FISSION
This page intentionally left blank
FISSION Corsendonk Priory, Belgium
1 8 - 2 1 September 2007
Editors Cyriel Wagemans University of Gent, Belgium
Jan Wagemans & Pierre D'hondt SCK'CEN, Mo/, Be/g/um
world Scientific ^^:«^^^^^?i^i*S^^^^^?S*^^fei^^^^^^'^^^^fe^^^^^^
Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA ofice: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
U K ofice: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-PublicationData A catalogue record for this book is available from the British Library.
SEMINAR ON FISSION VI Copyright 0 2008 by World Scientific Publishing Co. Re. Ltd.
All rights reserved. This book, or parts thereoj may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.
ISBN- 13 978-98 1-279-105-4 ISBN- 10 981-279- 105- 1
Printed in Singapore by World Scientific Printers
ORGANISING COMMITTEE P. D’hondt SCKCEN Mol Belgium M. Huyse University of Leuven Belgium C. Wagemans University of Gent Belgium
SCIENTIFIC ADVISORY COMMITTEE N. CPrjan University of Bordeaux France
J. Cugnon UniversitC de Li6ge Belgium
F. Gonnenwein University of Tubingen Germany D. De Frenne University of Gent Belgium F.-J. Hambsch IRMM Gee1 Belgium
0.Serot CEA Cadarache France vi
PREFACE This Seminar is the sixth of a series started in 1986. The five previous meetings took place in the historical castle of Pont d’Oye (Habay-la-Neuve), which we abandoned for several reasons. Our new location is the ancient priory of Corsendonk, which was a place of contemplation and wisdom during many centuries. The dimensions and the isolated situation of this priory constitute a nice environment for a small-size meeting like the present one and enable a good and relaxed working atmosphere. During this meeting, recent achievements in experimental and theoretical fission physics were discussed, giving special attention to low-energy fission and its traditional topics such as fission fragment characteristics, ternary fission, fission neutrons, fission barriers and fission cross sections. Also more specialised topics such as shape isomers, P-delayed fission, fission in spallation reactions and the importance of fission for nuclear astrophysics were discussed. Furthermore, review papers on the angular momentum in fission and on fission properties of heavy and super-heavy nuclei were presented. Finally, due attention was given to new facilities and detectors. This Seminar is strongly supported by three organisations: the Belgian Nuclear Research Centre in Mol (SCKCEN), the Fund for Scientific Research Flanders (FWO) and the University of Gent (UG). The organising committee is very grateful to these sponsors. Also the valuable help of the International Advisory Committee and of the Chairmen of the Sessions is gratefully acknowledged.
Cyriel Wagemans Conference chair vii
This page intentionally left blank
CONTENTS Organisations
V
Organising and Scientific Advisory Committees
vi
Preface
vii
Topical Reviews
1
Angular Momentum in Fission F. Gonnenwein, V. Bunakov, 0.Dowaux, A. Gagarski, I. Guseva, F. Hanappe, S. Kadmensky, J. Von Kalben, S. Khlebnikov, V. Kinnard, Yu. Kopatch, M Mutterer, V. Nesvizhevsky, G. Petrov, E. Prokhorova, V. Rubchenyu, M. Sillanpaa, G. Simpson, V. Sokolov, T. Soldner, L. Stuttgi, G. Tiourine, W. Trzaska, I. Tsekhanovich, C. Wagemans, H.-J. Wollersheim, T. Zavarukhina, 0.Zimmer
3
The Processes of Fusion-Fission and Quasi-Fission of Heavy and Super-Heavy Nuclei M G. Itkis, A. A. Boguchev, E. V. Chernysheva, I. M Itkis. G. N.Knyazheva, N. A . Kondratiev, E. M. Kozulin, L. Krupa. F. Hanappe, 0.Dowaux, N. Rowley, L. Stuttgi, A . M Stefanini, B. R. Behera, L. Corradi, E. Fioretto, A. Gadea, A. Latina, S. Sziher, M. Trotta, S. Beghini, G. Montagnoli, F. Scarlassara, V. A. Rubchenya, W. H. Trzaska
25
Fission Cross Sections and Fragment Properties
45
Minor-Actinides Fission Cross Sections and Fission Fragment Mass Yields via the Surrogate Reaction Technique B. Jurudo, G. Kessedjian, M. Ai'che, G. Barreau, A. Bidaud, S. Czajkowski, D. Dassii, B. Haas, L. Mathieu, B. Osmanov, L. Audouin, N. Capellan, L. Tussan-Got, J. N. Wilson, E. Berthoumieux, F. Gunsing, Ch. Theisen, E. Bauge, 0. Serot, I. Ahmad, J. P. Greene, R. V. F. Janssens
47
ix
x
Proton-Induced Fission on Actinide Nuclei at Medium Energy
55
S. Isaev, R. Prieels, Th. Keutgen, J. Van Mol, K El Masri, P. Demetriou
Fission Cross Sections of Minor Actinides and Application in Transmutation Studies A . Letourneau, 0. Bringer, S. Chabod, E. Dupont, G. Fioni, F. Marie, S. Panebianco, Ch. Veyssiere, L. Oriol, F. Chartier, P. Mutti, I. Almahamid
63
Systematics on Even-Odd Effects in Fission Fragments Yields: Comparison Between Symmetric and Asymmetric Splits F. Rejmund, M Caamano
71
Measurement of Kinetic Energy Distributions, Mass and Isotopic Yields in the Heavy Fission Products Region at Lohengrin A . Bail, 0. Serot, 0.Litaize, H. R. Faust, U. Koster, T. Materna. A . Letouvneau, E. Dupont
79
Ternary Fission
87
On the Ternary a Spectrum in 252Cf(sf) M Mutterer, Yu. N. Kopatch, S. R. Yamaledtinov, V. G. Lyapin, J. von Kalben, S. V. Khlebnikov, M Sillanpaa, G. P. Tyurin, W: H. Trzaska
89
Energy Degrader Technique for Light-Charged Particle Spectroscopy at LOHENGRIN A . Oberstedt, S. Oberstedt, D. Rochman
99
Ternary Fission of Cf Isotopes S. Vermote, C. Wagemans, 0.Serot, T. Soldner, P. Geltenbort, I. Almahamid, W. Lukens, J. Floyd
107
Systematics of the Triton and Alpha Particle Emission in Ternary Fission
117
C. Wagemans, S. Vermote, 0. Serot
Neutron Emission in Fission
123
Scission Neutron Emission in Fission F.-J. Hambsch, N. Kornilov, I. Fabiy, S. Oberstedt, A. Vorobyev
125
At and Beyond the Scission Point: What can we Learn from Scission and Prompt Neutrons? P. Talou
139
xi
Fission Prompt Neutron and Gamma Multiplicity by Statistical Decay of Fragments S. Perez-Martin, S. Hilaire, E. Bauge
147
Fission Theory
159
Structure and Fission Properties of Actinides with the Gogny Force H. Goutte, J.-F. Berger, J.-P. Delaroche. M. Girod, A. Dobrowolski, J. Libert
161
Fission Fragment Properties from a Microscopic Approach N. Dubrq, H. Goutte, J.-P. Delaroche
171
Smoker and Non-Smoker Neutron-Induced Fission Rates I. Korneev, I. V. Panov, T. Rauscher, F.-K. Thielemann
177
Facilities and Detectors
187
A Novel 2v2E Spectrometer in Manchester: New Development in Identification of Fission Fragments I. Tsekhanovich, J. A. Dare, A. G. Smith, B. Varley, D. Cullen, N.Lumley, T. Materna, U. Koster, G. S. Simpson
189
Development of PSD and ToF + PSD Techniques for Fission Experiments M Sillanpaa, M. Mutterer, W. H. Trzaska, G. Tyurin, Yu. N. Kopatch, S. Smirnov. S. Khlebnikov, J. von Kalben
197
MYRRHA, a New Fast Spectrum Facility H. Ail Abderrahim, P. D’hondt, D. De Bruyn
207
The BR1 Reactor: A Versatile Tool for Fission Experiments J. Wagemans
223
“Special” Fission Processes
23 1
Shape Isomers - A Key to Fission Barriers S. Oberstedt, F.-J. Hambsch, N. Kornilov, G. Lovestam, A. Oberstedt, M Gawlys
233
Fission in Spallation Reactions J. Cugnon, Th. Aoust, A. Boudard
24 1
xii
Conference Photo
259
List of Participants
261
Author Index
265
Topical Reviews
This page intentionally left blank
ANGULAR MOMENTUM IN FISSION F. GONNENWEIN’), V. BUNAKOV2),0. DORVAUX’), A. GAGARSK14),I. GUSEVA4),F. HANAPPE’), S. KADMENSKY6), J. VON KALBEN’), S. KHLEBNIKOV’), V. KINNARD’), W. KOPATCH9), M. MUTTERER’), V. NESVIZHEVSKY’”, G. PETROV4), E. PROKHOROVA9), V. RUBCHENYA”’ , h SILLANPU”), G. SIMPSONI” , V. SOKOLOV4),T. SOLDNER”) , L. STUTTGE I. TSEKHANOVICH’~),c. WAGE MANS]^), H.. G. TIOURINE’), w. TRZASKA~]), WOLLERSHEIM’’), T. ZAVARUKHINA4’, 0. ZIMMER”) 1) Physikalisches Institut, Universitat Tiibingen, 72076 Tiibingen, Germany 2) University of St. Petersburg, St. Petersburg, Russia 3) CNRS Strasbourg, Strasbourg, France 4) St. Petersburg Nuclear Physics Institute, Gatchina, Russia 5) Universiti Libre de Bruxelles, Bruxelles, Belgium 6)University of Voronezh, Voronezh, Russia 7) Technische Universitat Darmstadt, Darmstadt, Germany 8)Khlopin Radium Institute, St. Petersburg, Russia 9)Frank Laboratov, JINR Dubna, Russia 10) Institut Laue-Langevin. Grenoble. France I I ) University OfJyvaskyla, Jyvaskyla, Finland 12)University of Gent, Gent, Belgium 13)GSI Darmstadt, Darmstadt, Germany Three novel experiments in spontaneous and thermal neutron induced fission all with a bearing on angular momentum in fission are reviewed. In the first experiment it was observed that, in the reaction 215U(n,f)with incident polarized cold neutrons, the nucleus undergoing scission is rotating. This was inferred from the shift in angular distributions of ternary particles being dependent on the orientation of neutron spin. In the second study the properties of the angular momentum of spherical fission fragments was investigated. Current theories trace the spin of fragments to their deformations allowing for collective rotational vibrations at scission. However, in particular the spherical I3’Te isotope exhibits a large spin at variance with theory. Exploiting the specific properties of cold deformed fission it could be proven that, for ”’Te, single particle excitations instead of collective modes are responsible for the large spin observed. In a third project a pilot study was exploring the possibility to search for an evaporation of neutrons from fragments being anisotropic in their own cm-system. Due to fragment spin this anisotropy is claimed since decades to exist. It was so far never observed. A scheme has been devised and tested were triple coincidences between a fragment and two neutrons are evaluated in a way to bring the cm-anisotropy into the foreground while getting rid of the kinematical anisotropy in the lab-system due to evaporation from moving fragments. The test was run for spontaneous fission o f Z S 2 ~ f .
3
4
1. Introduction
Discussing angular momentum in fission one should distinguish between the spin of the fissioning nucleus and the spin of fission fragments. In the following three recent experiments in low energy fission, viz. spontaneous and thermal neutron induced fission, are considered addressing either compound or fragment spin. In the first experiment to be presented, ternary fission of 233Uand 235U induced by polarized neutrons was investigated. For a cold neutron beam being longitudinally polarized the two fission fragments and a ternary particle were detected in a plane roughly perpendicular to neutron spin. Two different effects in the emission of ternary particles were discovered. In the “ T N effect”, which is best pronounced in the reaction 233U(n,f), it is found that the emission probability is different relative to the oriented plane formed by neutron spin and light fragment momentum. In the “ROT effect”, which is dominant in the reaction 235U(n,f),the angular distributions of ternary particles are slightly shifted back and forth depending on the orientation of neutron spin. Both effects are linked to the angular momentum of the fissioning compound down to scission which, on one hand can lead to collective rotations of the fissioning compound and, on the other hand, give Coriolis forces in the rotating system a chance to become effective. In the past many studies have been devoted to the angular momentum of fission fragments. A most surprising fact here is that fragment spin can be much larger than the compound spin. Theory has explained these large spins by collective vibrating modes at scission carrying angular momentum. These modes require some deformation of the fission fragments at scission. In experiment, however, large (if not the largest) angular momenta are observed for fragments remaining spherical at scission like those in the vicinity of the doubly magic I3’Sn. For these fragments another mechanism must be at work. Evidently one may invoke single particle excitations. For the fission fragment ‘32Tefrom the 239Pu(n,f)reaction it could indeed be demonstrated that specific single particle levels with unusually large spins are excited either thermally or, more probably, by excitations in the non-adiabatic scission process. The experimental method relies on a comparison of cold compact and cold deformed fission at the Lohengrin spectrometer of the ILL. As to neutron emission from fission fragments it was conjectured since the early sixties that the large angular momenta of fragments should entail some anisotropy of neutron emission in the cm-system of the moving fragments. So far only indirect methods have been available to test this assumption. In an
5
experiment, dubbed the “CORA” experiment, a DEMON neutron detector array was combined with a twin ionization chamber for fission fragments, 252Cf(sf)serving as the fission source. The idea is to record triple events, one of the fission fragments and two neutrons, and to search for an angular correlation between the three momentum vectors. In the extreme case of a hypothetical 100% anisotropy in the cm-system the three vectors should lie on a single plane. In the evaluation of data this hypothesis is checked event-by-event. Of course, in reality the anisotropy is not loo%, but any anisotropy in the cm-system should become visible in the lab-system as a non-zero angular correlation between the above three vectors. So far a pilot experiment has been run which has proven necessary to optimize the setup of neutron detectors for an upcoming experiment. 2. The ROT effect
With the advent of intense polarized neutron beams, as e.g. at the high flux reactor of the Institut Laue-Langevin .in GrenobleRrance, studies of symmetry laws even for the rare ternary fission reaction have become feasible. Besides the study of parity violation in ternary fission, an intriguing proposition was that possibly also the violation of time reversal invariance could be scrutinized [ 11. In analogy to free neutron decay it was argued that the correlation B = PLCP ‘ [ax P L F l (1) may carry the signature of a T-odd reaction. In Eq. (1) pLcp,a and PLF are the momentum of the Light charged Particle (LCP), the spin of the incoming neutron inducing fission and the momentum of the Light Fission Fragment (LF), respectively. All vectors are normalized to unity. The triple correlation B becomes maximal in case the three vectors are at right angles relative to each other. This has suggested the experimental setup shown in Fig. 1. A thin target of highly enriched (299%) 23sUplaced at the center of a reaction chamber is irradiated by a beam of cold polarized neutrons. The beam is running - say - in the +z-direction and the polarization (295%) is longitudinal with either oz = +1/2 or oz= -1/2 in units of h. Fission fragments (Light and Heavy Fragment LF and HF) are intercepted by two multi-wire-chambers to the Left and Right, and the Light Charged Particles (LCP) are detected in two arrays of 4 Si-diodes each, mounted on top and bottom. To appreciate the meaning of the correlation it may be helpful to remark that in case the correlation B is non-zero, the yield of LCPs is different up and down relative to the oriented plane [a x pLF] defined by neutron spin and light fragment momentum. In experiment the spin of the neutrons in the beam is flipped every second and hence the asymmetry at issue
6
reverses signs. More quantitatively the asymmetry A is evaluated from the count rates N, and N-for the two spin orientations as A = (N+- N -) / (N + + N-), (2) Without going into further details here, it is seen that the asymmetry A is a reasonable measure for the strength of the correlation B.
Y/
DOWN
Figure 1. Experimental setup of detectors in reaction chamber with the 235Utarget at the center.
In a former experiment with the neighboring isotope 233Uas the target a non-zero correlation was indeed observed. For a given orientation of neutron spin and direction of flight of the light fragment, as anticipated, the asymmetries for all diodes of a plane had the same sign, but upper and lower plane had opposite signs [2, 31. However, it could be shown that this result does not put time reversal invariance into question, instead the effect is brought about by the Coriolis force acting in the polarized and hence rotating nucleus [4, 51. As a reminder of the triple correlation the effect was called the TRI-effect. In the present experiment with the isotope 235Ua similar result was expected. Surprisingly a completely different pattern of signs for the asymmetry A was observed. The distribution of signs is indicated in Fig. 1 for the individual LCP diodes. In addition, in contrast to the former experience, the signs did not change upon inversion of the light fragment momentum [ 6 ] . The idea popped up that this new phenomenon could be the signature for a rotation of the fissioning nucleus. To understand why, it has to be recalled that for a given capture state with compound spin J its polarization will change sign when the spin of the incoming neutron is flipped. In classical terms the vector J is decomposed into its components J = R + K perpendicular and along the symmetry axis of the fissioning nucleus, respectively. Hence, when the neutron spin is flipped, the collective rotation with angular momentum R is reverted. Consequently the ternary LCPs, mostly a-particles, are emitted from a rotating system whose sense of rotation is switched every second. Since the LCPs are
7
accelerated to their final velocity by the Coulomb forces provided by the two main fission fragments, the angular distributions of the LCPs should - at least to some degree - reflect the rotation of the force field. The idea is visualized in Fig. 2 for the specific case when the light fission fragment is moving to the left and the LCP is emitted upwards. wz>O at t = O
LCP
b
I
W,cO
at t = O
I Trajenoly of LCP
f
Lcp
Figure 2. Shift of LCP angular distribution for clockwise (left) and anti-clockwise (right) rotation.
It is seen that when the fission fragments have finally come in front of their detectors, the LCPs will have been slightly deflected compared to a situation without rotation. Synchronized with neutron spin flip the angular distributions of the LCPs will hence be shifted back and forth. For positive (negative) polarization of the compound nucleus the rotation is clockwise (anti-clockwise). The angular distributions exhibit a maximum at about 8 = 82" between the emission directions of light fragments and the LCPs, the half-width being about 20". On the slopes of these distributions the asymmetry A is expect'ed to be nonzero with the sign of A changing in going from small to larger angles of emission. This is made evident in Fig.3. Arrows in the Fig. label the centers of the LCP detector arrays. If now the compound nucleus capture state has e.g. J = 1 + %, its polarization will be positive (negative) for neutron spin orientation 0, = +1/2 (0 = -1/2) and correspondingly the sense of rotation will be clockwise (anti-clockwise). The asymmetry evaluated following eq. 2 will then be positive for angles 8 between the light fragment and the LCP smaller than 82" and negative for angles 8 2 82". With the labels for each Si-diode in Fig. 1 indicating the sign of the asymmetry A, it is seen that the pattern is precisely as anticipated from the rotational model.
*
8
Angle I d q .
Figure 3. Angular distribution of LCPs. For the configuration LEFT-UP of light fragment and LCP in Fig.2 the distribution is shifted to smaller (larger) angles for clockwise (anti-clockwise) rotation.
The above phenomenological reasoning has been corroborated by trajectory calculations of ternary particles emitted from the neck of a rotating mother nucleus. The usual expression for the collective energy of rotation perpendicular to the symmetry axis of the fissioning nucleus is E,,, = (h/20)[J(J + 1) - K2] (3) with 0 the moment inertia. In the spirit of a classical model we then have for the absolute size of the angular momentum R of the collective rotation RZ= w2 0' = [J(J +1) - K2). (4) With the angular momentum J and its projection K being constant, also the size of R will be constant and it is seen that the angular velocity w is decreasing very fast as time goes on since the moment of inertia 0 explodes when the two fission fragments are flying apart. The rotational motion is therefore not uniform but strongly non-uniform. The typical time constant may be guessed to be near l o 2 ' s. From Eq. 4 it is inferred that the rotation should be best pronounced for small K quantum numbers. Results from the actual calculations of LCP trajectories in the Coulomb field of a rotating system are on view in Fig. 4. With the assumption K = 0 and with the reasonable guess for the angular momentum R = l h the diagram to the left demonstrates that the angular velocity w in units of deg/10% is indeed fading away very fast. After a time lapse of 3*10-2'sthe initial angular frequency has decreased by some 90%. The rotation is hence strongly non-uniform. To the right are shown the angles of rotation for the light fragment (full points) and a ternary a-particle (open points) both as a function of time since scission. Very quickly the rotational angles are noticed to approach asymptotic values being surprisingly small and not exceeding some tenths of a degree. The small angles have to be traced to the virtual stop of the rotation after a couple of 10'21s.In the present context the main message of Fig. 4 is, however, the observation that the
angles through which fragments and ternary particles are rotating are not identical. The ternary particles are lagging behind the light fragment by an angle A which asymptotically is about 0.06°. It is precisely this non-zero lag angle A ^ 0 which enables assessing the rotation. The reason is that for opposite senses of rotation the angular distributions of LCPs are once shifted by +A and once by A (see Fig. 2). The relative shift is hence 2A and it is this shift which is visible in the sketch of Fig. 3. Evidently in Fig. 3 the shift is grossly exaggerated compared to reality. O.OW
9*14.
<MXW (MXM
0.000
Figure 4. To the left: angular velocity of a nucleus undergoing scission at time zero assuming that the collective angular momentum perpendicular to the fission axis is R = Ih. To the right: angle of rotation as a function of time both for the light fragment (full points) and for the ternary particle (open points), the latter being assumed to be an a-particle.
Results from a more dedicated study with more detectors covering more angles than shown in Fig. 1 fully confirmed the results of the discovery experiment. Asymmetries as a function of angle 0 are given in Figure 5.
60 70 80 90 100 110 120 Angle between pLf and pTp , deg
Figure 5. Asymmetry (data points, left scale) and ternary particle yield (right scale) versus the angle between the light fission fragment and the ternary particle for the reaction 235U(n,f). Preliminary data.
10
The data points represent the asymmetry as a function of angle 0 (scale to the left). The dotted curve is the angular distribution of the LCPs (scale to the right). While in Fig. 1 only the signs of the asymmetry were given, their size may be read off quantitatively from Fig. 5 . The asymmetry covers the range from A 5 + 0.006 to A =: - 0.006 with A being zero for the angle 0 =: 82" at maximum yield. Before transforming the asymmetries into a shift angle A (see Figs. 3 and 4) for comparison with theory it has to be taken into account that behind the dominant ROT effect a smaller TRI effect may be hidden. For the TRI effect the asymmetry should stay constant for all angles 0 of LCP emission. From a detailed analysis of the data it is concluded that a TRI asymmetry with a size of 0.98(15).10-3is present in the data. After correction for this TRI effect, the shift angle for the ROT effect in the reaction 23SU(n,f)is evaluated to be 2A = 0.20(1)". This figure is in surprisingly good agreement with the value of 2A = 0.12" found in trajectory calculations. More details may be found in Ref. [6]. As already outlined in the above, in the classical model of rotation the sign pattern of the asymmetry tells that the polarization of the compound nucleus is also positive. It is hence deduced that for the reaction 23sU(n,f) induced by thermal neutrons the fission cross section for the capture state J' = I + % should be larger than for the state J- = I - %. This is in full accord with experiment yielding for the ratio of fission cross sections crfi(.J+) 1 ofi(J-) = 553 b/ 323 b = 1.71. Besides being an interesting phenomenon, the ROT effect may also find applications in the search for scission neutrons and scission gammas. The fact is exploited that the rotation comes to a virtual stop in a few lO-"s. Therefore only particles emitted in this time span are liable to experience the rotation and exhibit a ROT effect. This enables to distinguish the conjectured emission of particles right at scission, neutrons andfor gammas, from those emitted much later. The bulk of neutrons in fission is indeed known to be evaporated at much later times and gammas appear even later only once neutrons have exhausted most of the excitation energy. Details on this particular aspect are given in Ref. [71. The classical interpretation of the ROT effect having been outlined is appealing since it does not only give a qualitative description of the phenomenon, but even allows predicting within a factor of 2 or so the size of the ROT effect and - as not discussed here - its dependence on the kinetic energy of the ternary particles. Nevertheless, ultimately the phenomenon has to be discussed in terms of quantum mechanics. A quantum-mechanical theory of the TRI and the ROT effect has indeed been worked out [8]. The gross features are
11
correctly reproduced though no quantitative results to be confronted with experiment are given. Probably more important is, however, that the na'ive sign correlations from classical physics are overthrown. It is stressed that, in the present (n,f) reaction at off-resonance energies, the T-odd effects at issue are due to the interference of s-resonances with amplitudes and phases being randomly distributed. In consequence, the correlations between the TRI and the ROT effect as to sizes and signs are random. Comparing results form the present 23SU(n,f)reaction and a re-evaluation of a former experiment with 233Uas the target [2,3], the claim for a random TRIROT correlation appears to be realized.
3. Angular momentum of fission fragments In the preceding section properties of the angular momentum of the compound nucleus undergoing fission were explored. We now turn to a discussion of the angular momenta of fission fragments once they have passed the scission point. Following scission and the relaxation of fragment deformation the primary fission fragments are highly excited. The excitation is released by neutron evaporation and gamma emission. At first neutrons are ejected in times shorter than 10-'5s to lO-I4s. The remaining excitation energy of about 5 MeV is exhausted by gammas. The first so-called statistical gammas are low-energetic and are usually not resolved in experiment. With the excitation energy approaching the ground state discrete gamma lines become observable. Virtually all what is known about the angular momentum of fission fragments comes from a study of these discrete gammas. Since isomeric transitions are readily identified in y-decay chains, a method often employed relies on the measurement of yields for two spin isomers, one at high and one at low spin. The spin distribution is then evaluated in a statistical model. An ansatz for the primary spin distribution is made where the rms spin of fragments I,, enters as a free parameter. Simulating neutron evaporation and emission of statistical gammas, the distribution at the entry points to the discrete gamma lines are found. This allows calculating the feeding of the two spin isomers. From a comparison of calculated and experimental values the rms spin I, of the fragments is assessed. Details on the evaluation technique may be found in Ref. [9]. Surprisingly large fragment spins are observed in spontaneous and low energy fission, sometimes exceeding 12 h. This sparks the question how these large angular momenta are generated. A model predicting angular momenta for the full range of fragment masses is the bending model [ 10, 1I]. It is argued that Coulomb and nuclear forces combine to form a potential pocket close to the
12
scission point. The two deformed fragments are aligned in this potential well along the fission axis. Angular vibrations carrying angular momenta are then excited as zero point oscillations or, in case of finite nuclear temperatures, thermally. More recently a microscopic model has been put forward where the constrained alignment of deformed fragments is pumping angular momentum into the motion of the nucleons [ 121. More generally also an orbital angular momentum between the two fragments has to be taken into account simply because the uncertainty relation forbids the alignment of the two centers of fragment mass on a fission axis fixed in space [13]. It should be noted that all above models predict fragment spin to be oriented perpendicular to the fission axis. This is indeed observed in experiment. Results from the bending model are shown together with experimental data in Fig. 6. T=3MeV
.
01: 80
:
:
100
:
'
120
:
:
140
:
: 160
Fragment Mass
Figure 6 . Experimental data points: average spin of fragments from "'U(n,f) [14], full lines: theoretical predictions from the bending model for various nuclear temperatures.
The experimental data are for thermal neutron induced fission 235U(n,f)[14]. It is seen that except for a fragment mass range near A = 135 the data are in very good agreement with theory for spins being generated by zero-point oscillations (these are labeled "adiabatic" in the figure). The model, however, fails to predict the large spins near A = 135. This becomes even more evident in Fig. 7 for photofission of 235U [15]. The gammas were produced by 12-30 MeV Bremsstrahlung. The staggering of data between huge fragment spins well outside the range of any bending-like model on one hand, and rather small spin values on the other hand is noteworthy. More in detail it was found that fragment spins are independent of y-energies in the above energy range. By contrast, there is a pronounced odd-even effect with odd-Z isotopes exhibiting large spins and even-Z isotopes small spins. This behavior is opposite to the dependence of fragment charge yields with even-Z elements being produced
13
more abundantly than odd-Z elements. It further appeared that fragment spins for several low energy and even spontaneous fission reactions are virtually identical. More recent comprehensive studies have confirmed the early findings depicted in Figs. 6 and 7 and brought to evidence finer details [16]. While evidently deformations of fragments at scission play an eminent role for pumping angular momentum, the large spins of spherical fragments clearly point to a different process of spin formation. In particular the staggering of spins in this fragment mass region has led to the conjecture that single-particle excitations are responsible for the observations at issue. However, this conjecture has so far never been proven directly. It is the purpose of the present work to prove it in experiment.
84
100
116
132
148
Fragment Mass
Figure 7. Experimental data points: average spins of fragments near A lines: theoretical predictions from bending models.
=
132 from 2’5U(y,Q[15],full
The basic idea to distinguish between the two mechanisms of spin formation - either deformation-driven or by single-particle excitation - is to exploit the properties of cold fission. In cold fission the system is pushed to limits of scission point configurations. At the point of rupture the two nascent fission fragments are facing each other, and in a static model, the scission point model, the total available energy is shared between the Coulomb energy of repulsion between fragments, the deformation energy and the intrinsic excitation energy of the fragments. In the most compact configuration nearly all energy goes into the Coulomb energy, with deformation and excitation energy virtually vanishing. In the opposite case of the most deformed configuration, a maximum of energy has gone into deformation while due to the increased distance between charge centers the Coulomb energy has become smaller and -for the present purpose most noteworthy - the excitation has again been virtually
14
consumed. The Coulomb energy of repulsion contributes the lions share to the kinetic energy of fragments as measured in experiment. In summary, cold compact fission is therefore characterized by no fragment deformation nor excitation energy and a maximum of fragment kinetic energy. By contrast, in cold deformed fission the fragments carry large deformation but no intrinsic excitation energy, and the kinetic energies are at a minimum. In cold compact fission spins are predicted to be close to zero whatever mechanism is at work. However, in cold deformed fission spins are expected to be small only in case the missing excitation energy does not allow feeding single particle states while they should be large for a deformation based mechanism. A measurement of fragment spin in cold deformed fission hence allows distinguishing between the two mechanisms being discussed. As a side remark it is worthwhile to demonstrate by experiment that the two types of cold fission indeed are really to be observed in fission. It is brought to evidence by studying the odd-even effect of fragment charges as a function of the kinetic energy of the fragments. Starting from a completely paired situation of an even-Z compound nucleus as e.g. uranium, the pairing in the fragments is broken the more efficiently the more they acquire excitation energy. Conventionally the charge odd-even effect 6 is measured as the difference in yields of even charge fragments minus odd charge yields normalized to the sum of yields. The dependence of the charge odd-even effect 6 on the lunetic energy of the light fragment is shown in Fig. 8 for the reaction 232U(n,f)induced by thermal neutrons [ 181.
Figure 8. Fragment charge odd-even effect given as 6 = - Yadd) / the kinetic energy of the light fragment. Reaction: 2'2U(nth,f).
+ Y d d ) as a function of
15
It is seen that both, in cold compact and in deformed fission, i.e. at the highest and the smallest fragment kinetic energies the odd-even effect is large indicating that the intrinsic excitation energy is at a minimum. As outlined in the above, the task to solve the question at hand is to measure angular momenta of fragments as a function of lunetic energy with special attention given to spin formation at the limits of the kinetic energy distributions. A facility well suited to this purpose is the Lohengrin spectrometer installed at the High Flux Reactor of the Institut Laue Langevin in GrenobleFrance. The technique to measure yields of isomers with high and low spins for given fragments and to deduce their spins was introduced already many years ago at this instrument [19]. However, the measurements were not pushed to the extremes of the energy distributions. This was precisely the aim of the experiment to be reported here. Referring to Fig. 7, the search for appropriate nuclides in the magic or nearmagic mass range A = 132 of fragments is nowadays greatly facilitated by recent comprehensive studies of ps isomers [19]. The experiments were performed at the Lohengrin spectrometer. For example, for several (e-e) Teisotopes ps isomers are reported, all with spin 10'. In a shell model approach these states are interpreted as the break-up of pairing of equivalent (n,lj) orbits of two neutron holes (hllZh1I2)-l. The level scheme of '32Teg0is shown on the left hand side of Fig. 9. The half-life for the 10' state is TI/*= 3.70(9) ps. This nucleus was selected for the present study because it has a second isomer with lower spin 7- and a half-life of T !h = 28 ps. Hence, it is well adapted to the technique of spin determination by comparing the relative yields of high and low spin isomers. The ps isomer technique has become popular because the isomeric transitions are readily identified even in a large y-background. By good fortune the Lohengrin spectrometer could not be better tailored for the task since the flight time of fragments form the target to the focal plane of the electromagnetic separator is between 1 and 2 ps depending on the fragments velocity. We will here not go into the details of the experimental setup and its evaluation and only sketch a few points. Very briefly, the target made available for our experiment was 239Pu.The target being placed close to the reactor core undergoes fission and fragments of given mass and energy having passed the spectrometer were stopped at the focal plane in a stopping foil. A series of Gedetectors surrounding the foil intercepted the gammas. The first transition 10' -+ 8' has only an energy of less than 23 keV available. It is therefore highly converted and not sensed by the Ge-detectors. All other transitions are readily
16
identified and their yield is determined. The yields of isomeric transitions have to be corrected for decay in flight in the spectrometer with the flight path from target to stopping foil being close to 23 m. The correction depends on fragment energy. Prompt y-decays occur near the target and are not registered by the detectors. Yields were taken for the full energy distribution of fragments.
10' 8'
76' 4'
(8
h97 / I
47.A I
I,
1
L
i
2 '
-+
E (6
7)+ (8 -+ 6)
-+
4
4
2 -+ 0)
0' Ekin I MeV
Figure 9. Left hand side: level scheme of '3ZTe80. Right hand side: ratio of transitions of the high spin isomer lo', i.e. {(8 -+ 7 + (8 -+ 6 ) ) ,to the sum of transition in the cascade { 6 -+ 4 -+ 2 -+ 0). Note that the transition (10 -+ 8) is highly converted and not observed as y-transition. Note further that both the high and the low spin isomer contribute to the cascade transitions.
For the presentation of results it was chosen to compare the sum of yields {(S --+ 7) + (8 --+ 6)} originating from the high spin isomer lof with the sum of yields from the cascade { 6 --+ 4 --+ 2 --+ 0}. The cascade receives contributions
from both, the high spin state 10' and the low spin state 7-. The ratio of the two sums of yields is visualized on the right hand side of Fig. 9. Remarkably, the ratio of yields drops at both the highest and the lowest kinetic energies of fragments, i.e. where count rates are just sufficient taking data. The decrease of the ratio tells that the population of the high spin state falls behind the population of the low spin state at the limits of the kinetic energy distribution or, in other words, in cold compact and cold deformed fission. As argued above, this behavior in compact fission is as expected from the two models for generating fragment spin. More telling is the decrease of the high spin contribution in cold deformed fission. In spite of large deformations the high
17
spin state is not fed. This is, however, as to be anticipated for states well above the ground state being populated by intrinsic excitation. Recalling the single-particle character of the state in question, it is concluded that for the example I3’Teg0chosen the large spin observed at average kinetic energies is not due to a collective zero-point oscillation but instead is generated by singleparticle excitation. This conclusion is reached even without going through a complete analysis yielding the initial spin distribution of fragments. For average kinetic energies a primary rms spin of about 9B has been reported [18]. More details concerning the present experiment are to be found in Ref. [20]. Further experiments are planned where spins of fragment nuclei neighboring the doubly magic I3’Sn will be scrutinized. For these spherical nuclei it will be intriguing to inspect the correlation between spin and level structure. The correlation will be interesting to analyze further, especially because the large spin of the presently studied (e,e) isotope I3’Teg0 does not follow the rule established in refs. [15] and [16] stating that generally odd-Z andor odd-N isotopes carry the larger spins. 4. Anisotropy of neutron emission in the cm system of fragments
It is well established that the bulk of neutrons being emitted in fission is due to evaporation from fully accelerated fragments. The evaporation from moving fragments is invoked in order to explain the anisotropy of neutron emission as observed in the lab system. Thereby it is generally assumed that neutron evaporation from the fragments is isotropic in their own cm system. With this ansatz the angular distributions of neutrons in the lab are reasonably well but not perfectly reproduced. The deficiencies of the model have led to the conjecture that either neutrons emitted right at scission are present or that besides the above kinematical anisotropy there might also be a dynamic anisotropy in the sense that even in the cm system of fragments the evaporation is anisotropic. The reasons behind the latter anisotropy are the large angular momenta of fission fragments. The majority of fragments are deformed at scission and therefore bending-like models apply where angular momentum is traced to rotations of nuclei. But for rotating nuclei one should expect that neutron evaporation is anisotropic with emission being favored perpendicular to fragment spin. Anisotropic neutron emission in the cm systems of fragments is therefore another facet of angular momentum in fission. The idea was put forward very long ago [21]. The experimental verification, however, is not evident because to disentangle in the lab the kinematical and the dynamic anisotropy is not feasible by simply observing the
18
angular distributions of neutrons. This was pointed out more recently while discussing new approaches to the problem at hand [22]. A novel technique had to be devised. The new concept to disentangle in the lab the contributions of kinematical and dynamic anisotropies is based on the analysis of triple correlations between one fission fragment and two neutrons from the same event. The method is best understood by inspecting Fig. 10.
X
Fission
v
,J*IS n
I
/
Figure 10. To the left: an extreme anisotropy is assumed where all neutrons are emitted in a plane perpendicular to spin I, the spin itself being perpendicular to the fission axis. To the right: project all neutron events on a plane perpendicular to the fission axis; all neutron events lie on the x-axis.
For the sake of argument assume a fixed orientation of the spin I which from the studies of fragment angular momentum is known to be perpendicular to the fission axis, in the figure the z-axis. For the sake of argument assume further an extreme anisotropy with all neutrons being emitted in a plane perpendicular to the spin I. Projecting all neutron events on a fictitious plane perpendicular to the fission axis, all events are seen to lie on a single line (in the specific choice of the coordinate system the x-axis). In actual practice it is evidently not feasible t o fix the orientation of fragment spin. Instead the fission axis in Fig. 10 can be imposed to be fixed in space by simply selecting those fission events where one of the fission fragments is flying along the z-axis of Fig. 10 in the lab system. Fragment spins will then be equally distributed on a plane perpendicular to this axis. For fixed spin the angular distribution of neutrons is disk-shaped with the symmetry axis coinciding with spin as visualized in Fig.10 for an extreme anisotropy. With the fission axis being fixed, the distribution is found by averaging over all possible spin directions on the plane perpendicular to the
19
fission axis. Thereby the distribution becomes cigar-shaped with the symmetry axis coinciding with the fission axis. When the distribution is finally transformed to the lab system, the dynamic anisotropy is superimposed to the kinematical anisotropy and it is virtually impossible to disentangle both. This was precisely the outcome of the calculations quoted above [22]. To overcome this difficulty it is proposed to analyze triple coincidences of one fission fragment and two neutrons, fission event by fission event. To explain the basic idea let us again assume for the sake of argument that the anisotropy is extreme like in Fig. 10. The fission fragment allows pinning down the fission axis and hence the origin 0 in the projection plane of Fig. 10. The location of the first neutron in the projection plane then tells how the fragment spin is oriented: it is perpendicular to the line joining the neutron point of impact with the origin. The second neutron then serves as a test for perfect anisotropy because only in this case the two neutron points of impact will lie on the same line through the origin. This will be true for any single fission event irrespective of the orientation of fragment spin, the inclination of the neutron line always staying perpendicular to spin just following the varying orientations of spin. Observing the pattern outlined, hence, corresponds to a measurement of extreme anisotropy. Of course, in nature the anisotropy is expected to be finite. With an obvious generalization of the analysis it can be shown that also finite anisotropies are to be assessed. How to proceed is illustrated in the very schematic drawing of Fig.11. In the left figure the (x,y)-projection plane with two neutron impact points is depicted. Note that the origin 0 marks the fission axis which is perpendicular to the plane. In the evaluation event by event the polar angles and @2 are determined and their difference A@ = (@2 - @ I ) is calculated. For an extreme anisotropy, as assumed in Fig. 10, the difference A@ of polar angles of neutron impacts should be strictly A@ = 0 or A@ = 180" depending on whether the two neutrons are observed in the same or in opposite quadrants of the (x,y) plane, respectively. By contrast, for perfect isotropy of neutron emission in the cm-system of the fragment the distribution of A@ angles will be equally distributed in the full range of angles from 0" to 360". In between, i.e. for an assumed non-zero but not extreme anisotropy the A@ distribution will exhibit bumps at angles A@ = 0" (or equivalently 360"), and at angles near A@ = 180". Schematically this A@ distribution is depicted on the right hand side of Fig. 11.
20
0
900
180
A@ = Q2
270
360
- 0,
Figure 11. Left figure: example of a two-neutron fission event as seen in the projection plane of Fig. 10; the fission axis is perpendicular to the plane and marks the origin 0. Right figure: distribution of the differences of polar angles A@ = (@2 - 0,) anticipated for a hypothetical non-zero dynamic neutron anisotropy.
The pattern is intermediate between the delta-like spikes for extreme anisotropy and the perfectly flat distribution in case of isotropic neutron emission. It has to be stressed that in the A@ distribution the influence of the kinematical neutron anisotropy is completely eliminated. As a reminiscence of kinematical anisotropy merely the density of neutron points in the projection plane is larger near the origin than farer away from it. A first experiment designed for the detection of the conjectured dynamic anisotropy was mounted at the CNRS in StrasbourgFrance. The layout is visualized in Fig. 12. A back-to-back Bragg ionization chamber served as the fission detector. The Frisch-gridded 4n chamber was called "CODIS". The common cathode is a segmented Cu plate with a thin foil carrying the 252Cf spontaneous fission source at the center. The chamber enables to measure both the polar and azimuthal angles of fragment emission relative to the chamber axis. For the evaluation the chamber axis (or rather a narrow cone around this axis) was singled out as the fixed fission axis. Details of the chamber construction and its operation are to be found in Ref. [23]. For the detection of neutrons the multi-detector spectrometer "DEMON' was used. It consisted for the present setup of 82 NE213 liquid scintillator cells (length 20 cm, diameter 16 cm) optically coupled to XP4512B photomultiplier tubes. The DEMON detectors were disposed in the form of two cylindrical walls. The front wall with the majority of detectors was mounted at a distance of =: 300 cm and the rear wall at = 140 cm from the center of the ionization chamber. Neutrons and gammas are separated by pulse-shape discrimination. Neutron energies are determined from their time-of-flight. Together with the
21
size of the scintillators pulse the time-of-flight also helped to identify background events. More details are given in Ref. [24]. The combination of the detector systems CODIS and DEMON was dubbed "CORA".
^
/ •'' 4+'""'»-'
OB*Nfert«*
1
^T^
=Up
^Sgi^^zz:^-^. ZB
"I 1
M
Figure 12. Left figure: top view of the experimental setup with the CODIS chamber at the center and the DEMON detectors arranged on cylindrical walls. Right figure: view of the DEMON front wall as seen looking along the chamber/fission axis.
In a run of about 3 weeks some 5-107 triple fission-neutron-neutron events were collected with the front wall neutron detectors. In the evaluation it turned out, however, that the geometrical positioning of modules having been optimized for a preceding experiment were not at all adapted to the problem at hand. The arrangement of neutron detectors is depicted on the right of Fig. 12. As brought to evidence by a Monte Carlo simulation of a CORA experiment with an assumed isotropic neutron evaporation in the cm system of fragments, the efficiency of detection was strongly biased in favor of events at AO angles of 0°, 180° and 360°. These preferred angles of detection may also directly be inferred from simply inspecting the position of modules in Fig. 12. Unfortunately they coincide with the angles characterizing cm anisotropies. The experimental data were of course corrected for efficiency, but due to the chance coincidence in AO angles the results are not considered to be reliable and, hence, are not quoted. More details of the evaluation are described in Ref. [25]. A new experiment is in preparation which above all will have to avoid the above mishap. To that purpose the positioning of DEMON detectors has to be elaborated where the efficiency for measuring angles A<1> is flat in 2n. But there still remains a further important task. Whether an anisotropic emission of neutrons in the cm system is present or not may at least for sizable anisotropies
22
be decided by merely having a look at the A@ dependence on angle. However, the transformation from the structure of the A@ distribution to the physically interesting degree of anisotropy in the cm system of fragments has to be calculated in a Monte Carlo simulation. Hereby the neutron detection efficiency of the DEMON modules has to be taken into account. The analysis is underway ". 5. Outlook The three novel experimental studies on angular momentum of fragments from fission having been discussed are each a starting point for future work. The ROT effect and in particular the ratio TRImOT is predicted to be very sensitive to changes in neutron energy at the eV level [S]. This prediction is a challenge for experiment to be confirmed. To test the dependence of these two T-odd effects on compound spin several fissile isotopes with different spins should be studied. The appealing application of the TFWROT effects in the search for fast scission neutrons and/or scission gammas will have to be pursued ~71. As regards the mechanisms for fragment spin generation in spherical nuclei, only a beginning has been made with so far only one isotope having been scrutinized. Here much work is still ahead clarifying the correlation between fragment spin and nuclear structure which - at least for the example at hand does not follow general rules having been reported. Finally the search for neutron anisotropies in the cm system of fragments should help to solve a long-standing issue. Whether this anisotropy exists or not is not a purely academic question. In fact, a more reliable knowledge of the size of the anisotropy together with a more reliable notion on the existence of scission neutrons will allow calculating with more confidence the neutron spectra of those fissile isotopes which are not directly accessible but of importance for future generation IV reactors. Acknowledgements
We gratefully acknowledge the support of the present studies by several national funding agencies in Russia, France. Belgium and Finland. However, of crucial importance for bringing together the many teams having contributed was the funding by two major INTAS project (99-0229,03-6417).
23
References 1. K. Schreckenbach et al, Proc."Time reversal invariance ", World Scientific, Singapore, 1994, C.R. Gould et al. eds, p 187. 2. P. Jesinger et al., Nucl. Instr. Methods, A 440, 618 (2000). 3. A. Gagarski et al., Proc. "Int. Sem. ISINN-l2", Dubna, Russia, 2004, p 255. 4. V. Bunakov, Phys. At. Nucl. 65,616 (2002). 5. V. Bunakov, S. Kadmensky, Phys. At. Nucl. 66, 1846 (2003). 6. A. Gagarski, Proc. "Int. Sem. ISINN-14", Dubna, Russia, 2006, p 93. 7. F. Gonnenwein et al., Phys. Lett. B 652, 13 (2007). 8. V. Bunakov, S. Kadmensky, Proc. " Int. Sem. ISINN-IS", Dubna, Russia, 2007, to be published. 9. D. De Frenne, in "The Nuclear Fission Process", C. Wagemans ed., CRC Press, Boca Raton, USA, 1991, p 475. 10. J. 0. Rasmussen et al., Nucl. Phys. 136,465 (1969). 11. M. Zielinska-PfabC, K. Dietrich, Phys. Lett. 49B, 123 (1974). 12. L. Bonneau et al., Phys. Rev. C 75,064313 (2007). 13. S . Kadmensky, Phys. At. Nucl. 67, 170 (2004). 14. J. L. Durell, Proc. "Dynamical Aspects of Nuclear Fission", eds. J. Kliman and B. Pustylnik, Dubna, Russia, 1997 p 270. 15. D. De Frenne, Phys. Rev. C 29,1777 (1984). 16. H. Naik et al., Phys. Rev. C 71,014304 (2005). 17. J. Kaufmann et al., 2.f Phys. A 341,3 19 (1992). 18. J.P. Bocquet et al., Proc. " Phys. and Chem. ofFission 1979 " IAEA Vienna, 1980, p 179. 19. J. A. Pinston, J. Genevey, J. Phys. G 30, R 57 (2004). 20. F. Gonnenwein et al., Int. J. Mod. Phys. E 16,410 (2007). 21. A. Gavron, Phys. Rev. C 13,2562 (1976). 22. V. Bunakov et al., Proc. "Int. Sem. ISINN-13", Dubna, Russia, 2005, p 293. 23. Yu. Kopatch et al., Phys. Rev. C 65,044614 (2002). 24. F. M. Marques et al., Nucl. Instr. Meth. A 450, 19 (2000). 25. E. Prokhorova et al., Proc. "Exotic Nuclei", Khanty-Mansiysk, Russia, 2006, AIP Conf. Proc. 912, 2007, eds. Yu.Penionzhkevich, E. Cherepanov, p 173. 26. I. Guseva, private communication. 27. G. Petrov, Proc. "Nuclear Fission and Fission-Product Spectroscopy", Cadarache, France, 2005, AIP Conf. Proc. 798,2005, eds. H. Goutte et al. p 205.
This page intentionally left blank
THE PROCESSES OF FUSION-FISSION AND QUASI-FISSION OF HEAVY AND SUPER-HEAVY NUCLEI* M.G. ITKIS, A.A. BOGACHEV, E.V. CHERNYSHEVA, I.M. ITKIS, G.N KNYAZHEVA,N.A. KONDRATIEV, E.M. KOZULIN, L. K R W A Flerov Laboratory of Nuclear Reaction, JINR. 141980, Dubna, Russia F. HANAPPE Universite Libre de Bruxelles. 1050 Bruxelles, Belgium 0. DORVAUX, N. ROWLEY, L. STUTTGE Institut de Recherches Subatomiques, F-67037 Strasbourg Cedex, France
A.M. STEFANINI, B.R. BEHERA, L. CORRADI, E. FIORETTO, A. GADEA, A. LATINA,S. SZILNER Istituto Nazionale di Fisica Nucleare. Laboratori Nazionali dilegnaro, I-35020 Legnaro, Padova, Italy M. TROTTA Istituto Nazionale di Fisica Nucleare,Sezione di Napoli, 1-80126. Napoli, Italyy S. BEGHINI, G. MONTAGNOLI, F. SCARLASSARA Dipartimento di Fisica and INFN-Sezione di Padova, Universita di Padova, I-35131, Padova, Italy
V.A. RUBCHENYA, W.H. TRZASKA Department of Physics, Universily of Jyvaskyla, FIN-4001 4 Jyvaskyla, P. 0.Box 35, Finland
*
This work is supported by the Russian Foundation for Basic Research under Grant No. 07-0200439.
25
26 Results of the experiments aimed at the study of fission and quasi-fission processes in the reactions 48Ca+144.'54Sm, I6'Er, '"'Pb, 23wU,244Pu, 24'Cm; 50Ti+208Pb,244pU;s'F&20'Pb, 244 Pu, 248Cm,and 64Ni+186W, 242Puleading to the formation of heavy and super-heavy systems with Z=82-122 are presented. Cross sections, mass-energy and angular distributions for fission and quasi-fission fragments have been studied at energies close and below the Coulomb barrier. The influence of the reaction entrance channel properties such as mass asymmetry, deformations, neutron excess, shell effects in the interacting nuclei and producing compound nucleus, the mechanism of the fusion-fission and the competitive process of quasi-fission are discussed.
