Modular eigenvalues of the variable exponent p-Laplacian
Transcrição
Modular eigenvalues of the variable exponent p-Laplacian
Modular eigenvalues of the variable exponent p-Laplacian Joint Meeting AMS-EMS-SPMM Porto, June 2015 Osvaldo Mendez University of Texas-El Paso Table of contents Section 1 Modular Spaces Section 2 Sobolev Spaces, Poincares inequality, Sobolev Embedding. Section 3 Euler-Lagrange equations Section 5 Existence of the rst eigenvalue Section no. 6 Stability Section 7 Remark Section 7 Open problems , D. Edmunds, J. Lang, O.M. Dierential Operators on Spaces of Variable Integrability , D. Edmunds, J. Lang, O.M. Dierential Operators on Spaces of Variable Integrability J. Lang, O. Mendez , D. Edmunds, J. Lang, O.M. Dierential Operators on Spaces of Variable Integrability J. Lang, O. Mendez , , Nonlinear An., 2014 Journal of Di. Eq, 2015 Journal d'Analyse Mathematique, 2015 Preliminaries Rn bounded, Lipschitz domain, Borel-measurable. p: ! (1; 1) Preliminaries Rn bounded, Lipschitz domain, Borel-measurable. p: ! (1; 1) Z p(); (f ) = jf (x )jp(x ) dx is a left-continuous convex modular on M( ); M( ) being the set of all Lebesgue-measurable, scalar-valued functions on ; f = g i f = g a.e.. Preliminaries Rn bounded, Lipschitz domain, Borel-measurable. p: ! (1; 1) Z p(); (f ) = jf (x )jp(x ) dx is a left-continuous convex modular on M( ); M( ) being the set of all Lebesgue-measurable, scalar-valued functions on ; f = g i f = g a.e.. kf kp(); := inf > 0 : p(); (f =) 1 is a norm on the space Lp() ( ) = f 2 M( ) : p(); (f ) < 1 for some > 0 : Particular case of Musielak-Orlicz spaces. Sobolev Spaces, Poincares inequality, Sobolev Embedding. Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) o Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). If p 2 C ( ) the embedding is compact E : W01;p() ( ) ,! Lp() ( ) Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). If p 2 C ( ) the embedding E : W01;p() ( ) ,! Lp() ( ) is compact and Poincares inequality holds: Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). If p 2 C ( ) the embedding E : W01;p() ( ) ,! Lp() ( ) is compact and Poincares inequality holds: kukp() C (p; n; )kjrujkp(): Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). If p 2 C ( ) the embedding E : W01;p() ( ) ,! Lp() ( ) is compact and Poincares inequality holds: kukp() C (p; n; )kjrujkp(): Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). If p 2 C ( ) the embedding E : W01;p() ( ) ,! Lp() ( ) is compact and Poincares inequality holds: kukp() C (p; n; )kjrujkp(): kuk = kjrujkp() Sobolev Spaces, Poincares inequality, Sobolev Embedding. n W 1;p() ( ) = u 2 Lp() ( ) : jru j 2 Lp() ( ) kukW 1 ( ) ( ) ;p o = kuk1;p() = kukp(); + kjrujkp(); : W01;p() ( ) is the closure of C01( ) in W 1;p()( ). If p 2 C ( ) the embedding E : W01;p() ( ) ,! Lp() ( ) is compact and Poincares inequality holds: kukp() C (p; n; )kjrujkp(): kuk = kjrujkp() is a norm on W01;p()( ) equivalent to the original one. The THERMISTOR