On the Equation $\operatorname({diy}(\nabla ul*{p-2}\nabla u) + \lambdalul Mp-2}u=0$ Peter Lindqvist Proceedings of the American Mathematical Society, Volume 109, Issue 1 (May, 1990), 157-164, Stable URL: hhtp//links,jstor.org/sic¥?sici=0002-9939% 28 199005%29109%3A 1%3C157%3AOTE%3E2.0,CO%3B2-2 ‘Your use of the ISTOR archive indicates your acceptance of JSTOR’s Terms and Conditions of Use, available at hhup:/www.jstororg/about/terms.html. JSTOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission, Proceedings of the American Mathematical Society is published by American Mathematical Society. Please contact the publisher for further permissions regarding the use of this work. Publisher contact information may be ob\ained at hutp:/swww tor org/joumalsams, hum Proceedings of the American Mathematical Society ©1990 American Mathematical Society ISTOR and the ISTOR logo are trademarks of ISTOR, and are Registered in the U.S. Patent and Trademark Office. For more information on JSTOR contact jstor-info@umich.edu, ©2002 JSTOR bupslwww jstor.org/ ‘Tue May 21 15:30:08 2002ON THE EQUATION div(|Vul?? Vu) + alul Pw PETER LINDQVIST (Communicated by Barbara L. Keyhitz) ApsTRACT. The firs eigenvalue A= 4, forthe equation divi|ul?"*0u) + Alul?2u = 0 is simple in any bounded domain. (Through the nonlinear coun- terpart to the Rayleigh quotient 4, is related to the Poincaré inequality.) 1. IntRopuction The first eigenvalue of the operator div(|Vul’~"Vu) is here defined as the least real number 4 for which the equation (uy div(\Vuf?? Vu) + alu” hhas a nontrivial solution w with zero boundary data in a given bounded domain in the n-dimensional Euclidean space. The first eigenvalue is the minimum of the Rayleigh quotient: Iu? dx 12) a= min Lu dx (1.2) me Tae ax Here 1 < p < oo, and in the linear case p = 2 one obtains “the principal frequency”; the associated first eigenfunction uw describes the shape of a mem- brane when it vibrates emitting its gravest tone, cf. [6]. (We shall often use the term principal frequency for the nonlinear cases as well.) 1.3. Theorem. The first eigenvalue is simple in any bounded domain Q in R" i.e, all the associated first eigenfunctions u are merely constant multiples of each other. In the classical linear case, ie., when p = 2, this is a well-known result [4, Theorem 8.38, p. 214]. This phenomenon is a reflection of the distinguish ing feature of the first eigenfunctions: they never change signs in 2. Higher eigenvalues are not simple even in the case Au+Au=0, cf. (6, §§5.9B and 7.8] Received by the editors February 16, 1989 and in revised form. July 4, 1989, 1980 Mathematis Subject Clastcation (1985 Revision). Primary 38370, 35360, 35P30, 49422. ©1999 anc nema ey1s DETER LINDQVIST ‘Various versions of Theorem 1.3 are known for general exponents but, so far as we know, only for a very restricted class of domains. Our contribution is, 10 prove the same result for any bounded domain. F, de Thélin has observed that there exists essentially only one first eigenfunc- tion among the radial functions in a ball [11]. An immediate consequence of the interesting work [8] of Sakaguchi is that the first eigenvalue is simple in convex domains.’ Theorem 1.3 has been extended to domains of Hélder class C?-" by Anane [1], We have been informed that Bhattacharya has further extended this to C?-domains. This simplicity of the first eigenvalue in C*-domains has ‘been proved in a slightly more general setting by Veron and Guedda, ef. [14]. Unfortunately, these proofs rely heavily on deep global regularity results for the gradients of the first eigenfunctions (3), [12]. As this kind of global estimate cannot hold in arbitrary domains, such an approach will artificially restrict the class of domains. The objective of this note is to present a natural proof for Theorem 1.3. Our ‘method is direct, and no advanced tools are needed: In fact, the proof is based on a few immediate observations and a refined choice of the test-functions in tu ‘The analysis can be readily extended to homogeneous eigenvalue problems of the form div A(x, Vu(x)) + Apx)lucx))P u(x p— 0 where A(x, w) © |wl?"?w fulfils certain restrictive structural conditions and p(x) 2 0. The proof is a simple adaptation of the arguments presented here. Finally, we want to mention that to any bounded domain Q there is a con- stant c(@) such that the Poincaré inequality (4) f ul’ dx < (a) [ oul? dx is valid for every w € C5°(Q) or, more generally, for every w € W,°?