VOL.- 17, 1931
MA THEMA TICS: M. MORSE
319
THE ORDER OF VANISHING OF THE DETERMINANT OF A CONJUGA TE BASE By MARSTON ...
11 downloads
359 Views
159KB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
VOL.- 17, 1931
MA THEMA TICS: M. MORSE
319
THE ORDER OF VANISHING OF THE DETERMINANT OF A CONJUGA TE BASE By MARSTON MORSE DEPARTMENT OF MATHEMATICS, HARVARD UNIVERSITY
Communicated April 9, 1931
In the present note I shall use the terminology of the paper entitled, "Sturm Separation and Comparison Theorems in n-Space."' In particular see §4 of that paper for the definition of a conjugate family of secondary extremals. I shall prove the following theorem of which an erroneous proof is given in a paper2 entitled, "The Foundations of a Theory of the Calculus of Variations in the Large in m-Space" (First paper) § 7. THEOREM. If D (x) is the determinant of a conjugate base the order of its vanishing at x = 0 equals the nullity s of D(O).
We suppose s > 0. Suppose || yij(x) || is the n-square matrix whose columns are the solutions of the given base. There exist s linearly independent solutions of the family (that is, solutions dependent on the given base) which vanish at x = 0, and no more. Without loss of generality we can suppose the first s columns of || yij(x) || are these vanishing solutions. For one could take a new base for the family for which this would be the case. The last n-s columns of yij(x) will then be of rank n-s at x = 0, for otherwise we could obtain additional independent solutions vanishing at x = 0 dependent on these last n-s columns. From each of the first s columns we can factor but an x and so write
I
D(x) = xSE(x). I say that E(0) ' 0. For E(O) is the determinant obtained from yij(x) | by differentiating the first s columns and then putting x = 0 in all columns, and we shall find that its vanishing is impossible. Using the summation convention of tensor analysis set (i = 1, ... ., n j; = 1 ... ., s) cjyij(x) = ui(x) (i = 1, .. ., n; h = s + 1, ..., n). chyih(x) = Zi(x) If E(O) were zero we could determine constants cl, ..., ct, not all zero, such that (i = 1, ... , n) ui (O) = zi(0). I say that [z(O)] s [O].
320
MA THEMA TICS: G. A. MILLER
PROC. N. A. S.
If [z(O) ] were null it would follow that cs+ 1 . . . c,n would be zero since the rank of the last n-s columns of E(O) is n-s. We would, moreover, have (i = 1,.. n) u' (O) = ui (O) = 0 so that [u(x)] would be identically null contrary to the assumption that the solutions of the base are linearly independent. Thus [z(O)] 5z [0]. Now the solutions ui(x) and zi(x) are conjugate. If we use the fact that [u(O)] = [O] the condition that ui(x) and zi(x) be conjugate solutions takes the form,3 Rik(O)zi(O)u'(0) = Rik(O)Zi(O)Zk(O) = 0. (i k = 1, ..., n) But this is contrary to the assumption that the problem is positively regular. Thus E(O) $ 0 and the order of D(x) at x = 0 equals the nullity s of D(O). 1 "A Generalization of the Sturm Separation and Comparison Theorems in n-Space," Math. Ann., 103, 52-69 (1930). 2 "The Foundations of the Calculus of Variations in the Large in m-Space. (First paper.) Trans. Am. Math. Soc., 31 (1930). Lemma § 7, p. 385. Lemma 2 of p. 387 should read, "Let D (x) be a determinant whose n columns are n conjugate, etc." The error in the proof in § 7 was first called to my attention by G. A. Bliss. The proof given in § 7 can be modified so as to be correct but the present proof seems simpler. 3 See Bolza, Variationsrechnung, p. 626.
A UTOMORPHISMS OF ORDER 2 OF AN 'ABELIAN GROUP By G. A. MILLER DEPARTMENT OF MATHEMATICS, UNIVERSITY OF ILLINOIS
Communicated March 31, 1931
1. INVARIANT AUTOMORPHISMS OF ORDER 2. It is known that an invariant operator of the group of isomorphisms of an abelian group G must transform every operator of G into the same power of itself. Hence it results that when the order of G is of the form pm, where p is an odd prime number, its group of isomorphisms involves one and only one invariant operator of order 2. When p = 2 this group of isomorphisms contains no invariant operator of order 2, one invariant operator of this order, or three such operators as G involves no operator of order 4, operators of order 4 but none of order 8, or operators of order 8. If we combine these facts with the known theorem that the group of isomorphisms of every abelian