symmetric and elliptic, $b_{i}¥in C^{0}(¥overline{¥Omega})$ ... $p¥geq 1$ , and finally (for some $¥delta¥in(0,1)$). (6) ... For arbitrary $¥epsilon>0$.
Funkcialaj Ekvacioj, 30 (1987) 111-114
Decay Properties of Global Solutions of
Reaction-Diffusion Equations By Tomasz DLOTKO (Silesian University, Poland)
This note is a supplement to our previous paper [2]. Certain generalizations associated with [2] are given concerning the asymptotic properties of global solutions of divergence parabolic equations. The subject here is the quasilinear parabolic Dirichlet problem, a special case of that considered in [2, Theorem 1];
(1)
$u_{t}=¥sum_{i,j}¥frac{¥partial}{¥partial x_{i}}(a_{ij}(x)¥frac{¥partial u}{¥partial_{X_{j}}})+¥sum_{i}b_{i}(x)¥frac{¥partial u}{¥partial x_{i}}$
$+f(t, x, u, ¥nabla u)$
(2)
on
$u=0$
where $C^{1}(¥overline{¥Omega})$
$¥partial¥Omega$
,
,
$u(0, x)=u_{0}(x)$
,
, , -smooth bounded domain in symmetric and elliptic, and a continuous function $t¥geq 0$
$R^{n}$
$¥mathrm{x}¥mathrm{e}¥mathrm{O}$
$b_{i}¥in C^{0}(¥overline{¥Omega})$
(3)
$u¥cdot f(t, x, u, ¥nabla u)¥leq Cu^{2}$
, with satisfying;
$¥nabla u=(u_{x_{1}},¥cdots, u_{x_{n}})$
$f$
$ a_{ij}¥in$
.
The growth rate [2] for $(u ¥cdot f¥leq Cu^{2}+D)$ has been sharpened to (3). This is needed for the iteration procedure (as in Theorems 1, 2 of [2]) to give decay instead of only boundedness. Moreover is assumed to be Lipschitz with respect to (constant ), (constant ), differentiate with respect to ¥ const. (M); both these assumptions are , and bounded with $K_{M}:=R^{+} ¥ times ¥ overline{ ¥Omega}¥times[-M, M]¥times R^{n}$ . uniformly satisfied in sets In accordance with our convention [2] all sums are taken from 1 to and all integrals over . norms in (4) replaced by It is shown in [2] (with norms this is not essential for the proof in [2] that: $f$
$L^{¥infty}(¥Omega)$
$f$
$L^{¥infty}(¥Omega)$
$N_{1}$
$t$
$¥nabla u$
$N_{3}$
$u$
$|f| leq$
$¥partial f/¥partial u¥leq N_{2}$
$n$
$¥Omega$
$L^{1}$
$L^{2}$
$)$
Proposition 1.
(4) $t¥geq 0$
(5)
Conjunction $||u(t, ¥cdot)||_{L^{1}(¥Omega)}¥leq$
, ensures estimates
uniform
$||u(t, ¥cdot)||_{L^{¥infty}(¥Omega)}¥leq$
of the const.,
two conditions $||u_{t}(t, ¥cdot)||_{L^{1}(¥Omega)}¥leq$
const.
in time
const.,
$||u_{t}(t, ¥cdot)||_{L^{p}(¥Omega)}¥leq$
const.
?
Tomasz DLOTKO
112 $p¥geq 1$
, and finally (for some
(6)
$¥delta¥in(0,1)$
$||u||_{c^{1/2,12}}l(R^{+}¥times¥overline{¥Omega})¥leq$
const.,
We want to study here the We have: For some
Theorem A. $||u(t, ¥cdot)||_{L^{¥infty}(¥Omega)}¥rightarrow 0$
as
$ t¥rightarrow¥infty$
)
$L^{¥infty}(¥Omega)$
$q¥in N$
$||¥nabla u||_{C^{¥delta/2,¥delta}(R^{+}¥times¥overline{¥Omega})}¥leq$
const.
decay to zero of the solution
let
when
$||u(t, ¥cdot)||_{L^{q}(¥Omega)}¥rightarrow 0$
$u$
of $(1)-(2)$ .
$ t¥rightarrow¥infty$
.
Then
.
convergence together with First it must be remembered that the convergence for all $p¥in N$ (but not necessarily boundedness (5) guarantees convergence. with the same rate of decay). We proceed to the proof of the , which together is compact in As a consequence of (6) the family , decay to zero ensures the existence of a sequence of times with the . This gives uniform in decay in both in such that
Proof.
$L^{q}$
$L^{p}$
$L^{¥infty}$
$L^{¥infty}$
$¥{u(t, ¥cdot)¥}_{t¥geq 0}$
$C^{0}(¥overline{¥Omega})$
$L^{q}$
$ t_{n}¥rightarrow¥infty$
$C^{0}(¥overline{¥Omega})(L^{¥infty}(¥Omega))$
$u(t_{n}, ¥cdot)¥rightarrow 0$
$L^{r}(¥Omega)$
$r$
; $||u(t_{n}, ¥cdot)||_{L^{r}(¥Omega)}¥leq|¥Omega|^{1/r}||u(t_{n}¥cdot)||_{L^{¥infty}(¥Omega)}$
$¥leq¥max¥{1;|¥Omega|¥}||u(t_{n}, ¥cdot)||_{L^{¥infty}(¥Omega)}¥rightarrow 0$
.