(Q). The best constant (Q) is the reciprocal of the first eigenvalue in Throughout this paper, © denotes an arbitrary bounded domain in R" 2. THE RAYLEIGH QUOTIENT Among the equations div(|Vul?7Vu) + aul 1 0, cf. [12] and [3].) The least eigenvalue, say 4, = 2,(2), is obtained as the minimum of the Rayleigh quotients, i.e., ing Lalu? dx M Jle dx the infimum being taken among all v € W/)"(), v #0. Alternatively, one can further restrict the admissible functions to those in C3°(Q).. It is easily seen that this minimization problem is equivalent to equation (2.2) with 2= 2, We say that 2, is she frst eigenvalue or the principal frequency and the cor- responding eigenfunction is called the first eigenfunction. In a bounded domain, 1 the existence of first eigenfunction (and of the principal frequency 2, > 0) is established via a minimizing sequence ,, u;,... for the Rayleigh quotient By homogeneity the normalization fiimtas 1 (k=1,2,.2.) ln is possible. In this standard procedure one usually uses the Rellich-Kondrachov imbedding theorem [9, §11, pp. 82-85] and the Radon-Riesz theorem [7, §37, p. 71] related to the uniform convexity of L’(Q). See [10]. If w minimizes (2.3), so does |u|. Thus the existence of a first eigenfunction not changing signs in Q is prov (23) 2.4, Lemma, In any bounded domain there is a first eigenfunction u, > 0 corresponding to the principal frequency 2, > 0 Furthermore, u, > 0 if u, > 0. This is a finer point following from the Harnack inequality [13, Theorem 1.1, p. 724]. 3. THE FIRST EIGENFUNCTIONS As before, @ denotes an arbitrary bounded domain. The crucial part of the proof for Theorem 1.3 is to establish that positive eigenfunctions are essentially unique. The general case can be reduced to this situation. To this end, note that if uw is a first eigenfunction, so is |u|. By Harnack’s inequality (13, Theorem 1.1, p. 724] either |u| > 0 in the whole domain or |u| = 0, the latter case being, excluded for eigenfunctions. By continuity, either « or ~w is positive in the whole domain, Hence Theorem 1.3 follows from the following lemma.160 PETER LINDQVIST 3.1. Lemma. Suppose that u > 0 and v > 0 are eigenfunctions both corre- sponding to A,. Then w and v are proportional, Proof. As Anane has observed in (1), the result would follow by certain balanced calculations, if the function = u—v°u'~? were, a priori, admissible as test- function in (3.2) fru yu. Und mM jut?Pun ax A lo and v—w’v'? in the similar equation for v. However, this seems to require some regularity of the boundary 9Q. For “irregular domains” this is the crucial point Therefore we use the modified test-functions WP WEP gg (wteh tute) (u+ ey (w+ey being a positive parameter. Then 04 ora {iron (2E8)"}oe-a (SE8)" ve, and, by symmetry, the gradient of the test-function in the corresponding equa- tion for v has a similar expression with u and v interchanged. Set (3.3) ” u=ute, Inserting the chosen test-functions into their respective equations and adding these, we obtain the expression (3.5) pr yl wi” ’ af |S - Sa] of eas ~flfie-n( 8) your o{.ea-0 (8) jr = flo GY reat ome5 +0 (te) eer ey eu] foe — of (|W logy, |” — [V logy,|”) dx a = ff motiwtogu! 7064, (Vlog, = Vogu,) dx In = [ontiv og? *7 tev, (Vlogu, — Vlogy,) dx, 2 and here the last member is clearly <0 by inequality (4.1) in the appendix. Iris apparent that rly (3.6) tp [5 Sa] —ras=0 at a8 Sa lw ?ON THE EQUATION div Vup*PH) + Jluf"u= 0 1 Let us first consider the case p > 2. According to inequality (4.3) in the appendix we have o< La A ml (3 +3) pu, u, —u,Vo,P dx < “rl, (eo - @)] (ud =v) dx for every ¢ > 0. (Inequality (4,3) was used with w, = Vlogu,, w, = Vlog, and vice versa.) Recalling (3.6), letting ¢ tend toward zero through any positive sequence ¢,,¢;,..., and using Fatou’s lemma in (3.7), we finally arrive at the conclusion that vVu=wVu ae. in Q. Hence there is a constant « such that u=nv ae. in Q. By continuity u=xv at every point in Q. This proves the case p22 The case 1 < p <2 is very similar. Applying inequality (4.4) in the ap- pendix on (3.5) we obtain (3.7) 0 C(p) fn, oe +p) Nee HT ble Jee ey eras for every ¢ > 0. Using (3.6), we again arrive at the desired dependence u = Kv for some constant. This concludes the proof. (38) Be Remarks. (1°) If v > 0 is any eigenfunction in Q corresponding to the eigen- value 2, then 2 = 2,, ie., only the first eigenfunctions