adsorption t o more closely resemble th a t on a clean metal, as electron transfer between the various surface traps may predominate over transfer between the adsorbed ions and the bulk semiconductor. If the number of these traps is small compared to the amount of adsorption, one would expect the adsorption characteristics to resemble those for the theory discussed above. Intermediate cases are also possible. A possible intermediate case discussed in Section IV,3 applies t o the adsorption of hydrogen on zinc oxide. The extra surface levels are considered t o be previously adsorbed oxygen. The above discussion of the band model and its application t o a description of equilibrium chemisorption has been brief and incomplete. It is not the principal purpose in this paper to discuss surface barrier effects in equilibrium chemisorption, but rather surface barrier effects in the irreversible region of chemisorption. However, before we begin the consideration of these latter effects, we will digress and examine the basic properties of zinc oxide, which material will be used both as motivation and as illustration in the following text. 111. BASICPROPERTIES OF ZINC OXIDE 1 . Introduction

Zinc oxide seems to be a semiconductor with many of the properties necessary for a comparison with the electron transfer theory partially described in the preceding chapter. The material has been found to be an N-type semiconductor a t all temperatures (24-27). The donor impurities, as has been quite well established (24,28),are interstitial excess zinc atoms. Their concentration depends on the method of preparation of the sample, ranging from 1Ol6 ~ m . -t o~ l O I 9 om.+ (24,28). Of the various types of surface states available, evidence will be presented below which indicates that the only deep surface traps are adsorption traps. The possibility exists th at some very shallow surface

S U R F A C E BARRIER EFFECTS IN ADSORPTION

267

traps arise from other causes, but they are shown t o be of the order of a few hundredths of an electron volt below the conduction band at most. Thus a small surface barrier will cause them to be completely ionized. As will be discussed later in this section, most of the investigations on zinc oxide have been done on specimens whose structure allows the surface effects to be large. 2 . Bulk Properties of Zinc Oxide The bulk properties of a material are best obtained from large single crystals. However, measurements on single crystals of zinc oxide of desirable purity and size are not yet available. Some work, however, has been done on small pure crystals and larger impure crystals. The forbidden energy region in zinc oxide is generally considered to be 3.1 or 3.2 e.v. in width. This value is obtained from the optical absorption edge, which was found to be at about 3850 8. (28,29) for both single crystals and for specimens prepared in other ways. Scharowsky (28) has made the most complete study of single crystals of zinc oxide, using small crystals whose length was 4 em. and whose thickness was the order of tenths of a millimeter. This is sufficiently large to eliminate most of the surface effects. He studied the conductivity and the absorption of light as a function of the concentration of excess interstitial zinc. The conductivity as a function of temperature (from 20" to 250") showed metallic beAavior (decrease in conductivity with increasing temperature) for samples with room-temperature conductivities greater than about 0.4 ohm-' cm.-l. Using Hahn's value for the mobility in single crystals, as referred to below, this conductivity, 0.4 ohm-' cm.-', would correspond to an impurity concentration of about 3 X em.+. For samples with less than this impurity concentration, the conductivity was found to increase with temperature, yielding an activation energy varying with concentration approximately according to Meyer's Rule (30). That is, as the concentration of interstitial zinc decreases, the activation energy for production of conduction electrons increases, attaining, for very low concentrations (about 1OI6 c m . 3 , the order of magnitude of a few tenths of an electron volt. The absorption of light increases as the concentration of interstitial zinc increases. Scharowski subtracts the "intrinsic " absorption, as found in relatively stoichiometric crystals, from the total absorption in a highly doped sample, and considers the excess absorption t o arise from the double ionization of interstitial zinc. This excess absorption peaks a t about 3.2 e.v., and from this Scharowski concludes that the energy of ionization of interstitial Zn+ is 3.2 e.v. The fact that this is equal to the forbidden gap width is considered to be coincidental.

268

S. ROY MORRISON

Hahn (24) has studied large impure single crystals, and from Hall effect and conductivity data has found the electron mobility a t room temperature t o be about 85 cm.2/volt-sec., with values deviating by about 20 percent from this one for some samples. 3. The Role of the Surface and Adsorbed Oxygen in Conductance

The electronic and optical properties of a solid may be influenced considerably by the state of the surface. The effects of the surface on optical properties will be discussed further in a later section; at this point we will attempt only t o demonstrate that the electrical conductivity may be partially controlled by the state of the surface. The simplest way to consider the problem is as follows. If we have a n N-type semiconductor upon which an acceptor is being adsorbed, the adsorption will remove conduction electrons from the semiconductor and the conductance of the semiconductor will decrease. The physical properties necessary for the control of the electrical conductivity by surface traps may be indicated by a brief calculation. If we assume t ha t the surface levels are deeper than the bulk donor levels, a large fraction of the electrons from the donors may be trapped on the surface, and the conductivity will be strongly dependent on the properties of the surface traps. If n is the density of donors in the sample, A is the surface area, V is the volume of the highly resistant parts of the safnple, and N , is the AN, number of surface traps per unit area, it is easily calculated th a t __ will nV be the fraction of the available electrons trapped from the donor atoms by surface traps. A typical bulk concentration of donors in zinc oxide is lo1' CM.-~.Using 5 X 1013 cm.-2 as a typical concentration of surface traps, it is seen that when the surface-to-volume ratio approaches the order of lo3cm.-' or more, surface trapping may become important in the analysis of conductivity measurements. I n terms of specific surface area units, familiar in catalytic work, this corresponds to a magnitude of about 0.1 m.2/g. of sample. This would be exhibited by particles of about l o p in linear dimensions. Most of the electrical and optical investigations on zinc oxide have been performed, not on single crystals, but on compressed powders, evaporated layers, and sintered samples, all of which constitute geometries of high potential surface-to-volume ratios. The work done on evaporated layers of zinc oxide has involved layers whose thickness is about one micron or less, for which surface effects are obviously large. I n order t o show that surface effects may be large in sintered samples

SURFACE BARRIER E F F E C T S I N ADSORPTION

269

also, the structure of sintered samples must be considered. Sintered samples are prepared by compressing zinc oxide powder into a pellet and heating it t o a temperature of approximately 1000" t o promote fusion of the powder granules. A photomicrograph of a typical sintered zinc oxide sample, sintered in air a t 1000" is shown in Fig. 2. The specimen was prepared for observation by boiling the sample in a solution of an organic dye, to expel the air in the pores, and allowing the sample to cool in the dye, which was then sucked into the pores. The sample was then cut and polished for observation. The white portions of the photomicrograph show the grain structure, the dark portions are the dye.

FIG.2. Photomicrograph of sintered zinc oxide. Magnification: 800 X.

It is observed in the photomicrograph that the grain diameter is on the average about 20 microns. The conductance of the grains thus will be affected only a few percent by surface trapping. However, i t has been shown by Hahn (24) that the electrical resistance in sintered zinc oxide arises, riot i n the grains, but a t the fused intergranular contacts, or ('necks." He demonstrated that the resistance decreased markedly with increasing frequency of an applied ac voltage. The intergranular resistance is in parallel with the intergranular capacity, and as the frequency becomes high, the intergranular impedance tends to zero. As may be seen in Fig. 2, the cross-sectional area of the necks is much less than that of the grains. Thus the highly resistant parts of the sample, the necks, have a high surface-to-volume ratio, and the conductance of the sample may be very much influenced by the state of the surface. As evidence that surface effects do control the resistance of sintered zinc oxide (and hence probably the resistance of evaporated layers),

270

S.

ROY MORRISON

Morrison (31) has compared measurements of the Hall effect and of the resistance. The Hall voltage is inversely proportional to the average concentration of carriers in the material (24), and so, for zinc oxide, will be inversely proportional to the concentration of carriers in the large grains (Fig. 2 ) of the material. Figure 3 shows an example in which the resistance and the inverse of the Hall voltage measured on a sintered sample of zinc oxide are plotted as functions of the time. This illustrates that the number of carriers in the bulk of the sample may remain relatively constant, while the conductance varies widely, all at constant temperature.

-

T?

.033-5

T=IOOOC.[From T=2 3 V )

bx-x=Condrrctonce in arbitmry unrts. KN.---.=yV, measure of number of carriers. V” =Hall voltage in arbitrary units.

).

--4---,.015 \X-xx

XX

.r

-

l : x Time in hours.

b i kii IhIh

’

l b ~ l ~ 4 ~ 7 $ O ~I 3I ~ I 6 i FIG.3. Variation with time of the conductivity of zinc oxide.

The conductance is proportional to the number of carriers in the neck (as was shown above) and t o the mobility of these carriers. Thus, unless one makes the unusual assumption that the bulk properties of the neck are very different indeed from those of the grain, or that the bulk electron mobility varies widely with time a t a constant low temperature, the conductivity must be controlled by surface effects. T o conclude this section, evidence will be presented that the surface effects observed are connected with adsorbed oxygen. The existence of other deep surface levels, for example “Tamm levels’’ (discussed in the preceding section), on the surface of zinc oxide is placed in doubt by a n experiment of Bevan and Anderson (32) on sintered zinc oxide. They observed that the activation energy of the conduction electrons (of the order of an electron volt when the sample is subjected t o high oxygen pressure) decreases t o a few hundredths of a n electron volt if the measurements are taken a t low pressure (less than mm.) and high temperature (the order of 600°C). Surface traps other than those asso-

SURFACE BARRIER BARRIER E E FF FF E EC CT T SS II N N ADSORPTION ADSORPTION SURFACE

2711 27

ciated with adsorption, such as Tamm traps or impurity traps) would not be expected t o be removed under these circumstances, and one has t o conclude that either they are not there in large concentrations, or that they have an energy level close t o the conduction band. We may therefore conclude that the surface levels controlling the conductance are associated entirely with adsorption, thus simplifying the problem considerably. Bevan and Anderson (32) deduced that adsorbed oxygen controlled the resistance of sintered zinc oxide a t temperatures between 500°C and 1000°C. This conclusion was based on the observations that (1) the oxygen pressure had a reversible controlling effect on the resistance down to 500"C, too low a temperature for thermodynamic equilibrium t o be Volume Adsorbed

0.10-

I'/

(Cm3 at N.T.P)

0.m:

3

//+

\ +\,

FIG.4. Adsorption of oxygen on zinc oxide. Sample was cleaned a t 600°C for 1 min. before lowering the temperature to T oand admitting oxygen.

attained between the gas and the bulk zinc oxide during the time of experimentation; (2) the conductance varied with the pressure of oxygen according t o the relation u = APo8-$i,as follows from the reaction

occurring at the surface; and (3) there exists a similarity between their results on zinc oxide, and those of Garner et al. (2) on copper oxide, where a similar mechanism was proposed. Morrison and Miller (33) have made some direct measurements on the adsorption of oxygen on zinc oxide powder (Fig. 4), which seem in accord with the conductivity measurements of Bevan and Anderson on sintered zinc oxide. The reversible region of chemisorption was shown t o be above about 450°C) corresponding t o the reversible region of conductivity versus oxygen pressure found by Bevan and Anderson t o be above about 500°C. Figure 1 is an energy level diagram showing a proposed model for the band structure of zinc oxide. The valence band and conduction band are shown separated by a forbidden gap. Two levels which correspond t o the trapping of two electrons by the interstitial zinc are indicated in the forbidden gap. Surface levels associated with adsorbed oxygen are shown.

272

S . ROY MORRISON

The oxygen ions, as drawn, have two acceptor traps, corresponding t o the possibility of double ionization. The evidence for this double ionization of the adsorbed oxygen, presented in Section IV, is, however, inconclusive. The energy values shown in Fig. 1 will be derived in the following sections. A brief summary of their origins is given in Section IV,2e. I n this section, we have presented evidence t o indicate that adsorbed oxygen on the surface may be a controlling factor in the electrical properties of zinc oxide. I n the following section, the electron transfer theory of adsorption will be discussed. The properties then will be examined in more detail, agreement of the results with those predicted by the adsorption theory will be emphasized, and the contribution of those that have been studied t o a more complete understanding of adsorption phenomena will be discussed. ADSORPTION AND THE ELECTRONIC STATE IV. IRREVERSIBLE OF THE SURFACE 1. Electron Transfer as the Rate-Limiting Step in Chemisorption

I n Section II,3 we have shown how a surface barrier affects equilibrium adsorption. I n the following, we will indicate the important role it may play in irreversible adsorption. Weisz (22) and Morrison (31) have pointed out that if there is a barrier at the surface of a semiconductor, and if electrons from the bulk of the material must cross this barrier for chemisorption to occur, then the rate of chemisorption may be limited by the rate at which electrons can cross the barrier. The rate a t which electrons can cross a barrier is proportional t o the number of electrons with energy greater than the barrier height, or R a exp (-E2 p ) / k T (6)

+

where R is the rate of adsorption, E Zis the height of the barrier (Fig. l ) , k Boltzmann’s constant, and T is the absolute temperature. At arbitrarily low temperatures and for large Ez, it is apparent that the rate may become negligibly small. I n the following discussion (Morrison, 31), we will consider again the adsorption of an electronegative gas on an N-type semiconductor, where the only surface levels are those of the adsorbed gas (Fig. 5 ) . At high temperatures, in the region of equilibrium adsorption, the rate of adsorption and desorption is high. As the temperature is lowered, the amount adsorbed in equilibrium becomes greater, so that Ez increases. From Equation (6), as EZ/T increases, the rate of adsorption decreases rapidly. There may be some temperature a t which equilibrium adsorption

SURFACE BARRIER EFFECTS I N ADSORPTION

273

will not be attained in the course of a n experiment, due to the decrease of R of Equation (6). This temperature will be denoted by T‘. Consider now temperatures below T’ in the nonequilibrium region of chemisorption, and let us assume that electron transfer over the surface barrier is rate-limiting. We will examine the case in which initially the surface is completely free of the adsorbed species; it has been heated t o high temperature, well above T’, a t a low pressure t o remove essentially all the adsorbed gas. The energy bands will be straight out to the surface, and no surface barrier will exist. If the sample is quenched t o a low temperature, well below T‘, and the gas pressure is increased, adsorption will commence. Initially it will be very fast, since the surface barrier, Ez, is

FIG.5. Dipole barrier formed by chemisorbed oxygen.

small. However, as adsorption occurs, E2 will increase, and the adsorption will pinch itseIf of, by increasing the height of the surface barrier until essentially the rate a t which electrons can cross is zero. If we consider the amount adsorbed by the techniques described in the preceding paragraph as a function of temperature, it is apparent that in the nonequilibrium region of chemisorption the amount adsorbed will increase with increasing temperature. As the temperature is increased, “pinch off” will occur a t a larger E2,and more adsorption is required t o produce the larger Ez. This, of course, is observed experimentally, as illustrated, for example, in Fig. 4. There are several adsorption phenomena which may readily be explained using this surface barrier model. Taylor’s activation energy (34), in “activated adsorption,” in some cases may be, or may be related to, the surface barrier. An increase in activation energy with surface coverage then corresponds to the increase in the surface barrier height Ez with adsorption. If quasi-equilibrium corresponding to pinch-off has been reached a t a given temperature, and the temperature is then lowered, no further adsorption will occur. The rate of adsorption, as given by Equation (6), will decrease as the temperature is lowered, a t a constant value for Ez. No desorption will occur, obviously, since the sample has less than a n equilibrium amount of gas adsorbed. It is of interest to consider one further effect of this model, which will

274

S. ROY MORRISON

be used in the following work. That is, if the quasi-equilibrium, due to the pinch-off effect, has been reached a t some temperature T , then a n increase in the temperature will allow further adsorption. This is because a t the temperature T the rate of adsorption as given by the above equation was slightly below that possible to observe in a reasonable time. Raising T will increase the rate of adsorption t o a n observable value, and more adsorption will be measured. This last effect is, of course, quite familiar in many cases of chemisorption. Our model indicates th a t the effect should also be observed in resistance measurements. A more quantitative analysis for the low-temperature irreversible region of chemisorption, has been made by Melnick (11) with the use of the same model (Fig. 5 ) . He has shown that the Elovich Rate Equation (35) follows from the model. The Elovich Equation

5%

dt

= a exp ( - b q )

is a n empirical relation which describes the rate of adsorption as a function of the amount adsorbed, q , and two constants, a and b. It has been shown (36) t o apply t o many cases of chemisorption. Melnick has shown that dq = B exp(=)- E2 dt

where B varies slowly with adsorption as compared with exp

(2)

The variation of E z with q is a function of the variation of the concentration of donor atoms with dist,ance from the surface, but Melnick assumes that for small changes of g, using q = go Aq, Ez may be expanded in a Taylor’s series in Aq, which, substituted in the above expression, yields

+

d(Aq)/dt

= C I C X (~- ~ z A g )

Thus if Aq is small compared with q, the adsorption follows the Elovich Equation. Melnick points out that the Elovich Equation agrees with experiment only if q is large, the adsorption rate being much too high for small p. Thus the model of Fig. 5 is in agreement with the Elovich Equation over approximately the range of its application. I n the above, we have discussed a theory of chemisorption. It is based on averg simple model of the electronic levels and of the possible reactions. It should be pointed out that in the above, the tacit assumption has been made that the electron transfer process is rate-limiting. This must be verified or assumed for each system studied, if the theory is t o be applied, A Schottky surface barrier was assumed throughout. The theory

SURFACE BARRIER EFFECTS I N ADSORPTION

275

cannot be expected t o apply t o adsorption on clean metals or on semiconductors, if no barrier is produced through adsorption. It has been assumed also that the adsorbed species of interest provide the only surface energy levels. Many variations of the theory are possible, if other surface levels of other types are present. 2. The Relation between Adsorption and Conductance in Z i n c Oxide

a. Introduction. I n Section I11 of this article, we have presented a brief outline of some of the properties of zinc oxide which have led to a belief that oxygen adsorbed on the surface has an important effect on the electrical properties of zinc oxide. As oxygen ions become adsorbed on the surface, creating surface energy levels and trapping conduction electrons from the interior, the conductance will decrease. The conductance of the sample will thus reflect the amount of adsorption which has occurred. I n this section, considerable space will be allotted to a n examination of the conductivity effects in order t o indicate how surface effects and adsorption may influence conductivity measurements, and to attempt t o correlate the many anomalous electrical effects on zinc oxide with similar anomalous effects often observed in adsorption studies. An examination of Fig. 4, an adsorption isobar of oxygen on zinc oxide, shows that equilibrium adsorption occurs at temperatures above about 450°C. I n the following we will discuss under separate subtitles the conductance characteristics (a) in the equilibrium adsorption region and (b) in the nonequilibrium, or irreversible adsorption region. I n general, zinc oxide which has been prepared for conductance experiments has been exposed to the atmosphere a t room temperature for a length of time sufficient for the quasi-equilibrium adsorption of oxygen characteristic of room temperature t o occur. When the sample has this amount of oxygen adsorbed and the temperature is lowered, no further adsorption or desorption should occur, as discussed in the preceding section. If, on the other hand, the temperature is raised somewhat, adsorption may occur. Thus it will be convenient to further subdivide the discussion of the irreversible adsorption region, and t o consider separately the temperature regions below and above room temperature. b. Low-Temperature Conductivity (Below 20°C). Work in this temperature range has been done mainly by Fritsch (37) and Harrison (26). The former measured the conductivity of evaporated layers and sintered zinc oxide as a function of temperature, as well as the Hall coefficient at 25°C and -190°C. The slopes of both of these parameters, plotted as logarithms against the inverse absolute temperature, agree within 20 percent. The slope, which gives the ionization energy of the donors, decreases with increasing conductivity, in rough agreement with

276

S. ROY MORRISON

Meyer’s Rule (30) for oxide semiconductors. As the concentration of interstitial zinc donors increases, the ionization energy decreases. The mobility, on the other hand, varies from sample t o sample, from 0.6 t o 30 cm.2/volt-sec. a t room temperature. Harrison found essentially the same results using single crystals and sintered samples. The slopes of the logarithms of the Hall constant and conductivity plotted against inverse absolute temperature were practically identical. The mobility varied in magnitude from sample t o sample, but not in temperature dependence. Little time dependence of the conductivity was observed and none for the Hall coefficient. These results indicate that surface effects were not affecting his results to any extent a t temperatures below room temperature. If surface traps had been involved, they would have affected the conductivity of sintered samples t o a larger extent than the Hall coefficient. However, the temperature dependence of the Hall coefficient corresponds closely t o th a t of the conductivity. Considering the method of preparation of these ZnO samples, these results correspond to what one would expect on the basis of the adsorption model. The samples were sintered or evaporated a t some high temperature, and then cooled to room temperature in air. As discussed in Section IV,l, adsorption will occur until the rate of electrons crossing the surface barrier is pinched off t o zero a t room temperature. If the temperature is now lowered below room temperature, no electron transfer will be possible between the surface level and the bulk of the solid due to this high surface barrier. Thus the surface levels will be isolated and unable t o affect the conductivity, which will therefore reflect bulk properties of the zinc oxide. The variation in the measured electron mobilities from sample to sample in sintered materials (also observed by Hahn, ref. 24), may be due to any of several effects. The most probable reason for this variation in the well-sintered samples studied is a difference in history; the individual samples are obtained with different numbers of conduction electrons per cm.3 “frozen in” in the necks. That is, the different history has allowed different amounts of oxygen to be adsorbed on the surface. Thus the concentration of electrons in the grain, as measured by the Hall coefficient, will have little relation to the concentration of electrons in the neck, as measured by the conductivity, and the “mobility,” obtained from the product of the Hall coefficient and the conductivity, will be neither the true mobility nor constant from sample to sample. The different samples may also end up with varying geometry of their necks, according to their previous treatment. It is of interest to note, a t this point, the observation of Miller (38) t o the effect t ha t the mobility of electrons in the large grains of sintered

SURFACE BARRIER EFFECTS I N ADSORPTION

277

zinc oxide (as determined by high-frequency conductivity measurements) is comparable t o the mobility in single crystals. I n both of these cases, surface effects may be considered unimportant, so the mobility measured in these cases will be the true mobility of electrons. The energy for single ionization of the interstitial zinc impurities, as determined from the slope of the graph of the logarithm of the conductivity or Hall coefficient against the inverse temperature, appears t o be a few hundredths of an electron volt, close t o 0.04 e.v. From the above discussion it would appear that a t these low temperatures only bulk processes are occurring, so that this activation energy, 0.04 e.v., is a true bulk property. The impurity concentration in the samples with which Harrison worked was of the order lo1' em.+ t o 10l8 ~ m . - ~As . was discussed earlier, the ionization energy of the interstitial zinc appears t o increase with decreasing concentration of impurities (14). c. High-Temperature Conductivity (Above 500°C). I n this temperature range, as is evident from Fig. 4, chemisorption of oxygen on zinc oxide is a t equilibrium. That is, adsorption is a reversible function of temperature and oxygen pressure, and the conductivity is also found t o be reversible. If the temperature is about 5OO0C, however, the conductivity is not instantaneously reversible as a function of pressure; it is time dependent and approaches its new equilibrium value slowly. Another anomalous feature, discussed below, is that a t certain pressures, even a t temperatures well above 500"C, the conductivity is not instantaneously reversible as a function of pressure. I n general, however, the conductivity parallels chemisorption of oxygen, and is reversible for both temperature and pressure changes above 500°C. According t o Stockmann (39) the conductivity of ZnO samples, studied a t the partial pressure of oxygen in air of one atmosphere, varies with the temperature in a fairly reproducible manner in this range. The variation of the Hall coefficient and the conductivity of zinc oxide as a function of oxygen pressure has been studied by many workers (27,32,40-42). The conclusions reached are t h a t the number of carriers varies with oxygen pressure approximately according to the equation

where A is a constant, and m is of the order of 4 for the conductivity measurements and of the order of 6 for the Hall effect. The conductivity of zinc oxide, when the sample is exposed t o high temperature and various gaseous ambients, may be affected not only by adsorption a t the surface, but also as a result of diffusion. I n a reducing atmosphere, or a t low oxygen pressure, oxygen a t the surface may become

278

S.