, which is the consequence of our We need also the simplified version (with sharpened condition (3) of the estimate (number (8) in [2]) shown in Theorem 1 in this is a dynamical system we may change 0 into of [2], Since estimate, to get: $¥tilde{D}=0$
$)$
$t_{0}¥geq 0$
$u(t, ¥cdot)$
(7)
$y_{k}(t_{0})¥leq¥max¥{||¥mathrm{u}(t_{0}, ¥cdot)||_{L^{2^{k}}(¥Omega)}; y_{k-1}(t_{0})(¥tilde{B}+¥tilde{C}2^{(k+1)3n/2})^{2^{-¥mathrm{k}}}¥}$
where For arbitrary
$y_{k}(t_{0}):=¥sup_{t¥geq t_{0}}||u(t, ¥cdot)||_{L^{2^{k}}(¥Omega)}$
(i) (ii)
take
$¥epsilon>0$
$n=n_{0}$
,
.
, such that ,
$¥max¥{1, |¥Omega|¥}||u(t_{n¥mathrm{o}}, ¥cdot)||_{L^{¥infty}(¥Omega)}¥leq¥epsilon$
$||u(t, ¥cdot)||_{L^{1}(¥Omega)}¥leq¥epsilon$
for in (7) to the value and (i), (ii) that $t¥geq t_{n_{0}}$
Enlarging the constant follows from (7) with
$¥tilde{C}$
$(||u||_{L^{1}(¥Omega)}¥rightarrow 0)$
$¥tilde{C}’$
.
, such that
$¥tilde{C}’¥geq 2¥max¥{¥tilde{B},¥tilde{C}, 1¥}$
, it
$t_{0}=t_{n_{0}}$
(8) Then clearly
$y_{k}(t_{n¥mathrm{o}})¥leq¥max¥{¥epsilon, y_{k-1}(t_{n_{0}})(¥tilde{C}’ 2^{(k+1)3n/2})^{2^{-k}}¥}$
$y_{k}(t_{n_{0}})¥leq y_{k}^{¥prime}$
, where the increasing
$ y_{1}^{¥prime}=¥epsilon$
,
$y_{k}^{¥prime}=y_{¥acute{k}-1}$
$(¥tilde{C}’¥geq 1)$
.
sequence
$(¥tilde{C}’ 2^{(k+1)3n/2})^{2^{-k}}$
,
or by induction
(9)
$¥sup_{t¥geq t_{n¥mathrm{o}}}||u(t, ¥cdot)||_{L^{¥infty}(¥Omega)}¥leq¥lim_{k¥rightarrow¥infty}y_{k}^{¥prime}=:y_{¥infty}^{¥prime}$
$=¥epsilon(C^{¥prime}2^{3n/2})(¥prod_{k=1}^{¥infty}(2^{k})^{2^{-k}})^{3n/2}=¥epsilon C^{¥prime}(2P)^{3n/2}$
$=$
const. , $¥epsilon$
$¥{y_{k}^{f}¥}$
is given by;
Reaction-Diffusion
where
$x$
$P:=¥prod_{k=1}^{¥infty}(2^{k})^{2^{-k}}$
. Hence the proof is completed.
For a subclass of the equation (1) (with (exponent ) in ):
$a_{ij}¥in C^{1+a}(¥overline{¥Omega})$
,
$¥partial¥Omega¥in C^{2+a}$
,
$f$
Holder in
$K_{M}$
$¥alpha$
(10)
113
Equations
,
$u_{t}=¥sum_{i,j}¥frac{¥partial}{¥partial x_{i}}(a_{ij}(x)¥frac{¥partial u}{¥partial x_{j}})+f(x, u, ¥nabla u)=:Lu+f$
with the conditions (2) we have the following theorem concerning the decay of $||u_{t}(t, ¥cdot)||_{L^{¥infty}(¥Omega)}$
;
independent be the solution of (10)?(2) (with Theorem B. Let begining satisfying the conditions at the of this note) belonging to when . . Then $f$
$u$
of
$t$
$C^{1+¥gamma/2,2+¥gamma(R^{+}}$
$ t¥rightarrow¥infty$
$(||u_{t}(t, ¥cdot)||_{L^{1}(¥Omega)}¥rightarrow 0)¥Rightarrow(||u_{t}(t, ¥cdot)||_{L^{¥infty}(¥Omega)}¥rightarrow 0)$
$¥times¥overline{¥Omega})$
Proof. As a consequence of
the estimate (16) in [2] (with $N_{1}=0$ as a , $(c_{1}2^{k(1+n/2)}$ result of the special form (10) we have with . Enlarging to the value $c_{1}^{¥prime}=2¥max¥{c_{1}, c_{2},1¥}$ we arrive at the estimate 2 (for roots): $ z_{k}^{2^{¥mathrm{k}}}(t_{0})¥leq¥max$
$)$
$+c_{2})z_{k^{¥_}1}^{2^{k}}(t_{0})¥}$
$¥{¥int u_{k}^{2^{k}}(t_{0}, x)dx$
$z_{k}(t_{0}):=¥sup_{t¥geq t_{0}}||u_{t}(t, ¥cdot)||_{L^{2^{k}}(¥Omega)}$
$c_{1}$
$k$
(11)
$ z_{k}(t_{0})¥leq¥max$
$¥{||u_{t}(t_{0}, ¥cdot)||_{L^{2^{k}}(¥Omega)}, (c_{1}^{¥prime}2^{k(1+n/2)})^{2^{-k}}z_{k-1}(t_{0})¥}$
