Abstract
In this paper, we prove uniqueness of renormalized solution for a class of doubly nonlinear parabolic problems.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
In the present paper, we establish the uniqueness for renormalized solutions for a class of doubly nonlinear parabolic equations, whose prototype:
In the problem (2) the framework is the following: \(\varOmega \) is a bounded open set of \({\mathbb {R}}^{N}\), with Lipschitz-continuous boundary \(\partial \varOmega \), \(N\ge 2\), \(T>0\), while the data \(f, F \, {\text {and}} \,b(u_{0})\) are respectively in \(L^{1}(Q_{T}), (L^{p'}(Q_{T}))^{N}\), \({\text {and}}\, L^{1}(\varOmega )\). b is strictly increasing \(C^{1}({\mathbb {R}})\)-function, unbounded of s. \(\triangle _{p}u\) is the p-Laplace operator, \(Q_{T}=\varOmega \times (0,T)\), data \(c(.,.),\ d(.,.), \gamma ,\, \delta \), will be defined later, see Sect. 2 (8–11).
Starting with the paper [9] the authors proved an existence result of a weak solutions for the non coercive problem 3 in the stationary case (\(b_t(u)=0\)) using the symmetrization method. More later Di Nardo et al. [10] have shown the existence of renormalized solution for the parabolic version, more precisely in the linear case (\(b(u) = u\)), and the uniqueness for such solution in the paper [6], Aberqi et al. [1, 2] have proved the existence of a renormalized solutions for 3 with more general parabolic terms \((b_t(x,u))\).
In the present work we prove the uniqueness or such solution, under some local control of Lipschitz coefficient (see Theorem 3.1).
In general, the concept of the weak solution is not sufficient to determine the solution physically observed due to the lack of uniqueness of the solution. It appears necessary to select among all the physically weak solutions feasible solution. The renormalized solutions allowed to have results of existence and uniqueness for certain equations that are not accessible within the solutions in the sense of distributions see the counter example given by Serrin [14] in the linear case, and Bénilan et al. [5] in the case of p-Laplacian operator. To overcome this difficulties we work in the framework of renormalized solutions (see 14–18), this notion was introduced by Diperna and Lions [12] in their study of Boltzmann equations, see also [4, 8].
The paper is concerned with giving a careful account on both existence and uniqueness of renormalized solution, we want to stress that, while the existence result follows a rather standard approximation argument see [1] due to the nonlinearity b(u) and non coercive terms \(c(x,t)|u|^{\gamma -1}u\) and \(d(x,t)|\nabla u|^{\delta -1}\) and the measure \(f -{\text {div}}(F)\).
In order to perform the uniqueness, the paper is planned in the following way. Section 2 is devoted to specify the assumptions on \(b,\, a,\, H,\, f,\) and \(b(u_0)\) and to give the definition of a renormalized solution of 3, and we prove some technical Lemmas whose play a crucial role to prove the uniqueness results. In Sect. 3, we prove that there exists a unique renormalized solution see Theorem 3.1.
2 Basic assumptions and definitions
In this section, we recall the definition of renormalized solutions to the following nonlinear parabolic problem:
where \(Q_{T}\)is the cylinder \(\varOmega \times (0,T)\), \(\varOmega \) is a bounded open set of \({\mathbb {R}}^{N}\), with \(N\ge 2\), \(T>0\), \(p>1\).
Throughout this paper, we assume the following assumptions hold true.
-
\( b:{\mathbb {R}}\rightarrow {\mathbb {R}}\) is strictly increasing \(C^{1}\)-function, \(b(0)=0\),
$$\begin{aligned} {\text { and }} 0< b_0 \le b'(s) \le b_1 \,\,\forall s \quad {\text {and with }}\quad b_1 < \left( {\frac{2}{\alpha }}\right) ^{\frac{p-1}{2}} \end{aligned}$$(4)where \(\alpha \) is a strictly real number defined below in (6).