1. Introduction
The interest to the study of the fission process of heavy and super-heavy nuclei mainly deals with the problem of synthesis of super-heavy elements. The knowledge of fusion-fission (FF) cross sections allows to estimate the cross sections of evaporation residue production in "cold" or "warm" fusion reactions. However for the reaction between heavy colliding nuclei the quasi-fission (QF) strongly inhibits the process of the compound nucleus (CN) formation. The competition between QF and complete fusion is influenced by the properties of the di-nuclear system at contact configuration, where entrancechannel effects play the major role in the reaction dynamics. A decrease in the entrance-channel mass-asymmetry along with an increase in CN-fissility are responsible for the appearance of QF manifested in the suppression of the fusion cross section for combinations, leading to strongly fissile compound nuclei [ 1, 2, 3, 41. Another circumstance that influences on the fusion probability is the relative orientation of deformed nuclei, which changes the Coulomb barrier and the distance between the centers of the colliding nuclei [5, 61. Nuclear shells in the interacting nuclei, as well as in the formed composite system and reaction fragments, influence strongly on the survival probability of the compound nucleus and the reaction dynamics in general. The characteristics of both processes (FF and QF), their manifestation in the experimental observables and the relative contribution into the capture crosssection in dependence on the energy of interaction and reaction entrance channel properties were investigated for the wide range of target-projectile combinations. Cross sections, mass-energy and angular distributions for fission and quasi-fission fragments obtained in the reactions 48Ca+144,154Sm, 168Er,208Pb, 238u, 244pu, 2 4 8 ~ 50Ti+208pb, ~; 244pU;58Fe+208pb, 244Pu, 248Cm,and 64Ni+'86W, 242Puleading to the formation of heavy and super-heavy systems with Z=82-122 have been studied at energies close and below the Coulomb barrier. The influence of the reaction entrance channel properties, such as mass asymmetry,
21
deformations, neutron excess, shell effects in the interacting nuclei and producing CN on the mechanism of the fusion-fission and the competitive process of quasi-fission are discussed. The experiments were carried out at the U-400 accelerator of the Flerov Laboratory of Nuclear Reactions (JINR, Russia), the XTU Tandem accelerator of the National Laboratory of Legnaro (LNL, Italy) and the Accelerator of the Laboratory of University of Jyvaskyla (JYFL, Finland) using the time-of-flight fission fragments spectrometer CORSET. 2. Results and Discussions 2.1. The influence of entrance channel mass asymmetry
The effect of mass-asymmetry in the entrance channel q = (M,,, - MiOn)/(Mu,+ Mion)is seen clearly in the 44Ca+ 206Pb(q = 0.648) and 64Ni+ 186W(q = 0.488) reactions leading to the same 25%0* compound nucleus.
mass, LI
Figure 1. Yields of fission-like fragments in the 44Ca+206Pband 64Ni+'86Wat CN excitation energies 40 MeV.
Figure 1 shows the mass distributions for the 44Ca+ 206Pband 64Ni+ '86W at energy close to the Bass barrier (the CN excitation energy is about 40 MeV). One can see that the mass distributions for these systems are very different. In the case of 44Ca, the mass distribution has a complicated structure: i) the asymmetric fission connected with the influence of the deformed shell near the heavy fission fragment mass 140; ii) the symmetric fission component; and iii)
the QF component, visible around Z = 28 and N = 50. In contrast to this reaction, the contribution of the QF component into the total mass distribution in the case of 64Ni+ lg6Wincreases greatly. The analysis of the mass-angular distributions of the fission fragments allow one to derive the QF and FF components from all fission-like products detected in the experiment. The angular distribution for the asymmetric (where we expected the QF process) and symmetric (where we assume the domination of CN-fission) mass splits for both reactions were extracted. In Figure 2 the angular distributions for the selected mass bins of fission-like fragments are shown.
Ni(311 MeY)+’86W+2”No
“Ca(227 MeY)+206Pb+2“No I000
61
------I 105 < m < 125 ( X I0)
I00
3
a
a
100
2
2
E
6 9 b a
E
10
0”
9 b
10
-0
I
1
20 40 60 80 100120140160180
0. I
0 20 40 60 80 100120140160180 @c.m.,
deg
Figure 2. Differential cross section for fission-like fragments for the reactions 44Ca+206Pband Ni+IE6Wfor different mass bins.
64
For the symmetric mass split the angular distribution is isotropic for the reactions with 44Caand 64Niand the value for the value of I& agrees well with expected for the CN-fission. It means that typically the nucleus probably rotates several times before scission and the main process, leading to the symmetric mass split, is CN-fission for both reactions. Experimentally, evidence for long CN-fission time scales comes from fission angular distributions. Pre-scission neutron measurements also indicate on the long fission time scale of several 102o s for CN-fission process [7, 81.
29
For the asymmetric mass region the significant forward-backward asymmetry in angular distribution is observed. From the mass-angular correlation it is evident that the composite system, which suffers the asymmetric mass split (A » 80 a.m.u) for both reactions, rotates less than one turn. From the mass-energy and angular distributions it follows that the main process for the symmetric mass split is CN-fission of 250No for both studied reactions. In that way we can estimate the fusion probability for the reactions 44Ca+206Pb and 64 Ni+186W. Only a small part around 30% of the fission cross section can be associated with CN-fission for the 64Ni + 186W system, the remaining part should be attributed to QF. In the case of 44Ca + 206Pb, the contribution of CNfission component into the total mass distribution is ~70%.
150
50 75 100125150175200 mass, u
100
150 200
mass, u
Figure 3. TKE-M matrixes for the reaction 58Fe+208Pb->266Hs at E'= 14-40 MeV (left panels) and 26 Mg+248Cm->274Hs at 32-64 MeV (central panels). The excitation functions for both reactions are shown at the right panel.
Figure 3 shows the formation of the isotopes of 266-274Hs in the "cold" 266 Hs) and "hot" fusion (26Mg+248Cm->274Hs) reactions. fusion (58Fe+208Pb Although in the former reaction 266Hs is produced at lower excitation energies which should increase the CN survivabiliry in the de-excitation process, the quasi-fission process is nevertheless the main reaction mechanism at the
30
asymmetry of the entrance channel q=0.56. In the 26Mg-inducedreaction a more neutron-rich isotope 274H~is formed with the asymmetry coefficient q=0.81. As one can see from the central part of Fig.3 TKE/mass matrixes have triangular shapes typical of the fission of heated nuclei, described by the Liquid Drop Model (LDM) [9]. Only at lower excitation energies E* = 31.7 - 35.3 MeV some difference appears in the matrix shape from the LDM predictions. The multimodal fission of 274Hs was revealed, and the QF component also appeared with decreasing excitation energy. The contribution of QF to oCap is oQF=12%at E*=35.3 MeV. The right-hand panel of Fig.3 shows the excitation functions for both reactions.
600
2
400 200
0 2500
2000 1500 *13=
1000
500 0
160 180 200 220 240 260 280 300 320 TKE, MeV
Figure 4. Mass yield, (TKE) crTm (M) and crM(TKE)for the reaction 26Mg+Z48Cmat energies Elab=129,143 and 160 MeV
2.2. Bimodaljission of 274Hs Figure 4 presents the mass yields, (TKE)(M) and variances o ~ ~ = ( M and ) o~~(TKE for) the reaction 26Mg+248Cm at energies El,b=129, 143 and 160 MeV. (E*=35.3, 48 and 64.3 MeV, respectively). Fig 4 a) shows the shoulders in the mass yield for MH=193-220amu and additional light fragment masses for the
31
Elab=129,143 MeV. The fit of the mass yield by the Gaussian according to the LDM model is shown by solid line. The experimental dependence (TKE)(M) was fitted by parabola (dash line in Fig4 b):
TKE =TKE(A/2)(1-p2)(1+pp2), where p=(M-A/2)/(A/2), 26
Mg +248Cm-+274Hs (E*=35.3 MeV)
J 601
40.
I
it i
1TKE
> 220 Me
Wd,
mass, u
Figure 5. Mass yields of the 274Hs(E*=35.3MeV) fission fragments for different TKE ranges.
p=O. 17 is the empirical parameter, defining the parabola width [101. One can see from Fig.4a) for all three excitation energies (TKE)(M) is higher for the masses MH=193-220amu than the LDM parabolic dependence. The similar increase of (TKE)(M) was already observed for QF component of 256N0[l 13. The variance 0 ~ , - ~ ( h 4increases ) for the mass region MH=200-220amu, revealing indirectly the presence of QF and FF processes. QF component on the edges of the mass distribution is caused by the closed shells with Z=28 in light fragment and Z=82, N=126 in the heavy fragment (Fig.4a). Fig. 4d) shows the mass variance oM as function of kinetic energy. For TKE=l60-220 MeV 0 2increases ~ with decreasing of the excitation energies of the compound nucleus. Only for
32
TKE>220 MeV ozMfor the lowest energy (E*=35.3 MeV) is reduced, and mass distribution becomes very narrow.
140
160
I80
260
220
240
260
280
300
Mtotal
120
140
160
180
200
220
240
260
280
300
TKE,MeV Figure 6 . T I E distributions of the 274Hsfission fragments for a) symmetric fragmentation M=A/2 +20 amu; b) for all masses.
Figure 5 shows the mass yields of the fragments at E*=35.3 MeV for different TKE ranges. For TKE>220 and TKE>208 MeV one can observe that mass yield consists of two fission modes - symmetric narrow and wider one, which manifests itself as a pediment in the mass distribution. In case of symmetric fission of 274Hs(Nm=166) both fragments are close to the magic shell N=82, that results to narrow mass distribution. The bottom panels of Fig.5 exhibits the wide mass distribution for TKE<208 MeV. The TKE distributions for all masses and for the symmetric mass region (A/2,20) are shown in Fig.6. It consists from two components with (TKE) = 198 MeV and 227 MeV. The strong increase of the yield of high-energetic symmetric fission mode when neutron number in both fragments is close to the shell N=82 is the general tendency for the super-heavy nuclei [12]. The effect of bimodal fission of 2 7 4 H ~ (N=l66) is observed even more obviously than that for 270Sg(N=l64) [13], probably due to the pre-fission neutrons.
33
2.3. The influence of deformation and fragment shell on quasi-fission
The mass-energy distributions of binary fission-like fragments produced in the reactions ‘60+’86W,40Ca+’54Smand 48Ca+’443’54Sm have been measured in the energy range from the Coulomb barrier to well above it, up to the excitation energy E*=100 MeV [14]. For all the reactions the main component of the distributions corresponds to a symmetrical mass division with typical values of the variance and mean total kinetic energies of fragments, which are inherent in CN fission.
fragment mass (a.m.u.)
Figure 7. Location of the closed shell (designated by arrows connected with horizontal lines for complementary fragments) in the fragment mass distributions obtained at the lowest CN excitation (upper panel), 40Ca+154Srn(middle panel) and 48Ca+’44Sm energies for three reactions 4RCa+154Sm (bottom panel).
For the 40,48Ca+’s4Smreactions an asymmetric component has been observed in the fission-fragment mass distributions. The analysis of massenergy distributions for these asymmetric “shoulders” points at the QF nature of this component. These QF “shoulders” peak around the masses having closed shells numbers of protons and neutrons (see fig. 7).
34
Figure 8. Differential cross sections for fission-like fragments in the reactions 48Ca+144,154 Sm for different fragment mass bins.
In Figure 8 the angular distributions for the selected mass bins of fissionlike fragments obtained in the reaction 48Ca+144,154Sm are shown. One can see that for the reaction 48Ca+’44Smangular distributions are symmetrical for all selected, whereas for the reaction with lS4Sma significant forward-backward asymmetry is observed in the angular distributions for light fragment masses. This forward-backward asymmetry is observed at both energies of 48Ca,but the contribution of the asymmetric-mass portion is larger at the lower energy. As one see, the symmetrical parts can be associated with pure CN-fission process, whereas the asymmetric-mass “shoulders” correspond to the pronounced forward-backward asymmetry in angular distributions and these events should be associated with the QF process. Preferential forward-peaking of the lightmass fission-fragments (AL=73-82) demonstrates that QF bypass the CN stage and occurs in time scales shorter than the rotational period for the 202Pb*.This leads to a broken forward-backward symmetry of the angular distribution in the c.m. system for the asymmetric masses of fission fragments. In contrast to 48Ca+’s4Sm,symmetric angular distributions for all masses of fission fragments are clearly observed in the 48Ca+’44Smreaction. In the latter case a CN nature of fission fragments is implied, i.e., one can state that the projectile-target system leading to lg2Pb’lives so long that is enough for several rotations. As the system rotates, it “forgets” entrance-channel conditions of its formation, i.e., moves to the CN production.
35
Thus in the 48Ca+144Sm reaction, no evidence of QF was found at the same CN excitation energy and angular momentum as in the case of the reaction with IS4Sm,where QF is evident. A small QF component has been also detected in the 40Ca+'s4Smreaction at energies close to the Coulomb barrier. So, the QF effect is manifested in the reactions with the deformed target nucleus IS4Sm, which correspond to greater values of the entrance-channel mass-asymmetry than in the case of the 48Careaction with spherical '44Sm.The fusion probability is reduced at the near-tip collisions below the nominal fusion (Bass) barrier and QF appears. The near-side collisions at these low energies lead to small fusion probability due to a higher value of the fusion barrier. At above-barrier energies all target orientations contribute and near-side collisions mainly determine the CN-formation. Qualitatively, this consideration corresponds to the model proposed for the explanation of a high angular anisotropy of fission fragments, which is observed in reactions leading to strongly fissile compound nuclei. 2.4. Reactions with heavy ions leading to the formation of super-heavy
systems The neutron-reach isotope of 48Cais used in experiments on SHE synthesis due to its undoubted advantages. Firstly, high neutron excess allows one to reach in the reaction with actinide targets the neutron number on the CN Nc+170-180 in contrast to cold fusion reaction, in which Nm=150-160. Secondly, the doubly magic structure of 48Ca leads to the excitation energy E*=: 30 MeV at the Coulomb barrier for elements Z= 112-116, that is by 10-15 MeV lower than in classical hot fusion reactions. Figure 9 presents MED of fragments of the elements with Z = 102-116, produced in the 48Cainduced reaction on the targets 208Pb,238U,244Puand 248Cm at the excitation energy E*= 33 MeV. On the top of Fig.9 two-dimensional matrixes of (TKEMass) are presented, the mass yields of the reaction products are shown in the bottom panel. The main peculiarity of the data is the sharp change of the MED triangular shape in the reaction 48Ca+208Pb, in which the FF process dominates, to the QF shape of MED in the 286112-2961 16 nuclei. The distinctive feature of the quasi-fission process for these super-heavy nuclei is the wide two-humped mass distribution with a high peak of heavy fragments near the doubly magic lead (MH=208).In spite of the dominating role of the quasi-fission process for these reactions we assume that in the symmetric region of the fragment masses (M2f20) the FF process coexists with QF. The mass yield of the fusion-fission process is shown in a frame at the bottom of Fig. 9. It is obtained as a difference between experimental spectra and quasi-fission peak
36
descriptions. One can see there that the mass distribution of the fusion-fission process is asymmetric in shape, with a nearly constant mass of the light fragment ML»132-134 amu. In the case of super-heavy elements the light spherical fragment with ML«132-134 amu plays a stabilizing role, whereas the heavy fragment with M H wl40 is playing the same role for actinide nuclei.
50
100
150
200
250
50
100
150
200
250
50
100
150
200
250
SO
100
150
200
250
mass, u Figure 9. Two-dimensional matrixes TKE-Mass (top panels) and Mass yields of the fragments (bottom panels) for the reactions 48Ca+ 208Pb, 238U, 244Pu, 248Cm at the excitation energy E*«33 MeV.
An experimental two-dimensional total kinetic energy (TKE) mass plot for the 48Ca+248Cm fusion-fission reaction is shown in Fig. 10 along with the corresponding potential energy surface [15] determining evolution of the nuclear system. When the nuclei are in contact, the nuclear system typically evolves in the asymmetric QF channels: with high probability path number 1 in Fig. 10(a), which populates the area of fragment masses near A =208 in Fig. 10(b). The asymmetric QF channels are closer to the initial state in the configuration space of collective degrees of freedom as compared with the configurations through which the system has to pass on the way to the compound nucleus: trajectory number 3 in Fig. 10(a). As a result, only a small part of the incoming flux reaches a compound nucleus configuration, and the fusion cross section turns out to be far less than the capture cross section. The distinction between GfM(E) and acapi(E) becomes still more evident at low excitation energies. The symmetric region of fission fragments [(,4cw/2)+20; area of dashed quadrangle in Fig. 10(b)] is shown in Fig. 10 (c), compared with the typical mass yield of 238U fission fragments [16]. As it can be seen from the figure, the mass spectra in the overlapped region practically coincide within the error bars.
37
In the case of 238U a two-humped fission mass distribution is regulated mainly by a doubly magic heavier fragment 132Sn, which plays the role of a lighter fragment in the case of fission of a 296116 nucleus at low excitation energies.
Figure 10. (a) Driving potential as function of mass asymmetry and distance between centers of two nuclei with zero deformation, (b) Two dimensional TKE-Mass matrix. Different regions are numerated in accordance with the most probable trajectories (shown on upper panel) contributed to these regions, (c) Mass distribution of fission fragments of 238U at the excitation energy E*«30 MeV (solid circles) and those of the SHE 296116 produced in the reaction 48Ca+248Cm (open circles).
The graphic example of the shell effect manifestation in the MED of the fission fragments is shown in Fig. 11. One can see (Fig. 11) that in the case of the heaviest targets 238U, 244Pu and 248Cm the ratio aQF/acap changes slightly, and the tendency was also observed in the aER excitation functions for the superheavy elements [15].
38 («.-, ,2s,,. as-, GJ + Pb-+ No
«„
,2»._,
2S8,,.,
Ca+ Cm-* 116
E=33MeV
E"-4fl.4MeV
am 8000
BOOO 4000 2000
100
150
50
100
150
50 100 150 200
50 100 150 200 250
Mass, u Figure 11. Two-dimensional matrices TKE-Mass (top panels) and mass yields (bottom panels) of fission fragments of "2Pb, 216Ra, 256No, 286112 nuclei produced in the reactions with 48Ca.
Fig. 12 shows the ratio of the QF to the capture cross-sections aQF/CTcap as a function of the composite nucleus mass for the reactions with 48Ca-projectiles at excitation energies E*=33-40 MeV. The solid circles represent the measured reactions; the question marks are the reactions to be investigated. Our prediction of this curve behavior is shown by the line. It is seen that the QF contribution increases for the targets lighter than 208Pb and decreases at the 144Sm target. The most probable explanation of such an unusual behavior of the ratio aQF/acap is the influence of spherical shell closures in the entrance channel and in the nascent fragments. This means, that symmetric region of fission fragment masses [(^CV2) ± 20] seems to originate mainly from the regular fusion-fission process in the reaction 48Ca+248Cm-»2%116. However, as shown in Ref. [15], evolving from the initial configuration of two nuclei in contact into the state of spherical or near-spherical compound nucleus [path number 3 in Fig. 10(a)], the system goes through the same configurations through which a compound nucleus goes in regular fission [path number 4 in Fig. 10(a)], i.e., configurations close to the saddle point. In such configuration and in a state of complete thermodynamic equilibrium, the nuclear system is much likely to go into the fission channel [path number 2 in Fig. 10(a)], without overcoming the saddle point, and producing a spherically symmetric compound nucleus.
39 100-
.I
80 -
-
E',,= 33 40 MeV
200
220
240
I
48Ca+238
260
280
300
Mass of Composite Nucleus, u
Figure 12. The ratio of oQF/oCap as function of the composite nucleus mass number for reactions with 4RCa-ionsand different targets.
This kind of process results in fragments that are practically not different from regular fission fragments, since in both cases the system follows the same path from the saddle point to the scission point. This means that among all the events resulting in the system going in regular near-symmetric fission channels, there are such events in which the system does not produce a true spherically compound nucleus. Columns (a), (b), (c) of Figure 13 show the experimental data for the reactions with "Fe projectile on 232Th,244Puand 248Cmtargets, leading to the formation of the compound system 290116 and the heaviest compound systems '02120 and '06122 (where N = 182-184), i.e. to the formation of the spherical compound nucleus, predicted by theory [17]. As seen from Fig. 13, in these cases we observe an even stronger manifestation of the asymmetric mass distributions of 306122and 302120fission fragments with the light fragment mass at about 132. The corresponding structures are seen well in the
distribution as a function of fragment mass. Only for the reaction 58Fe+ 232Th-+290116 (E*= 53 MeV) the valley in the region of M=A/2 disappears - this is seen from the mass yield distribution as well as from the (M) dependence. This fact is connected with a reducing of the shell effects influence at so high excitation
40
b)
a)
50
100
150
200
58r^
250
50
100
150
200
14%*-.
31*6.. it
Fe + Cm -» 122 E'= 31.5 MeV
*¥*+**?*-^"I E*=44MeV
E'= S3 MeV
d)
c)
250
50
100
150
200
E"-31.5 MeV
250
50
100
150
200
250
Mass, u Figure 13. Two-dimensional TKE-Mass matrices, the mass yields, average TKE and the variances cr2TKE as a function of the fission fragment mass for 290 116, 302 120, 306 122, produced in the reactions with 58Fe and M Ni projectiles.
Fig. 13(d) shows the data for the reaction 64Ni+242Pu-» 306122. In this reaction the same nucleus 306122 is formed at the same excitation energy as in the previous case (Fig. 13(c)), and the asymmetry of the entrance channel changes only slightly. Mass distributions look similar in both cases, with characteristic QF peaks in the region of doubly magic lead, however, energy characteristics of the fission fragments differ. In the case of the reaction with 58Fe-projectile the dependence (M) has a valley in the symmetric part and peaks corresponding to mass 132 and complimentary masses. For the reaction with 58Fe ions, (M) varies in the limit of 25 MeV. The 64Ni-induced reaction yields a totally different dependence: (M) changes only slightly, by about 5 MeV within the entire range of masses.
41
For the reaction with 64Niions the dependence 02TKE (M) is parabolic and is smaller in amplitude than that in the reaction with S8Feions, where it is seen (despite low statistics) that the variance ozTKE (M) increases in the central part, that indicates on the co-existence of both processes in this region- fission and QF. Thus, it indirectly suggests that in the 64Ni-inducedreaction the contribution of QF grows as compared with the reaction with 58Fe ions. This fact may possibly explain the reasons of failed attempts to synthesize element 293 118 in the 86Kr projectile reaction [18], since an increase in the asymmetry of the entrance channel in this case leads to the strong domination of the QF process.
2.5. Neutron and gamma-ray multiplicities in the fission of SHE Formerly emission of neutrons and y-rays in correlation with fission fragments in the decay of super-heavy compound systems at excitation energies near or below the Coulomb barrier had not been extensively studied. At the same time such investigations may be extremely useful for an additional identification of the fusion-fission and quasi-fission processes and consequently for a more precise determination of the cross-sections of these processes in the total yield of fragments. On the other hand, the knowledge of the value of the fission fragment neutron multiplicity can be used in the identification of SHE in experiments aimed at their synthesis. The results of such investigations are presented in Fig. 14 for the reactions 48 Ca + 244Pu,248Cmat energies near the Coulomb barrier. As seen from the figures, in all cases the total neutron multiplicity (v:) is considerably lower (by more than twice) for the region of fragment masses where the mechanism of quasi-fission predominates as compared with the region of fragment masses where, in our opinion, the process of fusion-fission prevails (in the symmetric region of fragment masses). Another important peculiarity of the obtained data is the large values of ( V y ) = 9.2 and 9.9 for the fission of 2921 14 and 2961 16 compound nuclei, respectively. Considerable differences have been observed in the values of y-ray multiplicities for different mechanisms of SHE decay as well as for (v:') .
42
60 90 120 150 180 210 240
60 90 120 150 180 210 240
Mass (u)
Mass (u)
Figure 14. Two-dimensional TKE-Mass matrices (top panels); the mass yields (the solid circles), neutron (the stars) and y-ray multiplicities (the open circles) as the dependences on the fission fragment mass (bottom panels) for the reactions 48Ca +248Cm and 4§Ca+M4Pu.
3. Conclusion and Outlook Mass and energy distributions of fragments, fission and quasi-fission cross sections, multiplicities of neutrons and gamma-rays have been studied for a wide range of nuclei with Z = 82-122 produced in the reactions with 12C, 22Ne, 26 Mg, 48Ca, 58Fe, MNi and 86Kr ions at energies close and below the Coulomb barrier. In the case of the fission process as well as in the case of quasi-fission, the observed peculiarities of fragment mass and energy distributions, the ratio between the fission and quasi-fission cross sections, in dependence of the nucleon composition and other factors, are determined by the shell structure of the formed fragments. It was found that the target deformation and mass asymmetry of entrance channel have a dominant role on the evolution of the composite system, whereas shell effects in the exit channel determine the main characteristics of the QF fragments. Strong manifestation of the shell effects were found for quasi-fission fragments of the nuclei with Z=l 12-122 caused by the nuclear shells Z=28, N=50 in light fragment and Z=82, N=126 in heavy fragment. Analysis of the fusion-fission components showed that the mass distribution of the fusionfission fragments is asymmetric in shape with the mass of the light fragment ML«132-134 amu. The multimodal fission phenomena were observed for superheavy nuclei 274Hs. It is important to note that in the case of the quasi-fission
43
process the influence of the shell effects on the observed characteristics is much stronger than in the case of classical fission of compound nuclei. A further progress in the field of synthesis of super-heavy nuclei can be achieved using hot fusion reactions between actinide nuclei and 48Ca ions as well as actinide nuclei and 58Fe or 64Ni ions. For future experiments on the synthesis of super-heavy nuclei of up to Z = 122, new research and more precise quantitative data obtained in the processes of fusion-fission and quasi-fission of these nuclei in reactions with 58Feand 64Niions are required.
References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.
R. Bock et al.,Nucl. Phys. A388, 334 (1982). J.Toke et al., Nucl. Phys. A440,327 (1985). B.B. Back, Phys. Rev. C31,2104 (1985). W.Q. Shen et al., Phys. Rev. C 36, 115 (1987). D.J. Hinde et al., Phys. Rev. C 53, 1290 (1996). J.C. Mein et a1.Jhy.s. Rev. C 55, R995 (1997). W.P. Zank et al., Phys. Rev. C33, 519 (1986). D.J. Hinde et al., Nucl. Phys. A452, 550 (1986). J. R. Nix and W. J. Swiatecki, Nucl. Phys. 71, 1 (1965) M.G.Itkis et al., Sov.J.Purt.Nucl., 19,301 (1988). M.G. Itkis et al., Nucl. Phys. A 734, 136 (2004). D.C. Hoffman et al., Radiochim.Acta 70/71, 135 (1995) M.G.Itkis et al., Phys.Rev. C59, 3172 (1999) G. Knyazheva et al., Phys. Rev. C75,064602 (2007). V.I. Zagrebaev, Phys. Rev. C64, 034606 (2001) ; J. Nucl. Radiochem. Sci., 3, No 1, 13 (2001). 16. A.A. Goverdovsky (private communication). 17. Z. Patyk, A. Sobiczewski, Nucl. Phys., A533, 132 (1991). 18. V. Ninov et al., Phys. Rev. Lett 83, 1104 (1999).
This page intentionally left blank
Fission Cross Sections and Fragment Properties
This page intentionally left blank
MINOR-ACTINIDES FISSION CROSS SECTIONS AND FISSION FRAGMENT MASS YIELDS VIA THE SURROGATE REACTION TECHNIQUE B. JURADO, G. KESSEDJIAN, M. AICHE , G. BARREAU, A. BIDAUD, S. CZAJKOWSKI, D. DASSIE, B. HAAS, L. MATHIEU, B. OSMANOV CENBG. CNRS/IN2P3, Univ. Bordeaux I, 331 7.5 Gradignan, France L. AUDOUIN, N. CAPELLAN, L. TASSAN-GOT, J.N. WILSON IPN, CNRS/IN2P3. Univ. Paris-Sud, 91405 Orsay, France E. BERTHOUMIEUX, F. GUNSING, CH. THEISEN CEA Saclay, DSMDAPNIAKPhN, 91 191 GiJsur-Yvette cedex, France E. BAUGE CEA. SPN, BP12, 91 680 BruyBres-le-Chitel, France
0. SEROT CEA-Cadarache. DEN/DER/SPRC/LEPh. 13108 Saint Paul lez Durance, France I. AHMAD, J.P. GREENE, R.V.F. JANSSENS Physics Division, Argonne National Laboratory, 9700 S. Cass Avenue, IL 60439, USA The surrogate reaction method has been used to determine neutron-induced fission cross sections for the short-lived actinides 242s "'Cm and 24'Am. The latest direct neutroninduced measurement for the 243Cmfission cross section is questioned by our results since there are differences of more than 60% in the 0.7 to 7 MeV neutron energy range. Our experimental set-up has also enabled us to measure for the first time the fission fragment "pseudo-mass'' distributions of 243s2443245Cm and 242Amcompound nuclei in the excitation energy range from a few MeV to about 25 MeV.
1. Introduction
The determination of minor actinides (mainly Np, Cm and Am isotopes) neutron-induced fission cross sections in the energy range from 1 to 10 MeV is of considerable interest for the transmutation of nuclear waste as well as for the determination of not yet well known fundamental quantities such as transition states, level densities and fission barrier parameters. Fission-fragment mass 47
48
yields of these nuclei are also of great importance for the understanding of fission and, in particular, for studying the role of shell effects on the fission process. Fission-fragment mass distributions are also very important for the operatiodsafety of a reactor as they are strongly related to the reactor neutron balance and to the radioactivity of the spent fuel. However, in the case of the Cm isotopes, the available data are rather scarce. For instance, the only available data for 242Cmare the cross-section measurements for fission induced by neutrons with energies of 0.1-1.4 MeV performed by Vorotnikov et al. [l] in 1984. No data are available on the fissionfragment yields produced in the 242Cm(n,f)reaction. The reason for this lack of data is the short half-life of 242Cm(163 days) which makes it very difficult to produce and to manipulate targets of this isotope. The experimental technique we present here allows one to overcome these difficulties. This indirect method, which is usually called the surrogate method, was developed in the 70‘s by Cramer and Britt [2]. It consists in measuring the decay probability of a compound nucleus (e.g. fission or radiative capture) produced via an alternative (surrogate) reaction, e.g., a few-nucleon transfer reaction. The chosen surrogate reaction is such that the resulting nucleus has the same mass A and charge Z as the compound nucleus that would be formed if a neutron would be directly absorbed by the minor actinide. The neutron-induced fission cross section is then deduced from the product of the measured fission probability and the compound nucleus cross section for the neutron-induced reaction obtained from optical model calculations [3]. The surrogate method relies on the validity of the Weisskopf - Ewing limit in which the fission probability is independent of the spin and parity of the compound nucleus. The conditions under which the Weisskopf - Ewing limit applies have been investigated in refs. [4, 51. It was stated in ref. [5] that this limit holds when the excitation energy is high enough for the decay widths to be dominated by the statistical level density, and when the angular momentum of the compound nucleus is not much larger than the spin-cutoff parameter of the level density distribution, which, for the actinide region, is around 6-7 h. Our group has already applied this technique to the measurement of the neutron-induced fission [6] and capture [7] cross sections of 233Pa(half-life of 27 days) via the transfer reaction 232Th(3He,p)234Pa.In the present contribution, we will concentrate on a recent experiment where we have applied the surrogate reaction method to determine the neutron-induced fission cross sections of 242, 243Cm and 24’Amand the fission-fragment mass yields of 243,244.245 Cm and 242Amfissioning systems.
49
2. Experiment In this experiment, the access to neutron-rich Cm isotopes via few-nucleon transfer reactions with a light projectile such as 'He implied the use of a 243Am target. Two targets, of approximately 106 pg/cm2, were prepared at the Argonne National Laboratory. Each target was deposited on a 75 pg/cm2 carbon backing. The 3He beam of 24 and 30 MeV was provided by the Tandem accelerator at the IPN Orsay. The 'He-induced transfer reactions on the 243Amtarget lead to the production of various heavy residues. Table 1 lists the different transfer channels we considered in the present experiment and the corresponding neutron-induced reactions that the surrogate method allows to "reconstruct". Table 1. Transfer channels investigated in the reaction 'He+243Amand the corresponding neutroninduced fission reactions. Transfer channel 243Am(3He,p)245Cm
Neutron-induced reaction 244~m(n,f)
Table 1 illustrates the advantage of using transfer reactions with respect to the standard direct method: the simultaneous access to several transfer channels allows one to determine cross sections of various nuclei from just one projectiletarget combination. Moreover, since there are two bodies in the outgoing reaction channel, the excitation energy of the heavy nucleus E* follows a broad probability distribution. The compound nucleus excitation energy can be translated into a neutron energy En via the relation E*=B,+(A-l).E,,/A, where A and B, are the mass and the neutron binding energy of the compound nucleus, respectively. Therefore, with fixed beam energy, the surrogate method allows the determination of cross sections as a function of neutron energy. 2.1. Experimental set-up
The detection set-up used to determine the fission probability and the mass yields of the different compound nuclei formed after a transfer reaction is displayed in fig. 1. Two sets of two Si telescopes placed at 90 and 130 degrees with respect to the beam axis served to identify the light charged particles emitted. If the corresponding heavy residue undergoes fission, one of the fission fragments is detected in coincidence with the light particle by means of a
50
fission-fragment multi-detector. This multi-detector consisted of 15 photovoltaic cells distributed among 5 units; each unit composed of 3 cells placed vertically one above the other. Four units were placed in the forward direction with an angular coverage from 14 to 125 degrees. The fifth unit was positioned at 180 degrees from the foremost unit. In this way, the fission fragments hitting the foremost unit were detected in coincidence with their complementary fragment in one of the cells of the fifth unit. The determination of the kinetic energies of the two fragments of a given fission event allows to infer the fission fragment mass distribution. The fifth unit also serves to add a point at backward angles to the measured angular distribution. More details on the experimental set-up can be found in ref. [ 6 ] . detectors
Si Telescopes
Figure 1 . Top view of the set-up for fission probability and mass distribution measurements.
3. Fission cross sections
The Si telescopes allowed for the identification of the light charged particles and for the determination of their kinematics parameters (energy and angle). With this information and the related Q-values, we could determine the excitation energy E* of the corresponding compound nuclei. The left part of fig. 2 shows the energy loss versus the residual energy measured in one of the telescopes. The typical hyperbolas corresponding to the different light charged particles can easily be distinguished. By selecting one type of light particle, for example tritons t, we can construct the spectrum represented by the solid line on the right of fig. 2, the so-called "singles" spectrum. This spectrum represents the number of tritons, i.e. the number of 243Cmnuclei, Nsingr as a function of their excitation energy. The broad peaks at the highest excitation energies stem from transfer reactions on the carbon backing and on I60impurities in the target. The background generated by reactions on the carbon support was measured separately. In this way, we could subtract from the singles spectrum the events arising from reactions on the carbon backing. The remaining singles spectrum
51
has been extrapolated under the l60peaks, introducing an additional source of uncertainty. If we now select the tritons detected in coincidence with a fission event, we obtain the spectrum represented by the dashed line in the right part of figure 2 which represents the number of 243Cmnuclei that have undergone fission, Ncoin.For each excitation energy bin, we can then determine the ratio between the fission events spectrum (dashed line) and the compound nucleus spectrum (full line). This ratio, corrected for the fission detector efficiency Eff(E*), gives the fission probability of 243Cmas a function of the excitation energy: P~E*)=Nco,,(E*)/(N,,,(E*).Eff(E*)). AEIch Counts
0
Ereslch
5
10
15
20
E*/MeV
Figure 2. Left: Energy loss versus residual energy in one of the Si telescopes. Right: Number of tritons as a function of the excitation energy of 243Cm(see text).
The geometrical efficiency of the fission detector is approximately 47%. This efficiency has been calculated with a Monte-Carlo simulation that reproduces the experimental efficiency obtained with a 252Cfsource. However, for fission induced by neutrons or light charged particles, the fission-fragment angular distributions can be forward peaked. This anisotropy depends on the angular momentum of the system undergoing fission. Therefore, we should actually use an effective detection efficiency that includes not only the geometry of the fission detectors, but also the fragment angular distribution effects. The arrangement of our fission detectors allows one to measure the angular distribution anisotropy; with this information and the Monte-Carlo simulation it is possible to calculate the effective efficiency. However, the effect of the angular anisotropy on the detector efficiency has not yet been determined; thus, our data have only been corrected for the geometrical efficiency. Nevertheless, as shown in [6], the angular distribution effect is estimated to be of only 2-3 %. We have determined the fission probabilities of 2443243Cm and 242Am, respectively. The associated neutron-induced fission cross sections have been deduced by multiplying the experimental fission probability by the corresponding calculated compound nucleus cross section [3]. The preliminary results are illustrated in fig. 3. The error associated with the compound nucleus
52
cross section of ~ 5% has not yet been included in the results. As shown in the left panel of fig. 3, from the fission threshold up to around 5 MeV the 24l Am(n,f) cross section is in very good agreement with all the evaluations, beyond 5 MeV it is not possible to say which evaluation reproduces our data best. The right part of fig. 3 shows our results for the Cm(n,f) in comparison with the data by Vorotnikov et al. [1]. There is excellent agreement between both sets of data. For neutron energies larger than 1.4 MeV, no other experimental data exist. This presumably explains the important discrepancies between the various international libraries. JENDL and JEFF present the best overall agreement with our data. The remarkable agreement found at the lowest neutron energies between our data and the neutron-induced measurements suggests that the spin-parity distributions populated through the transfer reactions used are similar to the ones populated through neutron-induced reactions.
ENDFB-Vn Fomnshlrin et al. Fursov et al. JENBL-3,3 JEFF-3.1/A CENBG
ENDFB-Vn • JEFF-3.J/A JENDL-3.3 CENBG
0
CENBG Vorotnitovetal ENDFB-VI1 JENDL-33 JEPF-3.1/A
2 4 6 0 Neutron Energy /MeV
Figure 3. Preliminary fission cross sections as a function of neutron energy in comparison with the available data and the evaluations.
The 243Cm(n,f) cross section is shown in the central part of fig. 3, and our results are compared with the most recent measurements by Fomushkin et al. [8] and by Fursov et al. [9]. At the lowest neutron energies the agreement between the three measurements is rather satisfactory. Beyond 0.7 MeV our data follow fairly well those of Fomushkin [8], but they clearly deviate from the results of Fursov [9]. Concerning the libraries, in contrast to JENDL, which closely follows Fursovs' data, ENDF and JEFF are in rather good agreement with our results. One may wonder whether this discrepancy may be viewed as an indication that the angular momentum induced by the nucleon transfer (3He,d) is much larger than the one induced by low-energy neutron absorption and, thus, that the surrogate method is not applicable in this specific case. However, ref. [5] shows that this effect should lead to an overestimation of the fission cross section. On the other
53
hand, Fursovs' cross section in the 1 to 6 MeV energy range is significantly higher than the experimental cross sections of neighbouring fissile isotopes such as Cm [10,11] and Cm [9] which are below 2 barns. Moreover, under the reasonable assumption that the neutron inelastic scattering cross section of 243 Cm ranges from 1 to 1.5 barn at 2 MeV neutron energy, the value of the fission cross section of 2.6 barn obtained by Fursov et al. at 2 MeV would give a total compound cross section (we neglect the capture contribution to the total compound cross section) varying from 3.6 to 4.1 barns, which is considerably larger than the 3 barns predicted by the optical model calculations by [3]. All these arguments suggest that Fursovs' results overpredict the 243Cm(n,f) cross section at neutron energies larger than 0.7 MeV. 4. Fission-Fragment mass yields Our experimental set-up allows one to measure the fission mass yields of 243,244,245Cm and 242Am jjy means ofme double-kinetic-energy technique [12].
is
at
Figure 4. Preliminary "pseudo-mass" yields of
as 244lM5
tm
im
PS
Cm at different excitation energies.
We have only considered the "pseudo masses"; i.e., the fragment masses uncorrected for prompt neutron emission from the fission fragments. The provisional fission-fragment mass distributions of 244Cm and 245Cm at different excitation energies are presented in figure 4. It is the first time that the mass yields of these actinides have been measured over a wide range of excitation energy extending from a few MeV to about 25 MeV. Note that, at the higher excitation energies, second-chance fission also contributes to the mass distributions. Even though these preliminary results should be considered qualitatively, they already reflect very interesting features such as the increase of symmetric fission with increasing excitation energy, which is caused by the vanishing of shell effects. Calculations will be done to determine the average
54
number of prompt neutrons as a function of fragment mass and to infer fragment masses before and after neutron emission.
5. Conclusions We have presented the first results of a recent experiment to determine the neutron-induced fission cross sections and fission-fragment mass yields using the surrogate reaction technique. The deduced 24’Am(n,f)and 242Cm(n,f)cross sections are in good agreement with the available data obtained via neutroninduced reactions. Our results for the fission cross section of 242Cmextend up to the onset of second-chance fission. None of the existing neutron-induced fission data for 242Cmgoes as high in neutron energy. For the 243Cm(n,f)cross section, our results are clearly below the latest results of Fursov et al. beyond 0.7 MeV 1191. We have determined for the first time the fission fragment “pseudo-mass” yields of 243*244v245Cm and 242Am as a function of excitation energy. The average number of prompt neutrons will be calculated in order to infer the mass distributions before and after neutron emission.
Acknowledgments We thank the tandem accelerator staff and the target laboratory of the IPN Orsay for their great support during the experiment. This work was partly supported by the CNRS programme PACEN/GEDEPEON, the Conseil RBgional d’Aquitaine, the U.S. Department of Energy, Office of Nuclear Physics, under contract DEAC02-06CHII357. The authors are also indebted for the use of 243Amto the Office of Basic Energy Sciences, U.S. Department of Energy, through the transplutonium element production facilities at Oak Ridge National Laboratory.
References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
P. E. Vorotnikov et al., Yadernaya Fizika 40, 1141 (1984). J. D. Cramer and H. C. Britt, Nucl. Sci. Eng. 41, 177 (1970). E. Bauge, private communication. W. Younes and H.C. Britt, Phys. Rev. C 67,024610 (2003). J. Escher et al., Phys. Rev. C 74,054601 (2006). M. Petit et al., Nucl. Phys. A 735,345 (2004). S. Boyer et al., Nucl. Phys. A 775, 175 (2006). E. F. Fomushkin et al., Atomnaya Energiya 69,258 (1990). B. I. Fursov et al., Con$ Nucl. Data for Sci. and Techn., Trieste 1997. E. F. Fomushkin et al., Atomnaya Energiya 63,242 (1987). R. M.White et al., Con$ Nucl. Data for Sci. and Techn., Antwerp 1982. H. W. Schrmtt et al., Phys. Rev. 141, 1146 (1966).
P R O T O N - I N D U C E D F I S S I O N ON A C T I N I D E N U C L E I A T
MEDIUM ENERGY S. ISAEV, R. PRIEELS*, T h . KEUTGEN, J. VAN MOL, Y. EL MASRI F N R S and UCL, Universitt catholique de Louvain, 8-1348 Louvain-la-Neuve, Belgium * E-mail:[email protected]
P. DEMETRIOU Institute of Nuclear Physics, N C S R Demokritos, 153.10 Athens, Greece E-mail: [email protected] 232Th, 237Np, 238U, 239Pu, and 241Am targets were bombarded with proton beam of 26.5 and 62.9 MeV energies. In this contribution some preliminary results are presented and discussed. Comparison with the predictions of the nuclear-reaction model code TALYS is also shown.
Keywords: Fission, Actinides, Neutrons, Multiplicity, Masses
1. I n t r o d u c t i o n Considerable effort has been devoted over the last decades t o feasibility studies of accelerator driven systems (ADS) for transmutation of long-lived radioactive wastes. In view of the increasing concern over the energy problem, novel nuclear techniques combining reactor and accelerator technology are currently of utmost interest for the nuclear industry. At the same time, the new technologies require a more extensive and reliable database of nuclear fission properties. High-energy neutron- and proton-induced fission data on sub-actinide and actinide nuclei are needed for new generation power plants (fast reactors) and ADS applications. On the theoretical front, fission data are also required for a better understanding of the nuclear processes involved and for the improvements of theoretical models and prediction codes. 55
56
2. Experiment
A proton beam of 26.5 and 62.9 MeV (Ep), delivered by the Louvain-laNeuve cyclotron CYCLONE, impinges on actinide targets located in the center of a reaction chamber which is surrounded by 69 large liquid- scintillator cells positioned in a 47r spherical arrangement of 4m diameter, constituting the so-called multidetector DEMON.' This setup allows the determination of the neutron energy and angular distributions and neutron multiplicity spectra associated with the fission process. Figure 1 displays the detector positions inside the reaction chamber. Masses of the fissions
Counters for
MWPC1,Z Large active area X,Y Multi Wire Proportional gas Counters
Fig. 1. Th e experimental set-up, top view. T h e zenith angle 0 runs from 0" t o 180° ( the beam direction corresponds t o 0 = 0), whereas the azimuthal angle CJ runs from Oo to 360" (CJ = 0 corresponds t o a direction in the horizontal plane at 90° left of the beam direction).
fragments (FF) were determined differently by two types of detectors. In each case, the time of flight (TOF) of the FF was used to get velocities. The start time was given by the Cyclotron radio-frequency, the stop times were given by the detector themselves. Fine tunning was applied separately for each detector to take into account of the time needed by the beam burst delivered by the Cyclotron to reach the target plus the electronic delays. The amplitude of these adjustments were determined experimentally in separate runs. First, two microchannel plate-Si diode detector assemblies G J 1 and GJ2 measure in single mode the velocity and the energy of each fragment as shown in the first panel of figure 2. These informations lead to FF masses which are the fragment products after all pre- and post-scission neutron emission. Next, two large active surface (20x20 cm2), X and Y position sensitive Multi-Wire Proportional gas Counters (MWPC), located
57
at mean angles of (6 = 45° with $ = 0°) and (9 = 135° with $ - 180°) on both sides of the beam axis detect in coincidence the FFs emitted backto-back in the center-of-mass frame (c.m.) and give angular distributions and velocities of both FFs with (0.4°) and (500 ps) resolution, respectively. In this case, from the kinematics one also obtains FF masses, but only for fragments produced at the scission time before post-neutron emission. Some typical raw measured spectra are shown in Fig. 2. Binary fission is a back-to-back process in the rest frame. When boosted by a proton projectile, the 180° folding angle between both the produced FFs is reduced. Due to the small proton mass compared to the actinide nuclear masses, the recoil velocity of the initial composite nucleus (CN) is very small and the FFs are emitted quasi back-to-back in the laboratory system (lab) (see the second panel of Fig. 2). In this panel one clearly observes that the transferred linear momentum (LMT) in these reactions is neither null or full. In fact the experimental fission folding angle is apparently in-between these two extreme cases clearly suggesting the occurence of incomplete fusion reaction mechanism and more specifically, of preequilibrium (PE) particle emission.
1 alpha activity
1000
1500
2000
2500
3000 3900 4000 Energy (channels)
40
45
50
55
60
65
70
faff
Fig. 2. First panel: Energy-TOF bidimensional spectrum derived from one microchannel plate-Si diode detector in the experiment p(62.9 MeV) + 239Pu. The start time is initiated by the cyclotron radio-frequency signal. The light charged particle (LCP), the fission fragments (FF) and the time-uncorrelated alpha activity are clearly distinguishable. Second panel: Correlation spectrum between the FF ©i and ©2 emission angles. The dotted line (or full line) corresponds to the hypothetic case of Zero (or Full) LMT of the proton at 62.9 MeV.
3. Analysis and discussion
As shown in Fig. 3, the experimental total fission cross sections (07) agree relatively well with the predictions of TALYS3 at both bombarding energies.
58
The small differences between the af measured at E, = 26.5 and 62.9 MeV and shown in Fig. 4 suggest that af has already reached a saturation value below our lowest E,.
i2.1[ p (62.9 MeV)
r
v
230
232
234
236
238
240
230
232
234
236
238
240
A @t-iu)244
A
f & ~ ) ~ ~ ~
Fig. 3. Experimental (uf)(circles), their error bars and the corresponding TALYS predictions (broken thick curve) as a function of the target mass numbers ( A T ) at both E,. The thin straight line represents the best X2-fit t o the experimental points assuming a function of type [cz.A$’~ 1.1. The Np target has shown an important inhomogeneity, reason why the related uf values (open circles) were not included in the x2 adjustment.
+
Mass distributions of primary FFs were reconstructed for all actinide targets and are plotted in Fig. 5 together with the mass yields obtained from TALYS. They correspond to fragment mass at scission. Mixed (asymmetric and symmetric) distributions are observed at both E, with a clear increase of the symmetric distribution at 62.9 MeV. However, they are not yet fully
59
;v 10
B
Fig. 4. Total fission cross sections for proton-induced fission on 232Th, 238U, 237NP, 239Puand 241Am available in EXFOR d a ta base (black symbol^)^ compared with the newly measured af (red stars) and TALYS calculations at E, = 26.5 and 62.9 MeV using the RFRM5 (open circles often hidden by a red star), ETFS16 (open squares), and RLDM7 (open triangles) fission barrier parameters.
symmetric which suggests that shell effects are not completely washed out even a t E, = 62.9 MeV. The case of 241Amis an exception since the theoretical mass yields turn out to be symmetric even at 26.5 MeV which is a t odds with the experimental results. Such a big contribution of symmetric fission is on the contrary experimentally observed for 232Thbut much less present in the TALYS prediction. More work is under way to understand these discrepancies. Note that the experimental FF mass distributions are subject to uncertainties arising from energy losses in materials, energy and angular stragglings, detector resolutions, and iterative approximations; all
60
these effects inducing a finite mass resolution of several mass units.
Fig. 5. FFs Mass distribution panels arranged in increasing mass number and corresponding to masses of fragments at scission. Upper panel: Experimental cross-section distribution of FFs at 26.5 MeV (full line) and 62.9 MeV (dotted line) proton beams. Distributions corresponding to 26.5 MeV exhibit the strongest asymmetric behaviour. Lower panel: Mass yields of FFs obtained with TALYS at 26.5 MeV (full line) and 62.9 MeV (dotted line) incident energy.
Neutron yields and energy spectra were measured in the 4vr geometry in coincidence with the MWPCs. The results of typical moving-source fits to the neutron kinetic energy and angular distribution spectra for 241Am target at different spatial angles and at both Ep assuming PE, CN and FF1 (fragments crossing MWPC1) and FF2 (fragments crossing MWPC2) contributing sources are shown in figure 6. The neutron contributions from the FFs dominate at around the angles of the detected FFs i.e. (0 = 45°, $ = 0°) and (9 = 135°, $ = 180°). This can be seen from the angular distribution in the reaction plane ($ = 0°) and from the contributions to the energy spectra where the neutrons emitted from FFs crossing the MWPC1 (FF1) dominates at Qn = 50° but are the weakest at 0n = 130°. At Ep = 62.9 MeV the PE is clearly present at forward angles (© = 50°) but absent at backward angle (0 = 130°). The weak contribution of the PE neutrons, even-though their energy shapes are not perfectly reproduced in the fits, does not significantly affect the quality of the general fit and of the neutron multiplicity results.
