Abstract
The method of Galerkin approximations is employed to prove the existence of a strong global (in time) solution of a doubly nonlinear parabolic equation in an unbounded domain. The second integral identity is established for Galerkin approximations, and passing to the limit in it an estimate for the decay rate of the norm of the solution from below is obtained. The estimates characterizing the decay rate of the solution as obtained here are used to derive an upper bound for the decay rate of the solution with respect to time; the resulting estimate is pretty close to the lower one.
Bibliography: 17 titles.
Export citation and abstract BibTeX RIS
This research was carried out with the support of the Russian Foundation for Basic Research (grant nos. 10-01-00118-a and 13-01-00081-a).
§ 1. Introduction
Let be a domain in , . We consider the first mixed problem
for a doubly nonlinear parabolic equation in the cylindrical domain .
Here,
The existence and uniqueness of the solution of this problem was examined by Lions [1], Raviart [2], Grange and Mignot [3], Bamberger [4], Alt and Luckhaus [5], Bernis [6] and others. For the most part, these problems were dealt with on a bounded domain and a bounded time interval with arbitrary . The strong solution of this problem in a bounded domain was found by Raviart by replacing the evolution derivative with a difference relation. The existence of a weak solution in an unbounded domain was proved by Bernis, who applied a limiting procedure to the solutions built by Grange and Mignot on bounded domains. However, working with a weak solution involves a difficulty consisting, for example, in examining the decay of the solution as . Bamberger proved the uniqueness of a strong positive solution of the problem.
In this paper, Galerkin approximations are employed in the conventional way to construct a strong solution of the problem across the whole time interval (the domain may be unbounded). In [7], this type of solution was constructed on a bounded interval for any . The ideas of [7] were used in [8] to construct a solution of an anisotropic parabolic equation for on a bounded time interval.
Derivation of the estimates we need for the Galerkin approximations is facilitated by the fact that they are smooth functions; then passing to the limit we can extend these estimates to the solution of problem (1.1), (1.2). In particular, for the equation
on a bounded domain , with , it will be shown that
Conceptually, these results continue the studies started by Gushchin [9] on the behaviour of solutions of linear parabolic equations for large . The left-hand side of estimate (1.4) with was obtained by Alikakos and Rostamian in [10] for the Cauchy problem. Tedeev [11] obtained analogous estimates for the solution of a high-order quasi-linear equation. Sharp, two-sided estimates for the decay rate of the norm of the solution to a linear and a quasi-linear parabolic equation in an unbounded domain were established by Kozhevnikova [12] and Karimov and Kozhevnikova [13], and for an anisotropic parabolic equation with , by Kozhevnikova and Mukminov [14]. Nearly sharp estimates for the solution of an anisotropic parabolic equation with were obtained in [8] (a solution on a bounded time interval).
§ 2. Statements of the main results
We define the space to be the completion of in the norm
where
We also let , where and , denote the completion of in the norm
Definition. A generalized solution of problem (1.1), (1.2) is a function in the space for all and satisfying the identity
for all .
Here and below, denotes the value of a distribution at an element , where is a domain in or in . However, in formula (2.1), it is assumed that .
We shall write a function as or when there is no ambiguity.
Throughout we assume that .
Theorem 1. Let for all . Then there exists a generalized solution of problem (1.1), (1.2) such that
for all . Here, .
Bamberger [4] defined a strong solution by the condition and proved, in particular, that if
where is a bounded domain, then the problem has at most one strong solution. He also showed that this solution vanishes for , provided that , .
It is easily checked that the solution in Theorem 1 is strong.
The paper [10] puts forward a series of assertions (not all of them are fully justified) about the behaviour of the solution of the Cauchy problem for the equation
The following upper and lower bounds for the solution of the Cauchy problem were obtained under the assumption that and .