-
\(a: \varOmega \times (0,T)\times {\mathbb {R}}\times {\mathbb {R}}^{N} \rightarrow {\mathbb {R}}^{N}\) be a Carathéodory function such that there is \(\alpha >0\), and for \(k>0\), there exists \(\nu _{k}>0\) and a function \(h_{k}\in L^{p'}(Q_{T})\) such that, \(\forall |s|\le k\), \(\forall (\xi ,\eta )\in {\mathbb {R}}^{2N}\), for a.e. \((x,t)\in Q_{T}\),
$$\begin{aligned}&|a(x,t,s,\xi )|\le \ \big (h_{k}(x,t)+|s|^{p-1}+|\xi |^{p-1}\big ) \end{aligned}$$(5)$$\begin{aligned}&a(x,t,s,\xi )\xi \ge \alpha |\xi |^{p}, \end{aligned}$$(6)$$\begin{aligned}&(a(x,t,s,\xi )-a(x,t,s,\eta )\cdot (\xi -\eta )>0,\ \xi \ne \eta , \end{aligned}$$(7) -
\(\phi : \varOmega \times (0,T)\times {\mathbb {R}}\rightarrow {\mathbb {R}}^{N}\) be a Carathéodory function such that
$$\begin{aligned} |\phi (x,t,s)|\le & {} c(x,t)|b(s)|^{\gamma }, \end{aligned}$$(8)$$\begin{aligned} c(.,.)\in (L^{r}(Q_T))^{N},\quad r= & {} \frac{p+N}{p-1}, \quad {\text {and}}\quad \gamma =\frac{(N+2)(p-1)}{N+p}, \end{aligned}$$(9) -
\(H: \varOmega \times (0,T)\times {\mathbb {R}}^{N} \rightarrow {\mathbb {R}}^{N}\) be a Carathéodory function such that
$$\begin{aligned}&|H(x,t,\xi )|\le d(x,t)|\xi |^{\delta }, \end{aligned}$$(10)$$\begin{aligned}&{\text {with}} \, \delta =\frac{p+N(p-1)}{N+2}, \, d(.,.) \,{\text {belonging a suitable Lorentz space}}\, L^{N+2,1}(Q_{T}),\nonumber \\ \end{aligned}$$(11)
Moreover we assume that the source terms
Definition 2.1
A measurable function u defined on \(Q_{T}\) is called a renormalized solution of (3) if:
and if, for every function S in \(W^{2,\infty }({{\mathbb {R}}})\) which is piecewise \(C^1\) and such that \(S'\) has a compact support, we have in the sense of distributions
and
Remark 2.1
Note that conditions (4), (14), and (15) allow to define \(\nabla u\) and \(\nabla b(u)\) almost everywhere in \(Q_{T}\).
Remark 2.2
Note that for a renormalized solution, due to (15), each term in (17) has a meaning in \(L^{1}(Q_{T})+L^{p'}((0,T);W^{-1,p'}(\varOmega ))\). Indeed, since \(|T_{k}(b(u))|\le k\), we can choose k such that \(supp(S')\in [-k,k]\). Then the properties of S, the functions \(S'\) and \(S''\) are bounded in \({\mathbb {R}}\). We have \(S(b(u))=S(T_{k}(b(u)))\in L^{p}((0,T);W_{0}^{1,p}(\varOmega ))\) and \(\frac{\partial S(b(u))}{\partial t}\in D'(Q_{T})\). The term \(S'(b(u))a(x,t,u,\nabla u)\) identifies with \(S'(T_{k}(b(u)))a(x,t,u,\nabla b^{-1}(T_{k}(b(v))))\) a.e. in \(Q_T\), where \(v=b(u)\) and \(u=b^{-1}(T_{k}(b(v)))\) in \(\{|b(u)|\le k\}\), by (4) and (5) we have
Using (6, 17) it follows that \(S'(b(u))a(x,t,u,\nabla u)\in (L^{p'}(Q_T))^{N}\). In view of (4, 6, 9, 10, 15, 19), we obtain:
Consequently, we have \(\displaystyle \frac{\partial S(b(u))}{\partial t} \in L^{p'}(0,T;W^{-1,p'}(\varOmega )) + L^1(Q_T)\) and \(S(b(u))\in L^{p}(0,T,W^{1,p}_{0}(\varOmega ))\cap L^{\infty }(Q_T)\). Which implies that S(b(u)) belongs to \(C([0,T];L^{1}(\varOmega ))\) so the initial condition (18) makes sense.