which is exactly similar to the estimate (8) for . Since we have also compactness of the family ( . estimate (22) in [2] or (12) below), then the same reasoning as in Theorem A finishes our proof. Remark 1. Sufficient conditions (related to the first eigenvalue of the , Dirichlet problem for the elliptic operator ) for which formulated in , were tend to zero when [2]. $u$
$¥{u_{t}(t, ¥cdot)¥}_{t¥geq 0}$
$¥mathrm{c}.¥mathrm{f}$
$L$
$||u(t, ¥cdot)||_{L^{2}(¥Omega)}$
$||u_{t}(t, ¥cdot)||_{L^{2}(¥Omega)}$
$ t¥rightarrow¥infty$
Remark 2. Often (cf. [1], [4]) solutions of parabolic problems tend to solutions of associated elliptic (stationary) problems. This situation holds for bounded solutions of (10)?(2) with independent of under additional . Compactness of trajectories ([2] Proposition 1) condition $L^{1}(¥Omega)$
$f$
$t$
$||¥mathrm{u}_{t}(t, ¥cdot)||_{L^{1}(¥Omega)}¥rightarrow 0$
(12)
$||u||_{C^{1+¥gamma l2,2+¥gamma}(R^{+}¥times¥overline{¥Omega})}¥leq$
const.
allows us to pass (on the sequence ) to the limit in the equation (10), to get consequence of Th. ): as a , , , ( ( ) $ t_{n}¥rightarrow¥infty$
$||u_{t}$
$t$
$¥cdot$
$||_{L^{¥infty}(¥Omega)}¥rightarrow 0$
$¥mathrm{B}$
$ t¥rightarrow¥infty$
$0=Lv+f(x, v)$ ,
where we denoted
$u(t_{n}, x)¥rightarrow v(x)$
in
$C^{2}(¥overline{¥Omega})$
$v=0$
on
$¥partial¥Omega$
,
.
There is also another way to prove the decay of the derivative solution. We have;
$u_{t}$
of the global
114
Tomasz DLOTKO
Proposition 2. Let be a global solution of the Dirichlet problem with continuous continuously differentiable in , , , and let . Then uniformly in when . $u$
$f$
$t$
$¥lim¥sup_{t¥rightarrow¥infty}¥partial f/¥partial u¥leq 0$
$u$
$¥nabla u$
$¥overline{¥Omega}$
$u_{t}¥rightarrow 0$
$(1)-(2)$
$¥lim_{¥mathrm{r}¥rightarrow¥infty}¥partial f/¥partial t=0$
,
$ t¥rightarrow¥infty$
Proof.
The proof follows from Theorem 1 p. 158 of [3]. It remains to note that since for global solutions the condition (12) holds, then composite , of the argument $(t, x, u, ¥nabla u)$ appearing in the equation , functions for are continuous and bounded uniformly in time. Then the proof is a direct consequence of the linear theory [3]. $¥partial f/¥partial t$
$¥partial.f/¥partial u$
$¥partial f/¥partial x_{i}$
$u_{tt}$
Below are listed recent publications close to our subject ([5], [6], [7]). References [1] [2]
Chafee, N., Asymptotic behaviour of a one-dimensional heat equation with Neumann boundary conditions, J. Differential Equations, 27 (1978), 266-273. Dlotko, T., Global solutions of reaction-diffusion equations, Funkcial. Ekvac., 30
(1987), 31-43. Friedman, A., Partial differential equations ofparabolic type, Prentice Hall, 1964. Hale, J. K., Stability and bifurcation in a parabolic equation, in Springer Lecture Notes 898, 1981, 143-153. [5] Lions, P. L., Structure of the set of steady-state solutions and asymptotic behaviour of semilinear heat equations, J. Differential Equations, 53 (1984), 362-386. [6] Lortz, D., Meyre-Spasche, R., and Stredulinsky, E. W., Asymptotic behavior of the solutions of certain parabolic equations, Comm. Pure Appl. Math., 37 (1984), 677-703. [7] Wiegner, M., On the asymptotic behaviour of solutions of nonlinear parabolic equations, Math. Z., 188 (1984), 3-22.
[3] [4]
nuna adreso: Institute of Mathematics Silesian University 40-007 Katowice Bankowa 14, Poland (Ricevita la 8-an de julio, 1985)