The upper estimates read as follows:
The lower estimates are as follows:
In [15] the following estimate for the solution of the Cauchy problem for equation (2.6) for any is given, under the assumption that :
The paper [11] looks at the problem
where
(In fact [11] deals with somewhat more general equations.) A solution in the space , , is considered under the assumption that . In particular, the following bounds were proved for a bounded domain :
Above we have only given basic results on the behaviour of the solutions of equations (2.6), (2.7). For more detailed results the reader is referred to the book [16].
The remaining assertions of the present paper are established under the assumption that . In Theorems 2–5, a domain may be bounded or unbounded. For an unbounded , it is assumed that the initial function has bounded support
where .
Theorem 2. Let . Assume that condition (2.9) is satisfied and that the initial function is bounded. Then there exist a positive number and a bounded solution of problem (1.3), (1.2) such that, for all ,
Below we shall need the characteristic function of the layer between two concentric spheres
when or the limit indices will be omitted: , .
The proof of the lower bound when is unbounded depends heavily on the following fact.
Theorem 3. Let , let (2.9) be satisfied, and let be a generalized solution of problem (1.3), (1.2) with bounded initial function . Assume that satisfies properties (2.2)–(2.5). Then is also bounded and there exist positive numbers and such that
for all , .
To prove Theorem 3 we shall employ the maximum principle (see § 5).
For , estimate (2.11) becomes trivial. Hence, we need a different estimate in order to bound the decay rate of the solution from above. We shall assume that the origin lies on the boundary of . Also let , .
Consider the following geometrical characteristic of the unbounded domain . We set
Assume that the domain satisfies the condition
For example, this condition is satisfied for a domain lying inside some cone.
Theorem 4. Let and let (2.9) and (2.13) be satisfied. Then there exist positive numbers , and a generalized solution of problem (1.3), (1.2) such that
for all , .
Let be an arbitrary positive function satisfying the inequality
for all . Clearly, such a function always exists.
Theorem 5. Let . Assume that (2.9) and (2.15) are satisfied. Then there exist a positive number and a solution of problem (1.3), (1.2) in a domain lying inside some cone (of aperture ) such that
If the domain satisfies the condition
then we can set , , and (2.16) assumes the form
Taking sufficiently small, can be made arbitrarily close to zero. The exponent in the last estimate is close to the exponent in the lower bound (2.10).
As an example, consider the domain of revolution
, with a positive function . The only requirement imposed on is that the set be a domain and that
For these domains, one can easily prove the relation
from which condition (2.17) follows.
§ 3. Proof of the existence theorem
Under the above assumptions on we have
In particular, .
3.1. The case .
Consider a sequence of linearly independent functions whose linear hull is dense in . We set . Galerkin approximations to the solution will be sought in the form
where the functions are determined from the equations
The numbers will be chosen later. We claim that equations (3.1) can be solved for the derivatives . Clearly, they are as follows:
For each , the coefficient matrix
is a Gram matrix of the system of linearly independent vectors , , and hence is invertible. From equations (3.1) with initial conditions , chosen so that in , we find the functions . At first, these functions are determined on a short time interval. Then, the fact that the Galerkin approximations are bounded (to be proved later) lets us define them on an infinite time interval. The numbers will be chosen so as to have as .
Now we proceed to estimate the Galerkin approximations. Multiplying equations (3.1) by and summing, this gives
Integrating with respect to shows that
The last integral on the right is bounded by a constant independent of , thanks to the above convergences. Further, for we have
Hence, from (3.2) and Gronwall's lemma we see that the sequence is bounded in the spaces and for all by the constant , which remains unchanged if the sequence is replaced by a larger one .
Now we multiply equations (3.1) by and sum:
Integrating with respect to shows that
Integrating the last term by parts, we obtain
Note that
Further, since is bounded in the space , we have
Now we find from (3.4) that, for any , the sequence is bounded in and that the sequence is bounded in . Using the diagonal process, we can now choose a subsequence which is weakly convergent in these spaces. For brevity of notation we shall omit the subscript :
The convergence holds for any , and the limit functions agree in their common domain of definition. Consequently, we have convergence for any .