are positive. See [1]. Indeed, if w > 0 denotes a first eigenfunction, then the same procedure as _ Ls ()'(2)7] (ul -vf)dx <0 and arguing as before, we arrive at a-a [oe - This leads to a contradiction, if 4 >A, , since w can be replaced by any of the functions 2u, 30, 4v,.... Thus 2= 4, (2°) The first eigenvalue is isolated. This was proved in [1] for sufficiently regular domains, but those proofs can be carried over to general domains. The necessary modifications are suggested by the constructions in the proof for Lemma 3.1 Pydx <0.162 PETER LINDQVIST 4, APPENDIX: AN INEQUALITY ‘The familiar inequality (4) wal? > wy? + ply? mw, » (w, — w,) for points in R", w, # w), p > 1, is just a restating of the strict convexity of the mapping w — |w/”. It is sometimes convenient to express this strictness in the following quantitative way: 4.2. Lemma. If p > 2, then ’ ’ p- fw, =v) 43) I >I w, (to, ~w,) + MT (43) (wal? > le |? + ply Pay «(wy ~ w,) + SEE {for all points w, and w, in R" If 2 inequality (4.3) follows from Clarkson’s inequality +2 oy mee (46) mf +m 22) 2242) 4] 2 To see this, use (4.1) to get (47) es > |w,l? + folw,|2w Combining the two inequalities we arrive at (48) Il? > furl? + plew, [P20 «(wy ~ w,) +2 This is the desired inequality but with the constant 2'~ in place of 1. Repeating this procedure, starting again with (4.6) but now using (4.8) instead of (4.1), we get the constant improved to 2'~? +4!" By iteration one finally obtains the constant, 1 ra in (4.3). (The best constant is irrelevant for our purpose here.) Paha Behu) + a 1a ON THE EQUATION divi Let us now consider the case 1 < p <2. In order to derive (4.4) we fix w, and w;. Expanding the function £0) according to Maclaurin’s formula jw, + (uw, ~ WP fa F0)+ £0} ['U-of"ode fi we have (4.9) |w,| Iw + aly, (ww) + fd -of"oae provided that |w, + r(w, ~ w,)| #0, when 0. <1 <1, But the case when ‘w, + t(w, —w,) = 0 for some 1, 0 < ¢ < 1, is easily checked. By direct calculation SL" (0) = (p= 2)lw, + t(w, = w,)/*{(w, + H(w, —'w,)) «(wy = wy) + plw, + (w, — w,)/" Iw, — WP, ‘and the Schwarz inequality yields (4.10) S(O 2 Pp = lw, + (w, = 4)" |w, = wy)". Returning to (4.9), we have [a —of"ndt> af roa and, since |u| -+ wal > |w, +¢(w,~w,)| we finally arrive at (4.4) with C(p 3p(p ~ 1)4-*. Here (4.10) was used. REFERENCES 1. A. Anane, Simplicté et isolation de la premiere valeur propre du p-laplacien avec poids, €.R-Acad, Sci, Paris Ser. I Math. 308 (1987), 725-728. 2. J, Gareia Azorero and I, Peral Alonso, Evistence and non-unigueness for the p-Laplacian: Nonlinear eigenvalues, Comm. Paral Differential Equations 12 (1987), 1389-1430. 3. E, DiBenedetto, C'** focal regularity of weak solutions of degenerate elliptic equations, Nonlinear Anal. 7 (1983), 827-850. 4. D. Gilbarg and N. Trodinger, Elliptic partial differential equations of second order, 2nd cd, Springer-Verlag, Berlin-Heidelberg, 1983. 5. G. Mostow, Quasi-conformal mappings in n-space and the rgilty of hyperbolic space forms, Inst Hautes Etudes Sc. Publ, Math, 34 (1968), 33-104, 6. G. Polya and G. Szepo,Isoperimetric inequalities in mathematical physics, Princeton Univ, Press, Princeton, NJ, 1951 7. F. Riesz and B. Sz-Naga, Vorlesungen ier Funkuionalanalysis, Deutscher Verlag der Wis- senschatien, Berlin, 1986 8 S. Sakaguchi, Concavity properties of solutions 10 some degenerate quasilinear eliptc Dirick ler problems, Ann, Seuola. Norm. Sup. Pisa Cl. Sei. (4) (10 appa.1 PETER LINDQVIST 9. S. Sobolev, Applications of functional analysis im mathematical physics, American Mathematical Society, Providence, RI, 1963, 10, F. de Thelin, Quelques résultats d'exstence et de non-existence pour une E. D. P.eliptique ‘nom linéaire, C. R. Acad. Sei. Paris. Ser. 1 Math. 299 (1984), 911-914. 11, —, Sur Vespace propre associé a fa premiere valeur propre du pseudolaplacien, C. R. ‘Acad. Sei. Pars Ser 1. Math, 303 (1986), 355-358, 12. P. Tolksdort, Regularity for @ more general clas of quas-linear elliptic equations, 3. Differ= ential Equations 51 (1984), 126-150 13, N, Trudinger, On Harnack type mequatities and ther application to quasilinear elliptic uae tions, Comm, Pure Appl. Math. 20 (1967), 721-747. 14, L, Veron and M. Guedda, Quasiincar elliptic equations involving critical Sobolev exponents, Université Frangois Rabelais (Tours), 1988. (10 appear in Nonlinear Anal) Hersivet UNIVERSITY oF TECHNOLOGY, INSTITUTE oF MATHEMATICS, SF-02150 EsP00, FINLAND,