nonlinear pdes in sobolev spaces with variable...
TRANSCRIPT
Preprint CMAF Pre-015 (2013)
http://cmaf.ptmat.fc.ul.pt/preprints.html
Nonlinear PDEs in Sobolev Spaces with Variable Exponents
Stanislav Antontsev and Sergey Shmarev
The slide presentation for the invited lecture given at the International Conference “DifferentialEquations, Function Spaces, Approximation Theory" dedicated to the 105th Anniversary of SergeySobolev, August 18-24, Novosibirsk, Russia
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 1
Nonlinear PDEs in Sobolev spaces with variableexponents
Stanislav Antontsev 1 Sergey Shmarev 2
1CMAF, University of Lisbon
2University of Oviedo
Differential Equations, Function Spaces, Approximation TheoryAugust 18-24, Novosibirsk, Russia
International Conference Dedicated to 105th Anniversary of Sergey Sobolev
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 2
Physical motivation
Electro-rheological fluidsrot E = 0, div E = 0, div v = 0
−div a(x,E,D(v)) +∇P = div (v⊗ v) + f
a(x,E,D(v)) ≈ (1+ |D(v)|)(p(E)−2)/2D(v)
Non-Newtonian fluids (v · ∇)v = div(µ(θ) + τ(θ)|D(v)|p(θ)−2D(v)
)−∇P+ f)
div v = 0, −∆θ+ v · ∇b(θ) = g(x)
Filtration of an ideal barotropic gas in a non-homogeneous porous medium:
ρt +∑i
Di
(ρ |Dip|
λi(θ)−2Dip
)= h p = ργ(θ)
Processing of digital images, e.g., Perona-Malik anisotropic diffusion models,
ut = div (G(|∇u|)∇u)) G(s) = e−s ≈
1 as s→ 0
0 as s→∞enhances edges big |∇u| ⇒ slow diffusion
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 3
Lavrentiev’s phenomenon
The classical minimization problem J(u) =
∫Ωf(x,∇u)dx, u ∈W1,p0 (Ω), Ω ⊂ Rn
−C1 + C2|r|p 6 f(x, r) 6 C3 + C4|r|
p, p = const
The classical Sobolev spaces
minJ(u)| u ∈W1,p0 (Ω)
= inf
J(u)| u ∈ C∞
0 (Ω)
Minimization of a functional with nonstandard growth
−C1 + C2|r|p(x) 6 f(x, r) 6 C3 + C4|r|
p(x) 1 < p− 6 p(x) 6 p+ <∞Generalized Sobolev spaces: W1,p(x)0 (Ω)
A model problem:
J(u) =
∫Ω
[1
p(x)|∇u|p(x) + ug(x)
]dx u ∈W1,p(x)0 (Ω)
div(|∇u|p(x)−2∇u
)= g(x)
Lavrentiev’s phenomen: M.A. Lavrentiev “Sur quelques problemes du calcul devariations". Annali di Matematica Pura ed Applicata, v.4, (1926) 107-124
Variable exponents Sobolevspaces
PDEs with variablenonlinearity
New mathematical modelsin continuum mechanics
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 4
Main issues
ut =
n∑i=1
Di(|Diu|pi(z)−2Diu) + f(z,u) z = (x, t) ∈ Q = Ω× (0, T)
u(x, 0) = u0(x) in Ωu = 0 on Γ = ∂Ω× (0, T)
existence of weak (energy) solutions
space localization:
u0 ≡ 0 in ω ⊂ Ω ⇒ u(z) ≡ 0 in ω ′ × (0, τ), ω ′ ⊆ ω
vanishing in a finite time:
∃t∗ : ‖u‖2,Ω(t) = 0 ∀ t > t∗
blow-up:
∃t∗, ω ⊆ Ω : ‖u‖∞,ω(t)→∞ as t→ t∗−
the influence of anisotropy on the propagation properties of solutions.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 5
Lebesgue and Sobolev spaces with variable exponents
Lebesgue-Orlicz spaces (Ω ⊂ Rn)Lp(·)(Ω) = f : Ω 7→ R| f is measurable in Ω,
∫Ω
|f(x)|p(x) dx <∞
‖v‖p(·),Ω = inf
λ > 0 :
∫Ω
(|v(x)|
λ
)p(x)dx < 1
- Luxemburg norm
Sobolev-Orlicz spaces (Ω ⊂ Rn)V(Ω) = v|v ∈ L2(Ω)∩W1,10 (Ω), |Div|pi(x) ∈ L1(Ω)