Below it will be shown that , and that the function is a generalized solution of problem (1.1), (1.2). We will proceed in three steps, which we will also need later in the subsequent analysis.
Step 1. The sequence is bounded in the space on any finite interval:
We fix a countable dense subset of . We can assume that . Given any bounded domain with smooth boundary, the embedding is compact. Hence, using the diagonal process we can choose a subsequence converging strongly in for all natural . Choosing, if necessary, a further subsequence (and dropping the subscripts), we assume that almost everywhere in for any . In particular, for we have almost everywhere in .
For the next step we require the following result.
Lemma 1. Let the sequence satisfy the following properties:
- 1)converges almost everywhere in for each and some ;
- 2)the sequence is bounded in for each .
Then it contains a subsequence converging to the function in the space , and almost everywhere in .
Proof. The sequence is equicontinuous in in the space :
Next, the sequence is bounded in the space . Consequently, it contains a subsequence converging weakly in for the same as above. Along with the almost everywhere convergence in , this implies strong convergence in for any (see [1], Ch. 1, § 12.2). We again write for .
For a bounded domain , it follows easily from (3.5) that is a uniformly Cauchy sequence in the norm :
Choosing a finite grid of small spacing and then increasing and , we ensure that the right-hand side is uniformly small in .
Thus, we see that in for any . The convergence also takes place in . Hence, one may select a subsequence converging in almost everywhere. The lemma is proved.
Step 2. We apply Lemma 1 to the sequence . Since is arbitrary and , we can apply the diagonal process to select a subsequence converging in almost everywhere. Hence, the sequence converges almost everywhere in to (see [1], Ch. 1, § 1.4, Lemma 1.3). We quote this result.
Lemma 2. Let be a domain in or in . Assume that a sequence converges to almost everywhere in and is bounded in . Then weakly in .
Clearly, this result, which is stated in [1] for bounded domains, also holds for unbounded domains.
From Lemma 2 it also follows that weakly in for any . Given a fixed , the convergence in implies that there is a subsequence convergent almost everywhere in for all . Hence,
The sequence is bounded in the space , and hence we can choose a subsequence such that
also, if , then the constant depends only on . The constant is monotone in , so this proves (2.2).
Further,
Passing to the limit, we get
It follows that .
We claim that the sequence is bounded in . Indeed,
Hence we can assume that weakly in . As a result, . We also have
In fact, in Step 1 we noted that converges almost everywhere in . Now, in , , by Lemma 1, thereby proving (3.8).
Step 3. We will now prove the equality . For this we need some integral relations. We multiply equation (3.1) by the smooth function , integrate over , and let , denoting by in the resulting expression:
We note that
as is bounded in and . Clearly, any function from can be approximated by linear combinations
Hence, (3.9) also holds for functions from the space . So, in view of (3.8), will be a generalized solution of problem (1.1), (1.2) if we prove that .
Note that implies that and that , . Substituting the function into (3.9), this gives
Our next argument depends on the operator being monotone. It is readily verified that
From equations (3.1) it is easily seen that
Hence,
Consequently, since , by (3.6)
Applying (3.10) shows that
Setting , , , we have
Letting , we see that for any . Hence, .
3.2. The case .
Galerkin approximations to the solution will be sought in the previous form, but now the functions are determined from the equations
Here, the functions are introduced for regularization; the numbers will be chosen later. We claim that equations (3.13) can be solved with respect to the derivatives . Clearly, they are as follows
For each , the coefficient matrix
is a Gram matrix of the system of linearly independent vectors , , and hence is invertible. From equations (3.13) with initial conditions , chosen so that in the space , we find the functions .
Now we estimate the Galerkin approximations. Multiplying equations (3.13) by and summing, we find that
We have and , and hence, integrating with respect to shows that
Since the sets are bounded, we may choose to satisfy
Now the last integral on the right of (3.14) is bounded by a constant independent of , thanks to the above convergences. As before, inequality (3.3) holds. Hence, using (3.14) and Gronwall's lemma it follows that the integrals are uniformly bounded in and , and hence so is the sequence in the spaces and for all .