PROBLEM (Zhikov, 1996) p()u := div jrujp(x ) 2ru = 0 The THERMISTOR PROBLEM (Zhikov, 1996) p()u := div jrujp(x ) 2ru = 0 Existence, uniqueness and regularity of the solution!!! The THERMISTOR PROBLEM (Zhikov, 1996) p()u := div jrujp(x ) 2ru = 0 Existence, uniqueness and regularity of the solution!!! Fan-Zhang, 2003 The THERMISTOR PROBLEM (Zhikov, 1996) p()u := div jrujp(x ) 2ru = 0 Existence, uniqueness and regularity of the solution!!! Fan-Zhang, 2003 Modeling of Non-Newtonian uid hydrodynamics, Electrorheological uids. The THERMISTOR PROBLEM (Zhikov, 1996) p()u := div jrujp(x ) 2ru = 0 Existence, uniqueness and regularity of the solution!!! Fan-Zhang, 2003 Modeling of Non-Newtonian uid hydrodynamics, Electrorheological uids. Electrorheological uids: Modeling and Mathematical Theory. R_ucika Modular Eigenvalue problem p()u = jujp() 2u (1) Modular Eigenvalue problem p()u = jujp() (Fan-Zhang-Zhao, 2005) 2u (1) Modular Eigenvalue problem p()u = jujp() (Fan-Zhang-Zhao, 2005) Z 8h 2 C01( ): Z 2u jrujp() 2rurh = jujp() 2uh (1) (2) Modular Eigenvalue problem p()u = jujp() (Fan-Zhang-Zhao, 2005) Z 8h 2 C01( ): 2u Z jrujp() 2rurh = jujp() 2uh If p is constant ! rst positive eigenvalue: 1 : = kE kp (1) (2) Modular Eigenvalue problem p()u = jujp() (Fan-Zhang-Zhao, 2005) Z 8h 2 C01( ): 2u Z jrujp() 2rurh = jujp() 2uh If p is constant ! rst positive eigenvalue: 1 : = kE kp If p has either a local min or a local max in .......... (1) (2) Modular Eigenvalue problem p()u = jujp() (Fan-Zhang-Zhao, 2005) Z 8h 2 C01( ): 2u Z jrujp() 2rurh = jujp() 2uh If p is constant ! rst positive eigenvalue: 1 : = kE kp If p has either a local min or a local max in .......... R p (x ) inf1 ( ) R jrjuu(x(x)j)pj(x ) = 0!!!!: 06=u2W0 ( ) (Fhang-Zhang-Zhao (2005).) ;p (1) (2) The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? Abstract theory of Ljusternik-Schnirelman: The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? Abstract theory of Ljusternik-Schnirelman: k = 1; 2; 3::: R k jrujp() : R = Cinf sup 2C v 2C ju jp() k The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? Abstract theory of Ljusternik-Schnirelman: k = 1; 2; 3::: R k jrujp() : R = Cinf sup 2C v 2C ju jp() k Ck =: fC W01;p()( ) ; C closed ; (C ) k g: The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? Abstract theory of Ljusternik-Schnirelman: k = 1; 2; 3::: R k jrujp() : R = Cinf sup 2C v 2C ju jp() k Ck =: fC W01;p()( ) ; C closed ; (C ) k g: : Krasnoselskii genus: The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? Abstract theory of Ljusternik-Schnirelman: k = 1; 2; 3::: R k jrujp() : R = Cinf sup 2C v 2C ju jp() k Ck =: fC W01;p()( ) ; C closed ; (C ) k g: : Krasnoselskii genus: Not a tangible description of the eigenfunctions. The modular eigenvalue problem as Euler-Lagrange Equations Does the modular eigenvalue problem have a solution (; u)? Abstract theory of Ljusternik-Schnirelman: k = 1; 2; 3::: R k jrujp() : R = Cinf sup 2C v 2C ju jp() k Ck =: fC W01;p()( ) ; C closed ; (C ) k g: : Krasnoselskii genus: Not a tangible description of the eigenfunctions. There are eigenvalues "per se" that are NOT obtained via Ljiusternik-Schnirelmann. (Bining-Rynne(2008)) Opt for an ad-hoc treatment (Sobolev embedding) F : W01;p() ( ) ! R R () F (u ) = jrup((xx))j dx p x Opt for an ad-hoc treatment (Sobolev embedding) F : W01;p() ( ) ! R R () F (u ) = jrup((xx))j dx p x G : Lp() ( ) ! R R () G (u ) = ju(px()xj ) dx p x Then hF (u); hi = 0 and Z hG (u); hi = 0 jru(x )jp(x ) 2ru(x )rh(x ) dx Z ju(x )jp(x ) 2u(x )h(x ) dx (3) (4) The modular eigenvalue problem is the Euler-Lagrange equation for the constrained problem max G (u) subject to F (u) = constant (5) Existence of the rst eigenvalue Existence of the rst eigenvalue r > 0 ) 9(u0 ; 0 ) solution of modular eigenvalue problem: Existence of the rst eigenvalue r > 0 ) 9(u0 ; 0 ) Z solution of modular eigenvalue problem: ju0(x )jp(x ) dx = p (x ) R jru sup( ) ( )j () x p x p x Z dx r ju(x )jp(x ) dx : p (x ) Existence of the rst eigenvalue r > 0 ) 9(u0 ; 0 ) Z solution of modular eigenvalue problem: ju0(x )jp(x ) dx = p (x ) R Z jru sup( ) ( )j () x p x p x Z dx r jru0(x )jp(x ) dx = r p (x ) ju(x )jp(x ) dx : p (x ) Stability ( p ; u p ) (q ; uq ) eigenvalues corresponding to p; q; Stability ( p ; u p ) (q ; uq ) eigenvalues corresponding to p; q; ) pq Stability ( p ; u p ) (q ; uq ) eigenvalues corresponding to p; q; ) pq p q ?? Stability ( p ; u p ) (q ; uq ) eigenvalues corresponding to p; q; ) pq p q ?? up uq ?? (Lindqvist (1990), p constant.) Holder's inequality: Holder's inequality: Z jf (x )g (x )j dx (1 + 1=p 1=p+) kf kp kg kp : 0 Holder's inequality: Z jf (x )g (x )j dx (1 + 1=p Not suitable for "stability" analysis. 1=p+) kf kp kg kp : 0 Holder's inequality: Z jf (x )g (x )j dx (1 + 1=p 1=p+) kf kp kg kp : Not suitable for "stability" analysis. However: If p (x ) q (x ) p (x ) + " a.e. in : 0 (6) Holder's inequality: Z jf (x )g (x )j dx (1 + 1=p 1=p+) kf kp kg kp : Not suitable for "stability" analysis. However: If p (x ) q (x ) p (x ) + " a.e. in : q (f ) 1; then p (f ) " j j + " " : 0 (6) Holder's inequality: Z jf (x )g (x )j dx (1 + 1=p 1=p+) kf kp kg kp : Not suitable for "stability" analysis. However: If p (x ) q (x ) p (x ) + " a.e. in : q (f ) 1; then p (f ) " j j + " " : Edmunds-Lang-Nekvinda, 2009 0 (6) Holder's inequality: Z jf (x )g (x )j dx (1 + 1=p 1=p+) kf kp kg kp : 0 Not suitable for "stability" analysis. However: If p (x ) q (x ) p (x ) + " a.e. in : q (f ) 1; then p (f ) " j j + " " : Edmunds-Lang-Nekvinda, 2009 This is the substitute of Holder's inequality for the stability analysis. (6) (pi )i C ( ) pi increases uniformly to q inf q > n; (pi )i C ( ) pi increases uniformly to q inf q > n; Then, there exists a solution to the q-eigenvalue problem, such that and (q ; uq ) up * uq i in W01;q ( ) (pi )i C ( ) pi increases uniformly to q inf q > n; Then, there exists a solution to the q-eigenvalue problem, such that and Moreover: (q ; uq ) up * uq