61
Fig. 6. Left: Some neutron energy spectra at 62.9 MeV for the 241Am target at different 0, and q5n angles adjusted by four moving neutron-emitting sources. T h e F F 2 contributions merge with the total in the 8, = 130° spectra, while FFl's merge with the 0, = 50° ones. Right: Typical neutron angular distribution fits for the same target at E, = 62.9 MeV for different 4, angles (in- and out-of the reaction plane. For graphic purpose the range of 0 runs here from -180° t o 180° taking a single value for a. T h e thin lines represent the neutron contribution of each source.
Fig. 7. Left: Neutron multiplicity versus the mass number of the five actinide targets at the two E,. T h e four lower lines represent the partial contributions of the moving emitting sources: vPE(triangles), v C N (full squares), vzlz(open circle) , vEz$(open squares. (For clarity the experimental errors are only shown on the d at a related t o the total neutron multiplicities (upper line with error bars). T h e thick upper curve without error bars is the TALYS expectation for the total multiplicity of neutron emission. Right: Experimental and corresponding RMS values and calculated mean values of FFs T K E following TALYS code (full line) and Viola's systematics (dotted line) are compared at 62.9 MeV beam energy.
The experimental values of different mean neutron multiplicities, i.e. u P E , u C N l uflg (us","m' f u&L2), u,Pi$,, and d o t a l a t the two E, are shown in Fig. 7. Only utotal is compared with the TALYS predictions obtained assuming Sierk's fission barrier^.^ One observes that, the PE contribution
62
remains relatively small, while the CN contribution is quasi-independent of AT and significantly increases (by a factor M 3) with increasing E,. The post-scission multiplicities weakly increase with AT and are slightly higher a t E, = 26.5 than a t 62.9 MeV but they remain quasi-identical whenever symmetric or asymmetric FF mass partitions are selected. The small reduction of the post-scission multiplicities at 62.5 MeV can be explained by the decrease of the CN mass which has emitted more neutrons before reaching its scission stage. The total neutron multiplicities (around 4.7 at 26 MeV) and (about 6.8 a t 62.9 MeV) are in agreement with TALYS calculations within the error bars. Also shown in Fig. 7 are the experimental and TALYS predictions for the total kinetic energy (TKE) mean values as a function of AT a t the two E,. The agreement is excellent even with the values obtained from the Viola et al. systematics.8 The analysis is still in progress, so all the results presented herein must be considered as preliminary.
References 1. T h . Keutgen et al., Phys. Rev. C 70, 014611 (2004). 2. S. Isaev et al., Phys. Rev. C , to be submitted, (2007). 3. A.J. Koning, S Hilaire, M.C. Duijvestijn, (AIP Conference Proceedings 769, 2004) p. 1154. 4. EXFOR database, (2005), www.nds-iaea.org. 5. A. J . Sierk,Phys. Rev. C 33,2039 (1986). 6. A. Mamdouh and J. M. Pearson and M. Rayet and F. Tondeur, Nucl. Phys. A 679, 337 (2001). 7. S. Cohen and F. Plasil and W. J. Swiatecki, Ann. of Phys. 82, 557 (1974). 8. V. E. Viola, K. Kwiatkowski, and M. Walker, Phys. Rev. C 31, 1550 (1985).
FISSION CROSS SECTIONS OF MINOR ACTINIDES AND APPLICATION IN TRANSMUTATION STUDIES A. LETOURNEAU~, 0. BRINGER, s. CHABOD, E. DUPONT, G. FIONI’, F. MARIE, S. PANEBIANCO, CH. VEYSSIERE DSWDAPNIA, CEA - Saclay Gif-sur-Yvette. 91 191, France
L. ORIOL DEN/DER/SPEX, CEA - Cadarache Saint-Paul-lez-Durance, 13108, France F. CHARTIER DEN/DANS/DPC/SECR/LANIE, CEA - Saclay GiJsur-Yvette. 91 191, France P. MUTT1 Institut Laue Langevin Grenoble, 38000, France I. ALMAHAMID Wadsworth Center, New York State Department of Health Albany, NY12201, USA
Fission cross sections of Minor Actinides are of great importance for the reduction of long-term nuclear waste radiotoxicity by transmutation. In this paper we present the results of measurements done on the fission cross sections of three minor actinides: and 245Cm,in the thermal energy range. These cross sections participate 238Np,242gs-mAm significantly to the incineration of 237Np, 24’Amand 244Cmisotopes and show some discrepancies with nuclear data libraries or previous experiments.
Corresponding author: [email protected] * Address : DSM/DR, CEA - Saclay, Gif-sur-Yvette, 91 191, France
63
64
1. Introduction Facing with the world demand in energy, in a near future, and the constraint to limit green-house gases emission, the development of nuclear energy seems to be unavoidable in this context. But in returns, nuclear industry has to demonstrate its ability to fulfill some requirements, such as safety and sustainability of resources as defined within the Generation IV forum. Before this transition to Gen-IV reactors, Pressurized Water Reactors are still in life for at least 40 years with, certainly, an increase in their capacities to better burn their fuel, but also with a production of minor actinides (MA) that have to be treated. In this context, it is clear that reliable nuclear data on MA are needed both in thermal and fast energy ranges. In particular fission cross sections, which are the main parameters for nuclear energy production, are of prime importance as far as uncertainties on their value could act directly on the efficiency of the incineration process and thus on the economy of dedicated incinerators. In this paper, we present recent results we have obtained on the fission cross sections of isotopes produced by double capture during the transmutation of the most abundant MA contains in a spent fuel. These cross sections were measured in thermal energy neutron fluxes at the High Flux Reactor (HFR) of the LaueLangevin Institut (Grenoble - France). 2. Impact of fission cross section uncertainties on incineration First of all, we have tried to quantify the impact of nuclear data and their uncertainties on transmutation process. For this, we have propagated nuclear data uncertainties into the calculation of quantities, such as radiotoxicity, related to the transmutation of Np and Am targets, in different reactor concepts: EFR, GT-MHR and HTFR [1].
Figure 1. Contributions to the fission rates resulting from the incineration of one gram of M7Np,
65
From our study, it turns out that uncertainties on fission cross sections have deep impacts on both the economy of the incineration process, i.e. the efficiency to incinerate, and on the production of alpha and/or neutron-emitters. The efficiency directly depends on the fission rate which is higher for thermal reactors than fast ones (see Figure 1). For instance, it takes about 9 years to incinerate about 20% of 237Npin the EFR reactor (high flux, fast neutrons), whereas only 50 days are needed in the HTFR concept (high flux, thermal neutrons). The induced uncertainty on the incineration time due to the 237Np(n,f) and 238P~(n,f) reactions is about 8% in the fast reactor. However, in thermal spectra these two cross sections do not contribute very much. But the 238Np(n,f) reaction induces an uncertainty amounting to 3%. On the other hand, the production of alpha and neutron-emitters could be strongly reduced by fission process which stops the mass flow towards heavier masses. It is the case for the fission of 238Npwhich contribute significantly to the reduction of 244Cmformation in the HTFR concept as compared to the GTMHR concept for which both the neutron flux and average fission cross sections are reduced (harder energy spectra). 3. Experiments at ILL 3.1. Experimental approach
Fission cross sections are measured using a new detector concept we have developed, called Triple Deposit Fission Chambers (TDFC) [2]. It is composed of three electrically independent micrometric ionisation chambers (i.e. fission chambers), mechanically connected and sharing the same gas (see Figure 2b). Each fission chamber is 2 cm long, has a diameter of 4 mm, and is filled with pure argon gas. Space charge effects due to high fission rates (high fluxes) are considerably reduced thanks to the thin gap between anode and cathode. Their functioning in current mode is detailed in reference [2]. This concept of detector allow to measure on-line the actinide fission current with respect to the fission current of the 235U(n,f)standard reaction. Parasitic currents induced by y-rays and neutron activation are measured thanks to the third chamber without deposit and subtracted. For irradiation, we use the V4 channel which is located about 10 cm from the fuel element, giving access to neutron fluxes as high as 1.5 lOI5 n/cm2/s. Moreover, due to the inclination of the tube (approximately 8') with respect to the vertical axis, different energy spectra could be used (see Figure 2a).
66
In such high fluxes, short-lived isotopes (as 238Np or 242Am) can be formed and studied on-line thanks to the developed detectors. They are produced by double capture from their fertile isotope. In that case the measured fission current is proportional to both the fission cross section of the fissile isotope and also on the neutron capture cross section of the fertile one. The latter cross section is measured independently by activation techniques: either by alphaand/or gamma-spectroscopy in the H9 channel [3] or by mass spectrometry (Thermal lonisation and Inductively Coupled Plasma Mass Spectrometry).
V4T ¥•4 i
Figure 2. a) (left) Neutron energy spectra in the,H9 and V4 channels, as obtained from a very detailed simulation of the HFR core with MCNP. The different V4 curves correspond to different irradiation positions where the numbers indicate the distance from the median plane corresponding to "0 cm" position. At this position about 85% of neutrons have less than 1 eV. b) (right) View of a Triple Deposit Fission Chamber (TDFC).
3.2. Analysis tools High fluxes are very useful to explore transmutation chains, i.e. to form and study short-lived isotopes for which targets are impossible to make. On the other hand, the evolution of MA is rather complex and experimental results can be difficult to interpret, especially for long-time irradiations. In this context, we have developed an evolution code called MERCS [4] and used it both for analysis (fit of nuclear parameters to experimental data) and to optimise irradiation conditions (suitable neutron flux and irradiation time). MERCS is a one-group evolution code, based on ROOT shared libraries, which solves numerically Bateman equations. Its originality lies within its capacity to compute absolute and relative sensitivity tables, which reflect the sensitivity of an experimental observable (i.e. fission chamber current) to nuclear parameters.
67
3.3. Integral measurements A global procedure is applied to extract the cross section value at Eo= 25.3 meV !?om the measured effective one ( omeasured ): N
where a(E) is the cross section value as a function of the neutron energy and @(E) the neutron energy distribution. In principle, ME) is taken !?om evaluated nuclear data libraries and @(E) has been calculated with a complete MCNP simulation of the core reactor. This procedure is accurate when resonances are well described or if they do not contribute too much to the total reaction rate. Nevertheless, errors on resonance parameters are taken into account and propagated into the final error. Just like for uncertainties on the neutron flux distribution which are fully propagated through the extraction procedure. In Table 1 are given the effective cross sections obtained for the standard reactions used to normalise the neutron flux. Table 1. Effective cross sections (in barns) of the two normalisation reactions and thermal-over-total neutron flux ratios in different irradiation positions.
Positions 25 cm 50 cm 75 cm 100 cm H9
5 9 ~ ~ ( n60, co y) 28.33 30.74 3 1.66 3 1.84 31.10
235
U(n,f) 424.68 466.83 484.17 487.29 473.93
RthRtot 0.898 0.966 0.997 1.ooo 0.98 1
4. Fission measurements 4.1. z38Np(n& measurement
The 238Np(n,f)cross section was measured by double capture starting from pure 237 Np target. TDFC, containing (42*1.3) pg of 237Npand (2.64*0.01) pg of 235U, was irradiated for 43 days in the V4 channel, 25 cm above the median plane (a 9 lOI4 n.cm-2.s-1,about 90% of thermal neutrons). Measured currents are shown in Figure 3 together with the fit analysis done with MERCS code.
-
68
In this analysis we have used, as input parameter, the 237Np(n,y)238Np cross section that we have measured separately in the H9 channel. The isotopic concentration of 237Nptarget (13.8fO.l pg) was determined before and after 2 hours of irradiation by CL and y-spectroscopy using the Mini-INCA chamber. The resulting cross sections are given in Table 2.
Figure 3 . Evolution of the measured fission current (symbols) of the TDFC containing 237Npas a function of time. The results of the fit and the contributions of fissile isotopes are also indicated (lines).
The 238Np(n,f)cross section is in good agreement with data libraries and with the value obtained by Abramovitch et al. (21 10*740) [5] but the value of Danon et al. (2638h58 b) [6] could be rejected. Table 2 . Effective and 25.3 meV measured cross sections.
Reactions 237Np(n,y)238Np 238NP(n,f)
(b)
0 0 (b)
150 f 4.2 1828 h 55
180 5 2165 h 70
*
4.2. 24ZgSAm(n,J and Z42mAm(n,fl measurements
The 242gSAm(n,f) and 242mAm(n,f) cross sections were also determined by double capture starting from a pure 241Am target and using fission chambers containing (103.1*03) pg of 241Am and (38.8*0.2) pg of 235U.The detectors were irradiated
69
-
for 44 days at the upper position of the V4 channel (@ 6 l O I 3 n.cm-2.s-',pure thermal neutrons). The isomeric branching ratio (i.b.r) and the 241Am(n,y)242Am reaction cross section were measured by precise mass spectrometry of several 241Amsamples irradiated in thermal fluxes (H9 and T4 channels) [7]. The samples were 97% in purity of 241Am,with 2.5% of 237Npand with some plutonium traces [7]. Two sets were irradiated for 11 and 24 days in the T4 channel, whereas four other sets were irradiated in the H9 channel for 1, 5, 11 and 19 days. After a cooling time of about 1 year, samples were analysed by thermal ionization mass spectrometry analysis (TIMS). The resulting cross sections are tabulated in Table 3. Table 3. Effective and 25.3 meV measured cross sections.
Reactions 241~(n,y)242~ i.b.r 2 4 2 ~ ~ ( 242m~(n,~ 3f)
(b)
609 f 27 0.8947f0.004 2224 f 215 6527 f 225
(b) 704 f 32 0.8947k0.004 2644 f 281 6856 f 656 0 0
The 242mAmfission cross section is compatible with data library values and with previous measurements [8-111, whereas the 242gsAm shows differences of about 22% with data libraries. It has to be noticed that it is the first measurement for this quantity. 4.3. 2 4 i ~ m ( measurements n~ The 245Cm(n,f)cross section was measured during the start-up phase of the reactor (see Figure 4) using upgraded TDFC containing (26.41t0.44) pg of 244Cm,(1.539f0.027) pg of 245Cm,and (4.22f0.004) pg of 235U.The detector was placed 25 cm above the median plane in the V4 channel. Thanks to the high fission cross section of 245Cmas compare to the 244Cmone, the fission current was dominated by the 245Cm(n,f)reaction at the beginning of irradiation. The resulting effective cross section is (1352k23) b and the 25 meV value is (1943f65) b. This value is 10% less than evaluated ones and than the most recent measurement done by Browne et al. (2143*58) b [12], but in good agreement with the oldest ones: Gavrilov et al. (1 900*100) b [ 131 and Benjamin et al. (2018k37) b [14].
70
Figure 4. Measured 245Cm and 235U fission currents as a function of irradiation time (left) and reactor power (right) during the reactor start-up phase.
Acknowledgments
The authors are grateful to the ILL personnel for cooperation and assistance. This work is partially supported by the GDR GEDEPEON (France). References 1. O. Bringer et al., Ph.D. thesis, INP Grenoble (2007); O. Bringer et al., Accepted for publication in Annals of Nuclear Energy (2007). 2. S. Chabod , Ph.D. thesis, University of Paris XI (2006) ; S. Chabod et al., Nuclear Instrument and Methods A566, 633 (2006). 3. F. Marie et al., Nuclear Instrument and Methods A556, 547 (2006). 4. http://www-dapnia.cea.fr/Sphn/MNM/Modelisation/ 5. S. Abramovich et al., Proceedings of the!3* Meeting on Physics of Nuclear Fission, Obninsk 303 (1995). 6. Y. Danon et al, Nuclear Science and Engineering 124, 482 (1996). 7. G. Fioni et al., Nuclear Physics A693, 546 (2001). 8. T. Kai et al, Annals of Nuclear Energy 28, 723 (2001). 9. J.C. Browne et al, Physical Review C29, 2188 (1984). 10. J.W.T. Dabbs et al., Nuclear Science and Engineering 84, 1 (1993). 11. C.D. Bowman et al., Physical Review 166, 1219 (1968). 12. J.C. Browne et al., Nuclear Science and Engineering 65, 166 (1978). 13. V.D. Gavrilov et al., Atomic Energy 41, 808 (1976). 14. R.W. Benjamin et al., Nuclear Science and Engineering 47, 203 (1972).
SYSTEMATICS ON EVEN-ODD EFFECTS IN FISSION FRAGMENTS YIELDS: COMPARISON BETWEEN SYMMETRIC AND ASYMMETRIC SPLITS' F. REJMUND, M. CAAMANOt GANIL, CNRYIN2P3. CEA/DSM, bd H. Becquerel Caen, 14076, France Even-odd effect in fission fragment yields is investigated for symmetric and asymmetric splits of different fissioning systems. Data in inverse kinematics obtained at GSI gave for the first time the complete elemental yields over a broad range of fissioning systems. These data indicate that the even-odd effect at symmetry is independent of the fissility of the fissioning system. New data in inverse kinematics are required to conclude on this topic that impacts strongly our understanding on fission dynamics.
1. Introduction Even-odd staggering in the fission fragment yield has been observed since the beginning of experiments investigating fission properties at low excitation energy. It has been reported with magnitude as important as 40% [l]. Its quantitative description within nuclear models has been somehow more puzzling and is still controversial. Historically, a statistical description of the even-odd staggering based on the level density of the final states (in the fission fragments) has been discarded, as the increased Q value for even splits is counterbalanced by the pairing gap in the level density. It was then quickly understood that the amplitude of the even-odd staggering in fragment yields is related to the energy dissipation along the fission path and the probability for a certain number of quasi-particle excitations at scission [2]. Indeed, the experimental observation of odd-Z fragments in the fission of and even element is the evidence for the existence of unpaired nucleons that may then split into two different fragments. From this general picture, the following properties lead to the relatively well-established rules:
Work partially supported by grant 448 16 of the EURATOM program of FP6.
71
72
Even-odd effect for neutron numbers in fission fragments has always been observed lower than for proton number; indicative of a smaller dissipation for neutrons than protons Even-odd effect is larger for asymmetric splits; associated to more dissipation in symmetric split Even-odd effect is decreasing with the fissioning nucleus fissility; reflecting an increasing path from saddle to scission when the fissility increases Even-odd effect decreases with increasing kinetic energy of the fission fragments; signifying less energy available for dissipation when kinetic energy increases. In recent experiments at GSI, it became possible using inverse kinematics to measure the full elemental yields (light, symmetric and heavy fragment) for electromagnetic-induced fission of long isotopic chains of actinides and preactinides [3]. Even-odd effects have been observed, and, even though smaller than observed in neutron-induced fission, following the same trend of decrease with fissility of the fissioning system. However, an important even-odd effect has been observed in odd-Z fissioning nuclei [4], forcing to revisit of the general understanding of even-odd staggering. Indeed, from a general understanding, once a nucleon is unpaired, as it is always the case in odd element fissioning system, there should not be any even-odd effect. Steinhauser et al. in [4] showed that the phase space available in the fission fragment has an influence on the choice of the unpaired particle, that will preferentially choose the heavy fragment. With the same analysis, it was possible to reproduce the even-odd effect in even-Z fissioning nuclei, showing a larger even-odd effect for asymmetric than for symmetric splits. It has been then proposed that only the even-odd effect at symmetry pictures the dissipation from saddle to scission [5]. Indeed, at symmetry the probability to share the unpaired particle is independent of the final fragment mass and reflects the probability that no pair is broken at scission. The probability that the fissioning system remains in a completely paired proton configuration has been derived considering the probability that the dissipated energy may be restricted to only neutron excitations, based on a rigorous description of the level density for quasi particle excitations at scission. Following this description, it has been possible to reproduce for the first time the difference between even-odd effect in proton and in neutron number: as the level density for neutron excitations is larger than the level density for proton excitation, the probability to break a pair of proton is smaller than for a pair of neutrons.
73
With the advent of new experimental techniques, the classical picture on even-odd effect origin and its link with fission dynamics has been revisited, and at least point 1 and point 2 of the above list may be explained by other means than differences in dissipation for proton and neutrons or symmetric and asymmetric splits. In the present contribution, we want to show that the same data tend to infirm the assumption 3 concerning the amount of dissipated energy and the fissility of the nucleus. 2. Inverse kinematics data
Figure 1 shows the even-odd effect observed in electromagnetic-induced fission of 230-234 U, 2'9-229Th, 219-Ra and Rn isotopes as a function of the fissility of the nucleus. The even-odd effect is compared to the even-odd effect observed in neutron-induced fission of heavier actinides. For Th and U isotopes, the same decrease trend with fissility is observed, coherent with the picture of an increasing path from saddle to scission when the fissility increases. The amplitude of the even-odd effect is hindered in GSI experiment, indicating a higher excitation energy gained in Coulomb-induced fission compared to neutron-capture. Global even-odd effect for Ra and Rn isotopes is somehow different, and the expected decrease with fissility is not observed. Indeed, compared with the trend of Th and U isotopes, the even-odd effect shows a constant value around 5%. It has been shown in [ 3 ] that the fission fragment distribution of Ra and Rn isotopes is symmetric, when it shows a double-hump structure of U and Th isotopes.
20
__
-5
1L 3F.L
4
3 5 6 358
360 362 3 6 L 388 368 3 7 0
Z2/A
Figure 1. Left part: Even-odd effect in electromagnetic fission (full symbols) of Th and U isotopes as a function of fissility. Open symbols show the even-odd effect measured in neutron-induced fission. Lines are to guide the eye. Right part: even-odd effect in electromagnetic fission of Ra and Rn isotopes, compared to the line guiding U,Th. From [Steinhauser] with permission.
14
Figure 1 suggests then that the even-odd effect measured in symmetric splits remains independent of the fissility of the nucleus. This assumption is enforced by Figure 2 that shows the local even-odd effect for symmetry and asymmetric splits of Th fissioning isotopes. The local even-odd effect is deduced using the recommandation of [ 6 ] . When a clear increase of the evenodd effect is observed with increasing number of neutrons of the fissioning Th (corresponding to decreasing fissility), the even-odd effect at symmetry remains constant between 5 and lo%, without any clear tendency.
30 75
1 'Thoriih t
1
I
I
I
I
I
I
As ymme trie I
T
131
132
I
I
I
I
I
I
I
133
131
135
136
137
138
139
NCN
Figure 2. Local even-odd effect at symmetric and asymmetric splits (2=45 and 2=54) for Th isotopes. From [Steinhauser] with permission.
3. Even-odd effect at symmetry for neutron induced fission
Figure 3 shows the elemental distribution of fission fragments measured in the neutron-induced fission of actinides from Th to Cf. The local-even-odd effect is displayed for each of these distributions. If a clear decrease of the even-odd effect for asymmetric splits is observed as the fissility Z2/A of the nucleus increases, it is difficult to conclude for symmetric scission. Indeed, in these data obtained at Lohengrin spectrometer [7,8,9,10,11], the element yields close to symmetric scission are difficult to observe due to the limits in the detection system. In addition, the local even-odd effect is averaged over 3 consecutive elements in order not to be sensitive to any local shell effect in the distribution [ 6 ] . As a consequence, the local even-odd effect is measured mainly in the asymmetric part of the distribution. Figure 4 summarizes the results of Figure 3: the global even-odd effect is plotted as a function of the fissility of the fissioning system. In comparison is plotted the local even-odd effect for asymmetric split, which is taken as the value of the local even-odd effect for
75
Z=ZCN-54, ZCN being the atomic number of the fissioning nucleus. The local even-odd effect for the
Figure 3. Left: Element yields in neutron-induced fission for different fissioning systems. Right: local even-odd effect deduced following [Tracy]
«HB
far
«
MSf-
IT
8S.S
S8
3&S
S?
j*«
Figure 4. Global even-odd effect (square) as a function of fissility. For comparison are shown the loca! even-odd effect for asymmetric scission (circles), and for most symmetric experimentally available scission (triangle). If triangles do not appear they fall behin a circles.See text for details.
76
most symmetric scission configuration is also plotted for comparison. If the even-odd effect at asymmetry shows the same trend as the global even-odd effect, the situation is less clear for the even-odd effect at symmetry. A decrease is observed, but much less pronounced than for asymmetry. Again, the absence of measure at symmetric scission prevents from further conclusion. 4. Perspectives
Data in inverse kinematics for which the complete element yields could be measured along a wide range of actinides and pre-actinides put forward that the even-odd staggering is independent of fissility for symmetric scission. This result is somehow puzzling, as it is known that the gain in potential energy from saddle to scission is increasing with fissility. Considering the even-odd effect at symmetry as the probability for the fissioning nucleus to remain completely paired at scission, the constancy in even-odd staggering reveals a constant dissipated energy for proton-pair breaking in the deformation process from saddle to scission. The rest of the increasing available energy may be removed by neutron evaporation, that shows systematically a more energetic spectrum and higher multiplicities for symmetric fission [ 121. Before any further speculative results may be reached, the data of GSI need to be confirmed by other measurements with a precise estimation of the excitation energy as well as a complete scan of the element yields (from light to heavy). Projects aiming at measuring the fission fragments in inverse kinematics such as ELISE at FAIR [ 131 or transfer-induced fission at GANIL [ 141 would be most favorable. In GANIL, it is proposed to use a uranium beam to induce multinucleon transfer fission on a C beam. The multi-nucleon process will populate a broad range of neutron-rich actinides from U to Cm that will be completely determined in mass, charge and excitation energy, using a Silicon telescope. The in-flight fission decay will be detected using the high resolution and large acceptance spectrometer VAMOS, which allows for an isotopic identification of the fission fragments before their beta-decay. High precision data will be obtained on the evolution of the pairing correlation with the excitation energy, and shell effects in the heavy fragments as a function of the fissioning system on a wide range of neutron-rich actinides. References 1. 2. 3. 4.
N. Boucheneb et al., Nucl. Phys. A502,261c (1989). J.-P. Bocquet, R. Brissot, Nucl. Phys. A502,213c (1989). K.-H. Schmidt et al., Nucl. Phys. A 665, 221(2000). S. Steinhauser et al., Nucl. Phys. A 634,89(1998).
77 5 . F. Rejmund, A. V. Ignatyuk, A.R. Junghans, K.-H. Schmidt, Nucl. Phys. A 678,215(2000). 6 . B. L. Tracy et al, Phys. Rev. C 5,222 (1972). 7. M. Djebara et al., Nucl.Phys. A 425, 120(1984). 8. W. Lang et al., Nucl.Phys. A 345,34(1980). 9. U. Quade et al., Nucl. Phys. A 487, 1 (1988). 10. C. Schmitt et al., Nucl.Phys. A 430,21(1984). 11. D. Rochman et al., Nucl. Phys. A 710,3(2002). 12. H.-H. Knitter, U. Brosa, C. Budtz-Jorgensen, The Fission Process, CRC
PRESS. 13. J. Taieb et al, these proceedings. 14. F. Rejmund et al., Proceedings of the 3rd Int. Workshop on Nuclear Fission and Fission-Products Spectroscopy, AIP 2005.
This page intentionally left blank
MEASUREMENT OF KINETIC ENERGY DISTRIBUTIONS, MASS AND ISOTOPIC YIELDS IN THE HEAVY FISSION PRODUCTS REGION AT LOHENGRIN A. BAIL*, 0. SEROT and 0. LITAIZE CEA Cadarache F-13108 Saint Paul-Lez-Durance, France *E-mail: [email protected]
H.R. FAUST, U. KOSTER and T. MATERNA Institut Laue-Langevin B.P. 156, F-38048 Grenoble, France
A. LETOURNEAU and E. DUPONT DAPNIA, CEA Saclay F-91191 Gif-sur- Yvette, France
Mass yields and kinetic energy distribution functions for heavy mass fission products from thermal neutron induced fission of 235U and 239Puhave been measured at the spectrometer Lohengrin at the high flux reactor of the Institut Laue-Langevin in Grenoble. In addition to these measurements where an ionization chamber was used for the mass identification, we also performed gamma spectrometry to quantify the isotopic and isomeric yields. This setup using Ge-detectors has been commissioned with the system 241Pu(nth,f). In order t o extend the data to less abundant fission products, a proportional counter for beta detection has been constructed, allowing to reduce the background by beta-gamma coincidences.
1. Introduction Even if nuclear fission was discovered almost seventy years ago, complementary data are still needed to improve our understanding of the process. Data already listed in data bases are not always very accurate, and the list of the possible reactions is not exhaustive. Moreover, considerable differences can be found between the various evaluated data libraries (fig. l),in particular between ENDF, JEFF and JENDL evaluations. To improve the efficiency of nuclear reactor operation, cumulative yields of fission products 79
80 0 09
.
f
008
1
EFFlJ IFF22 ENbFB68
-3ENDL33
0 07 0 06
ZOO5 .n
:004
‘I
I
i T
0 03
t
120
125
130
135
140
145
150
155
160
hkSA
Fig. 1. Heavy mass yields for the 239P~(nth,f) reaction from the main evaluated data libraries.
affecting reactivity need to be known with better accuracy for the minor actinides, but also for the most known actinides. Likewise the fission models need more and better nuclear data to be compared with [1,6]. For systems which undergo fission in a thermal neutron flux, detailed results for kinetic energy, nuclear mass, and nuclear charge have been measured on Lohengrin, in particular for 233U, 235U, 238Np, 239Np, 241Pu, 245Cm and 249Cf [7,12]. However, only the light fission yields for these actinides have been investigated (apart from the system 245Cm(nth,f)).We present here a study for fission product characteristics in the heavy mass region for 235U(nth,f), 23gP~( nt,f) h and 241Pu(nth ,f). 2. The Lohengrin mass spectrometer
The Lohengrin recoil-mass spectrometer is a nuclear facility which uses low-energy fission reactions for fragment production (fig. 2). It allows to study fragment distributions from thermal neutron induced fission with a very high resolution. The beam intensity at the separator allows also detection of gamma-rays, conversion electrons, beta-rays, delayed neutrons, and coincidences between these particles. The Lohengrin fission source is typically a thin layer of a fissile material placed close to the core of ILL’Sresearch reactor, in a high thermal-neutron flux of 5.3.1014 neutron/cm2/s. Fission products emerging from the target are selected by a combination
81
Fig. 2. Schematic view of the Lohengrin mass spectrometer.
of a magnetic and an electric sector field, whose deflections are perpendicular to each other. Both sector fields have focusing properties in their plane of deflection. Energetic fission products are leaving the target with high ionic charge states. The combined action of the two fields separates ions with the same velocity into different parabolas at the exit slit of the spectrometer, according to their (A/q) and (E^/q) values. Here A, q and E^ are mass, ionic charge state and kinetic energy of the ions, respectively. The energy dispersion in the direction along each parabola amounts to 7.2 cm for 1% difference in energy, and the mass dispersion perpendicular to each parabola is 3.24 cm for 1% mass difference. A focusing magnet (RED magnet in fig. 2) has been installed at the exit slit of the spectrometer, which increases the particle density by a factor of 7. The flight-path for the fission products is 23 m, and the separation time is of the order of 2 fj,sec, so that fission products reach the detector before undergoing betadecay. For spectroscopy work on neutron rich nuclei different equipments may be installed next to the exit slit of Lohengrin such as ionization chambers, surface barrier detectors, Si and Ge detectors, plastic scintillators and counters for delayed neutrons. Also a fast tape transport is available for different experiments. 3. Mass yields and kinetic energy distributions The thermal fission mass yields of most of the compound systems have been measured with Lohengrin in the light fission product region (fig. 3), but no heavy mass yields have been investigated so far, except for 245Cm(nt/,,f) for which the average kinetic energy distribution is available in [13]. The aim of the present work was the determination of yields and kinetic energies for heavy fission fragments. The system of detection was an
82
Fig. 3. Absolute mass yields from the neutron-induced fission of different nuclei: Data from the mass separator Lohengrin (connected symbols) completed with the ENDF/B-VI nuclear-library data (lines), figure taken from [14].
ionization chamber placed at the exit slit of the spectrometer. It allows to determine the kinetic energy Efc value of the separated fragments, giving thus the mass A and ionic charge state q values (fig. 4). The ionization
(GO
5GO
525
540
5SO
KtO
KO
S2a
I (channels)
Fig. 4. Example of acquisition for the mass yields of 239 Pu(n t ^ ,f) measurement. A clear separation between the different A/q values can be seen.
chamber is filled with isobutane at p~40 mbar pressure. The high voltage
83
applied on the anode and on the F'risch grid is 8OOV and 400V, respectively. The cathode is on ground. More details on the ionization chamber can be found in [15].The integration over kinetic energy and ionic charge distributions for every mass provides complete mass yields of the studied actinide. Two neutron induced fission reactions were investigated: measurements have been done on a set of mass for the determination of mass yields in 235U(njh,f)in order to validate the method, the heavy mass region for 239Pu(nth,f)were investigated. The fig. 5 shows our results of the heavy mass yields from 2 3 g P ~ ( n t h , f ) which are compared with the one from the JEFF3.1 library. Non negligible
"'I
129
134
119
144
h
119
A
) JEFF 3.1 (gray line) and Lohengrin (black line), Fig. 5. Mass yields of 2 3 9 P ~ ( n t h , ffor preliminary result. The incertities are only statisticals.
differences can be observed. In particular, more pronounced structures can be seen which could be due to the reminiscence of the odd-even effect after emission of prompt neutron. Kinetic energy properties (average and variance) are displayed on fig. 6 for 235U(nth,f)and 239Pu(nth,f).A good agreement is observed with the literature [16]. Some investigations concerning the explanation of the kinetic energy distributions shape have been performed. They aim at the understanding of the asymetric shape (compared to a pure gaussian function) for the
84
II
Avcmgr kinetic mew fa f39Pu(4.f)
I
Fig. 6. Average kinetic energy (top left) and variance (top right) from fission of 2 3 9 P ~ ( n t h , fand ) average kinetic energy (bottom left) and variance (bottom right) from fission of 235U(nth,f), preliminary result. T h e incertities are only statisticals.
kinetic energy distributions. Several hypotheses were advanced to explain this asymetry:
0
0
0
due to background from contaminations with fissile material inside the Lohengrin beam tube: background measurements without a target show that this contamination is negligible. due to the energy loss when traversing the target and the cover foil: the simulations carried out with TRIM [17] are not able to reproduce such a shape. due to the contribution of different isotopes present inside a mass line: an experiment including isotopic identification (see below) will be undertaken in the near future. due to the reaction mechanism (different fission modes [6]).
4. Isotopic yields
For low mass fission products with 2 1 4 2 the isotopic components within one mass line are determined by measuring the specific energy loss in a
85
split anode ionization chamber. This method is no longer applicable for fragments with nuclear charges Z higher than 42. Therefore we tried to determine isotopic yields by gamma-spectrometry on 241Pu(nth,f).As the beta-decays of fission products are often followed by gamma de-excitation, and because these decays occur after their fligh through the spectrometer, beta-gamma concidences can be used to determine isotopic yields. For this purpose the fission products are implanted in a moving tape which is traversing the ionization chamber at the focus point of the spectrometer. Two germanium clover detectors are used to measure the gamma-decay with high efficiency, whereas the beta emitted particle is detected by a proportional gas counter or a scintillator. An example of a gamma-spectrum without gama-beta coincidences obtained for mass A=134 is given in fig. 7. The different isotopes contributing for this mass are Sb, Te and I, and their characteristic gamma-decay lines are clearly visible. We can also see the contaminant of the background: 214Pb and '14Bi, and 1311 and 138Xe (diffusing through the beam line). From the respective strenghts of the gamma transitions and knowing the mass yields, the isotopic yields will be deduced.
134 51Sb 134 5llS
1400 '35Xe
1200
/
'
134 5ZTe
134 53I 134 53mI
'
'34Te
1000
>
y 800
a
h 600
+ u)
s 400 200
0
200
300
400
500
600
700 A.134,
800 900 9'23. E.80 MeV
Energy (keV)
Fig. 7. Example of ungated gamma-spectrum for isotopic yields determination from 241~~(nth,f).
86 5 . Conclusion and perspectives Kinetic energy distributions and m a s yields for 235U(nth,f) and 2 3 g P ~ ( n t,f) h have been evaluated and d a t a treatment for isotopic yields of 2 4 1 P ~ ( n t h , fis) in progress. T h e studies t o reproduce t h e kinetic energy distribution shape are under way.
References 1. H. Goutte et al., Mass and kinetic energy distributions of fission fragments using the time dependent generator coordinate method, Nucl. Phys. A734 (2004) 17. 2. H. Goutte et al., Microscopic approach of fission dynamics applied to fragment kinetic energy and mass distributions in 238U, Phys. Rev. C71 (2005) 024316. 3. S. Oberstedt et al., Fission-mode calculations for 239U, a revision of the multimodal random neck-rupture model, Nucl. Phys. A644 (1998) 289. 4. F.-J. Hambsch et al., Prediction of fission mass-yield distributions based on cross-section calculations, Ann. of Nucl. Ener. 32 (2005) 1297. 5. H.R. Faust, A model for fragment excitation and kinetic energy in nuclear fission, Eur. Phys. J. A14 (2002) 459. 6. P. Moller et al., Nuclear fission modes and fragment mass assymetries in a five dimensional deformation space, Nat. 409 (2001) 785. 7. U. Quade et al., Nuclide yields of light fission products from thermal neutron induced fission of 233U at different kinetic energies, Nucl. Phys. A487 (1988) 1. 8. W. Lang et al., Nuclear charge and mass yields for 235U(nth,f) as a function of the kinetic energy of the fission products, Nucl. Phys. A345 (1980) 34. 9. I. Tsekhanovich et al., Mass and charge distributions in the very asymmetric mass region of the neutron induced fission of 238Np, Nucl. Phys. A688 (2001) 633. 10. D. Rochman et al., Isotopic yields from the reaction 245Cm(nth,f) a t the Lohengrin mass separator, Nucl. Phys. A710 (2002) 3. 11. T. Friedrichs et al., Investigations of mass, charge, and energy of thermal neutron induced fission of 245Cm and 241Pu, in Proceedings of the Second International Workshop on Nuclear Fission and Fission-Product Spectroscopy (AIP Conference Proceedings 447, Seyssins, 1998, France) p. 231. 12. M. Djebara et al., Mass and nuclear charge yields for 24gCf(nth,f)at different fission-product kinetic energies, Nucl. Phys. A496 (1989) 346. 13. B. Weiss et al., Kinetic energy distributions in thermal neutron induced fission of 245Cm, in Proceedings of Nuclear fission and fission-product spectroscopy (AIP Conference Proceedings 798, Cadarache, 2005, France) p. 232. 14. http://www.ill.fr/nfp/npp/Pnl/Physics.htm. 15. A. Bail, Ph.D. Thesis, CEA Cadarache, 0. Skrot et al., in Proceedings of ND2OO7 (Nice, 2007, France) in presse. 16. C. Wagemans, The nuclear fission process, CRC Press, Boca Raton, 1991; 17. http://www.srim.org.
Ternary Fission
This page intentionally left blank
ON THE TERNARY a SPECTRUM IN 252Cf(sf)
M. MUTTERER W . N . KOPATCH 3 , S.R. YAMALEDTINOV V.G. LYAPIN J. VON KALBEN ', S.V. KHLEBNIKOV 2 , 5 , M. SILLANPAA 2 , G.P. T W R J N W.H. TRZASKA 'i2*,
214t,
215,
234
'Institute of Nuclear Physics, Univ. of Technology, Darmstadt, Germany; Department of Physics, University of Jyvaskyla, Jyvaskyla, Finland; Joint Institute for Nuclear Research, Dubna, Russia; Helsinki Institute of Physics, Helsinki, Finland; V.G. Khlopin Radium Institute, St. Petersburg, Russia
We have dedicated a new experiment t o remeasuring the ternary a-particle energy spectrum in 252Cf spontaneous fission using a n array of unshielded silicon detectors and unambiguously discriminating a-particles from neighboring isotopes by time-of-flight (TOF) techniques using fission fragments as the start. Particle detectors were placed at right angle to fission fragment direction, so mainly equatorial particles are registered. Compared t o many previous A E E experiments which feature detection thresholds at 6 to 9 MeV, the energy distribution of ternary &-particles could, for the first time, be measured down t o 1 MeV. Furthermore, the energy spectrum of 6He could be analyzed, albeit with weak statistics, and the yield ratio 6He/4He was deduced. For both, 4He and 6He, an excess in the yield as compared t o a Gaussian shape is observed at energies below 9 MeV. Corrections t o the measured a-spectrum were made for both, the wider than detected angular interval of equatorial particles ~ 130 ) and the full span of angles, including polar emission, (50 I: O a 5 eventually measured in experiments without fragment coincidences. For this purpose, we determined our constraint in detection angle by Monte-Carlo and employed d at a on the energy dependent angular distribution of the a-particles obtained by P. Heeg with DIOGENES. T h e emission angle was found t o affect mainly the width of the energy distribution, by up t o 1 MeV. We compare our results with literature data, and discuss prospects for future measurements. Keywords: Ternary fission; 2 5 2 C f ( ~ f )E; distributions of ternary 4He and 6He; T O F - E method. PACS: 24.75.+i, 25.85. -w
'corresponding author, e-mail: [email protected] t deceased
89
2,
90
1. Introduction
Since the discovery of ternary fission in the forties of last century there have been numerous experiments devoted to the energy distribution of the ternary a particles (see, e.g. reviews1i2). There is common agreement that the shape of the spectrum closely resembles a Gaussian but shows some yield in excess of at low energy.3 The yield at low-energy is particularly significant for probing initial a-particle energies and the main fragments’ configuration at scission. However, particle-unstable ternary particles (e.g., 5He and 8Be) may also give rise to low-energy a’s in sequential p r o c e ~ s e s ,~ and thus mask the true ternary particle emission. Without any doubt, precise experimental data are very much needed for addressing the problem further. Most experiments performed up to now were not optimized for the study of the lowest particle energies. In fact, A E E detector telescopes were usually applied for particle discrimination, and shielding foils were often placed in front of the AE detectors for preventing radiation damage by fission fragments. Thus, the spectral distributions measured have low-energy thresholds not below 6 to 9 MeV a energy (e.g., Wagemans et al.3). These rather high threshold values compared to the 16 MeV mean energy do not only cut away the interesting low-energy part of the spectrum but leave also substantial ambiguity in the energy assignment of the above threshold events due to uncertainty in absorber and AE detector thicknesses and related energy losses. The only measurement with unshielded energy detectors up to now has been the TOF-E measurement by Tishchenko et al.5 with the 4n Berlin Silicon Ball. Here, the detection threshold was pushed down to 2.0 to 2.5 MeV, but separation of a particles, tritons and 6He was only weak due t o the equal flight paths of 10 cm for fission fragments (FF) and ternary particles inside the ball. There have been also attempts to measure low-energy ternary a’s by non-electronic methods e.g., a mass spectroscopic measurement after using P b catcher foils for the reaction 235U(nth.f),6and solid state nuclear track detectors (SSNTD) for 252Cf(sf).7While in Ref. 6 a massive “short-range” component below 7.7 MeV was stated, no intensity in excess of a Gaussian shape was found in Ref. 7 in this region. We have dedicated a new counter experiment to remeasuring the ternary a-particle spectrum from 252Cf fission by improved time-of-flight (TOF) techniques with unshielded silicon detectors placed a t a distance of about 20 cm to the source and registering the FFs with a channel plate detector (CP) at a close distance of about 2.5 cm.
91
~MS?
;;
$'
Fig. 1. Experimental set-up for measuring the ternary o-particle spectrum from 252Cf fission by the time-of-flight-E method. The assembly of 252Cf source, channel plate start detector and fragment energy detectors (roof detectors) is seen at the left-hand side. Ternary particle detectors facing the open side of the sample are placed at 20 cm distance. Data from the sideward located detectors triplets have not been included in the present analysis.
2. Layout of Experiment
A photo of the experimental setup is shown in Fig. 1. The thin 252Cf sample, produced at the Radium Institute at St. Petersburg, had an activity of 500 fissions/sec. It was prepared by self-transfer method onto a 22 /zg/cm2 aluminum oxide support backing with a 10 /zg/cm2 evaporated layer of gold. Particles were measured with an array of 10 silicon p-i-n diodes (380 /um thickness, and 30 x 30 mm2 area) placed at a distance of w 20 cm from the source. Proper diaphragms promote FF background in the energy region of the ternary particles. With dedicated preamplifiers and low-noise timing filter amplifiers in the timing channels the energy threshold could safely be reduced to < 0.5 MeV. This is the lowest cut-off value ever achieved in a ternary fission experiment. As the start signals, fission fragments emitted from the source were registered in a 30 mm diameter channel plate (CP) detector placed at right angle to the particle detectors, at 2.5 cm distance to the source. Discrimination of fission fragment start signals from the 30 times more frequent 6 MeV alphas was achieved by coincident registration of the complementary fragments in 10 silicon p-i-n diodes of 20 x 20 mm2 area (roof detectors) mounted in a semicircular configuration at a distance of 6 cm to the sample, opposite to the CP. Data were collected over a period of about 6 weeks, with no significant deterioration of detector performance due to radiation damage. Energy calibration of the silicon detectors was
92
performed with alpha lines from a spectroscopic thin 226Rasource and a BNC PB5 precision pulser, to a precision of 5 50 keV (fwhm). We would like to note that, in the present experiment, there is not any material between the 252Cf source and the surface of the detectors. The silicon detectors in use have aluminium front windows of nominally 140 nm thickness. The corresponding effective dead-layer was determined with angular dependent cy spectroscopy t o be 369(11) nm of silicon equivalent thickness,12 which results in an energy correction of 110 keV for 1 MeV a particles, and 30 keV a t 10 MeV. 3. Data Analysis
Only ternary fission events with FF pairs registered in both, C P and roof detectors, have been analyzed. The measured difference in the fragment flight times to C P and roof detectors, respectively, vs. fission fragment energy EFF registered in the roof detectors was used to correct the measured TOF spectra of ternary particles for the difference in flight time between heavy and light fragment masses from source to CP. The resulting particle TOF vs. E pattern is shown in Fig. 2. The intense bunch in the centre corresponds to ternary alphas, and the weaker bunches of the neighboring isotopes 3H and 6He below and above the cx particle distribution, are nicely separated from it. The three bunches in the upper left corner are identified as 27Al, l60and 12C scattered off from the source backing or the roof detector surface by fission fragments. Between these groups and the ternary 6He a few events from heavier ternary particles, mainly 8He and "Be, are visible. It has to be noted that the TOF-E pattern in Fig. 2 is particularly clean of background, although FFs are hitting the detectors with M 300 times higher rate. The vertical line at 6.1 MeV represents random coincidences with the lo4 times more frequent a particles from 252Cfradioactive decay which are of low enough probability to be safely subtracted in the time window of the ternary a particle distribution. It is interesting to see also a small 6.1 MeV peak about 2 ns above the pattern for the ternary a's which is attributed to start signals from x-rays or conversion electrons in the CP when the 252Cfradioactive decay proceeds through the excited state of 248Cm.This peak falls accidentally into the TOF-E pattern of ternary 6He ruling out an analysis of the 6He spectrum in a small energy gap around 6 MeV. At the high-energy side the ternary a particles spectrum is cut off at 27.5 MeV due to the limited detector thickness of 380 pm. Since the cut-off takes place at a yield level of M 3 % relative to the maximum yield at 16 MeV there is only minor influence of it on the high-energy half of the
93
spectrum. For ternary 3H, the TOF-E pattern reaches its highest energy value due to the detector thickness already at 11.5 MeV (see Fig. 2) bending back at higher 3H energies and interfering with the respective pattern for the protons.
252
Cf (Sf)
41) 6,1 MeV alphas, random coincidences
ternary alphas
I
•20
10
15 K (Met)
20
Fig. 2. Scatter plot TOP vs. E of ternary particles in 252 Cf(sf), as measured with 10 silicon p-i-n diodes of 380 p,m thickness and 30 x 30 mm 2 in size, located at 20 cm distance from the source. Time is in ns with respect to the flight time of 16 MeV a particles; energy E is in MeV.
Figure 3 (left) is our measured energy distribution of ternary a particles, deduced by converting the TOF-E pattern of Fig. 2 into a mass (A) vs. energy representation and subsequent integration over the A = 4 line. Discrimination from the 3H and 6He lines was assured by inspecting the mass spectra related to narrow energy intervals of 0.5 MeV. The energy spectrum of ternary a particles covers the wide energy range from 1 to 27.5 MeV. It has to be noted that having the particle detectors at right angles to the direction of emission of the fission fragments does bias the experiment to detect mainly ternary particles emitted at the instant of scission, i.e. the so-called equatorial particles. Strictly speaking, any measured ternary a spectrum depends a little on the margin of QaL angles (between the ternary particle and the light group of fission fragment) covered experimentally due to the well known increase of the angular width of equatorial a particles with energy and the onset of the polar a par-
94
ticles at energies above about 20 MeV (e.g., Ref. 9). Thus any constraint in the a particle emission angle @,L causes part of the equatorial yield a t higher energy to be suppressed compared to the yield a t lower energy. On the other hand, if ternary particle spectra were measured without fragment coincidences, as was, e.g., the case in Ref. 3, the equatorial yield is completely covered for all energies, but the a spectrum is slightly enhanced a t high energy compared to the spectrum of equatorial a's due to the rare but mainly higher energetic polar ones.
40 400
30
.s.
_i
r:
C
20
20u
10
0
I
E MeV)
Fig. 3. Left-hand side: Measured energy distribution of ternary a particles from 252Cf fission. Right-hand side: Energy distribution of ternary 6He particles. Solid lines are Gaussian curves fitted to the data above 9 MeV.
The issue has been studied with the aid of a simulation calculation, ~ E, measured previously by P. Heeg et a1.loJ1 with using data on 0 , vs. DIOGENES. For this purpose, we determined our constraint in detection angle by Monte-Carlo. Our detector geometry shows to register a-particles with @,L angles from 55 ' to 130 ' . The efficiency function has a FWHM of 33 ' around a mean at 92 . Folding this angular dependent registration efficiency with the a-particle angular distribution determined by P. Heeg (Figure 5.3 at page 66 in Ref. 11) we have obtained the registration efficiency in our setup vs. E,. Finally, corrections to the measured a-spectrum were made for two cases: (a) the wider than detected angular interval of equatorial particles, best defined as 50 5 @,L 5 130 according to the DIOGENES data, and (b) the full span of angles, including polar emission, eventually measured in experiments without fragment coincidences (e.g., Ref. 3) and in near 47r coincidence experiments (e.g., Refs. 4,5).