Vt(Ω) = v(x)| v ∈ L2(Ω)∩W1,10 (Ω), |Div|pi(x,t) ∈ L1(Ω) ∀a.e. t ∈ (0, T)
V ′t(Ω) dual to Vt(Ω) with respect to the scalar product in L2(Ω)
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 6
Parabolic spaces
ut =
n∑i=1
Di(|Diu|pi(z)−2Diu) + f(z,u)
the main space
W =
u(x, t)∣∣∣∣∣∣∣u : [0, T ] 7→ Vt(Ω),
|u|2, |Diu|pi(x, t) ∈ L1(Q)
‖u‖W = ‖u‖2,Q +
n∑i=1
‖Diu‖pi(·,·),Q
the dual space W ′:
w ∈W′ ⇐⇒
w = w0,w1, . . . ,wn, w0 ∈ L2(Q), wi ∈ Lp′i(·)(Q),
∀φ ∈W 〈〈w,φ〉〉 =∫Q
(w0φ+
∑i
wiDiφ
)dz.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 7
Basic properties
1 if p(x) ∈ C0(Ω), then
C∞(Ω) is dense in Lp(·)(Ω) W1,p(·)0 (Ω) is separable and reflexive
2 C∞0 (Ω) is dense in V(Ω) if p(x) is Log-continuous: ∀ z, ζ ∈ Ω, |z− ζ| < 1
∑i
|pi(z) − pi(ζ)| 6 ω(|z− ζ|), limτ→0+ ω(τ) ln
1
τ= C < +∞.
3 min(‖f‖p
−
p(·) , ‖f‖p+
p(·)
)6∫Ω
|f(x)|p(x) dx 6 max(‖f‖p
−
p(·) ,‖f‖p+
p(·)
)4 Hölder’s inequality: f ∈ Lp(·)(Ω), g ∈ Lp′(·)(Ω), p(x) ∈ (1,∞), p ′(x) =
p(x)
p(x) − 1∫Ω
|f g|dx 6
(1
p−+
1
(p′−)
)‖f‖p(·) ‖g‖p′(·) 6 2‖f‖p(·) ‖g‖p′(·)
L. Diening, P. Harjulehto, P. Hasto, and M. Ruzicka, Lebesgue and Sobolev Spaceswith Variable Exponents, Springer, Berlin, 2011. Lecture Notes in Mathematics, Vol. 2017,1st Edition.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 8
Existence results for evolution p(x, t)-Laplace equation
Existence theorems for the evolution p(x, t) -laplacianut = div
(a(z,u)|∇u|p(x,t)−2∇u
)in Q
u = 0 on Γ , u(x, 0) = u0 in Ω0 < a− 6 a(z, s) 6 a+ <∞ (1)
1 S.Antontsev, S.Shm. (2006, 2009) - Galerkin’s approximations. Energy solution forp(x, t) continuous in Q with logarithmic module of continuity
2 Yu.Alkhutov, V.Zhikov (2010) - very weak solution for measurable and boundedp(x, t). The theory of monotone operators.
3 L.Diening, P. Nägele, M. Ružička (2011) Existence of energy solution for systems ofequations of the type (1). The theory of monotone operators + p(x, t) ∈ Clog(Q).
(For p independent of t: entropy and renormalized solutions with L1 data, variationalinequalities, equations with nonlocal terms, semi-group theory, stationary solutions,calculus of variations, .... )
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 9
Definition of energy solution
ut =
n∑i=1
Di(|Diu|pi(z)−2Diu) + f(z,u) in Q
u = 0 on Γ , u(x, 0) = u0 in Ω
(2)
Definitionu(x, t) is called energy solution of problem (2) if
1 u ∈W, ∂tu ∈W ′
2 ∀φ ∈W with ∂tφ ∈W ′
∫Q
(φ∂tu+
n∑i=1
|Diu|pi−2Diu ·Diφ− fφ
)dz = 0
3 ∀φ(x) ∈ C∞0 (Ω)
∫Ω(u(x, t) − u0(x))φ(x)dx→ 0 as t→ 0
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 10
Existence of energy solutions ut =
n∑i=1
Di(|Diu|pi(z)−2Diu) + f(z,u) in Q
u = 0 on Γ , u(x, 0) = u0 in Ω
(3)
Theorem (Global in time existence)
Let f(z, r) be a Carathéodory function, |f(z, r)| 6 c0|r|λ−1 + h(z), h ∈ Lλ′(Q) . Assume
that pi(z) are Log-continuous in Q. If one of the conditions
λ = max2,p− − δ with δ > 0, p− = mini
infQpi(z),
λ = max2,p− and c0 << 1
is fulfilled, then for every u0 ∈ L2(Ω) problem (3) has a solution u ∈W, ut ∈W ′.
