Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems

We are interested in the existence of distributional solutions for two types of nonlinear evolution problems, whose models are (1.1) and (1.2) below. In the first one the nonlinear reaction term depends on the solution with a slightly superlinear growth. In the second one we consider a first order t...

Повний опис

Збережено в:
Бібліографічні деталі
Дата:2004
Автори: Dall'Aglio, A., Giachett, D., Segura de Leon, S.
Формат: Стаття
Мова:English
Опубліковано: Інститут прикладної математики і механіки НАН України 2004
Назва видання:Український математичний вісник
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/124629
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems / A. Dall'Aglio, D. Giachett, S. Segura de Leon // Український математичний вісник. — 2004. — Т. 1, № 4. — С. 518-531. — Бібліогр.: 13 назв. — англ.

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-124629
record_format dspace
spelling nasplib_isofts_kiev_ua-123456789-1246292025-02-23T18:01:04Z Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems Dall'Aglio, A. Giachett, D. Segura de Leon, S. We are interested in the existence of distributional solutions for two types of nonlinear evolution problems, whose models are (1.1) and (1.2) below. In the first one the nonlinear reaction term depends on the solution with a slightly superlinear growth. In the second one we consider a first order term depending also on the gradient of the solution in a quadratic way. The two problems are strictly related from the point of view of the a priori estimates we can obtain on their solutions. We point out that no boundedness is assumed on the data of the problems. This implies that the methods involving sub/super-solutions do not apply, and we have to use some convenient test-function to prove the a priori estimates. 2004 Article Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems / A. Dall'Aglio, D. Giachett, S. Segura de Leon // Український математичний вісник. — 2004. — Т. 1, № 4. — С. 518-531. — Бібліогр.: 13 назв. — англ. 1810-3200 2000 MSC. 35B33, 35B45, 35K20, 35K55, 35K57. https://nasplib.isofts.kiev.ua/handle/123456789/124629 en Український математичний вісник application/pdf Інститут прикладної математики і механіки НАН України
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
language English
description We are interested in the existence of distributional solutions for two types of nonlinear evolution problems, whose models are (1.1) and (1.2) below. In the first one the nonlinear reaction term depends on the solution with a slightly superlinear growth. In the second one we consider a first order term depending also on the gradient of the solution in a quadratic way. The two problems are strictly related from the point of view of the a priori estimates we can obtain on their solutions. We point out that no boundedness is assumed on the data of the problems. This implies that the methods involving sub/super-solutions do not apply, and we have to use some convenient test-function to prove the a priori estimates.
format Article
author Dall'Aglio, A.
Giachett, D.
Segura de Leon, S.
spellingShingle Dall'Aglio, A.
Giachett, D.
Segura de Leon, S.
Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
Український математичний вісник
author_facet Dall'Aglio, A.
Giachett, D.
Segura de Leon, S.
author_sort Dall'Aglio, A.
title Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
title_short Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
title_full Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
title_fullStr Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
title_full_unstemmed Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
title_sort semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems
publisher Інститут прикладної математики і механіки НАН України
publishDate 2004
url https://nasplib.isofts.kiev.ua/handle/123456789/124629
citation_txt Semilinear parabolic equations with superlinear reaction terms, and application to some convection-diffusion problems / A. Dall'Aglio, D. Giachett, S. Segura de Leon // Український математичний вісник. — 2004. — Т. 1, № 4. — С. 518-531. — Бібліогр.: 13 назв. — англ.