95
4. Results and Discussion
With reference to providing a benchmark ternary a spectrum in 252Cf,we are listing in Tab. 1 mean energies and widths of Gaussians fitted above 9 MeV energy to the spectra corrected for the equatorial range 50 5 Oar, 5 130 , and the full 47r emission angle, respectively. The latter spectrum has been fitted for energies above both, 9 MeV and 12.5 MeV. It is obvious that taking the emission angle into account mainly affects the width of the energy distribution, by up t o 1 MeV, while the mean energy changes up to 0.3 MeV only. Finally, the spectral parameters presented in Tab. 1 have been compared with literature data. Within rather small experimental errors, our data with the full-angle correction applied compare favorably with data from Refs. 3 and 5 at a fitting threshold at 12.5 MeV, and with data from Ref. 4 a t the threshold at 9 MeV. On the other hand, the somewhat narrower spectral width obtained for a coincidence experiment by Grachev et aL15 is in good agreement with the present data being corrected for equatorial emission. So, the cited literature data tend to confirm the slight dependence of the spectral shape of ternary Q particles on emission angle, which has been analyzed, to our knowledge, for the first time in the present work. As to the low energy part of the a spectrum, a comparison with the result by Tishchenko et aL5 shows good agreement between 2.5 and 9 MeV (Fig. 4, right) while the data reported as early as 1974 by Loveland' overestimate the low-energy yield of ternary a particles. l 3 Shown in the left plot of Fig. 4 are also two Gaussian curves which result from a fit, above 9 MeV, to the present data corrected for equatorial emission, considering besides true ternary Q particles (the dominant Gaussian) also the about 17% contribution of residual a particles from the decay of ternary 'He, recently measured by Kopatch et al.4 It is obvious Fig. 4, that the ternary Q spectrum shows more low-energy Q particles than would be predicted by the two-Gaussian fit. Apparently, the spectral shape measured p r e v i o u ~ l y at ~,~ energies E > 9 MeV can not be extrapolated meaningful to low energies. We have finally also extract the energy spectrum of ternary 6He from our data shown in Fig. 2, leaving out the energy region around 6 MeV. The ternary 6He spectrum is plotted on the right-hand side in Fig. 3, the total number of events collected over the 6 weeks period of the measurement being 468. This is a rate of about 10 events per day. To our knowledge it is the first time that ternary 6He particles from 252Cf(sf)were measured over their full energy range. Because of the low statistics involved, we have passed
96 Table 1. Spectral parameters of the ternary a-spectrum in '"Cf Mean energy, MeV
FWHM, MeV
Gaussian fit range, MeV
Angular range for emL
Method (detectors)
Reference present work
15.4f0.1 15.5k0.1 15.6fO.l 15.7f0.1
10.OfO.l 10.5*0.1 10.9f0.2 10.6f0.2
9-27 9-27 9-27 12.5-27
experimenta equatorialb full full
TOF-E (MCP-silicon)
15.7f0.2
10.4f0.2
2 12.5
full
15.7f0.2
10.9f0.1
8-28
full
15.7f0.1
10.6
2 10
full
15.8f0.1
10.2f0.1
8-28
equatorial
AE-E (silicon-silicon) AEE (gas-silicon) TOF-E (silicon ball) AEE (gas-silicon)
Note: a for the current experimental setup.
~agernans~ Kopatch4 Tishchenko5 Grachev15
defined as within the range 50 ' 5 f 3 m<_~ 130 ' .
E (MeV)
Fig. 4. Left-hand side: Energy distribution of ternary a particles from 252Cf fission, corrected for equatorial angles 50 ' 5 Oar, 5 130 ' . Two Gaussian curves were fitted to the data above 9 MeV, taking residual a particles from 5He decay into account, according to the results of Kopatch et al. [4](see text). Right-hand side: Energy distribution of ternary a particles from 252Cf fission, corrected for full emission angles, in comparison with data by Tishchenko a t al. [5](open squares).
on without applying any correction for the emission angle. The spectrum turns out to be asymmetric as well, although some uncertainty remains here about a minor interference with some background in the analysis window. Summing up the spectrum shown in Fig. 3 (right), with interpolating
97
the missing values around 6 MeV, and relating it t o the sum of a particles shown in the left-hand side of Fig. 3, has yielded for the ratio 6He/4He a value of 0.049(3). Fitting the spectrum, for energies above 9 MeV, with a single Gaussian curves yields 12.5(5) MeV for the mean energy and 9.0(5) MeV FWHM for the width. Taking the area under the Gaussian as the estimate for the 6He yield and relating it t o the a yield fitted also above 9 MeV gives for the 6He/4He ratio a value of 0.041(5), which is in line with most values deduced earlier from experiments with similar threshold energy (e.g., Refs. 14-16). The analysis of the spectrum from the rare 6He particles gives us confidence that an essential interference of fragment background with the about 25 times more intense ternary a spectrum can safely be neglected.
5 . Summary and O u t l o o k
The present work has demonstrated that ternary a particle energy spectra can be precisely measured over the full energy range by the TOF-E method using unshielded silicon detectors. The present study validates the low-energy tailing and settles the issue experimentally, both in a- and 6Heaccompanied fission. More sophisticated measurements are needed to decide whether low-energy tailing is a common issue to most ternary particle species, and is thus possibly decoding particularities of the nuclear configuration at scission, or rather a special feature of ternary fission accompanied by He isotopes. Currently, a new experiment has been projected, with an about 10 times stronger '"Cf source and improved fragment trigger using the sample as the conversion foil of a micro channel plate (MCP) start detector. This will improve time resolution and statistical accuracy, and permit measuring the dependence of the spectral shape on particle emission angle. We also want to scrutinize whether combination of TOF with pulse-shape discrimination techniques in suitable reverse-mounted silicon surface barrier detectors17 can be applied for the discrimination of ternary particles according to both, their mass and nuclear charge. This would permit registration of ternary particle spectra up to carbon isotopes. Application of the present technique for neutron induced fission reactions where interference with cy particles from radioactive decay is generally less, e.g. in 235U(nth,f),is also envisaged.
98
Acknowledgments T h e present work was supported by t h e EU Integrating Infrastructure Initiative, Contract No. 506065 (EURONS), and by t h e Academy of Finland under t h e Finnish Centre of Excellence Programme 2006 - 2011, and by t h e INTAS Grant No. 03-51-6417. One of us (M.M.) wants to thank t h e Academy of Finland for a research grant. Fruitful discussions with F. Gonnenwein and C. Wagemans are gratefully acknowledged.
References 1. C. Wagemans, in The Nuclear Fission Process, ed. C. Wagemans (CRC Press, Boca Raton, F1. USA, 1991), Chap. 12. 2. M. Mutterer, and J. Theobald, in Nuclear Decay Modes, ed. D.N. Poenaru (IOP, Bristol, UK, 1996), Chap. 12. 3. C. Wagemans, J. Heyse, P. Jansen, 0. Serot, and P. Geltenbort, Nucl. Phys. A 742, 291 (2004). 4. Yu.N. Kopatch, M. Mutterer, D. Schwalrn, P. Thirolf, and F. Gonnenwein, Phys. Rev. C 65, 044614 (2002). 5. V.G. Tishchenko, U. Jahnke, C.-M. Herbach and D. Hilscher, Report HMI-B 588, Nov. 2002. 6. G. Kugler, and W.B. Clarke, Phys. Rev. C 5, 551 (1972). 7. H. Afarideh, K. Randle, and S.A. Durrani, Ann. Nucl. Energy 15, 201 (1988); also: Intern. Journ. of Radiation Applications and Instrumentation D 15, 323 (1988). 8. W. Loveland, Phys. Rev. C 9, 395 (1974). 9. F. Gonnenwein, M. Mutterer, and Yu. Kopatch, Europhysics News 36/1, 11 (2005). 10. P. Heeg, J. Pannicke, M. Mutterer, P. Schall, J.P. Theobald, K. Weingartner, K.H. Hoffmann, K. Scheele, P. Zoller, G. Barreau, B. Leroux, and F. Gonnenwein, Nucl. Znstr. Meth. in Phys. Research A 278, 452 (1989). 11. P. Heeg, Ph.D. Thesis, TU Darmstadt, 1990. 12. A. Spieler, Diploma Thesis, T U Darmstadt, 1992, unpublished. 13. M. Mutterer, Yu.N. Kopatch, S. Yamaletdinov, V. Lyapin, 3. von Kalben, S. Khlebnikov, M. Sillanpaa, G. Tjurin, W.H. Trzaska, Proc. Intern. Con!. on Dynamical Aspects of Nuclear Fission, Smolenice Castle, Slovak Republic, Oct. 2006, (World Scientific, Singapore, 2007), in press. 14. Z. Dlouhi, J. Svanda, R. Bayer, and I. Wilhelm, Proc. Znt. Conf. on Fifty Years Research in Nuclear Fission (Berlin, 1989), Report HMI-B 464, p. 43. 15. V. Grachev, Y. Gusev, and D. Seliverstov, Sou. J. Nucl. Phys. 47,622 (1988). 16. G.M. Raisbeck and T.D. Thomas, Phys. Rev. 172, 1272 (1968). 17. M. Sillanpaa, Master Thesis, Univ. Jyviiskyla, Finland, 2007; M. Sillanpaa, W. H. Trzaska,M. Mutterer, G. Tyurin, Y. Kopatch, S. Smirnov, S. Khlebnikov, J. v. Kalben, Proc. Seminar on Fission VZ, Corsendonk Priory, Belgium, Sept. 2007, (World Scientific, Singapore), this issue.
ENERGY DEGRADER TECHNIQUE FOR LIGHT-CHARGED PARTICLE SPECTROSCOPY AT LOHENGRIN A. OBERSTEDT* Department of Natural Sciences, Orebro Universitet, 5-70182 Orebro, Sweden *E-mail: [email protected]
S. OBERSTEDT EC- J R C Institute for Reference Materials and Measurements ( I R M M ) , B-2440 Geel, Belgium previously: Institut Laue-Langevin, 6 rue Horowitz, F-38042 Grenoble, France D. ROCHMAN Brookhaven National Laboratory, National Nuclear Data Center, Upton, N Y 11973-5000, USA previously: Institut Laue-Langevin, 6 rue Horowitz, F-38042 Grenoble, France Although the recoil mass-separator LOHENGRIN at Institute Laue-Langevin was originally designed for the spectrometry of binary fission fragments, it was also used in the past for measuring light-charged particles from ternary fission. However, due to limited electric field settings the energy distribution of the lightest particles was not completely accessible. In this contribution we report on an energy degrader technique that allows the measurement of the entire energy spectra of ternary particles with LOHENGRIN. We demonstrate how the measured particle spectra are distorted by the energy degrader and present results from a Monte Carlo simulation that shows how the original energy distributions are reconstructed. Finally, we apply this procedure to experimental data of ternary particles from the reaction 235U(nth,f ) . Keywords: Energy degrader; Mass spectrometer; light-charged particles; 235U(nth, f); excited ternary sLi*.
1. Introduction The unambiguous identification of light charged particles (LCP), e.g. produced in nuclear fission, is a necessary prerequisite for the full understanding of the nature of their formation. Solid state detectors, assembled as 99
AE-E telescope, may provide a separation of atomic numbers, while timeof-flight measurements may give information of the particles’ mass. A severe disadvantage, however, is the fact that particles with low kinetic energies are stopped in the AE-detector. Moreover, since the solid state detectors have t o be protected from the intense flux of binary fission fragments, thin absorber foils have to be installed in front of the detectors. Also there lowenergetic particles are stopped and, thus, remain unidentified.’ As an alternative, a mass-separator in combination with an ionization chamber with split anode, the latter serving as AE-E telescope as available at Institute Laue-Langevin (ILL)’ in Grenoble (France), may be employed. There the recoil mass-separator LOHENGRIN3 was originally designed for binary fission fragments produced at the high flux reactor, where the fission products are separated by electric and magnetic fields. However, the maximum electric field settings of the LOHENGRIN spectrometer correspond t o a limited measurable kinetic energy according to
E = 5.5MeV x q,
(1)
where E stands for the kinetic energy with which the particle passes the mass spectrometer and q denotes the particle’s ionic charge. Although appropriate for binary fission fragments, LOHENGRIN was in the past also used for the spectroscopy of light-charged particle^.^-^ However, in this case the high energy tail of the energy distribution sometimes remained undetected. Examples for energy spectra obtained with both methods are shown in Ref. 8. In a recent paperg we presented a technique that allows t o adapt the fission fragment mass spectrometer LOHENGRIN to the detection of ternary particles over practically the entire kinetic energy range. In order t o achieve that, we suggested t o apply an energy degrader to reduce the initially high kinetic energies to those in compliance with Eq. (1).Below we describe briefly the influence of the energy degrader on the measured particle spectra and show results from a Monte Carlo simulation that shows how the original energy distributions are reconstructed. We also present results on ternary fission products from the reaction 235U(nth,f ) measured with an energy degrader and address the question whether ternary particles may be emitted in an excited state. 2. The energy degrader In order to slow down high energetic light ions prior t o entering the mass spectrometer, we suggested to put a degrader foil (e.g. nickel) of appropriate thickness into the particle beam between target and spectrometer
101 300
J
250 200 150
100
50 0 0
5
10
15
20
25
30
E (MeV) Fig. 1. Result of a Monte-Carlo simulation of particles passing an energy degrader. The original energy distribution is shown as full drawn line and full black circles. The open diamonds denote the corresponding yield distribution obtained by calculating the residual energies for each event. The open circles show the same yields, but now plotted as a function of incident energy, whereas the recreated original distribution is shown as full grey circles.
entrance. From the choice of energy, corresponding to the spectrometer’s field settings for a given ionic charge state and the properties of the degrader material, the kinetic energy, with which the particles were emitted in the fission process, can be calculated according to their energy loss in the degrader foil. The use of an energy degrader affects the particle beam in two ways: for one, the particles will have a reduced kinetic energy before they enter the spectrometer; secondly, due t o the constant energy acceptance of the mass separator b E / E , less particles pass the separator because of their lower kinetic energy. As a consequence, we expect to observe a reduced yield at a reduced energy in the spectrum. Figure 1 shows the result of a Monte Carlo simulation for a typical LCP energy distribution. We have chosen a Gaussian with the following characteristics: an average energy of 15 MeV, a width of about 11 MeV (FWHM), and 1000 particles. The full drawn line shows the energy distribution and the full black circles the randomly obtained yield, i.e. number of particles per energy unit, for selected energies. For the energy loss we assume that events with E < 12 MeV are stopped, while those with E = 30 MeV leave the degrader foil with E,,, = 24 MeV. According to results of energy loss calculations,” the residual energy for incident energies between 12 and 30 MeV may be
102
obtained by linear extrapolation. If one calculates now the residual energies behind the degrader for each event and determines their number per MeV, one obtains the distribution shown as open diamonds in Fig. 1. Moving these points to their correct, i.e. initial, energies results in the open circles. Obviously, the yields are underestimated, which is due to the fact, that the spectrometer lets through fewer particles with reduced energy than it would be the case without energy degrader. This effect can be corrected for by multiplying these yields with the energy ratio E/E,,, at a given energy E. However, the transformation from E,,, to E in the plot leads to an effective binsize that is less than 1 MeV. Hence, the yields have t o be multiplied as well with the derivative (dE/dE,,,), which is less than 1. Applying both corrections and binning around the same energies as indicated by the black dots, leads finally t o the original energy distribution, now shown as grey dots. We observe a very nice agreement, of course only for events with energies large enough to pass the degrader foil. For a detailed description we refer t o Ref. 9.
+
3. Ternary fission from the system 23sU nth
High resolution measurements of ternary particles produced in the thermal neutron-induced fission of 236U*were performed at the high-flux reactor a t ILL (an=5 . 4 ~ 1 0 cm-' ~ ~ s-l). The target consisted of 235U02 of 500 pg/cm' thickness and a size of 7.0x0.5 cm', coated with a 0.25 pm thick Ni-cover to minimize sputtering of target material into the separator. In order to extend the kinetic energy of the LCPs measurable with the massseparator LOHENGFUN, we introduced a 10 pm thick Ni-foil as energy degrader. The measured particle yields as function of particle energy were corrected for burn-up, ionic charge (when necessary) and energy loss. In case LCPs were measured parasitically, field corrections were applied too. A detailed description of the data treatment is given in Ref. 4, however some facts are mentioned here as they demonstrate the value of the degrader technique. The target burn-up was monitored with ''Be4+ without energy degrader and with 6He2+ both without and with energy degrader a t their most probable energies, i.e. E = 18 MeV and 11 MeV,ll respectively. The latter one corresponds to a residual kinetic energy E,,, = 7.26 MeV after the energy degrader." A sum of two Exponentials was used to describe the measured burn-up, taking both neutron irradiation and additional material loss during heating into account (see Fig. 2). The half-lives, Tl/z, of the fast and slow component were 0.9 and 8.8 days, respectively. As shown in Fig. 2,
103 1000 900 800
,L-
700 600
7-
>, z
'
500 400
9
F
300
200 0
5
1 0
1 5
Time (d) Fig. 2. Measured burn-up curve of the 235Utarget. T h e different symbols correspond t o different LCPs, for which the observed yields (with and/or without) were normalized t o each other. T h e fuel drawn line represents the result of the fit with two Exponentials. The experimentally found normalization factors are interpreted in the text.
the ratio between the yields of 6He and l0Be a t their maximum energies was found to be 10.54. Calculating this ratio from published total yields and widths" gives 9.71, which is in good agreement with our findings. The ratio between the yields of 6He without and with energy degrader can be estimated according to section 2 as
@ / E m ) x (dE,es/dE) = 1.34,
(2)
while the experiment gives a value of 10.54/8.51 7z 1.24. Also here the agreement is nice. So far energy spectra of different ternary He, Li, Be and C isotopes have been analyzed, where the energy degrader was employed in all measurements except for the carbon isotopes. Figure 3 shows the measured energy distributions of 7Li and 'Li. The Li-ions were measured for charge states q = 3+ and 2+, depicted as open circles and open squares, respectively. The full circles show the total yield, integrated over all ionic charge states. Since the contributions of the charge states O+ and 1+ are negligible at the measured energies, the total yields were determined by adding those for 2+ and 3+. All data obtained without energy degrader are shown in black, the ones taken with energy degrader are depicted in grey. The thick, full drawn line corresponds to a Gaussian fitted to all experimentally obtained yields. The full drawn and dotted lines
104
|
E (MeV)
E {MeV)
Fig. 3. Experimental energy distributions for 7'8Li, measured with and without energy degrader (grey and black symbols, respectively). The open squares correspond to 7 ' 8 Li 2+ , the open circles to 7 ' 8 Li 3+ , and the full circles to the sum of all charge states. The thick, full drawn lines correspond to Gaussians fitted to all experimental data, while the full drawn and the dotted lines indicate the calculated 3+ and 2+ contributions, respectively.
indicate the calculated12 contributions of 7>8Li3+ and 7'8Li2+, respectively. The error bars contain statistical errors and, in case the energy degrader was used, systematic uncertainties from the different corrections applied. In the latter case an uncertainty in energy was estimated from the linear fit to the E-Ere,, dependence calculated with SRIM.10 The excellent agreement between data taken with and without energy degrader is obvious (cf. Fig. 3), in particular in the overlap region around 15 MeV, makes us confident in the presented energy degrader technique. Further LCP energy spectra taken with this technique are shown elsewhere.8 For all evaluated energy spectra their most probable energies, widths and yields were determined. They are all in excellent agreement with previously obtained results. This is depicted - for the yields - in Fig. 4, where results from this work (symbols) are compared to corresponding ones from Ref. 11 (lines). From the measured 7'8Li spectra we then made an attempt to investigate, whether the emission of excited, neutron-unstable ternary 8Li (Ex = 2.26 MeV) as observed in spontaneous fission of 252Cf13 may be corroborated in the system 236U*. Due to its short lifetime (r = 2 x 10~20s), this excited state of 8Li is detected as 7Li, but with different kinematic properties compared to directly emitted 7Li. Making the - reasonable - assumption that 8Li has the same mean energy and width, regardless whether in ground or excited state, the corresponding properties for the decay product 7 Li may be calculated. This results in an average energy E^ = (11.78±0.13) MeV, in contrast to E = (14.68 ± 0.15) MeV for the measured spectrum. The latter spectrum must then consist of two components: 7Li, either emitted or from the 8Li* decay. If the fraction of 8Li at Ex = 2.26 MeV here was
105
102
10' h
0 0 Q)
100
I L
0
.-0)
1 0-1
1o 2
1 0-3
10-4
0
5
10
15
20
A Fig. 4. Experimental yields for different He, Li, Be and C isotopes obtained in this work (symbols) together with corresponding values from Ref. 11 (lines). f82 = 0.25 like in " ' C f ( ~ f ) ~and ~ taking into account that the measured yield for 'Li amounts about 52% of the one for 7Li, 13% of the measured 7Li nuclei would result from the 8Li* decay. This, in turn, would result in an asymmetric energy distribution with an enhanced low-energy tail, which however is not observed (cf. Fig. 3). Calculating different sLi contributions in the 7Li spectrum by varying f82 and subtracting them from the measured 7Li spectrum gives the contribution from emitted ternary 7Li. Fitting a Gaussian to the difference spectrum and minimizing the X2-value, did not give any indication for the emission of excited 8Li either.
4. Summary and conclusion
In this paper we have described an energy degrader technique dedicated t o increase the kinematical limits of mass-separators by reducing the kinetic energy of an entering particle beam. We have also presented high resolution LCP spectroscopy results from the reaction 235U(nth, f ) with the
106 recoil mass-separator LOHENGRIN, where for the first time this technique was applied successfully. T h e measured LCP properties were in excellent agreement with previous results. T h e emission of neutron-unstable LCPs in ternary fission was searched for in t h e case of 8Li*, but no evidence was found at t h e present state of data treatment. Obviously, further investigations of this claimed phenomen seem t o be necessary.
References 1. Yu N. Kopatch, V. Tishchenko, M. Speransky, M. Mutterer, F. Gonnenwein, P. Jesinger, A. M. Gagarski, J. von Kalben, I. Kojouharov, E. Lubkiewics, Z. Mezentseva, V. Nezvishevsky, G. A. Petrov, H. Schaffner, H. Scharma, W. H. Trzaska, and H.-J. Wollersheim, in: H. Goutte, H. Faust, G. Fioni, and D. Goutte (Eds.), Nucl. Fission and Fission Spectroscopy: Third Int. Workshop, in: AIP Conf. Proc., vol. 798, 2005, p. 115. 2. J. P. Bocquet, R. Brissot, and H. R. Faust, Nucl. Inst. Methods A 267 (1988) 466. 3. E. Moll, H. Schrader, G. Siegert, H. Hammers, M. Asghar, J. P. Bocquet, P. Armbruster, H. Ewald, H. Wollnik, Kerntechnik 8 (1977) 374. 4. S. Oberstedt, A. Oberstedt, D. Rochman, F. Gonnenwein, I. Tsekhanovich, J. Becker, A. Sartz, H. Bax, F.-J. Hambsch, and S. Raman, Nucl. Phys. A 761 (2005) 173. 5. I. Tsekhanovich, Z. Buyukmumcu, M. Davi, H. 0. Denschlag, F. Gonnenwein, and S. F. Boulyga, Phys. Rev. C 67 (2003) 034610. 6. U. Koster, H. Faust, G. Fioni, T. Friedrichs, M. GroB, S. Oberstedt, Nucl. Phys. A 652 (1999) 371. 7. M. Wostheinrich, R. Pfister, F. Gonnenwein, H.O. Denschlag, H. Faust, S. Oberstedt, in: G. Fioni, H. Faust, F.-J. Hambsch, S. Oberstedt (Eds.), Nucl. Fission and Fission Spectroscopy: Second Int. Workshop, in: AIP Conf. Proc., vol. 447, 1998, p. 330. 8. A. Oberstedt, S. Oberstedt and D. Rochman, in: Proceedings of International Conference on Nuclear Data for Science and Technology 2007, in press. 9. A. Oberstedt and S. Oberstedt, Nucl. Instr. Meth. A 570 (2007) 51. 10. Computer code SRIM 2000.39, available from J. F. Ziegler, IBM Research, Yorktown, NY-10598, USA and J. P. Biersack, Hahn-Meitner Institut, Berlin39, Germany. 11. C. Wagemans, Chap. 'Ternary Fission' in The Nuclear Fission Process, CRC Press (1991) 546. 12. Y. Baudinet-Robinet, Nucl. Inst. Methods 190 (1981) 197. 13. Yu. N . Kopatch, M. Mutterer, D. Schwalm, P. Thirolf, and F. Gonnenwein, Phys. Rev. C 65 (2002) 044614.
TERNARY FISSION OF CF ISOTOPES S. VERMOTE', C. WAGEMANS University of Gent, B-9000 Gent, Belgium
0. SEROT CEA Cadarache, F-13108 Saint-Paul-lez-Durance, France T. SOLDNER, P. GELTENBORT Institute Laue-Langevin. F-38042 Grenoble, France I. ALMAHAMID Wadsworth Center, New York State Department of Health, Albany NY 12201, USA And Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA W. LUKENS, J. FLOYD Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA During the last years, different Cm and Cf isotopes have been studied by our research group in the frame of a systematic investigation of gas emission characteristics in ternary fission. In this paper we report on the energy distribution and the emission probability of 'H,4He and 6He particles emitted in neutron induced ternary fission of 249Cfand 251Cf. Both measurements were performed at the high flux reactor of the Institute LaueLangevin (Grenoble, France), using suited AE-E telescope detectors, consisting of wellcalibrated silicon surface barrier detectors. In this way, the available database can be expanded with new results for Z=98 isotopes, for which the information on neutron induced ternary fission is almost nonexistent. These measurements are important for the systematic investigation of gas emission characteristics in ternary fission.
1. Introduction
Roughly 2 to 4 times per thousand fission events, the two heavy fission fragments are accompanied by a light charged particle. This process is called ternary fission; it is an important source of helium and tritium gas in nuclear reactors and in used fuel elements. Therefore, accurate ternary fission yields for 4He and tritons are needed by nuclear industry. Furthermore, ternary fission data
' E-mail address: [email protected] 107
108
are of interest to the nuclear physics field to improve the understanding of ternary particle emission and to provide information on the fission process itself. In this paper, we describe new measurements of neutron induced ternary fission of 249Cfand 251Cf.After a description of the experimental setup, the analysis and results concerning Long Range Alpha particles (LRA), tritons (t) and 6He particles are reported. More specifically, characteristics of energy distributions and emission probabilities will be shown. 2. Experimental setup
Both 249Cfand 251Cfneutron induced fission measurements were carried out at the PFlb cold neutron guide installed at the Institute Laue-Langevin (ILL) in Grenoble, France. The neutron flux at the sample position was higher than lo9 neutrons/s.cm2. 2.1. Sample characteristics
The 249Cfsample was prepared at the Lawrence Berkeley National Laboratory (LBNL) in the United States. The 5.84 pg 249Cfsample had an activity of 0.9 MBq and an enrichment of 100%. The 251Cfsample was prepared at the Institute of Nuclear Chemistry of Mainz University in Germany. Its activity was 10.8 MBq with a weight of 5 pg. The isotopic composition of the 251Cfsample is given in Table 1. In both cases, the targets consisted of californium oxide deposited on Ti-foils. Deposit diameters were 6 mm and 4 mm for the 249Cfand the 251Cfsample, respectively. Table 1: Isotopic composition of the Zs'Cfsample (06/06/07). Isotope Abundance (%)
249Cf 20.09
2s0Cf 26.71
25'Cf 53.09
252Cf 0.1 1
2.2. Detection system
The same detection system was used for both 249Cfand 251Cfmeasurements. The sample was placed in the center of a vacuum chamber at an angle of 45 degrees with the incoming neutron beam. A polyimide foil was used to cover the sample in order to prevent contamination of the chamber by recoil nuclei. Measurements were performed in two separate steps. In a first step, ternary particles were detected allowing the determination of both energy distributions and counting rates. Therefore, two well-calibrated silicon surface barrier AE-E
109
telescope detectors were placed at both sides of the sample in perpendicular position to the beam as shown in Fig. 1.
detector AE detector
2491251 Cf-sample
I
Figure 1 : Experimental setup for 249Cfand '"Cf neutron induced fission measurements.
In addition, AE detectors were covered with thin aluminium foils of 30 pm to stop alpha decay particles and fission fragments from penetrating the detector. Depending on the ternary particles we wanted to detect, the sample was turned in order to face the suited AE-E telescope detector. For both experiments, detector characteristics were chosen in order to have an optimal setup for detecting a and 6He particles and measuring binary fission, or for detecting a particles and tritons (Table 2). Table 2: Thickness of surface barrier detectors used. 249Cf AE E
L W B and 6He [pm]
29.8 500
LRA and t [pm] 49.8 1500
25'Cf
AE E
29.8 500
55.1162.9 1500
Signals coming from the surface barrier detectors were sent through a preamplifier and an amplifier. These signals were digitized in an Analogue to Digital Converter (ADC), coincident AE and E signals were stored in a PC.
110
The AE-signal is proportional to the energy deposited by the ternary particle traversing the silicon surface barrier detector; the E-signal is proportional to the remaining particle energy. Typical examples of coincident AE and E spectra for 249Cf(nth,f) are shown in Fig. 2. 600-
250
200
c
n
s
0
1
2
3
4
AE [MeV
5
6
'50
100
0
16
24
32
E [MeV]
Figure 2: Example of typical AE (left) and E (right) spectra for 24gCf(nth,f)
In a second step, binary fission fragments were detected in order to determine the Binary Fission Yield (B). At this stage, the AE detector from the telescope suited to measure L W , was removed, together with the aluminium foil, and replaced by a dummy ring with exactly the same dimensions. In this manner, binary fission fragments could be measured with the E-detector with the same detection geometry as the ternary particles. 3. Analysis and results
3.1. Particle identification The procedure used to identify various ternary particles and separate them from the background is the one proposed by Goulding et al. [ 11. This method is based on the difference in energy loss of different particles in the same material using where T is the thickness of the AE the equation: T l a = ( E + AE)'.73detector and a is a particle and material specific constant. An example of a Tla spectrum for 249Cf(n,h,f)is shown in Fig. 3, together with a 3-dimensional view of E-AE.
Ill
LRA l\
M j i 6He
triton
4GQ-
.
aoo-
10
15
3)
23
M
36
«
Figure 3: T/a spectrum for all ternary particles (left) and corresponding E-AE spectrum (right).
14001200-
1 i i1
i&ffi)1 3
..
1, Is
1600-
fcr
?• ""
i *s
808-
iS ™10
15
20
x^
%,
V.
, ,..«•£, 5
a-
\
/
^00-
m
25
30
35
40
IS
20
35
30
Figure 4: T/a spectrum for LRA particles alone (left) and corresponding E-AE spectrum (right).
The selection of ternary particles was realized by placing a window on the region of interest of the T/a spectrum. This has been done for LRA, tritons and 6 He particles. In the case of tritons, an additional correction due to the background was needed, as can be seen in the T/a spectrum (Fig. 3). In Fig. 4, a T/a spectrum containing only LRA particles is shown together with the corresponding E-AE spectrum. After the selection, AE and E spectra were obtained for a given ternary particle and the total energy distribution could be deduced. The thresholds in energy for each ternary particle are due to the thickness of the AE-detector, the electronic noise and the presence of Al-foil. The average energy and the full width at half maximum of the energy distribution were obtained from a Gaussian fit performed using experimental data. 3.2.
Results for the
249
Cf(nlhJ) measurement
The spectrum obtained from the binary fission measurement is plotted in Fig. 5. The small alpha pile-up peak due to the radioactive decay of 249Cf has to be
112
separated from the two bumps of the fission fragments. Then, the remaining spectrum is obtained and the corresponding number of binary fission events was deduced after integration of the extrapolated spectrum, yielding 326.68 h 1.79 binary fissions per second. 1
,c ,,
5030
-
4030
-
I
3030-
2
0
I
I
-
2030
a pile-up
10
Figure 5: Binary fission spectrum for the 249Cf(n,h,f)measurement.
Fig. 6 shows energy distributions for LRA, 6He and triton measurements. The characteristics of these distributions are given in Table 3. The uncertainties given correspond to the sum of statistical and systematical errors. Emission probabilities relative to 4He are shown in Table 4 together with the absolute emission probabilities. For LRA particles, a Gaussian fit was performed for particles with energies above 12.5 MeV. In the case of 6He particles, statistics were not very satisfactory; still a fit was done starting at 10 MeV. For tritons, the fit was performed starting at 6.4 MeV. The yield for tritons was obtained after subtracting the contribution of the background. Table 3: Values for average energy (E) and full width at half maximum (FWHM) for the various ternary particles measured. 249Cf LRA 6He tritons
E [MeV] 16.09 f 0.18 1 1.39 f 1.09 8.52 f 0.26
FWHM [MeV] 10.64 f 0.27 9.52 f 1.43 8.47 f 0.56
15.89f0.12 1 1.05 0.27 8.50*0.10
10.57i0.18 10.42 f 0.37 8.37 0.16
25'Cf
LRA 6He tritons
*
113 20
I
E [MeV]
Figure 6: Energy distributions for LRA, 6He and tritons for 249Cf(n~,f).
Table 4: Values for relative and absolute emission probabilities (per fission) for the various ternary particles measured. 249Cf LRA 6He tritons 25’Cf LRA 6He tritons
rel. em. prob. [“/.I 100 2.59 0.65 7.93 f 0.87 100 3.26 f 0.72 9.00 f 0.93
abs. em. prob. (2.77 0.1 I)xlO-’ (7.18.t 1 . 7 9 ) ~ 1 0 - ~ (2.20 f 0 . 2 6 ) ~ 1 0 - ~
*
(2.40 *0.14)x1O1 (7.85 f 1 . 9 6 ) ~ 1 0 ~ ~ (2.17 * 0 . 2 9 ) ~ 1 0 ~ ~
3.3. Results for the zs’Cf(n,h?f) measurement
Due to the isotopic composition of the sample (Table l), two measurements had to be performed in each step. The first one with the neutron beam open, measuring both the neutron induced and the spontaneous fission of all isotopes present in the sample. The second measurement is done with closed neutron beam, in order to determine the contribution of the spontaneous fission of 250Cf and 252Cf.In this way, results are obtained for the neutron induced fission only.
114
However, a correction still has to be made for the amount of 249Cfpresent in the sample. This can be done easily since the measurement results of the 249Cf sample are available in part 3.2. The spectrum obtained from the binary fission measurement with open neutron beam is plotted in Fig. 7. The quality of the spectrum is not as good as in the case of 249Cf(Fig. 5) due to degradation of the sample with age. After removing the alpha pile-up peak, the remaining spectrum is extrapolated. The number of binary fission events for neutron induced fission is obtained after subtraction of the spontaneous fission contribution, yielding 2410.47 f 56.47 binary fissions per second. I
I
I
I
1800
Channel
Figure 7: Binary fission spectrum (beam on) for the 25'Cf(n,h,f)measurement.
The spectra for neutron induced fission for LRA, 6He particles and tritons are shown in Fig. 8. For LRA particles, a Gaussian fit was performed starting at 12.5 MeV. For 6He particles, the energy threshold is quite high, therefore, the fit was performed starting at 11 MeV. Tritons were measured using two different AE detectors (Table 2), resulting both in reasonable values. In both cases the fit started at 6.5 MeV, and the result is the weighted average of both measurements. The values for the energy distribution of ternary particles are given in Table 3, emission probabilities are shown in Table 4. These represent results for the 25'Cf isotope only, obtained after correction for the 249Cfamount present in the sample.
115
E [MeV
Figure 8: Energy distributions for LRA, 6He and tritons for 251Cf(nlh,f).
4. Discussion
Over the last decades, a variety of measurements was performed by our research group in order to expand the database relevant to the emission of light ternary particles and to search for their systematic characteristics. In order to complete the series of Cf isotopes, information on 250Cf(SF)and "'Cf(SF) is needed. When emission probabilities for the compound nuclei 250Cfand 252Cfare known, the influence of excitation energy on emission probabilities can be examined. Indeed, this effect can be measured by comparing the ternary particle emission probability for the same compound nucleus at zero excitation energy (in the case of spontaneous fission) and at an excitation energy corresponding to the neutron binding energy (in the case of neutron induced fission). Available results on 250Cfgo back to 1985, when Wild et al. [2] measured the characteristics of LRA particles and tritons, with poor statistics. In order to obtain more accurate results a new measurement on '"Cf is in preparation. The sample needed for this experiment was prepared at the Lawrence Berkeley National Laboratory (LBNL) and was just shipped to our institute. In the case of 252Cf,different measurements were performed in the past [351, reporting values for average energy and FWHM of various ternary particles
116
together with relative emission probabilities. Nevertheless, results regarding absolute emission probabilities are still unsatisfactory. 5. Conclusion and outlook
In the present paper the main characteristics (energy distribution and emission probability) of LRA, tritons and 6He particles emitted in neutron induced fission of 249Cfand 25'Cf are presented. In this way, the available database could be enlarged with new values for Z=98 isotopes, for which information on neutron induced fission was almost nonexistent. These results are needed in the frame of the systematic investigation of gas emission characteristics in ternary fission, which is discussed in Wagemans et al. [6]. As a next step, a measurement of 250Cf(SF)is in preparation. In the near future, further investigation on 6He particles is necessary in order to obtain better insights in their behavior. To complete the study of the Cf isotopes, a measurement of the absolute LRA emission probability of 252Cfshould be performed. Acknowledgments Part of this research was performed at the Lawrence Berkeley National Laboratory (LBNL) and the 249Cfisotope was provided by the Office of Sciences, Office of Basic Energy Sciences and the Division of Chemical Sciences, Geosciences and Biosciences of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. The authors also acknowledge the assistance of D.K. Shuh of LBNL. J.O. Denschlag and N. Trautmann of the Institute of Nuclear Chemistry of the University of Mainz are acknowledged for the preparation of the 251Cf sample.
References 1. F.S. Goulding, D.A. Landis, J. Cerny and R.H. Pehl, Nucl. Instr. Meth. 31, l(1964). 2. J.F. Wild et al., Phys. Rev. C32,488 (1985). 3. C. Wagemans et al., Nucl. Phys. A742,291-302 (2004). 4. M. Mutterer et al., these proceedings. 5 . C. Wagemans, The Nuclear Fission Process, CRC Press, Boca Raton, USA (1991). 6 . C. Wagemans, S. Vermote and 0. Serot, these proceedings.
SYSTEMATICS OF THE TRITON AND ALPHA PARTICLE EMISSION IN TERNARY FISSION C. WAGEMANS, S. VERMOTE University of Gent, B-9000 Gent, Belgium
0. SEROT CEA Cadarache, F-13108 Saint-Paul-lez-Durance,France During the past two decades, the energy distribution and the emission probability of 'H and 4He particles emitted in thermal neutron induced and spontaneous ternary fission were investigated systematically for more than 20 nuclides. This permitted to observe several phenomena: (1) the constancy of the average energy of the 'H and 4He particles, respectively; (2) the linear increase of the FWHM of the energy distributions with Z2/A of the fissioning nucleus; (3) a smaller FWHM for spontaneous fission compared to neutron induced fission; (4) the impact of the a cluster preformation probability S, on the ternary a emission probability and ( 5 ) the different behaviour of the ternary a and triton emission probability with increasing excitation energy.
1. Introduction
During the past two decades, we have been performing a systematic study of ternary fission characteristics for spontaneous and thermal neutron induced fission of a large series of nuclides (see [l-41 and the refs. therein). All measurements were performed under similar experimental conditions, which is important if one wants to compare the characteristics observed. A detailed description of the experimental method can be found e.g. in [4]. So far, the thermal neutron induced fission was studied at intense neutron beams of the Institute Laue-Langevin (ILL) in Grenoble, France, for 229Th, 2 3 3 . 2 3 5 ~ , 2 3 7 ~ 239.241pu, ~ , 241.243 243,245,241 Cm and 249.251 Cf. Spontaneous fission Am, was studied at the Institute for Reference Materials and Measurements (IFWM) in ~ ~~ ~ ~ 1 for l 238240,242,244 , ~ i ~pu, 244,246,248 ~ , Cm and 252Cf. Since these are mostly rare and very radioactive isotopes, the samples used contained only small quantities of the material under investigation. For the neutron measurements, the small number of atoms could be compensated by the intense neutron flux available at the high flux reactor of the ILL. In the case of 117
118
spontaneous fission however, the combination of few atoms with small decay constants for spontaneous fission resulted in very long measuring times. 2. Results and discussion
The main characteristics studied in the measurements mentioned above are the emission probabilities of ternary a's and tritons, the corresponding energy distributions and the influence of the excitation energy of the fissioning nucleus on both characteristics. 2.1. Average energy of the ternary a's and tritons Fig. 1 shows the average energy for the ternary a's and tritons resp. as a function of the fissility parameter Z21A of the fissioning nucleus. The average energy of the ternary a's is compatible with a constant value of 16 MeV, for neutron induced as well as for spontaneous fission. In first approximation this is also the case for the hitons, but here the average energy is 8.37 MeV.
Fz n
I i
4
4
so
,
, 355
,
, 360
,
, 305
,
, 37.0
,
, 375
,
, laa
,
, ~(LS
,
{ JPO
0
Z4A
Figure 1: Average energy for the ternary a's (left) and ternary tritons (right). In both cases, the straight line is a fit to the (nh,Q and the spontaneous fission points.
2.2. Widths of the energy distributions
The full width at half maximum (FWHM) of energy distributions always is a parameter that is very sensitive to experimental conditions such as detection geometry, sample thickness, detection lower limit, etc. Therefore comparable experimental conditions (as realized in our systematic study) are clearly needed to permit a reliable comparison. Fig. 2 shows the FWHM versus Z2/A for the ternary a's and tritons respectively. In both cases a linear increase of the FWHM with increasing Z'IA can be observed, which can be understood as follows: the broadening of the energy distribution is mainly due to the Coulomb field that
119
amplifies small differences in the initial kinetic energy distribution of the ternary particles. These small differences are due to fluctuations of the scission shape, which in turn increase with increasing deformation energy, which in turn is proportional to Z'IA. Moreover, the FWHM of the ternary a's is 0.2 MeV smaller for spontaneous fission than for neutron induced fission. This confirms the wellknown phenomenon already observed for fission fragments, that excitation energy enlarges kinetic energy distributions.
Figure 2: FWHM of the ternary a energy distribution (left); idem for the ternary tritons (right). For the tritons, only a few data are available, so the fit includes both (n,& and (SF) data points.
2.3. Ternary a and triton emission probabilities
The ternary a and triton emission probabilities L R A B and tA3 have been con-elated in the literature with many other fission observables. In the present paper we limit ourselves to two logic choices: (a) the fissility parameter Z2/A (the energy needed for ternary particle emission is taken from the deformation energy, which is proportional to Z'IA); (b) the Coulomb parameter Z2/A"3(the ternary particles are indeed ejected by the Coulomb field). Figs. 3 and 4 show the behaviour of LRA/B and tA3 versus Z'IA and Z21A1i3, respectively. These figures permit two straightforward observations: in both cases the triton emission probability linearly increases with increasing Zi/A or Z2/A'J3 values, as expected; for the ternary a's on the other hand the situation is not so clear. As we mentioned before [ 5 ] , the clue for the explanation of the different behaviour of ternary a's and tritons lies in the cluster preformation probability, and this becomes more evident when the data base becomes larger.
120
Figure 3: Emission probability as a function of the fissility parameter Z2/A for the ternary a's (left) and ternary tritons (right).
Figure 4: Emission probability as a function of the Coulomb parameter Z2/A"' for the ternary a's (left) and ternary tritons (right).
The a cluster preformation probability S, is defined as S,= hexp/A+,,,,; here hexpstands for the experimental value of the a decay constant and hcamOv is the corresponding value calculated with the WKB approximation [ 6 ] .This implies that S , values are only available for ground state decay. For the tritons on the other hand, no preformation is expected. Fig. 5 shows the a emission probability LRAA3 divided by S,, but using the same S, value for spontaneous and neutron induced fission leading to the same compound nucleus. It clearly illustrates that, after removing the contribution of the a preformation, also the ternary a emission probability increases with increasing Z'IA and Z21A''3.One can also observe that the corrected emission probability is always smaller for neutron induced fission than for spontaneous
121
fission. This indicates that the S, value used for neutron induced fission is overestimated, which is no surprise since we approximated this value by S, for spontaneous fission. I
. , . ( . , . , . , . , . , .
00 8 0
8 5
350
355
375
370
350
355
.
00 390
,
.
. ,
, 1 m
1350
,1
1450
,
, 1 m
, 1553
.fIAn3
Z'IA
Figure 5 : Ternary a emission probability corrected for the a cluster preformation probability.
2.4. Influence of the excitation energy In our data base we have several examples of fissioning nuclides at two different excitation energies: E,,, = 0 for the spontaneous fission of the nucleus AX and E,,, = the neutron binding energy (=: 6.5 MeV) for the nucleus formed by *-'X + thermal neutron. Fig. 6 shows the ratio of the ternary a (resp. triton) emission probability for neutron induced and spontaneous fission. 16-
16
g e 2
14:
14HeJ
14: 12-
12-
'
5
e
c5 1 0 -
08-.
I
0
06-
L !I
I
=
i a .
--
' 10-
c
!
. 3 082 . 06-
2
pJi I
I'
I
0 : . 04-
04-
Figure 6: Ratio of the emission probability for thermal neutron induced and spontaneous fission for ternary a's (left) and tritons (right).
122
A striking difference can be observed: the ternary a emission probability decreases by about 20% in the case of neutron induced fission, despite an increase in excitation energy. The triton emission probability on the other hand is almost independent of the excitation energy. This can be explained by a reduction of the a cluster preformation probability with about 20% at an excitation energy of = 6.5 MeV. 3. Outlook Another interesting particle emitted in ternary fission is 6He, which has an emission probability of about 3% relative to 4He. According to the theory of Blendowske [7], the 6He preformation probability S6He = [S,] 5’3, so with S, = 0.006 one expects that the 6He preformation probability is about 30 times smaller than S,. So we want to investigate its impact on the 6He emission probability: will it behave as the a emission probability or rather as the triton emission probability? In first instance we want to investigate this for the fissioning systems 244Cmand 250Cf.
References 1. C. Wagemans, The Nuclear Fission Process, CRC Press, Boca Raton, USA (1991). 2. C. Wagemans et al., Nucl. Phys. A742,291-302 (2004). 3. 0. Serot, C. Wagemans and J. Heyse, Proc. Int. Conf. on Nucl. Data for Science and Technology, Santa Fe, USA, AIP Conf. Proc. 769 (2005) 857. 4. S. Vermote et al., these proceedings. 5. C. Wagemans and 0. Serot, Proc. Int. Conf. on Dynamical Aspects of Nuclear Fission, Casta-Papiemicka (Slovakia) 200 1, World Scientific (2003) 301. 6. 0. Serot, N. Cajan and C. Wagemans, Eur. Phys. J. A8 (2000) 187. 7. R. Blendowske, T. Fliesbach and H. Walliser, Z. Phys. A339 (1991) 121.
Neutron Emission in Fission
This page intentionally left blank
SCISSION NEUTRON EMISSION IN FISSION F.-J. HAMBSCH, N. KORNILOV, I. FABRY, S. OBERSTEDT EC-JRC-IRMM,Retieseweg, B-2440 Geel, Belgium E-mail: franz-joseJhambsch@ec. europa.eu
A. VOROBYEV St Petersburg Nuclear Physics Institute (PNPI), 188350 Gatchina, Russia E-mail: alexander. [email protected]
New experimental results for the prompt fission neutron spectrum (PFNS) of '"U at 0.5 MeV incident neutron energy are discussed. These results being in good agreement with several other literature data do not reproduce the Los Alamos (LA) model prediction. The new results give evidence of an angular effect in the PFNS and the emission of scission neutrons (SCN). The last conclusion is supported by the comparison of experimental results with model calculations for the neutron multiplicity of "'Cf, 244.248Cm, and However, the nature of the angular effect is not clear as well as the contradiction between low neutron multiplicity and high average neutron energy for some mass splits. Also existing differences between microscopic and macroscopic data can not be explained with the available experimental data.
1. Introduction The prompt fission neutron spectrum (PFNS) plays an important role in various nuclear energy and non-energy applications. From a more fimdamental point of view, an accurate knowledge of the PFNS can shed light on the nuclear fission process itself. The properties of fission neutrons, their multiplicity and energy distribution, could give answers to related questions on the mechanism of neutron emission and the fission process itself, how the energy is distributed between the complimentary fragments and, what is the time scale of the fission process. The recent PFNS measurements ([ 1-41), being in reasonable agreement with each other, result in an average energy of the promptly emitted fission neutrons that is smaller than required by benchmark experiments, macroscopic experimental data and the predictions by the Madland-Nix (Los Alamos) model [ 5 ] . This problem was highlighted in Ref. [6] and stimulated a new interest in 125
126
PFNS investigations. It was concluded [7] that the present database of the neutron spectrum for 235U(n,f)at thermal incident neutron energy is insufficient and needs improvement. First measurements have been performed at IRMM for 0.5 MeV incident neutron energy because at this energy the LA-model can perfectly describe the experimental results in Ref. [8]. The results achieved so far [9,10] do not reproduce the behaviour of the LA-model, but are in agreement with several other literature values. An angular dependence of the measured neutron spectrum is also not completely excluded. A satisfactory description of the present results concerning the PFNS with the LA-model is only possible, when scission neutron emission is assumed. For spontaneous fission of 252Cf,2443248Cm and thermal neutron induced fission of 235U,the prompt neutron multiplicity v(A) and the neutron energy &(A)were calculated [l 11 as a hnction of fission fragment mass and v(TKE) and compared to available experimental data. Whilst v(A) can be very well reproduced &(A)cannot simultaneously pointing to a problem in the modeling of neutron emission. In spontaneous fission v(TKE) can only be well reproduced, if the existence of scission neutrons correlated with high TKE values of the fission fragments is assumed. For neutron induced fission of 235Uexisting experimental data cannot be reproduced. Too low v(TKE) values are measured for low TKE as well as a too broad mass distribution, both pointing to problems with the experimental data. Hence, in both investigations the incorporation of scission neutron emission plays a crucial role and improves very much the comparison between modeling and experimental results. In this contribution both investigations are linked by the same idea - that is scission neutron emission is an important ingredient of the fission process. 2. The prompt fission neutron spectrum
The present measurements were carried out at the 7 M V Van de Graaff accelerator of the IRMM in Geel, Belgium, using the fast neutron time-of-flight technique. A pulsed proton beam of about 1.3 ns FWHM at 1.25 MHz repetition rate and 0.5 pA average current was used. Mono-energetic neutrons of 0.52 MeV average energy were produced using the 7Li(p,n) reaction. A metallic 235Usample (93.15 % enrichment, 161.28 g) and a similar sized lead sample were applied for foreground and background measurements, respectively. The actual neutron energy range due to the LiF-target thickness and the geometrical factor extended from 0.41 MeV to 0.58 MeV.