Theorem (Local in time existence)Let |f(z,u)| 6 d0|u|λ−1 + h(z), λ > 2, h ∈ L1(0, T ;L∞(Ω)). Then ∀ u0 ∈ L∞(Ω) problem(3) has at least one energy solution in a cylinder QT with some T > 0. This solution can becontinued to the maximal interval [0, T∗],
T∗ = supT > 0 : ‖u(t)‖∞,Ω <∞ ∀ t < T S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 11
Sketch of proof: adaptation of Galerkin’s method
C∞0 (Ω) ⊂ Hs0(Ω) ⊂ Vt(Ω) (dense embeddings)
u(k) =∑ki=1 u
(k)i (t)ψi(x), ψi dense in W1,p
+
0 (Ω), (ψi,ψj)2,Ω = δij
monotonicity+
compactness+
integrationby parts in t
⇒ ∫
Q
[(ut + f)φ+
∑i
(|Diu|
pi(z)−2Diu)·Diφ
]dz = 0
integration by parts: finite differencesf(·, t+ h) − f(·, t)
h→ ∂tf wouldn’t work
f(x, t+ h) ∈ Vt+h(Ω), f(x, t) ∈ Vt(Ω), Vt(Ω) 6= Vt+h(Ω)
Lemma (Integration by parts)v,w ∈W(Q), vt,wt ∈W ′(Q), pi(z) are Log-continuous ⇒
∀a.e. t1, t2 ∈ (0, T ]∫t2t1
∫Ωvwt dz+
∫t2t1
∫Ωvtwdz =
∫Ωvwdx
∣∣∣t=t2t=t1
.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 12
Localization properties in the isotropic case
Isotropic p-Laplacian, p = const: the alternative
ut = div(|∇u|p−2∇u) in Qu(x, 0) = u0(x) in Ω, u = 0 on Γp = const ∈ (1, 2)∪ (2,∞)
Counterpart localization properties
Slow diffusion ⇒ space localizationp > 2 ⇒ u0(x0) = 0 ⇒ ∃ t0 : u(x0, t) = 0 ∀ t ∈ [0, t0]
Fast diffusion ⇒ time localizationp ∈ (1, 2) ⇒ u(x, t) ≡ 0 for t > T∗, x ∈ Ω
No longer true if the diffusion is anisotropic or variable!
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 13
Anisotropic diffusion: new issues in localization
Dirichlet problem for the anisotropic p(z)-Laplacian ut =∑i
Di
(|Diu|
pi(z)−2Diu)+ f(z,u) in Q
u(x, 0) = u0 in Ω u = 0 on Γ
Localization (vanishing) properties
traditional technique: comparison
no explicit super-solutions ⇒ analysis of local energy functions
directional (anisotropic) localization in space
time localization in equations of fast-slow diffusion
localized solutions of eventually linear equations
S. Antontsev, J. I. Diaz, S. Shmarev “Energy methods for free boundary problems.Applications to Nonlinear PDEs and Continuum Mechanics", Progress in Nonlinear PDEs,v.48, Birkhauser (2002)
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 14
Finite speed of propagation
1 Finite speed of propagation of disturbances from the initial data
ut =∑i
Di
(|Diu|
pi(z)−2Diu)+ f pi(z) > p
− > 2
Finite speed of propagation. pi(z) > 2 - slow diffusionu0 = 0, f = 0 en Bρ0(x0) ⇒ u(x, t) = 0 en Bρ(t)(x0), ρ(t) = ρ0 − Ctγ
