ON THE ASYMPTOTIC BEHAVIOR OF THE PRINCIPAL EIGENVALUES OF SOME ELLIPTIC PROBLEMS

Size: px
Start display at page:

Download "ON THE ASYMPTOTIC BEHAVIOR OF THE PRINCIPAL EIGENVALUES OF SOME ELLIPTIC PROBLEMS"

Transcription

1 ON THE ASYMPTOTIC BEHAVIOR OF THE PRINCIPAL EIGENVALUES OF SOME ELLIPTIC PROBLEMS TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA Abstract. This paper is concerned with nonselfadjoint elliptic problems involving inde nite weights and boundary conditions of the Dirichlet, Neumann or Robin type. We study the asymptotic behavior of the principal eigenvalues when the rst order term (drift term) becomes larger and larger. The basic results of [1] are extended to the present context. Moreover, answers are provided to some open problems raised in [1]. 1. Introduction This paper is mainly concerned with the study of the asymptotic behavior of the principal eigenvalues of the problem (1:1) L u := div (A (x) ru) + ha (x) ; rui + a 0 (x) u = m (x) u in ; u = 0 when the real parameter goes to in nity. Here is a bounded domain in R N ; A (x) is an uniformly elliptic matrix, a (x) is a divergence free vector eld, and m (x) is a (possibly inde nite) weight satisfying m + 6 0: Several works have been devoted to this type of question and we refer in particular to [1] for a recent survey as well as for interesting new results and applications. These applications concern for instance the behavior in nite time of the solution of a linear parabolic Cauchy problem with large advection, or the estimation of the speed of pulsating fronts in some nonlinear reaction di usion equation. See also [3] for another recent contribution with application to a competition model. The main result of [1], as far as the eigenvalue problem (1:1) is concerned, is the following. Assume the weight m (x) de nite, i.e. m (x) " > 0 in ; and let be the principal eigenvalue of (1:1). Then the limit of as! 1 always exists in R[f+1g, and this limit is nite if and only if the vector eld a (x) admits a rst integral w in H 1 0 () (i.e. w 2 H 1 0 () with w 6 0 and ha; rwi = 0 in ); moreover the limit of a ; when nite, is equal to the minimum of the Raleigh quotient of the associated selfadjoint problem (1:1) 0 over all such w 0 s; in addition this minimum for all : It is our purpose in this paper to extend this result of [1] to the case of an inde nite weight m (x) : In the process various complements will be obtained, which in particular answer three open problems raised in [1]. Our approach di ers from that in [1]; it is based on a minimax formula for the principal eigenvalue of (1:1). Date: December 3, Mathematics Subject Classi cation. 35J20, 35P15. Key words and phrases. Nonselfadjoint elliptic problems, principal eigenvalue, inde nite weight, minimax formula, asymptotic behavior, large drift. This work was carried out with the support of FNRS and SECyT. 1

2 2 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA The rst di culty one faces when dealing with the inde nite case is that a principal eigenvalue may fail to exist (cf. e.g. [14], [4], [6], [9]). Denoting by the largest principal eigenvalue of (1:1) (provided a principal eigenvalue exists), we will show that if exists for one 0; then exists for all (cf. Proposition 3.2). We will also give examples showing that may fail to exist for all (cf. Example 3.7), or may exist for some s but not for others (cf. Examples 3.4 and 3.10). In the case where a (x) is a nontrivial gradient eld, we will see that always exists for large, with moreover! +1 as! +1 (cf. Proposition 3.6). Our main result concerning the function! in the general case, beside its continuity (cf. Proposition 4.1) and evenness (cf. Proposition 3.2), is that it is either constant on R or strictly increasing on [0; 1) (cf. Proposition 4.3). Note that the mere monotonicity of answers a question raised in Section 5 of [1]. The case constant can be characterized in term of the associated selfadjoint problem (1:1) 0 : is constant if and only if 0 exists and its associated eigenfunction is a rst integral (cf. Proposition 4.4). In the case non constant we have the following extension of the result of [1] described above: remains bounded as! +1 if and only if a (x) admits a rst integral w in H0 1 () with R mw2 > 0 (cf. Theorem 5.1). Moreover, when remains bounded with in addition either non constant, or a 0 0; or m 0; then the limit of is equal to the minimum of the Rayleigh quotient of (1:1) 0 over all rst integrals w 2 H0 1 () with R mw2 > 0 (cf. Theorem 5.4). When is constant with a and m changing sign, a rst integral w 2 H0 1 () with R mw2 > 0 may not exist (cf. Example 5.3). Another di culty which occurs in the inde nite case is that when m (x) changes sign, there may exist two principal eigenvalues. Distinguishing between the two is easy when a 0 0 (since then one is > 0 and the other is < 0), but requires more care when a 0 is not 0: We will use for this purpose the -function associated to (1:1), which we denote by L;m or brie y by : () is de ned for 2 R as the (unique) principal eigenvalue of (1:2) L u m (x) u = u in ; u = 0 The principal eigenvalues of (1:1) are thus exactly the zeros of the function : It turns out that the largest principal eigenvalue of (1:1) can be distinguished from the smallest by the property that the derivative of at is 0 (cf. Lemma 3.1). This observation of a technical nature will be used in some of our arguments, and more generally the function will play an important role in many of our proofs. As brie y indicated earlier, our approach is rather di erent from that in [1]. A basic ingredient for us is a minimax formula of variational type for the principal eigenvalues of a problem like to (1:1). Such a formula, for m 1 or m " > 0, goes back to [5] and [12], where it was derived by arguments involving respectively semigroups theory or stochastic di erential equations. The form we use in our present inde nite problems is that obtained recently in [9], whose derivation relies on the study of some degenerate elliptic equations in the context of weighted Sobolev spaces (cf. Proposition 2.2 below). In fact, we will also need here a variant of the formula from [9], whose derivation from the formula in [9] involves Ekeland variational principle (cf. Proposition 2.3 below). Note that the fact that Holland s minimax formula can be used to study the asymptotic behavior of provides an answer to another question raised in Section 5 of [1].

3 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 3 Finally it turns out that our approach can be easily adapted to the case of the Neumann or Robin boundary conditions (cf. Section 6). Dealing with the Robin boundary conditions provides an answer to another question raised in Section 5 of [1]. The plan of our paper is the following. Section 2, which is independent from the rest of the paper, is devoted to Holland s minimax formula. In Section 3, we mainly concentrate on the question of the existence of and in Section 4 on the properties of the function! : Section 5 deals with the asymptotic behavior of as! 1. Finally, in Section 6, we consider the Neumann-Robin boundary conditions. 2. Minimax formula Let us start by stating the assumptions to be imposed on the operator L and the weight m in (1:1) throughout the paper. is a bounded domain in R N of class C 2 and the coe cients of L satisfy: A is a symmetric uniformly positive de nite N N matrix, with A 2 C 0;1 ; a 2 C 0;1 ; R N and a 0 2 L 1 () : The weight m belongs to L 1 () and we will always assume m (the case m 0 can of course be treated by changing m into m). Since the parameter plays no speci c role in this section, we will assume it equal to 1 and write L and (1:1) instead of L 1 and (1:1) 1 : Note that we do not require in this section a (x) to be divergence free. Our purpose in this section is to present a variant of the minimax formula derived in [9] for the principal eigenvalues of (1:1) : By a principal eigenvalue we mean 2 R such that (1:1) admits a solution u 6 0 with u 0: The following result is well known (cf. e.g. [11], [4], [14], [6], [9]). Proposition 2.1. Assume rst a 0 0: If m 0 then (1:1) admits exactly one principal eigenvalue which is >0; if m changes sign, the (1:1) admits exactly two principal eigenvalues, one is > 0 and the other is < 0: Without the assumption a 0 0; (1:1) may have zero or one principal eigenvalues if m 0; and zero, one or two principal eigenvalues if m changes sign. All these eigenvalues are simple, and the associated positive eigenfunctions belong to W 2;p ()\D () for any 1 p < 1 (where D () is de ned below). We now recall the minimax formula from [9]. Let us de ne the weighted Sobolev space H 1 ; d 2 := u 2 Hloc 1 () : d 2 u 2 + jruj 2 < 1 where d (x) = dist ; as well as the set of functions comparable to d (x): D () := fu :! R measurable: c 1 d u c 2 d a. e. in ; for some positive constants c i = c i (u)g: Proposition 2.2. (cf. [9]) Assume the existence of a principal eigenvalue for (1:1) and let be the largest of these eigenvalues. Then (u) (2.1) = inf u2u D inf Q u (v) v2h 1 (;d 2 )

4 4 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA where U D := u 2 H0 1 () \ D () : mu 2 = 1 ; (u) := haru; rui + ha; rui u + a 0 u 2 ; Q u (v) := u 2 (harv; rvi ha; rvi) : For any u 2 U D the in mum over v in (2:1) is achieved; moreover, when a 0 0 or when m (x) " > 0; the in mum over u in (2:1) is also achieved. In our study of the asymptotic behavior of the principal eigenvalue of (1:1), we will need a formula like (2.1) where u varies in the whole H 1 0 () instead of H 1 0 () \ D () : Proposition 2.3. Assume the existence of a principal eigenvalue for (1:1) and let be the largest of these eigenvalues. Then (u) (2.2) = inf u2u where U := u 2 H0 1 () : E u := v 2 H 1 () : inf Q u (v) v2e u mu 2 = 1 ; u 2 jrvj 2 < 1 : When a 0 0 or when m (x) " > 0; the in mum over u in (2.2) is achieved at some u 2 U D : The rest of this section is devoted to the proof of Proposition 2.3. A direct argument by approximation using Proposition 2.2 does not seem to work due to the lack of control on v in (2.2) when u lies only in H0 1 () : The proof below uses Proposition 2.2 as well as Ekeland variational principle. Let us start with some preliminaries concerning the space E u and the quadratic R expression Q u (v) : Fix u 2 H0 1 () with u 6 0: Then p u (v) := u2 jrvj is a seminorm on E u ; which becomes a norm on the quotient space G u of E u by F u := fv 2 E u : p u (v) = 0g : Call G u the completion of G u with respect to that norm and denote by p u the extended norm on G u : We will mainly deal below with G u ; which can be seen as the set of elements of E u de ned up to elements of F u : Our functional Q u (v) is made of a quadratic part R u2 harv; rvi and a linear part R u2 ha; rvi : It is clear that Q u (v) is well de ned for v 2 G u and extends by continuity into a functional Q u on G u : Moreover Q u is a C 1 functional on G u ; and its derivative at v 2 G u in the direction of h 2 G u is given by T u (v) h; where T u (v) h for v; h 2 G u denotes the extension by continuity of T u (v) h := u 2 (2 harv; rhi ha; rhi) where in this latter formula v; h 2 G u :