The existence theorem of renormalized solution of (3):
Theorem 2.1
Under assumptions (4)–(13) there exists at least a renormalized solution u of problem (3).
Proof of Theorem 2.1
The existence theorem of renormalized solution of (3) is proved in ([10]) in the linear case \((\mathrm{b}(\mathrm{u})=\mathrm{u})\) and by author in ([1, 2]). \(\square \)
Remark 2.3
To prove the uniqueness result for the problem (3), due to, the presence of a general monotone operator \(A(u)=-div(a(x,t,u,\nabla u)\), and the non-linearity b(u), a standard approach does not feasible. To overcome this difficulty we draw upon the idea included in ([6]), for which we recall some basic results that will be a key point.
Lemma 2.1
(see [7]) Let v be a function in \(W^{1,p}_{0}(\varOmega )\cap L^{2}(\varOmega )\) with \(p\ge 1\). Then there exists a positive constant C, depending on \(N,\ p\), such that
for every \(\theta \) and \(\gamma \) satisfying
An immediate consequence of the previous result:
Corollary 2.1
Let \(v\in L^{p}((0,T);W^{1,p}(\varOmega ))\cap L^{\infty }((0,T);L^{2}(\varOmega ))\), with \(p\ge 1\). Then \(v\in L^{\sigma }(\varOmega )\) with \(\sigma =p(\frac{N+2}{N})\) and
Lemma 2.2
Let \(\omega \) be an open subset of \({\mathbb {R}}^N,\) \(N \ge 1,\) \(F \in L^{p}(\varOmega ),\) and \(\overline{u}: \varOmega \rightarrow [0, + \infty ]\) and \(\overline{v} : \varOmega \rightarrow [0, + \infty ]\) be two measurable functions. Then there exists a sequence \(n_{j}\)(related for simplicity as n) of real numbers such that
Proof
see [6], Lemma 6. \(\square \)
Lemma 2.3
Under the assumptions (4)–(13), any renormalized solution u of (3) satisfies the following estimate for any \(n \ge 1\) and any \(0< \delta < 1\)
with \(\lim _{n \rightarrow + \infty } \lim _{\delta \rightarrow 0}\epsilon (n,\delta )=0\).
Proof
Using the same proof (Lemma 5, p. 356, [6]), adding the analysis of the two lower order terms \(\phi \) and H, and taking into account the non-linearity b(u). \(\square \)
Let \(S_n\in W^{1,\infty }({\mathbb {R}})\)be the function defined by
Since \(supp S'_n \subset [-n-2, n+2]\), by setting \(S=S_{n+1}\), \(\forall n>0\) in (17), we have in the sense of distributions
For a real numbers \(n>0\) and \(0<\delta <1\), using the admissible test function
in (22), we get
Remark that \(R_{n}^{\delta }(b(u))=R_{n}^{\delta }(S_{n+1}(b(u)))\) as soon as \(0<\delta <1\), and then we have
where \(\displaystyle \widetilde{R}_{n}^{\delta }(s)=\int _{0}^{S_{n+1}(b(u_0))}R_{n}^{\delta }(s)\,ds\) for any \(n>1\) and any \(0<\delta <1\).