ROY MORRISON

dissociated and the excess zinc may diffuse to the interior and occupy interstitial sites. The reverse may occur at higher oxygen pressure; interstitial zinc may diffuse to the surface and combine with oxygen to form ZnO. As pointed out by Bevan, Shelton, and Anderson (40),thermodynamic equilibrium is not attained between the bulk and the ambient at temperatures below about 850°C. Bevan and Anderson (32) conclude that at lower temperatures, it is the effect of adsorbed oxygen which controls the conductivity. At very low oxygen pressures, Bevan and Anderson found that the graphs of conductivity against inverse temperature yield an activation energy of a few hundredths of an electron volt, and that the conductivity becomes independent of oxygen pressure. This activation energy is similar to that for ionization of the electrons from the interstitial zinc atoms, as obtained from low temperature and single-crystal measurements. This may indicate that the surface effects become unimportant at very low pressures. Bevan and Anderson point out an interesting feature of the transition from surface-controlled conductivity to bulk-controlled conductivity. There are two segments of the pressure dependence of the conductivity. At pressures greater than mm., Equation (7) described above, is obeyed. At pressures less than mm., however, the conductivity is high and independent of the oxygen pressure. This is the region in which bulk effects are assumed to control the conductivity. The transition between the two states is slow, and is characterized by hysteresis. That is, if the pressure is lowered to the transition region, the conductivity slowly rises, possibly because the adsorbed oxygen is being desorbed slowly. If the initial pressure has been low and the sample in the state of high conductivity, and the pressure is then raised, the conductivity slowly drops, oxygen apparently becoming slowly adsorbed. It must be pointed out that this occurs a t temperatures at which the high-pressure changes in conductivity with pressure are practically instantaneous, so that the rates of adsorption and desorption of oxygen cannot be the limiting factors. This interesting effect may be related t o an effect observed by Morrison and Miller (33) in their studies of the adsorption of oxygen on zinc oxide. In order t o remove previously adsorbed oxygen, the zinc oxide was heated to the high-temperature and low-pressure range in which Bevan and Anderson observed the anomalous behavior of conductivities. Morrison and Miller observed that samples adsorbed less oxygen at a relatively low temperature and high oxygen pressure, when they had been preheated longer. This indicates that the number of “adsorption sites” was slowly decreasing while the sample was held a t a high temperature and at a moderately low pressure. They showed also that the adsorption sites

SURFACE BARRIER EFFECTS I N ADSORPTION

279

could be slowly regenerated by subjecting the samples t o a high oxygen pressure and a high temperature. It was suggested by Morrison and Miller th a t these losses and gains of adsorption sites may be due to the diffusion of zinc atoms in the following manner: high oxygen pressure causes some adsorption at the high pretreatment temperatures where the coulombic attraction between the adsorbed oxygen and the interstitial zinc may draw such interstitial zinc toward the surface. At the low temperature, more adsorption of oxygen will then occur on this surface enriched in interstitial zinc. Conversely, if the oxygen has been desorbed a t the high temperature, the adsorption centers, the interstitial zinc, may diffuse away from the surface or combine with adsorbed oxygen, and the number of sites for the low-temperature adsorption will decrease slow with time. An analysis of this concept produced reasonable agreement with the experimental observations of adsorption. However, this postulate, although it fits Bevan and Anderson’s effect below 10-2 mm., does not seem adequately t o explain the results above mm., where no time dependent effect was observed. It is of interest t o note, however, th at Bevan and Anderson suggest that their resistance-time curves in the transition region are indicative of a diffusion-controlled process. A more intensive experimental and theoretical study of the behavior of ZnO in the temperature range of L 500°C, where equilibrium is reached for the adsorption of oxygen, may lead to interesting conclusions regarding the character of this adsorption. For exampIe, it should be possible t o obtain a value for the energy Es. The theory a s described in Section I, however, is very complex, and further interpretation of the few results available will not be attempted here. d. Intermediate Temperatures (20°C to 50o0C). T h a t this temperature range should be most difficult to interpret from the basis of conductivity measurements alone is t o be expected from consideration of the adsorption-temperature curves of Fig. 2, and from a knowledge of adsorption effects. Here we are in the range of irreversible adsorption, where the amounts adsorbed depend upon the past history of the sample. I n certain investigations of this temperature range (24,43), no consideration was given t o adsorption effects, with the result th a t no reproducible values for the resistance of the ZnO samples were observed on raising and lowering the temperature. Moreover, the resistance was found in these studies to vary with the time a t a constant temperature. The first published attempt to obtain reproducible results despite this time dependence was that of Stockmann (39). He obtained reproducible results by keeping the rate of heating his samples (evaporated films 0.1 micron thick) constant. He found th at the conductivity steadily increased

280

S. ROY MORRISON

with temperature to 250"C, a t which temperature a maximum occurred followed by a decrease to a minimum at about 400°C. Intemann and Stockmann ( 2 5 ) repeated these measurements and simultaneously studied the Hall effect. The latter showed the same dependence of the number of carriers on time and temperature as did the conductivity. This parallelism between the Hall effect and conductivity is to be expected in very thin films, since the latter will be greatly influenced by surface effects. The mobility, of course, will be different from the electron mobility in a single crystal, due t o the smaller mean free path of electrons in very thin samples. If the ZnO sample was heated at 250°C in hydrogen before the measurement was made, a different curve was obtained, with the conductivity maximum a t 150°C. After the sample was heated a t 400°C in hydrogen (less hydrogen adsorbed, according to Taylor and Strother (44)), a curve showing two conductivity maxima, one at 250°C, and one a t 150"C, is obtained. Using ZnO films of the same thickness, a similar technique was employed by Fritzsche (45), except that this author made his measurements in a vacuum. I n these experiments, the conductivity increased continuously as the temperature was increased a t a constant rate. If the temperature was kept below 3OO0C, the conductivity was reversible as a function of the temperature. If the sample was heated t o 450"C, the conductivity remained high upon successive cooling in a vacuum. The longer the sample had been kept in a vacuum a t 450"C, the higher became the conductivity and the higher it remained upon cooling. Suggestions as to the interpretation of these measurements on the basis of the adsorption theory will be presented later. Morrison (31), using sintered zinc oxide, applied a different technique t o study the conductivity effects in the range between room temperature and 500°C. He studied the variation in conductance as a function of time with the temperature held constant. Figure 3 shows one such conductivity-time experiment. The sample used was a slab of zinc oxide cut from a pill which had been compressed and sintered in air for eighteen hours a t 1000°C. The sample was immersed in oil (the oil does not penetrate into the pores of the sample) a t the start of the run. The sample container was immersed in boiling water, the temperature reaching 100°C in the order of one-half of a minute. The conductance was recorded as a function of time while the sample was held a t 100°C. The results are shown in Fig. 3. The inverse of the Hall voltage is also plotted as a function of time. An interpretation of the Hall measurement is discussed in Section 111. It is seen that the conductance steadily increases to a maximum

SURFACE BARRIER EFFECTS I N ADSORPTION

28 1

within about two hours, then decreases to a value below the original. Curves showing this effect were obtained using a continuous ac current, as well as no current (except during the short intervals when readings were taken). Continuous dc current appears t o affect the curves slightly, causing the maximum to be observed earlier. The decrease in conductivity as a function of time is what one would expect from the theory given in this section, the decrease being attributable t o the adsorption of oxygen following the temperature change from 20°C t o 100°C. The initial increase is not expected on the basis of this simple theory . It was found th at the initial rate of increase was lower if, instead of raising the temperature from 20°C to 1OO"C, the temperature was raised by a smaller increment. Below about 8O"C, the initial increase in conductance is no longer observed, and is replaced by a slow monotonic decrease of the conductance with time. I n modifying the experimental technique, other conductance measurements were taken with the ZnO sample in air, instead of in oil. This allowed exchange of the oxygen in the pores of the sample with the atmosphere. I n this case the initial rate of rise of conductance was about equal t o that shown in Fig. 3, but the maximum of conductance occurred a t an earlier time (after about twelve minutes), and the change in conductance a t the maximum amounted t o only about 10 percent of th a t in Fig. 3. This may indicate that the factors tending to decrease the conductivity are more effective when the sample is exposed to the atmosphere, in accordance with the hypothesis that the decrease of the conductance is due t o chemisorption of oxygen. If the sample is in air, the oxygen chemisorbed by the zinc oxide from the pores of the sample can be replenished by the oxygen present in the surrounding atmosphere, whereas the sample immersed in oil cannot receive exterior oxygen, with the result that the oxygen pressure decreases as chemisorption proceeds. A simple calculation of the amount of oxygen present in the pores compared t o the approximate amount which must be chemisorbed to give the resistance change observed indicates that this assumption is reasonable. If this explanation is correct, it would appear that when a run of the type shown in Fig. 3 is concluded with the sample in oil, the oxygen pressure is very low, and with further heating of the sample, little chemisorption occurs. This is actually observed, as discussed below. Measurements also were made on samples sintered in a vacuum of cm. for eighteen hours a t lOOO"C, then cooled in a vacuum t o 500"C, and quenched t o room temperature in air. Samples prepared in this manner have a much smaller grain size, of the order of one micron. The surface density of chemisorbed oxygen would

282

8. ROY MORRISON

be expected t o be somewhat smaller than on the air-sintered samples, from the work of Morrison and Miller (33). As described earlier, these authors found that holding the sample in a vacuum at high temperatures apparently decreased the surface density of the adsorption sites. Using the same experimental procedure, it was found th a t the vacuumsintered samples produced curves similar t o that shown in Fig. 3. It was observed that if a vacuum-sintered sample was placed in oil or sealed in alundum and water-glass to limit exchange of oxygen in the pores with the atmosphere, the irreversible decrease in conductance was eliminated. This procedure consisted of heating the sample in the closed vessel to lOO"C, and allowing the sample t o remain a t th a t temperature for several hours to complete the conductivity cycle as shown in Fig. 3. When the sample in this condition was further heated t o a higher temperature, a slow increase of its conductance occurred, but little succeeding decrease. This behavior of the sample was reproducible on renewed heating and cooling. The suppression of the irreversible decrease in conductance under these experimental conditions was ascribed t o the adsorption, during the first heating period, of the oxygen left in the pores of the sample, when the sample was heated to the higher temperature. Such a temperature-time-conductance curve for a vacuum-sintered sample is shown in Fig. 5. A sample, pretreated a t 100°C as described above, was heated to 125°C at time zero. The conductance changes as a function of time were followed for 160 minutes (point I?), and then the temperature was suddenly raised to 150°C. The time-dependent conductance changes proved to be quite reversible, contrary t o the irreversible decrease of conductance with increasing temperature for samples kept in a n oxygen atmosphere. e. Theoretical Considerations. Three possible explanations to account for the surface conductivity on zinc oxide have been suggested. Hauffe and Engell (46) have suggested a hypothesis th a t may apply t o compressed powders. They suggest that the surface barrier due to a n adsorption layer must be traversed by conduction electrons. Thus, if many grains of powder are pressed together, each with a barrier layer on its surface, electrons would require an activation energy at least equal t o the height of the potential barrier t o pass between grains. This evidently does not apply to sintered zinc oxide, since a t low temperature its conductivity shows an activation energy of a few hundredths of a n electron volt, the same as is shown by the Hall coefficient (26). independent of the previous treatment of the sample. This, then, indicates tha t the grains of sintered zinc oxide are actually fused, rather than merely touching.

SURFACE BARRIER EFFECTS IN ADSORPTION

283

Fritzsche (45)has suggested th at the observed effects may be due to absorption of oxygen by the samples. This concept, as he points out, can explain most observed effects. The major difference between this concept and the adsorbed oxygen concept is the following. The adsorbed oxygen concept uses the barrier layer to explain the time dependence of the conductivity; the absorbed oxygen concept uses the rate of diffusion of oxygen into and out of the sample to explain the time dependence. Thus the absorbed oxygen theory allows only a small temperature range for time effects, namely the temperature range in which diffusion is measurable but not instantaneous. We will attempt t o show th a t this temperature range is not wide enough to include many of th e time-dependent effects observed. The activation energy for diffusion, calculated by Fritsche using his model, is 2 t o 4.5 e.v., averaging 3 e.v. The value of the diffusion constant a t 460°C is lO-'4 cm.2 set.-'. Then the diffusion constant, even if we use the most favorable activation energy, 2 e.v., becomes, as a function of temperature, D = lo-' exp ( - 2 / k T ) cm.2/sec. Using the relation X = (Dt)36as the average distance which a n oxygen atom will diffuse into the zinc oxide in time t, it is found th a t a t 100DC it would take 10" see. for such an atom to diffuse one Angstrom. Thus the absorption theory cannot explain either results of the type shown in Fig. 3 , or the experiments of Heiland (47) and Melnick (11) a t room temperature. I n these measurements, which will be discussed in more detail in Section V, it was found that if the sample has been prepared by subjecting it t o irradiation in a vacuum, the conductance is high. When oxygen is allowed to reach the sample a t room temperature, however, the conductance decreases considerably. According t o the above calculations, absorption of oxygen cannot occur a t room temperature. Yet a n ambient oxygen-containing atmosphere may produce large changes in resistance, and so it must be concluded that adsorption of oxygen is important. The third hypothesis t o explain the observed conductivity effect is that of Morrison (31) who used a modification of the adsorption theory presented in Section 111. As has been pointed out, the adsorption theory in its basic form is adequate t o explain the high-temperature conductivity and the low-temperature conductivity in zinc oxide. However, it must be expanded slightly t o present an adequate explanation of the conductivity effects in the intermediate temperature range. The adsorption theory predicts the slow irreversible fall in the conductance shown in Fig. 3, but does not predict the initial reversible rise, observed in experiments such as illustrated in Fig. 3 for temperatures above 80"C, and which is shown isolated in Fig. 6. From Fig. 3 it is evident, upon comparison with the Hall voltage, th a t the reversible time dependence of the conductance is also associated with

284

S. ROY MORRISON

the surface. If the initial rise were associated with bulk effects, one could expect the Hall voltage t o change by the same factor as the conductance. There are no deep Tamm traps on zinc oxide, as has been discussed in Section 111. It is therefore suggested that the surface levels involved are higher levels on the adsorbed oxygen, which makes possible the transfer of a second electron to the adsorbed oxygen. That two electrons are transferred t o the adsorbed oxygen was assumed by Bevan and Anderson (32), who concluded this from a hypothesis based on conductivity measurements carried out a t high temperatures. .020.019

-

.om0180-

7 0

-b2

,0175,0170,0165-

.0160.0155-,w-~--X

125°C.

FIG.6. First temperature-cycle variation of conductance with time and temperature.

It should be pointed out, however, that the results shown in Fig. G are also compatible with a model involving acceptor-type adsorption levels associated with a gas other than oxygen which, a t around 100°C: or somewhat lower, is reversibly adsorbed on zinc oxide. It is not obvious, however, that a gas with such characteristics is present in air. A possibility might be water vapor, which is adsorbed by zinc oxide in this temperature range (29), and particularly strongly a t room temperature. Melnick (11) has found, however, that a t room temperature water vapor adsorption has no effect on the resistance of zinc oxide. This seems t o indicate that there is no electron transfer involved in this specific chemisorption, in contrast to the adsorption of oxygen a t room temperature, which, according to Melnick, changes the resistance of zinc oxide considerably. I n the following discussion we will assume th a t the higher energy level on the surface is associated with the second electron of the adsorbed oxygen ion. The same arguments as brought forward for this specific model would apply with few changes if the level were due to a n adsorbed gas other than oxygen.

SURFACE BARRIER E FF E C T S I N ADSORPTION

285

The electron affinity of oxygen for its second electron is smaller than for the first. Accordingly, the trap level is assumed to be higher, a t a level E l below the conduction band, as shown in Fig. 7. A further assumption is made; namely th a t the 0-- traps are in equilibrium with respect t o the transfer of electrons between themselves and the conduction band. It is conceivable that the 0-- levels are in equilibrium, whereas the 0- levels are not. This is to be expected when there is a larger surface density of empty 0-- levels, N1, than of physically adsorbed atoms, N z . As a simple model we will consider that empty 0- levels are physically

0-

level. FIG.7. Postulated energy-level diagram for zinc oxide, showing oxygen levels.

adsorbed atoms, whereas empty 0-- levels are the 0- chemisorbed ions. I n case the cross sections of the two types of oxygen, physically adsorbed and singly ionized, for capture of an electron are equal, the ratio of the rate of filling 0-- levels to that of filling 0- levels will be N I / N 2 . Then, although chemisorption may have a very low rate, the rate of transfer of electrons between the 0-- ions and the bulk will be greater th a n chemisorption by a factor of N1/N2. Hence equilibrium in the number of 0-ions will be reached although chemisorption, and the creation of 0- ions will, essentially, cease. With these assumptions, the results of Figs. 3 and 6 are easily explained. As with any level in equilibrium, increasing the temperature causes the 0-- level to lose electrons to the conduction band, and the conductance rises. Figure 3 shows the two competing slow reactions, the one involving 0-- levels, and the other, the chemisorption of oxygen. The former tends to increase the conductance, the latter t o decrease the conductance of the catalyst. As the system approaches equilibrium with respect t o the 0-- levels, the influence of further chemisorption becomes important, and the conductance after passing through a maximum, decreases. The final value of the conductance is always smaller than the initial value, because a t the higher temperature, a higher value of the barrier height E2 is attained before pinch-off occurs, and with a higher E2, more electrons are trapped in the surface traps.

286

S. ROY MORRISON

As E l , the energy difference between the 0-- levels and the conduction band, is greater than Ez, the rates of the two competing processes have a different temperature dependence. It is found, as described earlier, th a t a t about 80°C the rates are equal, and below this temperature the initial rise in conductance is not observed. The curve of Fig. 6 shows the “model case” when there is no chemisorption t o compete with the attainment of equilibrium with respect t o the 0-- levels. The energy E l can be estimated from these results. Considering the model shown in Fig. 7, the rate a t which electrons will leave the 0-- levels will be

R

=

A exp (- E l / k T )

(8)

Let us assume, as a first order approximation, that A is a constant, and that when the temperature is suddenly increased, R is much larger than the expression describing the rate a t which electrons are trapped in the 0-- levels. Then, setting t equal to zero at every point a t which the temperature is suddenly raised (i.e., the points A , B, and C of Fig. 6), we can plot (log R)t=o against the inverse temperature, and obtain the activation energy E l. Similarly, considering an equation of the same type as Equation (8), one can, from points D , E , and F , obtain a value for EQ. The graphs of (log R)t=oagainst the reciprocal temperature for two such curves as in Fig. 6 are shown in Fig. 8. The change in electron mobility with temperature has been considered. As the zinc traps are a t a level only a few hundredths of an electron volt below the conduction band, they are nearly all ionized at these temperatures. The results show El = 0.7 e.v., and Ez = 0.55 e.v. A consideration of the variation of the “constant” A in the equation above indicates tha t the value for E l is too low, and that for Ez is too high. It is probable tha t E lis of the order of 0.8 e.v. and th at Ez for this particular sample is of the order of 0.5 e.v. It must be emphasized again that these discussions on the conductivity of zinc oxide have been based on the two assumptions (1) th a t the chemisorption of oxygen has a strong effect on the resistance of the zinc oxide and (2) that the rate-determining step in the chemisorption of oxygen is a n electron transfer. The former assumption has been based on indirect arguments, as well as on some direct experimental evidence. The latter assumption is based on the ability of the model t o explain in a simple consistent manner the various experimental effects, some of which have been described, and some of which will be, in later chapters. Studies of the conductivity of adsorbents as a function of the amount of chemisorbed gas promise to be of considerable value in adsorption

287

SURFACE BARRIER E F F E C T S IN ADSORPTION

studies. Conductivity measurements can serve as a simple and sensitive tool which is capable of following adsorption far more accurately than direct volumetric measurement. An advantage of the conductivity measurement over the direct adsorption measurements is the possibility of measuring reactions occurring a t surfaces which involve no actual adsorption or desorption, but merely change the state of excitation of adsorbed ions, and thus change their reactivity in a catalytic conversion.

15 -

J

1.0 -

0.5 I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

f.9 2.0 2.1 22 2.3 2.4 25 26 27 2.8 2 FIG.8. Rate of change of conductivity of ZnO vs. temperature, assuming

The energies are noted on the graph. Temperature was increased through l(a), then decreased through 1 ( b ) . Similarly, the temperature was increased through 2(a), then decreased through 2 ( b ) in steps of 25".

From the various concepts developed in this and the preceding section, it would appear that an energy-level structure as shown in Fig. 1 may explain the gross effects observed for the conductivity, fluorescence, and photoconductivity of zinc oxide. To summarize the various results which suggest the energy level diagram of Fig. 1, many authors have shown (24,26,28) t h a t zinc oxide has interstitial zinc as a donor impurity. As determined by conductivity and Hall effect measurements, the energy level for single ionization of this interstitial zinc is of the order of several hundredths of an electron volt below the conduction band when the concentration of donors is of the order of lo1' ~ m . 1 The ~ . energy level for double ionization, from optical absorption measurements, appears to be a t about 3.2 e.v. below the con-

288

S . ROY MORRISON

duction band (28)- For recombination, from measurements of the fluorescence, the level appears to be about 2.2 e.v. below the conduction band (58,30). These investigations of the fluorescence will be discussed further in Section V. The valence band appears to be 3.2 e.v. below the conduction band, as concluded from measurements of the absorption of light by single crystals (28,29). Evidence has been presented (31,32) that oxygen is chemisorbed on the surface of zinc oxide. The energy level for the first electron is denoted by EB. From conductivity investigations, a second surface energy level, associated with adsorption, has been indicated with an energy level a t about 0.8 e.v. below the conduction band a t the surface. The hypothesis has been presented that this level is associated with double ionization of the adsorbed oxygen. The possible application of this model t o explain the effects observed for the adsorption of hydrogen on zinc oxide is presented in Section IV,3. f. Further Comparison with Experiment. The model proposed here explains qualitatively most of the conductivity measurements reported to date for zinc oxide a t temperatures between 20" and 500°C. It indicates why no simple correlation exists in certain cases. The time dependent reactions involving chemisorption and those attributed t o 0-- levels can become hopelessly complex if measurements are made without proper consideration of the past history of the sample, the time factor, the temperature changes, and the oxygen pressure. As an example of how earlier work can be qualitatively interpreted on the basis of this model, let us consider the results of Stockmann et al. (25,39) described previously. The specimens of these authors, prepared by oxidizing a film of zinc, were probably covered with chemisorbed oxygen corresponding to quasi-equilibrium a t a temperature considerably higher than room temperature. Stockmann found the conductivity changes reversible and in approximate agreement with Scharowsky's results on single crystals, up t o 250". Thus for these samples, it would appear that quasi-equilibrium adsorption of oxygen has occurred corresponding to 250"C, and that below 250"C, no transfer of electrons between the surface levels and the bulk zinc oxide is possible. Stockmann's specimens, when heated above 250"C, decreased in conductivity, as would be expected as a result of adsorption of oxygen. Above about 450°C, the conductivity was found to increase, probably due to the beginning desorption of oxygen a t this temperature. Lowering the temperature of specimens previously heated to 550-600°C resulted in a gradual decrease of conductivity, obviously caused b y increasing chemisorption of 0 2 , on cooling. 1

SURFACE BARRIER EFFECTS I N ADSORPTION

289

It is not evident, from this qualitative correlation, th a t the concept of doubly ionized oxygen is necessary to explain these results. It may be that the surface barrier, under the experimental conditions used in Stockmann's work, was of such magnitude th a t the 0- levels did not release many trapped electrons upon moderate heating. If the surface barrier is small, the 0-- levels will not be in equilibrium, in the case of the 0- levels. If the surface barrier is very large, the 0-- levels will tra p few electrons. I n either case, the 0-- levels will play a less important role in the low-temperature conductivity, and the results obtained are approximately what would be expected. Stockmann found that tempering zinc oxide in a hydrogen atmosphere a t 250°C increased its conductivity and changed the shape of the resulting time-temperature-conductivity curve. The effect of hydrogen on zinc oxide will be discussed in the next section. Fritzsche (45) used a technique similar t o that of Stockmann, but held the sample in a vacuum while taking measurements. His ZnO samples were prepared by evaporation in an oxygen atmosphere a t about 300°C. Again, the conductivity was found to be reversible below 300°C. I n these experiments, as the temperature was raised, the conductance increased monotonicly. Due t o the vacuum, only a desorption of oxygen was possible in these experiments. When the temperature was lowered, still in a vacuum after reaching 460"C, the conductivity remained high, as little readsorption occurred, due to the low pressure. A time dependent rise in conductance was observed if the temperature was kept a t 460"CJ as observed before by Bevan and Anderson (31), and attributed to the desorption of oxygen. Fritzsche, using the formula u = A exp ( - E / k T ) for the conductivity, A being slowly dependent on temperature and E a n activation energy (which he assumed to be th at for diffusion of oxygen) found in his investigations that E averages 3 e.v. We will attempt to show a correlation between ES, the 0- energy level, and the value of E calculated by Fritzsche. Assuming that desorption alone accounts for the conductivity rise (the contribution of the 0-- traps is negligible), the rate of desorption will be given by

R

=

Adu/dt

=

Bdc/dT

=

CI exp (-E,/kT)

-

Cz exp ( - E 2 / k T )

where 3 and A are constants, A / 3 is the constant rate of increase of temperature, ES is the barrier height, and C1and Cz are assumed slowly varying with temperature. Because the temperature is raised rapidly, this latter assumption may be approximately correct, and the rate of adsorption may be assumed small compared with the rate of desorption

290

S. ROY MORRISON

a t all times. An approximate solution to the remaining equation,

du/dT

=

CBexp (- E 3 / k T )

was obtained by Fritzsche to be u =

C4 exp ( - E , / k T )

+ C5

where Cq and Cg again vary with temperature, but slowly compared t o the exponential factor. The assumption necessary to obtain this expression was that ES >> IcT. Over the greater part of the temperature range Cg must be small relative t o the first term, as the conductivity changes by many orders of magnitude with temperature. Hence, t o a first approximation, the apparent activation energy of the conductivity with temperature will be Ea.The apparent activation energy was found to be about 3 e.v. k 1 e.v. The above analyses show the qualitative agreement of the adsorption model with recent conductivity experiments, and indicate how the trapping energy in adsorption may be indirectly obtained by conductivity measurements in the temperature range where slow desorption occurs. The energy can be obtained accurately, however, only with more careful examination of the theory and experiment than is presented here. 3 . A Suggested Mechanism for the Adsorption of Hydrogen on Zinc Oxide

a. The Proposed Model. Most of the adsorption work of hydrogen on zinc oxide has been done below 500"C, in view of the fact that, according to Taylor and Kistiakowsky (48),zinc oxide was reduced in its activity for hydrogen adsorption, if heated above 500°C. I n general, the samples were outgassed a t 450°C in a vacuum. Thus, according t o the discussions in the preceding sections, the zinc oxide used for adsorption experiments with hydrogen contained chemisorbed oxygen on its surface. It would appear reasonable for a singly ionized oxygen atom to have a strong affinity for a hydrogen atom. For the approximate heat evolved from the reaction ,$5Hz O-+ OH-

+

a value can be calculated of approximately 2.6 e.v., by using the known heats of reaction for various similar reactions (49) and algebraically adding these reactions t o form the total process above. This heat of reaction indicates th at the adsorbed oxygen ions do provide attractive adsorption sites for the hydrogen. It will be assumed here that most of the hydrogen adsorbed on zinc

SURFACE BARRIER EFFECTS IN ADSORPTION

291

oxide is adsorbed by the formation of OH- ions, the hydrogen reacting with previously adsorbed oxygen. It will be seen th a t this postulate provides a simpleinterpretation (31) of many anomalous results (44,50,51). Assuming the validity of the energy-level diagram of Fig. 1, we see that together with the 0- ions adsorbed on the surface, there will be in general doubly ionized adsorbed oxygen. Two types of adsorption will occur. The first, which we will term Type A adsorption, is th a t occurring when hydrogen combines with 0- ions which are present on the surface when the hydrogen is injected. The second possibility, Type B adsorption, will be adsorption made possible by the reaction

0--

+ XHz -+

OH-

+ e-

where the electron returns to the zinc oxide. The latter reaction will occur (at temperatures sufficiently high t o overcome the activation energy E l in Fig. 1) because adsorption of hydrogen (Type A) will upset the electronic equilibrium between the number of electrons leaving the 0- sites and those returning from the zinc oxide. There will be fewer empty 0-- sites after the Type A adsorption, hence more electrons will leave the 0-- sites than return. As more 0- sites are formed, more hydrogen will be adsorbed. The activation energies for the two types of reactions may be qualitatively considered. Type A adsorption will have a certain activation energy, EA,not defined by our model, which will limit the rate of Type A adsorption. Type B adsorption will also have this activation energy E,, which may limit its rate. However, as seen from Fig. 8, it will in addition have a n activation energy of the electron transfer type, EB E E l (in Fig. 1) E 0.8 e.v. = 18 Kcal./mole. Thus if EA > EB, the rates of the two types of adsorption will be equal; if EB > EA, Type B adsorption will be slower, controlled by an activation energy of about 0.8 e.v. It will be shown, in the comparison with experiments t o follow, th a t there are two types of adsorption of hydrogen on zinc oxide, one much slower than the other; hence by this model EB > EA. Parenthetically, it must be pointed out th a t the expression for the rate of adsorption involves not only E l but also Ez, the barrier height, so that the apparent activation energy EB will be some complex function of El and Ez. The heats of adsorption of the two types of adsorption may be considered. The heat of adsorption of Type A and its variation with the surface covered is, of course, not given by this model. The model tells us, however, that the heat of adsorption of Type B will have the same variation as that of Type A, with a variation of the electron-transfer type (Section 111) superimposed. Thus there will be less heat of adsorption in

292

S. ROY MORRISON

+

Type B than in Type A by amount E l - (E2 p ) , where p is the Fermi level, and where E z varies with surface covered, yielding less heat of adsorption with increasing surface coverage. b. Comparison with Experiment. The decrease in the amount of hydrogen adsorption if the zinc oxide is heated above 5OO0C, as observed by Taylor and Kistiakowski (48), corresponds to the fact that heating above this temperature (to 600°C) in a vacuum removes adsorbed oxygen. I t was this fact that led to the above suggestion that hydrogen adsorbs by combining with adsorbed oxygen. The possibility that progressive sintering occurs at 6OO0C, with corresponding surface area changes, is repudiated by the observation of Bevan and Anderson (32) that at this temperature the conductance of the material does not vary with time.