5 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 5 Note the coercivity relation (2.3) Q u (v) c 1 p u (v) 2 c 2 8v 2 G u for some constants c 1 ; c 2 with c 1 > 0: It is this relation which will provide some control on v: One also clearly has (2.4) inf v2g u Q u (v) = inf v2e u Q u (v) : Moreover, if u 2 H 1 0 () \ D () ; then (2.5) inf Q u (v) = inf Q u (v) : v2e u v2h 1 (;d 2 ) Indeed, the inequality in (2.5) follows from the inclusion E u H 1 ; d 2 ; while the inequality follows from the density of C 1 (and so of E u ) in H 1 ; d 2 (cf. [15]). We are now ready to start the Proof of Proposition 2.3. We will rst prove that inequality holds in (2.2), i.e. (2.6) (u) inf v2e u Q u (v) for all u 2 U: Part of the argument will follow [9]. Let u be a positive eigenfunction associated to : Clearly it su ces to prove (2.6) for u 0: Take such a u and x " > 0: Multiplying both sides of the equation satis ed by u by u 2 = (u + ") and using the divergence theorem, one gets mu u 2 = (u (2.7) + ") Aru ; ru 2 = (u + ") h i = haru ; ru i = (u + ") 2 + ha; ru i = (u + ") u 2 + a 0 u u 2 = (u + ") : The use of the divergence theorem can be justi ed by observing that u 2 = (u + ") 2 W 1;s 0 () for some s > 1 and that div (Aru ) 2 L s0 () with s 0 the Hölder conjugate of s; it then su ces to approximate u 2 = (u + ") by Cc 1 () functions. Now, a simple calculation based on the idea of completing a square gives that for any x 2 and 2 R N ; h A (x) ; i ha (x) ; i = min h; i + 1 a (x) + ; A 1 (x) (a (x) + ) 2R N 4 (cf. (3.11) in [8]). We apply the inequality provided by the above relation with = ru = (u + ") to derive from (2.7) that for any measurable function :! R N with R u2 jj 2 < 1; one has mu u 2 = (u + ") Aru 2 (2.8) u 2 ; ru = (u + ") 1 a + ; A 1 (a + ) u 2 + a 0 u u 2 = (u + ") : 4 In order to select a suitable in (2.8) we recall the following consequence of Ekeland variational principle (cf. e.g. [18]): if a C 1 functional on a Banach space

6 6 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA is bounded from below, then there exists a minimizing sequence satisfying the (PS) condition; moreover the elements of this sequence can be taken in any pregiven dense subset of the Banach space. Applying this result to our functional Q u on G u ; with G u as a dense subset, we obtain a sequence v k 2 E u such that for k! 1; (2.9) Q u (v k )! inf v2e u Q u (v) ; (2.10) u 2 h2arv k a; rvi = o (p u (v)) for v 2 E u where the o (p u (v)) in (2.10) is uniform with respect to v 2 E u as k! 1: We then take = a + 2A (rv k + ru= (u + ")) in (2.8), which gives, after some simple calculations, mu u 2 = (u + ") 2 haru; ru i u (1 u= (u + ")) = (u + ") + u 2 h2arv k a; r log (u + ")i haru; rui u 2 = (u + ") 2 + u 2 h2arv k a; r log (u + ")i + u 2 ha; rui = (u + ") + u 2 harv k ; rv k i + a 0 u u 2 = (u + ") : Since log (u + ") and log (u + ") belong to E u ; we deduce from (2.10) that when k! 1; the third integral of the left-hand side and the second integral of the Rright -hand side are o (1) as k! 1; moreover one also derives from (2.10) that u2 harv k ; rv k i di ers from Q u (v k ) by a o (p u (v k )) as k! 1: It thus follows that mu u 2 = (u + ") 2 haru; ru i u (1 u= (u + ")) = (u + ") haru; rui + ha; rui u 2 = (u + ") + a 0 u u 2 = (u + ") Q u (v k ) + o (1) + o (p u (v k )) : Since by (2.3), v k remains bounded in G u ; one has that o (p u (v k ))! 0 as k! 1: Going to the limit as k! 1 in the above inequality thus leads to mu u 2 = (u + ") 2 haru; ru i u (1 u= (u + ")) = (u + ") haru; rui + ha; rui u 2 = (u + ") + a 0 u u 2 = (u + ") inf Q u (v) : v2e u Finally, letting "! 0 and using dominated convergence, we get (2.6). It remains to prove that inequality holds in (2.2). Let " > 0: By Proposition 2.2 there exists u 2 U D such that + " (u) inf Q u (v) : v2h 1 (;d 2 )

7 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 7 Since (2.5) holds for such a u (in fact only the trivial inequality in (2.5) is used here) and since U D U; we deduce that + " inf (u) u2u inf Q u (v) : v2e u Letting "! 0 leads to the conclusion. Finally, the result relative to the achievement of the in mum over u in (2.2) follows from the corresponding result in (2.1) using (2.5). Remark 2.4. In the applications of the minimax formula treated in this paper, any of the two versions above can be used, except in the proof of Proposition 4.4 (resp. Theorem 5.1) where the old (resp. new) version is needed. More comments on the need of this new version are given after the statement of Theorem Existence of The assumptions in this section as well as in Sections 4 and 5 are those stated at the beginning of Section 2, with in addition (3.1) div a = 0 in : As observed in Proposition 2.1, problem (1:1) may have zero, one or two principal eigenvalues. In case of existence, we denote by the largest principal eigenvalue. Our purpose in this section is mainly to investigate various situations of existence or nonexistence of : Existence for any certainly holds when a 0 (x) 0 (by Proposition 2.1), or when m (x) " > 0 (by writing L u = mu as L u + lmu = ( + l) mu; which allows to reduce for l large to the preceding case of a nonnegative zero order coe cient). We will use repeatedly in this paper the -function () de ned in (1:2). Clearly is a principal eigenvalue for (1:1) if and only if vanishes at : The following lemma (which does not depend on assumption (3.1)) collects some properties of this function. References include [13], [10], [2], [16]; the proof of most of the properties below can for instance be adapted from the proof of Lemma 2.5 in [8]. Lemma 3.1. (i) If a 0 0; then (0) > 0: (ii) If m + 6 0; then ()! 1 as! 1; if m 6 0; then ()! 1 as! 1: (iii) The function! () is real analytic; moreover a solution u of (1:2) can be chosen so that! u 2 C0 1 is real analytic. (iv) The function! () is concave. (v) If m 0; then! () is strictly decreasing. Proposition 2.3 can be used in (1:2) to get the following expression for () : (3.2) () = inf (u) u mu 2 inf v Q u (v)

8 8 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA where u varies in H0 1 () with R u2 = 1; v varies in E u ; and where (u) and Q u (v) denote (u) := haru; rui + ha; rui u + a0 u 2 ; Q u (v) = u 2 (harv; rvi ha; rvi) : Note that as this stage, Proposition 2.2 could be used as well (see Remark 2.4). A simple calculation based on the divergence theorem and hypothesis (3.1) shows that (3.3) (u) = 0 (u) 8u 2 H 1 0 () ; 8 2 R; moreover, replacing v by v in Q u (v) gives (3.4) inf v Q u (v) = 2 inf v Q1 u (v) 8 2 R; consequently, (3.5) () = inf u 0 (u) mu 2 2 inf v Q1 u (v) where u and v vary as above. These formulas (3.3), (3.4), (3.5) will be used repeatedly. We are now ready to start the study of the existence of : Proposition 3.2. (i) exists if and only if exists, and then = : (ii) If exists for some 0; then exists for all and : Proof. (i) By (3.5), () = () ; and the conclusion follows. (ii) We know that is the largest zero of : Using (3.5) and the fact that inf v Q 1 u (v) 0; one has that : Since by Lemma 3.1, ()! 1 as! +1; one deduces that has a zero which is : The conclusion follows. Corollary 3.3. If 0 exists, then exists for all ; and 0 for all. As will be seen in Example 3.10, the existence of for all 6= 0 does not imply the existence of 0 : Because of (i) from Proposition 3.2, we will from now on mostly restrict ourselves to the study of for 0: The rest of this section is mainly devoted to various examples related to the existence or nonexistence of : Our starting point will be the observation that if m changes sign, or if m 0 but vanishes on a ball B ; then, for l su ciently large, the problem (3.6) u lu = mu in ; u = 0 has no principal eigenvalue (cf. e.g. [9]). Our rst example shows that may exist for some but not for others. Example 3.4. Take m such that (3.6) has no principal eigenvalues for l large. Consider the problem (3.7) u + ha; rui lu = mu in ; u = 0

9 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 9 with a (x) = a a nonzero constant vector eld. Writing u = e ha;xi v; one easily sees that (3.7) is equivalent to (3.8) v + ( 2) ha; rvi + ( 1) jaj 2 l v = mv in ; v = 0 For = 2 we deduce from (3.6) that (3.8) (and so (3.7)) has no principal eigenvalue for l su ciently large. Fixing such a l; we then see that for su ciently large, the zero order coe cient of (3.8) is nonnegative, and consequently (3.8) (and so (3.7)) has a principal eigenvalue. Remark 3.5. By a change of unknown function as in the above example, one can see that if A (x) I and a (x) = a is a nonzero constant vector eld, then (1:1) always has a principal eigenvalue for su ciently large. This is in fact a particular case of the proposition below, which concerns gradient elds. Note that the proof of this proposition relies on some unique continuation property. Proposition 3.6. If a (x) 6 0 satis es (3.1) and is of the form a = 2Arg for some g 2 C 1;1 () ; then exists for su ciently large, and moreover! +1 as! 1: that Proof. It follows from (3.5), after completing the square in Q 1 u (v) = u 2 ha (rv) 2rg; rvi ; (3.9) () = inf u 0 (u) + 2 harg; rgi u 2 mu 2 : This latter expression is the -function associated to the selfadjoint operator L 0 u+ 2 harg; rgi u and the weight m: By Proposition 3.2,! () is nondecreasing on 0: We look at (0) : If (0) goes to +1 as! 1; then Lemma 3.1 implies that has a zero for large, and the existence of follows. If on the contrary (0) remains bounded as! 1; then we consider (3.10) L 0 u + 2 harg; rgi u = () u where u is an eigenfunction associated to (0) and normalized by ku k L 2 = 1: Taking u as test function in the weak form of (3.10), one sees that u remains bounded in H0 1 () : Going to a weakly convergent subsequence and dividing by 2 ; one deduces the existence of u 2 H0 1 () with kuk L 2 = 1 and (3.11) harg; rgi u 2 = 0: But since by (3.1), div (Arg) = 0 in ; one has that fx 2 : rg = 0g has measure zero (cf. [7]). This contradicts (3.11). We have thus proved the existence of for large. We now prove that! +1 as! 1: Since! is nondecreasing, we can assume by contradiction that remains bounded. We go back to (3.9), however now with = ; and obtain L o u + 2 harg; rgi u = mu

10 10 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA where here u is an eigenfunction associated to and normalized by ku k L 2 = 1: Arguing as above, one reaches again (3.11) for some u with kuk L 2 = 1; and the same contradiction follows. In opposition with the previous situation, in the following example fails to exist for all : Example 3.7. Take the unit disk in R 2 ; m (x) as in (3.6) with in addition m (x) radial, and x l such that (3.6) has no principal eigenvalues. Consider the - function l;m () corresponding to (3.6), which we abbreviate into () ; and let w be the positive normalized (by kw k L 2 = 1) eigenfunction associated to () : (3.12) w lw mw = () w in ; w = 0 Simplicity implies that w is radial; moreover the nonexistence of a principal eigenvalue for (3.6) implies () < 0 for all ; nally, multiplying (3.12) by w and integrating give (3.13) jrw j 2 lw 2 mw 2 = () < 0: We now take the vector eld a (x) = ( x 2 ; x 1 ) : Clearly (3.1) holds, and since w is radial, we have ha; rw i = 0 in : Thus w 2 H0 1 () is a rst integral for a; and so, by Lemma 3.8 below, (3.14) inf v2e w Q 1 w (v) = 0 where the matrix A (x) in Q 1 w is in the present example the identity matrix. Consider now the function () associated to the problem (3.15) u + ha; rui lu = mu in ; u = 0 Taking u = w in formula (3.5), using (3.14) and (3.13), one has () jrw j 2 lw 2 mw 2 2 inf v Q1 w (v) < R; 8 2 R. This implies that for any ; (3.15) has no principal eigenvalue. Lemma 3.8. Let w 2 H 1 0 () be a rst integral for a: Then inf Q 1 w (v) = 0: v2e w Proof. Using (3.1) and the fact that w is a rst integral, one obtains w 2 ha; rvi = 0 for any v 2 E w : The derivation of this equality involves truncating w and using the divergence theorem, which can be justi ed by standard approximation arguments. It follows that inf Q 1 w (v) = inf w 2 harv; rvi = 0: v2e w v2e w