series Український математичний вісник
work_keys_str_mv AT dallaglioa semilinearparabolicequationswithsuperlinearreactiontermsandapplicationtosomeconvectiondiffusionproblems
AT giachettd semilinearparabolicequationswithsuperlinearreactiontermsandapplicationtosomeconvectiondiffusionproblems
AT seguradeleons semilinearparabolicequationswithsuperlinearreactiontermsandapplicationtosomeconvectiondiffusionproblems
first_indexed 2025-11-24T06:18:49Z
last_indexed 2025-11-24T06:18:49Z
_version_ 1849651499139858432
fulltext Український математичний вiсник Том 1 (2004), № 4, 518 – 531 Semilinear Parabolic Equations with Superlinear Reaction Terms, and Application to Some Convection-diffusion Problems Andrea Dall’Aglio, Daniela Giachetti and Sergio Segura de León (Presented by A. E. Shishkov) Abstract. We are interested in the existence of distributional solutions for two types of nonlinear evolution problems, whose models are (1.1) and (1.2) below. In the first one the nonlinear reaction term depends on the solution with a slightly superlinear growth. In the second one we consider a first order term depending also on the gradient of the solution in a quadratic way. The two problems are strictly related from the point of view of the a priori estimates we can obtain on their solutions. We point out that no boundedness is assumed on the data of the problems. This implies that the methods involving sub/super-solutions do not apply, and we have to use some convenient test-function to prove the a priori estimates. 2000 MSC. 35B33, 35B45, 35K20, 35K55, 35K57. Key words and phrases. Nonlinear parabolic problems, convection- diffusion problems, superlinear reaction term, gradient term with quad- ratic growth, existence, a priori estimates. 1. Introduction In this paper we are interested in solving nonlinear parabolic problems of the type    vt − ∆v = f(x, t) ( 1 + |v| ∣∣ log |v| ∣∣α) in Ω×]0, T [; v(x, t) = 0 on ∂Ω×]0, T [; v(x, 0) = v0(x) in Ω. (1.1) Received 12.11.2003 ISSN 1810 – 3200. c© Iнститут прикладної математики i механiки НАН України A. Dall’Aglio, D. Giachetti and S. Segura de León 519 Here 0 < α < 1, f(x, t) ∈ Lr(0, T ;Lq(Ω)) for convenient r and q, and v0 ∈ L2(Ω). This type of problems is strictly related to parabolic convection- diffusion problems whose model is    ut − ∆u = β(u)|∇u|2 + g(x, t) in Ω×]0, T [; u(x, t) = 0 on ∂Ω×]0, T [; u(x, 0) = u0(x) in Ω, (1.2) where β(s) is a continuous function which grows like an arbitrary power of s at ±∞. In these model examples it is easy to see that it is possible to perform a change of unknown function in (1.2), i.e., v = Ψ(u) (see (2.6) below for the definition of the function Ψ(s)), and reduce problem (1.2) to a problem which is similar to (1.1). Nevertheless, we would like to consider also more general situations where one has a nonlinear pseudomonotone operator as a principal part in (1.2) and a general first order term which grows quadratically with respect to the gradient. In this case one cannot change the unknown function, but the previous