'4I 16

O

W Degrees Absolute.

FIG.9. Adsorption isobar of hydrogen on zinc oxide (after Taylor and Strother).

Taylor and Strother (44) were the first investigators t o observe that there were two types of chemisorption of hydrogen on zinc oxide. Their results are shown in Fig. 9. Physical adsorption occurs a t low temperatures; there is a peak for one type of chemisorption (it will be assumed t o be Type A adsorption) at about 350°K, and for Type B adsorption (again assumed) at about 500°K. It is seen that Type B adsorption becomes observable as chemisorption at about 400"K, or 120°C. This agrees fairly well with the results (discussed above) or Morrison (31), who in studying the conductivity, observed that a new release of electrons, probably from 0-- levels (as observed by the initial rise in conductance of the type in Fig. 3), first becomes observable in conductivity measurements at about 80°C. Taylor and Strother found that the activation energy of Type A adsorption from 0 to 56°C varied, with increasing surface coverage, from 3 to 6 Kcal./mole. The activation energy of adsorption from 184" to 218"C, with the same surface coverage, varied from 10 to 11 Kcal./mole. The latter may be qualitatively compared to the activation energy for Type B adsorption of 0.8 e.v. or 18 Kcal./mole. Two sources of error may be mentioned here: (1) the observed Type B adsorption may have

SURFACE BARRIER EFFECTS I N ADSORPTION

293

included also some Type A adsorption, and (2) a different model was used by Taylor and Strother in the theoretical analysis. Taylor and Thon (36), using the results of Taylor and Sickman (29), have recently applied the Elovich Equation to the adsorption of hydrogen by zinc oxide. The results indicate th at there are two types of adsorption at 184"C, and one a t 110°C. Smith and Taylor (50) studied the hydrogen-deuterium exchange on zinc oxide in the temperature range where Type B is predominant. They found t ha t the rate of the exchange corresponds t o Type A adsorption. This is readily explained by our model, as the exchange rate will depend on adsorption sites (0- ions) already prepared, hence on Type A reactions., Taylor and Liang (51) have studied adsorption of hydrogen on zinc oxide using a technique wherein the hydrogen was not removed between each reading. The results (Fig. 10) show effects which are consistent with the proposed model. Figure 10 shows observations when the temperature

Adsorbed, (G.G.)

4

FIG.10. Adsorption of hydrogen on zinc oxide (after Taylor and Liang). The temperature was raised as follows: a t A from 0°C t o 111°C; a t B from l l l ° C t o 154°C; a t C from 154°C to 184°C; and a t D from 184°C to 218°C.

of the zinc oxide is raised with the adsorbent in a hydrogen atmosphere, and the variation in adsorption with time is followed. As is indicated in Fig. 9, Type A adsorption reaches equilibrium a t about 350°K. Below this temperature the adsorption is irreversible. Above this temperature, as shown in Fig. 10, sudden heating will cause immediate desorption of hydrogen from the 0- sites (points B , C , D).However, more sites will slowly be prepared by electron transfer, and Type B adsorption will occur a t a slow rate. This model, of course, implies th at there should be a correlation between hydrogen adsorption and the conductivity of zinc oxide. Type A adsorption should cause no conductivity change in zinc oxide; Type B adsorption should cause a n increase in the conductivity. T o the knowledge of the author, no systematic study of this phase has been made. Stockman (39) has observed that subjecting an evaporated layer of zinc oxide t o a hydrogen atmosphere a t 250°C caused the conductivity of the material t o increase considerably, and caused the resistance-temperature-time curves he obtained t o change their characteristics (see previous section). This latter, however, is a rather complicated effect and requires a much more detailed analysis than will be attempted in this paper.

294

S. ROY MORRISON

It should be emphasized that the model suggested above is based upon the assumption that there may be two types of chemisorbed oxygen on the surface, singly and doubly ionized. The evidence which has been presented to support this assumption is as follows. First, the experimentally observed variation of conductivity a t high temperature with oxygen pressure is consistent with this model, as was shown using thermodynamic analysis by Bevan and Anderson (32). The required law relating conductivity to oxygen pressure is not obtained assuming the adsorbed oxygen is singly ionized. Secondly, the analysis of the reversible variation of the conductance with time and temperature at intermediate temperatures (the order of 150°C) suggests that the oxygen may be doubly ionized. This analysis is described more fully in Section IV,2. In conclusion, we would like to point out the possibility that an effect similar to that described above, a correlation between the adsorption of gases and surface imperfections present, may occur on other materials. For example, Schuit and de Boer (52) have found that slow adsorption of hydrogen on nickel is found only if there is oxygen on the surface. V. THE RELATIONSHIP OF

THE

PHOTOCONDUCTIVITY TO THE STATE

OF THE

SURFACE

When a semiconductor is illuminated, electrons may be excited into the conduction band and/or holes into the valence band, producing photoconductivity. This excited condition is not generally permanent, and when the illumination ceases, the excess current carriers will decay, or “recombine.” The average time which a photoelectron remains in the conduction band is termed the “lifetime.” As the lifetime increases, the photocurrent, for a given intensity of illumination, increases. In many semiconductors the majority of the photoelectrons are produced by excitation from the valence band, the process thus simultaneously producing holes. Surface traps may act as recombination centers for electron-hole recombination and a change in the number or energy of these surface traps, or a change in the height of the surface barrier, may change the rate of recombination. For example Bube (9,lO) has concluded that it is through this effect that the adorption of water vapor influences the photoconductivity of cadmium sulfide. Photoconductivity in zinc oxide, on the other hand, appears to be influenced by the surface through a different effect. Absorption of light effectively excites the electrons trapped in surface levels into the conduction band. This chapter will be primarily devoted to a consideration of this concept, proposed by Melnick (11), that photoconducting electrons are produced through the ionization of surface levels, specifically the adsorbed oxygen levels on zinc oxide. The decay of the photoconductivity,

SURFACE BARRIER EFFECTS IN ADSORPTION

295

on the other hand, occurs through chemisorption of oxygen. It will be demonstrated that a great deal of information concerning the characteristics and theory of adsorption may be accumulated by the study of the photoconductive process when it is influenced in this manner by adsorption. Photoconductive response (the rate of creation, or the “rise,” of the photocurrent, and the rate of “decay” of the photocurrent) appears to be divided into fast and slow responses. The fast responses, with time constants for rise and decay of the order of a second or less, have been adequately interpreted by Mollwo, et al. (53-55), Weiss (56), and Heiland (47,57) as bulk processes. These authors have concluded that the fast response processes are associated with the double ionization of interstitial zinc, and have proposed that the photon excites electrons from the valence band, and that the hole immediately recombines with the electron from an interstitial Zn+, producing double-ionized zinc ions. Evidence indicates that the slow process or processes may be associated with the surface. For example Heiland (47) has found that the irradiation of zinc oxide in a vacuum by light or electrons produces a photocurrent which remains stable, even after the irradiation has ceased, until oxygen is allowed t o reach the sample. Melnick (11) has made the most extensive study of the slow photoconductive response. Using zinc oxide sintered in air, he has studied responses in the time range from 0.3 t o lo7 see. A few of his experimental results will be briefly described. A typical photoconductive rise at room temperature was found to reach half of the maximum conductivity change in a time of the order of an hour after the light was switched on. A similar period was required for half the decay to occur after the light was switched off. The rates of the rise and decay decreased with increasing time, the conductance approaching (‘equilibrium.” It was observed that the increase of photoconductivity under illumination was retarded by high oxygen pressure; that is, the photoconductivity, a t a given time and intensity of illumination, has a lower value if the oxygen pressure is high. On the other hand, decay of the induced conductivity was hastened by increased oxygen pressure. Nitrogen, water vapor, or carbon dioxide had no effect. Studying the effect of temperature changes, Melnick observed that the photoconductivity could be ((frozenin.” For example, he found, while studying the decay of the photoconductivity a t room temperature, that if the temperature was suddenly lowered to 130°K, the slow decay ceased. The photoconductivity was frozen in. If, at some later time, the temperature was raised again to room temperature, the decay recom-

296

S. ROY MORRISON

menced, the conductivity and the rate of decay of the conductivity being the same as that a t the instant the temperature was lowered t o 130°K. These few observations will serve t o illustrate the effects observed, and to indicate the necessity and su$ciency of the adsorption model which Melnick proposed. The experiment of Heiland, and Melnick’s observations of the effect of oxygen pressure, illustrate the importance of adsorbed oxygen as it affects the photoconductivity. Particularly from the work of Heiland it may be concluded that the adsorption of oxygen is important in the decay of the slow photoconductivity. I n order that the effects observed be reversible, in this irreversible region of the adsorption isobar, it is apparent that the oxygen must be desorbed while the sample is illuminated. Melnick suggested that the adsorption of light creates a hole-electron pair, and the hole diffuses to the surface, combining with an electron in the 0- level. That is, the electron in the 0- level returns to the crystal, occupying the empty valence bond, which is the hole. The adsorbed oxygen, now a neutral atom, is desorbed. Thus we have a method of degassing the sample, and studying the rate of adsorption, all a t room temperature. The pressure dependence of the photoconductivity arises because the desorption of oxygen by light (increasing the conductivity) is opposed by the pressure-sensitive adsorption of oxygen (decreasing the conductivity). Hence a high oxygen pressure, increasing the rate of adsorption, will reduce the rate of photoconductivity rise, as observed, or increase the rate of decay after the illumination is removed. The frozen in photoconductivity, as was concluded by Melnick, will arise from effects of the surface barrier layer or, of course, would arise similarly from any other rate-limiting process in the adsorption of oxygen. For our model in this discussion we shall use electron transfer over the surface barrier as the rate-limiting reaction. I n this case, the rate a t which adsorption occurs is proportional to exp (- Es/kT), where Ez is the barrier height. Thus if we measure the decay in photoconductivity (the chemisorption of oxygen) a t room temperature, and then suddenly quench the sample to 130”K, it is obvious that the rate of decay in photoconductivity will decrease considerably. The change in the rate will be dependent on Ez and the temperature t o which the sample is quenched. From an analysis of the rate of decay of the photoconductance as a function of temperature, Melnick concluded that the barrier height was 0.43 e.v., as compared to Morrison’s (31) estimate on a different sample of about 0.5 e.v., the latter being based on measurements of the reversible portion of the time-dependent conductivity. Melnick analyzed his results quantitatively on the basis of the model as described above. As previously discussed in Section IV, he derived,

SURFACE BARRIER EFFECTS IN ADSORPTION

297

using this model, the Elovich Equation for adsorption

!%?d = c1 exp ( -cczAq) dt I n this expression, q is the surface concentration of the adsorbed gas, c1 and c2 are constants, and t is the time. By relating the conductance to the amount of adsorption, he derived an expression for the decay of the photoconductance due to adsorption

dK dt

=

-a exp ( b K )

where K is the conductance and a and b are constants. The expression for the rise of the conductance due to illumination is then given by

dK/dt

=

cZ - a exp ( b K )

where Z is the intensity of illumination at any point in the crystal. As the intensity of illumination Z decreases as a function of distance into the crystal, due t o absorption of light, the conductance K also will be a function of distance into the crystal. Excellent quantitative agreement was found between this theory and the experimental results, both those discussed above and others described further in his paper. Melnick has suggested that even the fast photoconductive rise and decay is a surface phenomenon, involving the same reactions and model as the slow responses. Again he shows agreement between the theoreticaI predictions and the experimental results. It must be emphasized that these results have been interpreted using two assumptions: (1) that the resistance of the zinc oxide is strongly dependent on the amount of chemisorbed oxygen, and (2) th a t electron transfer is rate-limiting in the chemisorption of oxygen. The former assumption has been strongly supported by the photoconductivity experiments, in which it was shown that oxygen, and no other gas studied, has a strong influence on the conductivity a t room temperature. The second assumption has been supported t o the extent th a t it has been shown th a t the Elovich Equation follows from the electron transfer model. Any ratelimiting step in the chemisorption of oxygen, which has a n activation energy varying with the fraction of surface covered in the particular manner predicted by the Elovich Equation, will also be compatible with these results. These investigations may open u p a new and fruitful method for adsorption studies. The use of photoconductivity as a tool for observing

298

S. B O Y MORRISON

adsorption rates appear to be a new technique, which may prove useful in many systems.

VI. THE RELATIONSHIP OF

THE

OF THE

FLUORESCENCE TO THE STATE

SURFACE

The evidence that adsorption has an influence on the luminescence of zinc oxide is not as clear-cut as it is in the case of photoconductivity and conductivity. No experiments have been carried out, to the author’s knowledge, directly correlating adsorption t o luminescence, such as the work by Ewles and Heap (7) on silica, which showed correlations between its fluorescence and the adsorption of the hydroxyl radical. There has, however, been indirect evidence, mainly in the work of Reboul (58). This author found, as have others (59,60), that spectroscopically pure zinc oxide has, in general, two lumineoscent peaks, one at 3950 A. and one a t 5100 The intensity of the 3950 A. emission seemed to depend on the method of preparation of the specimen, as would be expected if the peak were associated with adsorption states. For example, the fluorescenceoof single crystals, with their low surfac$to-volume ratio, shows no 3950 A. peak, although it exhibits the 5100 A. fluorescence. Zinc oxide powder, which has a low concptration of interstitial zinc for adsorption centers, showed a small 3950 A. peak, and also a small 5100 8.peak. If the powder was sintered,:t increase the concentration of excess zinc, the “green” peak a t 5100 A. was intensified. If the sintering was done in air, so thatoadsorption could occur upon cooling, the “ultraviolet” peak a t 3950 A. was also considerably increased in intensity. If the sintering and cooling was done in a vacuum (or in hydrogen), so that little adsorption of oxygen would occur, the ultraviolet peak disappeared. The intensity of the green luminescence on the other hand, was very much higher than for the air-sintered specimens. If a single crystal of ZnO which emits no ultraviolet radiation is ground up t o increase its surface area, i t still does not show any ultraviolet emission. According to Fig. 2, little oxygen adsorption can occur a t room temperature. If it is correct to assume (33) that in sintered samples or powders the density of interstitial zinc is higher near the surface than in the bulk, then it is obvious that in a single crystal, with its constant density, the adsorption will be even lower. Schrader and Leverenz (60) report th at the ultraviolet peak may be removed by reducing the sample in CO or Hz. The interpretation of these observations may be th a t the ultraviolet peak is associated with adsorbed oxygen, and the green luminescenre with the presence of interstitial zinc. The green luminescence, as mentioned

S U R F A C E BARRIER E F F E C T S I N ADSORPTION

299

above, has a n intensity higher in sintered specimens than in a powder, and higher in vacuum- or hydrogen-sintered than in the air-sintered specimens. The concentration of interstitial zinc would be expected to vary in the same direction as a function of the manner of preparation. The sintering process involves decomposition of ZnO, with diffusion of the excess zinc into the lattice in interstitial positions (24). The 3950 A. fluorescence, corresponding t o a n energy of 3.1 e.v., may be associated with a relaxation of the selection rules a t the surface due to the adsorbed oxygen, or it may be associated with a direct transition involving the adsorbed oxygen itself; for example, a transition from the conduction band t o the 0- level, corresponding t o the energy change Ea in Fig. 5 . At present the studies of the luminescence of zinc oxide give little precise information on the adsorption properties of this substance, in contrast t o the more fruitful studies of its photoconductivitg and conductivity. But there exists a possible correlation of the 3950 A. fluorescence with the adsorption of oxygen. Further study of the fluorescent properties may clarify this correlation. I n conclusion, we would like to emphasize the major concepts which we have attempted to bring under the reader’s consideration, and outline the evidence for these concepts. It has been suggested that electron transfer between adsorbate and adsorbent may be a rate-limiting process in adsorption, through the action of the surface barrier “pinching off ” the electron transfer. The surface barrier increases in height as ions are adsorbed, hence the rate of electron transfer decreases. The surface barrier was shown to have many of the characteristics of the activation energy as measured experimentally in activated adsorption. The heat of adsorption decreases, the surface barrier increasing, in the manner regularly observed experimentally, as a function of the fraction of the surface covered by adsorbed ions. It is proposed that on zinc oxide the adsorption of oxygen has an important effect on the electrical properties. This has been supported by direct evidence relating the electrical effects to oxygen pressure over a wide range of temperatures (20°C t o 800”C), and in addition has been supported by abundant indirect evidence. It has been further suggested that electron transfer is rate-limiting in the adsorption of oxygen on zinc oxide. This has been supported mainly by the consistency with which, on the basis of the assumption, many diverse observations may be simply and consistently explained. The characteristics of the adsorption of oxygen, with its attendant resistance changes, are consistent with the assumption over the entire temperature range. T he occurrence, a t temperatures of the order of 150°C, of a reversi-

300

S. ROY MORRISON

ble slow conductance change, arising from a surface energy level other than the first ionization level of oxygen, may be explained considering electron transfer as the rate-limiting process, and a value for the surface barrier height may be estimated. The model proposed for the adsorption of hydrogen on zinc oxide is based on the electron transfer model, and finally the experiments on the photoconductivity were consistently interpretable on the basis of the model. From the rate of chemisorption of oxygen, following the degassing of zinc oxide by light, another estimate of the barrier height is obtained, and found in good agreement with that obtained from the conductivity measurement. Quantitative comparison of the photoconductivity experiments with the characteristics of the electron transfer model yields good agreement. A hypothesis has been presented to describe the adsorption of hydrogen on zinc oxide, wherein the hydrogen is assumed to react with surface energy levels which are associated with adsorbed oxygen. The predictions of the model are shown to be qualitatively consistent with experiment. The author wishes to express his appreciation t o Professor P. H. Miller, Jr., of the University of Pennsylvania, who actively directed the author's research on zinc oxide, and who made many helpful suggestions during the preparation of this article. H e is also indebted t o Professor J. Bardeen, of the University of Illinois, and t o Dr. Paul Weisz, of the Sacony-Vacuum Company, for many valuable discussions and suggestions concerning the material presented above.

REFERENCES 1. Dowden, D. A., and Reynolds, P. W., Discussions Faraday SOC.No. 8, 184 (1950). 2. Garner, W. E., Gray, T. J., and Stone, F. S., Discussions Faraday SOC.No. 8, 246 (1950). 3. Garner, W. E., Stone, F. S., and Tiley, P. F., Proc. Roy. Soc. (London) A211, 472 (1952). 4. Clarke, E. N., Phys. Rev. 91, 756 (1953). 5. Morrison, S. R., J. Phys. Chem. 67, 860 (1953). 6. Suhrmann, R., and Schulz, K., Naturwissenschaften 40 (4), 139 (1953). 7. Ewles, J., and Heap, C. N., Trans. Faraday Soc. 48, 331 (1952). 8. Brattain, W. H., and Bardeen, J., Bell System Tech. J. 32, 1 (1953). 9. Bube, R. H., Phys. Rev. 83, 383 (1951). 10. Bube, R. H., J . Chem. Phys. 21, 1409 (1953). 11. Melnick, D., Thesis, Physics Dept., University of Pennsylvania, 1954; see also Bull. Am. Phys. Soc. 29 (3), 40 (1954). 12. Tamm, I., Physik. 2.Sowjetunion 1, 733 (1932). 13. Shockley, W., Phys. Rev. 66, 317 (1939). 14. Baldock, G. R., Proc. Cambridge Phil. SOC.48, 457 (1952). 15. Bardeen, J., Phys. Rev. 71, 717 (1947). 16. Boudart, M., J. Am. Chena. SOC.74, 1531 (1952).

SURFACE BARRIER E F F E C T S I N ADSORPTION

301

17. Nyrop, J. E., “The Catalytic Action of Surfaces.” Chemical Publishing, New York, 1937. 18. Dowden, D. A., J . Chem. Soc. 1960,242. 19. Volkenshtein, F. F., Zhur. Fiz. Khim. 23, 917 (1949). 20. Schwab, G. M., 2. Elektrochem. 66, 297 (1952). 21. Aigrain, P., and Dugas, C., 2. Electrochem. 66, 363 (1952). 22. Weisz, P. B., J . Chem. Phys. 21, 1531 (1953). 23. Weise, P. B., J . Chem. Phys. 20, 1483 (1952). 24. Hahn, E. E., J . Appl. Phys. 22, 855 (1951); see also Univ. Penn. Tech. Rept. No. 17 (1949). 25. Intemann, K., and Stockmann, F., 2. Physik 131, 10 (1951). 26. Harrison, S., Univ. Penn. Tech. Rept. No. 3 (1952); see also Phys. Rev. 93,52 (1954). 27. Hogarth, C. A., 2. physik. Chem. 198, 30 (1951). 28. Scharowsky, E., 2. Physik 136, 318 (1953). 29. Taylor, H. S., and Sickman, D. V., J . Am. Chem. SOC.64, 602 (1932). 30. Meyer, W., and Neldel, H., 2. tech. Phys. 18, 588 (1937). 31. Morrison, S. R., Thesis, Physics Dept., University of Pennsylvania, 1953; see also Uniu. Penn. Tech. Rept. NO.4 (1952). 32. Bevan, D. J. M., and Anderson, J. S., Discussions Faraday Soc. No. 8, 246 (1950). 33. See ref. 31; also Morrison, S. R., and Miller, P. H. Jr., Univ. Penn. Tech. Rept. No. 6 (1952). 34. Taylor, H. S., J. Am. Chem. SOC.63, 578 (1931). 35. Elovich, S. Y., and Zhabrova, J., J . Phys. Chem. (U.S.S.R.) 13, 1761 (1939). 36. Taylor, H. S., and Thon, N., J . Am. Chem. Soc. 74, 4169 (1952). 37. Fritsch, O., Ann. Physik [5] 22, 375 (1935). 38. Miller, P. H., Jr., in “Semiconducting Materials” (K. Henisch, ed.), p. 172. Academic Press, New York, 1951. 39. Stockmann, F., 2. Physik 127, 563 (1950). 40. Bevan, D. J. M., Shelton, J. P., and Anderson, J. S., J . Chem. Soc. 1948, 1729. 41. Baumbach, H. H., and Wagner, C., 2. physik. Chem. B22, 199 (1933). 42. Hauffe, K., and Block, J., Z. physik. Chem. 196, 438 (1951). 43. Miller, P. H., Phys. Rev. 60, 890 (1941). 44. Taylor, H. S., and Strother, C. O., J . Am. Chem. Soc. 66, 586 (1934). 45. Fritzche, H., 2. Physik 133, 422 (1952). 46. Hauffe, K., and Engell, H. J., 2. Elektrochem. 66, 366 (1952). 47. Heiland, G., 2. Physik 132, 354 (1952). 48. Taylor, H. S., and Kistiakowsky, G. B., J . Am. Chem. Soc. 49, 2468 (1927). 49. Latimer, W. M., “Oxidation Potentials.” Prentice-Hall, New York, 1938. 50. Smith, E. A., and Taylor, H. S., J . Am. Chem. Soc. 60, 362 (1938). 51. Taylor, H. S., and Liang, S. C., J. Am. Chem. Soc. 69, 1306 (1947). 52. Schuit, G. C. A., and de Boer, J. H., Nature 168, 1040 (1951). 53. Mollwo, E., Ann. Physik [6] 3, 230 (1948). 54. Mollwo, E., and Stockmann, F., Ann. Physik [6] 3, 240 (1948). 55. Mollwo, E., 2. physik. Chem. 198, 258 (1951). 56. Weiss, H., 2. Physik 132, 335 (1952). 57. Heiland, G., 2. Physik 132, 367 (1952). 58. Reboul, T. T., Thesis, Physics Dept., University of Pennsylvania, 1953; see also Uniu. Penn. Tech. Rept. No. 7 (1953). 59. Kroeger, F. A., and Dikhoff, J. A . M., J . Electrochem. Soc. 99, 144 (1952). 60. Schrader, R. E., and Leverenz, H. W., J . Opt. Soc. Amer. 37, 939 (1947).