The definitions (21) and (23) permit to obtain from (24) and (25) that for any \(n >1\) and any \(0<\delta <1\),
\(\star \) Estimates for the first lower order: Note that the terms involving \(\phi (x,t,u)\) in (22) not equal to 0 for any \(n>0\), and any \(\delta >0\) (as in equation 24 in [6]). By (8,9), (21, 24) and using Hölder inequality, Gagliardo–Nirenberg (see Corollary 2.1) together with Young inequality yields to
The coercive character (6) of a and choosing the norm of c(x, t) small enough, we get
in the same way
\(\star \) Estimates for the second lower order: By Hölder inequality (in Lorentz space), we have
and by using (Lemma A.1, see ([10]) in Appendix) we have
Finally by using Young inequality and the coercivity of a for the sixth term of the right hand (26), we obtain from (25) to (30) that for any \(n>1\) and any \(0<\delta <1\)
Since \(f\in L^{1}(Q_T)\), \(a(x,t,u,\nabla u)\nabla b(u)\in L^{1}(Q_T)\) and conditions (14), (16) we have
so that Lemma 2.3 is established.
3 Uniqueness of renormalized solution
In this section, we assume a local control of Lipschitz coefficients to prove the following uniqueness theorem
Theorem 3.1
Assume that assumptions (4)–(13) hold true and moreover that for any compact set D of \({\mathbb {R}}\), there exists \(L_D \in L^{p'}(Q_T)\) and \(\rho _D>0\) such that \(\forall s, \overline{s} \in D\)
for almost every \((x,t) \in Q_T\) and for every \(\xi \in {{\mathbb {R}}}^{N}.\) Then the problem (3) has a unique renormalized solution.
For the sake of shortness, we denote by \(\{|u| \le k\}\) (resp. \(\{|u| < k\}\)) the measurable subset \(\{(x,t)\in Q_T; |u(x,t)| \le k\}\) (resp. \(\{(x,t)\in Q_T; |u(x,t)| < k\}.\)) Moreover the explicit dependence in x and t of the functions a, \(\phi \) and H will be omitted so that \(a(x,t,u,\nabla u) = a(u,\nabla u),\) \(\phi (x,t,u)=\phi (u)\).
Proof of Theorem 3.1
Let u and v be two renormalized solutions of (3) for the same data f and F and initial condition \(b(u_0)\). We define a smooth approximation of \(T_n\) by \(T_n^\sigma \) and
\(\square \)
Using \(\frac{1}{k} T_k(T_n^\sigma (b(u))-T_n^\sigma (b(v)))\) as test function in the difference of Eq. (17) for u and v in which we take \(S=T_n^\sigma ,\) we obtain
where
For any \(k>0\), \(n>0\), \(\sigma >0\). Now we will pass to the limit of each term of (35) when \(\sigma \) and k tends to 0, and n tend to \(+\infty \).
-
For the first term in the right-hand sid of (35), upon of Lemma 2.4 ([3]), and due to
\(T_n^\sigma (b(u))(t=0)= T_n^\sigma (b(v))(t=0)= T_n^\sigma (b(u_0))\) a.e. in \(\varOmega ,\) we have
$$\begin{aligned}&\lim _{k \rightarrow 0} \lim _{\sigma \rightarrow 0}\frac{1}{k} \int _0^T \Big < \frac{\partial (T_n^\sigma (b(u))-T_n^\sigma (b(v)))}{\partial t} ; T_k(T_n^\sigma (b(u))-T_n^\sigma (b(v)))\Big > dt\\&\quad = \lim _{k \rightarrow 0} \lim _{\sigma \rightarrow 0} \frac{1}{k} \int _{Q_T} \overline{T_k} (T_n^\sigma (b(u))-T_n^\sigma (b(v))) \,dt \,dx\\&\quad = \int _{Q_T} |T_n(b(u))-T_n(b(v))| \,dt \,dx \end{aligned}$$where again \(\overline{T_k} (z) = \int _0^z T_k(r) dr.\) We deduce from the above equality that for almost any \(t \in (0,T)\)
$$\begin{aligned}&\lim _{n \rightarrow +\infty }\lim _{k \rightarrow 0} \lim _{\sigma \rightarrow 0}\int _0^T \Big < \frac{\partial (T_n^\sigma (b(u))-T_n^\sigma (b(v)))}{\partial t} ; T_k(T_n^\sigma (b(u))-T_n^\sigma (b(v)))\Big > dt \nonumber \\&\quad = \int _{Q_T} |b(u)-b(v)| dt dx \end{aligned}$$(36) -