remark suggests the use of convenient test functions which replace this technique. We give an existence result of distributional solutions for (1.1) via test-function method under the hypothesis that f(x, t) ∈ Lr(0, T ;Lq(Ω)), where the exponents r and q belong to a part of the so-called Aronson- Serrin region in the (r, q)-plane, part which also depends on the value of the parameter α in (1.1). As far as the initial datum is concerned, we assume that v0 ∈ L2(Ω). The a priori estimates that we will be able to obtain on a sequence {vn} of approximating solutions of (1.1) will also provide a priori esti- mates for a sequence of approximate solutions un for problem (1.2), and therefore an existence result of distributional solutions for this problem. Besides this application, the result obtained for problem (1.1) seems to have an autonomous interest. Let us recall some results concerning the existence of solutions of the nonlinear heat equation    vt − ∆v = λh(v) in Ω×]0, T [; v(x, t) = 0 on ∂Ω×]0, T [; v(x, 0) = v0(x) in Ω, (1.3) with 0 ≤ v0 ∈ L∞(Ω), while h(s) is a positive function. In [5] Brezis, 520 Semilinear Parabolic Equations with... Cazenave, Martel and Ramiandrisoa proved that if +∞∫ ds h(s) < +∞ , then there is no solution for large λ. This shows that one cannot hope to prove global existence for (1.1) if α > 1. On the other hand, if +∞∫ ds h(s) = +∞ , and if the data v0(x) and f(x, t) are bounded, it is easy to prove the existence of a global solution using sub/super-solutions independent on x. However this method does not work if either v0 or f is unbounded. One of the aims of this paper is to present some results in this case. Another interesting remark is the comparison with the case α = 0 and f(x, t) = f(x): it is well known (see, for instance, [11]) that if f(x) ∈ Lq(Ω), with q ≥ N/2, there exists a solution of the following linear heat equation with singular potential    vt − ∆v = f(x)v in Ω×]0, T [; v(x, t) = 0 on ∂Ω×]0, T [; v(x, 0) = v0(x) in Ω, while, if q < N/2, one can have instantaneous and complete blow-up (for instance this happens if f(x) = λ/|x|2, with λ large enough (see [2], and [6] for more general potentials f(x)). In this case our result (see Theorem 2.1 below) states that one can allow α > 0, but has to pay a price by assuming the stronger condition q > N/[2(1−α)] (see Remark 2.1 below). As far as the quasi-linear problem (1.2) is concerned, existence results have been given in [3] under the assumptions that the data f(x, t) and u0(x) are bounded, via a sub/super-solution method. For more general data f and u0, the result we obtain for (1.2) (see Theorem 2.2 in next Section) improves previous results proved in a wider framework in [7]. In that paper (see also [8] and [10]) a special condition is assumed which prevents one from considering functions β(s) which tend to +∞ for s→ ±∞. The plan of the paper is as follows: the next section is devoted to stating the assumptions and the main results of the paper. In Section 3 we will define the approximate