127
Up to three NE213-equivalent liquid scintillation detectors of 10.2 cm diameter and 5.1 cm length were used in the measurement. In a first experiment we used two liquid-scintillation neutron detectors placed at 90 degree (flight path 2.73 m) and at 120 degree (flight path 2.40 m). The distance from the neutron production target to the sample was ~5 cm. The results of this experiment are available with Refs. [9,10], and have been submitted to ND2007 [13]. In a second experiment we used three identical neutron detectors at a flight path of 2.24 ± 0.01 m placed at 90, 150 and 120 degrees. The distance from the neutron production target to the sample was ~8 cm. The detectors were shielded against direct and room-scattered neutrons. The traditional pulse-shape analysis was applied to reduce the gamma-ray background. A small Pilot-U scintillator was used as a proton pulse shape monitor. The detector efficiencies were measured relative to the 252Cf standard spectrum. A specially designed low mass, fast ionization chamber [12] was put at the place of the U-sample keeping the same geometry as during the experiments. The energy spectra were corrected for detector efficiency, for neutron multiple scattering in the sample, and for time resolution. A detailed description of the experimental procedure will be published elsewhere [9].
•M
m
1W9
«M
«0»
TOT channel (eft « 0.483 rss)
Figure I. Time-of-flight spectra for the 235U (full symbol) and the Pb- sample (open symbol) at 120 degree. Both spectra were normalized to the elastic neutron scattering peak.
TOF.n* Figure 2. Comparison between the distribution of the input neutrons samples for the present experiment symbols) and the one of Ref. [14] symbols). The data are from a Monte simulation.
TOF inside (full (open Carlo
The time-of-flight spectra for the uranium and lead runs for the detector at 120 degree are given in Fig. 1. Both spectra were normalized to the elastic neutron scattering peak. No time dependent peculiarities were found in the leadsample background run in the neutron energy range of interest from 0.7 MeV to 12 MeV.
128
Fig. 2 shows the TOF intensity distribution of the incident neutrons inside the sample for our experiment compared to the one in Ref. [14]. Clearly the double peaked intensity structure of our ring shaped sample is visible in contrast to the very broad distribution of the very large sample (7.7 cm diameter) of Ref. [14]. This factor together with a shorter flight path (1.63 m) in spite of a very short proton burst of 0.6 ns is very important for data comparison. Therefore, the experimental data of Ref. [14] were corrected for this time resolution. We also recalculated the time resolution correction for our measured spectra [9, 10, 131. An additional energy dependence in the neutron detector efficiency at >4
1.q
2
c9
N II A
-present experiment 120 degrea -0 - Staples et al.
-0
1.2 1.0 0.8
E, MeV Figure 3. Comparison between the present results at 120 degree (full symbol) and the data of Ref. [I41 (open symbol) at 90 degree.
MeV has been taken into account, too. Details are given in Ref. [9]. This factor slightly reduced the PFNS in the energy range 5-8 MeV and the average neutron energy up to -1 5 keV for the first run. Comparing now the spectrum shape of the present experiment to the one obtained in Ref. [14] we see a very good agreement to our result at 120 degree up to at least 8 MeV (see Fig. 3). The experimental PFNS were normalized to unity and the average neutron energy was calculated. A Maxwellian spectrum was fitted in the energy range of 0.7-1.5 MeV and 9-11 MeV to the measured spectrum and an extrapolation to zero and to 20 MeV was done. Based on our detailed analysis of all incorporated corrections and possible uncertainties, we conclude that the
129
average energy is estimated with an accuracy of + 0.015 MeV. The average energies measured in both runs are given in Table 1. Table 1. Average energies of the PFNS for all angles and runs.
Angle, degree
90 120 150
<E>, MeV Run 1 2.008 2.076
<E>, MeV Run 2 2.004 2.050 2.026
1.4
- W-deg
7
0,8-
*
0.6-
• 0.4
- 150-tteg - 12Q-d*g 120-deg(20C6) - Staples et af. —Starostev «t ai. ~ENDF/B-Vl
4
6 E,MeV
Figure 4. Comparison between the present results at several angles (full symbols) with data of Ref. [14] (open squares) and Ref. [1] (open triangles). Also ENDF/B-VI is given as full line.
The PFNS at all investigated angles and for both runs (runl in 2006 and run2 in 2007) are shown in Fig. 4 as a ratio to a Maxwellian distribution with the same average energy. A very good agreement of our data at 90 degree is observed with the data of Ref. [1] over the whole prompt fission neutron energy range. The data at 150 and 120 degree from the 2007 run deviate outside the error bars from the 90 degree data. The data at 120 degree of the 2006 run are deviating from the 120 degree data of 2007, but are in very good agreement with the data of Ref. [14] (already seen in Fig. 3). The ENDF/B-VI data do deviate mainly at low secondary neutron energies smaller 2 MeV.
130
Altogether, the present results point to the presence of an angular anisotropy in prompt neutron emission. An angular dependence of the PFNS shape and neutron emission yield was already discussed e.g. in Ref. [15]. Therein the PFNS was, however, never analyzed with an angular dependence in mind and the deviation may have been masked by statistical uncertainties. Actually, most of the previous experiments were not able to reveal an anisotropy in the PFNS mainly due to the low statistical accuracy. The observed difference for the data obtained at 120 degree between run 1 and 2 still needs some clarification. The only difference between the two experiments was the sample to source distance. The present results obviously show a tendency of an angular dependence of the PFNS, although unexpected and physically not understood. The observed angular anisotropy of the PFNS cannot be explained with the fission-fragment angular anisotropy in the laboratory system. For a fission-fragment angular anisotropy A = W(Oo)/W(9O0) - 1 = 0.06 [16] the spectrum ratio at 90 degree to the 120 degree is much closer to unity and does not correlate with the energy dependence of the present results. Hence, the presently observed anisotropic neutron emission can only be connected with neutron emission before scission of the compound nucleus, the so-called scission neutrons. In this case fission neutrons should be emitted from three sources: 1. Neutrons from fragments after fission of the compound nucleus A + 1
N,,, ( E l = (1 - a ). w,+, (El
9
(1)
where a is the share of scission neutron emission and WA+,is the spectrum which describes the neutron emission from accelerated fragments; 2. Neutrons from accelerated fragments after fission of the nucleus A , which is formed after the emission of one SCN:
N,( E ) = a . (V - 1) . W,( E )/ v .
(2)
3. Scission neutrons itself:
V
where is the share of the low energy component.
(3)
131
The spectrum WA+, was taken from the ENDFB-VI library, which coincides with the Los Alamos model, for an input energy Eo. The WAspectrum was also taken from ENDFB-VI but it was calculated with the extrapolation of the spectrum to the energy Eo-B,-<Escn>= Eo-8.5 MeV. The equation for Nscn(E),and the corresponding parameters a, 6, T I ,T, were taken from Ref. [17] introducing minor changes: a = 0.4, TI = 0.35 MeV, T2 = 1.25 MeV. The experimental data were described inside the error bars with variation of only the share of the low energy component 6. So ((90") = 0.28 ($/n = 0.6), ((120") = 0.13 (?/n = 0.8), 6(150") = 0.20 (?/n = 0.6). An example of a nearly perfect agreement with the experimental data at 90 degree emission angle from 0.7 MeV up to 10 MeV prompt neutron energy is shown in Fig. 5. The spectrum at any incident neutron energy may be calculated on the basis of the ENDFB-VI library and the parameters found from our experimental spectra. Assuming that the SCN spectrum has the same shape at thermal energy, we compare the model spectra with Starostov et al. [l], and Lajtai et al. [2] at 90 degree emission angle. Fig. 6 shows all experimental data and the "3 sources model'' result ( = C/E = 0.956 f 0.008. At the same time,
E, MeV Figure 5 . Modeling of the PFNS using the "3 sources model" according to eqs. (1-3).
132
J
.
1
.,,,,I
0.01
I
r
, , ,'.,,
, , , ,,.")
0.1
,
1
.
I
rrl.T
1
E,MeV Figure 6. Experimental data of Ref. [1,2] together with the results of the "3 sources model", the ENDF/B-VI evaluation and the approach of Kornilov et al. [I 81
this ratio calculated with the 120-degree spectrum ( = 0.994 f 0.007. In conclusion, an angular-dependence of the PFNS was measured for the first time using detectors at 90, 120 and 150 degree. None of the obligatory corrections (detector efficiency, multiple scattering in the sample, uncertainties of sample and neutron beam position, time resolution, time independent background) did reveal any systematic errors that may have produced the measured anisotropy. We may describe this effect changing the contribution of the low energy component, but we have no scientific explanation of this approach. In addition, we have no idea how to explain the discrepancy between available microscopic and macroscopic data. 3. Neutron multiplicity and average energy versus fission fragment split
For the correct calculation of the neutron multiplicity versus fragment mass and TKE, the two-dimensional distribution Y(A,TKE) should be known. Experimental data for Y(A,TKE) are only available for the major fissile isotopes. To overcome this deficiency the fission fragment yield was calculated in the framework of the multi-modal random-neck rupture fission model [19] using the same ideas as developed in Ref. [20]. Model parameters were adjusted iteratively by comparing the calculated with the experimental one-dimensional
133
Y(A) and TKE(A) distributions. Four fission modes were assumed - one symmetric mode and three asymmetric modes. Neutron emission has been calculated based on the total excitation energy (TXE) calculated by assuming an unchanged charge distribution. The TXE was distributed between the light and heavy fragments based on a thermodynamicalequilibrium assumption. For each target nuclei 235U(nlh,f),2449248Cm(sf) and z52Cf(sf)the dependence of neutron multiplicity and neutron energy on the fission fragment mass, v(A) and .(A) was calculated. The "'Cf results together with the experimental data of Ref. [21] are given in Fig. 7. With the level density systematic of Ignatyuk [22] we are unable to describe the v,,,(A)-data of Ref. [21] (see dashed line in Fig. 7a). However, one can reproduce reasonablly well the "bell like" dependence for ,$A) (see dashed line in Fig. 7b). To be able to reproduce the v,,,(A)-data in Fig. 3a we had to introduce a correction factor on the share of the excitation energy of the fragments of 0.40 < cor(A) < 1.25. The calculated result for v(A) including this correction is shown in Fig. 7a as full line. Consequences were observed on the disagreement for &(A)in the mass range with v(A) < 1. However, for mass ranges 90-105 and 145-170 a better agreement including the additional correction for &(A)was reached (full line in Fig. 7b). For 235U(nlh,f) the Nishio et al. [23] and M a s h et al. [24] experimental data for v,,,(A) were rather well described with cor(A) = 0.9 in the framework of our model (full line in Fig. 8a). However, the model failed again to predict &(A)(full line in Fig. 8b). Similar peculiarities were found in Ref. [25]. Moreover, our calculation and those of Ref. [25] seem to describe, with reasonable agreement, &(A)only for light fragments . The experimental data for heavy fragments show an inverse dependence on &(A)(see Fig. 8b). The v( TKE) for spontaneously fissioning isotopes 244*248Cm and 252Cfis given in Fig. 9 [26]. The full symbols represent the experimental data for heavy (squares) and light fragments (triangles), and for the total v(TKE) (circles). The open symbols correspond to model calculations. The solid line indicates the corresponding TKE distribution. All calculations for 244s248Cm and " ' ~ f were carried out with the same correction for the share of the excitation energy between fragments as applied before to 252Cf.
134
0,o 2.0
3 1.0 0.0
80
100
120
I40
leu
I80
A r==1
Figure 7 a Expenmental neutron multiplicity for 252 Cf [21] (full symbols) compared to the model calculations The dashed line is based on the level density systematic of Ignatjuk [22] and the full line on the additional correction (see text). b Expenmental average neutron energy [21] (full symbols) compared to model calculations. The dashed line is based on Ref [22] and the full line on the additional correction.
rpo
100
$20
I40
$60
1m
A [amq
Figure 8. a. Experimental neutron multiplicity for 2’5U(nth,f)of Ref. [23] (full symbols) and Ref. [24] (open symbols) compared to our model calculations (full line). b. The expenmental average neutron energy of Ref. [23] (full symbols) compared to our model calculation (full line).
It is obvious from Figs. 9, 10 that v(TKE) is overestimated for low TKE and the positive slope dv/dE observed for all isotopes cannot be reproduced. Actually, this is a very difficult energy region to assess in experiments based on the double energy (2E) method. An event from the range of maximum yield with moderate excitation and v 3-4, if false identified, will most probably be registered in either the symmetric or very asymmetric mass range [27]. This would effectively lead to a reduction of v at low TKE. Hence, a low v(TKE) in conjunction with a positive slope dv/dE may point to systematic errors in the experiment.
-
135
1«
180
180
22Q
Mi
169
186
Z0§
228
TMB
Figure 9. a. Experimental v(TKE) for 252Cf(sf) [26] (closed symbols) compared to the model calculation (open symbols). The solid line shows the corresponding fragment yield. The circles are the total v(TKE), the triangles are the contribution of v(TKE) for the heavy fragments and the squares for the light fragments. b. The same as in Fig. 9a but for 248Cm(sf) [26]. c. The same as in Fig. 9a but for 244Cm(sf) [26].
Figure 10. The same as in Fig. 9, but now the contribution of the scission neutrons is taken into account and is given by the dashed line.
A perfect agreement in slope and absolute value is found at higher TKE (see Fig. 9). However, at still higher TKE a constant difference between theoretical and experimental data becomes evident. This cannot be explained due to experimental problems. Introducing scission neutron emission [11] at higher TKE made the agreement between experiment and model calculation much better. A similar behaviour is observed in the case of 235U(nth,f) (see Fig. 11, 12). For this isotope the deviation at low TKE between experiment and theoretical calculation are even higher and point to the same experimental problems as in the case of spontaneously fissioning isotopes. Adding scission neutrons also here improves the agreement at higher TKE (see Fig. 11).
136
•us
«»
«s
aw
Figure 11. Experimental neutron multiplicity versus TKE data for 235U(nth,f) (closed symbols) compared to the model calculation (open symbols). The solid line shows the Y(TKE). The open circles are the total v(TKE), the open triangles are the contribution of v(TKE) for the light and the open squares for the heavy fragments. The dashed line shows the calculation taking into account a broadening of the experimental data, the dotted line represents the results of Ref. [25]. The experimental data are from Refs. [23,24, 28].
,^_. MB
«
tW
Figure 12. The same as in Fig. 11, but now the contribution of the scission neutrons is taken into account and is given by the dashed line. The experimental data are from Refs. [23, 24, 28].
4. Conclusions For the PFNS, the new measurements are not in agreement with calculations based on the LA model. We have no direct answer of the physical nature for the disagreement between new (old) microscopic spectra and macroscopic data (average cross sections). The angular anisotropy of the PFNS which has been observed in the present measurements may be responsible for this contradiction. Neutron emission in fission can be modeled for v(A) but does not simultaneously agree with e(A). For v(TKE) we believe that at low TKE experimental problems govern the data and hence they should not be regarded as reliable. At higher TKE introducing scission neutrons in the model calculation improved the agreement with experimental data for spontaneous and neutron induced fission. Both effects - the angular anisotropy and the behaviour of the neutron multiplicity at high total kinetic energy require incorporation of a scission neutron emission mechanism. References 1. B.I. Starostov et al., in Proc. 6th Conf. for Neutron Physics, Kiev, 1983, Obninsk 1984, pp. 285, EXFOR 40871, 40872, 40873. 2. A. Lajtai, IAEA-TECDOC-335, 1985, 312, EXFOR 30704.
137
3. 4. 5. 6.
W. Yufeng et al., Chin. J. Nucl.Phys. 11 (1989) 47. EXFOR32587. A. M. Trufanov et al., J. Nucl. Phys. 57 (1994) 4,606. D.G.Madland, J.R.Nix,Nucl. Sci. Eng. 81,213 (1982). N.V. Kornilov, A.B. Kagalenko, K.I. Zolatorev, in Proc. of Illh Int. Seminar on Neutron Interaction with Nuclei (ISI"-6), Dubna, 1998, 242. 7. D.G. Madland, ISBN-92-64-02134-5, NEA/WPEC-9,2003. 8. P.I., Johansson, B.Holmqvist, Nucl. Sci. Eng. 62 (1977) 695. 9. N.V. Kornilov et al., An experimental study of prompt neutrons from 235U fission, Internal report GE/NP/O1/2007/02/14, to be published. 10. N.V.Kornilov, F.-J. Hambsch, S. Oberstedt et al., JRC-IRMM, Neutron Physics Unit, Scientific report 2006, p. 37. 11. N.V.Kornilov, F.-J. Hambsch, A.S. Vorobyev, Nuclear Physics A789 (2007) 55-72; 12. N.V. Kornilov, I. Fabry, F.-J. Hambsch, et al., JRC-IRMM, Neutron Physics Unit, Scientific report 2005, p. 67. 13. N.V. Kornilov, F.-J. Hambsch, I. Fabry, et al., Proc. of the Znt. Conf for Nucl. Data for Sci. and Tech., ND2007, Nice, France, Apr 2007, to be published. 14. P. Staples, J. J. Egan, G. H. R. Kegel, A. Mittler and M. L. Woodring , Nuclear Physics A59 1 (1995) 4 1. 15. M.M. Islam, H.-H. Knitter, Nucl. Sci. Eng. 50 (1973) 108. 16. F.-J. Hambsch, Nuclear data standards for nuclear measurements, NEADC -31 1 "U", INDC(SEC)-101, 1992, p. 59. 17. N.V. Kornilov et al., Phys. of Atomic Nuclei. 62 (1999) 209 and Nucl. Phys. A686 (2001) 187. 18. N. V. Kornilov, A.B.Kagalenko, F.-J. Hambsch, Physics of Atomic Nuclei, 62 (1999). 19. U. Brosa, S. Grossmann, A. Muller, Physics Reports 197 (1990) 167. 20. N.V. Kornilov, A.B. Kagalenko, Proc. of l l t hInt. Seminar on Neutron Interaction with Nuclei (ISINN-1l), Dubna 2003, 144. 21. C. Budtz-Jrargensen, H.-H. Knitter, Nucl. Phys. A490 (1988) 307. 22. V. Ignatyuk, K. Istekov, G.N. Smirenkin, Sov. J. Nucl. Phys. 29 (1979) 450. 23. K. Nishio, Y. Nakagome, H. Yamamoto and I. Kimura, Nucl. Phys A632 (1998) 540. 24. E.E. Maslin, A.L. Rodgers, W.G.F. Core, Phys. Rev. 164 (1967) 1520. 25. S. Lemaire, P. Talou, T. Kawano, M. B. Chadwick, and D. G Madland, Phys. Rev. C72 (2005) 024601. 26. V. A. Kalinin, V.N. Dushin, B.F. Petrov, et al., Proc. Of 5" nt. Conf., Casta-Papernicka, Slovak Rep., 23-27 Oct., 2001, p. 252. 27. F.-J. Hambsch, S. Oberstedt, Nucl. Phys. A617 (1997) 347. 28. J .W. Boldeman, A.R. Musgrove, R.L. Walsh, Aust. J. Phys. 24 ( 971) 821.
AT AND BEYOND THE SCISSION POINT: WHAT CAN WE LEARN FROM SCISSION AND PROMPT NEUTRONS? P.TALOU Nuclear Physics Group, Theoretical Division, Los Alamos National Laboratory Los Alamos, N M 87545, USA E-mail: talouQlanl.gov Scission and prompt neutrons provide indirect clues of the physical processes a t play near the scission point when the two fission fragments are finally and forever separated. A scission model based on the approximation of a sudden rupture of the neck between the two nascent fragments is presented. It is used to compute the average number of scission neutrons per fission event in the symmetric fission of 236U. Once the fragments are fully accelerated, they will release their intrinsic excitation energy by emitting so-called prompt neutrons and gamma-rays. Monte Carlo simulations of this evaporation process are presented in the case of the first-chance fission of 235U. Future developments expected in the arenas of experiment, theory and evaluation are discussed. Keywords: Nuclear Fission; Scission; Scission neutrons; Prompt Neutrons
1. Introduction
The point at which nuclear scission occurs, or "scission point", is of crucial importance in our understanding of the nuclear fission process, since it dictates its final outcome, with two (or more) fragments in specific intrinsic and collective states. Although the direct observation of the nuclear configurations of the fragments at the scission point remains elusive", its consequences have nevertheless been well studied. Indeed, the primary fission fragment yields, the total number of prompt neutrons and gamma-rays, and their spectra, are all consequences of what happens near the scission point. anotwithstanding that even a precise and quantitative definition of the scission point is itself elusive.
139
140
One important and unsolved question in the nuclear fission puzzle is how the total excitation energy available at scission is shared among the light and the heavy fragments. Since this energy is believed to be released by the rapid emission of neutrons and gamma-rays from the excited fission fragments, the study of prompt neutrons and gamma-rays characteristics should help answer this question. From an evaluator’s point of view, modern libraries of evaluated nuclear data contain information on average quantities only, such as the average number of prompt neutrons F p , and the average prompt neutrons emission spectrum N ( E ) . Such data are commonly obtained by applying the so-called Los Alamos or Madland-Nix mode1.l However, recent nuclear technology applications (e.g., non-proliferation) are asking for more detailed informations, such as n-n correlations. The present paper deals with scission and prompt neutrons. In Section 2, the topic of scission neutrons is introduced, first by presenting experimental evidence indicating their existence, and next by introducing theoretical model calculations in the sudden approximation. Numerical results in the case of the scission of 236Uare presented. Section 3 presents Monte Car10 simulations of the evaporation stage of the excited primary fission fragments in the case of the first-chance fission of 235U.A discussion on future developments in experiment, theory and evaluation activities concludes this paper.
2. Scission Neutrons 2.1. Experimental Evidence The very existence of so-called ”scission neutrons”, i.e., emitted near or a t the scission point, remains highly controversial. From the viewpoint of nuclear applications (so far!) , the prompt neutrons fission spectrum and multiplicity are really the only two quantities of interest, and in this regard, the Madland-Nix model has been highly successful in describing and predicting those quantities for many isotopes and incident neutron energies. This model does not consider any contribution coming from a scission neutrons component though. But it does not imply that they do not exist either. In Ref.2 G.A.Petrov made a compilation of what seems to be the status of experimental data on scission neutrons. Several experiments indicate the existence of scission neutrons (up to 20% of the total number of prompt neutrons) while others are consistent with no scission neutrons at all. As
141
long as this question remains unsettled, we believe theoretical as well as experimental efforts should be devoted to it. 2 . 2 . Theoretical Modeling: The Sudden Approximation
Following the image of H a l ~ e r n we , ~ are making the extreme assumption that the rupture of the neck at scission happens suddenly. From a quantum mechanical point of view, this "sudden approximation" leaves singleparticle eigenstates of the hamiltonian "just before scission" as a superposition of the eigenstates of the "immediately after scission" hamiltonian: allstates
where aif =< @flQi > are the single-particle survival probabilities. In an axially symmetric configuration, single-particle wave functions are characterized by the projection R of the angular momentum on the axis of symmetry. If in addition, the nucleus has a symmetry of reflection, the parity 7r is also a good quantum number. The neutron emission probability will be given by the sum of all contributions present in the continuum: u n b o u n d states
bound states
Note that we are making the assumption of no reabsorption of the emitted neutrons by the just formed fragments. The total number of scission neutrons per fission is obtained by summing over all occupational probabilities of the initial eigenstates: i
where v: is the BCS ground-state occupation probability of IQi >. This formalism has first been applied to the symmetric fission of 23sU4 and later on t o asymmetric fission configuration^.^ Figure 1 exemplifies the application of this model with two typical situations. The left part of the figure represents two initial eigenstates, before scission, while the right part depicts their adiabatic partners after scission. In the top case, the initial wave function is largely present in the neck. Its adiabatic partner (top right) however cannot be present in the neck, and their overlap is therefore hindered. In this case, the emission probability will be large, and mostly localized in the neck region. On the contrary, in the scenario below, the initial state is already mostly present in the nascent
142
fragment (bottom left), and therefore quite similar to its adiabatic partner (bottom right). In this case, the overlap is large, and the emission probability small.
Fig. 1. A pair of initial wave functions before scission (on the left) and their corresponding adiabatic partners after scission (on the right). In the top scenario, the initial eigenstate is largely present in the neck, hence its overlap with its partner is hindered (only 0.25). The scenario below shows an initial state mostly present in the nascent fragment, hence its overlap with its partner is quite favored (0.75).
By application of Eq. (3), the total scission neutron multiplicity Nn can be calculated. Depending on the computational assumptions, Nn lies between 0.16 and 1.73 for the symmetric fission of 236U. In what we believe to be the most realistic situation, the scission neutron multiplicity would represent about 15% of the total number of prompt neutrons. We can expect similar numbers for asymmetric fission from the initial results shown in Ref.5 It is important to note however that due to the sudden approximation and to the neglect of any possible reabsorption of the neutrons by the moving fragments immediately after scission, those numbers represent upper limits only. 3. Prompt Neutrons
3.1. Theoretical Model and Computational Details We have studied the de-excitation of the fully accelerated fission fragments by performing Monte Carlo simulations of the evaporation process. Neu-
143
trons are emitted sequentially from the excited primary fission fragments following a Weisskopf evaporation spectrum. The complete decay is followed until a fission product is reached and the residual energy is too low for the emission of additional neutrons. In the present calculations, it is assumed that all prompt gamma-rays are emitted after the emission of prompt neutrons. A Monte Carlo sampling is performed over an initial yield distribution Y(A,Z,TKE) to select an initial pair of light (Al,Zl) and heavy ( A h ,Zh) fragments, and total kinetic energy TKE. The total excitation energy TXE available a t scission is obtained as Q-TKE. How does TXE get shared among the light and heavy fragments remains an important and unsolved question t o date. The excitation energies Eth in the light and heavy fragments at scission are the sum of an intrinsic and deformation energy components. So, even under the assumption of thermal equilibrium at scission, different deformation energies for the two fragments would lead to a non-equitemperature law, i.e., # Th.In the present work, we consider the ratio parameter
where (Tl,h) are the average effective scission temperature of the light and heavy fragment, resp. In the following, RT is used as a free parameter. 3.2. Numerical Results
We have applied this formalism to the study of the first-chance neutroninduced fission of 235Uand spontaneous fission of 252Cf.697 In the present work, we have used the experimental yields Y(A,TKE) from Hambsch' for the neutron-induced fission of 235Ufor Ei,,=0.5 to 6.0 MeV, and their interpretation as three fission modes as proposed by B r o ~ a and , ~ noted "S1" for Standard-I, "S2" for Standard-I1 and "SL" for super-long. Both S1 and S2 modes are characterized by asymmetric mass distributions and high-TKE values, while the SL mode is characteristic of a symmetric mass distribution and lower TKE values. Figure 2 represents these yields for 0.5 MeV incident neutron energy. Note that the S2 mode is largely dominant (74.25%), S1 representing 25.53% and the SL mode corresponding to very rare events (only 0.22%). We have considered two casesb: R ~ = l . 0and R ~ = 1 . 3/ 1.2 / 1.0 for the bfor more details, please read Ref.'
144 S2
SUM
00 S1
00 40 60 80 100120140160180200 Fragment mass (amu)
40 60 80 100120140160180200 Fragment mass (amu)
Fig. 2. Experimental yields Y(A,TKE) of n(0.5 MeV)+ 235 U and their decomposition in fission modes SI, S2 and SL. The scale on the z-axis is different for each plot. The S2 mode largely dominates (74.25%), with the SI following (25.53%) and the SL mode contributing very little (0.22%) to the full distribution, as can be seen by the shadow in the symmetric mass region in the upper-right plot.
SI, S2, and SL modes. The average prompt neutrons multiplicity v calculated with RT=1.0 is 2.49, close to the experimental data of Nishio10 (2.47) for thermal neutrons and Miiller11 (2.46) for 0.5 MeV incident neutrons. However, more neutrons are emitted from the heavy fragment (F/i=1.38) than from the light fragment (Fj=l.ll), in stark contrast with experimental evidence. By using a value of RT greater than 1.0, more neutrons can be emitted from the light at the expense of the heavy fragment. By following in detail the entire decay sequence, more detailed information is now made accessible, such as P ( v ) , v(A,TKE), n-n correlations, etc. As an example, Fig. 3 shows the calculated distribution v(A,TKE). More results can be found in Ref.7 and a full-length paper is in preparation.
4. Discussion 4.1. Experiment As the existence of scission neutrons remains highly controversial and debated, experimental efforts to answer this question are of great importance. Clever works in this direction have been discussed in this Conference, as
145 240 220 5- 200
S2 -
SUM -
S1 -
SL -
I 180 LU 160 H 140 120 100
240 220 y 200 1 180
m v; 160 i- 140 120 100
40 60 80 100120140160180200 Fragment mass (amu)
40 60 80 100120140160180200 Fragment mass (amu)
Fig. 3. Average prompt neutron multiplicity as a function of fragment mass and total kinetic energy, for each mode and their weighted sum, at En=0.5 MeV.
for instance by Gonnenwein.12 Detailed and more complete experimental data on prompt neutrons are also lacking. As demonstrated above, only detailed distributions such as Y(A, Z, TKE) and v(A, Z, TKE) can help understand the physics near and at scission. Average quantities such as V and N(E) are not sensitive enough, and can often be fitted with very few parameters within the Madland-Nix formalism. 4.2.
Theory
The " sudden approximation" model calculations presented above as a mechanism for the emission of scission neutrons can be extended in several ways: all possible fragmentations and scission configurations should be taken into account, and more "realistic" scission nuclear configurations such as given by Hartree-Fock or macroscopic-microscopic calculations could be studied. Monte Carlo calculations of the prompt neutrons emitted during the evaporation stage of the excited fission fragments can also be improved. In particular, the sensitivity of the results to the model input parameters values have to be carefully analyzed. A more rigorous account for the competition between neutron and gamma-rays emissions could be implemented within a Hauser-Feshbach formalism, as initiated by Perez-Martin.13 Obviously, this work would involve more assumptions regarding the initial spin distribution in the fragments at scission, and the low-lying levels structure
146
in the product fragments. The extension of this approach t o higher neutron incident energies would represent a challenge as the number of parameters dramatically increases passed the second-chance fission threshold. 4.3. Evaluation
So far, all libraries of evaluated nuclear data use the Madland-Nix model as a basis for calculating the average multiplicityi7p and spectrum N ( E ) .However recent applications (e.g., non-proliferation) are requesting evaluations on prompt neutrons multiplicity distribution P( v) and neutron-neutron correlations for instance. An extension of the current ENDF format used in evaluated libraries will have t o be introduced. A final but very important issue concerns the appropriate handling of existing experimental data. In regards to fission data, other than fission cross sections, the current status of the EXFOR database is really poor and disorganized. An important effort needs to be devoted t o cleaning up the database and making sure that most of the existing fission data get there.. . before some data go missing. References 1. D.G. Madland and J.R. Nix, Nucl. Sci. Eng. 81,213 (1982). 2. G. Petrov, in Third Int. Workshop on Nuclear Fission and Fission-Product Spectroscopy, Cadarache, France, p.205 (2005). 3. I. Halpern, in First Symposium on Physics and Chemistry of Fission, IAEA, Vienna, 1965. 4. N. Carjan, P. Talou and 0. Serot, Nucl. Phys. A792, 102 (2007).
5. N. Carjan, H. Goutte, M. Rizea, 0. Serot and P. Talou, to appear in Proc. of the Int. Conf. on Nuclear Data for Science and Technology, Nice, France (2007). 6. S. Lemaire, P. Talou, T. Kawano, M.B. Chadwick and D.G. Madland, Phys. Rev. C 72,024601 (2005); Phys. Rev. C 73,014602 (2006).
7. P. Talou, to appear in Proc. of the Int. Conf. on Nuclear Data for Science and Technology, Nice, France (2007). 8. F.-J. Hambsch, private communication (2006). 9. U . Brosa, S. Grossmann and A. Miiller, Physics Reports 197,167 (1990). 10. K. Nishio, Y. Nakagome, H. Yamamoto and I. Kimura, Nucl. Phys. A632, 540 (1998). 11. R. Miiller, A. Naqvi, F. Kappeler and F. Dickmann, Phys. Rev. C 29,885 (1984). 12. F. Gonnenwein, this Conference (2007). 13. S . Perez-Martin, this Conference (2007).
FISSION PROMPT NEUTRON AND GAMMA MULTIPLICITY BY STATISTICAL DECAY OF FRAGMENTS S. PEREZ-MARTIN*, S. HILAIRE, E. BAUGE Commissariat l’nergie Atomique, DAM Ile-de-France DPTA/Service de Physique Nuclaire Bruyres-Le-Chtel 91297 ARPAJON CEDEX, France * E-mail: [email protected] The fission prompt neutron and gamma multiplicities for neutron-induced fis sion of 235Uare calculated over a large range of neutron incident energies (0.08-30 MeV). We use the TALYS code to follow the statistical decay of all the produced fission fragments by using neutron-gamma emission competition. The neutron multiplicity probability distribution P ( u ) as well as the average prompt neutron energy are also computed. The quality of the hypothesis made for the partitioning of the excitation energy of the fissioning nucleus between the two fragments is challenged though comparison with both experimental data and theoretical results coming from other models. The influence of the parameters relevant to the statistical model (level densities, gamma-ray strength functions, optical model) is also assessed.
1. Introduction
It is well known that the knowledge of fission prompt neutron multiplicity and spectrum is a key-point in the development of nuclear applications. For that reason we consider of great interest the implementation of a systematics that evaluates both magnitudes as accurately as possible. The backbone of this study in the nuclear reaction code TALYS [l], which follows the statistical decay of all fission fragments. TALYS is based on the optical, the pre-equilibrium and the Hauser-Feshbach models. It calculates partial and total fission cross sections, gamma and neutron production cross sections, total exclusive ( n ,y), (n,n ) ,(n,2n),(n,3n), ..., cross section and their corresponding emission spectra, among other quantities. TALYS not only describes nuclear reactions, but it computes the decay of an initial populated nucleus. An important advantage of our procedure 147
148
compared t o other models is that we explicitly account for the ylneutron emission competition in the fission fragment decay. For the present work, fission mass distributions have been taken from Wahl’s evaluation because they are given in terms of the specific incident neutron energy. Another reason to choose Wahl’s evaluation vs other evaluations, is the multi-fission chance. When the incident neutron energy is high, for instance 30 MeV there are 6 open channels leading to fission. All those open fission channels should be taken into account to collect neutrons and gammas with their corresponding fragment distributions which are all provided by Wahl’s evaluation. The choice of the initial fission fragment population is a direct consequence of both the partition of the total available fission energy and the total kinetic energy (TKE). For sharing the total excitation energy between the two fragments the equal nuclear temperature approximation has been used. For TKE, Gaussian parameters were deduced from experimental data.
2. Methodology
Since the incident neutron energy ranges from 0.08 MeV to 30 MeV, the set of fissioning nuclei associated with each incident neutron energy should be determined first. By using TALYS with parameters adjusted to reproduce experimental results, we obtain the total fission cross section, as well as all the partial and the exclusive fission cross sections (n,f) , (n,nf) , (n,pf)7 (n,2nf) , (n,V f ) , (n,df) 7 etc. Those exclusive fission cross section are important when second, third, etc fission chance channels are open because, in this case, the neutron multiplicity must account not only for the neutrons emitted by fragments, but also for those emitted before fission.
2.1. Fission Fragment Population
In order to populate the initial fission fragment distribution, quantities such as the Total Excitation Energy (TXE) to be shared between the fission fragments and the Partitioning Energy must be determined. Both quantities are related to the fission energy released and to which fraction of that energy is turned into kinetic energy. In the fission reaction, the released energy is calculated from the mass difference between the compound nucleus and the two fission fragments. For the first chance fissioning nucleus 236U,the excited energy is
149
The above expression is generalized for the subsequent fissioning nuclei, E*
E*
2 3 4 ~=
2 3 5 ~ = Einc. -
(E($nf 1)
(2)
s ~ ( ~+ ~ ~ ~+ uBi)~ ~ ( ~ 2~ ($""") ~ ~ u )-. ~ ~
~ ( ~ 3(3)% )
where &?nf) is the average energy of evaporated neutrons prior t o fission. The energy released in fission for first fission chance reads
QY;:~
=M
+ ~+ i ~ ~ .
( ~ ~ ~ u ) B~ ~~ ( ~- ~ M ( F ~ F ~u ) C) ~- M ( F F ~ ) ~ ~(4).
Since the total excitation energy is the difference between the energy release in fission and the total kinetic energy, the TKE should be estimated as precisely as possible. For that purpose, we have chosen to use experimental results and to assume a Gaussian distribution, where both the mean value T K E = T K E ( A ) and the width u = (T ( A ) depend on the fragment mass. Such parameters can be deduced from the experimental results of Ref. [2-41. In Fig. 1, the energy released in fission and total kinetic energy are plotted as a function of the fission fragment mass for a 0.53 MeV incident neutron.
I
100'
70
"
80
90
n(0.53 MeV) +235U
" ' " 100 110 120 130 Fission Fragments Mass
I
"
140
150
'
160
Fig. 1. Fission Q-value and T k E as a function of the mass number of t h e FF.
Then the total excitation energy (TXE) available for fission fragments is:
TXE =Q -TKE.
(6)
150
P a r t i t i o n i n g a p p r o x i m a t i o n When TKE is smaller than the Q-value (energy conservation condition) the partition energy between light and heavy fragments is obtained by assuming equal nuclear temperature (V = a T z ) for the compound nucleus, light and heavy fragments. This leads t o the expression: E i = TXE-
1
l+Z
(7)
where a H and a L are the level density parameters of the heavy and light fragments, respectively. 2.2. Prompt Neutron and Gamma Multiplicities One way of estimating multiplicities is through production cross section. Therefore the interesting quantities for f17 and fln are the y and neutron production cross sections, as well as the total exclusive cross section. In order to also deduce the average energy of the emitted neutrons and gammarays, we need the exclusive emission reaction spectra, and the outgoing neutron and gamma spectra. There are two converging ways t o compute the average neutron and gamma multiplicities. The first one relies on the global neutron production cross section, first summed over all fission fragments of the corresponding fissioning nucleus, and then summed over all fissioning nuclei provided following fission chances are open,
where Y F F is the fission fragment yield, init.pop. is a normalizing factor, n$N is the number of pre-fission neutrons associated to the fissioning nucleus F N , and nTN is the number of pre-fission gammas of F N . Pre-fission N e u t r o n s For Uranium isotopes the number of evaporated neutrons prior to fission nFN is easy to compute, as the neutron number difference between the compound nucleus and the fissioning isotope; but in the case of another chemical element, we have to adopt an approximation. Let us for example look at the 233Pafissioning nucleus, that be reached trough three distinct reaction channels from the 236Ucompound nucleus :
151
+ 2n +233 Pa 236U -+d + n +233 Pa 2 3 6 + ~ t +233 Pa
2 3 6 + ~ p
The probability of each reaction to occur can only be estimated by looking at the exclusive (n,2 n p f ) , (n,n d f ) and (n,t ) cross-sections. Thus, the number of those pre-fission neutrons is the fraction 2 u c n 3 2 n pP ~a ~ ~ u c n 3 n d r ~ . ufiss
In that case, the prompt neutron multiplicity contribution of that fissioning nucleus will be:
Neutron Multiplicity I1 For the second calculation method, we deduce the prompt neutron multiplicity in terms of the multiplicity probability distribution P ( u ) which is deduced from the sum of exclusive reactions which generate u prompt neutrons. FN
pa,, ( 0 ) =
CYFF * QyFN-FF FF
FN
pa,, (l) = C
Y F F FF
*
Qn FN-FF
+ ,pFN-FF
+
QrN-FF
+ ...
init . p o p F N - F F
+ Q Fp nN - F F
+
,,rnN-FF.
init .pop.FN-FF
+ ...
(12)
(13)
In the above formulae, u z N - F Fis the exclusive reaction cross section for decay to the x channel.
Fission Multiplicity Distribution It is important t o notice that P i ? (Eqs. 12 and 13) is the fragment multiplicity distribution, but not the fission multiplicity distribution P F N ( u ). That difference means that for each fission, there are contributions from the two fragments, so that the global probability distribution results from the combination of the probability distributions of both fragments.
Another point worth mentioning is the fact the nkFpre-fission neutrons should be taken into account by offsetting the corresponding multiplicity distribution by nkF units. Therefore the probability to emit zero neutrons should be zero for chances higher than the first.
152
(15)
P F N ( 0 )= 0
PFN
n (nFN
+ ')
,FN-FF fiss
= ,TOTAL fiss
pFg,,('1
FN * 'FF,,, ('1 + pFg,, ('1 * 'FZ,,
(O)
(17)
Finally, when averaging over all fissioning nuclei, the global multiplicity distribution made P ( v ) is computed as a linear combination of the partial distribution P F N(v) coming from every FN. P(,) =
c
f iss *PFN(Y)
FN
~
u~~~~~
Then, the average neutron multiplicity can be deduced
c N'
D=
UP(,).
u=l
2.3. Neutron Average Energy The average prompt neutron energy of a given fission fragment is estimated by using the center of mass (C.0.M) emission spectra provided by TALYS, which is averaged over all energy bins:
bin
where
(
AuFN-FF
7;"
)
is the energy spectrum and
(
c[
A ~ F N - F F
7;"
)
AEbin] him
is nothing but the neutron production cross section. For the average pre-fission neutron energy c?Inf), we take the same procedure, using the neutron spectrum of the ( n ,nf),(n,npf), (n,2nf), etc reactions. Actually such pre-fission neutron energy is also needed t o estimate the excitation energy of the residual fissioning nucleus in second and successive fission chances, as we mentioned in Eqs. (2) and (3).
153
For gamma ray, only the global gamma production energy can be computed, and it is done in the same way as for neutrons, using the gamma auFN-FF
spectrum(
)
.
/ bin
\
Finally the global prompt neutron energy for the neutron induced reaction is an average over all the fission fragment and pre-fission neutron sources fiss
-
E n
=
C F N
+%+ * "TOTAL
cFN "TOTAL
FN-FN
vn
En
+ v Fp NN- F~ Np N + L /PF NN
,
where $ : is the average kinetic energy of evaporated neutrons prior to fission. When the number of pre-fission neutrons is not a integer value, that is, when there are different ways to produce the corresponding fissioning nucleus, the energy of those pre-fission neutrons is calculated as (example of 233Pa)
3. R e s u l t s We have performed calculations with neutron energies ranging from 0.08 t o 30 MeV. Since TALYS contains many options and parameters for driving its statistical model module, we have studied the influence of level density parameters and y-ray strength functions on prompt neutron and gamma emission. Therefore we have performed calculations with either the Fermi gas phenomenological level density model and a Kopecky-Uhl generalized Lorentzian as a y-ray strength function, or with microscopic level densities and Hartree-Fock-Bogoliubov y-ray strength functions. The first set of calculations is called TALYS-1 and the second one TALYS-2. In Fig. 2 we show (as squares) the mean value of the total excitation energy available for both fragments as a function of the fragments mass for a 0.53 MeV 235U neutron induced fission. The mean individual excitation energy as a function of fragment mass is drawn with crosses. We can see the effect of the energy partition approximation. The sharp drop in the A=120-130 region is due to the difference between the energy release in fission and the total kinetic energy as shown in Fig. 1, where the Q-value is almost constant while the TKE increases.
154
I n(0.53MeV) + *=Li I
45 40 35 30
2 rw 25 20 15 10 5 0 70
80
90
100
110
120
130
140
150
160
Fission Fragments Mass
Fig. 2. Total excitation energy and fragment excitation energy as a function of the mass number of the FF.
For 0.53 MeV incident neutrons, in Fig. 3 we plot the prompt average neutron multiplicity as a function of the fission fragment mass. We compare our results with evaluated data (Ref. [5,7]). The general trend of the data (the sawtooth shape) is accounted for by our results. There are however, sizable differences, but the largest disagreement occur in the symmetric fission mass region where the yield is low, thus limiting the effect of that mismatch. There are also some differences in the region of the light and heavy asymmetric fission, but they are smaller and of opposite directions (overestimation in the heavy asymmetric fission region and underestimation in the light fragment region). The effect of phenomenological or microscopic parameters in the TALYS calculation results in slight differences for fragments near A = 91 and 143. Although these changes seem quite small, they produce a change of D from 2.27 for the TALYS-1 calculation t o 2.34 for TALYS- 2. The prompt neutron multiplicity distribution is shown in Fig. 4. In the left panel, we compare our results with the experimental data of Diven et al. [6] for 0.08 MeV incident neutron. Notwithstanding the difference of D, both distribution look similar, and the maximum value of the neutron multiplicity distribution is located at the same value v = 2. In the right panel, results for 0.53 MeV incident neutron energy are shown. Here our results are compared with those coming from Monte Carlo simulation [8]. Large differences are observed between our results and those of Ref. [8].For example, our multiplicity distribution peaks at v = 2 while Lemaire’s peaks a t v = 3. Moreover, our calculation (TALYS-1) produces an average prompt neutron multiplicity value of D = 2.27 whereas the Monte Carlo model estimates D = 2.73 using the same equal temperature energy partition hypothesis. The experimental result (Ref. [4]) gives D = 2.46.
155
I
n(0.53 MeV) + 235U
,~TALYS D t ? i a ~ ~ l i o TALYS Microsc.
I
I\
"
80
70
100 110 120 130 140 Mass Number of Fission Fragments
150
90
160
Fig. 3. Average neutron multiplicity as a function of the mass of t h e FF for n ( 0 . 5 3 M e V ) +235 U . Experimental results are taken from Ref. [5] and [7]. FF yields are drawn in dashed line.
1 I
"
"
"
'
I
n(o.08 MeV)
"
~
I
+'%
I
[ "
"
"
'
n(0 53 MeV) + 235U
1
I
Fig. 4 . Neutron multiplicity distribution P ( v ) . In the left-hand side results with 0.08 MeV incident energy are compared. In right-hand side different calculations are plotted for 0.53 MeV neutron incident energy.
The prompt neutron and gamma average multiplicity is shown in Fig. 5 as a function of the incident neutron energy. As indicated above, the two ways of computing the average neutron multiplicity (though neutron production cross sections (method I) or though exclusive channels cross sections(method 11))coincide for all incident neutron energies. The effects of the statistical decay ingredients of TALYS (TALYS-1 vs TALYS-2 results) on the average neutron multiplicity are quite limited, but they are not negligible. The prompt gamma average multiplicity zZ7 seems to be the most
156
sensitive to statistical decay parameters, and TALYS-2 results in a larger number of emitted gammas. Overall, the use of microscopic parameters in TALYS produces neutron and gamma multiplicities closer to experimental data.
TALYS-1 Met. I TALYS-1 Met. II P(v) • TALYS-2 Met. I P(v) Evaluation (F.MANERO.V.A.KONSHIN)
Z 3
10
15
E/nc(MeV)
Fig. 5. Neutron and gamma multiplicity as a function of the incident neutron energy. Evaluation of neutron multiplicities [9] are also plotted.
The average prompt neutron energy is shown in Tab. 1. Our results are smaller than those of Los Alamos [10] and Monte Carlo [8]. As it stands now, our model produces less neutrons which are on average less energetic than the other models. Fig. 6 shows the average prompt neutron energy as a function of the fission fragments mass. Table 1. Prompt fission neutron average multiplicity and energy. We compare our neutron multiplicity calculation results with those of Monte Carlo [8] and experiment [4]. n(0.53MeV) +235 U TALYS-1 2.27 TALYS-2 2.34 Monte Carlo S. Lemaire 2.73 Mller et al. 2.46
n(0.53MeV) +235 U
C.O.M e(MeV)
TALYS-1 TALYS-2 Monte Carlo S. Lemaire Los Alamos model
1.02 1.10 1.241 1.269
157
We can see that, from A=115 to 130, the neutron multiplicity decreases and the average neutron emission energy has a small peak which starts to decline when neutron multiplicity increases from A=130. This is the region where a large difference exists between TALYS-1 and TALYS-2 calculations.
I
n(0.53 MeV) +235U
1
Neut. Mult. o TALYS-1
2
o TALYS-2
z 5 (D
a
3
12
-1g _. n.
u"
4 & 0.5
0
i
70
80
100 110 120 130 140 Mass Number of Fission Fragments
90
150
' 0 160
Fig. 6. Average neutron emission energy obtained with both sets of parameters used in TALYS. Neutron multiplicity is also drawn with a wide line.
4. Conclusions and Projects
We have developed a calculation scheme to compute neutron and gamma multiplicities which accounts for the competition between neutron and y emission. By letting the fission fragment population decay in the statistical model module of TALYS, we simultaneously collect the emitted neutron and gammas, and process them to produce the usual average prompt neutron(gamma) multiplicity, multiplicity distribution, and average energy. Our calculations of these quantities are compared with available experimental data as well as with results from other models. We have investigated the influence of the parameters of statistical decay (level densities and y-ray strength functions) and found that moving from phenomenological parameters to microscopic ones, results in differences that are not negligible. For the moment, although general trends seem to be well described by our model, we do not reach the level of precision needed for applications.
158
T h e main reason for this mismatch is most likely the hypothesis used for populating t h e initial fission fragment distribution. Our next task will be t o implement more realistic initial fragment populations in our model and investigate their influence on prompt neutron and gamma distributions.
References http://www.talys.eu/ TALYS-0.72 A nuclear reaction program A.J. Koning, S. Hilaire and M. Duijvestijn (2006) 2. H. W. Schmitt, J . H. Neiler, and F. J. Walter, Phys. Rev. 141, 1146 (1966) 3. J . W. Meadows Phys. Rev. 177, 1817 - 1825 (1969) 4. R. Mller, A. A. Naqvi, F. Kppeler, and F. Dickmann Phys. Rev. C 29, 885 905 (1984) 5 . K . Nishio, Y. Nakagome, H. Yamamoto, and I. Kimura, Nucl. Phys. A632, 540 (1998) 6. B. C. Diven, H. C. Martin, R. F. Taschek, and J. Terrell, Phys. Rev. 101, 1012 (1956) 7. E. E. Maslin, A. L. Rodgers, and W. G. F. Core Phys. Rev. 164, 1520 - 1527 (1967) 8. S. Lemaire, P. Talou, T. Kawano, M. B. Chadwick, and D. G. Madland Phys. Rev. C 72, 024601 (2005) 9. F.Manero, V.A.Konshin. Atomic Energy Review, IAEA; Vol.10, p.637 (1972) 10. D. G. Madland and J. R. Nix, Nucl. Sci. Eng. 81, 213 (1982) 1.