This Page Intentionally Left Blank

Electronic Interaction between Metallic Catalysts and Chemisorbed Molecules R. SUHRMANN Instit ut jiir Ph ysikalische Chemie und Eleklrochemie der Technischen Hoehschule Braunschweig, Germany Page I. Introduction .......................... . . . . . . . . 303 11. The Nature of Electronic Interaction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305 111. Experimental Metho . . . . . . . . 306 1. Methods for Deter . . . . . . . . 307 a. Photoelectric Method . . . . . . . . 307 . . . . . . . . . . 309

c. Electron Optical Method

2. Emission of Secondary Electrons . . . . . . . . . . . 315 3. Electric Resistance of Transparent Catalyst Films. . . . . . . . . . . . . . . . . . . . 316 IV. Results of Experiments.. . . . . . . . . . . . 320 1. Influence of Crystal S Molecules and Electro ..................... 320 a. Work Function and Different Crystal Faces 320 b. Adsorption of Foreign Molecules., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322 c. Migration of Adsorbed Molecules.. . . . . . . . . . . . . . . . . . . . . . . . . 322 2. Electronic Interaction on Polycrystalline Metal Surfaces, . . . . . . . . . . . . . 325 a. Foreign Metal Atoms.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325 b. Rare Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 ............................................... 332 .................................. nd Carbon Monoxide.. . . . . . . . . . . . . . . g. Precovered Metal Surfaces .... . . . . . . . . . . . . . . . . 342 h. Ammonia and Water Vapor.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 i. Hydrocarbons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344 V. Theoretical Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347 VI. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350

I. INTRODUCTION If a n electric field influences polarizable free gas molecules, their electric equilibrium will be disturbed, and their electron clouds will be shifted with respect t o their positive charges. The same phenomenon is to be expected if such molecules enter the electric field of a phase boundary, e.g., if they are adsorbed on an ionic lattice or on a metal surface. In the 303

304

R. SUHRMANN

latter case, too, the molecule will be polarized. Intensity and direction of the polarization depend on the intensity of the potential field, the polarizability of the molecule, and its electron affinity. The electric field a t a pure metal surface is the origin of the work function en * @ of escaping electrons. The potential @ which must be overcome is a t the same time a measure of the electron affinity of the metal surface. If adsorption of polarizable molecules occurs on the metal surface, 9 will be changed by an amount A@. For a sufficient electronic polarizability of the adsorbed molecule, A@ is negative if the electron affinity of the metal surface predominates so th at electrons are shifted toward the metal surface. A@ is positive if the electron affinity of the foreign molecule predominates, in which case metal electrons are shifted in the direction of the adsorbed molecule. As A@ can amount to several volts, the electron deformation of the adsorbed molecules can be expected to influence their chemical behavior substantially. Therefore, when reactions are catalyzed via the intermediate formation of boundary layers on a catalyst, we may assume th a t the activation of the reacting molecules is frequently correlated to their polarization on the catalyst surface. There are two effects of polarization: either it causes a strong but reversible adsorption, or the deformation of the electron shell of the adsorbed molecule is so thorough th a t the system-provided that it possesses sufficient activation energy- switches over irreversibly into a new quantized equilibrium position, forming a chemical bond (1) under liberation of energy. Intermediate states exist between these two extremes. Furthermore, the observed decrease or increase of the work function of metal surfaces on adsorption of foreign molecules is to be expected if only an electron shift occurs to or from the catalyst surface. If the electrons of the adsorbed molecules become part of the metal electron gas, or if the metal electrons become part of the electron shells of the molecule, not only a change of the work function but, in addition, a change of the electric resistance of the metal will be observed. The change in resistance will usually become of measurable magnitude when the adsorbing catalyst layer is not thicker than approximately a hundred or a thousand times the thickness of the layer undergoing electronic interaction. For such systems, measurements of the electric resistance of transparent catalyst layers can give valuable data on the type of electronic interaction. The connection between electronic work function and catalytic activity was shown some time ago, especially by Suhrmann (1-4) and Czesch (l),who found a simple relation between the changes of the work functions of different metal surfaces upon adsorption of H atoms, and the catalytic activities of such surfaces for the recombination of H atoms.

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

305

Schwah and co-workers (5-7) found a parallel between the electron concentration of different phases of certain alloys and the activation energies observed for the decomposition of formic acid into HB and Con, with these alloys as catalysts. Suhrmann and Sachtler (8,9,58) found a relation between the work function of gold and platinum and theenergy of activation necessary for the decomposition of nitrous oxide on these metals. C. Wagner (10) found a relation between the electrical conductivity of semiconducting oxide catalysts and their activity in the decomposition of NzO. Bosworth, Rideal, and Eley (11-13,13a) by measuring the contact potential investigated the change of the electron work function due t o gases adsorbed on metal surfaces. Mignolet (14) studied the change of the surface potential of metal layers caused by gas films adsorbed in two layers. De Boer, Kraak, and Verwey (15,16) studied the influence of oxygen on thin semiconducting layers of molybdenum a t 90°K. and observed an increase of electrical resistance, which they explained by the formation of a surface oxide. Suhrmann and Schulz (9,17,18) observed changes in the resistance of thin nickel layers on the adsorption of gases and vapors, the resistance increasing or decreasing depending on the types of adsorbed gases. Dowden (19) developed a theory of heterogeneous catalysis on the basis of electron exchanges between catalysts and adsorbates [see also Boudart (19a)l. Hauffe and Engell (20), Aigrain and Dugas (21), as well as Weisz (22), tried t o relate chemisorption on semiconductors to the charge distributions in the adsorbing semiconductors.

11. THE NATUREOF ELECTRONIC INTERACTION For the chemisorption of foreign molecules on a metal surface the following three limiting cases of electronic interaction are t o be expected (Fig. 1): 1. The electron shell of the foreign molecule is saturated in itself, for example in the atoms of a rare gas; it has no electron affinity, and its polarizability is small. Therefore it neither receives nor gives of electrons. But in the molecule itself, a small electron shift in the direction t o the metal is possible if the electron affinity of the metal surface is high. 2. The electron affinity of the metal surface is low in comparison with the tendency of the foreign molecule to receive electrons. This tendency is particularly high if the electron shell of the adsorbed molecule is incomplete (e.g. as in a n oxygen atom) or if the bond of the atoms in the molecule is affected by a n asymmetric electron shift (e.g. as in the molecules of nitrous oxide or carbon monoxide). I n such cases the metal electrons become part of t h e electron shell of the foreign molecule. 3. The electron affinity of the metal surface is high in relation to t h a t of the foreign adsorbed molecule whose electron shell is easily displaced (for example with hydrogen atoms or in the presence of r electrons) or asymmetrically distributed (if the

306

It. SUHRMANN

molecules have lone electrons, such as in the HZO or NHI). I n these cases the electrons of the molecule become part of the electrons in the metal surface.

Between these limiting cases, intermediate states are possible: The same molecule can give off electrons to a metal with high electronic work function (e.g., H atoms t o platinum) or receive electrons from a metal with low electronic work function (formation of hydrides between H atoms and alkali-metal surfaces).

FIG.1. Three extreme cases of electronic interaction between foreign molecules and adsorbing metal surface.

111. EXPERIMENTAL METHODS I n order t o study the electronic interaction under the simplest conditions, it is advisable to work a t the lowest possible temperature; i.e., a t room temperature or a t the temperature of liquid air. Since as little as one monolayer, or even less, of foreign molecules can prevent or modify electronic interaction between the adsorbent and the adsorbate, the surface of the adsorbent must be produced in the highest obtainable vacuum by use of vacuum-melted metals. Metals with a low melting point and those with a high one but with a low boiling point are preferably evaporated* in vucuo. The vacuum vessel used must maintain a vacuum of lower than 10-6 mm. Hg for 12 hr. after the connection to the pump has been cut off since, for instance, oxygen of p = mm. Hg can change the electronic properties of a pure metal surface. Greased stopcocks and joints must be avoided. Prior to its use, the vacuum apparatus must be outgassed a t 400" to 500°C. for severalehours. The vessel containing the evaporated metal sample must not be separated from the pumps by melting the glass connections because the melting glass develops gas contaminations, which may cover the metal surface and change its properties. According t o section I, the electronic interaction between an adsorbing metal surface and adsorbed foreign molecules is best studied by observation of the electronic properties of the catalyst surface before and after adsorption. T he work function can be measured directly by photoelectric,

* For the technique of evaporation see, for example, ref. 23; the contamination of evaporated films by the material of the source has been studied by Heavens (24).

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

307

thermionic, or field emission or indirectly by determination of the contact potential of the catalyst surface in reference t o a n unchanged electrode surface. Electron optical methods show how the distribution of the adsorbed molecules depends on the type of crystal face.* The investigation of the emission of secondary electrons from the catalyst surface, especially by use of slow primary electrons, can also give valuable results. Finally, the investigation of changes of the electric resistance of thin (transparent) layers can give useful information on the type of electronic interaction between catalyst and adsorbate. 1. Methods for Determining the Work Function of Catalysts

a. Photoelectric Method. Conditions favorable for obtaining metal specimens of very high purity can be best obtained in experiments designed for measuring the quantum yield for the emission of photoelectrons. This yield I is determined for monochromatic light of different wavelengths a t a temperature T , before and after the adsorption, on the metal surface, of foreign molecules ( l ) ,and the potential CI, is calculated by means of Fowler’s equations (28). According t o Fowler: I / T 2 = Mf(6)

(1)

where M , for a given surface, is a constant independent of the frequency v of the light used for excitation. M is a measure of the number of the centers emitting electrons. Under certain conditions (see section IV,2) the number of these centers, M , changes as the metal surface adsorbs foreign molecules. 6 is defined by 6

= (hv - eoCI,)/(kT)

where eo = elementary charge k = Boltzmann’s constant h = Planck’s constant e26

For 6 S 0;

e36

h v $ e o ( a : f ( 6 )= e 6 - -22+ - -3f2

* .

(3)

for 6 2 0; hv 2 e&: f(6) = * Changes on the crystal surfaces can also be studied by diffraction of slow electrons. In this way Rupp (25) and Germer (85) observed changes of the diffraction maxima of Ni crystal planes due to gas adsorption. Suhrmann and Hayduk (27) found that the diffraction maxima obtained with faces of pyrite and galenite disappear during a long bombardment with slow electrons and reappear some hours later. This method will respond to very small changes in the nature of crystal surfaces. I t s application, however, is difficult, and other uses have not become known.

308

R. SUHRMANN

By taking logarithms we have log ( I / T 2 ) = F ( 6 )

+ log M ; F ( 6 ) = logf(6)

(5)

I n order t o obtain @, F ( 6 ) is calculated for arbitrary positive and negative values of 6, and the results are plotted on transparent coordinate paper as functions of 6 (curve a ) . Then the measured values of log (I/T2) are plotted in the same diagram as a function of hv/(/cT) (curve b ) . The experimental curve b then is shifted vertically and horizontally (see Fig. 31).

FIG.2. Photocell for the investigation of the electronic interaction between chemisorbed foreign molecules and the metallically conducting catalyst [according t o (9)].

Thus both curves can be made to coincide, provided that Fowler’s equations (3) and (4) apply to the experimental results. Finally the shift is expressed by the position of the origin of system a in system b. The horizontal shift is e o @ / ( k T )and gives the potential @; the vertical shift gives the constant M (see Fig. 31 in which the origin of system a is marked by a cross). Figure 2 shows a photoeIectric ceII suitable for the direct determination of the photoelectric work function of a metal under t h e conditions of chemisorption. The electrically conducting catalyst, used as a cathode, either is inserted as a metal foil B or is evaporated from E to B . A metal layer coating the inside of the photoelectric cell serves as the anode, with a lead wire G . B can be heated electrically; the leads b are sealed into quartz and connected to a n instrument (electrometer or amplifier) for

ELECTRONIC INTERACTION O N METALLIC CATALYSTS

309

measuring the photoelectric current. At D there is a n intermediate graduated seal joining quartz and glass. E shows diagrammatically several incandescent filaments, e.g., a tungsten wire delivering thermionic electrons for bombarding the catalyst surface and a platinum filament used to dissociatc thermally molecules into atoms or other fragments. The capsule F contains the substance to he tested for its clectronic interaction with the surface of B . It is separated from the cell by a thin glass membrane which ran be smashed magnetically by a glass covered iron plunger. The monochromatic light passes through a quartz window A and strikes R. I t s cnergy can be measured by a calitxated photocell. G and B can be cooled by liquid air. At J the cell is connected to the vacuum apparatus.

After B has been freed from adsorbed foreign molecules hy heating and by bombardment with electrons, the photoelectric yield as a function of the wavelength is measured and @ calculated according to Equations (31, (4), and ( 5 ) . Then the substance in F is admitted to influence the catalyst surface. The photoelectric sensitivity I of B at a definite wavelengthis controlled. If the electrons of the adsxbed molecules are drawn t o the catalyst surface, I increases, passes through a maximum when almost a monolayer of molecules has been adsorbed, and decreases again when several layers of adsorbed molecules prevent the passage of electrons. At the maximum sensitivity the photoelectric yield is measured, and @a is calculated. The difference gives a measure for the electronic interaction. The photoelectric method cannot be employed at pressures higher than lop4or lop3mm. Hg, on account of ionization by collision. b. Thermionic Method. Determination of the electronic interaction by thermionic emission can be carried out only if the heat of adsorption of the chemisorbed molecules is very high or if, even though adsorbed in small amounts, they lower CP considerably. A sufficient emission must occur a t such a low temperature that no desorption takes place. This is the case, for example, with the adsorption of alkali atoms or alkaline earth atoms on tungsten. The potential between cathode and anode must be lower than the ionization potential of the substance t o be adsorbed if the free path of the electrons in the vapor is smaller than or equal t o the electrode distance. c. Electron Optical Method. The thermionic emission method is especially useful for the electron optical investigation of the distribution of an adsorbate that lowers the work function of the catalyst surface. Figure 3 shows as an example the electron optical reproduction of a crystalline nickel plate in cesium vapor, published by Schenk (31). The plate, which was indirectly heated t o 400" or 630"C., emitted thermionic electrons only from the crystallites covered with cesium atoms. These electrons reproduced the crystal faces on a fluorescent screen in the vessel. If an image of an emitting catalyst surface is t o be produced by the

3 10

I<. S U H R M A N N

photoelectron optical method, the picture must riot be disturbed h y the radiation used for the liberation of electrons. Figure 4 shows a photocell constructed h y Mahl (32) applicable for this purpose. For measuring the work function of single spots on the catalyst surfa(ae, the image of a definite point can be directed tty magnetic fields to a collector iri t h e cvacwited vessel. Then the emission current is measured as t func~tioriof light freqwn(3y or of the temperature of the cathode

FIG.3. Electron-optical reproduction of a crystalline nickel plate emitting thrrmionic clcctrons in cesium vapor [according to Schenk (31)].

without and with adsorbate. Figure 5 shows a corresponding arrangement in which the image of an incandescent cathode K is projerted on a fluorescent, screen S by a magnetic lens 1,. I n S there is a diaphragm H of 1- to 2-mm. diameter. At twelve times linear enlargement the diaphragm opening corresponds to an area of 5 . sq. mm. on the cathode. The electrons passing R fall into the collector A connected to the electrometer. The cathode image can be deflected magnetically by the lateral coil C so that the investigated point of the cathode appears on R. The distribution of chemisorhed molecules on monocrystals has been stJudied by electron optics mainly by E. W. Muller (33,34),,Johnson and Shockley (35), Martin (36), 13enjamin and ,Jenkins (30), arid ,J. A. Uecker (37). Field emission, first used by Muller, has proved especially suitable. Figure 6 shows the arrangement used by this author. The tip of a tungsten filament has been converted by corroding and electrical heating in vacuo into a rounded moriocrystal serving as a cathode. With a radius of the

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

311

curvature of mm. an accelerating voltage of 5 t o 10 kv. is necessary to give an image of the tip on the fluorescent screen with an enlargement of almost 500,000. The field electron microsrope permits determination of the distribution of emission and thereby of the covering of the monocrystal with foreign molecules a t room temperature. By a method analogous to that shown in Fig. 5, the work function can be measured too. I n the anode, covered by a luminous substance, a small opening is made in front of a collector (Fig. 7). By a lateral magnetic field or by adjustment of the cathode, a certain point of the monocrystal is focused on the opening. The field electron microscope has the disadvantage that the eleotrons impinging the FIG.4. Arrangement for the fluorescent screen can decompose the lumielectron optical reproduction nous substance, and the molecules or atoms of a photo cathode [according set free from the latter can be adsorbed by t o Mahl (32)]; K photocaththe catalyst surface, esperially if they escape ode, S fluorescent screen, A as cations from the fl uorescent-screen sur- quartz lamp, B quartz window, face. Therefore, experiments with the field M magnetic coils, C window electron microscope must be done under for the direct observation of the cathode. continuous pumping. This is often possible, because the capsule with the substance to be evaporated on the-monocrystal can be installed near the monocrystal tip. Pump

I

I

S

EB H C FIG. 5. Arrangement for the electron optical reproduction of an incandescent cathode and for mcasuring the emission point by point [according to Schenk (82)l.

d. Method of Contact Potential. The adsorption of a monolayer of molecules, the electron shells of which withdraw electrons from the metal surface, can increase the electronic work function by more than 1 or 2

312

11. SUHRMANN

Pump

FIG.G. Arrangement for the electron optical reproduction of a monocrystal point by field emission [according to E. W. Muller (34)].

700 v

FIG.7. Arrangement for measuring the field emission current of a single plane of a monocrystal [according to Drechsler and E. W. hfiiller (83)].

ELECTRONIC INTERACTION ON MET'ALLIC CATALYSTS

313

volts. This increase results in raising the normal work function of 4 to 5 volts for most catalyst metals t o more than 5.6 volts, giving a photoelectric threshold of less than 220 mp. In this case the determination of the spectral sensitivity meets with considerable experimental difficulties and the work function cannot be evaluated photoelectrically. I n this case, too, it is possible to determine Q, or the increase of Qi caused by the adsorption of foreign molecules, for the contact potential U1,,of two metal surfaces isgiven by the difference of their work functions:

Lrl,z

= - (Q,l

- Q,z)

(6)

If the state of the surface of the reference electrode 2 remains unchanged, the change of the contact potential of the surface 1 by the adsorption of foreign molecules with respect t o the reference electrode is AUl,2 = UI,z*- Ul,z =

- ax);

Q,, =

const.

(7)

where U 1 , 2 * and are the contact potential and the electron work function of the adsorbing surface after adsorption. If AL'1.2 is negative, the work function has increased during adsorption; i.e., in the boundary layer metal-vacuum electrons have been shifted to the outside, the molecule receiving metal electrons. If A U , , , is positive, an electron shift took place toward the metal. But a change of the contact potential of a metal surface covered with foreign molecules may also occur if the surface adsorbs ions or if it is struck by electrons. There are two methods for measuring ACT,,,. According t o the first, an electrically heated tungsten filament is used as a reference electrode, and the parallel shift of the current-voltage curve is investigated for the electrons emitted by the incandescent wire and passing over to the adsorbing surface. This method, first published by Rothe (38) and Germer (26), was used by Langmuir and Kingdon (39) for measuring the contact potential of tungsten wires covered with cesium. Bosworth, Rideal, and Eley (11-13a) developed it for the investigation of adsorbed gases arid it was described by Eley in Advances in Catalysis (40). The parallel-shift method of the current-voltage curve has the disadvantage that the surface covered with foreign molecules is struck by electrons, which may cause an additional charge of the adsorption layer. The gas adsorbed by the tungsten filament can also be changed chemically (e.g., by dissociation) during glowing, and the changed molecules or molecule fragments can strike the electrode under investigation and produce properties different from those caused by the adsorbed gas molecules. In the presence of gases, this method cannot be used because of the glowing tungsten cathode. The second method, described by W. Thomson (41), has none of these

disadvantages. It is based on the change of capacity connected with a change of distance of two electrodes. If the work functions of two opposite electrode surfaces differ from each other, the condenser is charged according t o the contact potential. Changing the plate distance causes part of the charge to flow from the condenser t o the connected electrometer, which is deflected. The deflection cannot be observed if the contact potential has been compensated by a potential a t one of the plates. The comneiisating potential is ryual to the contact potential U 1 . 2 . r - - -- - - - Dewar seal

Iron

Vibrating electrode

FiPump -@

Electromagnet Non electrode vibrating

20 mm.

FIG.8. Arrangernrnt for measuring the contact potential by the vibration method [according to Migriolet (14)].

Thom~on’smethod has been used for the investigation of adsorption layers, especially in the arrangement described by Zisman (42) [cf. also Potter (43), Frost and Hurka (44), and Rosenfeld and Hoskins (45)l. In Zisman’s method the mobile electrode vibrates mechanically, causing periodic variations of the above-mentioned electrometer charge. If the electrometer is replaced by a n amplifier, a signal is heard in a telephone a t the output of the amplifier, which vanishes if the difference of the contact potential is compensated. Mignolet (14,45a) has improved this method in the following way. A glass pipe serves as vibrating fork (Fig. S), and a n electromagnet a t one

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

315

branch of the fork excites the vibrations. An extension of the main branch, which holds the vibrating electrode, is cemented into a 40-kg. iron block by means of several side tubes, the nonvibrating electrode being fitted to this extension. The metal to be investigated is evaporated on the vibrating electrode; b y a special arrangement the nonvibrating electrode is protected against metal vapor during evaporation. This ingenious method has the disadvantage that the surface of the nonvibrating electrode, serving as a reference electrode, is also in contact with the gas or vapor to be adsorbed. Even if the reference electrode can be outgassed after admission of the gas) it is possihle that adsorbed molecules or molecule fragments are given off, strike the surface under investigation, and change its electron work function. The main requisite of this method-that the work function of the reference electrode remain constant and that only the surface of the investigated electrode be influenced by the adsorbed gas-cannot always be completely fulfilled. Doubtless it is possible t o examine the constancy of the work function of the reference electrode by using a measuring bulb of fused quartz and by checking this work function photoelectrically. 2. Emission of Secondary Electrons

Metal surfaces bombarded by primary electrons with energies higher than 10 volts emit secondary electrons with an average energy of about 2 t o 5 volts, coming from a depth of up to about 100 8. Since the primary electrons, with growing energy E,, penetrate deeper into the surface, the yield of secondary electrons a t first increases with increasing E,. The deeper the primary electrons penetrate, the more difficult it becomes for the secondary electrons -with small energies only-to reach the surface. Therefore the yield passes through a maximum between 100 and 1000 volts with pure metal surfaces and then decreases sIowIy. The position and the height of the maximum, therefore, are connected with the depth a t which the secondary electrons have been liberated. If the yield of secondary electrons of a catalyst surface is markedly changed by the adsorption of foreign molecules, the interaction between foreign molecules and the lattice or between secondary electrons and adsorbed molecules does not remain limited to the outer layer of the catalyst but extends into several lattice planes [cf. ref. 461 of its interior. Figure 9 shows an arrangement for measuring secondary electron emission. An electron gun with an incandescent tungsten cathode and the electrodes B 1 and BZ is situated in the lower part of the cell. The secondary emitting target P carries the catalyst. A layer of the latter can also he evaporated from E l or E z , and deposited on P , which can be heated by radiation or electron bombardment from D. The leads FF of a

310

rt. SUHRMANN

thermocouple serve for measuring the secondary electron current going from 'I to A A , the leads of the anode. The substance t o he adsorbed is placed in the capsule G . The emission of secondary electrons can he measured a t low and high temperatures, hut, only in a high vacuum.