For a fixed \(n>0\), we studied the behavior of \(I^{\sigma }_{1,n}\) when \(\sigma \) and k tends to 0:
We have, when \(\sigma \rightarrow 0\), \(T_n^\sigma )'(r) \rightarrow \chi _{|r|\le n}\) a.e. in \(Q_T\) and strongly in \(L^{q}(Q_T)\) for any \(q<+\infty \) and \(T_n^\sigma )(r) \rightarrow T_n(r)\) a.e. in \(Q_T\) and strongly in \(L^{p}(Q_T)\). Since supp \((T_n^\sigma )' \subset [-n-\sigma , n + \sigma ] \) we have
Rewritten \(I_{1,n}\) as \(I_{1,n} = I_{11} + I_{12} + I_{13} + I_{14} + I_{15},\) where
We use the monotonicity of \(a(s, \xi )\) with respect to \(\xi \) and \(b'(s) > 0\) for all \(s \in {\mathbb {R}},\) we obtain
It remains to prove that \(I_{12}\) goes to zero as k goes to zero. Indeed using the local Lipschitz condition (31) and (35) on a we get
Due to regularity of u, v, and \(L_{D}\) we have
Since \(\chi _{\{|b(u) - b(v)| \le k\}}\) tends to zero almost everywhere in \(Q_T\) as k goes to zero, the Lebesgue dominated convergence allows us to conclude that, for all \(n\ge 1\):
We denote by \(C_n\) the compact subset \([b^{-1}(-n-1), b^{-1}(n+1)],\) and remark that due to the local Lipschitz character of \(b',\)there exists a positive real number \(\beta _n\) such that
for all \(r_1\) and \(r_2\) lying in \(C_n.\) Using now (4) again leads to
for all \(r_1\) and \(r_2\) lying in \(C_n,\) then \(|b'(u) - b'(v)| \le \frac{ k \beta _n }{b_0} \) on
\(\{|b(u) - b(v)| \le k, |b(u)|\le n, |b(v)|\le n\},\) and in view (4) we obtain
Due to regularity of u, v, \(\nabla u\) and \(\nabla v\) we have
Since \(\chi _{\{|b(u) - b(v)| \le k, b(u)\ne b(v), |b(u)|\le n, |b(v)|\le n\}}\) tends to zero almost everywhere in \(Q_T\) as k goes to zero, the Lebesgue dominated convergence allows us to conclude that, for all \(n\ge 1\):
In view of the definition of \(T_n,\) we have
It is possible to obtain
Using (35), (6) we conclude that
Similarly to the argument of limit \(I_{14},\) we conclude
We obtain from (37) to (42) that
then
The limit of \(I_{2,n}^\sigma \): In view of the definition of \(T_n^\sigma \) it is possible to obtain
Using (44) and the estimates of Lemma 2.3, then we obtain
The limit of \(I_{3,n}^\sigma \): We first write that for almost any \(t\in (0,T)\)
where
and
We estimate \(I_{31}\) and \(I_{32}\) by (8) we obtain
and similarly
Applying Lemma 2.2 in (46) and (47), we obtain:
Finally, since the function \(\phi \) is locally Lipschitz continuous, we have for some positive \( L_{D}\) element of \(L^{p'}(Q_T)\)
by (4) we obtain
Since \( L_{C}\) belongs to \(L^{p'}(Q_T)\) and due to (15) the function \( L_{D}(x,t) (|\nabla T_n(b(u))|+ |\nabla T_n(b(v))|)\) belongs to \(L^1(Q_T).\) Using \(\chi _{\{|T_n(b(u))- T_n(b(v))| \le k \}}\) tends to 0 almost everywhere in \(Q_T\) as k goes to 0 and is bounded by 1, the Lebesgue dominated convergence theorem leads to
Using (48) and (49) we obtain:
The limit of \(I_{4,n}^\sigma \): We have for any \(\sigma \) and \(k > 0\)
Using the Lemma 2.2, we get
The limit of \(I_{5,n}^\sigma \): Using Lebesgue’s theorem, the definition of \(T_{n}^{\sigma },\) it is possible to conclude that
Then
The limit of \(I_{6,n}^\sigma \): We have for almost every \(t\in (0,T)\)
where
and similarly
Applying Lemma 2.2 in (53) and (54), we obtain:
The limit of \(I_{7,n}^\sigma \): We have for any \(\sigma \) and \(k > 0\)
Using the Lemma 2.2, we get
The proof of Theorem 3.1 is complete.