This page intentionally left blank
Fission Theory
This page intentionally left blank
STRUCTURE A N D FISSION P R O P E R T I E S OF A C T I N I D E S WITH THE G O G N Y FORCE H. GOUTTE, J.-F. BERGER,
J.-P. DELAROCHE, and M. GIROD C E A / D A M Ile de France B P 12, 91680 BruyZlres-le-Chdtel, France [email protected] A. DOBROWOLSKI Maria Curie-Sktodowska University, Lublin, Poland J. LIBERT Institut de Physique Nucldaire INA'PS-CNRS/UniversitC Paris-Sud, 91406 Orsay Cedex, France Structure properties of many even-even actinides have been calculated using the Gogny D1S force and the Hartree-Fock-Bogoliubov approach as well as the configuration mixing method. Results on rotational states, shape isomers and fission barriers are here discussed. Keywords: Nuclear Fission; Actinides; Mean-field approach
1. I n t r o d u c t i o n
One major issue in nuclear physics is to develop a consistent model able to describe on the same footing the different aspects of the fission process, i.e. properties of the fissioning system, fission dynamics and fragment distributions. Among the different approaches available nowadays, microscopic ones are well-suited as they may provide predictions in mass regions where there are no experimental data. They can also serve as a guide for reaction models. Fission studies have been performed, in which the sole input is the nucleon-nucleon effective force of finite-range D1S of Gogny. Results for even-even actinides are discussed, namely rotational states, shape isomers and fission barriers. 161
162
2. Formalism The Hartree-Fock-Bogoliubov (HFB) approach and its extensions represent a convenient and flexible tool for studying structure properties of heavy nuclei. Applied with the finite-range effective Gogny force D1S [1,2], such methods allow one t o treat on the same footing mean-field properties and pairing correlations. When constraints are employed, deformed and/or rotational states (from the so-called cranking approach) can be obtained from the minimization principle:
(4(tq},1)Ifi - A N G -
-
c
- wzjz14({q),1)) = 0,
AijQij
(1)
ij
with (4({q},l) lfi(-al4(tq},1)) = N ( Z ) ,
(2)
= qij,
(4((4},1)IQijl4({q},l))
m.
( 4 ( { q } , l ) l j z l 4 ( { ~ ~= ,l~)
In Eqs. (1) and (2), I? is the Hamiltonian, Qij a multipole moment operator, j zthe z-component of the angular momentum operator, I the total angular momentum, and Ak and w, are Lagrange multipliers. Beyond mean-field calculations have also been performed by use of a 5-dimensional collective Hamiltonian (5DCH) derived from the Generator Coordinate Method with the Gaussian Overlap Approximation for the five quadrupole coordinates, that is for axial q20 and triaxial q 2 2 quadrupole deformations and the three rotation degrees of freedom [3]. 3. Rotational states
Rotational states have been obtained by the two approaches presented above, namely cranking HFB and 5DCH [9]. Kinetic moments of inertia J ( ' ) have been deduced from J(1) =
m W
1
(3)
where the rotational frequency is taken as
E(I
+ 2) - E ( I - 2) ,
(4) 4 with E(1) the excitation energy of the rotational state with angular momentum I. Results for kinetic moments of inertia for the yrast normal deformed and superdeformed bands of even-even Plutonium and Fermium isotopes W =
163
are shown in Figs. 1 and 2, respectively. Predictions for cranking HFB and 5DCH approaches are compared with experimental data [4-81. Each theoretical method has its own figure of merit: the cranking approach handles the weakening of pairing correlations under the influence of increasing rotational frequency, while 5DCH includes implicitly the effects of long range correlations due to the rotation-vibration coupling, but ignores Coriolis antipairing correlation. At normal deformation, both approaches give good results a t low frequency. The main disagreement with experimental data is found at high spin, where theoretical back-bending takes place at too low frequencies. For superdeformed rotational bands, we observe in Fig. 2 a 10% difference between theory and data for the 240Puisotope. More experimental data are needed to challenge the reliability of our' approaches for kinetic moments of inertia for superdeformed bands in actinides.
>
;120 80
5 m
f 120
:80
y1
120
I
c
80
E E 120
-E
80
Y
120 80
8
00
0
t
005
, 01
M
,
,
, ,
,,,, ,
015 02 025 03 w (rotational frequency in MeV)
00
005
01
015 02 025 03 w (rotationalfrequency !n MeV)
Fig. 1. Kinetic moments of inertia for the yrast normal deformed bands of even-even 232-24sPu (left panel) and 242-258Fm (right panel) isotopes as functions of the rotational frequency. Results from cranking HFB and 5DCH calculations are shown as black dots and open circles, respectively. Experimental data are shown as stars.
164 -
,
160.
Pu
&ae*b.
120I
200 160.
,
.................. .................... .................. * ...................: .................. ..... ............... ................ u2
,
I
,
I w
,
I
>
I
,
I
Pu
6bbbb
120-,
-~-
I
1120-,
I
I
,
200
,
I
,
I
,
I
,
I -
"P U
> 160-8 e b b . o . .
. -.
200-
~
5
,
6
160:
c,
ZIZO-,
I
200-
,
-.-I 1 2 0 - ,
-.
200-
160 . 6
E;%:'
5
,
I
I
,
,
I
,
I
I
,
,
I
I
,
I
,
1 -
u120-
;200-
Y 1 6 0-
I
,
I
,
24*
' Fm
I
,
I
I
-
I
,
I
,
1 -
-
%n
'
;120-, al 200-
I
'
'
246
I
'
'
'
Fm
a a$.
q60-a
-
I
"pu
-
160
-
..................... ....................... ................... .......!?.......... I ....................
' 200. 160-eb'.
,
I
I
I
,
,
I
248
I
-
,
I
,
I
,
I -
,
I
,
I
,
I -
=inFm
-
a&bo I
,
I
,
I
,
I
,
I
,
1
-
"^pU
-
I**+*
m120-,
1
:200-
,
I
,
I
242
,
I
,
I
,
I -
pu
-
~160-~.a8bb............ I
2 200-
,
I
,
I
,
I
,
I
,
1 -
I
,
I
,
1:
I
,
-
%
I
160-e6eee..
~
1
2
-
0
~I
I
,
I
,
%6Pu
160-beeee 120-, 00
,,
I
005
,
I
01
,
I
015
,
02
1
025
~
17
03
Fig. 2. Same as Fig. 1 for the superdeformed yrast bands of even-even 232-246Pu (left panel) and 242-258Fmand 250-254No (right panel) isotopes.
Improving predictions may imply: i) exact projection on particle number and/or ii) treatment of octupole correlations, which may delay the frequency at which backbending occurs. 4. Shape isomer
Superdeformed positive parity collective states have been obtained for 55 even-even actinides from 5DCH calculations [9]. Results for the shape isomers are shown in Fig. 3, where the excitation energy of the shape isomers is expressed with respect t o normal deformed lowest energy states. A good agreement between predictions and experimental data is found in T h , U, Pu, and Cm isotopes. A global lowering of isomer energies is predicted as A increases. Superdeformed states are even found to be lower in energy than normal deformed states in 242,244Fm and 2 5 0 N ~However, . as these states are only a few hundred keV below the octupole unstable outer barrier, they may not survive as bound states. Such issues are of contemporary interest since superdeformed ground states of superheavy nuclei have been
165
Fig. 3. Excitation energy (in MeV) of the shape isomers above respective normal deformed ground states for even-even actinides.
predicted in relativistic mean field calculations [10,11] and their stability discussed [ 121. From WKB calculations, we have found that: i) shape isomers in T h and U decay by y-emission, ii) fission and y-back decay are competing for Pu and Cm, and finally iii) shape isomers in Cf, Cm and No decay through fission [9]. 5 . Fission barriers
Nuclear fission is a complex process which involves the evolution of a nucleus over a multidimensional landscape. The problem of determining the most probable one-dimensional effective nuclear path to be used in reaction models has been the subject of scientific effort since the discovery of the nuclear fission phenomenon. In most self-consistent mean-field studies of fission barriers, nuclear shapes are generated by only the axial mass quadrupole moment, and all other multipoles take on values that minimize total energy. A large-scale study along this line has been made recently to refit a new Skyrme-typeinteraction on fission barrier heights (with a rms deviation on the primary barriers of 772 keV) [13]. In Ref. [9],we have calculated such static least energy fission barriers for 55 even-even actinides from "'Th to 2'2No. Results for 250-262N~ isotopes are plotted on Fig. 4 as functions of the quadrupole deformation parameter
166
,b’ defined in [9]. In these calculations, at low elongation the triaxial degree of freedom has been left free, whereas parity has been broken at large elongation. Solid lines correspond to axial shapes, dashed lines to triaxial ones and dotted ones to asymmetric shapes.
Fig. 4. Potential energy curves as function of the axial quadrupole deformation parameter for No isotopes. Thin solid lines are for axial barriers; dashed lines for triaxial inner barriers and dotted lines correspond to mass asymmetric outer barriers.
The main features are: i) triaxial inner barriers are systematically lowered by up to 4 MeV when compared to the axial ones, ii) the outer barrier i s found to be asymmetric for systems with N < 152 and symmetric for more neutron-rich systems, and finally iii) super deformed minima appear to be washed out for N > 156. These features are illustrated here in the case of Nobelium isotopes, but they are common features of all the seven studied isotopic chains. In order to take into account mass parameters in the definition of the fission barriers and eventually to smooth out discontinuities in high-order muitipoles, least action paths (LAPS) have been determined. For that purpose, two-dimensional landscapes have been first obtained from HFB cal-
167
culations with constraints on i) axial and triaxial quadrupole deformations between the first well and the superdeformed one, and ii) axial quadrupole and octupole deformations from the superdeformed minimum up to scission.
0.0
Q.2
0.4
0.6
0.8
1.0
1.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Fig. 5. Potential energy surfaces as functions of quadrupole axial (/3o) and triaxial (Pi] deformation parameters for even-even isotopes of Th (left panel) and U (right panel) isotopes. Least energy and least action paths from the normal deformed to the superdeformed wells are shown as thick-grey and thick-black lines, respectively.
Least action paths in the axial and triaxial plane between the first well and the superdeformed one are displayed in Fig. 5 as black lines for Thorium and Uranium isotopes and compared with least energy paths displayed as grey thick-lines. In heavy Uranium with A > 232 LAPs go through axial saddles, whereas they go through triaxial shapes in the lighter isotopes. Axial LAPs are also predicted in 232-236Thj 238-244pUi 240-246Cm and in all the Cf, Fm and No isotopes exhibiting shape isomers. Least energy and LAP fission barriers plotted in Fig. 6 for 238U and 226Th also illustrate how important is the role played by collective masses. In the LAP approach presented above, fluctuations around the path are not taken into account. In a more elaborated approach, vibrational collective movements in the direction perpendicular to the fission path have been considered. In practice we have i) defined a curvilinear absciss along the static path, ii) defined the transverse directions all along the path, iii) calculated the inertia and potential energies in these directions, iv) calculated the transverse vibrations using the Generator Coordinate Method and the Gaussian Overlap Approximation, and finally v) defined the barrier as
168
-1780
F
-1784
-
-1788
-
E
F
W
-1792l
0
40
,
,
80
,
I
120
,
I
160
.
l
~
I
,
240
2W
280
I
l
0
!
l
40
.
l
80
.
l
120
.
l
'
160
l
2W
,
l
240
'
I
280
Fig. 6. Fission barriers as functions of the axial quadrupole moment for 238U(left panel) and 22sTh (right panel) isotopes. Least energy paths (solid lines) are compared to least action paths (dashed lines).
10
g4
'-W--
0
1 ..
d-. 2 0
40
80
120
% (b)
160
200
240
qm (b)
Fig. 7. HFB energies as functions of the quadrupole moment for 238U (left panel) and zzsTh (right panel) isotopes. Least energy paths (solid lines) are compared with dynamical barriers, which take into account transverse vibrations (dashed lines). Curves have been arbitrarly shifted in energy to coincide at the superdeformed minimum.
the lowest energy vibrational states for all curvilinear abscisses. Results are plotted in Fig. 7 for 238Uand 226Th.In the two studied nuclei, the dynamical barrier is found to be the lowest one, the energy difference being almost 1 MeV for large elongations. 6. Conclusion
Mean-field based calculations have been performed in many even-even actinides with the D1S Gogny force as sole input. Experimental results collected over the years have been used to systematically challenge these predictions, in particular rotational states and shape isomers. The role played
169 by t h e mass parameters a n d by t h e transverse vibrations along t h e fission p a t h has been discussed. Lifetime calculations will be performed in a near future.
References J. Decharg6 and D. Gogny, Phys. Rev. C21,1568 (1980). J.F. Berger, M. Girod, and D. Gogny, Comp. Phys. Comm. 63,365 (1991). J. Libert, M. Girod, and J.-P. Delaroche, Phys. Rev. C60,054301 (1999). R.B. Firestone, Table of Isotopes, 8th edition, edited by V.S. Shirley, John Wiley and sons inc., New York (1996). 5. R.D. Humphreys et al., Phys. Rev. C69,064324 (2004). 6. T. Ishii et al., Phys. Rev. C72,021301(R) (2005). 7. U. Goerlach, D. Habs, V. Metag, B. Schwartz, H.J. Specht, and H. Backe, Phys. Rev. Lett. 48, 1160 (1982). 8. M. Hunyadi et al., Phys. Lett. B505,27 (2001). 9. J.-P. Delaroche, M. Girod, H. Goutte and J. Libert, Nucl. Phys. A771, 103 (2006). 10. Z. Ren, Phys. Rev. C65,051304(R) (2002). 11. Z. Ren, H. Toki, Nucl. Phys. A689,691 (2001). 12. I. Muntian, A . Sobiczewski, Phys. Lett. B586,254 (2004). 13. M. Samyn, S. Goriely, and J.M. Pearson, Phys. Rev. C72,044316 (2005). 14. A . Dobrowolski, H. Goutte, and J.-F. Berger, Znt. J. Mod. Phys. E16,431 (2007).
1. 2. 3. 4.
This page intentionally left blank
FISSION FRAGMENT PROPERTIES FROM A MICROSCOPIC APPROACH N. DUBRAY", H. GOUTTE and J.-P. DELAROCHE DPTA, CEA/DAM he-de-France, Bruy&res-le-Ch6tel, 91680, f i n c e *E-mail: [email protected] www. cea.fr We use the Hartree-Fock-Bogoliubov method with the Gogny nucleonnucleon effective interaction D1S to calculate potential energy surfaces in the elongation-asymmetry plane, from sphericity to very large deformations, for the 226Thand 256,258,260Fm fissioning systems. Using a criterion based on the nuclear density, many scission configurations are identified, and are used for the calculation of several fragment properties, like fragment deformations, deformation energies, energy partitioning, neutron binding energies at scission, neutron multiplicities, charge polarization and total fragment kinetic energies. Keywords: Hartree-Fock-Bogoliubov approach; fission; fission fragment prop erties.
1. Introduction
Fission is a unique laboratory for studying nuclear configurations far from equilibrium. Recent Hartree-Fock-Bogoliubov calculations using the Gogny force have been performed with constraints on quadrupole and octupole moments from spherical shape up to large deformations. In this subspace of collective coordinates, many scission configurations have been identified using a criterion based on the nuclear matter density, and several fission fragments properties have been calculated from these configurations, and compared with experimental results when available [l]. 2. Theoretical framework
We use the Hartree-Fock-Bogoliubov (HFB) method [2] with the Gogny effective nucleon-nucleon interaction. This finite range and density dependent interaction allows the simultaneous treatment of the nuclear and pairing 171
172
mean fields. We have used the D1S set of parameters [3,4], which is known for its good reproduction of nuclear properties [5,6]. In order to obtain the total energy landscape of the nuclear system in the (elongation, asymmetry) coordinates, we introduce in the main HFB equation additional terms called constraints. This leads to the constrained HFB equation:
d(Plfi- A N f i - A Z 2
- AlO$lO
- A 2 0 0 2 0 - A30$301'p)
= 0.
(1)
The purpose of these constraints is to fix the mean values of the numbers of neutrons and protons of the nuclear system as well as those for the usual multipole operators $10, $20 and $30, namely
The constraint on the dipole moment $10 aims a t imposing a fixed position to the center of mass of the system. By letting q20 and 430 take regularly spaced values, we are able to draw the total energy map of the system in the ( 4 2 0 , 430) plane. Eq. (1) is solved by expanding the quasi-particle operators onto axiallysymmetric harmonic oscillator bases. In this calculations, the conservation of the z-axis symmetry of the system is enforced. For each ( 4 2 0 , 4 3 0 ) couple, the parameters describing the bases are optimized, i.e. they are chosen in order to minimize the total energy. 3. Definition of the scission line
Scission corresponds to the splitting of the compound nucleus into two fragments. In our study, we use the following definition for the scission points: if a point from the fission valley leads to a point in the fusion valley by a small increase of one deformation parameter, this point is called a scission point. Since we are working in a 2-dimensional deformation space ( 4 2 0 , q 3 0 ) , the ensemble of considered scission points forms a line, which we call the scission line. In Fig. 1, the evolutions of the binding energy ( E ) ,the hexadecapole moment ( q 4 0 ) and the density in the neck ( p ~ are ) plotted in
173
Fig. 1. Example of very different symmetric scssion transitions. qg)and represent the quadrupole moment and the HFB energy of the first post-scission point for each fissioning system.
the vicinity of the symmetric scission transitions of 226Thand 256Fm.This clearly shows that the scission transition can be either smooth or sudden, and that a criterion based on the matter density in the neck is a good way to distinguish pre-scission from post-scission configurations. 4. Results
Calculations have been performed for 226Th,256Fm,258Fm and 260Fmnuclei. The corresponding scission lines are plotted on Fig. 2. The total kinetic energies of the fragments computed for 226Th(see Fig. 3) are in rather good agreement with experimental data [7],and the neutron emission multiplicities for 256Fmreproduce the general structure of the sawtooth experimental data [8],even if there seems to be a slightly underestimation of the theoretical values. 5 . Conclusion
The results obtained in the present study of Th and Fm nuclei clearly show that our fully microscopic approach is able to provide a quantitative account of scission properties of actinide nuclei. Several fragment properties have being computed for many different scission configurations, and are found to be in good agreement with experimental data. The description of timedependent fission dynamics and of fragment mass distributions with the method of Ref. [9] will be carried out in a near future. Finally, extensions
174
900
250
300
350
400
450
550
500
,q (b)
Fig. 2. 2001.
i
-
1
Scission lines in the (420, 4 3 0 ) plane.
' I
',
"
1
"
"I
'
I"
1 ' .
1
'
I '
"
' I '
"
1
I
'
1
"
I '
"
'
I
180-
160:
140:
Th exp. 70
Fig. 3.
80
90
100 110 120 130 140 150 A
fw
(am&)
Total fragment kinetic energies in 22sTh.
175
Fig. 4. Neutron emission multiplicity in 25sFm.
of the present calculations to a three-dimensional mesh (qzo, to non-axial nuclear shapes are under consideration.
q30, q4o)
and
References 1. N . Dubray, H. Goutte and J.-P. Delaroche, submitted to Phys. Rev. C (2007). 2. P. Ring and P. Schuck, The Nuclear Many Body Problem Springer-Verlag, New York, 1980, p. 267. 3. J . Dechargh and D. Gogny, Phys. Rev. C21,p. 1568 (1980). 4. J.-F. Berger, M. Girod and D. Gogny, Comp. Phys. Comm. 63,p. 365 (1991). 5. G. Bertsch, M. Girod, S. Hilaire, J.-P. Delaroche, H. Goutte and S. PBru, Phys. Rev. Lett. 99,p. 032502 (2007). 6. J.-P. Delaroche, M. Girod, H. Goutte and J. Libert, Nucl. Phys. A771,p. 103 (2006). 7. K.-H. Schmidt, J. Benlliure and A.R. Junghaus, Nucl. Phys. A693, p. 169 (2001). 8. J.E. Gindler, Phys. Rev. C19,p. 1806 (1979). 9. H. Goutte, P. Casoli, J.-F. Berger and D. Gogny, Phys. Rev. C71,p. 024316 (2005).
This page intentionally left blank
SMOKER AND NON-SMOKER NEUTRON-INDUCED FISSION RATES* I. KORNEEV, I. V. PANOVt Institute for Theoretical and Experimental Physics, B. Cheremushkinskaya 25, Moscow, 117259, Russia. E-mail: Igor.PanovOitep.m T . RAUSCHER, F.-K. THIELEMANN University of Basel, Klingelbergstr. 82, CH-4056 Basel, Switzerland ch E-mail: fktOquasar.physik.unibas. Theoretical neutron-induced fission rates in the temperature range los 5 T(K) 5 10.0 x lo9 are calculated within the framework of the statistical model for targets with atomic number 84 5 Z 5 110 (from Po t o Ds) and for a mass range defined by the neutron and proton drip-lines. Three complete sets of rates have been calculated, utilizing the nuclear data self-consistently for neutron separation energies and fission barriers from Thomas-Fermi, ETFSI and FRDM predictions. A comparison of rates predictions on the basis of different realizations of Hauser-Feshbach formalism was made. Neutron-induced fission rates as a function of temperature are given by analytic seven parameter fits. Keywords: Nuclear reactions; Nucleosynthesis; Fission
Introduction The knowledge of nuclear reaction rates is very important for all investigations of nucleosynthesis or energy generating processes in astrophysics. Highly unstable nuclei are produced in such processes and can then participate in subsequent reactions. Cross sections and astrophysical reaction rates for a large number of nuclei are required to perform complete network calculations which take into account all possible reactions. Despite concerted experimental efforts, most of the involved nuclei are currently not accessible in the laboratory and therefore theoretical models have to *This work is partially supported by grant IB7320-110996 of SNF +I. P. thanks for hospitality Basel University (grant NPA 1538).
177
178
be invoked in order to predict reaction rates, in particular for r-process nucleosynthesis. The prediction of cross sections and reaction rates for astrophysical applications has led to many efforts (for the details see refs.1>2and cited references therein). Results of the present work will complete existing nuclear rate sets, calculated on the basis of different mass predictions for neutroninduced fission rates, which are very important for r-process nucleosynthesisl3p4and in particular for the formation of the superheavy elements. The statistical model (Wolfenstein-Hauser-Feshbach a p p r ~ a c h ) can ~>~ be applied for these calculations as well as for the majority of nuclear reaction rates in astrophysics. The compound nucleus picture will only dominate when the energy of the incident particle is low enough. This point is practically always satisfied in astrophysical environments. 1. The Statistical Model and its application for calculations
As in previous w o r k ~ we , ~ have ~ ~ applied the statistical Hauser-Feshbach formalism for the calculation of beta-delayed and neutron-induced fission rates, so important in modelling the r-process. It has been shown7 that the statistical model is very well applicable for astrophysical neutron-induced rate calculations, as long as there exists a sufficiently high density of excited states, which is the case for heavy nuclei. Near the neutron drip-line the systematic errors of the approach in neutron-induced fission rate calculations can rise, underlining on the one hand that reliable mass predictions are absolutely necessary for r-process applications far from stability and on the other hand that direct capture has to be introduced as well. Early r-process calculations made use of the mass predictions by Hilf et a1.l' and the fission barriers of macroscopic-microscopic modelll for fission rate evaluations. For a consistent treatment of the nucleosynthesis the fission rates, however, should be calculated with the neutron separation energies, reaction &-values and fission barrier heights, derived from the same mass model. The cross-section for a neutron-induced reaction io(n,f ) from the target ground state io with center of mass energy Ei, and reduced mass pin is given by
The total transmission coefficient Ttot = C,,,T,"
(1) describes the trans-
179
mission into all possible bound and unbound states Y in all energetically accessible exit channels o (including the entrance channel i). The total transmission coefficients were calculated utilizing up to 19 experimentally known excited states (if available). The data are taken from,12 up to the first level for which the spin assignment was not known. Ground state spin and parities are known for many unstable nuclei. Far off stability, ground state spins and parities are taken from13 if experimental values are not available. The transmission coefficient T f ( E , J " )includes the sum over all possible final states. Since the work of Strutinski15 fission has been generally described within the framework of double-humped fission barriers. Similar to previous workg we followed the approximation of a twehump barrier. The calculation of the fission probabilities was performed in the complete damping approximation which averages over transmission resonances, assuming that levels in the second minimum are equally spaced. The statistical model can be applied provided that the use of averaged transmission coefficients is permitted. This will be the case for high level densities with completely overlapping resonances, typical for the compound nucleus reaction mechanism. For light nuclei, decreasing particle separation energies or at shell closures, level densities will eventually become too low for the application of the statistical model at astrophysical temperatures. In those cases, single resonances and contributions from the direct reaction mechanism have to be taken into account.16 Based on the level density description' a quantitative criterion for the applicability was derived re~ e n t l yIn . ~the present work we produced the fits of neutron-induced fission rates regardless of applicability and the allowed temperature range will be specified. The estimate is quite conservative and thus the rates can still be accurate even below the given lower limits of the temperature. The important ingredients of statistical model calculations as it was discussed in detail earlier' are the particle and y-transmission coefficients T and the level density of excited states p. Therefore, the reliability of such calculations is determined by the accuracy with which these components can be evaluated (often for unstable nuclei). It is in these quantities that various statistical model calculations differ. The reaction rates given in this paper are calculated with the code SMOKER, l7 but detailed comparison of calculated values was made either with the code NON-SMOKER," extended from the SMOKER code.17 The unified and modified input data such as experimentally known level characteristics were used. The challenge is in the goal to provide them in as reliable a way as possible, also for
180
unstable nuclei for which no experimental information is available. Thus, global descriptions are employed which minimize the overall error and are trusted to be reliable also far from stability. Although the dependence of the astrophysical rates on the utilized level density formula on the back-shifted fermi-gas formalism (see7?l7and cited therein) usually do not change the results more than lo%, we applied more advanced level density formulae.16 2. Calculations of neutron-induced fission rates on the basis of different mass predictions.
The nuclear reaction rate for a specific reaction a t a given stellar temperature T* is determined by folding the stellar reaction cross section u*(E), which is a superposition of the cross-sections of Eq. (1) over a MaxwellBoltzmann distribution of relative velocities between projectiles and targ e t ~ : ~ ~
\.--
I
It has to be emphasized that only the use of the stellar cross section u* thermally populated of target states, yields a reaction rate with the desired behavior that the inverse reaction can be calculated by using detailed balance. Laboratory rates measure only dab= C , sou, i.e. the cross section with the target being in the ground state. For astrophysical applications, such rates have to be corrected for the stellar enhancement effect due to the thermal excitation of the target.20 The values of stellar enhancement factor for a range of temperatures and nuclei close t o stability can be found in previous paper' and in a recent compilation of neutron cross sections for the s process2 as well. To derive energy-dependent and averaged cross-sections, different massand fission barrier predictions were used. But Thomas-Fermi mass predict only one fission barrier. In order to apply the generally used doublehump barrier approach for the Thomas-Fermi fission barrier predictions22 we made use of the relative height of the first and second barrier of ref.,g based on the Howard-Moller (HM) fission barrier predictions" considering the original TF-barrier as the highest one. To consider the sensitivity of the rates to the height of the second barrier we also performed calculations with the original ratio rather than the difference of the relative height. The comparison shows, that differences are very small, for the majority of cases
181
it did not exceed 1%. An extreme case is C f 294, where it amounts t o 9%. Such an insensitivity of the results on the smaller barrier height values shows the reliability of the calculations with described evaluation of second barrier. It is important for the fission TF-rates of the isotopes of elements with 2 > 100 for predictions of the second barrier are absent in ref.ll
Figure 1. The ratio of rates with different fission barriers and mass predictions under T g = 0.3 MeV (left) and T g = 10 MeV (right). The numbers I,J define the set of the rates, calculated with utilization of different models: ETFSi (I,J=l), Thomas-Fermi (I,J=2) or Thomas-Fermi for fission-barriers and FRDM for masspredictions (I,J=3).
Fig.1 show the comparison of rates, calculated on the basis of different fission barrier and mass predictions. The difference of the rates values for the majority of nuclei calculated with different data can achieves 8 orders of magnitude, when one of 2 comparable sets is based on ETFSi. The difference between rate sets calculated on the basis of FFtDM14 and TF-modelz1 are rather small. The extreme difference of ETFSi and other rates was obtained for nuclei with neutron number close to 184, for which the EXTFi model gives very high fission barrier values resulting in small fission rates. The extrapolation of rate calculations in regions of very exotic nuclei is a hard task and only further investigations can answer which kind of predictions is more reliable. Up to that moment probably only r-process calculations can give the answer whether one or another data set can fit the observed the r-process abundances. The next figure (Fig.2) show the cross-section (left panel) and neutxoninduced fission rates (right panel) calculated with use of different data, Difderived from different mass- and fission barriers predictions. ferent arrows in the left plot show difference between highest fission barrier 11114922-24
182
and neutron separation energy Bf - S, by predictions of TF-mode121>22 (red arrow a t the top of the left panel) and by ETFSi mode123t24(dashed arrow at the bottom). Table 1. S, and fission barrier predictions for zslU.
I
Models:
I
TF
I
I
ETFSI
FRDM
I
The cross-sections values are very sensitive not only t o the values of fission barriers and neutron separation energies (see Table l),but also t o the difference of these quantities. The slow decreasing of S, results in decrease of cross-sections when FRDMI4 mass-predictions were used instead of TF ones.22The cross-sections in this case come closer to the cross-sections with higher barriers and Sn23 for the low energies E < S,, and for higher S, almost coincide with rates when TF22predictions were used. The temperature averaged rates (Fig.2, right panel) show the same dependence.
lxlO;!l
'
'
1
'
' ' " " "10
T9
Figure 2. Dependence of neutron-induced fission cross-sections anf ( E ) (left) and rates Xf, = (av)NA(right) on mass- and fission barrier predictions for 261U. Arrows show the difference between highest fission barrier and neutron separation energy Bf - SN for ETFSi(dashed line) and T F (line) predictions.
From the presented plots it is clear that the ETFSi model gives lower rates, mostly and especially for nuclei with neutron numbers close t o N x 126,160,184, because it predicts higher fission barriers, than other models,
183
especially for the closed shells. With increasing temperatures the difference increases. 3. The fit for neutron-induced fission rates
Reaction rates have been calculated for a temperature grid of 24 temperatures: Tg=O.l, 0.15, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0, 6.0, 7.0, 8.0,9.0, 10.0. For easy application in astrophysical investigations, the rates were fitted with the same parameterization as fits for other reaction types were made’
with the seven open parameters ao-a6 and the stellar temperature T g given in lo9 K. This parameterization proves to be flexible enough to accommodate the different temperature dependencies of the various reaction types across the fitted temperature range of 0.01 5 TS 5 10. Parameterizations of the present rates in the form used in25 will be available soon upon request. For finding the best fit the well known code FUMIL126 for the function of any kind was used. The flexibility of the fitting function makes it prone to numerical problems outside the calculated range at low temperatures. In some cases they tend to diverge strongly. This difficulty can be avoided by providing fit data at low temperatures additionally to the calculated values by appropriately extrapolating the rates to lower temperatures. However, it has to be emphasized that the considered parameterization is only valid within the temperature range of 0.01 5 T g 5 lo., although many fits will show a “proper” behavior down to lower temperature. As a measure of the accuracy of a given fit, the quantity is considered. It is defined by
with r being the original rate value as calculated at each of the 24 temperatures T g = 0.1, 0.15 . . .10.0, and fi is the rate calculated from the fit a t these temperatures.The rates with T < cm3 s-l mole-’ are neglected and were not considered for fits. The small value of C is indicative of an accurate fit over the entire temperature range, large C generally signify deviations of the calculated from the fitted rate at the lowest temperatures
184
-2
0
I
I
Figure 3. The examples of fits - for Cf273 and CmZg1.
<
only. For the majority of nuclei the value of is less than 1 and lies in the range 0.1 For all cases it is recommended to use the fits only down to the temperature TCL. The estimated lower temperature limit of the validity of the statistical model, we considered as 0.01. Below that limit the calculation of the rate by means of the statistical model may not be justified, although the fit to the calculated rate will still be accurate. At temperatures below the applicability limit, rates may be overestimated and should be compared to calculations considering single resonance and direct reaction contributions. Especially close t o the drip-lines, fits of reactions with low Q-value cannot be applied a t low temperatures. Although the fit may be valid, it should not be used at low temperature because the statistical model will not be applicable anymore.
TFF,
4. Conclusions
The provide reaction rate predictions on the basis of different realizations of theoretical Hauser-Feshbach method the neutron-induced fission rates and ( n ,7)-rates for the wide range of astrophysical temperature ( 10' 5 T(K) 5 10.0 x lo9 ) with different mass- and fission barriers are made. The of tabulated coefficients in standard form are in preparation now and can be applied for the r-process calculations.
Acknowledgements The authors thank E. Kolbe, K.-L.Kratz, K.Langanke, G.Martinez-Pinedo, P.Moller, D.K. Nadyozhin, B. Pfeiffer for useful discussions.
185
Bibliography 1. T. Rauscher and F.-K. Thielemann, Atomic Data Nucl. Data Tabl. 75, 1 (2000). 2. Z. Y. Bao et al., Atomic Data Nucl. Data Tabl. 76,70 (2000). 3. I. V. Panov, E. Kolbe, B. Pfeiffer, T. Rauscher, K.-L. Kratz and F.-K. Thielemann, Nuclear Physics A . 747,633 (2005). 4. G. Martiriez-Pinedo et al. Progress in Particle and Nuclear Physics. 59,199 (2007). 5. L. Wolfenstein, Phys. Rev. 82,690 (1951). 6. W. Hauser and H. Feshbach, Phys. Rev. 87,366 (1952). 7. T. Rauscher and F.-K. Thielemann and K.-L. Kratz, Phys. Rev. C56,1613 (1997). 8. D. Mocelj and T. Rauscher et al. Nuclear Physics A . 758,154 (2005). 9. F.-K. Thielemann, J. Metzinger and H.V. Klapdor-Kleingrothaus, Zt. Phys. A 309,301 (1983).
10. E. R. Hilf, H. V. Groote and K. Takahashi, Proc. 3 Int. Conf. on Nucl. f a r from Stability, 1976, CERN-76-13, 142. 11. W. M. Howard and P. Moller, Atomic Data Nucl. Data Tabl. 25,219 (1980). 12. R. B. Firestone, in Table of Isotopes, 8th edn., edited by V. S. Shirley (Wiley, New York, 1996). 13. P. Moller, J. R. Nix, and K.-L. Kratz, Atomic Data Nucl. Data Tabl. 66,131 (1997). 14. P. Moller, J. R. Nix, W. D. Myers and W. J. Swiatecki, Atomic Data Nucl. Data Tabl. 59,185 (1995). 15. V. M. Strutinsky, Nucl. Phys. A 95,420 (1967). 16. T. Rauscher et al., Phys. Rev. C. 57,2031 (1998). 17. F.-K. Thielemann and M. Arnould and J. W. Truran, in Advances in Nuclear Astrophysics, edited by E. Vangioni-Flam et al. (Editions frontikre, Gif sur Yvette, 1987), p. 525. 18. T. Rauscher and F.-K. Thielemann, in Stellar Evolution, Stellar Explosions and Galactic Chemical Evolution, edited by A. Mezzacappa (IOP, Bristol, 1998), p. 519. 19. W. A. Fowler, Quarterly Journ. Royal Astron. SOC. 15,82 (1974). 20. M. Arnould, Astron. Astrophys. 19,92 (1972). 21. W. D. Myers and W. J. Swiatecki, Nuclear Physics A , 141 601,1996 22. W. D. Myers and W. J. Swiatecki, Phys. Rev. C. 60,014606-1 (1999). 23. A. Mamdouh, J. M. Pearson, M. Rayet, F. Tondeur, Nucl. Phys. A . 679,337 (2001). 24. Y. Aboussir, J. M. Pearson, A . K. Dutta and F. Tondeur. Atomic Data Nucl. Data Tabl. 61,127 (1995). 25. J. Holmes, S. Woosley, W. Fowler and B. Zimmerman, Atomic Data Nucl. Data Tabl. 18,305 (1976). 26. S. N. Dymov, V. S. Kurbatov, I. N. Silin, S. V. Yaschenko, Nuclear Instuments and Methods i n Physics Research A . 440,431 (2000).
This page intentionally left blank
Facilities and Detectors
This page intentionally left blank
A NOVEL 2V2E SPECTROMETER IN MANCHESTER: NEW DEVELOPMENT IN IDENTIFICATION OF FISSION FRAGMENTS I. TSEKHANOVICH', J.A. DARE, A.G. SMITH, B. VARLEY,
D. CULLEN, N. LUMLEY Department of Physics and Astronomy, University of Manchester, Oxford road, Manchester M13 9PL, UK * E-mail: igor. [email protected] T. MATERNA,
u. KOSTER
Institut Lam-Langevin, 6 rue J. Horowitz, 38042 Grenoble, France
G. S. SIMPSON Laboratoire de Physique Subatomique et de Cosmologie, 53, avenue des Martyrs, 38026 Grenoble, France A 2v2E spectrometer for identification of fission products by their masses and nuclear charges is under construction in Manchester. It is aimed for the use in prompt and delayed y-ray spectroscopy experiments in conjunction with photon detectors. T h e fragment identification is made on the analysis of pulse traces from gas detectors, aquired and processed with digital readout electronics. Keywords: 2v2E Spectrometer; Fission Fragment Identification; Pulse Trace Analysis.
1. I n t r o d u c t i o n
At present, much effort is being devoted to produce, and to study the properties of, nuclei with high neutron excess. Low-energy fission of very heavy nuclei is one of the best methods to create neutron-rich nuclear matter. The spectrum of products from fission is broad and comprises more than 500 nuclei. A fraction of them produced most abundantly (-100) has been intensively studied with large arrays of y-ray detectors such as Euroball or Gammasphere (see review in Ref. 1 and references therein). The selection of 189
190
nuclei wich such arrays is based on the y-ray analysis and requires at least triple y-ray coincidences. However a clear identification of a new nucleus can only be done if the structure of the complementary partner(s) is known; this is practically impossible in cases where complementary fragments are not present in spectrum (e.g. decay of isomeric states). To proceed, and t o extend spectroscopic studies to nuclei with higher-than-average neutron excess (i.e., produced less frequently in fission) one wishes to have direct information about an excited nucleus, such as its mass and atomic number. A possibility of mass and, in some cases, nuclear charge selection of fission products, when combined together with lower-fold y-ray detection should then allow better use of the rich fission reaction potential, by extending the range of nuclei, accessible for spectroscopy, to very exotic species. This is also where interesting phenomena are expected, such as break up of magic numbers or nuclear shape changes. The device to provide this extra information on fission products has to fulfill the following requests: 0 0
0
Large solid angle acceptance for fragment detectors; High efficiency for fission fragments as the latter should serve as a trigger for data acquisition; Reliable identification of fission products by mass and ideally by nuclear charge.
A device which meets these criteria and is currently under construction in Manchester is briefly described in this paper. An emphasis is made on the issue of nuclear charge determination. 2. Spectrometer Layout
A novel detector system built in Manchester is aimed at efficient identification of fission products along with their associated y rays. The general scheme of the experimental setup is shown in Fig. 1. The SpecTrometer for Exotic Fission Fragments ( S T E F F ) consists of two fragment detection arms, each comprising a time-of-flight ( T O F ) detector coupled to a wide-angle axial ionization chamber ( I C ) . A secondary-electron multi-channel plate ( M C P ) serves as a common start detector for both the T O F sections. The stop detectors are home-built position-sensitive large area secondary electron detectors utilizing multiwire proportional counters ( M W P C s ) .A T O F measurement is performed over a flight path of one l m with a resolution expected to be close to 1%.The M C P and electrostatic mirrors of the MWPCs operate under secondary
191
\
/
MWPC etectwstatic mlfror Fig. 1.
Schematic layout of the STEFF spectrometer.
vacuum conditions (better than 10~5 mbar). The MWPCs require a few mbar of isobutane (C^H\o) as counting gas, to achieve the best performance. Wide-angle acceptance axial ionization chambers with multi-segmented anodes serve as energy detectors for complementary fission products. To improve signal-to-noise ratio, each 1C houses a board of charge-sensitive preamplifiers (PAs). The PAs are made of surface-mounted components specially designed for the use in vacuum. The ionization chambers are filled with isobutane and operate at constant pressure of about 65 mbar. The isobutane is chosen for its low ionization potential and relatively high density which results in quite low operating pressure of the /Cs; the latter determines the thickness of the 1C entrance windows separating the detectors from the secondary vacuum volume of STEFF. The energy resolution of the ICs depends mainly on the thickness of this window and can be as good as 0.5%.2 As large angle acceptance of the ICs comprises the use of large area windows, a compromise between window area/thickness has to be made. Therefore more modest value of 1% in energy resolution is set as a goal to achieve for the ICs performance. The fission chamber has two Al hemispheres (2 mrn thick) attached to the hexagonal ring. The chamber contains a fission-fragment source made of a thin layer (~50-7-100 fj.g/cm'2') of fissile material. The source is mounted on the rod of the source changer, which allows the former to be put into or removed from the chamber (from the beam) without braking vacuum. At present, the fission chamber is surrounded by an array of 12 large-volume Nal detectors placed at a distance of 20 cm from the target. Each detector has 5 mm tungsten shielding, for passive supression of Compton scattered 7 rays. The detectors will record 7-ray irradiation emitted by fission products.
192
The total absolute photopeak efficiency of the array is 6.7% and energy resolution is 7%. The inferior resolution of NaI relative t o Ge detectors will not prevent gathering of at least rudimentary spectrosopy information for the lowest few excited states of a nucleus of interest. Depending on the aim of experiment, S T E F F may be run in either a single-arm configuration, for use with a backed target, or in the doublearm configuration for use with a target on a thing substrate so that both fragments built in one fission event can escape. The selectivity of the axial ion chambers is increased through the use of digital electronics to process the pulse shape. 3. Digital DAQ and Data Rates
A data acquisition system based on digital V M E and standard N I M electronics and a PC will acquire the multi-parameter data. The Daresbury GRT4 V M E module is its basic component originally designed to do digital processing for y-ray tracking. With some minor modifications, the card can also be used for digital sampling and processing of signals from slow energy detectors such as ionization chambers. Each GRT4 card includes four acquisition channels operating in parallel, with a common trigger. Each channel has a 14 bit 80MHz flash ADC. Two FPGAs per channel provide data processing and buffering. Circular buffering allows programmable pretrigger delay which is an important feature for the correct base line identification and subtraction. Detailed information about GRT4 modules and the software for data acquisition ( M I D A S ) can be obtained from Daresbury webpage (http://npg.dl.ac.uk). The DAQ will be triggered by a coincidence of the three secondaryelectron detector signals with a t least two y rays (FFGG event). With the V M E bus limiting data rate to 4 M B / s and l k fission-fragment count rate in each I C , the data flow will be on the level of lMB/s which means that full traces can safely be acquired and stored for all 30 segments of the ICs anodes. A special algorithm for the on-line analysis of the FFs pulse shapes can be implemented to the FPGAs of GRT4 modules. This will decrease the length of the data recorded (by a factor of 50) thus allowing measurements at much high fission rates in the target; this work is in progress. The upper limit on the count rates comes from pile-up limitations in the ICs. Another limitation to the IC count rate comes from the prompt neutron production in fission target and dead time of photon detectors. Ge detectors are known to be sensitive to fast neutron exposure. To assess nuclei with low production probabilities (on the level of 0.05%) long measuring campaigns
193
are necessary, at high count rate conditions and at high neutron fluxes. On the contrary, the N a I are better for the use in these harsh conditions. 4. Mass Determination of Fission Products Each T O F arm of the setup (position at the entrance to the I C , given by the M W P C , start and stop time signals) will be used to determine velocities 2, of fission products. Velocity measurements combined together with energy signals &in from the ICs will allow masses of fission products to be deduced:
Eq. 1 shows that good timing properties of the device are of extreme importance for the reliable mass determination. Resolving properties of the T O F detectors themselves are difficult to improve; the easy way to proceed is to increase the distance between T O F start and stop detectors ( T O F length). The T O F length of each arm of S T E F F is set therefore to l m . The resolution of energy detectors of the setup is a subject of constant improvement. The best values achieved at present are somewhat above 1%. Efforts are being undertaken to produce very thin secondary-electron foils for the T O F detectors and decrease the thickness of the entrance window foil, which should bring the energy resolution to the expected 1% level. This will give the uncertainty of less than 2% for the mass determination of fission products. 5. Nuclear Charge Determination and Ranges of FFs In an axial ionization chamber, charged particles (fission fragments) leave an ionization track parallel t o the electric field. The column of electrons is drifted by this field toward the anode, inducing on it a signal which reflects the charge density distribution along the reverted particle track in the chamber. This is demonstrated in Fig. 2. The energy of the particle is given by the amount of the charge collected on the anode (integrated charge density distribution, see Fig. 2 ) . As full pulse traces are recorded, any fraction of them can be used for nuclear charge detemination. Therefore the use of digital readout electronics allows a single gas detector to be exploited like a very flexible E - A E telescope where the depth of the transmission layer can be effectively selected by varying the times relative to the trigger. Determination of ranges of charged particles becomes also possible; ranges are just the lengths of the induced anode signals.
194 1so0
isobutane 6Smbar
2
Y
8
x'
..... '0
20
40
60
80
Depth, mm
lo0 120
Time, GRT4 samples
Fig. 2. Left: Energy loss per unit of length along the particle track in the ionization chamber as extracted from S R I M calculations, averaged for 1000 ions of each sort and smoothed over 5 mm of depth. Right: Example of a fission-fragment pulse shape (baseline substracted), sampled by a GTR4 card.
Nuclear charge determination is based on the analysis of specific energy loss dEldx which is to be derived from the recorded pulse shapes. As known, charge density distribution along the particle track in an IC is described by the Bethe-Bloch equation
Taking Z , ff = v G / v o and combining together universal and detector-related constants one obtains
_ _dE
= clze\fln(c2v2) (3) dx Eq. 3 states that specific energy loss is a function of the F F nuclear charge and velocity. Therefore the information on Z can only be assessed if fission fragments of the same velocities are considered; this again stresses the importance of precise timing measurements. An experiment has been made at the Lohengrin mass separator3 of the I L L in Grenoble, aimed t o test the performance on an axial IC of S T E F F and to develop algorithms for treatment of digitized pulse traces. Fission fragments after Lohengrin are filtered according to their velocities and therefore perfectly suit for testing purposes. An example of nuclear charge determination is given in Fig 3. Ratios between isobars could be correctly reproduced. Nuclear charge resolution deduced from the fit parameters of Fig 3 was however 1/17; it is a sum result of energy straggling in the T O F and IC foils. To improve the IC energy resolution, commercially produced Mylar foils (0.9pm) used in secondary electron detectors and in the ICs during the experiment are being
195
replaced with home-made thinner Formvar foils ( ~ 2 and 0 50pg/cm2, correspondingly). Even if the final 2 resolving power of the energy detectors of S T E F F may not still be high enough to resolve individual fission fragments cleanly it should be however possible to determine nuclear charge using a statistical ensemble (center of gravity) of ions detected in coincidence with particular y rays. A = 80; Ekin = 90MeV
A = 83; Ekin = 9SMeV
200 150 rn
+
5 100
8
so 0
690
720
720 750 780 750 780 SPECIFIC ENERGY LOSS, ADC CHANNELS
810
840
Fig. 3. Deconvolution of specific energy loss signals deduced from digitized pulse traces into single isobars. Difference in positions of identical nuclear charges is due t o the difference in FF velocities.
The range R of a fragment stopped inside the chamber is dependent on its mass number A , its atomic number 2 and fragment velocity 21. The relationship between A E , the ratio of fragment range to mass number RIA, mean atomic number (2) and velocity 21 has been investigated. The results are shown in Fig. 4, with the (2) values taken from Refs. 4,5. Having the A E and R parameters extracted from the pulse shapes it becomes possible to improve nuclear charge selection by setting appropriate gates in the twodimentional space. It should be noted that range is a very sensitive parameter to the signalto-noise ratio of the preamplifiers (see Fig. 2). Currently, the PAS have been adapted, what improved the signal-to-noise of the circuitry by a factor of 3.5. This will result in better resolution on both z and y axis of Fig. 4 and will therefore give better selectivity for fission products. 6 . Conclusion
A novel double-energy double-velocity spectrometer S T E F F is under construction a t the Manchester University. S T E F F is aimed to identify fission
196
|5,ooe ||8.8oa !4M 4100
240
Range/A
Fig. 4. Specific energy loss versus reduced range for different velocities of fission fragments. Lines are just to demonstrate the trend.
products by their masses and nuclear charges and constructed for the use in prompt and delayed 7-ray spectroscopy experiments. Modern digital readout electronics is used for acquisition and processing of pulses from ionization chambers. Appropriate algorithms have been developped for the pulse trace analysis. Results obtained from a test experiment with one 1C with known beams of fission fragments have demonstrated the power of digital electronics in determination of fragments' ranges and energies in gas detectors. It was also possible to obtain information on nuclear charges of light fission products, from the analysis of specific energy losses deduced from the pulse traces. The factors limiting the resolving properties of STEFF detectors are understood and steps are being undertaken to improve resolution of the apparatus. Once fully operational, STEFF may find a range of applications going beyond the scope of 7-ray spectroscopy. Acknowledgment This work is supported by the EPSRC grant 2-4570.5, UK. References 1. P.J. Nolan, F.A. Beck, and D.B. Fossan, Ann. Rev. Nud. Part. Sc. 44, 561 (1994). 2. A. Oed, P. Geltenbort, F. Gonnenwein, T. Manning and D. Souque, Nud. Instr. Meth. 205, 455 (1983). 3. E. Moll, H. Schrader, G. Siegert, et ai, Kerntechnik 19, 374 (1977). 4. J.-L. Sida. Ph.D. Thesis, Paris (1989). 5. W. Lang, H.G. Cleck, H. Wohlfahrt, H. Schrader, and K.-H. Schmidt, Nud. Phys. A345, 34 (1980).
DEVELOPMENT OF PSD AND TOF+PSD TECHNIQUES FOR FISSION EXPERIMENTS M. SILLANPAAI', M. MUTTERER',',
W . H. TRZASKAl, G. TYURIN4v1, YU. N
KOPATCH3, S. SMIRNOV3, S. KHLEBNIKOV4i1, J. VON KALBEN' Department of Physics, University of Jyvaskyla, Jyvaskyla, Finland University of Technology, Darmstadt, Germany Joint Institute for Nuclear Research, Dubna, Russia V.G. Khlopin Radium Institute, St. Petersburg, Russia
'Institute of Nuclear Physics,
Identification of charged particles by mass (A) and nuclear charge (Z) is one of the key requirements in many nuclear physics experiments. As an alternative t o the traditional setup involving A E E telescopes we have combined pulse shape discrimination from a silicon detector with a time-of-flight technique (ToF+PSD). This has revealed a noticeable advancement over using PSD method alone. Recent measurements at the Accelerator Laboratory of the University of JyvLkyla (JYFL) have demonstrated that by adding PSD information t o ToF the discrimination power of the method improves and provides also charge identification down to the low energies. For good PSD, high homogeneity of the silicon material and fast low-noise front-end electronics are mandatory. We have used novel surface barrier Si detectors made from homogeneously neutron-transmutation doped (n-TD) silicon. For fission experiments two applications of PSD and ToF+PSD techniques are proposed. First, in the measurements at JYFL, using "Ne beam on 238U target, the discrimination of fission fragments from lighter products of similar energies was evaluated showing good performance. A second application could be the identification of light charged particles from ternary fission. Especially, the ToF+PSD technique is a very interesting alternative t o A E E method for a ternary particle measurement at low kinetic energies. The discrimination thresholds achieved with ToF+PSD for elements up to carbon were almost half of the thresholds usually obtained with the AE-E method. Keywords: particle identification, semiconductor detectors, fast electronics, pulse shape discrimination, time of flight, nuclear reactions
*corresponding author, e-mail: [email protected]
197
198
1. Introduction It has been known since the early 1960’s that the shape of the current pulse from the interaction of charged particles inside of a solid state silicon detector, carries information about the particle’s A and Z.l The two main processes behind this effect are the differences in electron/hole mobility and plasma delay resulting from charge erosion of the plasma column along the track of the particle in ~ i l i c o n .Currently ~?~ pulse shape discrimination (PSD) is becoming an important tool for particle separation especially for large detector arrays such as CHIMERA4 or FAZIA.5 In fact all large detector systems for the next generation of nuclear beam facilities, such as EURISOL,6 intend to use PSD for particle identification. Over the last decades, different kind of PSD techniques like integration gate^,^ zero-crossing’ and pulse ~ t r e t c h i n g ,have ~ been developed t o scrutinize the information from the pulse shape. Most of them were based on probing the rise time part of the current pulse. It was found, that the key parameter for good quality discrimination is the homogeneity of the resistivity profile of the silicon material.lO~llQuite recently very good results where obtained by Mutterer et a1.12 made of very homogeneously neutron transmutation doped (n-TD) silicon surface barrier detectors (SB) and rear side injection of particles. Clear Z and partly even A discrimination was obtained down to relatively low energies. Having quite low energy threshold for discrimination the PSD method has become an alternative t o the traditionally AE-E telescopes, which often suffers from the limitations in homogeneity, active area and minimum thickness of thin silicon layers. This work is the continuation of the research presented in Ref. 12. The main difference is the addition of TOF information t o the spectra. Combining ToF with PSD techniques to extend the dynamical range to the lowest energies has been introduced and tested by Pausch et al.3 However, this attempt was limited by the modest time resolution of the detectors and of the start signal. In the measurements presented here we have used high quality micro-channel plate (MCP) detectors and SB detectors with fast amplifiers both giving time resolution around 100 ps (FWHM of a T O F peak). As a result we have achieved impressive discrimination of A (ToF), and Z (PSD) over a wide energy range for all light particles. We intend to apply this technique to our studies of ternary and higher modes of fission.