F

A

'$ F'Ic:. 9. Arrarigprnrnt for the investigation of the interaction between catalyst and foreign moirculrs by Ineasnreinent of the scrondary rlertron yield of t h r vatalyst sinface [according to (84) ].

3. Electric Resistance of Transparent Catalysl F i l m s

The photoelectric method and the method of the contact potential show only whether, and i n which direction, ari electron shift has occurred a t the surface, owing to an interaction of the catalyst and the adsorbed molecules. By measuring the electric resistanve of transparent catalyst films, however, one can firid out whether ciomplet,e transfer of electrons between the electron gas of the catalyst and foreign molecules has taken place (9). Figure 10 shows a c d l appropriate for such measurements. The adsorbing metal is evaporated on the inside of the bulb from a filament or

ELEC TR ONIC I N T E R A C T I O N ON ME T A LLI C CATALYSTS

317

spiral B in the center. The screens SS prevent metal condensation in the upper tube. From K K the resistance of the thin metal film can be measured. The gas to be adsorbed is admitted from A . The bulb C is connected t o two diffusion pumps 1 and 1’ (Fig. 11) for measuring the amount of gas adsorbed. The part of the apparatus separated by the mercury seals 3 and 3’ is free of stopcorks or greased glass joints.

Pump t

FIG. 10. Vessel for studying the inflrience of adsorbed foreign niolecriles on tlic electric resistance of evaporated transparent layers [according to (IS)].

During an adsorption measurement a certain quantity of gas passes through 7’ into the part of the apparatus separated by 3, pump I’ being cold. The decrease of pressure in 2’ is determined. Part of the entering gas is adsorbed by the thin metal film in C a t a pressure measured with the manometer 4 and a t a constant temperature. The quantity of gas adsorbed a t a given pressure is calculated from the difference between the actual pressure measured with 4 and the pressure computed under the assumption that no adsorption had occurred. After the gas in the separated part of the apparatus had been pumped back into bulb 2’ by means

318

It. SUHRMANN

of pump l’, the quantity of irreversibly adsorbed gas is computed froni the pressure difference in 2‘ before and after the gas contacted the metal film. Like Thornson’s method for measuring contact potentials, the method of resistance measurement permits the investigation of the electronic interaction under pressure and in vucuo. And this latter method has the additional advantage that the influence of the gas on the catalyst is observed directly, while in the (.oritact-potential determinations a referenre elertrodc is int,erposed.

1

I<’I(;. 1 1 . ArrangPrnrnt for mc~rtsuringthe c h n g c of rlertric rcsiatsncc and the amount of gas stdsorhcd [acrording to (IS)].

I n order t o avoid the formation of crystal aggregates during deposition of the metal film, the film is condensed a t low temperatures (80” to 90°K.) arid then warmed slowly to room temperature. A cadmium film e.g. (about 50 mp thick) condensed a t room temperature in a high vacuum appears

dull gray. If the film, however, is condensed at 90°K. arid then warmed t o room temperature, a homogeneous mirror of high reflectivity is obtained. Before warming, the mirror is dark and its elevtric resistance a t 90°K. is about ten times as high as after warming, and cooling back to 90°K. The disordered metal film condensed a t 90°K. passes into an ordered state by warming. Its resistance decreases, and its reflectivity increases (47). If the gas is t o act on the film surface a t room temperature, the film must be annealed at a somewhat higher temperature, so that its own structure will remain frozen during the experiment. When the electron interaction takes place only on the outer surface of the metal film, the film thickness must not exceed 30 to 80 atom layers t o allow of a measurable resistance change. Such films, too, can be repro-

319

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

duced if the vacuum conditions described a t the beginning of section I11 are fulfilled. Table I (18) shows, as an example, the resistance and the thickness of transparent nickel films, the condensation of which was interrupted a t nearly the same resistance (95sl). The decrease of the resistance a t 90"or 293°K. by warming to room temperature and annealing a t 110°C. is evident. The quantity of nickel contained in each layer has been deterTABLE I Resistance and Thickness of Transparent Nickel Films Resistance in After Condensation -

Before annealing - ____

After annealing a t 110°C.

90°K.

293°K.

90°K.

293°K.

1/R298 .dR/d'l'

95.6 95.3 95.0 97.4 97.5 94.8

72.2 78.0 76.3 67.9 69.1 61.1

26.15 26.33 44.6 29.85 30.50 26.20

36.80 38.10 58.3 43.1 43.2 36.10

0.0014 0.0015 0.0012 0.0015 0.0014 0.0014

Film thickness Milliin 1016 Ni gram Ni/ atoms/ total film sq. cm. 1.28 1.22 1.18 1.28 1.22 1.28

92 88 85 92 88 92 -

mined chemirally. After annealing, the values deviate more than immediately after condensation. This can be explained by small inequalities of the metal film covering the inside of the bulb and is probably due t o the facts that the evaporating nickel spiral had not the same distance from all parts of the bulb wall and that the bulb was cut off after each experiment and installed again after the metal layer was dissolved. Nevertheless, layers of reproducible properties can be formed in the described manner, as may be seen from the constancy of the temperature coefficient of the resistance (Table I). The number of superimposed atom layers can be estimated on the following basis. According to Beeck and others (29), metal films condensed a t low temperature in a high vacuum and warmed t o room temperature are composed of crystallites with different orientations. The mirrors obtained in this way show the same optical behavior as crystalline compact material (47), especially after annealing in a high vacuum a t an elevated temperature. The temperature coefficient of resistance, too, even that of the transparent nickel films in Table I, has the same order of magnitude as that of the compact metal; therefore, it seems correct to use the atomic volume of the crystalline metals for estimating the

320

R. S U H R M A N N

film thickness. For Table I, this volume is 1.094 X cb. cm.; 90 x 1 0 1 6 atoms/sq. cm. give with this number a film thickness of 98 A. or 43 atom layers. I n order t o obtain a measure for the density e of an adsorbed film on the metal film (in numbers of foreign molecules per atom of metal surface) the number of the surface atoms must he estimated, assuming that certain lattice planes are predominant. The average surface area per metal atom ran be calculated from the middle density of these lattice planes, and thc number of the surface atoms follows from the area of the macrosurface. This number is a lower limit, and the number 6 calculated from this and from the number of adsorbed molecules is an upper limiting value.

IV. RESULTSOF EXPERIMENTS 1. InJluence of Crystal Structure on W o r k Function, Coverage by Foreign Molecules and Electronic Interaction

a. W o r k Function and Specific Surface Energy of Diflerent Crystal Faces. Since the interaction between chemisorbed molecules and catalyst surface depends on the magnitude of the work function, it is very important t o emphasize that different planes of a monocrystal have different electron work functions [Table I1 (48)]. With the same metallic catalyst, the energy of activation therefore depends on the nature of the most abundant crystal faces. TABLE 11 Numbers of Neaghbor Atoms, SpecajiL Surface Energaes, and Work Fiinctions of Dz.erent Planes of a Tungsten (Iiyslal (48) Number of neighbor atoms

Work Specific surfarr cnergy, functions nj erg/sq. em. 9,volts -

n,

n2

in

G

ad in ad i n ad i n ad in ad

2 5 3 4 4 4 4

4 2 3 3 5 1

i 5

3 3 3 3

9

1'1 an 6,

01' '12

'11 '16

4 4

7 5 8 4 3 7 5

55 10

5.53*

6340

4.691

6430

4.5Gt

6690

4.39t

6690

4.391

*According to Sinirnow and Shuppe (49), ~(Oll)/~(lll)= 5.3/4.2 = 1.26; with 9(111) = 4.39 volts results in (P(Ol1) = 5.53 volts. t According to Nichols and Herring (50).

ELECTR ONIC I N T E R A C T I O N ON ME T A LLI C CATALYSTS

32 1

The second column of Table I1 shows the numbers of the first (nl), and third (ns)neighboring atoms for an atom situated inside second (4, (in) or located outside (ad) the plane under consideration. Planes having the same numbers of nl and n 2 for in-atoms show identical values of a. @ increases with n l and 122 from the octahedral to the rhombic dodecahedral faces. As the work function increases with packing density of the crystal planes, their specific surface energy u decreases [Stranski and Suhrmann (48)] in a reverse linear relation to CP (see Table 11, column 3. and Fig. 12). erglcrn

2

U

6600 -

>

6400 -

?? 0)

6200m UI

a m

1

5.6

5.4

5.2

5.0

4.8

4.6 4-9

4.4 eV 4.2-

Work function

FIG.12. Specific surface energy u as function of the electron work functions

of

various planes of a tungsten crystal.

From the correlation between u and 9 and experimental measurements of CP the following remarkable conclusion can he drawn (48). If atoms of the same kind are condensed on a crystal plane a t such low temperature that they stay a long time a t the points where they struck the surface, the crystal planes with high specific surface energy do not change their fine structure and therefore retain their u value. Planes with low specific surface energy, however, become especially rough under such a bombardment with atoms of the solid phase. Their u value increases considerably and can even exceed the u value of planes with the highest specific surface energies. Since a high u value corresponds to a low value of CP) the work function of a plane with small surface energy is considerably lowered by the condensation on it, of atoms of the same kind. Accordingly, these

322

H. SUHRMANN

crystal planes, are activated for catalytic surface reactions that require an activation of the reacting molecules by transfer of electrons from the catalyst surface (e.g., for the N& decomposition see section IV,2.f). b. Adsorption of Foreign i2.lolecules. According to section I, the desorption energy p of a chemisorbed foreign molecule should depend on the work function @J of the adsorbing surface, on the polarizability or electron affinity of the ad-molecule, and on the number of neighboring atoms of the adsorbed particle [Stranski and Suhrmann (51,52)]. Hence p should be separated into p p and pad, pp corresponding t o the first two types of influence and pLcIto the last one. If the iiork func%ion of the crystal fare and the polarizability of the adsorbed particsle are relatively high, as with cesium atoms on the tungsten planes (011 ) atid ( 1121 (Table 11),ppwill prevail. If the polarizability of t,he ad-molecules is relatively low -as with thorium atoms on tungsten the influence of pad will predominate. As a matter of fact, cesium atoms are adsorbed especially on 1 I 1 2 ) arid { 01 1} tungsten planes, where @ is high, and riot on the {001}and (111) planes, where @ is low. This has been shown by Johnson and Shockley ( 3 5 ) . Thorium atoms, however, are adsorbcd particularly on { 111] tungsten planes according t o Benjamin and Jenkins (30), because n1and n2for ad-atoms on these planes are high (Table 11), and therefore so is pad. The (011 ) planes of tungsten are not covered hy thorium atoms beeause n1 and nz of ad-atoms and therefore p,,,, arc sninll (Tahlt 11) on this plane. The cooperation of pp and pIlcl is illustrated very impressively in the above-mentioned work of Johnson and Shockley, who investigated the distribution of the t hermionir emission of a tungsten moriocrystal wire from the zone [ O l l ] i n cesium vapor. I n the center of the polar diagram (Fig. 13) the tungsten wire is shown vertical t o the drawing plane. Cesium atoms are not yet adsorbed a t 2000°K.; therefore only the planes with the lowest work functions (Table 11) emit, i.e., the (001) and (111) planes ( A in Fig. 13). If the temperature of the filament is lowered t o 900"K,, however, only those planes that have work functions lowered by the adsorption of cesium atoms (.an emit; these are the ( 112) faces ( B in Fig. 13). A t 850°K. (C in Fig. 13) the { O l l ) planes also emit; hence they are able to adsorb caesium atoms a t this temperature, whereas the (001) and { 1 11 } planes do not adsorb cesium atoms a t all in this temperature region. Consequently, cesium atoms adhere most strongly to the planes of the highest work functions, (011) and (112) (Table 11). But because of the influence of pa,l, they adhere more strongly t o the { 112} than to the (011) planes, as the number of neighbor atoms for ad-atoms a t { 112) planes is higher than a t { 01 1} planes (Table 11). c. Migration of Adsorbed Molecules. In order t o understand the ad~

ELEC TR ONIC INT E RACT ION ON ME T A LLI C CATALYSTS

323

sorption process at the surface of a crystalline catalyst it is essential to know the characteristics of migration on the surface. When foreign molecules are adsorbed on a crystal face, those situated in hollow corners of the catalyst surface (c in Fig. 14) are tied most 110

FIG. 13. Distribution of thermionic electron emission of a tungsten-monocrystal wire in cesium vapor in the [ O l l ] zone at A 2,000”, B 900”, C 850°K. Cesium vapor pressure corresponds to room temperature [according t o Johnson and Shockley (35)l.

FIG. 14. Position of adsorbed foreign molecules on a crystal surface with incompletely formed lattice planes [according to Stranski and Suhrmann (51)l.

strongly. As the number of adsorbed molecules increases, these positions are occupied first. This is followed by molecules adsorbed on the border of a n incomplete lattice face ( h ) , which are tied less strongly. Those situated on a crystal plane (a) are most loosely bound. With increasing covering, molecules are also found in the position c’. Finally (at e ‘v 1) those foreign molecules prevail which are surrounded by similar molecules a t four sides and are kept a t the catalyst by the sublayer only. The differentialheat of adsorption of foreign molecules, therefore, is high a t a small coverage e and decreases with increasing e first very quickly, then

324

R. SUHRMANN

more slowly. When 8 approaches 1 , the heat of adsorption becomes constant. The energy necessary for place changes can be calculated for the movement of foreign atoms on a layer of equal foreign atoms adsorbed on the (011 ] tungsten crystal planes in the following manner [Stranski and Suhrmann (51)]. On the (011 } planes an ad-atom (Table 11) can take the positions rharacterized by the numbers of neighboring atoms 2 121 5 and 3 101 3 . The energies of addition pa are therefore, acwrding to Kossel (52a) and Strariski ( 5 3 ) :

+ + 5,s

2~1

and

~ P Z

3p1

+

~ P S

Since p 2 1 p I = 0.5 arid p J / p I ‘v 0, pa is about 3 . 0 ~in ~both cases. The activation energy for place changes consists of the energy of addition and of the energy necessary to surmount the saddle point. With the latter one gets for the place-change energy on the { 01 11 planes 0 . 5 1 ~ 1‘v

Paa ‘v

PI/^

A t o m situated in the second layer can take the positions shown in Fig. 14. Their energies of addition are = 2P1

Pa

3Pl p~ = 4pi pb =

+ + 3PZ + 3pz

~ PN z N

3Pl 4.5p1 5.5~1

Hence we have the place-change energies Pan C%

P1/2;

Pbh

paa ‘v

Pba

Pb -

P1/2;

Pa

+

Peb

Pna ‘v

2pl;

Pc

-

ph

+

Paa

Pca ‘v P c

3PlVL;

- pa

+

Pan ‘v

3pl

Since the heat of evaporation of cesium atoms from the second atom layer is 17,500 cal., acrording to Taylor and Langmuir (54), arid since evaporation occurs from the half cryst,al position c we find pc =

Thus

p1 =

17,500 ral. N 5 . 5 ~ ~

3180 ml., and pn

‘v

pb ‘v

p, ‘v

9500 caal.

14,300 (’al. 17,500 cal.

1600 (d. 4800 cal. Pba ‘v 6400 cal. pra ‘v 9500 cal.

Paa N Pbb ‘v

P c b ‘v

Because of the high value of pc the cesium atoms will remain in the position c for most of the time. Since the (011} planes on the tungsten wires used by Taylor and Langmuir might have had several steps, diffusion may not have occurred on the planes themselves but along the

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

325

steps with the energy P c b , and not with the place-change energy Pbb. For many spots c (hollow corners) may have been interspersed amorig the steps of the crystal face. The transition from the position c to a in Fig. 14 will happen very seldom because of the high value of P , ~ .Indeed, the measured place-change energy of 4600 cal. of cesium atoms is very near the calculated value for pcb = 4800 cal. The activation energy of the migration of tungsten atoms on a tungsten crystal was calculated by E. W. Muller (34) in the same way and was compared with the values obtained by the field electron microscope. I n this case, too, the agreement between measured and computed values is very good. Recently Drechsler (53a) has calculated the adsorption and migration energies of arbitrary systems of atoms 011 faces of cubic body-centered crystals. Muller observed that the migration velocity on the surface a t lower temperatures was greatly accelerated by adsorbed water molecules. The energy required for a place change, and for the migration of tungsten atoms on a crystal surface of tungsten can be decreased by adsorbed water molecules t o one third of its original value. At the same time the constant indicating the number of migrating atoms decreases, according to Muller. As a result, the lowering of the place-change energy is overcompensated a t a higher temperature, and the migration on the surface occurs more slowly. 2. Electronic Interaction on Polycrystalline Metal Surfaces

Since the electron work functions and the specific surface energies of different planes of the same crystal may have different values, it would be interesting to study the electronic interaction during adsorption of foreign molecules on monocrystals. Investigations of monocrystals, however, encounter many difficulties; therefore, one has to restrict oneself in general t o pobycrystalline surfaces, which also give remarkable results because the force of interaction essentially depends on the nature of the metal and differs for the same metal from one species of adsorbed molecules to the other. a. Foreign Metal Atoms. The conditions are comparatively simple for adsorption of foreign metal atoms on metallic surfaces. In this case the interaction energy should be determined on the one hand by the work function of the metal surface, on the other hand by the polarizability, the ionization potential, or one of the excitation potentials of the foreign metal atom. Systematic investigations are available for adsorption of only such atoms as lower the electron work function of the adsorbing metal surface, e.g., for alkali atoms whose valence electrons are drawn into the metal surface on account of their high polarizability.

326

R. SUHRMANN

With increasing covering by foreign atoms, CP decreases a t first linearly and then more slowly t o a minimum value a0,somewhat below that for monatomic covering, 0 = 1. Then the work function increases again t o the value found for the pure foreign metal itself, which is reached when about 5 atom layers (55) of the adsorbed metal have been deposited. The increase of @ shortly before the coverage reaches the value 0 = 1 is usually explained by the mutual interaction of adsorbed dipoles, but it can also be related t,o the difference of the electronic interaction on different crystal planes of the sublayer surface (see Fig. 3). The value of % depends on the work function of the adsorbent and on the polarizability of the foreign metal atoms. Table I11 shows that, after TABLE 111 Work Functions 4 of Pure Tungsten and Platinurn; Work Functions +o at Optimuin Covering N O ,in Alkali Atoms per Square Centimeferof Metal Surface According t o H. Mayer (54a) -

Metal 4,volts

w

4.53

Pt

5.36

No

Alkali metal

K

cs I<

cs

3.4

x

@o,

1014

3 . 1 5 x 1014 3 . 5 x 1014 2 . 5 x 1014

volts

I .76 1.70 I .68 1.60

A@, volts

2.77 2.83 3.68 3.76

the adsorption of potassium and cesium atonis on platinum, the @O values are smaller by 0.1 volt than they are after adsorption of the same atoms on tungsten. The diminution of @, the work function of the pure metal, caused by the alkali deposit is greater by 1 volt for platinum, the metal with the highest work function, than for tungsten. The diff erewe between the @,, values found after adsorption of cesium and potassium atoms is about 0.1 volt for both platinum and tungsten. For thc value of the decrease of the work function of tungsten caused by its being covered with sodium atoms, Hosworth arid Rideal (12) obtained A+ = 2.78 volts by the contact-potential method. This value agrees well with the values of Table 111. The electronic interaction is small if the metal used as an adsorbent has a work function low in relation t o the polariaability of the adsorbed atoms. On adsorption of sodium atoms on an aluminum surface of CP = 4.08 volts, for instance, Brady and Jacobsmeyer (56) obtained a noticeable increase of the photoelectric emission only after five atom layers of sodium had been condensed. I n this rase the alkali layer itself and riot the metal of the sublayer emitted the electrons.

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

327

A systematic investigation of this type of electronic interaction would probably give valuable results, particularly if it could be made with respect t o the filling of the electron bands of the adsorbent. It may be expected that adsorbed alkali atoms lower @, especially in the rase of transition metals with incomplete electron levels, but for metals with filled d levels, a smaller effect can be expected. An exception may result from preadsorption with 0 atoms, which create new vacancies in the bands owing t o electron transfer (see section IV,2c). b. Rare Gases. Rare gases with closed electron shells should not give electronic interaction on adsorption. Mignolet ( 14), however, observed a shift, AU,,,, of the contact potential to the positive side (decrease of the work function, i.e. shift of electrons to the metal) for adsorption of argon (1 mm. Hg) and xenon (2 x mm. Hg) on evaporated, nontransparent nickel layers : AU1,2(A) = +0.03 volt;

AUl,,(Xe)

=

+0.85 volt

On the other hand, no resistance change is obtained if one brings argon of lo-, mm. Hg in contact with evaporated transparent nickel layers, a t 90°K. or a t room temperature, provided that the gas was cleaned before by a glow discharge in the presence of a potassium film (18). The ability of the nickel film t o adsorb argon is very small. The covering a t a pressure of mm. Hg is e = 0.13 and after the free gas is pumped off e corresponds to 0.06 argon atom per surface nickel atom. The electron shells of the adsorbed argon atoms do not become part of the electron gas of the metal surface, although, according to Mignolet, a slight shift of their electrons to the metal take place. c. Oxygen. On account of the high electron affinity of the 0 atom, decomposition into atoms of 0 2 molecules striking the surface is to be expected if the work function of the metal surface is not too high and if the necessary energy of activation is available. The atoms should draw out electrons from the metal surface and raise the work function. A surface oxide will be formed if the thermal energy is sufficient for penetration [cf. de Boer, Verwey, and Kraak (15,16)]. But even without formation of an oxide, adsorbed oxygen will considerably influence the electron gas of the metal. I n fact, the oxidation of metals with very low work functions (alkali metals, alkaline earth metals) proceeds very quickly a t room temperatures and more slowly a t 83°K. If the work function of the metal has a medium-sized value, it is increased by oxygen, e.g., that of tungsten from 4.53 t o 6.27 volts, or b y 1.74 volts, according t o the measurements of the contact potential by Bosworth and Rideal (12). I n this case oxygen is adsorbed in the form of atoms. No oxidation was observed at low tem-

328

11. HTJHRMANN

peratures [cf. Rideal and Trapnell (57)) With molydenum, however, the work function of which amounts to 4.24 volts, being below that of tungsten, de Boer and Kraak (15) observed the formation of a surface oxide when the surface, (*overedwith oxygen a t 90"K., was warmed to room temperature. This was borne out by the increase of the electric resistance of transparent (semiconducting) molybdenum films by about 80 %, under these conditions. 90.5'K

44.000 -

I

I

I

I

I

I I

I

I

I I

I At,>=H2admitted I B - Ozpurnpedoff I BH2-H2pumped off

I

I

I

I I

I I I

I

I

I

I

0

B

60

120 180 Time, minutes

240

300

FIG.15. Increase of resistance of a transparent nickel film (90 X 1016atoms per sq. rm.) a t oxygen adsorption; 'I' = 90.5"K. according to (18).

If oxygen acts on a metal with a work function somewhat higher than that of molybdenum, e.g., nickel [@(Ni)= 4.91 volts] the change of resistance is smaller although still measurable a t 90°K. Figure 15 shows the increase of resistance of a transparent nickel film a t 90.5"K. (18). At the beginning the admitted oxygen is adsorbed at once arid irreversibly. If the pressure of oxygen exceeds 10P mm. Hg, the resistance increases more slowly but also irreversibly. At pressures higher than mm. Hg more oxygen is adsorbed, but a further increase of resistance is not observed, as may be seen from Table IV. The oxygen additionally adsorbed shows no electronic interaction with the nickel surface and can be pumped off. The strongest interaction with the electron gas of the metal takes place a t the beginning of the adsorption. If the specific increase of the resistarm A R / B is plotted as a function of the covering e (number of 0 2 molecules per surface Ni atom), the values a t first lie near 10% and

329

ELECTRONIC INTER.4CTION O N METALLIC CATALYSTS

TABLE I V Adsorption of Oxygen at Constant Temperature and Digerent Pressures (18) Quantity of adsorbed oxygen (molcculcs Before pumping off Pressure of oxygen (mm. H d

per cm.2 macrosurface

p1 < 10-6 p2 = 3 x 10-5 p 3 = 1 . 3 x 10-4 p4 = 9 . 5 x 10-4

0.62 x 1015 I , 54 x 1016 1 . 7 x 1015 1 . 8 5 x 1015

p1 = 3 x 10-6 p 2 = 1 x 10-5 pa = 1 . 3 x 10-3 p4 = 1.95 X 10-2 p s = 3.0 X 1V2

0.53 x 2.33 x 5.4 x 5.2 x 5.15 X

1015

10'5 10'5

:12.0

After pumping off

per surfacc Ni atom, 0

101b 1015

o

0 2 )

per cm.2 macrosurface

90.5"K. 0.42 0.36 x 1.03 1.40 X 1.14 1.69 X 1.24 1.69 X 294.9"K. 0 . 3 6 0,122 x 1.56 2.16 X 3.64 4 . 9 x 3 . 4 8 5.1 x 3.46 5.1 x

Increase of resistance AR after pumping off (in percentage of the original value)

per surface __ Ni atom, 0 A R

AR/0

1015 l0lb 1015 1016

0.242 0.94 1.13 1.13

2.16 5.34 5.57 5.57

1015

0.082 I .45 3.28 3.42 3.42

0.87 10.62 4.33 2.99 9 . 7 3 2.97 2.90 9.93 9 . 9 3 2.90

1015 1015 1015

8.98 5.68 4.93 4.93

90 5'K 2949'K

i l

U,

0

I

1

I

2 -+Q

,

I

3

a

FIG.16. Specific increase of resistance A R / 0 as function of the covering 0 of the Ni film with oxygen, corresponding t o Fig. 15 [according t o (IS)].