Change history
31 May 2017
An erratum to this article has been published.
References
Aberqi, A., Bennouna, J., Mekkour, M., Redwane, H.: Renormalized solution for a nonlinear parabolic equation with lower order terms. Aust. J. Math. Anal. Appl. 10(1,15), 1–15 (2013)
Aberqi, A., Bennouna, J., Redwane, H.: A nonlinear parabolic problems with lower order terms and measure data. Thai J. Math. 14(1), 115–130 (2016)
Boccardo, L., Murat, F., Puel, J.-P.: Existence of bounded solutions for nonlinear elliptic unilateral problems. Ann. Mat. Pura Appl. 152, 183–196 (1988)
Boccardo, L., Giachetti, D., Diaz, J.-I., Murat., F.: Existence and regularity of renormalized solutions of some elliptic problems involving derivatives of nonlinear terms. J. Differ. Equ. 106, 215–237 (1993)
Bénilan, Ph, Boccardo, L., Gallouët, Th, Gariepy, R., Pierre, M., Vázquez, J.L.: An \(L^{1}\) theory of existence and uniqueness of solutions of nonlinear elliptic equations. Ann. Sc. Norm. Super. Pisa Cl. Sci. 22(4), 241–273 (1995)
Blanchard, D., Murat, F., Redwane, H.: Existence and uniqueness of a renormalized solution for fairly general class of non linear parabolic problems. J. Difer. Equ. 177, 331–374 (2001)
Brézis, H., Browder, F.E.: Strongly nonlinear parabolic initial-boundary value problems. Proc. Natl. Acad. Sci. USA 76, 38–40 (1976)
Dal Maso, G., Murat, F., Orsina, L., Prignet, A.: Renormalized solutions of elliptic equations with general measure data. Ann. Sc. Norm. Super. Pisa Cl. Sci. 28(4), 741–808 (1999)
Del Vecchio, T., Posteraro, M.R.: An existence result for elliptic equations with measure data. Adv. Differ. Equ. 1, 899–917 (1996)
Di Nardo, R., Féo, F., Guibé, O.: Existence result for nonlinear parabolic equations with lower order terms. Anal. Appl. 9(2), 161–186 (2011)
Di Nardo, R., Perrotta, A.: Uniquness results for nonlinear elliptic problems ith two lowers order terms. Bull. Sci. Math. 137, 107–128 (2013)
Diperna, R.-J., Lions, P.-L.: On the cauchy problem for the Boltzmann equations: global existence and weak stability. Ann. Math. 130, 285–366 (1989)
Landes, R.: On the existence of weak solutions for quasilinear parabolic initial-boundary problems. Proc. R. Soc. Edinb. Sect. A89, 321–366 (1981)
Serrin, J.: Pathological solution of elliptic differential equations. Ann. Sc. Norm. Super. Pisa Cl. Sci. 18, 385–387 (1964)
Simon, J.: Compact set in the space \(L^{p}(0, T, B)\). Ann. Mat. Pura Appl. 146, 65–96 (1987)
Author information
Authors and Affiliations
Corresponding author
Additional information
The original version of this article was revised: the first name of the second author was incorrect. Now, it has been corrected.
An erratum to this article is available at https://doi.org/10.1007/s11587-017-0328-x.
Rights and permissions
About this article
Cite this article
Aberqi, A., Bennouna, J. & Hammoumi, M. Uniqueness of renormalized solutions for a class of parabolic equations. Ricerche mat 66, 629–644 (2017). https://doi.org/10.1007/s11587-017-0317-0
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11587-017-0317-0