The set x ∈ Bρ(t)(x0) : u(x, t) = 0 shrinks as t grows
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 15
The waiting time effect
1 Finite speed of propagation of disturbances from the initial data2 The waiting time effect.
ut =∑i
Di
(|Diu|
pi(z)−2Diu)+ f pi(z) > p
− > 2
The waiting time effect. pi(z) > 2 - slow diffusionu0 = 0, f = 0 en Bρ(x0) ⇒ u(x, t) = 0 en Bρ(x0)× [0, t∗]
y ∈ ∂x ∈ Bρ0(x0) : u(x, t) = 0 is immobile on [0, t∗(y)]
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 16
The local energy methodReduction to a nonlinear ODI for the energy function
E(r) =
∫t0
∫Br(x0)
∑i
|Diu|p− dz
E ′(r) =
∫t0
∫∂Br(x0)
∑i
|Diu|p− dSdt for a.e. r ∈ (0, dist (x0,∂Ω))
ut =∑i
Di(|Diu|pi(z)−2Diu) + f(z) pi(x, t) ∈ (p−,p+)
u0 ≡ 0 in BR(x0). Multiply by u(z) and integrate by parts in Br(x0)× (0, t):
1
2‖u‖22,Br(t) +min
Ep+
p− (r), E(r)6
t∫0
∫∂Br
|u|∑i
|Diu|pi(z)−1 + F
trace-interpolation inequality + small oscillation of pi(z)⇒∀a.e. r ∈ (0,R) E(r) 6 C(t) (E ′)
1α (r) + F(r) α ∈ (0, 1) if p− > 2
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 17
Analysis of the ODI for the energy function
ut =∑i
Di
(|Diu|
pi(z)−2Diu)+ f
E(r) =
∫t0
∫Br(x0)
∑i
|Diu|p− p− > 2
⇒
Eα(t) 6 C(t)E ′(r) +Φ(r)
E ′(r) > 0 0 6 E(r) 6M
α ∈ (0, 1)
if Φ = 0,then
0 6 E1−α(r) 6M1−α −r− ρ0C(t)︸ ︷︷ ︸→ 0 as r grows
Big domain+
small total energy M =∑i
∫Q|Diu|
p− dz
⇒
localization:
u(x, t) ≡ 0 in Bρ(t)(x0)× [0, t∗]
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 18
Anisotropic diffusion: the choice of the energy function
ut =∑iDi
(|Diu|
pi(z)−2Diu)+ f(z), x = (x1, x ′)
E(s) =
t∫0
∫Ω∩ x1 > s
∑i
|Diu|pi(z) dz
E ′(s) = −
t∫0
∫Ω∩ x1 = s
∑i
|Diu|pi(z) dx ′ dt ∀a.e. s
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 19
Directional localization•
ut =∑iDi
(|Diu|
pi(z)−2Diu)+ f(z) pi(z) ∈ (p−,p+) ⊂ (1,∞)
• Multiply by u and integrate by parts in Ω∩ x1 > s× (0, t)
12‖u‖
22,Ω∩x1>s
+ E(r) 6∫t0
∫Ω∩x1=s
|u|∑i
|Diu|pi(z)−1 + F
embedding theorems in anisotropic spaces on Ω∩ x1 = s
• Eβ(s) + C(t)E ′(s) 6 Φ(s)
0 6 E(s) 6M, E ′(s) 6 0
1
p−1<
1
n− 1
n∑i=2
1
p+i6
1
n− 1⇒ β ∈ (0, 1) E(s) ≡M for x1 > s∗
Localization in the direction x1The fast diffusion equation admits localized in space solutions
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 20
Anisotropic diffusion: infinite waiting time
ut =∑i
Di
(|Diu|
pi−2Diu)+ f(z) pi(z) > 1
Eβ(s) + CE ′(s) 6 Φ(s) ∼ ‖u0‖22,Ω∩x1>s + ‖f‖22,(Ω∩x1>s)×(0,t)∫s0
s
Φ(z)
(z− s0)1/(1−β)dz <∞ 1
p−1<
1
n− 1
n∑i=2
1
p+i6
1
n− 1
Anisotropic diffusion ⇒ nonpropogation of disturbances !