problems and prove the related a priori estimates. In the final section we will study the limiting process. A. Dall’Aglio, D. Giachetti and S. Segura de León 521 2. Assumptions and main results Let Ω be a bounded open set in RN , T > 0, QT = Ω×]0, T [, ΣT = ∂Ω×]0, T [. We will denote by Lq(Ω), 1 ≤ q ≤ +∞, the usual Lebesgue spaces. If X is a Banach space, we will denote by Lq(0, T ;X) the usual evolution spaces (see, for instance, [4]). We will sometimes write ‖f‖q instead of ‖f‖Lq(Ω), and ‖f‖r,q instead of ‖f‖Lr(0,T ;Lq(Ω)). Moreover C will denote a positive constant which only depends on the data of the problem. Its value may be different from line to line. We are interested in studying the following two types of nonlinear evolution problems:    vt − div (a(x, t, v)∇v) = F (x, t, v) in QT ; v(x, t) = 0 on ΣT ; v(x, 0) = v0(x) in Ω, (2.1) and    ut − div (a(x, t, u)∇u) = B(x, t, u,∇u) in QT ; u(x, t) = 0 on ΣT ; u(x, 0) = u0(x) in Ω, (2.2) where in both cases the principal part satisfies • The function a : QT × R → RN2 is a Carathéodory function; that is, it is measurable with respect to (x, t) for all s ∈ R and continuous in s for almost all (x, t) ∈ QT ; moreover there exist two positive constants ν and M such that [a(x, t, s) · ξ] · ξ ≥ ν|ξ|2 (2.3) and |a(x, t, s) · ξ| ≤M |ξ| hold for almost all (x, t) ∈ QT for all (s, ξ) ∈ R × RN . We first consider problem (2.1) and state the assumptions and our results. We will assume that: • The function F : QT × R → R satisfies the Carathéodory condi- tions; moreover there exist a constant α, with 0 < α < 1, and a positive measurable function f(x, t) ∈ Lr(0, T ;Lq(Ω)), with q, r > 1 , q > N 2 max { 1 1 − α , r r − 1 } , (2.4) 522 Semilinear Parabolic Equations with... such that |F (x, t, s)| ≤ ( 1 + |s| ∣∣ log |s| ∣∣α ) f(x, t) ; • v0(x) ∈ L2(Ω). Remark 2.1. If f ∈ L∞(0, T ;Lq(Ω)), with q > N/[2(1 − α)], and in particular if f(x, t) = f(x) ∈ Lq(Ω), then obviously there exists r < ∞ such that (2.4) holds. We can now state the first of our main existence results: Theorem 2.1. Under the above hypotheses, problem (2.1) admits at least one distributional solution v ∈ L∞(0, T ;L2(Ω)) ∩ L2(0, T ;H1 0 (Ω)). Remark 2.2. The condition q > Nr 2(r − 1) is the same one which ensures the local boundedness of the solutions of the equation vt − ∆v = f(x, t) , f ∈ Lr(0, T ;Lq(Ω)) , as shown by Aronson and Serrin in [1]. We now turn our attention to the second quasi-linear problem (2.2). We will assume that: • The function B : QT × R × RN → R satisfies the Carathéodory conditions; moreover there exist two positive constants λ and C1, and a positive measurable function g(x, t) ∈ Lr(0, T ;Lq(Ω)), with q, r > 1 , q > N 2 max { λ+ 1 , r r − 1 } , such that |B(x, t, s, ξ)| ≤ C1(1 + |s|λ)|ξ|2 + g(x, t) ; (2.5) In order to state the assumption on the initial datum, we define two auxiliary functions by γ(s) = C1 ν s∫ 0 (1 + |σ|λ) dσ = C1 ν ( s+ |s|λ+1 λ+ 1 sign s ) , A. Dall’Aglio, D. Giachetti and S. Segura de León 523 and Ψ(s) = s∫ 0 e|γ(σ)| dσ . (2.6) The assumption