We now multiply equations (3.13) by and sum:
Integrating over gives
As before, . In addition, . Setting
it follows from (3.16) that the sequence is bounded in and the sequence is bounded in . In view of the above results, we can choose a subsequence that is weakly convergent in the spaces specified below. For brevity, the subscript will be dropped:
Repeating Steps 1–3, as adapted to the new setting, we will be able to prove that .
Proceeding as in Step 1, we can assume (dropping the subscripts) that the sequence converges almost everywhere in for each .
At Step 2, Lemma 1 is applied to the sequence to produce a subsequence converging almost everywhere in . Since , are arbitrary, the diagonal process can be applied to extract a subsequence converging almost everywhere in . The sequence of inverse functions converges pointwise. Now, Lemma 2 shows that the sequence converges almost everywhere to in . Hence, we have proved that
Moreover, .
As in the case , (3.6) and (3.7) are satisfied for a fixed .
For , the sequence is bounded in . Indeed, using (3.16),
Hence, we can assume that weakly in , , and so . This proves (2.4).
At Step 3, our aim is to show that . For this purpose we shall need some integral relations. We multiply equations (3.13) by a smooth function , integrate with respect to , and then integrate the first term by parts. Now, writing for , we have
We note that
since is a bounded sequence in . Hence, we can choose a subsequence to ensure that weakly in and weakly in . That the limit functions have exactly this form is justified by the fact that, as we said above, the subsequence converges almost everywhere in and almost everywhere in for (see (3.6)). Now making , in view of (3.8) it follows that
As in the case , we show that this equality is satisfied for all functions . Relation (3.17) also means that is a generalized solution of problem (1.1), (1.2), provided that .
Substituting into (3.17), we see that
Our next argument depends on being monotone. Using equations (3.14), it is easily seen that
Employing inequality (3.15), by (3.19) we have
Now (3.12) follows because and . Finally, a similar argument as in the case shows that .
The proof of Theorem 1 is complete.
§ 4. Proof of Theorem 2
Assume now that and that a domain is bounded. Our aim is to estimate the decay rate of the solution to problem (1.3), (1.2) from below as .
4.1. The case .
We define
dropping the subscripts where possible. After differentiating with respect to , for formula (3.2) assumes the form
Differentiating (3.4), we obtain
The following estimates hold:
Applying the Cauchy-Schwarz inequality for the inner product in , this gives
Hence,
Using (4.1), we rewrite this as follows:
As a result, , or, on integrating,
Consequently,
or
It follows that
Thus,
In view of (3.7), for fixed and if the domain is bounded one may choose a subsequence which converges strongly in the space . The functions
lie in the linear hull of the functions . In a finite-dimensional space all the norms are equivalent, and hence
We choose numbers so as to have . Now
Passing to the limit in (4.4) as , we obtain
4.2. The case .
In this case we set
Note that (3.15) implies that .
Differentiating (3.14) with respect to and rewriting it for , this establishes
From (3.16) it follows that
For any , clearly
In view of (3.15),
Now (4.3) follows by minimizing the right-hand side over .
The rest of the proof is similar to the case .
We claim that the estimate (4.5), in the case of a bounded domain , is sharp. To prove this, we need the following Friedrichs-Steklov-type inequality
We note in passing that the following Friedrichs-Steklov inequality holds for an unbounded domain located inside a cone:
For , we have
Differentiating (3.10) or (3.18) with respect to and employing (4.9) for , we see that
Solving this differential inequality, we obtain the estimate
which proves that (4.5) is sharp.
4.3. An unbounded domain.
Our aim here is to derive (4.5) for the solution of problem (1.3), (1.2) in the case when is unbounded.