330

R. SUHRMANN

decrease t o 3% a t about 6 = 1.5 (Fig. 16). The sudden increase of resistance at the beginning of adsorption shown in Fig. 15 depends on the covering of the most active crystallite planes, which cause a sudden decomposition of O2 molecules into atoms owing to the high activity (low work function) of these planes. The O2 molecules later adsorbed on less active spots need a higher energy of activation for their decomposition and therefore a longer time. The infliience of a difference in work functions on the rovering of different csrystni planes by oxygen atoms is seen easily from pictures of a,

FIG.17. Ilistribution of field emission of a tungstcn-monocrystal point a t room temperaturr 1)c.forr (a) and after: (h) the action of oxygen a t room temperature [acrording to E. W. Miillrr (33))

tungsten monocrystal (Fig. 17) obtained by E. W. Muller (33) with the field elevtron mkroscope: The octahedron and the cube faces, the work furictioris of which are lowest, emit most strongly before the adsorption of oxygen ( a ) . Afterward they appear dark ( b ) . At room tomperature the energy of activation for the decomposition of 0 2 molecules is more readily available than at 90.5"K. Therefore, an additional increase of A R is observed even at mm. Hg (Table IV). High pressures do not cause any further increase. T h e specific increase of resistance AR/6 is independent of the temperature, as shown in Fig. 16. Considering the high values of 6 obtained at room temperature, it is probable that 0 atoms adsorbed above 8 = 1 also enter the surface and interact electronically, Rcheuhle's (57a) adsorption measurements support t,his view. Adsorption of hydrogen atoms by a metal with a high work function,

ELECTRONIC I N T E R A C T I O N O N METALLIC CATALYSTS

331

e.g. platinum (a = 5.36 volts), lowers the work function, as will be explained in section 2d. After adsorption of hydrogen by a metal surface, its interaction with molecular oxygen is considerably lowered. The work function of a platinum surface may be lowered by adsorbed hydrogen atoms t o 4.0 voltls, which is below the value for pure nickel (4.91 volts). Oxygen a t a pressure of 10-l t o lop2 mm. Hg does not change the work function of this platinum surfacc; only the photoelectrically measured constant M (see section II1,la) decreases a t room temperature by 11% and a t 83°K. by 21 % upon the adsorption of O2molecules ( 5 8 ) . Atomic oxygen, however, strongly interacts with a platinum surface

I

0

I I I

40

,

I

I

80 120 Time, minutes

I 1 1 . 1 -

L

160

-C.

2 0

FIG.18. Influence of molecular ( B )and atomic ( C ) oxygen on a Pt surface covered with H atoms, investigated by the change of photoelectric emission. Ordinate: photoelectric yield I in electrons per light quantum. X = 302.2 mfi; 1’ = 293°K.; ( a ) 0 2 admitted, p o 2 = 0.34 mm. Hg; (b) O2pumped off; ( e ) O2admitted, p o 2 = mm. Hg, Pt spiral kept a t yellow heat for 2 min.; ( d ) O2pumped off [according t o (58)l.

precovered with H atoms, as shown by the results of Fig. 18. I n the beginning ( A ) the work function of the hydrogen-containing surface was found t o be 3.91 volts. Molecular oxygen of 0.3 mm. Hg lowered the photoelecmm. tric sensitivity but slightly ( B ) . After pumping off, oxygen of Hg was added, and a platinum spiral in the test bulb ( E in Fig. 2) was kept glowing for 2 min. t o produce 0 atoms. Then the photocurrent due t o illumination in the quartz ultraviolet a t once became unmeasurable ( C ) . The influence of oxygen on metal surfaces, obtained by scraping solid metal pieces in vacuo (10-5 mm. Hg), was investigated by Fianda and Lange (59). After scraping, for all metals (Ag, Cu, Sn, Zn, Cd, Pb, Al,

332

R. SUHILMANN

Mg), the work function decreases first quickly then more slowly, passes through a minimurn, and then increases again. The gradient of the derreasc and the depth of the minimum are more strongly marked for metals with low work functions and high tendencies t o form an oxide. Unfortunately, these interesting results cannot be fully accepted because of the insufficicnt conditions of evacuation and the presence of several large greased joints in the cquipment used by thcse investigators. d. Hydrogen. If molecular hydrogen a t low temperatures contacts a metal surfac.e of low work function, eg., potassium, neither an interac+tionbetween the Hz molecule and the surface nor a decomposition into H atoms is to be expected. Actually it was proved long ago (60) and later verified by other authors [e.g., (6l)l that molecular hydrogen does not change the spectral photoelectric. sensitivity of pure potassium surfaces. If, on the other hand, H atoms come in contact with an alkali metal surface, they draw out electrons on account of their electron affinity and if there is sufficient energy of activation available form a surface hydride. This is seen from the considerable change of the spectral photoelectric sensitivity of a potassium surface on contact with atomic hydrogen (62). While in this rasp the electron affinity of the H atoms prevails over that of the metal surface, the electron affinity of the latter predominates in cases of high work function. With sufficient thermal energy for activation, the striking H z molecule decomposes into atoms, which give their valence elcctrons t o the metal. The decomposition will occur first on the crystallite planes with the highest work function. This electron shift to the metal is best observed from the behavior of the electric resistance of thin nickel films under the influence of hydrogen. As Fig. 19 shows, small quantities of hydrogen added at 90°K. and a t pressures helow 10-6 mm. Hg cause a clearly measurable decrease, AIi, of the elevtrir resistance. At 10-6 mm. Hg a stronger, sudden decrease oc*curs.At mm. Hg the rate of the resistance change diminishes, and with furthcr pressure increase, the resistance no longer decreases. At e = 0.34 molecule Hz per surface Ni atom, the surface is saturated with hydrogen ; that is, the nickel film is completely covered with hydrogen atoms. The maximum decrease of resistance for the experiment under consideration is 1% of its original value. The hydrogen adsorbed a t the pressures mentioned cannot be pumped off. Different changes of the resistance of a nickel film are observed, if hydrogen is adsorbed at room temperature (Fig. 20). In addition t o the sudden diminution a gradual decrease of the resistance takes place. On pumping, part of the hydrogen adsorbed under pressure is released, and a t the same time the resistance increases again t o a limiting value corresponding to that part of the hydrogen which is irreversibly bound t o the

333

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

26.600

I

I I

c: .-

B7-Ar pumped off

B2

I

I

I 26.500

a,

c

R1

v) .-

In

a,

a

26 400 I

26 300

I

0

I

I

I

,

I

I

I

I

I

60

120

180 240 Time, minutes

300

360

,

420

FIG. 19. Decrease of resistance of a transparent nickel film (90 X 1015 atoms/ sq. cm.) on adsorption of hydrogen; T = 90.6"K. [according to (lS)].

-

51.000 -

A H2 admitted B = H2 pumped off

c: 50.500 -

c

I<

I& lb.

I,

1

' ( 1

a, 0

I < I d4.

I x IT

. A

m. + Lo

%

CT

50 000 -

49.500

'

, B

+, B

180

240

Time, minutes

FIG. 20. Decrease of resistance of a transparent nickel a m (90 X 10l6atoms/ sq. cm.) on adsorption of hydrogen; T = 293.4OK. [according to (18)].

334

R. SUHRMANN

metal surface. The highest irreversible covering is 0 = 1.5 molecules HZ per surface Ni atom. The maximum decrease of the resistance is 3.4%. This behavior a t room temperature can probably be explained b y assuming that H atoms enter into the Ni film. On pumping off, some of the H atoms again diffuse to the surface, combine to Hz molecules, and leave the metal. Rideal and Trapnell (62a,63), too, observed that part of the hydrogen is adsorbed very quickly on nickel or tungsten films but that a smaller part is adsorbed slowly. These authors, too, consider the slow process t o be a solution of hydrogen in the nickel or tungsten lattice. At room temperature Hz molecules striking those crystallites of a platinum surface which have the highest work function (I) decompose into atoms. If a pure platinum surface is contacted by hydrogen, crystallites I will be covered by H atoms, those with a low work function, 11, b y Hz molecules, because the decomposition of Hz molecules on crystallites I1 requires a higher energy of activation than on crystallites I. The work function of crystallites I is lowered by the polarized H atoms; th a t of crystallites I1 will not be changed essentially. All effects combined, the photoelectric emission therefore increases if hydrogen is adsorbed on a pure platinum surface. When such a hydrogen-covered surface is bombarded with electrons of low energy amp., 20 to 300 volts) (Za), both types of crystallites will bestruck. The work function of crystallites I1 islowered, because their Hz molecules decompose into atoms under bombardment; that of crystallites I is increased, because their H atoms are shot off. If the first effect predominates, the photoelectric emission of the platinum surface increases; if the second effect prevails it decreases. The combination of both effects is seen in the results shown in Fig. 21. By adsorption of hydrogen, the sensitivity is increased (D-+ E ) . The electron bombardment causes at the beginning a further increase ( E --+ F ) , because the Hz molecules decompose into atoms; the later bombardment causes the photoemission t o decrease again ( F -+ G ) , because H atoms are shot off. The work functions 9,in volts, in the different states are

D 4.53

E 4.35

F 4.25

G 4.40

On a platinum surface covered by H atoms, additionally adsorbed hydrogen will not find any crystallites on which t o dissociate into atoms. It is adsorbed molecularly and hinders the passage of electrons (see Fig. 22). The original state A is marked by a high sensitivity and a low work function, = 4.05 volts. The influence of molecular hydrogen makes the photoelectric yield decrease a t 302.2 mp, i.e. near the thresh-

335

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

Bombardment

0

20

40

60

80 1000

10

20

30

40

Time, minutes

FIG.21. Influence of molecular hydrogen on a platinum surface, partly liberated E, F , G influence of hydrogen from adsorbed hydrogen by electron bombardment (D); adsorption ( E ) and following electron bombardment (F, G ) . Ordinate: photoelectric yield I in electrons per light quantum. = 265.5 mp. T = 293°K. (a) Hz admitted, = 1.35 mm. Hg; ( b ) Hz pumped off [according t o (58)].

15

~10-7 A,, -1

I I I

I

I I

Z*"

I

I

I

I

10 -

I

I

I

I

I

I

I

lo

I

I

I

1

I 1 I

I I

I I

I I I I

I

I

I I

I

0

Ib

la

I

B

I

I

5-

-

I I

I

I

I

336

R. SUHRMANN

old, by 30% ( B ) .The evaluation of the spectral sensitivity curves measured in A and B shows, however, the same work function in both. B u t the constant M , which is a measure of the number of emitting centers (see section III,lu), has decreased by 30%, because the highly emitting crystallites were partially covered by additionally adsorbed Hz molecules. If a platinum foil electrolytically furnished with hydrogen is flashed in vacuo, it gives off one part of the adsorbed gas. I n this case after cooling, the H atoms will be situated on crystallites with a high work function, the HZmolecules on those with a low one. The measured work function and also the constant M are therefore low. The admission of hydrogen does in this case not essentially change the sensitivity (Fig. 23, A + B ) ,because x

l0I iz

10-6 I

I

I

'

I

I

!

I

-J--

I

I

I I

I

I

I

B

I

I

0

L

<

I

40

I

I I

I

I

I

I

I

I

1 . 1 1

80

I

I

120

160

- . I 200

240

Time, minutes

FIQ.23. Influence of molecular hydrogen and single H atoms (formed by thermal dissociation) on the photoelectric yield of a platinum foil, electrolytically loaded with hydrogen and brought to red heat in BUCUO. Ordinate: photoelectric yield I in electrons per light quantum. X = 280.3 mp; T = 293°K. (a) H Z admitted, pa, = 5.3 X

( b ) Hz pumped off; (c) Hz admitted, p~~ = 0.5 heated for 10 sec. [according to (58)].

mm. Hg;

x

10-3mm. Hg; ( d ) tungsten filament

the surface is already covered so that striking H, molecules do not find pairs of sites where they can dissociate into atoms (see refs. 63al63b,64). Single H atoms, however, reaching the platinum surface at this state still find sites where they can be adsorbed and can lower the work function in the same way as the atoms formed by the dissociation of H2 molecules a t pairs of sites. I n fact, a tungsten spiral glowing in hydrogen gas increases the photoelectric sensitivity (Fig. 23, B + C) considerably, indicating a decrease of the work function from 4.04 to 3.94 volts ( A to C in Fig. 23). The work function of pure tungsten (a = 4.53 volts) is smaller b y

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

337

0.4 and 0.9 volt than that of pure nickel and platinum respectively. Therefore it would be understandable that hydrogen atoms adsorbed on a tungsten surface would be polarized, a t least partly, in the sense that they increase the work function of the metal surface. Indeed, by using the contact-potential method and measuring the current-voltage characteristics Bosworth and Rideal(l2) observed that a tungsten surface ( B ) covered by hydrogen had a work function higher by 1.26 volts than the surface of a n incandescent tungsten filament. I n the apparatus used by these authors, this hot tungsten filament A , was located a t a distance of 1 mm. from the cold tungsten filament B, served as reference electrode. After the adsorption of hydrogen on the cold wire, the filament A was flashed (2400°K.) t o remove all hydrogen from its surface. At this temperature the adsorbed hydrogen was given off in a n atomic, thermally activated state; therefore it may be expected that the covering of the surface of the cold wire B changed during this flashing of filament A in its immediate neighborhood, While the current-voltage curve is being measured, the current flows t o a point of wire B nearest to wire A. Hence it is doubtful whether the potential shift of 1.26 volts found b y Bosworth and Rideal was actually caused b y the original hydrogen covering of the cold wire or by a hydride formed during flashing. Mignolet (65) observed by the same method a change in contact potential of the surface of an evaporated tungsten film, by AU,,, = -0.65 volt, i.e., by a value that is about half of that found by Bosworth and Rideal. I n this case hydrogen was adsorbed on the metal film a t room temperature and a t mm. Hg. If Mignolet then cooled the bulb t o 77”K., permitting it to adsorb additional hydrogen, lAUl,2/ decreased by 0.2 volt. The author explains this result by the formation of a “volatile” positive hydrogen film. Using his method of measuring potentials, explained in section III,ld , the same author made similar observations for the adsorption of hydrogen on evaporated nickel films (14). A theoretical discussion of these interesting observations might be deferred until direct measurements of the work function, e.g., with the photoelectric method, gives analogous results. e. Nitrogen. As with oxygen, a strong electronic interaction between nitrogen and a metal surface is t o be expected, when the Nz molecule dissociates into atoms on the metal surface, or has been decomposed, prior t o adsorption. This, however, will happen less easily with Nz molecules than with O2 molecules because the energy of dissociation of the former is approximately twice as high as that of the latter, The decomposition does not occur if it is sterically hindered, as e.g. with potassium (KaN!). Actually nitrogen is not chemisorbed on potassium, whilst activated chemisorption of nitrogen on calcium (CaaN2!) a t - 78°C. and

338

R . SUHRMANN

at room temperature was observed by Trapnell (23). I n this case the energy of activation increased with growing covering from 1500 t o 3800 cal. Besides the steric factors, electron transfer to the adsorbate seems t o play a part, because metals chemisorbing nitrogen (23) have relatively low work functions (in volts) : W 4.53

Ta 4.13

Mo 4.24

Ti 4.16

Zr 3.93

Fe 4.63

Ca 3.20

Ba 2.52

In contrast, the work functions of the metals which do not chemisorb Nz a t room temperature (23) are relatively high: Ni 4.91

Pd 4.98

Rh (4.65)

Pt 5.36

At 90"K., however, Nz has already been adsorbed a t low equilibrium pressures by these metals, but a t room temperature it is given off again (23). Mignolet (14) observed a change of AUl,z = +0.21 volt of contact potential when he had Nz adsorbed on a nickel film a t low temperatures. Hence electrons are shifted from the molecule to the metal on adsorption of the N z molecule, which probably is t o be attributed t o the lone electrons of the Nz molecule. Nitrogen atoms, however, when adsorbed on platinum or tungsten surfaces catch electrons. Nitrogen gas acting on platinum at room temperature did not effect any change in the photoelectric sensitivity. The electron yield of the platinum surface a t 265.5 mp clearly decreased, however, if a Pt filament near the Pt surface was flashed to 1500°C. in nitrogen of 1.3 X mm. Hg, and a few N atoms were formed. The yield dropped t o unmeasurably low values if between the Pt surface and the Pt filament a t a nitrogen pressure of 1.3 X low2mm. Hg a glow discharge took place for 2 sec., in order t o produce N atoms (66). Bosworth and Rideal (67,68), too found with the contact-potential method a n increase of the work function by 1.38 volts when nitrogen was adsorbed on tungsten a t 90°K. Since the method described in section III,ld, was used, where a second tungsten filament is heated for the emission of thermionic electrons, it is to be expected that in this case, too, N atoms were formed, which reached the cold wire near the hot one and were adsorbed. f . Nitrous Oxide and Carbon Monoxide. If the bond of the atoms in a molecule is associated with an asymmetric electron shift as, e.g., with nitrous oxide and carbon monoxide (Fig. l), the molecule is able t o receive electrons from the adsorbent during adsorption. The electron

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

339

defect of single atoms in the molecule is thereby balanced, and the bond within the molecule is weakened. I n the metal-catalyzed decomposition of nitrous oxide-corresponding to this conception-not the liberated 0 atom [C. Wagner (lo)], but the NzO molecule receives electrons on adsorption (8,9,58). As the metal surface thereby loses electrons, it is to be expected that an increase of the work function will occur upon adsorption of Nz0 molecules, e.g., on platinum, even a t such low temperatures that the thermal decomposition of NzO does not occur.

Time. minutes

FIG.24. Adsorption of nitrous oxide on a platinum surface. Ordinate: photoelectric yield I in electrons per light quantum. = 265.5 mp; T = 83°K.; (a) smashing of the NzO capsule; ( b ) complete removal of the liquid air from NzO [according to ( 5 8 ) ] .

The results shown in Fig. 24 indicate that this assumption is correct. The photoelectric emission of a platinum surface a t X = 265.5 mb diminishes after the NzO capsule, still cooled with liquid air, has been smashed (F in Fig. 2). The emission current becomes entirely unmeasureable after removal of the cooling medium. This corresponds t o a considerable increase of the work function in the presence of NzO. Electrons bound by the NzO molecules on adsorption come fromthe metal-electron gas, as is seen from Fig. 25, which shows the influence of adsorbed NzO on the resistance of a transparent nickel film a t 90.3"K.

R. SUHRMANN

340

The maximum increase of resistance is 3.1 %, which is nearly the same as with oxygen if a covering of e = 1 is assumed. Furthermore, it is irreversible. According to this work the catalytic decomposition of nitrous oxide molecules proceeds in the following way: a N2O molecule adsorbed by the catalyst binds metal electrons, and thus the bond between the 0 atom and N2 in the molecule is loosened, and N2 is thermally dissociated from 0 at sufficiently high temperature. The 0 atom is held t o the surface through the influence of the metal electrons. It can combine with a neighboring

25 070

c

90.3'K

I

I

1

24 870 -

i

i

1.42

IB2

I

I

1

I

1 I

I

I

I

I

5 U al

5

c

24670 -

5

0) VI

a

24.470

24 270

i1

___)I I

0

, I

I

;

20

I

,

40 Time, minutes

I

I ,

60

I

I

80

FIG.25. Resistance increase of a transparent nickel film (90 X 10l6atoms/sq. cm.) on the adsorption of nitrous oxide a t T = 90.3"K. [according to (18)].

0 atom into an O2 molecule provided the energy of activation for such migration is available. The newly formed O2 molecules are desorbed. On recombination of 0 atoms, the previously bound metal electrons return to the electron gas. This mechanism requires binding of metal electrons t o the chemisorbed molecule without formation of a chemical bond (oxidation). The work function of the metal must be of medium height if the most favorable activation of the N 2 0 decomposition is desired. The energy of activation on gold (@ = 4.71 volts) is 29.0 kcal., but on platinum (a = 5.36 volts) 32.5 kcal., according to Hinshelwood and Prichard (69). The relative increase of the energy of activation (12.1 %) from platinum t o gold is about the same as that of the work function (13.8%). The magnetocatalytic effect, found by Hedvall and co-workers (70), may be similarly explained. I n this case the energy of activation for the

341

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

N 2 0decomposition on nickel increases suddenly a t the Curie point, owing t o a smaller electronic interaction in accordance with the foregoing explanation. On the other hand, the electronic work function of nickel, according t o Cardwell (71), above the Curie point is 0.2 volt higher than below it. The transfer of metal electrons to the NzO molecules, therefore, occurs less readily above the Curie point. Hence the bond between a n 0 atom and N2 in the adsorbed N2O molecule is not weakened so much a t temperatures above the Curie point. Thermal decomposition of NzO, therefore, requires a higher energy of activation. *

1

90.4'K

52.360

A,

.A,-COadmitted

I

A o 2 - 0 2admitted

im Ib

B1. . B5=CO pumped off Bo,- O2 pumped off

0

60

120

I

180 240 Time, minutes

300

360

420

FIG.26. Resistance increase of a transparent nickel film (90 X 10l6atoms/sq. cm.) on the adsorption of carbon monoxide a t T = 90.4"K. [according to (18)].

I n carbon monoxide the bond between the atoms depends, as in the NzO molecule, on a n asymmetrical electron shift, electrons of th e 0 atom moving toward the C atom, and the CO molecule having a dipole character. I n this case, too, metal electrons are displaced toward the adsorbed molecule and taken from the electron gas, as shown by the change of the electrical resistance of thin nickel films on carbon monoxide adsorption (18). Figure 26 shows the increase of resistance of such a film a t 90.4"K. with adsorption of CO. While a pressure of 10-6 mm. Hg does not give a measurable effect, the resistance increases suddenly a t mm. Hg. An increase of the pressure to lop3mm. Hg causes a n additional slow increase

* Dowden (19) has given a theory of the magnetocatalytic effect considering the increase of work function a t the Curie point and the disorientation of the spin of the metal electrons.

342

R . SUHRMANN

of resistance. Therefore, as the covering increases, an activation energy for the chemisorption is necessary. This activation energy would be produced by the place changes a t the nickel surface. After the surface has been covered with 0 = 0.89 molecule per surface Ni atom there is no further increase in resistance even if the pressure is increased to mm. Hg. The over-all increase in resistance is 0.83%. On pumping, the resistance change due to adsorption is found to be irreversible. g. Precovered Metal Surfaces. The method of measuring the resistance of thin films makes it possible to determine whether there is an electronic interaction between the metal surface and molecules which may be additionally adsorbed on an already adsorbed layer. I n Fig. 26, for example, the surface is covered with carbon monoxide ( 0 = 0.89). If at AOPoxygen gas at 6.1 X mm. Hg pressure is admitted, its molecules are still adsorbed under pressure to a covering of e = 0.02; however no further increase in resistance is observed and the oxygen can be pumped off completely. At a pure nickel surface a n oxygen pressure of 6.1 x mm. Hg would lead t o a resistance increase of 5.6%. Apparently the carbon monoxide layer of 0 = 0.89 blocks the surface so completely that a t 90.4"K. oxygen molecules do not dissociate into atoms, which could interact with the electron gas of the metal. Similarly a precovering of the nickel surface with oxygen to e = 1.13 molecules per surface Ni atom a t 90.5"K. blocks the surface against the influence of hydrogen. I n Fig. 15 hydrogen a t a pressure of 3.7 x lop3mm. Hg was added at AHZ.No resistance effect was observed when the surface had been precovered with oxygen. At the same pressure a hydrogen film on a pure nickel surface would have decreased the resistance by 1%. Even the precovering with hydrogen is able to block the surface against electronic interaction. In Fig. 27 a nickel surface was precovered with hydrogen at 3 X mm. Hg to saturation ( 0 = 0.39), causing an irreversible resistance decrease of 1%. After pumping off at B , carbon mm. Hg was added a t Ace. At a pure nickel surface monoxide of 6 X the carbon monoxide influence would have effected a n increase of the resistance by 0.8%. At the surface precovered with hydrogen, neither a resistance effect nor a carbon monoxide adsorption is t o be observed. Though no interaction of the molecules adsorbed in the second layer with the metal electron gas occurs, a polarization of the molecules adsorbed in this layer is possible by an induction influence of the first layer. For instance, Mignolet (14) observed a positive nitrogen film by contactpotential measurements when this gas was adsorbed on a nickel surface precovered with a negative hydrogen film. With the increase of the negative hydrogen potential the positive potential A U L of ~ the nitrogen film arose :

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

Surface potential of Ni-H, volts -0.072 -0.28 -0.30

AUl.2

( P N ~= 5 X

343

mm. Hg), volts

+O .36 4-0.45 f0.41

Hydrogen or xenon when adsorbed on a primary layer of CzH4 showed no polarization. Similarly, nitrogen was not polarized when adsorbed on a primary layer of CzHz (14). 26,5001

7~ 90.5'K

--

A H2 admitted Aco CO admitted B -gas pumped off

-

w

26400

5 0

t

I

i i

26.300

26.200

i

1 I

I" 14 I

! I I

0

I

60

120 180 Time, minutes

240

FIG. 27. Absence of resistance increase of a transparent nickel film (90 X 10l6 atoms/sq. cm.) a t carbon monoxide action ( A c o ) after precovering with hydrogen; T = 90.5"K. [according to (IS)].

k. Ammonia and Water Vapor. I n the case of adsorbed NHI molecules on conducting surfaces, the unpaired electrons of the nitrogen atom can join the metal electrons. The electron affinity of the metal surface is thereby partly saturated, and @ is lowered. Hallwachs (72), Leupold (73), and Brewer (74) therefore found a considerable increase of the photoelectric effect for platinum and iron surfaces when ammonia was present. The lone electrons of the 0 atom in the H2O molecule can also become part of the electron gas in the metal surface and reduce its work function. So Schaaff (75) observed an increase of the photoelectric emission of platinum in the presence of water vapor. On the other hand a n adsorbed layer of H 2 0 molecules on the surface of a thin nickel film decreases the electric resistance of the film (18). From the nature of the interaction between NHs and HzO molecules with platinum, iron, and nickel surfaces, the conclusion can be drawn that

344

R. SUHRMANN

the N atom of the adsorbed NH3 molecule and the 0 atom of the H2O molecule are directed toward the metal, while the H atoms are outside (9). i. Hydrocarbons. With saturated hydrocarbons no electronic interaction between the metal surface and the adsorbate is to be expected. Indeed Mignolet (14) has not observed any change of the contact potential upon contacting ethane with nickel at room temperature. If the adsorption takes place a t 90"K., however, the contact potential unexpectedly changes by the relatively large amount of A U l , z = +0.77 volt; i.e., the work function decreases. In other words the electrons of the C2He 32

; - 24 r

--u 01

5

0

16

E

a

0

--

Benzene, adsorbed on platinum Liquid air 1 mtn. removed from CsH6 Liquid air 2 min. removed from C6H6

180 240 30 360 . Time. minutes FIQ.28. Change of the photoelectric yield at X = 265.5 mp. of a platinum foil (in uacuo) on the adsorption of benzene molecules at T = 83°K. At the points marked 60

120

by arrows, benzene vapor of low pressure was admitted [according to (76)].

molecule are moved to the metal, an observation which is just as unexpected as the polarization of the Xe atoms on Ni surfaces, mentioned in section IV,Sb. More easily t o be understood are the effects observed when R electrons are present in the adsorbed molecule. Figure 28 shows the change of the photoelectric emission of a platinum surface covered with benzene (76). The benzene was contained in a capsule, which could be smashed magnetically (see F in Fig. 2). The tube, G in Fig. 2, was cooled by liquid air. At the points of the curve marked with arrows, the cooling of G was interrupted for 1 or 2 min., so that a small quantity of benzene molecules might be adsorbed a t the platinum surface. The sensitivity increased (Fig. 28) a t first and then decreased after passing a maximum, which was reached in the vicinity of the monomolecular covering ( B , C in Fig. 28).