on u0 reads as follows • Ψ(u0(x)) ∈ L2(Ω). Theorem 2.2. Under the above hypotheses, problem (2.2) admits at least one distributional solution u such that Ψ(u) ∈ L∞(0, T ;L2(Ω)) ∩ L2(0, T ;H1 0 (Ω)) . (2.7) Remark 2.3. Recalling that Ψ′(s) ≥ 1, the estimate (2.7) implies u ∈ L∞(0, T ;L2(Ω)) ∩ L2(0, T ;H1 0 (Ω)) . Remark 2.4. It is easy to show that for every δ > λ + 1 there exists Cδ > 0 such that |Ψ(s)| ≤ Cδ(1 + e|s| δ ) for every s ∈ R. It follows that the assumption on the initial datum of problem (2.1), i. e., Ψ(u0) ∈ L2(Ω), is satisfied if ∫ Ω e2|u0|δ dx <∞ for some δ > λ+ 1, and, a fortiori, if u0 ∈ L∞(Ω). 3. A priori estimates We first consider problem (2.1). For n ∈ N, we define the following approximate problems    (vn)t − div (a(x, t, vn)∇vn) = Tn(F (x, t, vn)) in QT ; vn(x, t) = 0 on ΣT ; vn(x, 0) = Tn(v0(x)) in Ω, (3.1) where Tn(s) = min { n,max{s,−n} } is the usual truncation at levels ±n. It is well known that problem (3.1) admits at least one weak solution vn ∈ C0([0, T ];L2(Ω)) ∩ L2(0, T ;H1 0 (Ω)). Proposition 3.1. Under the assumptions of Theorem 2.1, there exists a constant C, depending only on the data of the problem, such that ‖vn‖L∞(0,T ;L2(Ω)) + ‖vn‖L2(0,T ;H1 0 (Ω)) ≤ C . 524 Semilinear Parabolic Equations with... Proof. We first observe that, under the hypothesis (2.4), we can assume that the exponent r satisfies rα ≤ 1. Indeed, if this is not the case, we can replace r by a smaller value which still satisfies (2.4), and apply the usual inclusions between Lebesgue spaces. We multiply the equation (3.1) by vn and integrate on Ω, for t fixed. Using the assumptions (2.3) and (2.4) we obtain 1 2 d dt ∫ Ω v2 n dx+ ν ∫ Ω |∇vn|2 dx ≤ ∫ Ω ( 1 + |vn| ∣∣ log |vn| ∣∣α ) |vn| f dx ≤ C [∫ Ω f + ∫ Ω v2 n ∣∣ log |vn| ∣∣αf dx ] . The last integral can be estimated as follows. For δ ∈]0, 1[, to be chosen hereafter, we can write ∫ Ω f v2 n ∣∣ log |vn| ∣∣α dx = ∫ Ω f |vn|2δ|vn|2(1−δ) ∣∣ log |vn| ∣∣α dx ≤ ‖f(t)‖q [∫ Ω |vn|2 ∗ dx ] 2δ 2∗ [∫ Ω |vn|2ρ(1−δ) ∣∣ log |vn| ∣∣αρ dx ] 1 ρ , where ρ > 1 is defined by 1 ρ + 2δ 2∗ + 1 q = 1 . Using Young’s and Sobolev’s inequalities, one obtains, for every ε > 0, ∫ Ω f v2 n ∣∣ log |vn| ∣∣α dx ≤ ε [∫ Ω |vn|2 ∗ dx ] 2 2∗ + C(ε) ‖f(t)‖ 1 1−δ q [∫ Ω |vn|2ρ(1−δ) ∣∣ log |vn| ∣∣αρdx ] 1 ρ(1−δ) ≤ εC ∫ Ω |∇vn|2 dx+C(ε) ‖f(t)‖ 1 1−δ q [∫ Ω |vn|2ρ(1−δ) ∣∣ log |vn| ∣∣αρ dx ] 1 ρ(1−δ) . A. Dall’Aglio, D. Giachetti and S. Segura de León 525 Choosing ε = ν/(2C), and δ such that 1 1−δ = r, that is, δ = r−1 r , we obtain: 1 2 d dt ∫ Ω v2 n dx+ ν ∫ Ω |∇vn|2 dx ≤ C ( ‖f(t)‖1 + ‖f(t)‖rq [∫ Ω |vn| 2ρ r ∣∣ log |vn| ∣∣αρ dx ] r ρ ) . (3.2) Since q > Nr 2(r−1) , from the definition of δ and ρ one can check that ρ < r. Therefore there exists an increasing and convex function η(s) : [0,+∞) → [0,+∞) such that η(0) = 0 and η ( |Ω| s 2ρ r (log s)αρ ) ∼ Cs2 for s→ +∞. By Jensen’s inequality η (∫ Ω |vn| 2ρ r ∣∣ log |vn| ∣∣αρ dx ) ≤ 1 |Ω| ∫ Ω η(|Ω| |vn| 2ρ r ∣∣ log |vn| ∣∣αρ) dx ≤ C (∫ Ω v2 n dx+ 1 ) . This implies ∫ Ω |vn| 2ρ r ∣∣ log |vn| ∣∣αρ dx ≤ η−1  C (∫ Ω v2 n dx+ 1 )  . (3.3) It is easy to see that the function η satisfies η−1(s) ∼ s ρ r (log s)αρ for s→ +∞. Therefore, (3.3) implies [∫ Ω |vn| 2ρ r ∣∣ log |vn| ∣∣αρ dx ] r ρ ≤ CH (∫ Ω v2 n dx ) , (3.4) where H(s) = 1 + s| log s|αr . (3.5) If we define ξn(t) = ∫ Ω [vn(t)] 2 dx , 526 Semilinear Parabolic Equations with... we have proved that ξ′n(t) ≤ C [ ‖f(t)‖1 + ‖f(t)‖rqH(ξn(t)) ] ≤ C ( ‖f(t)‖1 + ‖f(t)‖rq ) [1 +H(ξn(t))] . Since ‖f(t)‖1 + ‖f(t)‖rq is an integrable function on ]0, T [, the last in- equality yields G(ξn(t)) −G(ξn(0)) ≤ C , (3.6) where G(s) = s∫ 0 dσ 1 +H(σ) . Recall than we are assuming that αr ≤ 1, therefore the function 1/(1 + H(s)) is not integrable on [0,+∞). Since G(ξn(0)) is bounded, the esti- mate (3.6) provides a uniform estimate of vn in L∞(0, T ;L2(Ω)). Using (3.2) and (3.4), after integration with respect to time, one obtains an estimate in L2(0, T ;H1 0 (Ω)). We can now turn our attention to proving an estimate for problem (2.2). Once again we have to define a sequence of approximate problems:    (un)t − div (a(x, t, un)∇un) = Tn(B(x, t, un,∇un)) in QT ; un(x, t) = 0 on ΣT ; un(x, 0) = u0,n(x) in Ω. (3.7) Notice that the prescribed datum at time t = 0 is not simply the trun- cation of u0 as in the previous proposition, but is a function u0,n ∈ L∞(Ω) ∩H1 0 (Ω) such that 1 n ‖u0,n‖H1 0 (Ω) → 0 as n→ ∞ , Ψ(u0,n) → Ψ(u0) a.e. and strongly in L2(Ω) . The existence of such a sequence may be proved by truncation and con- volution. These assumptions are required in order to prove the strong convergence of the gradients ∇un (see [8] and [7]). Proposition 3.2. Under the assumptions above, there exists a constant C, depending only on the data of the problem, such that ‖Ψ(un)‖L∞(0,T ;L2(Ω)) + ‖Ψ(un)‖L2(0,T ;H1 0 (Ω)) ≤ C . A. Dall’Aglio, D. Giachetti and S. Segura de León 527 To prove Proposition 3.2, we need the following elementary lemma: Lemma 3.1. For every α > λ λ+1 , there exists a constant Cα such that e|γ(s)| ≤ Cα ( 1 + |Ψ(s)| ∣∣ log |Ψ(s)| ∣∣α) for every s ∈ R. Proof. It suffices to show that lim s→+∞ eγ(s) Ψ(s) (log Ψ(s))α < +∞ , and this follows easily from a repeated application of De L’Hôpital’s rule. Proof of Proposition 3.2. Let us take e|γ(un)|Ψ(un) as test function in (3.7). Recalling the assumptions on the terms of the equation and inte- grating on Ω, we obtain for every fixed t: 1 2 d dt ∫ Ω Ψ(un) 2 dx+ ν ∫ Ω |∇Ψ(un)|2 dx + C1 ∫ Ω |∇un|2e|γ(un)|Ψ(un)(1 + |un|λ)sign un dx ≤ C1 ∫ Ω |∇un|2e|γ(un)||Ψ(un)|(1 + |un|λ) dx + ∫ Ω ge|γ(un)||Ψ(un)| dx . Under our hypotheses on g, we can find α such that λ λ+1 < α < 1 and such that (2.4) holds. Then, using Lemma 3.1, after cancellation we can estimate the last integral in the previous formula as follows: ∫ Ω ge|γ(un)||Ψ(un)| dx ≤ Cα ∫ Ω g|Ψ(un)| (1 + |Ψ(un)|) ∣∣ log |Ψ(un)| ∣∣α dx ≤ C (∫ Ω g dx+ ∫ Ω g|Ψ(un)|2 ∣∣ log |Ψ(un)| ∣∣α dx ) . From here, reasoning exactly as in the proof of Proposition 3.1, we obtain the differential inequality ξ′n(t) ≤ C ( ‖g(t)‖1 + ‖g(t)‖rqH(ξn(t)) ) , 528 Semilinear Parabolic Equations with... where