199
2. Experimental Method 2.1. Physical setup and geometry The measurements have been carried out at the JYFL using 400 MeV 20Ne beam and A1 targets (0.14 mg/cm2and 0.40 mg/cm2).We also used a 238U target to measure how fission fragments are separated from lighter reaction fragments. The SB detectors were made from n-TD silicon (resistivity 5kR * cm). The detectors had active areas of 150 mm2 and 300 mm2,both sizes having two different thicknesses of 250 pm and 800 pm. Depletion voltage was 45 V and 455 V for 250 pm and 800 pm thick detectors respectively. We overbiased our 250 pm thick SB detector approximately 2-3 times providing efficient charge collection and resolution. Detectors were cooled down to 15 "C to stabilize detector current. Detectors were located a t 20" with respect to the beam axis. The geometry is shown schematically in figure 1. Installation of two MCPs allowed
Fig. 1. Schematic scheme of t h e physical setup used in t h e ToF+PSD experiment
us to measure simultaneously ToF between MCPl and MCP2 (see figure), timing between MCPl and SB detector (ToF+PSD) and timing between
200
MCP2 and SB detector (risetime with external start). The ToF path between M C P l and SB detector was -60 cm. MCP2 was placed just in front of the SB detector. 2.2. Electronic s y s t e m The block diagram of the electronics used in the present work is shown in figure 2. We have used standard commercial units except for the chargesensitive preamplifier (model CSTA2) for SB detector and the low-noise wide-band timing filter amplifier (model TFA2000), both manufactured by the electronic laboratory of TUD. The slow energy output from CSTA2, was used for standard energy measurement. The fast timing output from
MCP
CFD
b
iFiFo
SB = Surface Barrier detector MCP = Multi-Channel-Platedetector PA = PreAmplifier SPA = Spectroscopy Amplifier TFA = Timing Filter Amplifier CFD = Constant Fraction Discriminator TAC = Time-to-Amplitude Converter ADC = Analog-to-Digital Converter FIFO = Fan in-Fan out
F i g . 2.
\
~ M C P Z ~ S B Relative riselime" ~
siop
TAC
ADC ~
B l o c k diagram of the electronics used in t h e ToF+PSD experiment
CSTA2 (marked as T in the figure 2) was driven to TFA2000 were it was amplified and shaped with 2 ns integration. To probe the rise time differences, we used constant fraction discriminator (CFD, ORTEC CFSOOO) operating in the true-constant-fraction mode (TCF).13 In order to work in the T C F mode, the settings of the CFD should fulfill the equation t d > tr(l - f ) , where t d and f are the intrinsic delay and fraction settings of the CFD,
201
respectively, and tr is rise time. In case of too small td and f values, the former condition fails resulting in an amplitude-and-risetime-compensation (ARC) mode of the CFD. In the ARC mode rise time differences are compensated reducing sensitivity of Z discrimination of particles. The signal from CFD was measured in relation to signals coming from MCP detectors using time-to-amplitude converters (TAG) as shown in figure 2 (marked as "ToF+risetime" and "Relative risetime"). We also measured ToF with two MCPs, MCP1 and MCP2 (flight path 61.5 cm) using TAG (marked as "ToF" in the figure 2). 3. Results
In the measurements to evaluate the combination of ToF and PSD techniques, it was shown that the discrimination power on the particle's A and Z reaches over a wide energy range, even down to relatively low energies. Figure 3 shows an example of obtained spectra. The main contribution
Fig. 3. a) ToF+PSD-energy matrix obtained with 400 MeV 20 Ne beam on 27A1. 250 pm/150 mm 2 SB detector was 3.3 times overbiased with 150 V. b) Low energy region from a). Supplement of PSD discrimination (mainly Z, but also A) to ToF is visible as a separation of neighbouring isobars 7 Li and 7Be
to ToF+PSD is coming from ToF, giving the family of hyperbolas corresponding to different mass values A. In ToF+PSD the rise time supplements ToF, generating a time shift between neighbouring Z values in the isobars. This effect can be clearly seen in the figures 4a and 4b, for mass condition A=7 and 15, respectively. The energy values where the lines are bending
202
back correspond to the maximum energy loss of particles with sufficient energy to penetrate through the detector. For higher energetic particles the mass dependent ToF is still measured, but now as a function of mainly charge dependent energy loss in the silicon detector, resulting in a clear charge separation. The obtained charge separation by measuring ToF as a function of energy loss is very interesting. As far as we know, nobody has applied this method before for charge separation. Contribution of the ToF
a)
I
7Be 7Li
Energy AfeV
EnergyMe"
Fig. 4. Isobar separation obtained with ToF+PSD between a) 7 Li and 7Be b) 15N and 15 O. Beam: 400 MeV 20Ne, target: 27A1. Detector: 250 /xm/150 mm2 SB detector with 150 V bias voltage.
to ToF+risetime timing is depending on the length of the used flight path, while the contribution of the rise time part is depending on the bias voltage applied to the SB detector. Considering the ToF+PSD method, bias voltage has two main effects. Strong over-biasing eventuates to more rapid charge collection from the plasma column. For heavier particles, which are having shorter penetration depths and so stronger plasma effect, this reduces the probability for charge carriers to recombine or trap inside the column resulting in better energy and time resolution. On the another hand, stronger over-biasing reduces the rise time, and consequently rise time differences between particles of equal energy on absolute scale. This shortens the distance between neighbouring ToF+PSD isobar lines, which is depicted in the figure 5 for the measurements with 105 V and 150 V SB detector bias voltages. Furthermore the shortening of the distance between lines affects also the energy threshold for isobar separation. Depending on the range of particles
203
of interest, a compromise in bias voltage has to be made to optimize both, energy resolution and separation power. By over-biasing our SB detector (250 /j-m thick, 150 mm2 in area) approximately three times, we obtained with ToF+PSD technique excellent 125 ps time resolution (FWHM).
a)
Energy
Energy MsV
Fig. 5. a) ToF+PSD matrix obtained with 400 MeV 20Ne on 27A1 and a) 105 V and b) 150 V SB bias voltage.
It was shown in the present study, that with the ToF+PSD discrimination technique, energy thresholds for charge Z and mass A discrimination could be pushed down to relatively low values. Threshold values below 1.2 MeV/A for masses A<8 and 1.9 MeV/A for masses 8
Applications of ToF+PSD and PSD to fission experiments The measurement with 20Ne on 238U revealed that two possible applications of ToF+PSD and PSD techniques to fission work could be imagined. First, the discrimination of fission fragments (FF) from lighter particles (LP) of the same energies with ToF+PSD was evaluated in heavy-ion induced fission reaction showing good performance (see the figure 6a). This separation is achieved already by the ToF contribution. In addition to separation of FF from LP with ToF+PSD, the discrimination with rise time measured with external start from MCP2 (see the figure 1) was also cogent (see the figure 6b). ToF+PSD technique could be also applied as a new tool for ternary fission studies especially at low energies, where the observed asymmetry for
204
Energy M»V 20
238
Fig. 6. Matrixes obtained with 400 MeV Ne on U using a) ToF+PSD (MCP1-SB) and b) Rise time with external start (MCP2-SB) techniques
energy distributions of ternary particles, e.g. 4He and 6He, has been an issue for decades. Several experiments have been dedicated to study this deviation from the Gaussian shape, probing the low energy part with different methods. Recently the energy spectra, attesting the spectral asymmetry, for ternary 4He and 6He in 252Cf spontaneous fission were measured down to 1 MeV using ToF-E technique.14 Challenge for the future is to study the low energy part also for other light charged ternary particles such as Li or Be isotopes. ToF+PSD could offer a good solution to achieve not only A, but also Z discrimination, over wide energy range. 4. Discussion Our results have demonstrated the good performance of the ToF+PSD technique. This technique gives an opportunity to reach mainly A but also Z separation at very low energy region, where traditional AE — E telescope, usually having thresholds >2 MeV/A,15'16 is not working anymore. By using the rear side injection of particles and strong over-depletion of a detector, one can achieve very good time resolution. Good resolution and discrimination power are enabling a wide applicability of ToF+PSD technique for new experimental setups. Studies on nuclear structure of exotic nuclei through direct and compound nuclear reactions, particular elastic and inelastic scattering, transfer reactions, fusion evaporation and multifragmentation reactions with ToF+PSD could be imagined. Most promising development for PSD studies in common is the step
205
towards digital pulse shape analysis with flash ADCs. T h e combination of PSD with digital pulse shape analysis could offer a solution for A and Z discrimination of particles in big detector setups such as CHIMERA4 or
FAZIA.5 Further experiment in JYFL is projected t o evaluate the limits of the ToF+PSD technique with the present SB detectors, preamplifiers and other analogue electronics. In this experiment also the applicability of the method for ternary fission measurements will be studied more detailed.
Acknowledgments This work has been supported by the EU Integrating Infrastructure Initiative, Contract No. 506065 (EURONS), and by the Academy of Finland under the Finnish Center of Excellence Programme 2006-2011, and by the INTAS Grant No. 03-51-6417. We are deeply indebted t o the accelerator staff of the University of Jyvkkyla.
References 1. C. Ammerlaan et al., Nucl. Inst. and Meth. 22, 189 (1963). 2. S. Aiello et al., Nucl. Inst. and Meth. in Phys. Res. A427 (1999). 3. G. Pausch et al., Nucl. Inst. and Meth. in Phys. Res. A337, 573 (1999). 4. M. Alderighi et al., IEEE Trans. Nucl. Sci. 52, 1624 (2005). 5. 6. 7. 8. 9. 10. 11. 12. 13.
FAZIA, web: http://fazia.in2p3.f. J. Vervier, The EURISOL report web: http://www.ganil.fr/eurisol. W.-D. Emmerich et al., Nucl. Inst. and Meth. 83, 131 (1970). G. Pausch et al., Nucl. Inst. and Meth. in Phys. Res. A349, 281 (1994). G. Pausch et al., IEEE Trans. Nucl. Sci. NS-43, 1097 (1996). F. Henari et al., Nucl. Inst. and Meth. in Phys. Res A288, 439 (1990). G. Pausch et al., IEEE Trans. Nucl. Sci 44, 1040 (1997). M. Mutterer et al., IEEE Trans. Nucl. Sci. A385, 756 (2000). ORTEC, Principles and applications of timing spectroscopy, tech. rep. Application Note AN42. 14. M. Mutterer et al., Energy distribution of ternary a particles in 252Cf(sf), in Proc. Intern. Conf. on Dynamical Aspects of Nuclear Fission, (Smolenice Castle, Slovak Republic, 2006). World Scientific, Singapore, 2007, to be published. 15. Y . Kopatch et al., Studies on particle-accompanied fission of 252Cf(sf) and 235U(ntt,f),in 3rd Int. Workshop on Nuclear Fission and Fission-Product Spectroscopy, (Chateau de Cadarache, St. Paul des Durance, France, 2005). 16. V. Tishchenko et al., Recent multi-parameter studies on particleaccompanied fission of 252Cf(sf)and 235U(nthf),in Proc. 14th Int. Seminar on the Interaction of Neutrons with Nuclei, ISINNl3, (Dubna, Russia, 2005).
This page intentionally left blank
MYRRHA, A NEW FAST SPECTRUM FACILITY H. AIT ABDERRAHIM, P. D'HONDT, D. DE BRUYN SCPCEN, Boeretang 200 B-2400 Mol, Belgium Since 1998, SCK-CEN, Mol, Belgium, - in partnership with many European research laboratories - is designing a multipurpose Accelerator Driven System for R&D applications - MYRRHA. In parallel, an associated R&D support program is being conducted. MYRRHA aims to serve as a basis for the European experimental ADS providing protons and neutrons for various R&D applications. In its "Draft-2" version, MYRRHA consists of a LMAC proton accelerator delivering a 350 MeV*5 mA proton beam to a windowless liquid Pb-Bi spallation target that in turn couples to a Pb-Bi cooled, sub-critical fast core of 50 MW thermal power. In this paper we report briefly first on the status of the MYRRHA design, then on its current evolution towards the XTADS machine in the framework of the integrated project EUROTRANS and finally on the R&D related programme.
1. MYRRHA within the mission of SCKCEN The mission of SCKCEN has been described in its statutes as follows: With a view to sustaining development by research and development, training, communication and services, SCKCEN contributes to:
0
nuclear safety and radiation protection; medical and industrial applications of radiation; the backend of the nuclear fuel cycle.
This general statement was further detailed and complemented by the Belgian parliament where it was specified that SCKCEN should "have confined activities regarding the reactor- and fuel cycle systems of the fourth generation and maintain, and later on replace, the BR2-reactor". In this sense, MYFWHA responds clearly to the mission given to SCK-CEN: 0
first with regard to the replacement of the BR2 reactor by a new material testing reactor (MTR) corresponding to today's needs be it as a tool for material and fuel research or as a machine ensuring the production of radioisotopes for the medical industry; 207
208
then with relation to the backend of the nuclear fuel cycle, by investigating the transmutation of minor actinides resulting from the existing power plants; and finally with regard to GenIV reactors by proposing MYRRHA as a lead fast reactor (LFR) concept demonstration. 2. Scope of MYRRHA
The activities in MYRRHA are related to the engineering design of a fast spectrum irradiation facility, to the related supporting R&D programme and to the set-up of an international organisational structure for the preparation and construction of the facility. For the engineering part: the detailed design has to be concluded and the construction to be prepared. S C P C E N and its partners will make use of competences in the domains of neutronics, reactor design, fuel technology, material technology, thermohydraulics, instrumentation, safety and civil engineering. A part of this work is executed in synergy with projects of FP6 and probably FP7 of the European Commission. The R&D support programme contains the following components: the development of the liquid lead technology, the development of advanced instrumentation and control mechanisms, the characterization and validation of the fuel and materials and the development of in-service inspection and repair (ISIR), robotics and teleoperations. The organisational structure of the project will change significantly in 2008-2009, installing an International Advisory Council setting up an Owner Consortium. 3. Overview of activities in the past 3.1. The ADONIS-project (1995-1997)
The coupling between an accelerator, a spallation target and a subcritical core has been studied for the first time at SCKCEN in the frame of the ADONIS project (1995-1997). The first purpose of ADONIS being the production of radio-isotopes for the medical industry, mainly 99M0 as fission product from highly enriched 23sU targets. The means used were of limited size with an accelerator of 150 MeV and a core with a power of around 1.5 MW,h to be evacuated. The sub-critical core was made of the 23sUtargets for producing the YY Mo without other driver fuel. The system was a thermal spectrum machine and therefore water was used as coolant.
209
3.2. From ADONIS to MYRRHA (1998-2005) The ad-hoc scientific advisory committee recommended then to extend the purpose of the ADONIS machine to become a Material Testing Reactor (MTR) for material and fuel research, to study the feasibility of transmutation of the minor actinides and to demonstrate at a reasonable power scale the principle of the ADS. It should be mentioned that this decision was taken around the same time of the last BR2 refurbishment. It was then clear that a next BR2 refurbishment would be compulsory around 2020. If such refurbishment remains technically possible, a new machine should be better adapted to the new needs, in particular for GEN IV and Fusion research. The project, since 1998 named MYRRHA, has then evolved (Figure 1) to a larger installation: the accelerator sends protons of 350 MeV energy with 5 mA current, on a spallation target installed in a 50 MWt,, sub-critical core. The socalled "Draft-2" design, published early 2005, is summarized in reference [ l], for which a "Business plan" [2] has also been written A significant part of the design and of the R&D programme has been covered by several projects of the fifth (2001-2005) and sixth (2005-2009) framework programmes of the European Commission. For FP5, we can mention the following projects: the thematic network ADOPT, the pre-design studies within PDS-XADS, spallation studies within MEGAPIE and ASCHLIM, neutronic studies within MUSE; material studies within SPIRE and TECLA; and fuel studies within FUTURE. Most of these studies have been regrouped in FP6 in the integrated project IP-EUROTRANS (see below under item 3.4).
21 1
It should also be mentioned that the previous version of the MYRRHA design (called Draft- 1) has been examined in 2002 by an international technical guidance committee (ITGC). If some domains like safety and R&D have been mentioned for requiring deeper investigation, no show-stoppers have been identified by the experts.
3.3. Current organisation of the MYRRHA project The MYRRHA project being a large project, even internally since 1999, we found necessary to organise it as a multi-disciplinary project organised per large components (namely; accelerator, spallation target, core design, primary system design, buildings and siting) or specific topics (namely; fuel design, safety studies, operation and control) of the facility. Long-term collaboration agreements have been signed with other research institutes or contractors; the most active collaborations being: UCL (spallation target); FZK (spallation, materials, corrosion); NRG (safety); CNRS (accelerator, neutronics, spallation target); PSI (spallation target, MEGAPIE experiment); SUEZ (safety, building) OTL (Robotics) UI-TUK (ultrasonic visualisation). Several other collaboration agreements have also been signed in a more strategic view with CEA, CIEMAT, ENEA and more recently JAEA.
3.4. From MYRRHA to XT-ADS (2005-2009)
The MYRRHA Draft-2 being available right at the beginning of the FP6 EUROTRANS integrated project, it has been proposed to the EUROTRANS partners as a starting point for the study of the XT-ADS machine. Instead of starting from a blank page, this allowed optimizing an existing design towards the XT-ADS (eaerimental facility demonstrating the technical feasibility of Transmutation in an Accelerator Driven System) objectives within the limits of safety requirements. The EUROTRANS project is an integrated project in the context of Partitioning and Transmutation. It aims to deliver an advanced design of a short-
-
212
term, small-scale Accelerator Driven System (ADS), XT-ADS, as well as the conceptual design of the mid-term, larger scale European Facility for Industrial Transmutation, EFIT. This project, started in April 2005, is scheduled for 48 months. The EUROTRANS project is the logical follower of all FP5 projects clustered under the FP5 ADOPT Thematic Network except those related to partitioning; it has been divided in five sub-projects, called "Domains" and an important aspect is the coherence between the different domains. I. 2.
3.
4.
5.
The design domain, called "DM1 DESIGN", is the follower of PDSXADS, The domain related to the experimental coupling of an accelerator, a spallation target and a sub-critical system, called "DM2 ECATS" is the follower of the MUSE project, The domain related to the study and development of advanced fuels for transmutation, called "DM3 AFTRA" is the follower of the FUTURE, CONFIRM and THORIUM advanced fuels FP5 related projects, The structural materials and HLM technology domain, called "DM4 DEMETRA", follows also logically the SPIRE, TECLA and ASCHLIM FP5 projects, While the nuclear data relevant for transmutation studies are covered by a fifth domain, called "DM5 NUDATRA" which is the follower of n-TOFADS and HINDAS
SCKCEN is the coordinator of DM1; many partners, from universities (KTH, UPM, IAP-UF.. .) through research centres (CEA, CNRS, FZK, ENEA, CIEMAT.. .) to industrial partners (ANSALDO, AREVA, SUEZ-Tractebel Engineering, Empresarios Agrupados, OTL.. .), are involved in this domain. The EUROTRANS project has also been the opportunity to add key partners in our global collaboration picture, we may mention among others:
0 0
CNRS (FR), CEA (FR), INFN (IT) and IAP-FU (D) for the accelerator; Empresarios Agrupados (SP) for the plant layout; Ansaldo Nucleare (IT) and AREVA (FR) for the primary system design.
Different components of the MYRRHA Draft-2 file have been considered and possibly simplified in the XT-ADS design: primary systems, sub-critical core, spallation loop and general layout of the MYRRHA building including remote handling [3]. It should be mentioned that some hypotheses taken in the MYRRHA Draft-2 have been questioned by the industrial partners (fuel loading
213
from the bottom of the core is a good example) but after comparison with more "classical" solutions, all partners had to admit that the logic behind the Draft-2 was absolutely coherent and that feasibility was not compromised at all. The aim in DM1 is to obtain at the end of the EUROTRANS project an "advanced" design of XT-ADS while the MYRRHA Draft-2 was limited to a "conceptual" level. By advanced design, we mean that it is mature (among others that all engineering specialities from the neutronics through the mechanics to the civil engineering have been correctly addressed) enough to be the basis to a "detailed engineering design" phase.
3.5. The R&D activities (2005-2009) The R&D programme that is required to support the design of MYRRHA is divided into several major topics [4]. These include structural materials and fuel research, Pb-Bi coolant chemistry control and support experiments for the development of specific components such as e.g. the spallation target, in-vessel remote handling, in service inspection, etc. It is clear that the full programme is too extensive to be carried out by employing only the limited resources at SCK-CEN. To resolve this issue, the MYRRHA project has been embedded in different European 6'h framework projects. The largest of these is the integrated project EUROTRANS described above in section 3.4. Some aspects of the support R&D programme are included in other ADS and GenIV related European initiatives like e.g. VELLA, MTR+13 and ELSY. The strategy of the SCK-CEN R&D effort in these is to focus on issues that are directly linked to design work at SCKCEN or on key technologies that are either not covered elsewhere or are of basic nature for the exploitation of ADS device. In this respect, the SCK*CEN participation in structural materials and fuel research are embedded in strategic lines outside our ANS intitute and are not discussed in this paper. It may be noted here that the proof-of-principle experiments are a first step in the development of technological components that can be employed in the MYRRHA ADS. The proof of principle of ADS as such has been conducted at SCKCEN through its participation in the MUSE FPYCEA project and the verification of its soundful and controllable operation is to be continued through the GUINEVERE project embedded in the EUROTRANS DM2 domain. 3.5.1. Spallation target support R&D A significant part of the R&D efforts at SCKCEN is aimed at supporting the design of the MYRRHA spallation target. Indeed, the target design can largely
214
draw upon the valuable information gained from the MEGAPIE project. However, because of the specific option of having a windowless target, significant addition R&D work is required. This work is concentrated on understanding and controlling the flow dynamics of a windowless liquid metal target on the one hand and on demonstrating the compatibility of the free target surface with the vacuum requirements of the accelerator beam. 3.5.2. Instrumentation development
The particular nature of a heavy liquid metal cooled ADS system requires the development of specific instrumentation. The R&D efforts at SCK-CEN are concentrated on two devices. The first is a so called LIDAR (LIght Detection And Ranging) system that will be employed for accurate and fast determination of the level of the LBE free surface in the MYRRHA spallation target. The second is an ultrasound camera to be used for imaging components that are submerged in the LBE pool of the MYRRHA main vessel for in-service inspection. 3.5.3. LBE submerged remote handling
The operation of MYRRHA relies heavily on remote handling. For applications in a gas atmosphere, albeit in harsh radiation conditions the feasibility of remote handling has been demonstrated in fusion applications. However, robotic devices capable of reliably operating within LBE are presently not available. To develop a functional apparatus for integration into the remote handling scheme it is first necessary to identify a suitable (representative) type of arm actuators and joints to be used for remote handling inside the MYRRHA vessel and to identify the critical components (gears, harmonic drives, bearings, etc.) The harshness of the LBE operating environment means that candidate joint/actuator components must be proven experimentally under representative operating conditions. Once a complete set of components has been successfully selected, a full robotic arm joint and actuator can be constructed to be used to demonstrate the feasibility of liquid LBE submerged remote handling in representative experimental conditions. At present the first two steps of this programme have been completed in collaboration with "Oxford Technologies ltd" (OTL) [ 5 ] . In addition a concept design of a test rig including appropriate instrumentation to be submerged in a LBE container has been produced. The next steps in the programme are the construction of the test rig and the performance of the components test.
215
3.5.4. In-pile LBE materials test loop In the materials test programme for MYRRHA, it is important to address the issues related to materials compatibility and integrity with Pb(-Bi) under the combined effects of elevated temperatures, irradiation, corrosion and mechanical load. For this purpose, it is planned to develop a design and investigate the technological feasibility of a heavy liquid metal loop that can be readily inserted in one of the available experimental channels of the material test reactor BR2. This facility will be unique in Europe and will be serving the community to address the issues related to materials compatibility and integrity with Pb(-Bi) under irradiation, corrosion issues on cladding and reactor internals, crack growth test under constant load, creep behaviour etc. The development of this type of device, however, is not straightforward. For this reason we will construct an out-of-pile mock-up of the in-pile materials test loop. The mock-up loop has to provide the necessary out-of-pile testing setup to perform experiments in order to check the safe operation of such a loop and to address the issues needed for the licensing procedure in accordance with the rules of the Belgian safety authority. Obviously, in order to create a representative model, at least the conceptual design of the in-pile loop should be completed. Since the project only started late 2006, the development of the inpile loop is presently limited to the investigation of the external and operational boundary conditions and the definition of the conceptual layout. 4. Objectives for the next five years 4.1. Engineering design
We foresee four main objectives or milestones in the engineering part: end of 2008: a preliminary decommissioning plan is submitted to the waste management authorities (NIRASIONDRAF). early 2009: completion of the EUROTRANS project. Here the objective is to obtain an advanced design file, in the continuation of the Draft-2 published in 2005. end of 2009: a preliminary safety assessment report and an environment impact assessment are drafted, in order to be submitted to the authorities (FANC) in the early 2010. The aim here is to obtain in 2013 the authorization of construction. in the years 2009 - 2011: the various technical issues of different components have to be answered
216
end 201 1 - early 2012: completion of the "detailed engineering design". The aim is to be able to write down the technical specifications of the different MYRRHA / XT-ADS components (and that those components are delivered in 20 13 - see the business plan for more details on the key dates). 4.2. RLD programme The main objectives for the next five years are listed below: Completing the support experiments for the spallation target design to a confidence level that the feasibility of a windowless spallation target can be demonstrated followed by the construction of a spallation target mock-up loop for component testing and validation. Development of a working LIDAR system operating with radiation hard electronic components. Demonstration of a working ultrasound camera that is capable of reconstructing images in a submerged MYRRHA like geometry Proof of principle of the feasibility of liquid metal submerged remote handling by validation of o critical components of a representative robotic arm o testing of a full robotic arm joint and actuator in MYRRHA like conditions. Demonstrating the feasibility of an in-pile HLM materials test loop by long term validation of the operation of an out-of-pile mock-up. Principal activities required to reach the objectives The objectives mentioned above will be reached by executing the following projects and activities: 5.1. Engineering design
First until 2009, we need here to finalise the work started in the domain DM1 of the EUROTRANS project: core design, primary systems, spallation design, safety calculations are the SCK-CEN activities in the share of the whole design work. Then we need to set-up a "Central Design Team" and submit a proposal of such kind for the coming call of FP7 by the European Commission where a "Central Design Team for a fast spectrum transmutation device" is foreseen. This "Central Design Team" will consist of about 13 to 15 persons in a threeyear project covering the period 2008-2009-2010. The different members for the
217
Central Design Team will be coming from different European research institutes and industrial partners and will most of them be stationed at the Molsite. When the owner consortium has been put in place and financing is secured (around 2009-2010), additional engineering design manpower will be used to prepare the detailed engineering design by the end of 201 1 -beginning 2012. In the years 2009 - 2011, the following technical issues of different components have to be answered: primary pumps: appropriate impeller material or coatings, resistant to LBE velocities up to 10 d s ; geometrical sizing of axial impellers for LBE has to be performed heat exchangers (HX): oxidation layer on LBE side of tubes (influence on heat conductivity); boiling water (waterhteam flow stability?); consequences of HX tube rupture. in-vessel manipulators materials o constraints: operation at temperatures up to 350°C in LBE; all mechanisms and bearings are immersed in LBE - the lubricating effect of LBE is not known but it is assumed to be poor as it has a low viscosity. o consequences: the construction materials must be tolerant to the corrosive effects of immersion in LBE and have useful strength at 350°C; the use of most forms of lubricants and polymers is precluded. The use of ceramics is attractive for bearing elements. They can operate without lubrication. Corrosion, wear resistance and low friction for the metal to metal moving parts could be improved by the use of coatings. (TiN, WS2, Nitron MC, Tantalum, DLC). manipulators: some major aspects of manipulator design: appropriate transmission elements (gearboxes, linkages); bearing technologies (ball bearings, plane bearings and wire race bearings); the need to accommodate the 50 mineral insulated cables for the ultrasonic camera; design and manufacturing of primary system components (HX, pumps) and the in-vessel remote handling manipulators rigs for tests to be performed; testing PHX, PP and different technologies for the manipulators; final design primary system components; in parallel with primary system components: perform the design of the secondary and tertiary cooling system, vessel, core support structure, diaphragm.
218
In the years 201 1-2012 we foresee the completion of the detailed engineering design, the technical specifications of the different components (and that those components are delivered in 2013 - see the business plan for more details on the key dates) and the call for tenders 5.2. R&D programme 5.2.1. Spallation target support R W The Flow dynamics experiments performed up to now have shown that a stable compact windowless spallation target design using a vertical confluent flow in a funnel-type geometry is feasible. Further experiments, ongoing until 2009 are planned to investigate new design features of the spallation target nozzle that are aimed at providing additional flow stability at the free surface. These include detailing the hydraulically behaviour of the feeder line, flow detachment, swirl and the design of the return line. In parallel, the vacuum interface compatibility programme should be completed by 2009 as well. This includes finalisation of the evaporation experiments mentioned above and Monte Carlo modelling of the emanation of volatile spallation products in the spallation target loop. The final outcome of this work is an optimised design of the vacuum and spallation product confinement system of the spallation target. The next step is the construction of a spallation target mock-up in which the different components of the spallation target loop (target nozzle, feeder line, LBE pumps, vacuudgas confinement system, LBE chemistry control, LIDAR, etc) can be validated. This should be completed by 2012. 5.2.2.
Instrumentation development
The instrumentation development activities are concentrated around the development of two major components being a LIDAR set-up for the detection of the free surface level in the spallation target and an ultra-sound camera for "under LBE" imaging in the main vessel of the ADS. 5.2.3. LBE submerged remote handling The validation of remote handling in a LBE environment requires a number of steps:
219
identify a suitable (representative) type of arm actuators and joints to be used for remote handing inside the MYRRHA vessel identify the critical components (gears, harmonic drives, bearings, etc.) Experimentally validate operation of candidate joint/actuator components under representative operating conditions o design test rig o construct test rig o perform validation experiments on representative components a full robotic arm joint and actuator can be constructed to be eventually used to demonstrate the feasibility of liquid LBE submerged remote handling in representative experimental conditions. The first two steps and the conceptual design of the test rig have been completed. The engineering design, the construction of the test rig and the component tests are planned for 2008-2009 whereas the final step of constructing and validating a representative full robotic arm joint and actuator will be undertaken in 20 10-20 1 1. 5.2.4. In-pile LBE materials test loop The support activity for the development of the in-pile LBE materials test loop consists of the design and construction of an out-of-pile mock-up loop. Its purpose is to investigate andor demonstrate its long term stable and safe operation. The loop will be based on the conceptual design of the in-pile materials loop. Its design and construction is scheduled for 2008-2009. 5.3. Project organisation
From 2009 on MYRRHA will be organised as an International Entity with: An International Advisory Council (IAC), An Owner Consortium (OC), A MYRRHA Central Project Management Team (CPMT), A Users/Customers Group, The work of the International Advisory Council (IAC) would be to guide the hrther development of the facility, the settlement of the Owner Consortium Group (OE) and the establishment of a Users/Customers Group of the facility. The Owners Consortium (OC) will consist of partners that are interested in the realisation and later on in the operation of MYRRHA. These partners are acting
220
in their name or as representative of their countries. Both possibilities are considered presently but will be fixed in the coming two years after negotiation with the representatives of the Belgian government and some European research institutions interested in the MYRRH A project.
MYRRHA Project Central Team | Assisted by an Owner Engineering Team }
The OC will be taking the overall responsibility of the project and will be led by SCK'CEN as it is the hosting organisation of the facility on its site. The OC will establish a MYRRHA Central Project Management Team (CPMT) based at Mol and responsible for the daily project management, control and follow-up of the project. The CPMT is composed of a Project management team (PMT) and an Owner Engineering team (OET) as indicated above. The PMT will be a team of 25 to 30 persons, whereas the OET will consist of 30-35 persons. Also, a Users/Customers group will be set up to define the future needs for the installation and to survey its possible realisation in the facility. At present time, the organisation (and the contacts at high level) at Belgian level is under way.
22 1
6. Conclusion MYRRHA is in our opinion essential for the existence of SCK-CEN. Our large infrastructure BR2 is today more than 40 years old. Building a new facility guarantees the future of our research centre. The MYRRHA project answers the requirements of such a new facility. As a fast spectrum facility, it will enlarge the current irradiation capabilities and place SCKCEN in a central position or as a key partner in new research programmes requiring irradiations. Such project needs huge resources and a strong support from our government is absolutely essential. Also a change in the mentality from a national like research institution towards an open-users facility and as such as an open organisation towards the international research community is mandatory and MYRRHA can be the trigger to reach this objective. Anyway, even if for any reason the MYRRHA facility would not be built, all the effort spent in design as well as in R&D support programme would not be jeopardised as it can be valorised within the Gen IV International Forum. References 1. H. Ait Abderrahim, D. De Bruyn, e.a., MYRRHA Project - Technical description, SCK-CEN Report ref. ANS/HAA/DDB/3900.B043000/85/0717bis, April 2007,57 p. 2. H. Ait Abderrahim, D. De Bruyn, M. Giot, P. Van Doorslaer, F. Hardeman, MYRRHA Project, Multi-purpose h Ybrid Research Reactor for High-tech Applications at Mol (Belgium), Business Plan 2007, SCKCEN report ANS/HAA/DDB/3 1.B043000/85/07- 17. 3. D. De Bruyn, D. Maes, L. Mansani, B. Giraud, From MYRRHA to XTADS: the design evolution of an experimental ADS system, 8th International Topical Meeting on Nuclear Applications & Utilization of Accelerators (AccApp'07), Pocatello (Idaho), United States, 30 July - 2 August 2007. 4. P. Schuurmans, K. Van Tichelen, M. Dierckx, H. Ait Abderrahim, A. Guertin, T. Kirchner, e.a., Design and Supporting R&D of the XT-ADS Spallation Target, 9th Information Exchange Meeting on Partitioning and Transmutation.- Nimes, France, 26-29 September 2006. 5. P. Schuurmans, S. Mills, D. Locke & R. Meek, ISI & R remote handling proof of principle. Phase 1final report and design of test rig for phase 2, Deliverable D4.17 for the EUROTRANS project, November 2006,29 pg.
This page intentionally left blank
THE BR1 REACTOR: A VERSATILE TOOL FOR FISSION EXPERIMENTS J. WAGEMANS SCK*CEN,Boeretang 200 2400 Mol,Belgium The BRl reactor located at the Belgian Nuclear Research Centre SCKCEN in Mol, Belgium, is a research reactor with a variety of irradiation possibilities. Thanks to its large reactor core, its flexible operation and its different irradiation facilities, this reactor is particularly suited for in-core and ex-core neutron physics experiments. This paper gives a general description of the BRI reactor, with special emphasis on the available irradiation possibilities. Then some examples of fission experiments that have been performed in the past will be referred to and two ongoing projects related to fission will be presented.
1. Introduction Despite its relatively low neutron fluxes, the BR1 reactor is a valuable tool for neutron physics experiments. The main advantages of this reactor are: 0
0 0
0
Its very flexible and stable operation. The reactor can be easily started up and shut down. Moreover thanks to the thick heavy concrete shielding human presence on and near the reactor is allowed during operation. The good thermalisation of the neutrons in the graphite moderator. Low neutron flux gradients for in-core experiments thanks to its large reactor core. A variety of irradiation possibilities, including a series of irradiation channels, a pneumatic rabbit system, standard neutron and gamma radiation fields and a neutrography installation.
This paper first presents some general characteristics of the BR1 reactor. Special attention will be given to the different irradiation possibilities. Then some past and ongoing fission related experiments will be highlighted. 223
224
2. General description of the BR1 reactor
The BR1 reactor is a research reactor that uses natural metallic uranium as fuel, graphite as moderator and circulation of air for the cooling. The maximum power of the reactor is 4 MW but since several years BR1 is operated on a daily basis at a maximum power of 700 kW or during a few hours at 1 M W . The reactor consists of a matrix of 14500 blocks, with a total volume of 6.7 x 6.7 x 6.8 m3. In this volume 829 fuel channels are foreseen, 569 of which are loaded with fuel. The core loading is approximately cylindrical with a diameter of 4.7 m and a length of 4.9 m. The fuel channels have a square section of 5 x 5 cm2 with a lattice pitch of 18 cm and contain 23 fuel rods each. These rods have a length of 2 1.4 cm and are made of natural uranium contained in an aluminium cladding. After more than 50 years of operation the BR1 reactor is still operated with its original fuel, the burnup being less than 1 %. 3. The main irradiation facilities of the BR1 reactor
The most important irradiation facilities of the BR1 reactor nowadays are schematically shown in Figure 1: About 50 irradiation channels penetrating the reactor core, parallel or perpendicular to the fuel. A vertical thermal column containing a spherical cavity with a thermalised neutron spectrum that is mainly used as standard neutron field. A pneumatic rabbit system that allows in-core irradiations with a very short transfer time. The irradiation channels can be used for in-core and ex-core experiments. Two types are available: channels with a cylindrical section of 8 cm diameter and channels with a square section of 10 x 10 cm2. These channels can be equipped with a throughput in the beamstop which allows on-line monitoring of the experiments. The maximum thermal neutron flux (in-core) at 1 MW is about 3.5 10" cm%'. In the centre of the core the neutron flux gradient is about 10 % per meter. Moreover the neutron beams at the entrancelexit of the channels are very thermalised and therefore particularly suited for experiments with thermal neutrons. The thermal neutron fluxes close to the channel entrance/exit are of the order of lo6 cm-2s-'at 1 MW.
225
Different standard neutron fields are available in the BR1 reactor. As it can be seen on Figure 1 the graphite (checkered part) is extended vertically to constitute a thermal column above the reactor core. In this column a spherical cavity of 1 m diameter was machined in the graphite. The fast neutrons coming from the fission reactions in the core are completely thermalised before arriving at the cavity. Therefore the neutron flux in this cavity follows a purely Maxwellian distribution at room temperature. The thermal neutron flux in the centre of the cavity at 1 MW is 1.0 109 cm'V.
Large Cavity
Pneumatic rabbits
Irradia Channel
'" Graphite Reactor Core (natural U) Figure 1. Schematic view of the main irradiation facilities at BR1.
Inside the cavity three other types of standard irradiation fields can be obtained. For this purpose converters have to be loaded in the cavity. We distinguish three different types:
226 0
0
0
Thermal neutron induced 235U(n,f)fast neutron fields. For this purpose cylindrical converters made of 235U enriched foils mounted on a cadmium tube are loaded in the cavity. The most frequently used converter is the so-called MARK I11 tube. This is a 93 wt.% enriched 235Ufoil wrapped around a cadmium tube with an inner diameter of 50 mm. The 235U(n,f)fast neutron flux at 1 MW is 3.4 lo8 cm-2s-1. Prompt capture gamma-ray fields. For this purpose cylindrical converters made of so-called thermal neutron black absorber materials are loaded in the cavity. The black-absorber foils (Cd, B or Co) are mounted on a pyrex tube that effectively stops p and Xrays. Mixed neutron-gamma fields. For this purpose spherical converters are loaded in the cavity through a large access plug. Various configurations of spherical shells of Fe, Cu and U can be chosen depending on the desired gamma and neutron spectrum.
The normalisation of the neutron irradiations performed in the cavity (empty or with the Mark I11 converter loaded) is performed using a calibration factor that is constant for all reactor powers. This has been validated for a reactor power ranging from lo2 to lo6 W. Using such a calibration factor is more accurate than using the reactor power level. This calibration factor has been derived from simultaneous neutron activation dosimetry measurements in the centre of the cavity and count rate records obtained from a fission chamber. The count rate of this fission chamber (that is located outside and close to the cavity) is permanently recorded and is independent of any device that is loaded in the cavity. For the neutron activation measurements a series of gold and indium foils (pure and alloyed) are used and their activity was accurately determined using high-purity germanium detectors. Various measurement campaigns have been performed at different reactor powers, resulting in a calibration constant known with an accuracy of 1.8 % (lo). BR1 is equipped with a pneumatic rabbit system that permits injection of samples into different positions in the reactor core. One of these rabbits has a very short transfer time of about one second and samples can be ejected directly into a germanium detector measuring system. This allows measuring neutron activation products with half-lives down to only a few seconds. The main application of this facility is neutron activation analyses. Applying this technique using the fast rabbit system at BR1 allows detection of a wide range
221
of elements with high resolution, even down to the ppb level depending on the isotope. 4. Fission experiments at the BR1 reactor
Since the start of its operation in 1956 the BR1 reactor has been actively used for a variety of physics experiments, including fission studies. These studies focused on cross section data, identification of new isotopes in fission, fission fragment spectroscopy, etc. In order to illustrate different types of fission studies performed at the BR1 reactor a few references are cited [ 1-51. Currently two experimental campaigns related to fission are ongoing at the BR1 reactor: 0
0
Measurement of fission rate ratios in U02 of short-lived fission fragments. Determination of the neutron induced fission cross sections of a number of Cm isotopes at one of the BR1 irradiation channels.
Both experiments will be briefly explained below. 4.1. Measurement offission rate ratios
Within the framework of a trilateral collaboration between SCK-CEN (Belgium), PSI (Switzerland) and CEA (France), a series of experiments is performed at the BR1 reactor. These experiments are a first step towards developing a system to measure fission rates in fresh and burnt fuel irradiated in the PROTEUS reactor at PSI [6]. Whereas for fresh fuel suited techniques exist (e.g. [7]), this is not yet the case for burnt fuel. The approach followed here is to investigate the feasibility of using high-energy y-rays (> 2 MeV) from newly produced (short-lived) fission products in burnt fuel. The first step is to identify suited short-lived fission products after irradiation of a fresh UOz pellet in the BR1 reactor. A series of irradiations of fresh U 0 2 fuel pellets with 235Uenrichments of 3.3 ‘YOand 4 % is performed in the BRl reactor core using its fast rabbit system. The irradiation time and reactor power were varied for the different irradiations. After the irradiation the pellets were immediately sent to an installed germanium detector for measurement of the y-activity. Different filters and source-todetector distances were used, as well as a I3’Cs source in the lead castle to simulate the background of a fuel rod. The gamma-spectra that have been recorded are currently being analysed to identify suited short-lived fission products. Taking into account the different
228
required criteria, preliminary analyses indicate that the fission products "Y and Cs are the most promising candidates for the envisaged application. Probably new experiments will take place at BR1, possibly deviating the fast rabbit to the empty cavity. [8] 138
4.2. Determination ofCm(nJ) cross sections
The Interuniversity Attraction Pole (IAP) programme is an initiative of the Belgian Government to support basic research teams from Belgium to work in networks. The first phase of this IAP programme started already in 1987. SCK-CEN joined this IAP programme for the first time in Phase VI which started in January 2007 and will last five years. Together with its partners, University of Leuven (KUL), University of Brussels (ULB) and University of Ghent (UGent), the so-called BriX network was established [9], in which seven research objectives are identified. One of these projects is the investigation of Cm(n,f) cross sections that will be partly performed at the BR1 reactor. The knowledge of accurate cross section data is of major importance for a wide variety of applications. Cross section data are of interest for basic nuclear physics and astrophysics studies and are a crucial input in reactor physics calculations. For a large series of nuclides accurate cross section data are readily available, but this is less the case for some short-lived nuclides or rare actinides. An example of important but yet poorly known cross section data are the neutron induced fission cross sections of the Cm isotopes. Indeed, the current understanding of the astrophysical r-process nucleosynthesis process indicates that it is terminated by spontaneous, P-delayed and neutron-induced fission. The Cm(n,f) cross section data are therefore part of the input of r-process nucleosynthesis network calculations. Moreover burnt nuclear fuel contains important quantities of Cm isotopes hence the Cm(n,f) cross section data are crucial input for nuclear waste transmutation calculations. The objective of the project is to experimentally determine neutron induced fission cross sections on a series of Cm isotopes. The major part of the work will be the determination of the thermal 243Cm,245Cm,247Cmand 248Cm(n,f) cross sections at the BR1 reactor. A dedicated setup will be installed at one of the BR1 irradiation channels. The thermal neutron flux at the sample position will be of the order of lo6 cm%-' at a reactor power of 1 MW. The measurements will be performed using well-characterised thin Cm deposits. The fission fragments will be detected in a 27c geometry using a gridded ionisation chamber.
229
References 1 . A. J. Demytter, Reactor Science and Technology 15, 165 (1961). 2. P. del Marmol, F. Hanappe and M. Monsecour, Journal of Inorganic and Nuclear Chemistry 35,4323 (1973). 3. P. del Marmol and M. N6ve de MCvergnies, JournaZ of Inorganic and Nuclear Chemistry 29,273 (1967). 4. P. del Marmol and P. Fettweis, Nuclear Physics A194, 140 (1972). 5. L. Dematt6, C. Wagemans, R. BarthClCmy and A. J. Demytter, Nuclear Physics A617,331 (1997). 6. F. Jatuff, PSI Project Proposal LIFE@PROTEUS. 7. U. C . Bergmann, R. Chawla, F. Jatuff and M. F. Murphy, Nucl. Instr. And Methods A556,33 1 (2006). 8. H. Krohnert, M. F. Murphy, G . Perret, M. Plaschy, J. Wagemans and R. Chawla, to be presented at the Int. Conf: on Reactor Physics, Interlaken (Switzerland), September 2008. 9. http :Iliks32.fy s.kuleuven.behap-wikilindex .phpiMain-Page .
This page intentionally left blank
“Special” Fission Processes
This page intentionally left blank
SHAPE ISOMERS
- A KEY TO FISSION
BARRIERS
S. OBERSTEDT', F.-J. HAMBSCH, N. KORNILOV, G. LOVESTAM EC-J R C Institute for Reference Materials and Measurements (IRMM), B-2440 Geel, Belgium *E-mail: stephan. oberstedt ec.europa. eu A. OBERSTEDT Department of Natural Sciences, Orebro Universitet, S-70182 Orebro, Sweden E-mail: andreas.oberstedt nut. oru.se M. GAWRYS Department of Fundamental Science, Chalmers University of Technology, S-41296 Goteborg, Sweden E-mail: gawrys post.utfors.se Quantitative predictions of fission product yields are relevant for the reliable operation of different modern nuclear applications. This concerns the realistic characterizations of the radio-toxicity of the fuel elements after the envisaged extended irradiation, as well as sub-critical assemblies, where the number of delayed neutrons from minor actinides is determined by the characteristic emission yields of the corresponding so-called pre-cursor isotopes. However, to be able to make more reliable quantitative predictions of fission characteristics requires the better understanding of the fission process itself. For this purpose a better knowledge about the distinct structure of the nuclear energy landscape around the fission barrier is indispensable. In particular, the question should be answered, whether the fission barrier is either double- or triple-humped or even multi-humped as been proposed within the multi-modal neck rupture model. Despite quite some effort based on different experimental techniques and theoretical approaches, this question remains still unanswered. There is still no consistent picture of the fission barrier available and hence, different sets of barrier parameters are in use, unable to describe the different observed phenomena in a coherent way. With the systematic investigation of shape isomer population, its decay modes as well as the branching ratio, precise information can be obtained to resolve the puzzling situation. The experimental approach will be discussed and results from first experiments presented. Keywords: Fission isomers; shape-isomeric decay half life; fission half life sysodd-A uranium isotopes. tematic; fission barriers; barrier parameters; 235fU;
233
234
1. Int r oduc t i on
Quantitative predictions of fission-fragment yields are relevant for a successful design and reliable operation of nuclear applications. In the next generation nuclear power devices with the envisaged extended irradiation of the fuel element the radio-toxicity must be well characterized. For subcritical assemblies correct delayed-neutron pre-cursor isotope yields are of utmost importance for a reliable and save operation. In order to meet the requirements on knowledge about fission-fragment characteristics a better understanding of the fission process as such is required. For this purpose the distinct structure of the nuclear energy landscape around the fission barrier has t o be known better. In particular, the question should be answered, whether the fission barrier is either double- or triple-humped, or even multi-humped as proposed within the multi-modal neck rupture Despite quite some effort based on different experimental techniques and theoretical approaches, this key question remains still unanswered. The idea that fission must be described in a multi-dimensional parameter space of the nuclear energy land, ~ the concept scape has been brought forward already in the early 1 9 5 0 ~and of multi-modal fission was fully developed in the late 1980s by Brosa and co-workers,' leading to a qualitative understanding of fission-fragment emission yields. The existence of local minima of the nuclear energy at larger deformations, namely super- and hyper-deformation, has been deduced from the observed sub-threshold fission cross-section, the so-called intermediate structure (IS )4,5 as well as from vibrational resonance structures around fission threshold.6-8 The observation of delayed fission has been explained as the decay of a meta-stable ground state populated at super-deformation (SD),9 being again a confirmation for the at least double-humped structure of the fission barrier. In the following different experimental approaches to reveal the structure of the fission barrier will be discussed. In particular, the systematic investigation of shape-isomer population and decay modes will be emphasized as a very promising strategy. However, in the following discussion we suppose, as experimentally proven, that the fission barrier is at least double- humped. 2. Experimental approach to fission b arriers
In principal the fission barrier is well characterized by the following parameters: the heights of the inner and outer barrier(s), EA and Eg (Ec), re-
235
v
neutron energy lev)
n
c
DEIO(IMA'I0N
Fig. 1. Left: Intermediate structure in the neutron-induced fission cross-section of 240Pu (upper part) and 234U(lower part). DI is the average spacing of resonances located above the normal ground state (class-I), D I I is the average spacing between two intermediate structures located around a class-I1 state at energy EcII; right: Schematic representation of the population of a shape isomer via charged particle-induced neutron evaporation reactions (adapted from Ref. 15) used to measure the relative delayed fission probability.
spectively, their corresponding barrier penetrabilities as well as the ground state energies at super- and possibly hyper-deformation. The most simple probe to investigate fission barrier properties in more detail is the photon. In photon-induced fission essentially only electric dipole and quadrupole excitation occurs. From the analysis of fission fragment angular distributions the relative height of inner and outer barriers may be deduced." As a result, the outer barrier decreases relative t o the inner barrier with increasing nuclear charge Z. Photons also permit the investigation of fission at excitation energies well below the neutron separation energy, S,, which is typically between 5 - 6 MeV. From the shape of the (r,f ) crosssection as a function of excitation energy the energy of the shape isomeric ground state may be determined." The extremely low cross-section a t very low photon energy in conjunction with a limited photon flux makes those experiments rather difficult. From investigating delayed fission induced by photons some conclusions have been drawn about the existence of a third minimum in 232Thand 238U.11112 A fingerprint of intermediate minima in the nuclear energy as a function of nuclear deformation is the observed sub-threshold resonance structure in the neutron-induced fission cross-section of non-fissile isotopes as e. g. for 240Pu4and 237Np.5This phenomenon is explained as the coupling of class-I states, excited above the normal ground-state, to so-called class-I1 states located above the super-deformed ground-state (see Fig. 1). From the av-
236
Fig. 2. Left: Example for the model-dependent extraction of the shape-isomeric ground state energy, EII, in 238U and the introduced uncertainty (see text). The broken lines towards zero-yield is a guide to the eye; right: Fission isomer half life systematic according to Ref. 17(upper part) and, schematic representation of the fission barrier and possible decay modes after neutron absorption (lower part, see text).
erage distance of IS fission resonance clusters, DII, relative t o the spacing DI between individual resonances belonging to one particular IS, the excitation energy E* above the super-deformed minimum may be deduced.13 The interesting shape-isomeric ground-state energy EII = S, E, - E* is very much dependent on the assumptions made on the level density and, therefore, model dependent. In some cases it is difficult to identify the position of the underlying class-I1 state of an IS and to determine even the distance between two fission resonance clusters as shown in Fig. 1 for the reaction 234U(n,f). The analogous investigation of resonance structures in (d, p) reactions at E* below neutron threshold has been successfully used to reveal the nature of hyper-deformed states at even larger nuclear deformations (see e. g. Ref. 8,14. However, information about barrier heights and the corresponding penetrability from analyzing neutron-induced fission crosssections around first chance fission can neither explain sub-threshold fission nor give the right transmission factors for fission shape-isomers. l5 Furthermore, from cross-section analysis a shallow third minimum is suggested for thorium and protactinium isotopes with a depth smaller than 1 MeV, whereas the resonance structures observed in (d, p) reactions on uranium isotopes are consistent with a depth of at least 2 MeV.