11 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 11 We conclude this section with the study of the following question: when exists for some but not for others, it seems natural to look at 0 := inf f 0 : existsg and to ask wether 0 exists or not. The following two examples show that no general answer can be given. We will use below the easily veri ed fact that 0 exists if and only if remains bounded when! 0 ; > 0 : Example 3.9. If m changes sign, then 0 exists. Indeed, picking u 2 H0 1 () such that R mu2 > 0; one deduces from (3.5) that for 2 ] 0 ; 0 + 1[ ; () c 1 for a constant c 1 > 0 and for all > some 1 > 0; with c 1 and 1 independent of : Picking u 2 H0 1 () such that R mu2 < 0 one similarly deduces that () c 2 for some c 2 > 0 and all < some 2 < 0; This implies that remains between 2 and 1 when! 0 ; > 0 ; and the conclusion follows. Example If m 0 vanishes on some ball B ; then 0 may not exist. To build an example, we start from (3.6) with l = l 0 ; where l 0 is chosen such that (3.6) with l = l 0 has no principal eigenvalue but (3.6) with any l < l 0 has a principal eigenvalue. The existence of such a l 0 easily follows by looking at the shape of the corresponding function () (cf. Lemma 3.1). We then consider (3.16) u + ha; rui l 0 u = mu in ; u = 0 where a is a nonzero constant vector eld. Writing u = e ha;xi v; one easily veri es that (3.16) is equivalent to v + ( 2) ha; rvi l jaj 2 jaj 2 v = mv in ; v = 0 For = =2; this latter problem becomes a problem of the form (3.6): v l jaj2 v = mv in ; v = 0 We can then conclude that (3.16) has a principal eigenvalue for > 0 but does not have a principal eigenvalue for = 0: 4. Function We will limit ourselves to the study of the function! on 0; which, by Proposition 3.2, is no real restriction. The domain A R + of this function is thus either empty or of the form [r; +1[ or ]r; +1[ for some r 0: We start with a continuity result which, for m 1; is a particular case of Proposition 5.1 of [2]. Proposition 4.1. Suppose A non empty. Then is continuous in A. Proof. We rst show that if 0 exists for some 0 0; then is continuous from the right at 0 : Let " > 0 and consider 0 : By Proposition 2.3, there

12 12 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA exists u = u ("; 0 ) 2 H 1 0 () with R mu2 = 1 such that 0 + " 0 (u) inf v Q0 u (v) = (u) inf v Q u (v) inf v inf v Q1 u (v) ; Q1 u (v) where we have used (3.3), (3.4), and where the in ma are taken for v varying in E u : So " inf v Q1 u (v) ; and the continuity from the right follows. We will now show that if exists in a neighborhood of some 0 ; then is continuous from the left at 0 : Consider a sequence k! 0 ; with k < 0 : By monotonicity, the limit of k exists and is 0 : Continuity from the left will follow if we show that for any given such that k for all k; one has 0 : Fix such a and call u k the positive eigenfunction associated to k ; normalized by ku k k C 1 0 () = 1: Since k and k is the largest principal eigenvalue of (1:1) k ; one has k 0: Moreover by Lemma 4.2 below, k remains bounded as k! 1; and consequently, taking l su ciently large in the equation satis ed by u k : (L 0 + l) u k = mu k + k u k k ha; ru k i + lu k ; one deduces that u k remains bounded in W 2;p () for any nite p: It follows that for a subsequence, k! 0; u k! u in C 1 0 with kuk C 1 0 () = 1 and u 0; with moreover, from the equation above, L 0 u = mu + u: This implies that = 0 ; and consequently 0 0: To go on we will now use repeatedly the properties of the function 0 () as described in Lemma 3.1. If m 0; we immediately deduce from 0 0 that 0 ; and we are nished. If m changes sign and (1:1) 0 has two principal eigenvalues, then we observe that the preceding argument can be repeated for any 0 and gives 0 ( 0 ) 0; this now implies 0 : Finally if m changes sign and (1:1) 0 has only one principal eigenvalue 0 ; then we argue di erently: since k 0 implies k 0 (cf. the proof of Proposition 3.2), one necessarily has in this situation k = 0 for all k; and the conclusion a0 follows. Lemma 4.2. () remains bounded when and remain bounded. Proof. For any u 2 H0 1 () with R mu2 = 1; one has, by (3.5), () 0 (u) mu 2 2 inf v Q1 u (v) const (u) when and remain bounded. On the other hand, for any u as above, 0 (u) mu 2 2 inf v Q1 u (v) 0 (u) mu 2 L0;m () and consequently () L0;m () : The conclusion follows.

13 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 13 We now turn to the study of the strict monotonicity of the function! on A R + : Proposition 4.3. Suppose A nonempty. Then either is strictly increasing on A; or is de ned and constant for all 0: Proposition 4.3 follows from Proposition 4.4. below, which in particular characterizes the constant case in terms of the associated selfadjoint problem (1:1) 0 : is constant if and only if 0 exists and its eigenfunction is a rst integral. It will convenient from now on to call I 0 the set of rst integrals of a in H 1 0 () ; i.e. I 0 := w 2 H 1 0 () : w 6 0 and ha; rwi = 0 a.e. in : Proposition 4.4. (i) If = for some 0 < ; then 0 exists and is equal to = ; moreover the eigenfunction u associated to 0 belongs to I 0 : (ii) Conversely if 0 exists and if its eigenfunction belongs to I 0 ; then exists and = 0 for all 0: The rest of this section is mostly devoted to the proof of Proposition 4.4. Proof of Proposition 4.4. We rst deal with (i) in the particular case where a 0 0: Let 0 < and assume = : Our starting point will be the minimax formula (2.1) for ; where, since a 0 0; the in mum over u 2 U D is achieved at some eu 2 H0 1 () \ D () with R meu2 = 1: Using (3.3) and (3.4), formula (2.1) for gives (eu ) inf v Q eu (v) = (eu ) inf v Q eu (v) 2 2 inf v Q1 eu (v) = 2 2 inf v Q1 eu (v) where the in ma are taken for v varying in H 1 0; the above implies (4.1) inf v Q1 eu (v) = 0: ; d 2 : Since = and inf v Q1 eu (v) We now apply Lemma 3.2 from [9]: the in mum in (4.1) is achieved at some W eu 2 H 1 ; d 2 ; which satis es (4.2) 2ArWeu a; r' = 0 8' 2 H 1 ; d 2 ; eu 2 R and the value of this in mum is eu2 ArWeu ; rw eu : Consequently, by (4.1), this last integral has value zero, and so W eu is constant on : Replacing in (4.2) gives (4.3) eu 2 ha; r'i = 0 8' 2 H 1 ; d 2 : We now apply another result from [9]: denoting by u the positive eigenfunction q of (1:1) associated to ; one can write eu as u G for some function G 2

14 14 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA ; d 2 \ L 1 () which is some " > 0 and which satis es (4.4) u 2 ArG + ag ; r' = 0 8' 2 H 1 ; d 2 H 1 (cf. Lemma 3.3 and the second part of the proof of Theorem 3.1 in [9]). Using (4.3) in (4.4) yields u 2 ArG ; r' = 0 8' 2 H 1 ; d 2 which implies, taking ' = G ; that G is a constant on : Consequently (4.3) becomes u 2 ha; r'i = 0 8' 2 H 1 ; d 2 : Applying the divergence theorem as well as (3.1) to this last relation gives a; ru ' = 0 8' 2 H 1 ; d 2 u D E and so a; ru = 0 in ; i.e. u belongs to I 0: The equation in (1:1) thus becomes L 0 u = mu ; which shows that is a principal eigenvalue for (1:1) 0 ; with u as associated positive eigenfunction. Since 0 and 0 is the largest principal eigenvalue of (1:1) 0 ; we conclude that = 0 : The proof of (i) in the case a 0 0 is thus completed. We now turn to the proof of (i) in the general case where a 0 is not necessarily 0: Let as before 0 < with = and call this common value = : Suppose rst > 0 and take l su ciently large so that the zero order coe cient of L +l is 0 and m+l is 0: Denoting as before by u the positive eigenfunction of (1:1) associated to ; one has L + l u = (m + l) u ; and it follows from Proposition 2.1 that is the largest principal eigenvalue of L +l for the weight m+l: Of course a similar conclusion holds for L + l u = (m + l) u : is the largest principal eigenvalue for L + l for the weight m + l. By the part of (i) which is already proved, we thus conclude that L 0 + l with the weight m+l admits as the largest principal eigenvalue and that the corresponding positive eigenfunction u belongs to I 0 : Clearly is then a principal eigenvalue for (1:1) 0 with u as positive eigenfuinction. Moreover = 0 ; and we conclude that = 0 : The proof of (i) is thus completed in the case under consideration. Suppose now < 0: The preceding argument can be easily adapted by writing L u = mu as L l u = ( m + l) u and by taking l su ciently large. Suppose nally = 0: Again the preceding argument can be easily adapted by writing L u = 0 as (L + l) u = lu and by taking l su ciently large. We now turn to the proof of (ii). By assumption 0 exists and its positive L 2 normalized eigenfunction u satis es ha; ru i = 0: One thus has L u = L 0 u = 0 mu ; which shows that for any 0; 0 is a principal eigenvalue for (1:1) : The proof of (ii) will be completed if we show that 0 is the largest principal eigenvalue of (1:1) :

15 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 15 This is easily veri ed in the cases a 0 0 or m 0; by using Proposition 2.1. Indeed, if a 0 0; then 0 is > 0 and consequently is the largest principal eigenvalue of (1:1) ; and if m 0; then (1:1) has only one principal eigenvalue. Our purpose in the general case is to prove that (4.5) 0 ( 0 ) 0: The properties of the function () (cf. Lemma 3.1) then clearly imply the desired conclusion that 0 is the largest principal eigenvalue of (1:1). To prove (4.5) we apply Lemma 4.5 below with = 0 : Here u ; = u ; moreover, since u is a rst integral for a; one has L u = L u and so u ; = u : It follows that 0 ( 0 ) = R m (u ) 2 R (u ) 2 ; which in particular implies that 0 ( 0 ) does not depend on : Since 0 is the largest principal eigenvalue of (1:1) 0 ; one has 0 0 ( 0 ) 0; and consequently 0 ( 0 ) 0. The following lemma was used in the proof above, and will be used again later. Consider (1:2) and let u ; be a positive eigenfunction associated to () : L u ; mu ; = () u ; : We will also consider (1:2) ; () and u ; : Recall that by Proposition 3.2, () = () Note in connection with the lemma below that by (3.1), the adjoint of L (with zero Dirichlet boundary condition) is L (with zero Dirichlet boundary condition). Lemma 4.5. Let be a principal eigenvalue of (1:1) ; i.e. 0 R = R mu ; u ; u ; u : ; = 0: Then m and let v be an asso- Proof. Call () the -function associated to L ciated positive eigenfunction: L m v mv = () v : Clearly () = + so that (0) = 0 and 0 (0) = 0 ; moreover, up to a multiple, v 0 = u ; : We recall that v can be chosen so that () 2 R and v 2 C0 1 are analytic functions of ; moreover the above equation for v can be rewritten, for l su ciently large, as v = T [mv + () v + lv ] where T is the continuous operator T := L m + l 1 : L 1 ()! C0 1 : Taking the derivative with respect to and making = 0 = T (mv 0 ) + (0) T (v 0 ) + lt : =0 Now a simple calculation using (3.1) and the divergence theorem shows that R R ut v = (T u) v for u; v 2 L 1 () : Multiplying (4.6) by u ; ; integrating and using