i up i in W01;q ( ) ! uq in Lq ( ) (pi )i C ( ) pi increases uniformly to q inf q > n; Then, there exists a solution to the q-eigenvalue problem, such that and Moreover: and (q ; uq ) up * uq i up i in W01;q ( ) ! uq in Lq ( ) R jrujq dx q L R q q+ ju j dx L; q 2 L q;1 L: q+ (7) (pi )i C ( ) pi increases uniformly to q inf q > n; Then, there exists a solution to the q-eigenvalue problem, such that and Moreover: and (q ; uq ) up * uq i up i in W01;q ( ) ! uq in Lq ( ) R jrujq dx q L R q q+ ju j dx L; q 2 L q;1 L: q+ Similar result for (pi ) decreasing uniformly to p. (7) For p; q 2 the convergence is actually strong in the corresponding Sobolev Space. Remarks The norm on W01;p()( ), u ! kjrujkp() = kuk1;p() is Frechet dierentiable with derivative given by p (x ) kjrf jkp(p)(x ) jjrf (x )jjp(x ) 1 sgn rf (x ) (grad kjf jk1;p())(x ) = R : p (x ) kjrf jkp(p)(x ) 1 jrf (x )jp(x ) dx Remarks The norm on W01;p()( ), u ! kjrujkp() = kuk1;p() is Frechet dierentiable with derivative given by p (x ) kjrf jkp(p)(x ) jjrf (x )jjp(x ) 1 sgn rf (x ) (grad kjf jk1;p())(x ) = R : p (x ) kjrf jkp(p)(x ) 1 jrf (x )jp(x ) dx Dinca-Mathei (2009) Remarks The norm on W01;p()( ), u ! kjrujkp() = kuk1;p() is Frechet dierentiable with derivative given by p (x ) kjrf jkp(p)(x ) jjrf (x )jjp(x ) 1 sgn rf (x ) (grad kjf jk1;p())(x ) = R : p (x ) kjrf jkp(p)(x ) 1 jrf (x )jp(x ) dx Dinca-Mathei (2009) Therefore: Z Z grad ku0k1;p() (x )rh(x ) dx = p grad ku0kp() (x )h(x ) dx ; with p = kE1 k : Remarks The norm on W01;p()( ), u ! kjrujkp() = kuk1;p() is Frechet dierentiable with derivative given by p (x ) kjrf jkp(p)(x ) jjrf (x )jjp(x ) 1 sgn rf (x ) (grad kjf jk1;p())(x ) = R : p (x ) kjrf jkp(p)(x ) 1 jrf (x )jp(x ) dx Dinca-Mathei (2009) Therefore: Z Z grad ku0k1;p() (x )rh(x ) dx = p grad ku0kp() (x )h(x ) dx ; with p = kE1 k : is the Euler-Lagrange equation corresponding to the maximal functions of the Sobolev embedding!!!! Remarks The norm on W01;p()( ), u ! kjrujkp() = kuk1;p() is Frechet dierentiable with derivative given by p (x ) kjrf jkp(p)(x ) jjrf (x )jjp(x ) 1 sgn rf (x ) (grad kjf jk1;p())(x ) = R : p (x ) kjrf jkp(p)(x ) 1 jrf (x )jp(x ) dx Dinca-Mathei (2009) Therefore: Z Z grad ku0k1;p() (x )rh(x ) dx = p grad ku0kp() (x )h(x ) dx ; with p = kE1 k : is the Euler-Lagrange equation corresponding to the maximal functions of the Sobolev embedding!!!! Bennewitz (2007), Lindqvist-Francina (2013) Open problems Open problems Strong convergence in general? Open problems Strong convergence in general? Uniqueness of positive eigenfunction? Open problems Strong convergence in general? Uniqueness of positive eigenfunction? Implied by the uniform convexity of W01;p ( ) with respect to the norm u ! kjru jkp : Open problems Strong convergence in general? Uniqueness of positive eigenfunction? Implied by the uniform convexity of W01;p ( ) with respect to the norm u ! kjru jkp : Other boundary conditions? (e.g. Neumann, Robin??) MUITO OBRIGADO! MUITO OBRIGADO! THANKS!