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

345

While the adsorption of benzene molecules before the maximum was reached increased the sensitivity, the molecules condensed on the platinum surface after the maximum had been reached decreased the sensitivity (6, D,and F in Fig. 28). The excess of the benzene molecules, however, can be desorbed in about 30 min. if no further molecules strike the surface ( E and G in Fig. 28). The work function was lowered by the adsorption of the optimum benzene layer from 4.54 volts (in A ) t o 4.11 volts (in B). The 7r electrons were therefore displaced to the metal surface by the adsorption. 90.3'K

0

10

20 30 Time, minutes

40

50

FIG.29. Decrease of resistance of a transparent nickel film (90 X 10l6 atoms/ sq. cm.) on the adsorption of benzene molecules at T = 90.3"K. [according t o (IS)].

The electronic interaction between benzene and the metal surface may be made up of two effects: the polarization of t h e molecule, which may be concluded from the above-described research, and the shifting of the ?r electrons t o the metal surface t o become part of the metal electron gas, which has been hypothesized by Polanyi (77). The first effect has been shown in Fig. 28, the second apparently can be seen from the research (18) illustrated by Fig. 29, in which the change of resistance of a transparent nickel film was studied during the adsorption of benzene molecules.* As the temperature of the benzene capsule was 90"K., the evaporation velocity was so low that only a small number of benzene molecules struck the surface in unit time. The resistance therefore diminished only

* It is not impossible that the resistance of the nickel film in this research is diminished by H atoms dissociated from the adsorbed molecules by the influence of the nickel catalyst.

346

R. SUHRMANN

slowly after the smashing of the benzene capsule. It was lowered with increasing coverage by about 0.053 %. A similar behavior of the resistance of transparent nickel films was observed with the adsorption of triphenylmethane and naphthalene. I n these cases the resistance decreased by 0.033 and 0.035% (18). As does the adsorption of benzene, the adsorption of triphenylmethane and of copper phthalocyanine lowers the electronic work function of a platinum surface (76). 1 8 r 10-6

0

0

4

8

12 22 24 Time, hours

28

FIQ.30. Change of photoelectric yield a t X = 265.5 rnfi by cooling of a platinum foil covered with copper-phthalocyanine molecules; (a)cooling by liquid air; ( b ) liquid air evaporated [according to (77a)l.

The electronic interaction of the relatively large molecules of phthalocyanine shows (Fig. 30) a considerable temperature effect (77a). I n an experiment demonstrating this effect, the platinum foil (B in Fig. 2) was covered by the dye molecules until the work function was lowered t o 4.32 volts a t room temperature. If B was cooled by pouring liquid air into the upper tube of the photocell (a in Fig. 30), the photoelectric sensitivity increased and remained constant as long as liquid air was added. If the liquid air evaporated (6 in Fig. 30), the photoemission dropped to th e original value at room temperature. This effect was arbitrarily reproducible. The calculation of the work function @ and the constant M by the curves of Fowler [see Equation (5) in section III,la] in Fig. 31 gives Q., = 4.32 volts, log M = -12.17 at room temperature (curve I), arid 9 = 4.15 volts, log M = -12.17 a t low temperature (curve 11). While

347

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

-8.

-10

5 M

0 1

- 12

l

.

120 140

I

1

.

,

.

,

.

,

.

160 180 200

t

.

I

.

I

520 540 560 580

4

,

.

I

.

I

.

I

.

I

.

I

I

I

.

I

.

280 300

220 240 260 1

I

.

I

.

600 620 640 660 680 6

1

I

.

320

.

700

I

.

I

720 740

FIG. 31. Fowler curves of platinum surface covered with copper-phthalocyanine molecules as in Fig. 30 at room temperature (I) and a t low temperature (11)[according to (77a)l.

the constant M and therewith the number of the emitting centers remains unchanged, the work function decreases by 0.17 volt upon cooling. The lowering of the heat motion of the adsorbed dye molecules, which are bulky, favors the interaction of the a electrons with the metal surface.

V. THEORETICAL CONSIDERATIONS From the results of the investigation discussed in section IV one can conclude that for the type and the intensity of the electronic interaction between the metal surface and the adsorbed molecule, the following factors have t o be considered: magnitude of the electronic work function

348

R. SUHRMANN

of the metal surface, structure of the adsorbing crystal plane,* amount of the dissociation energy in the case of molecules containing two equal atoms, electron affinity of the species to be adsorbed, asymmetry of the electron configuration of the adsorbed species, presence of unpaired electrons and a electrons, and effects due to steric hindrance. I n addition the electron configuration of the metals plays an important part; the best catalysts for hydrogenation are to be found among the transition metals with unfilled electron levels in their 3d, 4 4 and 5d bands, which are able t o take up the valency electrons of the H atoms [Dowden (19)I.t The relative inactivity of copper, silver, and gold in this reaction is thus accounted for, because, in these metals those bands are filled. Moreover it is t o be expected that any factor which tends t o fill up the d bands in Group 8 metals will result in a corresponding decrease of their activity (79a); thus the activity of nickel declines as a function of the filling of its 3d band by the valency electrons of the copper in Ni-Cu alloys [Reynolds (SO)]. I n the same manner within the domain of the a phases of Hume-Rothery alloys, the activation energy increases with the excess electron number of the solute. According to Schwab (6) the activation energy depends on the degree of electron saturation of the first Brioullin zone of the metal. Electron transfer from the adsorbed particle toward the metal is t o be expected especially with transition metals serving as catalysts and adsorbed particles having unpaired or a electrons. Electron transfer from the metal toward the adsorbate requires only that the metal be able t o deliver electrons coming from s or p bands. Regarding the influence of the electron work function a, ionization potential of the metal atoms, and the gradient d In g(c)/dE of the electron level density with electron energy a t the Fermi surface, Dowden (19) has given a theory on electronic interaction showing that positive ion formation is favored on metals and alloys possessing a large value of Q and a large positive value of d In g(E)/de. The entropy of activation is A S * = T(d In g(E)/de)*=-@

(8)

and the rate of ionization k+

=

Kf exp [-b(I* ipeM)/(IcT)I

(9)

* According to the investigation of Rhodin (78) on the adsorption of nitrogen on single crystal copper surfaces a t low temperatures, the heat of adsorption depends on the type of the crystal plane even in the case of van der Waal’s adsorption, as is to be expected according to Stranski and Suhrmann (51). It has the highest amount a t the monomolecular covering on (011) planes and decreases in the succession of (0111, fool), (111). t See also ref. (79).

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

349

where I* = I - aAU+, K+ = a transmission factor, and b is a constant. I is the ionization potential of the neutral particle, AU+ the adsorption energy of the ion and peM the thermodynamic potential per metalelectron per unit volume: peM = - @ - /I6/ S 2k 2 T2(d In g(e)/de)*=-* (10)

A negative ion is formed by removal of an electron from the highest occupied level -@ ‘v weM of the crystal face to the lowest unoccupied level ( - E = electron affinity a t the surface) of the adsorbed particle. The entropy of activation is A&*

01

- T(d In g[E)/de),=-a

(11)

and the rate constant for ionization

k-

=

K- exp [ - d ( E

+ cAU- - p e M ) / ( k T ) ]

(12)

c and a! are constants. Negative-ion formation is favored by a low value of @ and a large negative value of d In g ( c ) / d e . Besides the formation of positive and negative ions, covalent bonding can arise. Electron exchange bonding effected by resonance energy results in either a metallic bond or a covalent bond. I n the first case the bonding electrons are more or less mobile; in the second, they are localized between the bound atoms to give a directed bond. As we saw in section IV,2 these and intermediate bond types can arise. According to Dowden, the localized bond properties follow from Pauling’s (81) valencies for the transitional series of metals; this means that covalent-bond formation should be favored on metals with unfilled “atomic orbitals.” The strongest covalent bonding of a particle would result, when the density of d levels is largest compared with that of the s, p levels a t the Fermi surface. This maximum in d-level density occurs at pure nickel and the corresponding metals in the transitional series. Covalent bonding in chemisorption is favored at surfaces with large electronic work functions, large values of g( -@), large positive values of (d In g(e)/dE)e,-a, and of the presence at the surface of “atomic orbitals” in the Pauling sense. VI. CONCLUSION The discussions of section V indicate that the theoretical considerations concerning the interaction between the metallic catalyst and the substratum have been advanced far enough in some cases to allow certain predictions to be made about the effects to be expected. The experimental foundation, described in section IV, necessary for testing and developing the theory, however, is rather incomplete in relation t o the theory. For

350

R. SUHRMANN

this reason the methods of investigation of the electronic interaction have been t,reated in some detail in section 111. Future application of these methods may be expected to clarify considerably the fundamental processes connected with heterogeiious catalysis on metallic surfaces.

REFERENCES 1. Suhrmann, It., and Csesch, H., 2.physik. Chem. B28, 215 (1935). 2. Suhrmann, R., 2.Elektrochem. 36, 681 (1929). 2a. Suhrmann, R., Physik. Z. 30, 939 (1929). 3. Suhrmann, R., 2. anorg. u. allgent. Chem. 203, 235 (1931). 4. Suhrmann, R., Ergeb. erakt. Naturwiss. 13, 148 (1934). 5. Schwab, G.-M., and Hols, G., Z. anorg. u. allgem. Chem. 262, 205 (1944). 6. Schwab, G.-M., Trans. Faraday SOC.42, 689 (1946). 6a. Schwab, G.-M., 2.Elektrochem. 63, 274 (1949). 7. Schwab, G.-M., Discussions Faraday Sac. No. 8, 166 (1950). 8. Suhrmann, R., Angew. Chem. 63, 244 (1951). 9. Suhrmann, R., 2. Elektrochem. 66, 351 (1952). 10. Wagner, C., J. Chem. Phys. 18, 69 (1950). 11. Bosworth, R. C. L., Proc. Cambridge Phil. Sac. 33, 394 (1937). 12. Bosworth, R. C. L., and Rideal, E. K., Proc. Roy. Sac. (London) A162, 1 (1937). 13. Eley, D. D., and Rideal, E. K., Proc. Rag. Soc. (London) A178, 429 (1941). 13a. Bosworth, R. C. L., J . Proc. Roy, Sac. N. S . Wales 79, 53 (1946). 14. Mignolet, J. C. P., Discussions Faraday Sac. No. 8, 105, 326 (1950). 15. de Boer, J. H., and Kraak, H. H., Rec. trav. chim. 66, 1103 (1937). 16. Verwey, E. J. W., and de Boer, J. H., Rec. trav. chim. 66, 675 (1936). 17. Suhrmann, R., and Schula, K., Naturwissenschaften 40, 139 (1953). 18. Suhrmann, R., and Schula, K., 2. physik. Chem. N.F. 1, 69 (1954); J . Colloid Science Suppl. 1, 50 (1954). 19. Dowden, D. A., J . Chem. Sac. 1960, 242. 19a. Boudart, M., J. Am. Chem. Sac. 74, 3556 (1952). 20. Hauffe, K., and Engell, H. J., 2.Elektrochem. 66, 366 (1952). 21. Aigrain, P., and Dugas, C., 2.Elektrochem. 66, 363 (1952). 22. Weiss, P. B., J . Chem. Phys. 20, 1483 (1952); 21, 1531 (1953). 23. Trapnell, B. M. W., Proc. Roy. Sac. A218, 566 (1953). 24. Heavens, 0. S., Proc. Phys. Sac. B66, 788 (1952). 25. Rupp, E., 2. Elektrochem. 36, 686 (1929). 26. Germer, L. H., Phys. Rev. 26, 795 (1925). 27. Suhrmann, R., and Hayduk, H., 2. Physik 96, 726 (1935). 28. Fowler, R. H., Phys. Rev. 38, 45 (1931); see also L. A. Dubridge, “New Theories of the Photoelectric Effect,” Hermann, Paris, 1935. 29. Beeck, O., Smith, A. E., and Whecler, A., Proc. Roy. Sac. A177, 62 (1941). 30. Benjamin, M., and Jenkins, D. R., Proc. Roy. SOC.176,262 (1940); 180,225 (1942). 31. Schenk, D., Z. Physik 98, 753 (1936). 32. Mahl, H., Mineral. u. petrog. Milt.46, 289 (1935). 33. Muller, E. W., 2.Physik 106, 541 (1937); 108, 668 (1938). 34. Muller, E. W., 2.Physik 126, 642 (1949); Ergeb. exakt. Naturwiss. 27, 290 (1953). 35. Johnson, R. P., and Shockley, W., Phys. Eev. 49, 436 (1936). 36. Martin, S. T., Phys. Rev. 66, 947 (1939). 37. Becker, J. A., Bell System Tech. J . 30, 907 (1951).

ELECTRONIC INTERACTION ON METALLIC CATALYSTS

35 1

38. Rothe, H., Z. tech. Phys. 6, 633 (1925). 39. Langmuir, I., and Kingdon, K. H., Phys. Rev. 36, 129 (1929). 40. Eley, D. D., Advances i n Catalysis 1, 163 (1948). 41. Thomson, W., Phil. Mag. [5] 46, 91 (1898). 42. Zisman, W. A., Rev. Sci. Instr. 3, 367 (1932). 43. Potter, J. G., Phys. Rev. 68, 623 (1940). 44. Frost, A. A., and Hurka, V. R., J . Am. Chem. Sac. 62, 3335 (1940). 45. Rosenfeld, S., and Hoskins, W. M., Rev. Sci. Znstr. 16, 343 (1945). 45a. Mignolet, J. C. P., Bull. soc. roy. sci. LiBge Nos. 8, 9 and No. 10, 401 (1950). 46. Suhrmann, R., and Kundt, W., 2. Physik 121, 118 (1943). 47. Suhrmann, R., and Earth, G., Z . Physik 103, 133 (1936); Physik. Z. 36,971 (1934). 48. Stranski, I. N., and Suhrmann, R., Ann. Physik [6] 1, 153 (1947). 49. Smirnow, B. G., and Shuppe, G. N., J . Tech. Phys. (U.S.S.R.) 22, 973 (1952). 50. Nichols, M. H., Phys. Rev. 67, 297 (1940); Herring, C., and Nichols, M. H., Revs. Mod. Phys. 21, 185 (1949). 51. Stranski, I. N., and Suhrmann, R., Ann. Physik [6] 1, 169 (1947). 52. Suhrmann, R., Naturwissenschaften 37, 329 (1950). 52a. Kossel, W., Leipzig. Vortruge 1928, 18. 53. Stranski, I. N., 2. physik. Chem. B11, 342 (1931); Stranski, I. N., and Kaischew, R., 2. Krist. 78, 373 (1931); 2. physik. Chem. B26, 312 (1934). 53a. Drechsler, M., 2. Elektrochem. 68, 327, 334, 340 (1954). 54. Taylor, J. B., and Langmuir, I., Phys. Rev. 44, 423 (1933). 54a. Mayer, H., Ann. Physik [5] 33, 419 (1938). 55. Mayer, H., Naturwissenschaflen 26, 28 (1938). 56. Brady, J. J., and Jacobsmeyer, V. P., Phys. Rev. 49, 670 (1936). 57. Rideal, E. K., and Trapnell, B. M. W., Proc. Roy. SOC.A206, 409 (1951). 57a. Scheuble, W., 2. Physik 136, 125 (1953). 58. Suhrmann, R., and Sachtler, W., Proc. Intern. Symposium Reactivity of Solids, Gothenburg (1952); 2. Naturforsch. 9a, 14 (1954). 59. Fianda, F., and Lange, E., Z. Elektrochem. 66, 237 (1951). 60. Suhrmann, R., and Theissing, H., Z. Physik 62, 453 (1928). 61. Klauer, F., Ann. Physik [5] 20, 909 (1934). 62. Suhrmann, R., Z. Elektrochem. 37, 678 (1931). 62a. Trapnell, B. M. W., Proc. Roy. SOC.A206, 39 (1951). 63, Rideal, E. K., and Trapnell, B. M. W., J. chim. phys. 47, 126 (1950). 63a. Eucken, A., 2. Elektrochem. 63, 274 (1949); 64, 108 (1950). 63b. Smith, J. W., Science Progr. 59, 669 (1951). 64. Miller, A. R., Discussions Faraday Sac. No. 8, 57 (1950); Proc. Cambridge Phil. SOC.46, 292 (1946). 65. Mignolet, J. C. P., J . Chem. Phys. 20, 341 (1952). 66. Suhrmann, R., Hermann, H., and Goldmann, G., not yet published. 67. Bosworth, R. C. L., and Rideal, E. K., Physica 4, 925 (1937). 68. Bosworth, R. C. L., J . Proc. Roy. Soc. N . S. Wales 79, 166 (1946). 69. Hinshelwood, C. N., and Prichard, C . R., Proc. Roy. Sac. (London) A168, 211 (1925). 70. Hedvall, J. A., Hedin, R., and Persson, O., Z. physik. Chem. B27, 196 (1934). 71. Cardwell, A. B., Phys. Rev. 76, 125 (1949). 72. Hallwachs, W., Physik. 2. 21, 561 (1920). 73. Leupold, H., Ann. Physik [4] 82, 841 (1927). 74. Brewer, A. K., J . Am. Chem. SOC.64, 1888 (1932).

352

R. SUHRMANN

75. Schaaff, E., 2. physik. Chem. B26, 413 (1934). 76. Suhrmann, R., and Sachtler, W., “Arbeitstagung Festkorperphysik, Dresden, 1952,” p. 74. Verlag der Wissenschaften, Berlin. 77. Polanyi, M., 2.Elektrochem. 36,561 (1929). 77a. Suhrmann, R., and Goldmann, G., “ Arbeitstagung Festkorperphysik, Dresden, 1954,” p. 188. Verlag der Wissenschaften, Berlin. 78. Rhodin, T. N., J. Am. Chem. SOC.72, 5691 (1950). 79. de Boer, J. H., Chem. Weekblad 47, 416 (1951). 79a. Couper, A., and Eley, D. D., Discussions Faraday SOC.No. 8, 172 (1950). 80. Reynolds, P. W., J. Chem. SOC.1960, 265. 81. Pauling, L., Phys. Rev. 64,899 (1938); Proc. Roy. SOC.(London) A196, 343 (1949). 82. Schenk, D., Ann. Physik 151 23, 240 (1935). 83. Drechsler, M., and Miiller, E. W., 2. Physik 134, 208 (1953). 84. Suhrmann, R., and Berger, W., 2.Physik 123,73 (1944). 85. Germer, L. H., 2. Physik 64,408 (1929).

Author Index Numbers in italics refer to pages _ - on which complete references are listed at the end of each article.

A Adam, N. K., 200, 2 f O Adkins, H., 250(85), 267 Aigrain, P., 219, 244(22), 266, 264, 301, 305, 360 Allen, J. A., 18, 44, 65, 73 Almquist, J. A., 37(72), 44 Alpert, D., 132, 134 Anderson, J. S., 33, 44, 53, 73, 215, 219 (25), 225, 254, 266, 266, 267, 270, 271, 277(32, 40), 278, 284, 288(32), 289, 292, 294, 301 Antipina, T. V., 228, 266 Ashworth, F., 189(14), 210

B Balandin, A. A., 76(6), 78, 91 Baldock, G. R., 263, 277(14), 300 Baldt, R., 80(15), 82(15), 88(15), 91 Bardeen, J., 219(33, 34, 35), 229, 866, 260, 263, 300 Barnes, G., 97(6), 133 Barth, G., 318(47), 319(47), 362 Baumbach, H. H., 35,44, 277(41), 301 Becker, J. A., 141(3), 142(3), 145(3), 160 (7), 169(9), 178(12), 180(12), 190 (16), 197, 210, 310, 360 Bedworth, R. E., 16, 43 Beeck, O., 13(19), 16, 18, 21, 22, 30, 38, 43, 44, 141(2), 159(2), 208, 210, 319 (29), 360 Bbnard, J., 216, 266 Benjamin, M., 310, 322, 360 Benton, A. F., 37, 44 Berger, W., 316(84), 362 Bevan, D. J. M., 33, 44, 53, 73, 219(25), 225(36), 266, 266, 270, 271, 277(32, 40), 278, 284, 288(32), 289, 292, 294, SO1

Bharucha, N. R., 16, 43 Birckenstaedt, M., 248(81), 267 Bizette, H., 66(46), 73 Block, J., 69, 74, 219(30), 238(64), 242, 246, 247, 266, 267, 277(42), 301 Blue, R. W., 59, 73, 254, 267 Born, M., 83, 91 Bosworth, R. C. L., 13, 43, 118, 134, 206, 210,305,313,326,327,337,338,SBO, 361 Bouch-Bruevich, V. L., 236, 266 Boudart, M., 13, 21, 43, 44, 52(16), 60 (16), 64(40), 73, 216(14), 238(63), 239, 266, 267, 263(16), 500, 305, 360 Brady, J. J., 326, 361 Brandes, R. G., 178(12), 180(12), 210 Brattain, W. H., 219(33, 34), 229, 266, 260, SO0 Brauer, P., 225, 266 Bray, W. C., 37, 44 Bremer, H., 243(69), 267 Brewer, A. K., 1, 13(6), 43, 343, 361 Brotz, W., 89, 91 Brownlee, L. D., 67(53), '74 Bube, R. H., 260, 294, 300 Burmann, R., 248(81), 267 Burwell, R., 51(11), 73, 234(52), 266

C

Cabrera, N., 233(43), 266 Calvert, J. C., 230(40), 266 Cardwell, A. B., 341, 361 Chapman, P. R., 254(99), 267 Clark, A., 38, 42(77), 44, 253, 267 Clarke, E. N., 260, 300 Cole, W. A., 16, 43 Conrad, F., 80(16), 82(16), 91 Constable, F. H., 76(2), 77, 86, 90 Couper, A,, 13, 24, 43, 52(18), 73, 348 (79a), 366 363

354

AUTHOR INDEX

Cremer, E., 76(7), 77(4), 78, 79, 80(3, 7,

D Davidov, B., 219, 266 Davis, J. W., 250(85), 267 Davis, R. T., Jr., 203, 211 de Boer, J. H., 294, 301, 305, 327, 328, 348(79), 360, 552 Dell, R. M., 35, 36, 44, 64(43), 73, 242, 257 Dikhoff, J. A. M., 298, 502 Dilke, M. H., 19, 44 Dixon, J. K., 41(84), 45 Dodson, R. W., 84, 91 Dolan, W. W., 97(6), 133 Dowden, D. A., 14, 16, 21, 26, 32, 40, 42, 43, 44, 45, 52(18), 73, 215, 239, 255, 260,264,300, 301,305,341, 348, 360 Drechsler, M., 100, 133, 156, 210, 312 (83), 325, 351, 352 Drikos, G., 37(74), 44 Dubar, L., 219(35), 256 Duhridge, L. A., 307(28), 350 Dugas, C., 219, 244(22), 255, 264, 301, 305,350 Dushman, S., 200, 210 Dyke, W. P., 96, 07, 133