16 16 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA the fact that T u ; = 1 l u ; ; one obtains mv 0 u ; + 0 (0) u ; v 0 = 0: This is the desired relation since v 0 = u ; : Here is a consequence of Lemma 4.5 which has independent interest. recall that we systematically assume m + 6 0: Let us Corollary 4.6. Consider the selfadjoint problem (1:1) 0 and assume it admits a principal eigenvalue. Call 0 its largest principal eigenvalue and let u be an associated positive eigenfunction. Then R m (u ) 2 0: More precisely R m (u ) 2 = 0 in the case when m changes sign and has a unique principal eigenvalue, and R m (u ) 2 > 0 in all the other cases (i.e. m changes sign and (1:1) 0 has two principal eigenvalues, or m 0). Proof. One has by Lemma 4.5 m (u ) 2 = 0 0 ( 0 ) (u ) 2 ; and the conclusion follows from the shape of the function 0 () (cf. Lemma 3.1). We conclude this section with an example illustrating Propositions 4.3 and 4.4. Example 4.7. Let be the unit disc in R 2 ; m 1; and let 0 be the (unique) principal eigenvalue of on H0 1 () ; with u the associated positive eigenfunction. By simplicity, u is radial. So if we take a (x) = ( x 2 ; x 1 ) ; we see that (3.1) holds and that u belongs to I 0 : Moreover, by Proposition 2.1, the problem u + ha; rui = u in ; u = 0 has a unique principal eigenvalue, which is clearly 0 : Thus = 0 for all : This conclusion of course also follows from part (ii) of Proposition Asymptotic behavior of The results in this section provide a complete extension to the inde nite case of the main result of [1] recalled in the introduction. The new feature is the condition R mw2 > 0 which involves the weight. Theorem 5.1. Assume A nonempty and nonconstant. Then remains bounded as! 1 if and only if the vector eld a (x) admits a rst integral w in H 1 0 () with R mw2 > 0: Proof. We start with the necessary condition. So we assume bounded as! 1: Call u the L 2 normalized positive principal eigenfunction associated to in (1:1) : Multiplying both sides of L u = mu by u and integrating, using (3.1), one sees that u remains bounded in H 1 0 () : Consequently, for a subsequence, u converges to some w 2 H 1 0 () ; weakly in H 1 0 () and strongly in L 2 () ; and one has w 0; kwk L 2 = 1: Dividing both sides of L u = mu by ; one

17 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 17 obtains that ha; ru i! 0 in D 0 () : Since on the other hand ha; ru i! ha; rwi in D 0 () ; one concludes that w 2 I 0 : The proof of the necessary condition would be completed if we could show that (5.1) mw 2 > 0: Inequality (5.1) is easily veri ed when m (x) " > 0 or when a 0 0: Indeed, the rst case is trivial. In the second case, multiplying as above the equation by u ; one obtains (5.2) harw; rwi + a 0 w 2 lim mw 2 ; since a 0 0; the left-hand side is > 0; as well lim (because! is nondecreasing and each is > 0); (5.1) thus follows. In the general case, using the assumption nonconstant, we will construct another rst integral which will satisfy an inequality like (5.1). For this purpose we consider the -function () de ned in (1:2) ; we will denote below by u ; the associated L 2 normalized positive eigenfunction. Using the fact that remains bounded as! 1 and is nonconstant, one has the following Claim. There exist 1 and a constant c > 0 such that (5.3) 0 ( ) c for 1 : Accepting this claim for a moment, one deduces from the concavity of the function! () that (5.4) () c ( ) for 1 and : It then follows from the equation satis ed by u ; that (5.5) haru ; ; ru ; i + [a 0 + c ( )] u 2 ; mu 2 ;: Since remains bounded, one can select with > 0 and 1 := sup in such a way that the bracket [a 0 + c ( )] in (5.5) is 0 for all : Call this value of : By (5.5) u ; remains bounded in H0 1 () as! 1; and so, for a subsequence, u ; converges to some w 2 H0 1 () ; weakly in H0 1 () and strongly in L 2 () ; and one has kwk L 2 R = 1: One then deduces from (5.5) that harw; rwi R mw2 ; which implies R mw2 > 0: We will now show that w is a rst integral. For this purpose observe that remains bounded as! 1 since ( ) = 0 and is non decreasing by Proposition 3.2. It then easily follows, as before, by dividing by the equation that R ha; rwi = 0; i.e. mw2 > 0: L u ; mu ; = u ; ; w 2 I 0 : So w provides a rst integral in H 1 0 () with Proof of the Claim. The proof goes in 3 steps: (i) 0 ( ) 0 for all 2 A; (ii) 0 ( ) < 0 for 2 Int (A) ; (iii) inequality (5.3) holds for su ciently large.

18 18 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA Step (i) is a direct consequence of the fact that is the largest principal eigenvalue of (1:1) : To prove step (ii), let us x any 0 2 A and consider, for 2 R, ( 0 ) ; with v its associated L 2 normalized positive eigenfunction: (5.6) L v 0 mv = ( 0 ) v : By Proposition 3.2, the function! ( 0 ) is nondecreasing in R + ; moreover, by Proposition 4.4, it is strictly increasing if ha; rv 0 i 6 0: Let us rst consider this case. We thus have, for > 0 ; ( 0 ) > 0 ( 0 ) = 0; the shape of the function (cf. Lemma 3.1) then implies the conclusion 0 ( ) < 0: To complete the proof of step (ii), we will show that the situation where (5.7) ha; rv 0 i 0 is impossible. Indeed, if (5.7) holds, then, by Proposition 4.4, ( 0 ) as a function of is constant on R; and so ( 0 ) = 0 ( 0 ) = R: It follows that (5.6) becomes (5.8) L v = 0 mv ; which shows that 0 is a principal eigenvalue of L for the weight m: We will prove that it is the largest principal eigenvalue for the weight m; i.e. that 0 = 8 2 R, contradicting the assumption that the function! is nonconstant. That 0 is the largest principal eigenvalue of L for the weight m follows if we show (5.9) 0 ( 0 ) 0: To prove (5.9) we rst deduce from (5.8) and (5.7) that L v 0 = L 0 v 0 = 0 mv 0 ; comparing with (5.8) implies that v = v R: Applying now Lemma 4.5 to (1:1) with = 0 and observing that with the notations of Lemma 4.5, u 0 ; = v and u 0 ; = v ; we obtain R 0 ( 0 ) = R mv v v v R = R mv 0 v 0 v = 0 0 v 0 ( 0 ) 0; 0 where the latter inequality holds by de nition of 0 : Thus (5.9) holds and the proof of step (ii) is complete. To prove step (iii), x 0 2 Int (A) : Since by step (ii) 0 0 ( 0 ) < 0; one can nd < 0 such that 0 > 0: Let now > 0 : By Proposition 3.2, the concavity of the function! () and the fact that 0 ( ) 0; one has Consequently 0 < 0 0 ( ) 0 ( ) 1 : 0 ( ) 0 1 ; and the conclusion of the step (iii) follows. This concludes the proof of the claim and of the necessary part of Theorem 5.1.

19 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 19 It remains to prove the su cient part in Theorem 5.1. So suppose the existence of w 2 I 0 with R R mw2 > 0: Normalizing we can without loss of generality assume mw2 = 1: By Proposition 2.3, (3.3) and (3.4), one has (5.10) = inf u2u 0 (u) 2 inf v2e u Q 1 u (v) 0 (w) ; where the last inequality follows from Lemma 3.8. This shows that remains bounded as! 1: The proof of Theorem 5.1 is now complete. Remark 5.2. When is constant and either a 0 0 or m 0; then there still exists a rst integral w in H0 1 () satisfying R mw2 > 0: Indeed, by Proposition 4.4, the positive eigenfunction u 0 associated to 0 belongs to I 0 : Moreover one derives from the equation for u 0 that (5.11) haru 0 ; ru 0 i + a 0 u 2 0 = 0 mu 2 0: When m 0; then R mu2 0 > 0 since u 0 (x) > 0 in : When a 0 0; then 0 > 0 and the left-hand side of (5.11) is > 0; which again imply R mu2 0 > 0: The example below shows that a rst integral w in H0 1 () satisfying R mw2 > 0 may fail to exist when is constant, a and m changes sign. Example 5.3. Let be the unit disc in R 2 and denote by 1 the principal eigenvalue of on H0 1 () with ' 1 a corresponding positive eigenfunction. For a (x) = ( x 2 ; x 1 ) ; we consider the problem (5:12) u + ha; rui + a 0 u = mu in ; u = 0 where m (x) := 1 + a 0 (x) and a 0 is chosen so that m (x) = 1 and the half-disk x 1 < 0 and m (x) = 1 on the half-disk x 1 > 0: Clearly div (a) = 0 in ; and since ' 1 is radial one also has ha; r' 1 i = 0 in ; so that ' 1 2 H0 1 () is a rst integral. Moreover, for = 0; 1 is clearly a principal eigenvalue of (5:12) 0, with ' 1 as associated eigenfunction. We now show that 1 is the largest principal eigenvalue of R (5:12) 0 : For that purpose we look at 0 0 (1) which, by Lemma 4.5, is equal to m'2 1= R '2 1; by construction of the weight, (5.13) m = 0 jxj=r for any 0 r 1; which implies, since ' 1 is radial, that R m'2 1 = 0; consequently 0 0 (1) = 0 and we see that, in fact, (5:12) 0 has a unique principal eigenvalue. Proposition 4.4 thus applies and yields that is constant. We now claim that for any rst integral w 2 H0 1 () ; one has (5.14) mw 2 = 0: Take a rst integral w: Since hrw; ai = 0; we see that for a.e. x 2 ; rw (x) is a multiple of x; i.e. of the form f (x) x; going to polar coordinates and writing w (r cos ; r sin ) = ew (r; ) ; one easily veri es ew=@ = 0; and so w is radial. Using (5.13), one concludes that (5.14) holds.

20 20 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA We nally turn to the study of the limiting value of as! 1 and investigate in particular the validity of (5.15) lim = inf harw; rwi + a 0 w 2 : w 2 I 0 and mw 2 = 1 : Theorem 5.4. Assume A nonempty and bounded as! 1: Then (5.15) holds if either is nonconstant, or a 0 0; or m 0: Moreover, in each of these cases, the in mum in (5.15) is achieved. Remark 5.5. Note that Example 5.3 shows that (5.15) may fail to hold when is constant, a and m changes sign. Note also that Theorem 5.1 implies that (5.15) still holds, in the trivial form 1 = 1; when is unbounded as! 1: Proof of Theorem 5.4. Theorem 5.1 (when is nonconstant) and Remark 5.2 R (when is constant with a 0 0 or m 0) imply the existence of w 2 I 0 with mw2 = 1: Moreover (5.10) shows that inequality always holds in (5.15). When is constant with either a 0 0 or m 0; then the argument in Remark 5.2 yields equality in (5.15) as well as the achievement of the in mum. We now consider the case where is nonconstant. We start by observing the general fact that (5.15) holds with the achievement of the minimum whenever remains bounded and m (x) " > 0: This is a simple consequence of (5.2). We apply this observation for each 2 R to () (for which the weight is 1). This is possible since, as observed at the beginning of the proof, there exists w 2 I 0 with R mw2 = 1; and so (3.5) and Lemma 3.8 imply that for each 2 R, () remains bounded as! 1: Consequently, writing 1 () := lim () where the limit exists by Proposition 3.2, one has (5.16) 1 () = inf harw; rwi + a 0 w 2 mw 2 : w 2 I 0 ; kwk L 2 = 1 ; and the in mum is achieved at some w : The idea of the proof of Theorem 5.4 in the case nonconstant is now the following. We will approximate 1 := lim from the right by and show that for a subsequence, w converges to some ew satisfying R ew 2 I (5.17) 0 ; k ewk L 2 = 1; m ew2 > 0; 1 R m ew2 = R har ew; r ewi + a 0 ew 2 : The existence of such a ew clearly completes the proof of Theorem 5.4. Claim 1. Let! 1 with > 1 : Then, for a subsequence, w! ew weakly in H0 1 () ; strongly in L 2 () ; and one has ew 2 I 0 ; k ewk L 2 = 1; R m ew2 > 0 and (5.18) 1 ( 1 ) = har ew; r ewi + a 0 ew 2 1 m ew 2 : Note that the existence of ew satisfying the conditions of Claim 1, with the exception of R m ew2 > 0; follows directly by taking = 1 in (5.16). The key point is that the approximation procedure from Claim 1 yields a ew which also satis es R m ew2 > 0: Claim 2. 1 ( 1 ) = 0:

21 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 21 Claims 1 and 2 clearly yield a function ew with the properties stated in (5.17). Proof of Claim 1. The function 1 : R! R is concave and so continuous, and consequently 1 ()! 1 ( 1 ) as! 1 : It then follows from (5.16) that w remains bounded in H0 1 () and so, for a subsequence, w converges to some ew 2 H0 1 () ; weakly in H0 1 () and strongly in L 2 () : One has k ewk L 2 = 1 and ew 2 I 0 (this latter point can be seen by going to the limit in R ' ha; rw i = 0 for ' 2 Cc 1 ()). One also obtains by going to the limit in (5.16) that 1 ( 1 ) har ew; r ewi + a 0 ew 2 1 m ew 2 : But one also has the inequality in the above relation by applying (5.16) with = 1 : So (5.18) follows. It remains to show that R m ew2 > 0: For this purpose we use (5.16) twice to get, for > 1 ; 1 () = harw ; rw i + a 0 w 2 mw 2 ; 1 ( 1 ) harw ; rw i + a 0 w 2 1 mw 2 : Substracting yields 1 () 1 ( 1 ) mw : 2 1 On the other hand, applying successively the mean value theorem, the concavity of and the claim from the proof of Theorem 5.1 (which can be used since we are dealing here with the case bounded and nonconstant), one obtains, for some with < 1 < < ; () ( 1 ) 1 = 0 () 0 ( ) c for su ciently large, with c a constant > 0. Going to the limit as! 1 and comparing with the previous inequality give mw 2 c: The conclusion R m ew2 > 0 now follows by letting! 1 : This completes the proof of Claim 1. Proof of Claim 2. We rst prove the existence of a constant c 0 > 0 such that (5.19) 0 ( 1 ) c 0 for su ciently large. Indeed, for some with < 1 < < 1 + 1; and consequently when! 1; since 1 ( 1 + 1) ( 1 + 1) ( 1 ) = 0 () 0 ( 1 ) ; lim inf 0 ( 1 ) 1 ( 1 + 1) 1 ( 1 ) ; 1 ( 1 ) is a xed number, (5.19) follows.

22 22 TOMAS GODOY, JEAN-PIERRE GOSSE, AND SOFIA PACKA We now look at 1 ( 1 ) and rst observe that since < 1 ; one has ( 1 ) < ( ) = 0; and so 1 ( 1 ) 0: Assume by contradiction 1 ( 1 ) = M < 0: Since is concave, one has 0 = ( ) ( 1 ) + 0 ( 1 ) ( 1 ) : This implies, for su ciently large, using (5.19), 0 M 2 + c0 ( 1 ) ; i.e. 1 M=2c 0 ; which contradicts the fact that! 1 : This completes the proof of Claim 2 and of Theorem 5.4. We conclude this section with several examples illustrating various situations which can occur in relation with the asymptotic behavior of as! 1: When a is a gradient eld,! +1 as! 1 (cf. Proposition 3.6). The case constant together with the existence (resp. nonexistence) of w 2 I 0 satisfying R mw2 > 0 is illustrated by Example 4.7 (resp. Example 5.3). The following example describes a situation where is nonconstant and has a nite limit as! 1: Example 5.5. Let be the unit disk in R 2 ; a (x) = ( problem (5.20) u + ha; rui + a 0 u = m (x) u in ; u = 0 x 2 ; x 1 ) and consider the By the choice of a; any nontrivial radial function w 2 H0 1 () belongs to I 0 : Fix Rsuch a w: We then take a 0 2 L 1 () with a 0 0; m 2 L 1 () with m 0 and mw2 > 0; with moreover either m radial and a 0 nonradial, or m nonradial and a 0 radial. With these choices, (5.20) has a unique principal eigenvalue for each ; is nonconstant (because the eigenfunction associated to 0 is nonradial and so does not belong to I 0 ; which allows the application of Proposition 4.4), and remains bounded (by Theorem 5.1). 6. Neumann-Robin boundary conditions Our purpose in this last section is to indicate how the preceding results can be adapted to deal with the problem L u = m (x) u in ; (6:1) A + b 0 (x) u = 0 Here A := ha; rui denotes the conormal derivative associated to the operator L ; with the exterior unit normal The assumptions on ; the operator L and the weight m are the same as those stated at the beginning of Section 2 (in particular m + 6 0); moreover b 0 2 C 0;1 (@) with b 0 0 It is well known that (6:1) admits 0, 1 or 2 principal eigenvalues, which are simple (cf. e.g. [10], [17], [8]). In case of existence we will again denote by the largest principal eigenvalue of (6:1). Proposition 6.1. Assume the existence of : Then (6.2) = inf (u) + b 0 u 2 inf u2u 0 v2h 1 () Q u

23 ASYMPTOTIC BEHAVIOR OF PRINCIPAL EIGENVALUES 23 where ; Q u have the same meaning as in Section 2 and where U 0 := u 2 H 1 () \ L 1 () : essinf (u) > 0 and mu 2 = 1 : For any u 2 U 0 the in mum over v in (6.2) is achieved; moreover when a 0 0 with either a or b 0 6 0; or when m (x) " > 0; the in mum over u in (6.2) is also achieved. Proposition 6.1 is proved in [8] when a 0 0 (cf. Theorems 3.1 and 3.6). Its derivation in the general case can be carried out by adapting to the present boundary conditions the arguments of the proof of Theorem 3.5 from [9]. The -function associated to (6:1) is de ned for 2 R as the unique principal eigenvalue = () of (6:3) L u mu = u in ; Bu = 0 This function enjoys properties similar to those stated in Lemma 3.1. The only changes occur in (i) where now (0) is > 0 if a 0 0 and either a or b 0 6 0; and in (iii) where now u should be considered as varying in C (cf. Lemma 2.5 in [8]). From now on we will assume, in addition to the conditions stated at the beginning of this section, that the vector eld a (x) satis es (6.4) div (a) = 0 in and ha; i = 0 It then follows from (6.2) that (6.5) () = inf u 0 (u) b 0 u 2 mu 2 2 inf v2h 1 () Q1 u (v) where u varies in U 0 0 where U 0 0 is de ned as U 0 with however now m 1. This formula for () implies the following Proposition 6.2. (i) exists if and only if exists, and then = : (ii) If exists for some 0; then exists for all ; and : As in the Dirichlet case, we will limit ourselves to the study of the function! for 0; and denote by A R + its domain. Arguing as in Section 4, we obtain Proposition 6.3. Assume A nonempty. Then is continuous on A. Moreover either is strictly increasing on A, or is de ned and constant for all 0: In addition the constant case occurs if and only if 0 exists and its eigenfunction belongs to I := w 2 H 1 () : w 6 0 and ha (x) ; rw (x)i = 0 a.e. in ; the set of rst integrals of a in H 1 () : The proof of Proposition 6.3 is similar to that of Propositions 4.1 and 4.4. The main di erence consists at the beginning of the proof of (i) from Proposition 4.4 to replace the condition a 0 0 by the condition a 0 0; a 0 6 0: This guarantees that the in mum over u in the minimax formula is achieved. Of course references should

SPECTRAL PROPERTIES OF THE LAPLACIAN ON BOUNDED DOMAINS

SPECTRAL PROPERTIES OF THE LAPLACIAN ON BOUNDED DOMAINS SPECTRAL PROPERTIES OF THE LAPLACIAN ON BOUNDED DOMAINS TSOGTGEREL GANTUMUR Abstract. After establishing discrete spectra for a large class of elliptic operators, we present some fundamental spectral properties

More information

Near convexity, metric convexity, and convexity

Near convexity, metric convexity, and convexity Near convexity, metric convexity, and convexity Fred Richman Florida Atlantic University Boca Raton, FL 33431 28 February 2005 Abstract It is shown that a subset of a uniformly convex normed space is nearly

More information

The Dirichlet s P rinciple. In this lecture we discuss an alternative formulation of the Dirichlet problem for the Laplace equation:

The Dirichlet s P rinciple. In this lecture we discuss an alternative formulation of the Dirichlet problem for the Laplace equation: Oct. 1 The Dirichlet s P rinciple In this lecture we discuss an alternative formulation of the Dirichlet problem for the Laplace equation: 1. Dirichlet s Principle. u = in, u = g on. ( 1 ) If we multiply

More information

A MINIMAX FORMULA FOR THE PRINCIPAL EIGENVALUES OF DIRICHLET PROBLEMS AND ITS APPLICATIONS

A MINIMAX FORMULA FOR THE PRINCIPAL EIGENVALUES OF DIRICHLET PROBLEMS AND ITS APPLICATIONS 2006 International Conference in Honor of Jacqueline Fleckinger. Electronic Journal of Differential Equations, Conference 15, 2007, pp. 137 154. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu

More information

UNIQUENESS OF POSITIVE SOLUTION TO SOME COUPLED COOPERATIVE VARIATIONAL ELLIPTIC SYSTEMS

UNIQUENESS OF POSITIVE SOLUTION TO SOME COUPLED COOPERATIVE VARIATIONAL ELLIPTIC SYSTEMS TRANSACTIONS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 00, Number 0, Pages 000 000 S 0002-9947(XX)0000-0 UNIQUENESS OF POSITIVE SOLUTION TO SOME COUPLED COOPERATIVE VARIATIONAL ELLIPTIC SYSTEMS YULIAN

More information

SYMMETRY RESULTS FOR PERTURBED PROBLEMS AND RELATED QUESTIONS. Massimo Grosi Filomena Pacella S. L. Yadava. 1. Introduction

SYMMETRY RESULTS FOR PERTURBED PROBLEMS AND RELATED QUESTIONS. Massimo Grosi Filomena Pacella S. L. Yadava. 1. Introduction Topological Methods in Nonlinear Analysis Journal of the Juliusz Schauder Center Volume 21, 2003, 211 226 SYMMETRY RESULTS FOR PERTURBED PROBLEMS AND RELATED QUESTIONS Massimo Grosi Filomena Pacella S.