ξn(t) = ∫ Ω Ψ(un(t)) 2 dx , and H(s) is defined by (3.5). In view of the assumptions on the initial datum, the desired estimate follows immediately. Remark 3.1. From Proposition 3.2 and Gagliardo-Nirenberg’s interpo- lation result (see, for instance, [9], Chapter I, Proposition 3.1), it follows that Ψ(un) is bounded (uniformly with respect to n) in Lρ(0, T ;Lσ(Ω)) for every ρ and σ such that 2 ≤ σ ≤ 2N N − 2 , 2 ≤ ρ ≤ ∞ and N σ + 2 ρ = N 2 . Proposition 3.3. There exists a positive constant C, depending only on the data of the problem, such that ∫ {|un|>k} ∣∣TnB(x, t, un,∇un) ∣∣ ≤ C‖gχ {|un|>k} ‖r,q + C ∫ Ω∩{|u0,n|>k} |Ψ(u0,n)| (3.8) for every n∈N and k≥0. In particular the sequence {TnB(x, t, un,∇un)} is bounded and equi-intergable in L1(QT ). Proof. The estimate (3.8) follows easily from the following inequality: ∫ {|un|>k} (1 + |un|λ)|∇un|2 ≤ C‖gχ {|un|>k} ‖r,q + C ∫ Ω∩{|u0,n|>k} |Ψ(u0,n)| (3.9) To prove (3.9), we multiply the approximate problems (3.7) by hk(un), where hk(s) = χ{|s|>k}(s)sign (s) ( e|γ(s)|−γ(k) − 1 ) , and integrate over QT . If we define φk(s) = s∫ 0 hk(σ) dσ , A. Dall’Aglio, D. Giachetti and S. Segura de León 529 we obtain, using the assumptions (2.3) and (2.5), ∫ Ω∩{|un(T )|>k} φk(un(T )) − ∫ Ω∩{|u0,n|>k} φk(u0,n)+ + C1 ∫ {|un|>k} (1 + |un|)λ|∇un|2e|γ(un)|−γ(k) ≤ C1 ∫ {|un|>k} (1 + |un|)λ|∇un|2 ( e|γ(un)|−γ(k) − 1 ) + ∫ {|un|>k} g ( e|γ(un)|−γ(k) − 1 ) . Dropping positive terms, this implies C1 ∫ {|un|>k} (1 + |un|)λ|∇un|2 ≤ ∫ {|un|>k} ge|γ(un)| + ∫ Ω∩{|u0,n|>k} φk(u0,n) ≤ ∥∥g χ{|un|>k} ∥∥ r,q ∥∥∥e|γ(un)| ∥∥∥ r′,q′ + ∫ Ω∩{|u0,n|>k} φk(u0,n) . From Lemma 3.1 it follows that e|γ(s)| ≤ C ( 1 + [Ψ(s)]2 ) , so that ∥∥∥e|γ(un)| ∥∥∥ r′,q′ ≤ C ( 1 + ‖Ψ(un)‖2 2r′,2q′ ) . It is easy to check that the exponent ρ = 2r′ and σ = 2q′ satisfy N σ + 2 ρ > N 2 . Therefore, applying the usual embeddings between Lebesgue spaces and Remark 3.2, the last norm is bounded. Moreover 0 ≤ φk(u0,n) ≤ |Ψ(u0,n)|, therefore (3.9) is completely proved. 4. Proof of main results Proof of Theorem 2.1. By Proposition 3.1, the sequence {vn} is bounded in L∞(0, T ;L2(Ω)) and in L2(0, T ;H1 0 (Ω)). Moreover, by the equation, 530 Semilinear Parabolic Equations with... {(vn)t} is bounded in L2(0, T ;H−1(Ω)) + L1(QT ). Using standard com- pactness results for evolution spaces (see for instance [13]), we can extract a subsequence (still denoted by {vn}) which converges to some func- tion v strongly in L2(QT ) ∩ C0([0, T ];W−1,s(Ω)), for every s < N/(N − 1), and weakly in L2(0, T ;H1 0 (Ω)). It is easy to pass to the limit in the weak formulation of (3.1), thus showing that v solves (2.1) in the sense of distributions. Moreover, since there is strong convergence in C0([0, T ];W−1,s(Ω)), the initial datum has a meaning. Proof of Theorem 2.2. As in the previous proof, taking Propositions 3.2 and 3.3 into account, we can assume that, passing to a subsequence, un ⇀ u weakly in L2(0, T ;H1 0 (Ω)), un → u strongly in L2(QT ) ∩ C0([0, T ];W−1,s(Ω)), for every s < N N − 1 , Ψ(un) ⇀ Ψ(u) weakly in L2(0, T ;H1 0 (Ω)) and ∗-weakly in L∞(0, T ;L2(Ω)). Moreover it can be proved (exactly as in Propositions 6.2 and 6.3 of [7], where a technique introduced in [12] is developed) that ∇Tkun → ∇Tku a. e. and in L2(QT ), for every k > 0, ∇un → ∇u a. e. and in Lq(QT ), for every q such that 1 ≤ q < 2, and, using Proposition 3.3, TnB(x, t, un,∇un) → B(x, t, u,∇u) in L1(QT ). The existence result follows easily. We remark that the initial datum has sense since u ∈ C([0, T ];W−1,s(Ω)). The authors have recently proved that the solutions obtained in Theo- rem 2.1 and Theorem 2.2 are bounded if the initial datum is also bounded. This fact agrees with the result by Aronson-Serrin (see Remark 2.2) and its proof will be published elsewhere. References [1] D. G. Aronson and J. Serrin, Local behavior of solutions of quasilinear parabolic equations // Arch. Rat. Mech. Anal. 25 (1967), 81–122. [2] P. Baras and J. Goldstein, The heat equation with a singular potential // Trans. Amer. Math. Soc. 294 (1984), 121–139 A. Dall’Aglio, D. Giachetti and S. Segura de León 531 [3] L. Boccardo, F. Murat and J.-P. Puel, Existence results for some quasilinear parabolic equations // Nonlinear Anal. T. M. A. 13 (1989), 373–392 [4] H. Brezis, Analyse Fonctionnelle. Masson, Paris, 1983 [5] H. Brezis, T. Cazenave, Y. Martel and A. Ramiandrisoa, Blow up for ut −∆u = g(u) revisited // Adv. Diff. Eq. 1 (1996), 73–90 [6] X. Cabré and Y. Martel, Existence versus explosion instantanée pour des équations de la chaleur linéaires avec potentiel singulier // C. R. Acad. Sci. Paris, Série I 329 (1999), 973–978 [7] A. Dall’Aglio, D. Giachetti, C. Leone and S. Segura de León, Quasi-linear parabolic equations with degenerate coercivity having a quadratic gradient term, Ann. Inst. H. Poincaré Anal. Non. Linéaire, to appear. [8] A. Dall’Aglio, D. Giachetti and J.-P. Puel, Nonlinear parabolic equations with natural growth in general domains, Boll. Un. Mat. Ital. Sez. B, to appear. [9] E. DiBenedetto, Degenerate parabolic equations. Springer Verlag, New York, 1993 [10] V. Ferone, M. R. Posteraro and J. M. Rakotoson, Nonlinear parabolic problems with critical growth and unbounded data // Indiana Univ. Math. J. 50 (2001), No 3, 1201–1215 [11] O. A. Ladyzhenskaya, V. A. Solonnikov, and N. N. Ural’tseva, Linear and Quasi- linear Equations of Parabolic Type. American Mathematical Society, Providence, 1968 [12] R. Landes and V. Mustonen, On parabolic initial-boundary value problems with critical growth for the gradient // Ann. Inst. H. Poincaré Anal. Non Linéaire 11 (1994), 135–158 [13] J. Simon, Compact sets in the space Lp(0, T ; B) // Ann. Mat. Pura Appl. 146 (1987), 65–96 Contact information A. Dall’Aglio and D. Giachetti Dipartimento di Metodi e Modelli Mate- matici, Università di Roma “La Sapienza”, Via Antonio Scarpa 16, I-00161 Roma, Italy E-Mail: aglio@dmmm.uniroma1.it, giachetti@dmmm.uniroma1.it S. Segura de León Departament d’Anàlisi Matemàtica, Universitat de València, Dr. Moliner 50, 46100 Burjassot, València, Spain E-Mail: Sergio.Segura@uv.es