y
x
u(x,t)=0
a
b
l x*
supp fsupp u0
0
t<T*
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 21
Nonpropagation of disturbances: a simple example
ut = uxx +(|uy|
q−2uy)y
q > 2
Separation of variables: u = X(x, t) · Y(y)
Xt
Xq−1−
Xxx
Xq−1︸ ︷︷ ︸const
=(|Yy|
q−2Yy)yY︸ ︷︷ ︸
constXt = Xxx + µXq−1 in (a,b)× (0, T)(|Y ′|q−2Y ′) ′ = µY in (c,d)µ = const > 0
Y(y) = C (L− y)α+ α =q+ 1
q− 1C = C(µ,q)
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 22
Extinction in a finite time
ut =
∑iDi
(|Diu|
pi(z)−2Diu)− c |u|σ(z)−2u+ f(z)
u = 0 on ΓT , u(x, 0) = u0(x) in Ω
Theorem (Diffusion/absorption balance)If f ≡ 0 for t > tf and
either c(z,u) > c0 > 0 and
1 <1
n
n∑i=1
1
p+i (t)+
1
σ+(t)61
n
n∑i=1
1
p−i (t)+
1
σ−(t)6 1+
2
n,
or c(z,u) > 0, and
1 <2
n
n∑i=1
1
p+i (t)62
n
n∑i=1
1
p−i (t)6 1+
2
n,
then every energy solution vanishes at a finite moment tu > tf.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 23
Reduction to nonlinear ODI for the energy function
The energy function E(t) =
∫Ωu2(x, t)dx
ut =∑i
Di
(|Diu|
pi(z)−2Diu)− c |u|σ(z)−2u+ f(z)
Multiply by u and integrate over Ω
1
2E ′(t) +
∫Ω
(∑i
|Diu|pi(z) + c|u|σ(z)
)6∫Ω
|f| |u|
embedding theorems in anisotropic spaces ⇒E ′(t) + CEβ(t) 6 F(t) = K
∫Ωf2
1
β(t)=1
n
n∑i=1
1
p+i (t)+
1
σ+(t)or
1
β(t)=2
n
n∑i=1
1
p+i (t)
β(t) ∈ (0, 1), F(t) ≡ 0 for t > tf ⇒ E ≡ 0 for t > tu
Fast diffusion in one direction may cause extinction!
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 24
Conditions of extinction. Model equation: n = 2ut = Dx
(|Dxu|
p1−2Dxu)+Dy
(|Dyu|
p2−2Dyu)
p1,p2 = const
Anisotropic diffusion Isotropic diffusion
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 25
Equations which become linear as t→ ∞ut =
∑i
Di
(|Diu|
pi(z)−2Diu)− |u|σ(z)−2u,
supΩ pi(z)→ 2, supΩ σ(z)→ 2 as t→∞E(t) = ‖u(·, t)‖22,Ω :
E ′(t) + CEβ(t)(t) 6 0
1
β(t)=1
n
n∑i=1
1
p+i (t)+
1
σ+(t)→ 1 as t→∞
Theorem (Extinction in eventually linear equations)β(t) is monotone increasing
Eqn. C∫ t∗0‖u0‖
2(β(t)−1)2,Ω dt =
∫∞0
ds
es(1−β(s))has a root ⇒ u(x, t) ≡ 0 for t > t∗
Examples
1 ut = ∆u− |u|σ(t)−2u σ(t)→ 2 :
∫R
ds
es(σ(s)−1)<∞
2 ut =(|ux|
p(t)−2ux
)x
p(t)→ 2 :
∫R
ds
es(2−p(s))/2<∞
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 26
Space/time localization: an example
ut =
(|ux|
p−2ux)x+(|uy|
q−2uy)y
in QT = (0,a)× (0,b)× (0, T),
u = 0 on Γ , u(x,y, 0) = u0.
1
p+1
q> 1 ⇒
u(x,y, t) vanishes in a finite timeu(x, t) ≡ 0 ∀ t > T∗
1 < q < p ⇒ u(x,y, t) is localized in the direction x:u0 = 0 for x > x0 ⇒ u ≡ 0 for x > x ′
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 27
Blow-up: Kaplan’s method for semilinear equations
ut = ∆u+ b(z) uσ(z) − 1
u|Γ = 0, u(x, 0) = u0(x) > 0 in Ω
−∆φ = λφ in Ω,
φ = 0 on ∂Ω,∫Ωφdx = 1
DefinitionWe say that u(x, t) blows-up in a finite time if ∃ t∗ : ‖u(·, t)‖∞,Ω →∞ as t→ t∗.