More information

FENGBO HANG AND PAUL C. YANG

FENGBO HANG AND PAUL C. YANG Q CURVATURE ON A CLASS OF 3 ANIFOLDS FENGBO HANG AND PAUL C. YANG Abstract. otivated by the strong maximum principle for Paneitz operator in dimension 5 or higher found in [G] and the calculation of the

More information

Nonlinear Programming (NLP)

Nonlinear Programming (NLP) Natalia Lazzati Mathematics for Economics (Part I) Note 6: Nonlinear Programming - Unconstrained Optimization Note 6 is based on de la Fuente (2000, Ch. 7), Madden (1986, Ch. 3 and 5) and Simon and Blume

More information

Economics Bulletin, 2012, Vol. 32 No. 1 pp Introduction. 2. The preliminaries

Economics Bulletin, 2012, Vol. 32 No. 1 pp Introduction. 2. The preliminaries 1. Introduction In this paper we reconsider the problem of axiomatizing scoring rules. Early results on this problem are due to Smith (1973) and Young (1975). They characterized social welfare and social

More information

Optimal Boundary Control of a Nonlinear Di usion Equation y

Optimal Boundary Control of a Nonlinear Di usion Equation y AppliedMathematics E-Notes, (00), 97-03 c Availablefreeatmirrorsites ofhttp://math.math.nthu.edu.tw/»amen/ Optimal Boundary Control of a Nonlinear Di usion Equation y Jing-xueYin z,wen-meihuang x Received6

More information

Krein-Rutman Theorem and the Principal Eigenvalue

Krein-Rutman Theorem and the Principal Eigenvalue Chapter 1 Krein-Rutman Theorem and the Principal Eigenvalue The Krein-Rutman theorem plays a very important role in nonlinear partial differential equations, as it provides the abstract basis for the proof

More information

ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS

ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS Bendikov, A. and Saloff-Coste, L. Osaka J. Math. 4 (5), 677 7 ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS ALEXANDER BENDIKOV and LAURENT SALOFF-COSTE (Received March 4, 4)

More information

Mean-Variance Utility

Mean-Variance Utility Mean-Variance Utility Yutaka Nakamura University of Tsukuba Graduate School of Systems and Information Engineering Division of Social Systems and Management -- Tennnoudai, Tsukuba, Ibaraki 305-8573, Japan

More information

Global minimization. Chapter Upper and lower bounds

Global minimization. Chapter Upper and lower bounds Chapter 1 Global minimization The issues related to the behavior of global minimization problems along a sequence of functionals F are by now well understood, and mainly rely on the concept of -limit.

More information

Stochastic Processes

Stochastic Processes Stochastic Processes A very simple introduction Péter Medvegyev 2009, January Medvegyev (CEU) Stochastic Processes 2009, January 1 / 54 Summary from measure theory De nition (X, A) is a measurable space

More information

COHOMOLOGY AND DIFFERENTIAL SCHEMES. 1. Schemes

COHOMOLOGY AND DIFFERENTIAL SCHEMES. 1. Schemes COHOMOLOG AND DIFFERENTIAL SCHEMES RAMOND HOOBLER Dedicated to the memory of Jerrold Kovacic Abstract. Replace this text with your own abstract. 1. Schemes This section assembles basic results on schemes

More information

4.3 - Linear Combinations and Independence of Vectors

4.3 - Linear Combinations and Independence of Vectors - Linear Combinations and Independence of Vectors De nitions, Theorems, and Examples De nition 1 A vector v in a vector space V is called a linear combination of the vectors u 1, u,,u k in V if v can be

More information

NONTRIVIAL SOLUTIONS TO INTEGRAL AND DIFFERENTIAL EQUATIONS

NONTRIVIAL SOLUTIONS TO INTEGRAL AND DIFFERENTIAL EQUATIONS Fixed Point Theory, Volume 9, No. 1, 28, 3-16 http://www.math.ubbcluj.ro/ nodeacj/sfptcj.html NONTRIVIAL SOLUTIONS TO INTEGRAL AND DIFFERENTIAL EQUATIONS GIOVANNI ANELLO Department of Mathematics University

More information

EIGENVALUES AND EIGENVECTORS 3

EIGENVALUES AND EIGENVECTORS 3 EIGENVALUES AND EIGENVECTORS 3 1. Motivation 1.1. Diagonal matrices. Perhaps the simplest type of linear transformations are those whose matrix is diagonal (in some basis). Consider for example the matrices

More information

Math 413/513 Chapter 6 (from Friedberg, Insel, & Spence)

Math 413/513 Chapter 6 (from Friedberg, Insel, & Spence) Math 413/513 Chapter 6 (from Friedberg, Insel, & Spence) David Glickenstein December 7, 2015 1 Inner product spaces In this chapter, we will only consider the elds R and C. De nition 1 Let V be a vector

More information

A Perron-type theorem on the principal eigenvalue of nonsymmetric elliptic operators

A Perron-type theorem on the principal eigenvalue of nonsymmetric elliptic operators A Perron-type theorem on the principal eigenvalue of nonsymmetric elliptic operators Lei Ni And I cherish more than anything else the Analogies, my most trustworthy masters. They know all the secrets of

More information

NEW SIGNS OF ISOSCELES TRIANGLES

NEW SIGNS OF ISOSCELES TRIANGLES INTERNATIONAL JOURNAL OF GEOMETRY Vol. 2 (2013), No. 2, 56-67 NEW SIGNS OF ISOSCELES TRIANGLES MAKSIM VASKOUSKI and KANSTANTSIN KASTSEVICH Abstract. In this paper we prove new signs of isosceles triangles

More information

Some Properties of the Augmented Lagrangian in Cone Constrained Optimization

Some Properties of the Augmented Lagrangian in Cone Constrained Optimization MATHEMATICS OF OPERATIONS RESEARCH Vol. 29, No. 3, August 2004, pp. 479 491 issn 0364-765X eissn 1526-5471 04 2903 0479 informs doi 10.1287/moor.1040.0103 2004 INFORMS Some Properties of the Augmented

More information

8 Periodic Linear Di erential Equations - Floquet Theory

8 Periodic Linear Di erential Equations - Floquet Theory 8 Periodic Linear Di erential Equations - Floquet Theory The general theory of time varying linear di erential equations _x(t) = A(t)x(t) is still amazingly incomplete. Only for certain classes of functions

More information

Proof. We indicate by α, β (finite or not) the end-points of I and call

Proof. We indicate by α, β (finite or not) the end-points of I and call C.6 Continuous functions Pag. 111 Proof of Corollary 4.25 Corollary 4.25 Let f be continuous on the interval I and suppose it admits non-zero its (finite or infinite) that are different in sign for x tending

More information

Symmetry breaking for a problem in optimal insulation

Symmetry breaking for a problem in optimal insulation Symmetry breaking for a problem in optimal insulation Giuseppe Buttazzo Dipartimento di Matematica Università di Pisa buttazzo@dm.unipi.it http://cvgmt.sns.it Geometric and Analytic Inequalities Banff,

More information

Columbia University. Department of Economics Discussion Paper Series. The Knob of the Discord. Massimiliano Amarante Fabio Maccheroni

Columbia University. Department of Economics Discussion Paper Series. The Knob of the Discord. Massimiliano Amarante Fabio Maccheroni Columbia University Department of Economics Discussion Paper Series The Knob of the Discord Massimiliano Amarante Fabio Maccheroni Discussion Paper No.: 0405-14 Department of Economics Columbia University

More information

Measuring robustness

Measuring robustness Measuring robustness 1 Introduction While in the classical approach to statistics one aims at estimates which have desirable properties at an exactly speci ed model, the aim of robust methods is loosely

More information

It follows from the above inequalities that for c C 1

It follows from the above inequalities that for c C 1 3 Spaces L p 1. In this part we fix a measure space (, A, µ) (we may assume that µ is complete), and consider the A -measurable functions on it. 2. For f L 1 (, µ) set f 1 = f L 1 = f L 1 (,µ) = f dµ.

More information

Volume 30, Issue 3. Monotone comparative statics with separable objective functions. Christian Ewerhart University of Zurich

Volume 30, Issue 3. Monotone comparative statics with separable objective functions. Christian Ewerhart University of Zurich Volume 30, Issue 3 Monotone comparative statics with separable objective functions Christian Ewerhart University of Zurich Abstract The Milgrom-Shannon single crossing property is essential for monotone

More information

Chapter 1. GMM: Basic Concepts

Chapter 1. GMM: Basic Concepts Chapter 1. GMM: Basic Concepts Contents 1 Motivating Examples 1 1.1 Instrumental variable estimator....................... 1 1.2 Estimating parameters in monetary policy rules.............. 2 1.3 Estimating

More information

SHARP BOUNDARY TRACE INEQUALITIES. 1. Introduction

SHARP BOUNDARY TRACE INEQUALITIES. 1. Introduction SHARP BOUNDARY TRACE INEQUALITIES GILES AUCHMUTY Abstract. This paper describes sharp inequalities for the trace of Sobolev functions on the boundary of a bounded region R N. The inequalities bound (semi-)norms

More information

Eigenvalues and Eigenfunctions of the Laplacian

Eigenvalues and Eigenfunctions of the Laplacian The Waterloo Mathematics Review 23 Eigenvalues and Eigenfunctions of the Laplacian Mihai Nica University of Waterloo mcnica@uwaterloo.ca Abstract: The problem of determining the eigenvalues and eigenvectors

More information

Partial Di erential Equations

Partial Di erential Equations Partial Di erential Equations Alexander Grigorian Universität Bielefeld WS 205/6 2 Contents 0 Introduction 0. Examples of PDEs and their origin.................... 0.. Laplace equation..........................

More information

Second Order Elliptic PDE

Second Order Elliptic PDE Second Order Elliptic PDE T. Muthukumar tmk@iitk.ac.in December 16, 2014 Contents 1 A Quick Introduction to PDE 1 2 Classification of Second Order PDE 3 3 Linear Second Order Elliptic Operators 4 4 Periodic

More information

Topics in Mathematical Economics. Atsushi Kajii Kyoto University

Topics in Mathematical Economics. Atsushi Kajii Kyoto University Topics in Mathematical Economics Atsushi Kajii Kyoto University 25 November 2018 2 Contents 1 Preliminary Mathematics 5 1.1 Topology.................................. 5 1.2 Linear Algebra..............................

More information

EXISTENCE RESULTS FOR QUASILINEAR HEMIVARIATIONAL INEQUALITIES AT RESONANCE. Leszek Gasiński

EXISTENCE RESULTS FOR QUASILINEAR HEMIVARIATIONAL INEQUALITIES AT RESONANCE. Leszek Gasiński DISCRETE AND CONTINUOUS Website: www.aimsciences.org DYNAMICAL SYSTEMS SUPPLEMENT 2007 pp. 409 418 EXISTENCE RESULTS FOR QUASILINEAR HEMIVARIATIONAL INEQUALITIES AT RESONANCE Leszek Gasiński Jagiellonian

More information

1 Directional Derivatives and Differentiability

1 Directional Derivatives and Differentiability Wednesday, January 18, 2012 1 Directional Derivatives and Differentiability Let E R N, let f : E R and let x 0 E. Given a direction v R N, let L be the line through x 0 in the direction v, that is, L :=

More information

MAT 570 REAL ANALYSIS LECTURE NOTES. Contents. 1. Sets Functions Countability Axiom of choice Equivalence relations 9

MAT 570 REAL ANALYSIS LECTURE NOTES. Contents. 1. Sets Functions Countability Axiom of choice Equivalence relations 9 MAT 570 REAL ANALYSIS LECTURE NOTES PROFESSOR: JOHN QUIGG SEMESTER: FALL 204 Contents. Sets 2 2. Functions 5 3. Countability 7 4. Axiom of choice 8 5. Equivalence relations 9 6. Real numbers 9 7. Extended

More information

Notes on Mathematical Expectations and Classes of Distributions Introduction to Econometric Theory Econ. 770

Notes on Mathematical Expectations and Classes of Distributions Introduction to Econometric Theory Econ. 770 Notes on Mathematical Expectations and Classes of Distributions Introduction to Econometric Theory Econ. 77 Jonathan B. Hill Dept. of Economics University of North Carolina - Chapel Hill October 4, 2 MATHEMATICAL

More information

Discreteness of Transmission Eigenvalues via Upper Triangular Compact Operators

Discreteness of Transmission Eigenvalues via Upper Triangular Compact Operators Discreteness of Transmission Eigenvalues via Upper Triangular Compact Operators John Sylvester Department of Mathematics University of Washington Seattle, Washington 98195 U.S.A. June 3, 2011 This research

More information

Some SDEs with distributional drift Part I : General calculus. Flandoli, Franco; Russo, Francesco; Wolf, Jochen

Some SDEs with distributional drift Part I : General calculus. Flandoli, Franco; Russo, Francesco; Wolf, Jochen Title Author(s) Some SDEs with distributional drift Part I : General calculus Flandoli, Franco; Russo, Francesco; Wolf, Jochen Citation Osaka Journal of Mathematics. 4() P.493-P.54 Issue Date 3-6 Text

More information

Topics in Mathematical Economics. Atsushi Kajii Kyoto University

Topics in Mathematical Economics. Atsushi Kajii Kyoto University Topics in Mathematical Economics Atsushi Kajii Kyoto University 26 June 2018 2 Contents 1 Preliminary Mathematics 5 1.1 Topology.................................. 5 1.2 Linear Algebra..............................