+
237
Another approach is to populate the fission isomer in a particle-induced neutron-evaporation reaction and t o investigate the relative delayed fission probability (see Fig. 1).Extrapolation to zero-yield gives the shape-isomeric ground-state energy. Fig. 2 shows the relative delayed fission yield from 238U excited by means of inelastic neutron scattering. It is evident how the extracted isomeric ground state energy, EII, again depends on the particular model calculation indicated by different lines to guide the eye.15 Although results obtained from different reactions leading to the same compound nucleus, e. g. ( a ,2n) and (n, y), are in reasonable agreement with each other, uncertainties in the underlying model parameters do not allow t o determine EII t o a precision better than 0.5 MeV.15 Since the early 1960s, 33 fission isomers have been discovered in nuclei ranging from 236U t o 245Bk and the half lives measured. Only recently the fission of a shape isomer in an odd uranium isotope, 235U, has been identified.16 Half life systematic trends based on the experimental data are able to describe reasonably well shape isomer half lives in the range between Pu and Bk isotopes. However, in light actinide nuclei, where ydecay competes with fission, the predicted half lives may be wrong by orders of magnitude, since only the partial fission half life may be calculated. This is shown in the upper part of Fig. 2, where the vertical dashed line separates nuclei, where shape isomer decay essentially proceeds through fission (right side), from those showing also shape-isomeric y-decay (left side). The big full symbols show the few known shape isomers in this region and how much the measured decay half life may deviate from the partial fission half life obtained from systematic trends. The precision of the extracted barrier parameters must obviously be rather limited. Even worse for the design of an experiment is that different systematic approaches may differ by orders of magnitude for the same isotope.17-lg Together with the extremely low production cross-section of shape isomers, typically smaller than 20 pb, the measurement of shape isomer decay data is a challenging venture for the experimentalist. 3. Direct survey of fission barriers
In order to avoid uncertainties inherent to the different methods discussed above, we suggest t o systematically investigating the population and decay of shape isomers despite their low production cross-section. As it will be discussed in the following, this systematic approach will provide all characteristic barrier parameters, a s the inner and outer barrier height, EA and Eg, as well as the corresponding penetrability and, last but not least, the
238
0
,
I,
,
,
,j
I
I
600
f 400 f I: 200
238”
Fig. 3. Left: Direct determination of the shape-isomeric ground state energy in 238U by measuring the delayed y-decay back t o the normal ground state; right: Determination of the shape-isomeric ground state energy in 239Uby measuring the y-decay populating the ground state after resonant neutron-capture into a class-I1 state.
energy of the super-deformed and possibly hyper-deformed, EII and E I I I . The most straight forward strategy is to measure the decay half life after the shape isomer has been populated in a charged-particle induced reaction or after neutron capture. As depicted in the lower part of Fig. 2, after absorption of the neutron the excited state may, if not directly fission, decay t o lower lying states. If the state couples to a class-I1 state in the second potential well, this state may decay to the ground state and populate the shape isomer. Using neutrons with an energy coinciding with a resonance of an IS then the difference between the neutron separation energy and the total y-ray energy directly gives EII. Alternatively, the delayed fission of a shape isomer can be measured in coincidence with the prompt decay y-rays to obtain EII. If the isomer may also decay by y-ray emission t o a level above the normal ground state the subsequent y-ray cascade to the ground state, is a direct measure for EII as well. From the branching ratio between shape-isomeric fission and y-decay the relative height of EA and Eg may be determined. After determination of the parameters above, the absolute barrier heights and corresponding penetrabilities may be obtained from fission cross-section measurements. If sticking to the approximation of inverted parabolas, a smaller value for the outer penetrability would point directly to a triple-humped barrier.
239
In recent years several successful experiments with some of the above outlined methods were performed, but only for three even-even isotopes the strategy has been applied more or less completely, 236U, 238U, 240Pu,viz. For the uranium isotopes the branching ratio, the half lives and EII were measured including detailed spectroscopy in the second potential well. The same is the case for 240Pufor which, however, no y-decay branch exists. These experiments and results are excellently summarized and discussed in Ref. 8. Fig. 3 shows the measured decay scheme of the shape isomer level in 238Ufixing Err to 2.559 MeV. Comparing this precise result with the data shown in the left part of Fig. 2 demonstrates the strength of this experimental approach. It does not only give a more precise value, but also rules out that particular model calculation, which seems to describe much better the experimental relative delayed fission yield data. 239U is the only odd-uranium isotope for which y-decay populating the shape isomeric ground state has been observed in a neutron-capture experiment with resonance neutrons at around 721 eV (see Fig. 3, right). From this experiment Err could be determined very precisely.20>21 In a recent experiment the y-decay towards the normal ground state has been searched for,22apparently well confirming the results from the previous experiment. However, to date no decay half life could be determined from the experimental data. Very recently a fission shape isomer has been identified in 235U for the first time in an odd-A uranium isotope.16 This observation was only possible by using the new isomer spectrometer NEPTUNE built-up at the EC-JRC IRMM, which allows a flexible adjustment of a low-frequency beam pulsing device.23The measured half life Tl12 = (3.6 f 1.8) ms turns out to be higher than what may be expected from systematic trends. Taking all existing barrier parameter data found in literature, the penetrability of the outer barrier seems to be lower than the trend, which may be interpreted as the existence of a third barrier in this isotope similar to the neighboring even-A isotopes 2343236U.8914 The next step, i. e. the consequent search for a y-decay mode or the direct determination of EII will allow proving the above interpretation. 4. Conclusion
In this paper we have shown that the detailed investigation of shape isomer properties, i. e. decay half life, branching ratio of the possible decay modes and ground state energy at super-deformation, may provide direct
240
and, therefore, precise information about the structure of t h e fission barrier. Some first experiments have been presented and the promising results discussed. In a subsequent step model calculations can be tested against the experimentally obtained barrier parameters and may provide key information for a better understanding of the fission process. Still, the complementary investigation and modeling of the fission crosssection in t h e energy region of sub-threshold fission resonances and around fission threshold, and possibly the measurement of fission-fragment distributions at excitation energies well below the height of the outer barrier, will be indispensable in order t o prove t h e predicted multi-modality in the nuclear energy landscape from ground state t o scission.
References 1. 2. 3. 4. 5. 6.
7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
23.
U. Brosa, S. Grossmann, A. Mller, Phys. Rep. 197 (1990) 167 S. Oberstedt, F.-J. Hambsch, F. Vivks, Nucl. Phys. A644 (1998) 289 A. Turkevich and J . B. Niday, Phys. Rev. 129 (1963) 2239 E. Migneco and J. P. Theobald, Nucl. Phys. A112, 603 (1968) A . Michaudon, Symp. On Nuclear Structure, Dubna (USSR), 1968, 483 Paya et al., Proc. Conf. on Nuclear Data, Microscopic Cross Sections and other Data Basic to Reactors, Paris, 1966 (IAEA, Vienna, 1967), Vol. 11, p. 128 J . Blons et al., Phys. Rev. Lett. 41, 1282 (1978) P. G. Thirolf and D. Habs, Prog. Part. And Nucl. Phys. 49 (2002) 325 Polikanov et al, Zh. Eksp. Teor. Fiz. 42, 1016 (1962) E. Jacobs and U. Kneissel, The Nuclear Fission Process (Ed. C. Wagemans), ISBN 0-8493-5434-X (1991) 103, and Refs. therein J. D. T. Arruda-Net0 et al., Phys. Lett. B248 (1990) 34 J . Drexler et al., Nucl. Phys. A437 (1985) 253 S. F. Mughabghab, Atlas of Neutron Resonances, ISBN-13 : 978-0-444-520357 (2006) 95ff M. Csatl6s et al., Phys. Lett. B615 (2005) 175 S. Bjmnholm and J. E. Lynn, Rev. Mod. Phys. 52 (1980) 725 A. Oberstedt et al., Phys. Rev. Lett. 99, 042502 (2007) V. Metag, in Ref. [15], p. 784 ff H. Weigmann and J. P. Theobald, Nucl. Phys. A187, 305 (1971) R. Zhongzhou and Chang Xu, Nucl. Phys. A759, 64 (2005) S. Oberstedt, F. Gunsing, Nucl. Phys. A589 (1995) 435 S. Oberstedt, F. Gunsing, Nucl. Phys. A636 (1998) 129 S. Oberstedt et al., 3'* Int. Workshop on Nuclear Fission and Fission-Product Spectroscopy, Cadarache, France, 11 - 14 May 2005 (eds. H. Goutte, H. Faust, G. Fioni, D. Goutte), AIP Conf. Proc. 798, ISBN 0-7354-0288-4 (2005) 273 S. Oberstedt et al., Expl. Research at IRMM 2004, Final Report No. GE/SCIRMM/ER/2005, 2005, D1 (to be published)
FISSION IN SPALLATION REACTIONS J. CUGNON' AGO Department, University of LiBge, allbe du 6 AoCt 17, bit. B5, B-4000 LiBge 1, Belgium *E-mail: cugnon9plasma.theo.phys.ulg.ac.be Th. AOUST
SCK-CEN, Boeretang 200, B-2400 Mol, Belgium A . BOUDARD DA PNIA/SPhN, CEA-Saclay, F-91191 Gif-sur- Yvette Cedex, France Some properties of fission in spallation reactions in the GeV range are examined. It is shown on theoretical grounds that the charge, mass and excitation energy are strongly fluctuating. The range of accessible excitation energies is determined. The ability of a particular intranuclear plus evaporation model, namely the INCL4SABLA model to describe the existing data is demonstrated. In view of the numerous parameters used in the fission model, the sensitivity of the results to these parameters is investigated. It is shown that, due to the complexity of the fission modeling, it is hard to get reliable information on the level density parameters at high excitation energy. Finally the influence of the nature of the incident projectile is shortly discussed.
Keywords: Spallation reactions; Fission at high excitation energy; Intranuclear cascade; Evaporation-fisssion model.
1. Introduction There has been recently a revived interest in spallation reactions triggered in particular by the advent of projects for accelerated driven systems for transmutation of nuclear waste. In such devices, a spallation target is located at the center of a subcritical core and is bombarded by a high energy (typically 1 GeV) proton beam. Rapid neutrons emitted by the spallation source are propagating in the core and transform long-lived radioisotopes in stable or short-lived ones through nuclear reactions. In the last years an important program of theoretical and experimental studies [l]has pro241
242
duced a lot of good quality data concerning spallation reactions, which are expected to be utilized by engineers to design and optimize spallation sources. In particular inverse kinematics experiments conducted by the GSI group has allowed to have comprehensive and precise measurements of the residue production cross-sections [2]. This includes so-called fragmentation (or evaporation) residues, but also fission products. As we discuss below, an investigation of the fission process has so been realized for excitation energy of the order of 100 to a few hundred MeV. Previous data are rather scarce and often limited to fission probability. See Refs. [3-61 for recent reviews. In this paper, we want to make a short survey of the information which can be learned from this new set of data. We will first try to identify the conditions for fission occuring in spallation reactions induced by protons in the 200 MeV-1 GeV range. We will see that the charge and mass of the fissioning nucleus is not fixed and can fluctuate sizably. This is true for the excitation energy as well and we will identify with model calculations the range of this excitation energy. Despite of these aspects, that complicate the analysis, it turns out that the intranuclear cascade plus evaporationfission model described in Ref. [7] is able to reproduce successfully the bulk of the experimental results. Finally we will discuss the sensitivity of the calculations to the input of the model, especially for the fission model and try to identify the aspects that should be refined at high excitation energy. The paper is divided as follows. In Sec. 2 we give a brief general description of spallation reactions. Sec. 3 is devoted to model calculations showing the particular conditions under which fission is occuring in the course of spallation reactions. A short review of the main experimental data concerning fission is given in Sec. 4. After a description of the generic features of intranuclear cascade (INC) and evaporation-fission models, a short description of the specific model used in this paper, namely the Libge INC model, coupled to the ABLA evaporation-fission model, is given in Sec. 5 . The next Section shows how the standard version of this combined model can describe the existing experimental data. Sec. 7 is devoted to an analysis of the sensitivity of the model to the input data and to the behaviour a t high excitation energy. In Sec. 8, the dependance upon the nature of the projectile is briefly examined. Finally, Sec. 9 contains our conclusion. 2. Spallation reactions
This expression is used to designate both the interaction of a highly energetic particle with a macroscopic target (thick target experiments) and the one with a nucleus (corresponding to thin target experiments). Obviously,
243
the (nuclear) interaction in a macroscopic target starts with the interaction of the high energy particle with a nucleus. Generally fast nucleons with a energy smaller than, but comparable with, the incident particle energy are emitted. These particles can induce further interactions, which in turn produce other (less) fast particles inducing other interactions, and so on. The interation process with a thick target can be viewed as an iteration of microscopic spallation reactions, with a progressive decrease of the energy of the propagating particles. Charged particles are slowed down in the spallation target by Coulomb interactions with the electrons and are largely stopped in the target. Neutral particles, mainly neutrons, but also neutral pions and gamma’s, escape from the target. For transmutation applications, the important quantity E is the number of escaping neutrons per incident proton divided by the energy of the proton. This quantity is proportional to the ratio of the number of neutrons divided by the “price” to pay for producing them, since the price (in energy) of an incident proton is proportional to the energy E p a t which it is accelerated. It turns out that, for typical spallation targets, the quantity [ displays, as a function of Ep, a maximum around 1 GeV [S]. That is the reason why the above-mentioned studies of spallation reactions were focused on Ep values in the 100 MeV to 1 GeV incident energy range. We will limit ourselves to the same energy range here. It is interesting to mention the main features of microscopic spallation reactions. During this process several neutrons are emitted. Their multiplicity increases with the incident energy and with the target nucleus mass number. To give an idea, on the average, about 15 neutrons are emitted in p +’08 Pb reactions a t 1 GeV. Most of these neutrons are evaporation-like. Only a small fraction ( ~ 1 5 % have ) an energy above 20 MeV, but their spectrum may extend up to the incident energy. Protons, pions and light clusters may also be emitted, but less abundantly. A final residue (two in the case of fission) is ultimately produced. For light targets the residue mass spectrum extends from zero to the original target mass number. For heavy targets, the spectrum, though having strictly the same property, shows two prominent peaks corresponding to fission and evaporation (see below). In the energy range under interest, the spallation reactions are well described by the INC+evaporation-fission model. The latter views the reaction as a two stage process. The first stage corresponds to a sequence of binary nucleon-nucleon collisions leading to the emission of a few fast particles and to a more or less thermalized nucleus, called the remnant. The second stage consists in the de-excitation of this remnant through the
244
evaporation of slow light particles competing possibly with fission. 3. Conditions of fission in spallation experiments
Conditions of fission in spallation reaction experiments are somehow peculiar. First, the charge and mass of the fissioning nucleus are not fixed, since fission can occur in the course of the second stage. It is thus preceeded and followed by sequences of elementary processes, which can fluctuate from one event to the other. The identification of the fissioning nucleus is practically impossible, except if final products are detedted in coincidence, which is not yet the case (see Sec. 9). One has thus to rely on model calculations to have an idea about the identity of the fissioning nucleus and its other properties. We are going to illustrate this point, using the model of Ref. [7], for three typical cases. Table 1. Average mass number, charge and excitation energy of the remnant and of the fissioning nucleus.
238U
< E* > (in MeV)
-~~ ~~
zo8Pb
< A>
< E* > (in MeV) 181Ta
< E* > (in MeV)
Remnant
Remnant/fission
Fissioning nucleus
23 1 91 143
237 92 156
225.6 90.2 81.4
~
~
~
202.7 80.5 136
202 81 217
192.5 80.1 109
180.5 72 128
175 72 241
168 71 120
Table 1 gives the average charge, mass number and excitation energy in the cases of reactions of 1 GeV protons with lS1Tu,'08Pb and 238U. The third column refers to the remnants for all events, whereas the fourth column refers to fission events only (i.e. the events where fission occurs). The last column refers to the fissioning nucleus and the average is made on fission events only, of course. Several observations are in order. First of all, the average excitation energy of the remnant is much smaller than the available energy (1 GeV). It is increasing with the target mass, but the excitation energy per nucleon is decreasing with the mass number of the remnant. One can also see, as announced before, that only a few nucleons are emitted in the cascade stage and they are predominantly neutrons. More remarkably,
245
fission occurs in events where the mass loss a t the end of the cascade is modest and where the excitation energy is substantially larger than for all events. For the latter point, the lighter the target, the larger is the excitation energy. In fission events, the fissioning nucleus is substantially lighter than the remnant, but the charge is not diminished correspondingly. Also, the last column shows that the remnant nucleus has lost considerable excitation energy before fissioning, roughly half of its initial excitation energy, for all cases. On the average, the excitation energy per particle of the fissioning nucleus is rather modest (in view of the available energy): from -0.4 MeV for U to -0.7 MeV for T a . A 0 0
>
CI
0
50 "26
28
30
32
34
36
38
ESTFIS VS I2FlSrlZll5/IAFlS
X
ESTFIS YS ILFIS.IZFIS/IAFlS
X
$00,
CI
0
50 B
26
Fig. 1. Contour plots for the joint distribution of the excitation energy and the fissility parameter 2 = Z 2 / A . The upper panel refers to the remnant and the lower part refers to the fissioning nucleus in p ( 1GeV) zosPb reactions.
+
However, there are considerable fluctuations around the mean values of all the quantities discussed above: charge and mass numbers of the remnant, of the fissioning nucleus and excitation energy. We give an illustration in
246
Fig.1 for the case of z08Pb.One can see that fission can occur a t very low excitation energy and up to very high excitation energy (around 300 MeV), altough the average excitation energy of the fissioning nucleus is 109 MeV. The dispersion of the fissility parameter for the fissioning nucleus indicates that fission can occur a t fissility parameters substantially larger than the one of the target nucleus (z M 32 for this case). Summarizing this section, analysis of fission in spallation reactions is made somehow difficult because the charge and mass of the fissioning nucleus are subject to large fluctuations. This drawback will be alleviated when experiments detecting both fission products and accompanying light particles are performed. The interesting point is that spallation reactions allows to study fission at high excitation energy, owing to the fluctuations. 4. A short review of the experimental data
Fission cross section data are rather scarce and rather imprecise. One often refer to the Prokofiev systematics [6] which is a rough fit of existing data. This systematics shows a fission probability rising to a plateau around 400 MeV incident energy, where it assumes, for Pb, a value of 77%. There exist three extensive measurements of isotope production cross sections. Two of them are based on radioactive decay mesurements and are thus partial [4,5]. The GSI measurements [9,10] use the inverse kinematics and thus measures production cross sections for individual isotopes, be they radioactive or stable. These complete measurements have been done at very few incident energies, whereas the measurements of Refs. [4,5] offer excitation functions. There are practically no measurement of the neutron multiplicity, except a t low incident energy [11,121. The inverse kinematics experiments allowed also some crude measurements of the velocity of the fragments. We give a little more information below when we compare with models. 5. Theoretical models 5.1. The general model
As stated above, the general model for describing spallation reactions is the INC+evaporation model. Of course, for the description of fission events, a competition with fission, and thus a fission model, are required. For the moment, several INC models are used. They are defined by the acronyms of the numerical tools in which they are translated: INCL [7], Isabel [13], CEM2k [14], QMD [15],etc. Evaporation is usually based on the WeiBkopfEwing model [16]or the GEMINI model [17],which treats emission of 2 > 3
241
clusters by the transition-state model [18].Current fission models are in fact parametrizations of the various fission observables, largely based on systematic studies and sometimes on theoretical arguments. The most popular fission models are named: ABLA [19-211, Atchison [22], Dresner [23] and GEM [24,25]. We give some information below on their common features and differences. 5.2. The INC model
In this model, the impinging nucleon initiates a multiple collision process with nucleons in the target. This is realized by simulation. There are two kinds of models. In the first kind, the target is initially taken as a collection of nucleons distributed at random and with momenta taken in a Fermi sphere; then nucleons are set in motion and nucleon-nucleon collisions occur when two nucleons come close enough to each other, i.e. when their minimum relative distance is smaller than the square root of the NN cross section divided by T . In the second kind of INC models, the target is seen as a continuum providing the cascading particles with a path whose mean is the inverse of the product of the NN cross section and the nuclear density. In both kinds of models, collisions can be elastic or inelastic and are subject to Pauli blocking. Nucleons can be emitted if they reach the surface of the nucleus with a sufficient energy and if they pass the test for avoiding reflexion, based on transmission formula for (Schrodinger) waves on a potential step. The simulation process is stopped when the energy of the cascading particles is low enough (the precise criterion may vary from model to model). Below, we give numerical results for the INC model developped in Liege (usually denominated as INCL4a). Therefore we give a little more information on several other features of this model: (a) nucleons move in an average potential well, (ii) relativistic kinematics is used, (iii) isospin symmetry is respected, (iv) nucleon-nucleon collisions can be elastic or inelastic (in the last case a A-resonance is produced which can further decay into a nucleon and a pion), (v) pions can escape or can further interact with a nucleon to form a A-resonance, (vi) A-resonances can scatter elastically on nucleons and on other &resonances, (vii) the model can accomodate nucleons and light clusters (up to 4He) as incident particles. An original feature of the model is that the stopping time, i.e. the time a t which the cascade process this paper we stick with the standard version of INCLI, as described in Ref. [7]
248
is stopped to give place to evaporation, is determined self-consistently, as explained in [7]. Although classical in nature, the model accounts for some quantum aspects: existence of a mean field, Pauli blocking of collisions, quantum transmission through the nuclear surface and stochastic determination of the final states in NN collisions. Finally, we want to stress that the INCL4 model is basically a parameter-free model. It rests on assumptions, of course, but the input are the NN cross sections, which are taken from experiment and the nuclear density, which is taken from electron scattering measurements. The only real parameter is the Fermi momentum, the depth of the potential well being then related to the separation energy. 5 . 3 . The evaporation-fission model
The usual evaporation models rest on the Weii3kopf-Ewing formula
for the emission of a light particle b of kinetic energy E from an excited nucleus A*, leaving an excited nucleus B*. However, they differ by the choice of the parameters. The main ones are the inverse cross sections o C N , the Coulomb barriers (cutting down the cross sections a t low energy) and the level density parameter a entering in the level density p . The latter is basically written under the form
p ( ~ *= ) pexp(2&F),
(2)
where p is some slowly-varying prefactor. Quite sophisticated expressions for a are used, which often include the Gilbert-Cameron parametrization at low energy, correction of the excitation energy for collective (pairing, rotation, etc) effects and the Ignatyuk formulation for the dissapearance of shell effects a t increasing excitation energy [26]. More interesting for us here is the behavior at high excitation energy, which is usually assumed as a M A/<, where is a constant, possibly corrected by some surface term. The fission width is usually given by the transition state method
<
where EZ, is the excitation energy of the system a t the fission barrier and where E is the kinetic energy (in the collective motion) a t the barrier. The crucial parameters are the height of the fission barrier (and its angular momentum dependance) and the level density parameter a f a t the barrier
249
(entering p ~ ) At . high excitation energy, the ratio between the fission and the neutron widths, which basically controls the evaporation-fission competition, is given by
Due to the presence of the exponential, this quantity is quite sensitive to the difference of the level density parameters a t high excitation energy. In evaporation-fission codes, the cascade of emission/fission is governed by the relative probabilities
for decays in different channels, which are reevaluated a t each step. This procedure is justified as long as the (average) time separating two successive emissions, of the order of ~i = h/I'i, is large compared to the time ti required for the emission itself. This time is not well known. For neutron emission, it is of the order of Rlv, where R is the nuclear radius and v the neutron velocity. For fission, this time is much longer. At low excitation energy, ti, both for neutron emission and for fission, are smaller than the T ~ ' s . As excitation increases, T~ decreases substantially and eventually becomes smaller than t j . Even if the fission process starts, emission of neutrons may occur before the fission process ends. The independence of successive emissions breaks down and the net effect is an hindrance of fission. Eventually, when the system is sufficiently cooled down, owing to neutron emission, the condition for independant emissions is restored and fission can occur. This phenomena is often related to viscosity effects in the collective motion. It can be taken into account by simply correcting the fission width [27] by the Kramers formula [28]. It is worth to say that this effect can alternatively taken care of by introducing a time delay to fission in the cascade of successive decays [29]. In order to describe the result of the fission process, a partition model should be supplemented. It corresponds to the three following steps: (1) The neutron number according to the law
N
of one of the fission fragment is generated
where U = E* - B , B being the height of the fission barrier, and where V ( N )is the so-called conditional potential barrier, i.e. roughly the dif-
250
ference of the potential energy between the configuration of the two touching fragments and the intial one, for the partition under consideration.
(2) The charge 2 of the fragment, for a given N, is usually taken from a Gaussian law, with a mean which corresponds t o the same NJZ ratio as for the fissioning nucleus and with a variance which is related to the curvature of the potential energy of the touching-sphere configuration for a fixed N and varying Z. (3) Finally, the quantity E e x = E* - V(N)
+ Ediss,
(7)
i.e. the excitation energy of the touching-sphere configuration increased by the energy dissipated from fission barrier to scission, is spread in excitation energy of the fragments proportionally to their masses, and the quantity
K = E* iQ - Eex
(8)
is spread in kinetic energy of the fragments, proportionally to their inverse masses. In the following, we will use the ABLA model, which accomodates n,p and CY emission only, but which has a sophisticated fission model, including viscosity effects. Furthermore, many ingredients, especially the quantity V ( N ) , are inspired from microscopic calculations with inclusion of shell effects. The Atchison model does include delay to fission. Its parameters are largely phenomenological and have been essentially fitted on low-energy fission data. Let us also mention that the GEMINI model use the transition state model for any partition except for Z 5 3. 6 . Results with the INCL4+ABLA model
We here just want to give a taste of the results of the INCL4+ABLA model of Ref. [7] for fission in spallation reactions. This model predicts a fission cross section of 104 mb for p(500MeV) +208 Pb, to be compared to the experimental value of 132f10 mb. At lGeV, the respective numbers are 165 and . Fig.4 below shows the residue mass spectrum in p ( 1GeV) +208 Pb reactions. The nice agreement obtained for the fission part is of course largely due to the ABLA part of the model, but results also from the predictions of INCL4 for the properties of the remnants. Fig. 2 displays more detail in the
25 1
c
c
I
+
Fig. 2. Isotope production cross sections in p(1GeV) "'Pb reactions. Data from Ref. [9] (dots) are compared with the predictions of INCL4+ABLA calculations (curves).The left panel shows spallation residues whereas the right panel refers to fission residues. Adapted from Ref. [7].
form of isotope production cross sections. One can see that a nice agreement prevails in a large mass and charge range, although the calculations underpredict deep spallation products ( 2 M 65 - 70). Fig. 3 shows excitation functions for the production of some fission isotopes. The good agreement obtained in these calculations indicates that fission is correctly modelized in ABLA on a large domain of energy, although some shortcomings are observed locally. Good agreement is also obtained for recoil energy of the spallation residues (not shown here) , which demonstrates the potentialities of INCL4, evaporation playing a minor role in this respect.
7. Sensitivity to the input data Although INCL4fABLA reaches a good agreement with the description of fission in spallation reactions, one may wonder whether the results are sensitive to the input of the model. This is a sound question, since there are alternative models and since part of the input data (for evaporationfission) are largely based on phenomenology. In this paper, we just want to illustrate a few points, a general sensitivity study being quite involved
252
p
+ "'Bi -,
10-2 10
~
i0-2
103
10'
lo-? l:
7l 0-=
0
101 -
' 103
u
1 2
10'
1 3
10'
1 o=
1 0'
~
1 o-2
1
9
s
~
"Rb
~
lo-'
1 6 '
lo-'
10-2 1 02
10'
E (MeV)
Y
E (M-V)
Fig. 3. Excitation functions for the production of a few fission isotopes in p + ""tBi reactions. Data from Ref. [30] (dots) are compared with the predictions of INCL4+ABLA calculations (curves). Adapted from Ref. [31].
and outside the scope of this paper. Let us first address the question of the sensitivity to the INC input. As we have said, there is no adjustable parameter in INCL4. However, one may give a partial answer to this question by using another INC model with the same ABLA evaporation-fission. This has actually been done in Ref. [31], where the Isabel model has been used as an alternative to INCL4. As far as fission is concerned, there is practically no difference between the two calculations, whereas there are sometimes noticeable differences for other observables. The study of the sensitivity on the angular momentum dependence of the fission barriers have been initiated in Ref. [7]. Fig. 4 gives a typical result. One can see that fission cross sections are well reproduced by
253
0
.25..
50
75
100
A
125
150
175
200
Fig. 4. Residue mass spectrum in P b + p reactions at 1 GeV/nucleon. Data from Ref. [9] (dots) are compared with theoretical predictions (curves). The dotted line corresponds to the pure INCL4+ABLA predictions, whereas the full curve corresponds to the same predictions after the experimental selection is applied (not all isotopes are measured for some isobar. The dashed curve is further obtained when the INCL4 predictions for the angular momentum of the remnant has been replaced by the de Jong systematics (see text for detail). Adapted from Ref. [7].
INCL4SABLA (full curve) after the same cuts as in the experiment are applied. When the INCL4 predictions for the angular momentum of the remnants are replaced by the phenomenological distribution proposed by de Jong [32], the dashed curve is obtained. To give an idea of the sensitivity, let us notice that the average angular momentum in INCL4 lies around 1% wheras it is roughly 10h in the de Jong systematics for this particular case. One can see that the effect is an overall reduction of the fission cross section. This is rather annoying since an overall reduction of the fission probability (by modification of the fission level parameter e.g.) would have basically the same effect. Let us turn to the sensitivity of the level density parameters and let us assume that we are sufficiently close to the conditions of validity of Eq.4 to simplify the discussion. The calculations above have been done with the standard version of ABLA in which the asymptotic value of a (ordinary states) and a j (fission barrier states) are given by
254
A
+
Fig. 5 . Residue mass spectrum in p(1GeV) zosPb reactions. Data from Ref. [9] (dots) are compared with theoretical predictions of INCL4+ABLA with the standard values of the asymptotic level density parameters shown in Eqs.9 and 10 (full curve) and with B s = l (dotted curve).
u = A/13.7
+ 0.095A2I3
(9)
and
a f = A/13.7
+ 0.095B~A'/~,
(10)
respectively. In the last equation, the quantity Bs is a tabulated function decreasing regularly from 1.5 to unity when the fissility parameter z increases from 0 to 50. To give an idea, for fission in reactions on Pb, this yields u f / u M 1.03. In Fig.5, we show the results of a calculation where this ratio is put to unity (this is an option of the ABLA code). One finds a considerable change in the fission cross section although the change in the ratio a f / u is rather small. This result should be confronted with the one of Ref. [31], where a calculation using INCL4 and the Dresner evaporation code which includes the Atchison fission model (altough not the last version) is reported. In this case, the ratio a f / u is close to 1.08 and a reduction of the fission cross section is also observed. There is probably an explanation to this paradox. The level densities in ABLA and Dresner codes
255
differ at lower energy in a way which is not similar to one indicated for the asymptotic regime. This is consistent with the observation made in Ref. [7] that the Dresner model evaporates sizably more than ABLA, resulting in an enhanced production of deep spallation products. This indicates that the different input data influence in an intricate way the evaporation-fission process. We also look at the sensitivity of the results to the viscosity in fission. We make two calculations, one for the recommended value of the parameter controling this viscosity ( p = 1.5 x 1021s-') and for a modified value ( p = 1 x 1021s-l). We observed only minor changes. This result is a t variance with the results of Ref. [33], in which however a rather crude ablation model is used instead of a cascade model.
Fig. 6 . Fission probability as function of the excitation energy. Th e circles are the d at a from Ref. [34] relative t o fission occuring after annihilation of 1.2 GeV on 238Unuclei. The dotted histograms refers to the results of INCL4+ABLA for p(3.08GeV) +236 Th reactions. See text for detail.
256
8. Importance of the incident particle One may wonder whether the fission properties are very dependent upon the nature of the incident particle. We recently made a calculation illustrating this point. In Ref. [34], the fission probability has been measured as a function of the excitation energy in annihilation of 1.2 GeV antiprotons on 238U (the authors claim that a reconstruction of the excitation energy of the remnant is possible from the number of neutrons that they detect). In order to make a meaningful comparison with proton-induced reactions, one has to correct for the annihilation process: for a ( A , Z ) target, the total system is characterized by (A - 1,Z - 1) in the case of antiprotons and by ( A + l , Z+1) in the case of protons. Therefore, the ij+238U system should be compared with the p 236Th system. Furthermore, the annihilation brings roughly 2 GeV extra available energy and transfers less angular momentum than a proton (roughly half, see Ref. [35]).Therefore we compare the results of a INCL4+ABLA calculation for 3.08 GeV protons on 236Th,including a one-half reduction of the remnant angular momentum, with the data of Ref. [34]. This is illustrated by Fig.6. One can see that the agreement is remarkably good, indicating that the nature of the incident particle does not play an important role. What really matters is the available energy, which determines the excitation energy of the remnant, more or less independently of the projectile. This observation was already made in Refs. [35,36].
+
9. Conclusion
We have tried to deem the interest of fission studies in spallation reactions in order to have more insight into fission at special conditions. Of course fission in spallation present an intrinsic interest since it corresponds to the formation of special isotopes in addition to the usual ones coming from the evaporation of the remnant. We have tried to exhibit the conditions on which fission occur in this kind of reactions. We have shown that there are large fluctuations of the charge and mass numbers of the fissioning nucleus as well as of its excitation energy. If this complicates the study, it also allows the study of fission at high excitation energy, typically up to a few hundreds of MeV for 1 GeV incident protons. It should be stressed however that future experiments plan to detect event by event fission fragments, light charged particles and slow neutrons [37]. The charge and possibly the mass and the excitation energy of the fissioning nucleus will so be reconstructed, alleviating the drawback just mentioned above.
257
We have shortly described the experimental situation and made a review of the theoretical tools. We have shown that the combination of the INCL4 cascade model and the ABLA model for evaporation-fission stage gives accurate description of representative experimental data. We have indicated that the results for fission are not very sensitive to the choice of the INC model, but in view of the partially phenomenological input of the ABLA model, we initiated a short investigation of the sensitivity of the theoretical results on the input data. We have found that the results are very sensitive to the asymptotic values (at high excitation energy) of the level density parameters, more precisely of the ratio between the fission and neutron level density parameters. The hope to determine this ratio is however mitigated by the observation that results are also sensitive to angular dependence of the fission barriers, which is badly known. On the other hand, we did not find a large dependence upon the viscsosity parameter as used in ABLA. Finally we have given indications showing that the fission properties do not depend really on the nature of the incident particle, for the same excitation energy.
Acknowledgments This work has been partially done in the frame of the EU I P EUROTRANS project (European Union Contract No FI6W-CT-2004-516520). We acknowledge the EU financial support.
References 1. J.-P. Meulders et al, HINDAS Detailed Final Report, EU Report, 2005. 2. A. KeliC et al, in Proceedings of the Workshop on Nuclear Data for the Transmutation of Nuclear Waste, ed. by A . Kelic and K.-H. Schmidt, GSI publications, ISBN 3-00-012276-1,2004. 3. Proceedings of the International Conference on Nuclear Data for Science and Technology, Nice, April 2007, ed. by R. Jacqmin et al., t o be published. 4. M. Gloris et al, Nucl. Instrum. Meth. Phys. Res. A 643 593 (2001). 5. Yu. Titarenko et al, Nucl. Instrum. Meth. Phys. Res. A 414 73 (1998). 6. A. V. Prokofiev et al, Nucl. Instrum. Meth. Phys. Res. B 463 577 (2001). 7. A. Boudard, J. Cugnon, S. Leray and C. Volant, Phys. Rev. G 66,044615 (2002). 8. R. G. Vassil’kov and V. Yurevich, in Proc. of ICANS-11, KEK Report 90-25, Tsukuba, Japan, p.340 (1991). 9. T . Enqvist et al, Nucl. Phys. A 686,481 (2001). 10. P. Armbruster et al, Phys. Rev. Lett. 93,212701 (2004). 11. T . Ethvignot et al, Phys. Lett. B 575, 221 (2003). 12. T . Ethvignot et al, Phys. Rev. Lett. 94,052701 (2005).
258
13. Y . Yariv and Z. F'raenkel, Phys. Rev. C 20,2227 (1979). 14. S. Mashnik, AIP Conf. Proceedings 768,1188 (2005). 15. J. Aichelin, Phys. Rep. 202, 233 (1991). 16. V. F. WeiDkopf and D. H. Ewing, Phys. Rev. 57,472 (1940). 17. R. J. Charity et al, Nucl. Phys. A 483,371 (1988). 18. N. Bohr and J. A. Wheeler Phys. Rev. 56 426 (1939). 19. J.-J. Gaimard and K.-H. Schmidt, Nucl. Phys. A 531,709 (1991). 20. A. R. Junghans et al, Nucl. Phys. A 629,635 (1998). 21. J . Benlliure, A. Grewe, M. de Jong, K.-H. Schmidt and S. Zhdanov , Nucl. Phys. A 628,458 (1998). 22. T. Atchison, in Proceedings of the Meeting on Targets for Neutron Beam Spallation Sources, KFA-Julich, Julich Conf. 1980, p.17 (1980). 23. L. Dresner, Oak Ridge Report, ORNL-TM-196, 1962. 24. S. Furihata, Nucl. Instrum. Meth. Phys. Res. B 171,251 (2000). 25. S. Furihata and T . Nakamura, J . Nucl. Sci. Technol. Suppl. 2,758 (2002). 26. A. V. Ignatyuk et al, Sou. J . Nucl. Phys. 21,255 (1975). 27. B. Jurado , K.-H. Schmidt and J. Benlliure, Phys.Lett.B 553,186 (2003). 28. H. A. Kramers, Physica VII 4,284 (1940). 29. B. Jurado, C. Schmitt, K.-H. Schmidt, J . Benlliure, T. Enqvist, A. R. Junghans, A. Kelib and F. Rejmund, Phys. Rev. Lett. 93,072501 (2004). 30. R. Michel et al, cited in Ref. [31], t o be published. 31. J.-C. David, A. Boudard, B. FernAndez-Dominguez, S. Leray and C. Volant, in Proceedings of the Conference on Nuclear Fission, Obninsk, (2003). 32. M. de Jong, A. V. Ignatyuk and K.-H. Schmidt, Nucl. Phys. A 613, 435 (1997). 33. J. Benlliure et al., Nucl. Phys. A 700,469 (2002). 34. S. Schmid et al, 2.Physik A 359,27 (1997). 35. J. Cugnon, Phys. Atomic Nuclei 57, 1705 (1994). 36. J. Galin and U. Jahnke, J. Phys. G 20, 1105 (1994). 37. A . Boudard and J. Benlliure (spokesmen), Proposal S293 at SPALLADIN, GSI: Proton-induced fission in the GeV domain, 2007.
This page intentionally left blank
LIST OF PARTICIPANTS
ANDREYEV Andrei Katholieke Universiteit Leuven Instituut voor Kern- en Stralingsfysica Celestijnenlaan 200D B-3001 Leuven Belgium andrei.andrevev@ fvs.kuleuven.ac.be
D'HONDT Pierre SCK-CEN Boeretang 200 B-2400 Mol Belgium pierre.dhondt @ sckcen.be
CUGNON Joseph Universitt de Libge Dept. AGO Allte du 6 AoQt 17, bbt. B5 B-4000 Libge 1 Belgium cumon @plasma.theo.phvs.ulg.ac.be
DUBRAY Noel CEA Bruybres-le-Chkel DPTNService de Physique Nucltaire B.P. 12 F-9 1680 Bruykes-le-Chbtel France [email protected]
DE FRENNE Denis Universiteit Gent Vakgroep Subatomaire en Stralingsfysica Proeftuinstraat 86 B-9000 Gent Belgium [email protected]
GOENNENWEIN Friedrich Universitat Tubingen Auf der Morgenstelle 14 D-72076 Tubingen Germany [email protected]
DELAROCHE Jean-Paul CEA Bruybres-le-Chbtel DPTNService de Physique Nucltaire B.P. 12 F-9 1680 Bruybres-le-Chbtel France [email protected]
GOUTTE HBloise CEA Bruybres-le-Chbtel DPTNService de Physique Nucltaire B.P. 12 F-9 1680 Bruybres-le-Chbtel France [email protected] 261
262
HAMBSCH Franz-Josef EC-JRC-IRMM Retieseweg 111 B-2440 Gee1 Belgium [email protected]
JACOBS Etienne Universiteit Gent Vakgroep Subatomaire en Stralingsfysica Proeftuinstraat 86 B-9000 Gent Belgium [email protected]
HANAPPE Francis UniversitC Libre de Bruxelles CP229, av. F.D. Roosevelt 50 B-1050 Brussel Belgium fhanappe @ulb.ac.be
JURADO Beatriz CEN Bordeaux-Gradignan Chemin du Solarium B.P. 120 F-33175 Gradignan France jurado @cenba.in2~3 .€r
HUYSE Mark Katholieke Universiteit Leuven Instituut voor Kern- en Stralingsfysica Celestijnenlaan 200D B-3001 Leuven Belgium Marc .Huvse@ fvs.kuleuven.ac.be
KORNEEV Ivan Institute for Theoretical and Experimental Astrophysics Dept. of Astrophysics B. Cheremushkinskaya 25 R- 117259 Moscow Russia Ivan [email protected]
ITKIS Mikhail Joint Institute for Nuclear Research R-141980 Dubna Moscow region Russia [email protected]
LETOURNEAU Alain CEA Saclay F-91191 Gif-sur-Yvette France [email protected]
263
MATERNA Thomas Institut Laue-Langevin 6, rue Jules Horowitz B.P. 156 F-38042 Grenoble France thomas.materna (3ill.fr
PEREZ-MARTINSara CEA Bruybres-le-Ch2tel DPTNService de Physique NuclCaire B.P. 12 F-9 1680 Bruybres-le-Chbtel France saraaerez-martin (3cea.fr
MUTTERER Manfred Technische Hochschule Schlossgartenstrasse 9 D-64289 Darmstadt Germany [email protected]
PRIEELS RenC UniversitE Catholique de Louvain 2, Chernin du Cyclotron B- 1348 Louvain-la-Neuve Belgium [email protected]
OBERSTEDT Andreas Orebro University Dept. of Natural Sciences S-70182 Orebro Sweden andreas.oberstedt @nat.oru.se
REJMUND Fanny GANIL Boulevard H. Becquerel F- 14076 Caen France [email protected]
OBERSTEDT Stephan EC-JRC-IRMM Retieseweg 111 B-2440 Gee1 Belgium [email protected]
SILLANPAA Mikko University of Jyvaskyla Letkutie 2 A14 FIN-40700 Jyvaskyla Finland miolsill @cc.ivu.fi
264
TAIEB Julien CEA Bruybres-le-Chbtel DPTNService de Physique NuclCaire B.P. 12 F-9 1680 Bruybres-le-Chbtel France julien.taieb @?cea.fr
WAGEMANS Cyrillus Universiteit Gent Vakgroep Subatomaire en Stralingsfysica Proeftuinstraat 86 B-9000 Gent Belgium [email protected]
TALOU Patrick Los Alamos National Laboratory Theoretical Division T- 16, Nuclear Physics Group Los Alamos, NM 87545 USA talou @ lanl.gov
WAGEMANS Jan SCKCEN Boeretang 200 B-2400 Mol Belgium jan. [email protected]
VERMOTE Sofie Universiteit Gent Vakgroep Subatomaire en Stralingsfysica Proeftuinstraat 86 B-9000 Gent Belgium [email protected]
AUTHOR INDEX Ahmad, I., 47 Aiche, M., 47 Ait Abderrahim, H., 207 Almahamid, I., 63, 107 Aoust, Th., 241 Audouin, L., 47
Dobrowolski, A., 161 Dorvaux, O., 3,25 Dubray, N., 171 Dupont, E., 63,79
Bail, A., 79 Barreau, G., 47 Bauge, E., 47, 147 Beghini, S., 25 Behera, B. R., 25 Berger, J.-F., 161 Berthoumieux, E., 47 Bidaud, A., 47 Bogachev, A. A., 25 Boudard, A,, 241 Bringer, O., 63 Bunakov, V., 3
Fabry, I., 125 Faust, H. R., 79 Fioni, G., 63 Fioretto, E., 25 Floyd, J., 107
El Masri, Y., 55
Gadea, A,, 25 Gagarski, A., 3 Gawrys, M., 233 Geltenbort, P., 107 Girod, M., 161 Gonnenwein, F., 3 Goutte, H., 161,171 Greene, J. P., 47 Gunsing, F., 47 Guseva, I., 3
Caamano, M., 71 Capellan, N., 47 Chabod, S., 63 Chartier, F., 63 Chemysheva, E. V., 25 Corradi, L., 25 Cugnon, J., 241 Cullen, D., 189 Czajkowski, S . , 47
Haas, B., 47 Hambsch, F.-J., 125, 233 Hanappe, F., 3,25 Hilaire, S., 147 Isaev, S., 55 Itkis, I. M., 25 Itkis, M. G., 25
D’hondt, P., 207 Dare, J. A., 189 DassiC, D., 47 De Bruyn, D. 207 Delaroche, J.-P., 161, 171 Demetriou P., 55
Janssens, R. V. F., 47 Jurado, B., 47 265
266
Kadmensky, S., 3 Kessedjian, G., 47 Keutgen, Th., 55 Khlebnikov, S., 3, 197 Khlebnikov, S. V., 89 Kinnard, V., 3 Knyazheva, G. N., 25 Kondratiev, N. A., 25 Kopatch, Yu., 3 Kopatch, Yu. N., 89, 197 Korneev, I., 177 Kornilov, N., 125,233 Koster, U., 79, 189 Kozulin, E. M., 25 Krupa, L., 25 Latina, A., 25 Letourneau, A., 63, 79 Libert, J., 161 Litaize, O., 79 Lavestam, G., 233 Lukens, W., 107 Lumley, N., 189 Lyapin, V. G., 89 Marie, F., 63 Materna, T., 79, 189 Mathieu, L., 47 Montagnoli, G., 25 Mutterer, M., 3, 89, 197 Mutti, P., 63 Nesvizhevsky, V., 3 Oberstedt, A., 99, 233 Oberstedt, S., 99, 125, 233 Oriol, L., 63 Osmanov, B., 47 Panebianco, S., 63 Panov, I. V., 177 Perez-Martin, S., 147 Petrov, G., 3
Prieels, R., 55 Prokhorova, E., 3 Rauscher, T., 177 Rejmund, F., 71 Rochman, D., 99 Rowley, N., 25 Rubchenya, V., 3 Rubchenya, V. A,, 25 Scarlassara, F., 25 Serot, O., 47, 79, 107, 117 Sillanpaa, M., 3, 89, 197 Simpson, G., 3 Simpson, G. S., 189 Smirnov, S . , 197 Smith, A. G., 189 Sokolov, V., 3 Soldner, T., 3, 107 Stefanini, A. M., 25 StuttgC, L., 3,25 Szilner, S., 25 Talou, P., 139 Tassan-Got, L., 47 Theisen, Ch., 47 Thielemann, F.-K., 177 Tiourine, G., 3 Trotta, M., 25 Trzaska, W., 3 Trzaska, W. H., 25, 89, 197 Tsekhanovich, I., 3, 189 Tyurin, G., 197 Tyurin, G. P., 89 Van Mol, J., 55 Varley, B., 189 Vermote, S., 107, 117 Veyssiere, Ch., 63 von Kalben, J., 89, 197 Von Kalben, J., 3 Vorobyev, A., 125
267
Wagemans, C., 3, 17, 107 Wagemans, J., 223 Wilson, J. N., 47 Wollersheim, H.-J., 3
Yamaledtinov, S . R., 89 Zavarukhina, T., 3 Zimmer, O., 3
This page intentionally left blank