A sufficient condition of blow-up:
µ(t) =
∫Ωu(x, t)φ(x)dx 6 ‖u‖∞,Ω
∫Ωφ(x)dx = ‖u(·, t)‖∞,Ω →∞ as t→ t∗
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 28
Blow-up: Kaplan’s method for semilinear equations
ut = ∆u+ b(z)︸ ︷︷ ︸>0
uσ(z) − 1
−∆φ = λφ in Ω, φ = 0 on ∂Ω
µ ′(t) > −λµ(t) + α(t)µσ−−1(t) − β(t), µ(0) > 0, µ(t) =
∫Ωu(z)ψ(x)dx
0 < b− 6 α(t) =
(∫Ωb
12−σ− (z)φ(x)dx
)2−σ−0 6 β(t) =
∫Ωb(z)φ(x)dx 6 b+ <∞
A− = mint>0
(α(t) −
λσ−(t)−1
σ−(t) − 1
), B+ = max
t>0
(β(t) +
σ−(t) − 2
σ−(t) − 1
)<∞.
Theorem (Nonexistence of global solutions)
Every weak solution blows-up if A− > 0, B+ <∞, A−µσ−(t)−1(0) > B+ for all t > 0 and
either σ−(t) = minx∈Ω σ(x, t) > σ− > 2, or
µ(0) > 1, σ−(t) > 2, σ−(t) 2 as t→∞,∫∞lnµ(0)
es(2−σ−(s))ds <∞
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 29
ExtensionsAn example:
ut = ∆u+ u1+ε(t)∫∞lnµ(0)
dτ
eτε(τ)<∞ ε(τ) = α
ln ττ
α > 1
Regional blow-up
µ(t) =
∫Du(x, t)φ(x)dx
−∆φ = λφ in D ⊂ Ω ,φ = 0 on ∂D
Equations with nonlocal reaction terms
ut = ∆u+ b(z)uσi(z)−1 + c(z)
∫Ωd(z)uσ2(z)−1 dx
interaction between b(z) and σ(z): every solution of the equation
ut = ∆u+ b(z)uσ(z)−1
blows-up if I =
∫Ω
ds
bqγ1−q (s, t)
<∞ with some q > 1σ−−1
, γ > 1.
For example:
|Ω| = |x| < 1, b = |x|−α ⇒ I =
∫10sn−1−α qγ
1−q ds <∞ ⇒ α <n(1− q)
γq
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 30
Blow-up in solutions of quasilinear equations ut = div
(a(z)|∇u|
p(x) −2∇u
)+ b(z)|u|
σ(x) −2
u(x, 0) = u0(x) in Ω, u = 0 on ΓT
Theorem1) σ− > 2, p+ = supΩ p(x) 6 σ− = infΩ σ(x), at(x, t) 6 0, bt(x, t) > 0 ,
2)
∫T0
(maxx∈Ω
|at(x, t)|+ |bt(x, t)|)dt <∞, |u0|
σ(x) ∈ L1(Ω), |∇u0|p(x) ∈ L1(Ω)
3)
∫Ω
(a(x, 0)p(x)
|∇u0|p(x) −b(x, 0)σ(x)
|u0|σ(x)
)dx 6 0
⇒ every nonstationary solution blows-up in a finite time:
∃ t∗ ≡ t∗(Ω,‖u0‖∞) <∞ : ‖u(·, t)‖∞,Ω →∞ as t→ t∗.
Reduction to the 2nd-order differential inequality for f(t) =1
2
∫t0
∫Ωu2(z)dx :
1 p+ = supΩp(x) > 2 ⇒ f(t) 6 2λf ′′(t)
1
σ−6 λ 6
1
p+,
2 p+ ∈ (1, 2], σ− > p+ ⇒ K (f ′(t))σ−
2 6 f ′′(t) K =
(λ−
1
σ−
)b−
λ
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 31
Wave equation with p(x, t)-Laplacian
utt = div
(|∇u|p(z)−2∇u+ ε∇ut
)+ b|u|σ(z)−2u+ f
u(x, 0) = u0(x), ut(x, 0) = u1(x), x ∈ Ω,u|ΓT = 0, ΓT = ∂Ω× (0, T), ε = const > 0.