More information

Nonlinear Analysis 71 (2009) Contents lists available at ScienceDirect. Nonlinear Analysis. journal homepage:

Nonlinear Analysis 71 (2009) Contents lists available at ScienceDirect. Nonlinear Analysis. journal homepage: Nonlinear Analysis 71 2009 2744 2752 Contents lists available at ScienceDirect Nonlinear Analysis journal homepage: www.elsevier.com/locate/na A nonlinear inequality and applications N.S. Hoang A.G. Ramm

More information

A Necessary and Sufficient Condition for a Unique Maximum with an Application to Potential Games

A Necessary and Sufficient Condition for a Unique Maximum with an Application to Potential Games Towson University Department of Economics Working Paper Series Working Paper No. 2017-04 A Necessary and Sufficient Condition for a Unique Maximum with an Application to Potential Games by Finn Christensen

More information

Maths 212: Homework Solutions

Maths 212: Homework Solutions Maths 212: Homework Solutions 1. The definition of A ensures that x π for all x A, so π is an upper bound of A. To show it is the least upper bound, suppose x < π and consider two cases. If x < 1, then

More information

Stochastic integral. Introduction. Ito integral. References. Appendices Stochastic Calculus I. Geneviève Gauthier.

Stochastic integral. Introduction. Ito integral. References. Appendices Stochastic Calculus I. Geneviève Gauthier. Ito 8-646-8 Calculus I Geneviève Gauthier HEC Montréal Riemann Ito The Ito The theories of stochastic and stochastic di erential equations have initially been developed by Kiyosi Ito around 194 (one of

More information

Convexity in R n. The following lemma will be needed in a while. Lemma 1 Let x E, u R n. If τ I(x, u), τ 0, define. f(x + τu) f(x). τ.

Convexity in R n. The following lemma will be needed in a while. Lemma 1 Let x E, u R n. If τ I(x, u), τ 0, define. f(x + τu) f(x). τ. Convexity in R n Let E be a convex subset of R n. A function f : E (, ] is convex iff f(tx + (1 t)y) (1 t)f(x) + tf(y) x, y E, t [0, 1]. A similar definition holds in any vector space. A topology is needed

More information

Lecture Notes on PDEs

Lecture Notes on PDEs Lecture Notes on PDEs Alberto Bressan February 26, 2012 1 Elliptic equations Let IR n be a bounded open set Given measurable functions a ij, b i, c : IR, consider the linear, second order differential

More information

SYMMETRY OF POSITIVE SOLUTIONS OF SOME NONLINEAR EQUATIONS. M. Grossi S. Kesavan F. Pacella M. Ramaswamy. 1. Introduction

SYMMETRY OF POSITIVE SOLUTIONS OF SOME NONLINEAR EQUATIONS. M. Grossi S. Kesavan F. Pacella M. Ramaswamy. 1. Introduction Topological Methods in Nonlinear Analysis Journal of the Juliusz Schauder Center Volume 12, 1998, 47 59 SYMMETRY OF POSITIVE SOLUTIONS OF SOME NONLINEAR EQUATIONS M. Grossi S. Kesavan F. Pacella M. Ramaswamy

More information

It follows from the above inequalities that for c C 1

It follows from the above inequalities that for c C 1 3 Spaces L p 1. Appearance of normed spaces. In this part we fix a measure space (, A, µ) (we may assume that µ is complete), and consider the A - measurable functions on it. 2. For f L 1 (, µ) set f 1

More information

ON NONHOMOGENEOUS BIHARMONIC EQUATIONS INVOLVING CRITICAL SOBOLEV EXPONENT

ON NONHOMOGENEOUS BIHARMONIC EQUATIONS INVOLVING CRITICAL SOBOLEV EXPONENT PORTUGALIAE MATHEMATICA Vol. 56 Fasc. 3 1999 ON NONHOMOGENEOUS BIHARMONIC EQUATIONS INVOLVING CRITICAL SOBOLEV EXPONENT M. Guedda Abstract: In this paper we consider the problem u = λ u u + f in, u = u

More information

GALOIS THEORY I (Supplement to Chapter 4)

GALOIS THEORY I (Supplement to Chapter 4) GALOIS THEORY I (Supplement to Chapter 4) 1 Automorphisms of Fields Lemma 1 Let F be a eld. The set of automorphisms of F; Aut (F ) ; forms a group (under composition of functions). De nition 2 Let F be

More information

Positive Political Theory II David Austen-Smith & Je rey S. Banks

Positive Political Theory II David Austen-Smith & Je rey S. Banks Positive Political Theory II David Austen-Smith & Je rey S. Banks Egregious Errata Positive Political Theory II (University of Michigan Press, 2005) regrettably contains a variety of obscurities and errors,

More information

Congurations of periodic orbits for equations with delayed positive feedback

Congurations of periodic orbits for equations with delayed positive feedback Congurations of periodic orbits for equations with delayed positive feedback Dedicated to Professor Tibor Krisztin on the occasion of his 60th birthday Gabriella Vas 1 MTA-SZTE Analysis and Stochastics

More information

NONLINEAR FREDHOLM ALTERNATIVE FOR THE p-laplacian UNDER NONHOMOGENEOUS NEUMANN BOUNDARY CONDITION

NONLINEAR FREDHOLM ALTERNATIVE FOR THE p-laplacian UNDER NONHOMOGENEOUS NEUMANN BOUNDARY CONDITION Electronic Journal of Differential Equations, Vol. 2016 (2016), No. 210, pp. 1 7. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu NONLINEAR FREDHOLM ALTERNATIVE FOR THE p-laplacian

More information

Scattering for the NLS equation

Scattering for the NLS equation Scattering for the NLS equation joint work with Thierry Cazenave (UPMC) Ivan Naumkin Université Nice Sophia Antipolis February 2, 2017 Introduction. Consider the nonlinear Schrödinger equation with the

More information

SPRING 2006 PRELIMINARY EXAMINATION SOLUTIONS

SPRING 2006 PRELIMINARY EXAMINATION SOLUTIONS SPRING 006 PRELIMINARY EXAMINATION SOLUTIONS 1A. Let G be the subgroup of the free abelian group Z 4 consisting of all integer vectors (x, y, z, w) such that x + 3y + 5z + 7w = 0. (a) Determine a linearly

More information

Time Series Models and Inference. James L. Powell Department of Economics University of California, Berkeley

Time Series Models and Inference. James L. Powell Department of Economics University of California, Berkeley Time Series Models and Inference James L. Powell Department of Economics University of California, Berkeley Overview In contrast to the classical linear regression model, in which the components of the

More information

Introduction to Real Analysis Alternative Chapter 1

Introduction to Real Analysis Alternative Chapter 1 Christopher Heil Introduction to Real Analysis Alternative Chapter 1 A Primer on Norms and Banach Spaces Last Updated: March 10, 2018 c 2018 by Christopher Heil Chapter 1 A Primer on Norms and Banach Spaces

More information

Lecture 8: Basic convex analysis

Lecture 8: Basic convex analysis Lecture 8: Basic convex analysis 1 Convex sets Both convex sets and functions have general importance in economic theory, not only in optimization. Given two points x; y 2 R n and 2 [0; 1]; the weighted

More information

A CONVEX-CONCAVE ELLIPTIC PROBLEM WITH A PARAMETER ON THE BOUNDARY CONDITION

A CONVEX-CONCAVE ELLIPTIC PROBLEM WITH A PARAMETER ON THE BOUNDARY CONDITION A CONVEX-CONCAVE ELLIPTIC PROBLEM WITH A PARAMETER ON THE BOUNDARY CONDITION JORGE GARCÍA-MELIÁN, JULIO D. ROSSI AND JOSÉ C. SABINA DE LIS Abstract. In this paper we study existence and multiplicity of

More information

The harmonic map flow

The harmonic map flow Chapter 2 The harmonic map flow 2.1 Definition of the flow The harmonic map flow was introduced by Eells-Sampson in 1964; their work could be considered the start of the field of geometric flows. The flow

More information

Maximum Principles and Principal Eigenvalues

Maximum Principles and Principal Eigenvalues Ten Mathematical Essays on Approximation in Analysis and Topology 1 J Ferrera J López-Gómez F R Ruíz del Portal Editors c 004 Elsevier BV All rights reserved Maximum Principles and Principal Eigenvalues

More information

Robust Estimation and Inference for Extremal Dependence in Time Series. Appendix C: Omitted Proofs and Supporting Lemmata

Robust Estimation and Inference for Extremal Dependence in Time Series. Appendix C: Omitted Proofs and Supporting Lemmata Robust Estimation and Inference for Extremal Dependence in Time Series Appendix C: Omitted Proofs and Supporting Lemmata Jonathan B. Hill Dept. of Economics University of North Carolina - Chapel Hill January

More information

Functional Analysis. Franck Sueur Metric spaces Definitions Completeness Compactness Separability...

Functional Analysis. Franck Sueur Metric spaces Definitions Completeness Compactness Separability... Functional Analysis Franck Sueur 2018-2019 Contents 1 Metric spaces 1 1.1 Definitions........................................ 1 1.2 Completeness...................................... 3 1.3 Compactness......................................

More information

Max-Min Problems in R n Matrix

Max-Min Problems in R n Matrix Max-Min Problems in R n Matrix 1 and the essian Prerequisite: Section 6., Orthogonal Diagonalization n this section, we study the problem of nding local maxima and minima for realvalued functions on R

More information

SPECTRAL PROPERTIES AND NODAL SOLUTIONS FOR SECOND-ORDER, m-point, BOUNDARY VALUE PROBLEMS

SPECTRAL PROPERTIES AND NODAL SOLUTIONS FOR SECOND-ORDER, m-point, BOUNDARY VALUE PROBLEMS SPECTRAL PROPERTIES AND NODAL SOLUTIONS FOR SECOND-ORDER, m-point, BOUNDARY VALUE PROBLEMS BRYAN P. RYNNE Abstract. We consider the m-point boundary value problem consisting of the equation u = f(u), on

More information

Contents: 1. Minimization. 2. The theorem of Lions-Stampacchia for variational inequalities. 3. Γ -Convergence. 4. Duality mapping.

Contents: 1. Minimization. 2. The theorem of Lions-Stampacchia for variational inequalities. 3. Γ -Convergence. 4. Duality mapping. Minimization Contents: 1. Minimization. 2. The theorem of Lions-Stampacchia for variational inequalities. 3. Γ -Convergence. 4. Duality mapping. 1 Minimization A Topological Result. Let S be a topological

More information

An introduction to Mathematical Theory of Control

An introduction to Mathematical Theory of Control An introduction to Mathematical Theory of Control Vasile Staicu University of Aveiro UNICA, May 2018 Vasile Staicu (University of Aveiro) An introduction to Mathematical Theory of Control UNICA, May 2018

More information

Appendix A. Sequences and series. A.1 Sequences. Definition A.1 A sequence is a function N R.