We let , , denote the solutions in with fixed initial function . We can assume that these solutions are extended by zero outside . Using (3.7), we have the estimate
Using the properties of the solutions listed in Theorem 1, we can choose a subsequence such that weakly in as for all , and then, applying Steps 1–3, prove that the limit function is a generalized solution of problem (1.3), (1.2).
Further, given a fixed we can assume that weakly in as . As the embedding is compact, strongly in as for any .
By (2.11), for any there exists such that
Estimate (4.5) holds for . Now
Using the strong convergence in , we pass to the limit as and then as . Thus, we have now proved (2.10) for an unbounded domain as well.
The proof of Theorem 2 is complete.
§ 5. Proof of Theorem 3
For the sake of completeness we recall the maximum principle to be used in the proof of Theorem 3. Note that the inequality is not used in the proof of the maximum principle.
Proposition. Let , , . Then a generalized solution of problem (1.1), (1.2) with properties (2.2)–(2.5) is bounded:
Proof. The truncation of a function is known to lie in the same space (see, for example, [17], Ch. 2, § 4, Lemmas 4.1–4.3).
Assume first that . That the function can be substituted into (3.17) is justified by the passage to the limit; here is a Lipschitz function with bounded support. We have
We choose , where
In this case, . Note that . We have , and hence
Further,
where . Thus, (5.2) implies that
Without loss of generality we may assume that . Since , we have , . Hence, , (see, for example, [17], Ch. 2, § 1, inequality (2.12)). So, the right-hand side of (5.3) tends to zero as . Therefore,
for almost all . Hence, almost everywhere in . This proves (5.1) since and are arbitrary.
When , we have to substitute into (3.9) and proceed according to the above scheme. For example, we transform the first integral in (3.9) (putting ):
here , , is a monotonic function: . The rest of the argument proceeds as above.
Proof of Theorem 3. Assume first that . We set , where the function will be chosen later. Substituting into (3.17) with , we have
By choosing the functions so that the supports of the functions and are disjoint, we can assume that the right-hand side of (5.4) is zero. Next, using Young's inequality,
For , the same result is obtained by substituting into (3.9).
Let , for , for , and for . Also let be such that . Then
where we have set , and the superscript will be omitted. It follows that
Setting
we have
In view of (3.10), the function is bounded; that is, for all , . Hence, by (5.6),
Proceeding by induction and applying Stirling's formula, this gives
Let . Then
We choose so that . Then
As a result,
Consequently,
which implies (2.11). If cannot be chosen as required, the inequality holds, and estimate (2.11) is trivial.
The proof of Theorem 3 is complete.
§ 6. An upper estimate for the norm of the solution
Our aim in this section is to prove Theorems 4 and 5.
6.1. Proof of Theorem 4.
Let , , be an absolutely continuous function equal to 1 for , to 0 for , linear for and satisfying the equation
(the constant will be specified below). Solving this equation, we find, in particular, that
For any function it follows from the definition of that
Multiplying by and integrating over , we derive the inequality
which now holds for any function . For and , we rewrite the last inequality as follows
Substituting into (5.5), we have
Using (6.1), (6.2), one easily reduces (6.4) to the form
Applying (4.8) with and using (3.18), it follows that
Using (6.3) shows that
Taking and combining (6.5)–(6.7), we obtain
thereby proving inequality (2.14).
The proof of Theorem 4 is complete.
6.2. Proof of Theorem 5.
We choose a positive number . Setting
in view of (2.14), we have
By (4.8),
Setting with an appropriate factor , we have
Let denote a point in the interval such that . If for any , we let . Since the function is monotone nonincreasing for , we have . Now (6.8) implies that
Differentiating (3.18) with respect to , we see that
Next, in view of (6.9),
Solving this differential inequality, we obtain
The substitution into the last inequality shows that, for ,
Note that, for , the inequality holds and estimate (6.11) is also valid.
In (6.11) we set and use (2.15):
This proves inequality (2.16).
The proof of Theorem 5 is complete.