(4)
The main function space:
W =
u : u,∇ut ∈ L2 (QT ) , ut, |∇u|
p(z)2 , |u|
σ(z)2 ∈ L∞ (0, T ;L2 (Ω)
), u|ΓT = 0
‖u‖W = ‖u‖2,QT + ‖u‖σ(·),QT + ‖ut‖L∞(0,T ; L2(Ω))
+‖∇u‖L∞(0,T ; L2(Ω)) + ‖∇ut‖2,QT + ‖∇u‖p(·),QT
Definition (The energy solution u ∈ W)
1 u(·, t) u0 in W1,20 (Ω)∩W1,p(·,0)(Ω), ut(·, t) u1 in L2(Ω)
2 ∀ϕ ∈ C∞ (0, T ;C∞0 (Ω)
)such that ϕ(x, T) = 0
∫QT
(−utϕt +
(|∇u|p(·)−2∇u+ ε∇ut
)· ∇ϕ− b |u|σ(·)−2 uϕ
)dxdt
=
∫Ωu1ϕ(·, 0)dx+
∫QT
fϕdx
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 32
Local and global in time existence
utt = div
(|∇u|p(z)−2∇u+ ε∇ut
)+ b|u|σ(z)−2u+ f
u(x, 0) = u0(x), ut(x, 0) = u1(x), x ∈ Ω,u|ΓT = 0, ΓT = ∂Ω× (0, T), ε = const > 0.
(5)
u0 ∈ L2(Ω)∩W1,20 (Ω)∩W1,p(·,0)(Ω), u1 ∈ L2(Ω), f ∈ L2(QT )p(z) ∈ Clog(QT ) 1 < p− 6 p(x, t) 6 p+ <∞, |pt| = −pt 6 Cp ,
1 < σ− 6 σ(x, t) 6 σ+ 6∞, 0 6 σt 6 Cσ
(6)
Theorem (Global in time existence)
Let conditions (6) be fulfilled and one of the following conditions holds: a) b < 0 ,
b) σ+ 6 2, or σ+ < p− . Then problem (5) has at least one global in time energy solution.
Theorem (Local in time existence)
Let conditions (6) be fulfilled and b > 0, 2 < σ− 6 σ+ <n+ 2
np−,
2n
n+ 2< p− . Then
problem (5) has at least one solution on a small time interval [0, T0).
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 33
The energy relations
The energy functional: E(t) =
∫Ω
[|ut(·, t)|2
2+
|∇u|p(·,t)
p(·, t)− b
|u|σ(·,t)
σ(·, t)
]dx
E ′(t) + ε
∫Ω
|∇ut(·, t)|2 dx = Λ1 +Λ2 +Λ3
Λ1 =
∫Ω
[−
|∇u|p
p2(1− p ln |∇u|) pt
]dx,
Λ2 =
∫Ω
(b |u|σ
σ2(1− σ ln |u|) σt
)dx, Λ3 =
∫Ωfutdx.
1 pt = σt = f = 0 ⇒ E(t) + ε
∫t0
∫Ω
|∇ut(x, s)|2 dxds = E(0) ∀t > 0
2 f = 0 ⇒ E(t) + ε
∫t0
∫Ω
|∇ut(x, s)|2 dxds 6 E(0) + Ct
with the constant C = ( bσ2−Cσ + 1
p2−Cp)|Ω|.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 34
Conditions of the finite time blow-up
utt = div
(|∇u|p(z)−2∇u+ ε∇ut
)+ b|u|σ(z)−2u+ f
u(x, 0) = u0(x), ut(x, 0) = u1(x), x ∈ Ω,u|ΓT = 0, ΓT = ∂Ω× (0, T), ε = const > 0.
(7)
Theorem
If b < 0, 2 < p− 6 p+ < σ−, E0 < 0, Z(0) = 2(u0,u1)2,Ω + ε‖∇u0‖22,Ω > 0 , then
∃tmax <∞ :
∫Ω
|u|σdx > CZ(0)
(1−
t(µ− 1)
C(Z(0))µ−1
)− 1µ−1 →∞ as t→ tmax,
where C > 0 and µ > 1 are constants depending only on the data.