Appendix A. Sequences and series. A.1 Sequences. Definition A.1 A sequence is a function N R. Appendix A Sequences and series This course has for prerequisite a course (or two) of calculus. The purpose of this appendix is to review basic definitions and facts concerning sequences and series, which

More information

1 Which sets have volume 0?

1 Which sets have volume 0? Math 540 Spring 0 Notes #0 More on integration Which sets have volume 0? The theorem at the end of the last section makes this an important question. (Measure theory would supersede it, however.) Theorem

More information

LECTURE 12 UNIT ROOT, WEAK CONVERGENCE, FUNCTIONAL CLT

LECTURE 12 UNIT ROOT, WEAK CONVERGENCE, FUNCTIONAL CLT MARCH 29, 26 LECTURE 2 UNIT ROOT, WEAK CONVERGENCE, FUNCTIONAL CLT (Davidson (2), Chapter 4; Phillips Lectures on Unit Roots, Cointegration and Nonstationarity; White (999), Chapter 7) Unit root processes

More information

THE STOKES SYSTEM R.E. SHOWALTER

THE STOKES SYSTEM R.E. SHOWALTER THE STOKES SYSTEM R.E. SHOWALTER Contents 1. Stokes System 1 Stokes System 2 2. The Weak Solution of the Stokes System 3 3. The Strong Solution 4 4. The Normal Trace 6 5. The Mixed Problem 7 6. The Navier-Stokes

More information

Time is discrete and indexed by t =0; 1;:::;T,whereT<1. An individual is interested in maximizing an objective function given by. tu(x t ;a t ); (0.

Time is discrete and indexed by t =0; 1;:::;T,whereT<1. An individual is interested in maximizing an objective function given by. tu(x t ;a t ); (0. Chapter 0 Discrete Time Dynamic Programming 0.1 The Finite Horizon Case Time is discrete and indexed by t =0; 1;:::;T,whereT

More information

CHAPTER 7. Connectedness

CHAPTER 7. Connectedness CHAPTER 7 Connectedness 7.1. Connected topological spaces Definition 7.1. A topological space (X, T X ) is said to be connected if there is no continuous surjection f : X {0, 1} where the two point set

More information

Stanford Mathematics Department Math 205A Lecture Supplement #4 Borel Regular & Radon Measures

Stanford Mathematics Department Math 205A Lecture Supplement #4 Borel Regular & Radon Measures 2 1 Borel Regular Measures We now state and prove an important regularity property of Borel regular outer measures: Stanford Mathematics Department Math 205A Lecture Supplement #4 Borel Regular & Radon

More information

Revisiting independence and stochastic dominance for compound lotteries

Revisiting independence and stochastic dominance for compound lotteries Revisiting independence and stochastic dominance for compound lotteries Alexander Zimper Working Paper Number 97 Department of Economics and Econometrics, University of Johannesburg Revisiting independence

More information

SOME MEASURABILITY AND CONTINUITY PROPERTIES OF ARBITRARY REAL FUNCTIONS

SOME MEASURABILITY AND CONTINUITY PROPERTIES OF ARBITRARY REAL FUNCTIONS LE MATEMATICHE Vol. LVII (2002) Fasc. I, pp. 6382 SOME MEASURABILITY AND CONTINUITY PROPERTIES OF ARBITRARY REAL FUNCTIONS VITTORINO PATA - ALFONSO VILLANI Given an arbitrary real function f, the set D

More information

Alvaro Rodrigues-Neto Research School of Economics, Australian National University. ANU Working Papers in Economics and Econometrics # 587

Alvaro Rodrigues-Neto Research School of Economics, Australian National University. ANU Working Papers in Economics and Econometrics # 587 Cycles of length two in monotonic models José Alvaro Rodrigues-Neto Research School of Economics, Australian National University ANU Working Papers in Economics and Econometrics # 587 October 20122 JEL:

More information

A NOTE ON THE EXISTENCE OF TWO NONTRIVIAL SOLUTIONS OF A RESONANCE PROBLEM

A NOTE ON THE EXISTENCE OF TWO NONTRIVIAL SOLUTIONS OF A RESONANCE PROBLEM PORTUGALIAE MATHEMATICA Vol. 51 Fasc. 4 1994 A NOTE ON THE EXISTENCE OF TWO NONTRIVIAL SOLUTIONS OF A RESONANCE PROBLEM To Fu Ma* Abstract: We study the existence of two nontrivial solutions for an elliptic

More information

Mixed exterior Laplace s problem

Mixed exterior Laplace s problem Mixed exterior Laplace s problem Chérif Amrouche, Florian Bonzom Laboratoire de mathématiques appliquées, CNRS UMR 5142, Université de Pau et des Pays de l Adour, IPRA, Avenue de l Université, 64000 Pau

More information

The Kuhn-Tucker Problem

The Kuhn-Tucker Problem Natalia Lazzati Mathematics for Economics (Part I) Note 8: Nonlinear Programming - The Kuhn-Tucker Problem Note 8 is based on de la Fuente (2000, Ch. 7) and Simon and Blume (1994, Ch. 18 and 19). The Kuhn-Tucker

More information

MATH 220: INNER PRODUCT SPACES, SYMMETRIC OPERATORS, ORTHOGONALITY

MATH 220: INNER PRODUCT SPACES, SYMMETRIC OPERATORS, ORTHOGONALITY MATH 22: INNER PRODUCT SPACES, SYMMETRIC OPERATORS, ORTHOGONALITY When discussing separation of variables, we noted that at the last step we need to express the inhomogeneous initial or boundary data as

More information

Near-Potential Games: Geometry and Dynamics

Near-Potential Games: Geometry and Dynamics Near-Potential Games: Geometry and Dynamics Ozan Candogan, Asuman Ozdaglar and Pablo A. Parrilo September 6, 2011 Abstract Potential games are a special class of games for which many adaptive user dynamics

More information

REGULAR LAGRANGE MULTIPLIERS FOR CONTROL PROBLEMS WITH MIXED POINTWISE CONTROL-STATE CONSTRAINTS

REGULAR LAGRANGE MULTIPLIERS FOR CONTROL PROBLEMS WITH MIXED POINTWISE CONTROL-STATE CONSTRAINTS REGULAR LAGRANGE MULTIPLIERS FOR CONTROL PROBLEMS WITH MIXED POINTWISE CONTROL-STATE CONSTRAINTS fredi tröltzsch 1 Abstract. A class of quadratic optimization problems in Hilbert spaces is considered,

More information

Theory of PDE Homework 2

Theory of PDE Homework 2 Theory of PDE Homework 2 Adrienne Sands April 18, 2017 In the following exercises we assume the coefficients of the various PDE are smooth and satisfy the uniform ellipticity condition. R n is always an

More information

Convex Functions and Optimization

Convex Functions and Optimization Chapter 5 Convex Functions and Optimization 5.1 Convex Functions Our next topic is that of convex functions. Again, we will concentrate on the context of a map f : R n R although the situation can be generalized

More information

Threshold behavior and non-quasiconvergent solutions with localized initial data for bistable reaction-diffusion equations

Threshold behavior and non-quasiconvergent solutions with localized initial data for bistable reaction-diffusion equations Threshold behavior and non-quasiconvergent solutions with localized initial data for bistable reaction-diffusion equations P. Poláčik School of Mathematics, University of Minnesota Minneapolis, MN 55455

More information

1 Lyapunov theory of stability

1 Lyapunov theory of stability M.Kawski, APM 581 Diff Equns Intro to Lyapunov theory. November 15, 29 1 1 Lyapunov theory of stability Introduction. Lyapunov s second (or direct) method provides tools for studying (asymptotic) stability

More information

Notes on Integrable Functions and the Riesz Representation Theorem Math 8445, Winter 06, Professor J. Segert. f(x) = f + (x) + f (x).

Notes on Integrable Functions and the Riesz Representation Theorem Math 8445, Winter 06, Professor J. Segert. f(x) = f + (x) + f (x). References: Notes on Integrable Functions and the Riesz Representation Theorem Math 8445, Winter 06, Professor J. Segert Evans, Partial Differential Equations, Appendix 3 Reed and Simon, Functional Analysis,

More information

ATTRACTORS FOR SEMILINEAR PARABOLIC PROBLEMS WITH DIRICHLET BOUNDARY CONDITIONS IN VARYING DOMAINS. Emerson A. M. de Abreu Alexandre N.

ATTRACTORS FOR SEMILINEAR PARABOLIC PROBLEMS WITH DIRICHLET BOUNDARY CONDITIONS IN VARYING DOMAINS. Emerson A. M. de Abreu Alexandre N. ATTRACTORS FOR SEMILINEAR PARABOLIC PROBLEMS WITH DIRICHLET BOUNDARY CONDITIONS IN VARYING DOMAINS Emerson A. M. de Abreu Alexandre N. Carvalho Abstract Under fairly general conditions one can prove that

More information

Advanced Economic Growth: Lecture 21: Stochastic Dynamic Programming and Applications

Advanced Economic Growth: Lecture 21: Stochastic Dynamic Programming and Applications Advanced Economic Growth: Lecture 21: Stochastic Dynamic Programming and Applications Daron Acemoglu MIT November 19, 2007 Daron Acemoglu (MIT) Advanced Growth Lecture 21 November 19, 2007 1 / 79 Stochastic

More information

Nonlinear eigenvalue{eigenvector problems for STP matrices

Nonlinear eigenvalue{eigenvector problems for STP matrices Proceedings of the Royal Society of Edinburgh, 132A, 1307{1331, 2002 Nonlinear eigenvalue{eigenvector problems for STP matrices Uri Elias and Allan Pinkus Department of Mathematics, Technion, IIT, Haifa

More information

Local disaggregation of demand and excess demand functions: a new question

Local disaggregation of demand and excess demand functions: a new question Local disaggregation of demand and excess demand functions: a new question Pierre-Andre Chiappori Ivar Ekeland y Martin Browning z January 1999 Abstract The literature on the characterization of aggregate

More information

g 2 (x) (1/3)M 1 = (1/3)(2/3)M.

g 2 (x) (1/3)M 1 = (1/3)(2/3)M. COMPACTNESS If C R n is closed and bounded, then by B-W it is sequentially compact: any sequence of points in C has a subsequence converging to a point in C Conversely, any sequentially compact C R n is

More information

T. Godoy - E. Lami Dozo - S. Paczka ON THE ANTIMAXIMUM PRINCIPLE FOR PARABOLIC PERIODIC PROBLEMS WITH WEIGHT

T. Godoy - E. Lami Dozo - S. Paczka ON THE ANTIMAXIMUM PRINCIPLE FOR PARABOLIC PERIODIC PROBLEMS WITH WEIGHT Rend. Sem. Mat. Univ. Pol. Torino Vol. 1, 60 2002 T. Godoy - E. Lami Dozo - S. Paczka ON THE ANTIMAXIMUM PRINCIPLE FOR PARABOLIC PERIODIC PROBLEMS WITH WEIGHT Abstract. We prove that an antimaximum principle

More information

Simple Estimators for Monotone Index Models

Simple Estimators for Monotone Index Models Simple Estimators for Monotone Index Models Hyungtaik Ahn Dongguk University, Hidehiko Ichimura University College London, James L. Powell University of California, Berkeley (powell@econ.berkeley.edu)

More information

Math 350 Fall 2011 Notes about inner product spaces. In this notes we state and prove some important properties of inner product spaces.

Math 350 Fall 2011 Notes about inner product spaces. In this notes we state and prove some important properties of inner product spaces. Math 350 Fall 2011 Notes about inner product spaces In this notes we state and prove some important properties of inner product spaces. First, recall the dot product on R n : if x, y R n, say x = (x 1,...,

More information

is a weak solution with the a ij,b i,c2 C 1 ( )

is a weak solution with the a ij,b i,c2 C 1 ( ) Thus @u @x i PDE 69 is a weak solution with the RHS @f @x i L. Thus u W 3, loc (). Iterating further, and using a generalized Sobolev imbedding gives that u is smooth. Theorem 3.33 (Local smoothness).

More information