F

Fianda, F., 331, 351 Flugge, S., 87(29), 91 Fowler, R. H., 94, 133, 174(10a), 210, 305,360 Franck, J., 83, 91 Frankenburg, W. G., 203, 211, 244, 267 Frazer, J. C . W., 37, 44 Fritsch, O., 275, SO1 Fritzche, H., 53, 54(23), 55, 73, 280, 283, 289, 301 Frost, A. A., 314, 551 Frost, A. V., 228, 266 Fujita, H., 118(19), 134 Fujita, Y., 246(76, 77), 257

G

Garner, W. E., 32, 33, 34, 44, 64, 73, 215, 243, 256, 257, 260, 271, 300 Gensch, C., 56(30), 75, 254, 267 Germer, L. H., 307, 313, 360, 352 Glang, R., 71, 74, 216(14a), 238(14a), 239 (14a), 2566 Glasstone, S., 76(1), 90 Goldmann, G., 338(66), 346(77a), 347 (77a), 351, 366 Comer, R., 100(8), 101(9), 102(10), 104 (8, 10, 13a), 105(13a), 109(13a), 112 (14), 113(14, 16), 117(18), 118(9), 125 E (13a), 126(13a), 130(21), 133, 134 Gray, T. J., 32, 33, 34, 44, 64, 73, 215(7), Eckell, H., 78, 91 234, 243(71), 256, 256, 267, 260(2), Eley, D. D., 13, 16, 18, 19, 24, 26, 43, 44, 27 1(2), 300 52(18), 73, 305, 313, 348(79a), 360, Griffith, R. H., I, 43, 254, 267 351 Crirnm, H. G., 76(5), 78, 91 Elovich, S. Y., 233, 256, 274, 301 Gryder, J. W., 84, 91 Emmett, P. H., 1 3 , 16, 43 Gurevich, D. B., 236(57), 266 Engell, H. J., 61, 62(37), 63, 71, 73, 74, Gurney, R. W., 48(1), 72, 214(2), 246(2), 216(14a), 218, 219, 227, 233(44), 234, 256 239(14a, 17), 244(16), 855, 256, 282, H 301, 305, 360 Eucken, A., 237(62), 250, 256, 257, 336 Haayman, P. W., 67(52), 74 Hager, G. F., 250(85), 257 (63a), 361 Hahn, E. E., 219(27), 225, 266, 266(24), Evans, S. W., 230(41), 256 268, 269, 270(24), 276, 279(24), 287 Ewles, J., 260, 298, 300 (24), 299(24), 301 Eyring, H., 70, 85, 90, 91

355

AUTHOR INDEX

Haijman, P. W., 239(66), 657 Hallwachs, W., 343, 351 Hannan, M. C . , 230(41), 256 Hansen, M., 115, is4 Hansford, R. C., 40, 41, A5 Harrison,. S., 266(26), 275, 282(26), 287 (26), 301 Hartman, C. D., 160(7), 210 Hass, K., 66, 73 Hauffe, K., 48, 56(30), 61, 62(37), 63, 67 (xi), 68(56), 71, 76, 73, 74, 213(i), 215, 216, 217(1), 218, 219(30), 233 (44), 234, 238(1, 3, 14a, 64), 239, 242, 244(16), 254, 255, 266, 267, 277(42), 282, 301, 305, 350 Hayduk, H., 307, 350 Heap, C. N., 260, 298, 300 Heavens; 0. S., 306, 350 Hedin, R., 340(70), 351 Hedvall, J. A., 340, 351 Heiland, G., 283, 295, 301 Henisch, H. K., 219(28), 265 Herington, E. F. G., 19, 44, 254, ZC57 Hermann, H., 338(66), 351 Herring, C., 100, 112, 1-93, 190, 210, 320, 351

Heuer, K., 250(89), 267 Hinshelwood, C. N., 340, 351 Hogarth, C. A., 266(27), 277(27), $01 Holm, V. C . F., 59, 73, 254, 257 Holz, G., 305(5), 350 Hoskins, W. M., 314, 361 Houghton, S., 250(87), 257 Hoyle, K., 250(85), 267 Hiittig, G. F., 55, 73, 236(61), 250, 266 Huizenga, J. R., 84, 91 Hulm, J. K., 117(18), 134 Hume-Rothery, W., 2(11), 5, 2 8 ( l l ) , 43 Hurka, V. R., 314, 361

I Inghram, M. G., 104(13a), 105(13a), 109 (13a), 125(13a), 126(13a), 133 Intemann, K., 266(25), 280, 288(25), SO1

J Jacobsmeyer, V. P., 326, 361 Jenkins, D. R., 310, 322, 323, 360

Johnson, D., 254, 257 Johnson, R. P., 310, 322, 550 Jones, H., 2(10), 5, 6, 8(10), 43

K Kahn, M., 84, 91 Kaischew, R., 324(53), 361 Keier, N. P., 51, 73 Kellerl N . ~35, 4 4 p 257 Kernball, C.1 23,441 521 T5 Kerber, 80(19)J 82, 91 Kingdom, K. H., 313, 351 Kingman, F' " 32J I 4 Kinuyama, T., 246(76), 257 Kirchner, F., 106(13b), 133 Kirkpatrick, W. J., 250(84), 257 Kistiakowsky, G. B., 50(3), 72, 290, 292, 239j

R'j

J.j

$01

Klauer, F., 332(61), 351 Klein, R., 113, l S 4 , 197, 199, 210 Klemm, W., 61(34), 66, 73 IiohIschiitter~ H. w . ~ 527 " Kossel, W., 324, 351 H. H.l 3057 327, 328~"O Krauss, W., 61, 64, 65(36), 73, 244(11), 216, 255 Kroeger, F. A., 298, 301 KropaJ .' L., 41(84), 46 Krucke, E., 252, 257 Krutter, H. M., 18, 44 Krylova, E. S., 236(57), 266 Kullich, E., 80(8),81(8), 91 Kundt, W., 315(46), 351 Kwanr T . ~ 2457 246, 257

L Laidler, K. J., 76(1), 90, 244, 257 Lange, E., 331, 351 Laugmuir, I., 1, 43, 140, 141(4), 142(4), 174(10), 210, 313, 324, 351 Latimer, W. M., 290, 301 Lauder, I., 65, 7S Le Blanc, M., 60(33), 73 Lennard-Jones, J. E., 1, 12, 43, 156,168, 210, 218, 655 Leupold, H., 343, 361 Leverene, H. W., 298, 301 Liang, S. C., 50(4), 78, 234, 266, 291(51), 293, 301

356

AUTHOR INDEX

Lindars, P. R., 254(99), 267 Ljaschenko, W. J., 219(29), 266

M MacNevin, W., 230(40), 256 Magnusson, L. B., 84, 91 Mahl, H., 310, 311, 360 Manning, M. F., 18, 44 Marcus, R. J., 85, 91 Markham, M. C . , 230(41), 256 Marschall, E., 77, 80(3), 91, 243, 267 Martin, S. T., 144, 210, 310, 350 Maxted, E. B., 19, 20, 44 May, D. R., 41(84), 46 Mayer, H., 326(55), 361 Mayer, J. E., 94(3), 133 Mayer, M. G., 94(3), 133 Melnick, D., 260, 274, 283, 284, 294, 295, 300

Meyer, E. G., 84, 91 Meyer, W., 267, 276, 288(30), 301 Miesserov, K. G., 39, 46 Mignolet, J. C . P., 13, 43, 305, 314, 327, 337(14), 338, 342, 343(14), 344, 350, 361

Mikovsky, R. J., 81(31), 81 Miller, A. It., 205(24), 211, 336(64), 361 Miller, P. H., Jr., 219(26), 225, 255, 271, 270, 278, 279(43), 282, 298(33), 301 Milliken, T. H., 40, 41, 46 Mills, G. A., 40, 41(81), 46 Mitchell, E. W. J., 67(53), 74 Mitchell, J. W., 18, 44 Molinari, E., 55(28a), 73, 79, 91, 253, 254 (94), 267 Mollwo, E., 295, 301 Mooi, J., 32, 36, 37, 44 Morin, F. J., 66, 67(49), 68(49), 73 Morrison, S. R., 229, 266, 260, 270, 271, 272, 278, 280, 282, 283, 288(31), 291 (31), 292, 296, 298(33), 300, 301 Mott, N. F., 2(10), 6, 8(10), 43, 48(1), 72, 214(2), 219(31), 246(2), 266, 266 Muller, E. W., 93, 97, 100, 102(1, lo), 103(1, 11, 12), 104(1, lo), 105(1), 112, 117, 119, 130, 133, 175(11), 192, 193, 197, 198, 210, 310, 312, 325, 330, 360, 362

Muscheid, E., 228, 266

N Nakata, S., 71, 74 Neldel, H., 267(30), 276(30), 288(30), 301 Neuhaus, A., 61(35), 73 Nichols, M. H., 100, 112, 133, 320, 361 Nordheim, L. W., 94, 96, 133 Nyrop, J. E., 1, 13(8), 43, 264, 301 0

Oatley, C. W., 13, 43 Oblad, A. G., 40, 41(81), 46

P Pace, J., 52, 73 Parravano, G., 37, 44, 55(28, 28a), 58 (28a), 62, 66(48), 67(55), 68(28, 55), 70(55), 73, 74, 246, 248, 253, 254,267 Patat, F., 79, 85(27a), 91 Pauling, L., 8, 10, 17, 25, 40(83), 43, 46, 349, 362 Pearson, G. L., 219(34), 266 Persson, O., 340(70), 361 Pilling, N. B., 16, 43 Pitaer, E. C . , 37(73), 44 Polanyi, M., 83(21), 91, 345, 362 Potter, J. G., 314, 361 Prater, C. D., 252(92), 667 Prichard, C. R., 340, 361 Pyahev, V., 61(39), 64, 73

R Rabes, I., 236(60), 266 Rae, D. S., 250(85), 267 Ramanathan, K. V., 234(49), 266 Rankin, G. T., 230(40), 266 Reboul, T. T., 288(58), 298, 301 Rees, A. L. G., 67(51), 74 Reid, W. D., 22, 44 Reynolds, P. W., 26, 44, 260, 300, 348, 362 Rhodin, T. N., 348, 368 Richter, G., 102(10), 104(10), 133 Rideal, E. K., 1, 13, 16, 19, 42, 43, 44, 46, 206, 211, 254, 267, 305, 313,326, 327, 328, 334, 337, 338, 360, 361 Rienacker, G., 78, 91, 243, 248, 267

357

AUTHOR INDEX Rittenhouse, K. D., 252(92), 267 Roberts, J. K., 205, 211 Roberts, L. E. J., 254(95), 867 Roginskil, S. Z., 1, 42, 43, 46, 51, 62, 73, 246, 254, 867 Romeijn, F. C., 67(52), 74, 239(66), 267 Rooksby, K., 66(47), 73 Rose, D., 188, 810 Rosenfeld, S., 314, 351 Rothe, R., 313, 361 Rubin, T. R., 230(40), 256 Ruedl, E., 79, 82(12), 91 Rupp, E., 307, 360 Russell, W. W., 29, 44

S Sachsse, H., 60(33), 73 Sachtler, W., 305, 331(58), 335(58), 336 (58), 339(58), 344(76), 346(76), 362 Sarry, B., 78(11), 91 Sastri, M. V. C., 234(49), 866 Saunders, K. W., 41(84), 46 Schaaff, E., 343, 361 Scharowsky, E., 266(28), 267, 287(28), 288(28), 301 Schenck, R., 236, 866 Schenk, D., 309, 310, 311, 360, 368 Scheuble, W., 330, 351 Schmid, G., 35, 44, 239, 267 Schmidt, O., 1, 13(7), 43 Schottky, W., 219, 866 Schrader, R. E., 298, 301 Schuit, G. C . A., 22, 44, 294, 301 Schultes, H., 35,44,236(58), 238(58), 866 Schulz, I<., 230(42), 266, 260, 300, 305, 317(18), 318(18), 319(18), 327(18), 328( 18), 329 (18), 333 (18), 340( 18), 341(18), 343(18), 345(18), 346(18), 360 Schwab, G. M., 27, 28, 35, 37, 44, 69, 74, 78, 87(28), 91, 236, 238, 243, 246, 247,255,267,264,351,305, 348,360 Schwamberger, E., 76(5), 78, 91 Seitz, F., 214(2), 856 Selwood, P. W., 32, 36, 37, 44, 249, 267 Shelton, J. P., 219(25), 266, 277(40), 278, 301

Sheridan, J., 22, 44 Shive, J. N., 219(35), 266

Shockley, W., 219(34), 866, 263, 300, 310, 322, 323, 360 Shul'tz, E., 1, 43 Shuppe, G. N., 320, 361 Sickmann, D. V., 50(7), 73, 234, 266, 267 (29), 284(29), 288(29), 293, SO1 Silvermann, J., 84, 91 Smirnow, B. G., 320, 361 Smith, A. E., 18, 44, 319(29), 360 Smith, E. A., 51(13), 57, 73, 235,866, 291 (50), 293, 301 Smith, J. W., 336(63b), 361 Smoluchowski, R., 100, 133 Speer, D. A., 112(14), 113(14), 133 Stadlmann, W., 80(18), 82(18), 91 Staegar, R., 35, 44 Steiner, H., 250, 867 Stepko, J. J., 219(29), 266 Stockmann, F., 54, 73, 266(25), 277, 279, 280, 288(25, 39), 293, 301 Stone, F. S., 32, 33, 34, 35, 36, 44, 64(43), 73, 215(7), 242, 243(72, 73), 244, 866, 267, 260(2, 3), 271(2), 300 Stowe, R. A., 29, 44 Stranski, I. N., 320(48), 321, 322, 323, 324, 348, 361 Strother, C. O., 51(9), 73, 233, 866, 280, 291(44), 292, 301 Suhrmann, R., 230,'866, 260, 300, 304, 305, 307(1), 308(9), 315(46), 316 (9, 84), 317(18), 318(18, 47), 319 (18, 47), 320(48), 321, 322, 323, 324, 327(18), 328(18), 329(18), 331(58), 332(60, 62)) 333(18), 334(2a), 335 ( 5 8 ) , 336(58), 338(66), 339(8, 9, 58), 340(18), 341(18), 343(18), 344(9,76), 345(18), 346(18, 76, 77a), 347(77a), 348, 550, 361, 362

T Tamele, M. W., 40, 41, 46 Tamrn, I., 219, 266, 263, 300 Taylor, H. A., 50(5), 73, 234, 236, 266 Taylor, H. S., 50(3, 4, 7), 51(8, 9, 11, 13), 52(16), 57, 60(16), 72, rs, 168, 194, 810, 216, 233, 234(52, 53), 235, 666, 266, 267(29), 273, 274(36), 280, 284 (29), 288(29), 290, 291(44, 50, 51), 292, 293, 301

358

AUTHOR INDEX

Taylor, J. B., 141(4), 142(4), 210, 324, 351 Teller, E., 13, 16, 43 Temkin, M. I., 61(39), 64, 73 Theissing, H., 332(60), 351 Thomas, C. L., 39, 40, 44 Thomson, W., 313, 361 Thon, N., 50(5), 73, 234, 236, 256, 274 (36), 293, 301 Tiley, P. F., 32, 33, 34, 35, 36, 44, 242, 243(72, 73), 244, 257, 260(3), 300 Tolstoi, N. A., 236(57), 256 Trapnell, B. M. W., 1, 13(18), 18, 42, 43, 44, 45, 306(23), 328, 334, 338, 350, 351 Trolan, J. K., 96(5), 97(6), 133 Tselinskaya, T. S.,62, 73, 246, 257

U Unger, S., 243(69), 257

V Van Cleave, A. B., 205, 211 Verwey, E. J. W., 67, 74, 239(66), 267, 305, 327, 350 Vierk, A. I,., 56, 73 Volger, J., 225, 256 Volkenshtein, F. F., 60, 73, 235, 236, 256, 264, 301 Voltz, S. E., 38, 44, 52(17), 73

von Baumbach, H. H., 53, 73, 236(59), 238(59), 256

W Wagner, C., 33, 34, 44, 48, 53, 55, 67, 72, 73, 74, 213, 215, 217(1), 238(3, 4), 255, 277(41), 301, 305, 339, 350 Wansbrough-Jones, 0. H., 1, 16, 43 Ward, A. F . H., 16, 43 Waters, R. F., 81(31),91 Weidlich, P., 79, 91 Weiss, H., 295, 301 Weisskopf, V., 83, 91 Weisz, P. B., 219, 224, 239, 244(23, 24), 252,255, 257, 264, 272, 301, 305, 350 Weller, S., 38, 44, 52(17), 73 Weyl, W. A., 215, 254, 266, 267 Wheeler, A., 16, 18, 43, 44, 208, 211, 319 (as), 350 Whitsell, W. A., 37(73), 44 Wicke, E., 51, 73, 89, 91, 250, 252, 257 Williamson, A. T., 234(53), 256 Wilson, A. H., 30, 44 Winter, E. R. S., 250(86), 257 Z Zhabrova, G. M., 233, 256, 274(35), 301 Zisman, W. A., 314, 351 Zwolinsky, B. J., 85, 91

Subject Index A Activation of catalyst, 62 Adsorption activated, 51 effect of crystal plane, 140 in catalysis, 138 of molecules or atoms, 136 slow, 50, 173 Alkali metals on W and Pt, 326 on Al, 326 Alloys compensation effect in, 78 electron concentration in, 5 Ammonia deuterium exchange, 52 oxidation, 64 Atoms visibility in field emission microscope, 184, 188

B Band structure, 4,214, 261 Barium field desorption of, 104, 194 Barrier layer, electronic, 263, see also Boundary layer Beeck review of work by, 208 Boundary layer, electronic, 72, 219, 262, 272 due to 0 2 on ZnO, 273 Brillouin zone, 3

C Cadmium sulfide conductivity of, 229 photoconductivity of, 294 Calcium Nz on, 337 Carbon on W, 113, 198

Carbon monoxide on CuzO, 245 on metals, 338 on W, 200 oxidation on inverse spinels, 243 oxidation on NiO, 63, 68, 215, 248 oxidation on semiconductors, 243 Cesium on W, 137, 141, 323 Chromium oxide conductivity, 252 in Hz-DZexchange, 52 rate of Hz chemisorption on, 235 type of defects in, 242 Cobalt electron configuration, 10 Compensation effect, 75 Conduction electrons, 60 Conductivity, electrical, 261 of CrZOb252 of thin films. 316 of ZnO, 53, 260 Contact potential, 311 Contamination of surface, 138, 197 Copper electron configuration, 10 in ethanol dehydrogenation, 76 Coverage maximum-of semiconductor, 224 Cupric oxide in ethanol dehydrogenation, 76 in NzO decomposition, 77, 242 Cuprous oxide films of, 227 in NzO decomposition, 242 water on, 225 Curie point, 341 of antiferrornagnetism in NiO, 66

D d-bands in metals, 7, 349 d-character, 11 359

360

SUBJECT INDEX

Deactivation of catalyst, 62 Dehydration, mechanism of, 251 Dehydrogenation of ethanol, 76 Diffusion ionic, in ZnO, 53 of O2 in NiO, 65 of Si in Ni, 115 Dipole layers, due to adsorption, 101 see also Boundary layer

E Electron affinity, 305 Electron concentration, 5, 264 Electron exchange reactions in solution, 84 Electron holes, 214 Electron volt, 137 Electrons free, in semiconductors, 218 Elovich equation, 274 Energy scheme for adsorption, 217 of Ce on W, 156 of Nf on W, 168 Equilibrium adsorption effect of barrier layer, 264 Etching technique in field microscopy, 131 Ethanol dehydrogenation, 76 Ethylene chloride decomposition of, 84 Evaporation of W, 140

F Fermi level, 94, 262 effect of support on, 250 in NiO catalyst, 71 Field corrosion, 110 Field emission of electrons, 93, 178, 311 of ions, 103 * Fluorescence of ZnO, 298 Formic acid decomposition, 77, 81 Fowler Nordheim relation, 96, 100, 182

Frankenburg review of work by, 202

G Germanium conductivity of, in gases, 229 water on, 260

H Hall effect, 216 in ZnO, 276 of 0 1 containing tungsten surface, 198 Heat of adsorption, 140, 264 of Ce on W, 141 Helium liquid, technique for high vacua, 116 Hydrocarbons on metals, 344 Hydrogen atoms on W, 105, 128, 337 on metals, 12, 304, 332 on ZnO, 50, 290 para-ortho, conversion, 82 rate of chemisorption, 234 Hydrogen deuterium exchange on CrzOa, 52 on ZnO, 51, 56, 253 Hydrogen ion emission, 145, 193 Hydrogen peroxide decomposition, 243

I Impurities, valence in NiO, 67 in ZnO, 55 Ion emission, 103, 145, 193 Ionization energy of Ce on W, 157 on crystal surface, 48 Iron, electron configuration, 10 HzO and NHs on, 343

L Langmuir review of work by, 200 Langmuir isotherm, 174, 200 Luminescence, 230, 236, 260

361

SUBJECT INDEX

M Magnesite, 81 Magnesium oxidc, 77 Magnetocatalytic, effect, 340 Mass spectrometer use in field emission studies, 125 Metal catalysts compensation effect, 78 Metal sulfides, 250 Methyl alcohol on W, 126 Migration of absorbed molecules, 116, 323 Mobility study of, by field emission, 116 of adsorbed molecules, 323 of ions in solids, 216 of Ce on W, 137 of 0 2 0 n W, 118 Molecular images, 112 Molecules, visibility in field emission microscope, 183 Momentum space, 3 Monolayer, 144

N Nickel electrical conductivity of-films, 319, 328, 332 electron configuration, 10 H 2 0 on, 343 NHI on, 343 Nickel oxide decomposition of NzO on, 215, 239 electronic properties, 66 equilibrium with 0 2 , 217 O2 on, 60 oxidation of, 60 oxidation of CO on, 63, 68, 215, 248 rate of Oz adsorption, 233 Nitrogen on metals, 337 on W, 137, 159, 338 Nitrous oxide decomposition, 215, 236, 339 on Cu, 77, 242 on CuO, 77, 242

on Cuz0, 242 on metals, 338 on Pt, 339 on Pt-alumina, 81 on ZnO, 215 Non-stoichiometry in NiO, 66 in ZnO, 53 N-type conductivity, 214, 261

0 Oxidation at room temperature, 123 of metals, 327 Oxidation catalysis on inverse spinels, 243 on NiO, 60, 63, 68, 248 on semiconductors, 243 Oxygen analysis on solids, 61 effect on ZnO conductivity, 268 equilibrium with metal oxides, 216 desorption from W, 182 mobile, 64 on metals, 327, 331 on NiO, 60, 63 on semiconductors, 215 on W, 118, 128, 137, 175, 197 on ZnO, 271

P Photoconductivity, 260, 294 7-electrons, 344 Platinum alkali metals on, 326 benzene on, 344 copper phthalocyanine on, 346 Hz on, 331 HzO on, 343 N z on, 338 N 2 0 on, 339 0 2 on, 331 triphenylmethane on, 346 Pretreatment compensation effect in, 80, 82 of ZnO, 59 P-type conductivity, 214, 261

SUBJECT INDEX

R Rare gases on metals, 327 Rates limited by electronic barriers, 272 of adsorption and desorption, 264 Resolution of field emission microscope, 101 Resonating valence bond, 8 Rideal review of work by, 205 Roberts review of work by, 205 S

Secondary electron emission, 315 Semiconducting catalysts, 213 Silica, luminescence of, 260 Silicon diffusion of Si in Ni, 115 on Ni, 113 Sintering process, 254 Sticking probability, 137, 173 of N ~ o nW, 165 of 0 2 on W, 181 Surface energy, 190, 320 Surface phases, 113 Surface states, 219, 262 see also Tamm states Susceptibility magnetic, in mctals, 7

T

Tungsten adsorption of W on, 192 alkali metals on, 326 carbon on, 113, 198 carbon and oxygen on, 199 Ce on, 137, 141, 323 CO on, 200 contact potential of, 337 desorption of Ba from, 194 evaporation of, 140 H P on, 105, 128, 336 HzO on, 129 methyl alcohol on, 126 N Pon, 137, 159, 338 O 2 on, 118, 128, 137, 175, 197, 330 Tunnel effect, 83

v Vacuum technique, in field emission work, 132 Valence induction, 250

W Water on metals, 343 on Ni, 343 on Pt, 343 on W, 129, 198 Work function, 94, 304, 307 determination of, 132, 151 effect of crystal structure on, 320 of semiconductors, 230 of W with Cc, 152 of W with On, 182

z Tamm states, 219, 263 on ZnO, 270 Temperature dependencc of equilibrium, 75, 272 of reaction rate, 75 of slow desorption, 51, 272 Thermoelectric effect, 216 Transition metal alloys, 8 electron structure, 6

Zinc oxide, 47, 259 basic properties, 266 boundary effects, 225, 260 dehydration over, 251 equilibrium with 0 2 , 217, 285 fluorescence in, 298 Hz adsorption on, 234, 260, 290 H2-D2exchange, 51, 253 photoconductivity, 294

ERRATUM

Advances in Catalysis 6, 176 (1954); Equation 30 should read:

z-x-(2+*) S

db/dn dl

-

q/dicoth q/di - 1. qcothq-1 ’ s

=

kl/kz; p

=

R

This Page Intentionally Left Blank