Weak dependence of the exponents p(z), σ(z) on t (the constants Cp,Cσ are small):
δ = max (Cp,Cσ) 6 |E(0)| (tmax
(1
p2−+
b
σ2−
)|Ω|)−1
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 35
BibliographyS. Antontsev, J. I. Diaz, S. Shmarev “Energy methods for free boundary problems", Progress inNonlinear PDEs, v.48, Birkhauser (2002)
S. Antontsev, S. Shmarev. "Parabolic equations with anisotropic nonstandard growth conditions"Free boundary problems, 33–44, Internat. Ser. Numer. Math., 154, Birkhäuser, Basel, 2007.S. Antontsev, S. Shmarev. "Anisotropic parabolic equations with variable nonlinearity". Publ.Mat. 53 (2009) 355–399.
S. Antontsev, S. Shmarev. "Localization of solutions of anisotropic parabolic equations".Nonlinear Anal. 71 (12) (2009) e725–e737.
S. Antontsev, S. Shmarev. "Vanishing solutions of anisotropic parabolic equations with variablenonlinearity". J. Math. Anal. Appl. 361 (2) (2010) 371–391.
S. Antontsev, S. Shmarev. "Blow-up of solutions to parabolic equations with nonstandard growthconditions". Journal of Computational and Applied Mathematics, 234 (2010) 2633–2645.
S. Antontsev, S. Shmarev. "Blow-up of solutions to parabolic equations with nonstandard growthconditions". Journal of Computational and Applied Mathematics 234 pp.2633–2645 (2010)
S. Antontsev, S. Shmarev. "On the Blow-up of Solutions to Anisotropic Parabolic Equations withVariable Nonlinearity". Proceedings of the Steklov Institute of Mathematics, 2010, Vol. 270, pp.27–42.S. Antontsev, S. Shmarev. "Energy Solutions of Evolution Equations with Nonstandard GrowthConditions". Monografías de la Real Academia de Ciencias de Zaragoza 38: 85–111, (2012).
S. Antontsev, S. Shmarev. "Doubly degenerate parabolic equations with variable nonlinearity I:Existence of bounded strong solutions." Adv. Differential Equations 17 (2012), no. 11-12,1181–1212.S. Antontsev, S. Shmarev. "Elliptic equations with anisotropic nonlinearity and nonstandardgrowth conditions". Handbook of Differential Equations. Stationary Partial Differential Equations,V.3 Edited by M. Chipot and P. Quittner, Elsevier (2006), pp. 1–100.
S. Antontsev, S. Shmarev. "Parabolic equations with double variable nonlinearities". Mathematicsand Computers in Simulation, 81 (2011) pp. 2018–2032.
S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 36
Bibliography
S. Antontsev, S. Shmarev. "Doubly degenerate parabolic equations with variable nonlinearity I:existence of bounded strong solutions". Adv. Differential Equations 17 (2012), 11-12, 1181–1212.
S. Antontsev, S. Shmarev. "Uniqueness and comparison theorems for solutions of doubly nonlinearparabolic equations with nonstandard growth conditions". CPAA, 12 (4) (2013), pp. 1827–1846.
S. Antontsev, S. Shmarev. "On localization of solutions of elliptic equations with nonhomogeneousanisotropic nonlinearity". Siberian Mathematical Journal, 46 (5) (2005), pp. 765–782.
S. Antontsev, S. Shmarev. "Elliptic equations and systems with nonstandard growth conditions:Existence, uniqueness and localization properties of solutions". Nonlinear Analysis, 65 (2006)728–761.
S. Antontsev. "Wave equation with p(x, t)- Laplacian and damping term: existence and blow-up".Differ. Equ. Appl., 3 (2011), pp. 503–525.
S. Antontsev. "Wave equation with p(x, t)-Laplacian and damping term: Blow-up of solutions".C.R. Mecanique, 339 (2011), pp. 751–755.
P. Amorim and S. Antontsev. "Young measure solutions for the wave equation with p(x,t)-Laplacian: existence and blow-up". Nonlinear Analysis, 92 (2013), pp. 153-167.
S. Antontsev and J. Ferreira. "Existence, uniqueness and blow-up for hyperbolic equations withnonstandard growth conditions". Nonlinear Analysis, Series A: Theory, Methods, Applications, 93(2013), pp. 62–77.
Thank you!S. Antontsev, S. Shmarev PDEs with Variable Nonlinearity 37