Abstract
In this paper we study the regularity and the behavior in time of the solutions to a quasilinear class of noncoercive problems whose prototype is
In particular we show that under suitable conditions on the vector field E, even if the problem is noncoercive and although the initial datum \(u_0\) is only an \(L^{1}(\Omega )\) function, there exist solutions that immediately improve their regularity and belong to every Lebesgue space. We also prove that solutions may become immediately bounded. Finally, we study the behavior in time of such regular solutions and we prove estimates that allow to describe their blow-up for t near zero.
Similar content being viewed by others
1 Introduction
Let us consider the following parabolic problem
where \(\Omega \) is an open bounded subset of \({\mathbb {R}}^N\), \(N > 2\), and \(T>0\). We assume that \(a(x,t,s) : Q_{T} \times {\mathbb {R}}\rightarrow {\mathbb {R}}\) is a Carathéodory function (i.e., measurable in (x, t) for every s in \({\mathbb {R}}\), and continuous in s for almost every (x, t) in \(Q_{T}\)) such that
for almost every (x, t) in \(Q_{T}\) and for every s in \({\mathbb {R}}\), where \(\alpha \) and \(\beta \) are positive constants. On the initial datum we assume
The main difficulty in studying this problem is the presence of the lower order term \(-\mathrm{div}(u\,E(x,t))\) which makes the problem noncoercive. In addition, the initial datum has a very low regularity, so that the study of such an equation is more complicated.
In the stationary case this problem was studied in the sixties by Stampacchia (see [14, 15]) when E “is not too large”, and then by several authors until nowadays (see for example [2, 3, 6, 8] and the references therein).
In the evolution case existence and regularity results can be found in [4] (see also [9]).
We recall that if E is zero then a very surprising phenomenon appears: even if \(u_0\) is only a summable function, there exists a solution u of (1.1) that becomes immediately (i.e., for every \(t > 0\)) bounded and satisfies the same decay (or ultracontractive estimate) of the solution of the heat equation, i.e.,
The aim of this paper is to understand what happens when E is a nonzero measurable vector field; in particular, we want to understand if an immediate improvement in the regularity of the solutions appears or not. We will prove (see Sect. 2) that if E satisfies
for almost every (x, t) in \(Q_{T}\), then there is an immediate regularization since there exists a solution u of (1.1) which belongs to \(L^\infty (t_0,T;L^{q}(\Omega ))\) for every \(1< q <+\infty \) and \(t_0\) in (0, T). Besides, we will derive quantitative estimates “near \(t=0\)” (see Sect. 2.1) and for t large (see Sect. 2.2).
Moreover, we show that under stronger assumptions on E solutions become immediately bounded (see Sect. 3); furthermore, near “\(t = 0\)” the same behavior (1.4) of the solution of the heat equation holds true, but in this case the constant c in (1.4) depends also on E.
Finally, in Sect. 4 we study the less regular case of a vector field E such that
which may not belong to \(L^{\infty }(0,T;L^{N}(\Omega ))\) if 0 belongs to \(\Omega \).
Remark 1.1
We would like to thank the referee of the paper for pointing out that in the autonomous case, i.e., for the equation
with E in \(L^{\infty }(\Omega )\), the following holds:
-
the operator
$$\begin{aligned} A(u) = - \mathrm{div}(A(x)\nabla u) + \mathrm{div}(u\,E(x)) \end{aligned}$$is quasi accretive in \(L^{q}(\Omega )\) spaces (for every \(1 \le q \le \infty \)), and quasi m-accretive in \(L^{q}(\Omega )\) spaces (for every \(1 \le q < \infty \));
-
every mild solution u of (1.6) satisfies an \(L^{1}\)–\(L^{\infty }\) estimate similar to (1.4) thanks to Theorem 1.2 of [7].
2 An improvement of regularity
In this section we will show that even if the initial datum \(u_0\) is only assumed to be a summable function, if E satisfies (1.5) there exists a solution of (1.1) which belongs to every Lebesgue space \(L^{\infty }(t_0,T;L^{q}(\Omega ))\) for every \(1< q < +\infty \) and \(t_0 \in (0,T)\) (see Theorem 2.1 in Sect. 2.1 below). Moreover, we describe the blow-up of the \(L^{q}(\Omega )\)-norm of u(t) as t tends to zero and the behavior of the solution for t large.
We recall that here by a solution of (1.1) we mean a function u in \(L^{\infty }(0,T; L^{1}(\Omega ))\) and \(L^{1}(0,T;W^{1,1}_{0}(\Omega ))\) such that
for every \(\varphi \) in \(W^{1,\infty }(0,T;L^{\infty }(\Omega )) \cap L^{\infty }(0,T;W^{1,\infty }_{0}(\Omega ))\) such that \(\varphi (T) = 0\).
We recall that, thanks to the results of [4], under assumptions (1.3) and (1.5) there exists a solution u of (1.1), with u in \(L^{\infty }(0,T;L^{1}(\Omega )) \cap L^{q}(0,T;W^{1,q}_{0}(\Omega ))\), with \(q = \frac{N+2}{N+1}\) (see Lemma 3.2 of [4]). Furthermore, if |E| belongs to \(L^{\infty }(Q_{T})\) and \(u_{0}\) belongs to \(L^{\infty }(\Omega )\), there exists a bounded solution u of (1.1) such that (see Lemma 3.1 of [4])
2.1 Behaviour near zero
Theorem 2.1
Assume (1.2), (1.3) and (1.5). Then there exists a solution u of (1.1) satisfying
Furthermore, the following estimate holds:
where \(c_0\) is a constant depending only on \(t_0\), q, N, \(\alpha \), T and B. Moreover, for every \(q \ge 2\) and \(t_0 >0\) it results
Finally,
where \(c_1\) is a constant depending only on N, q, \(\alpha \), \(\gamma \), T and B (see formula (2.21)).
Proof
In [4] it is proved the existence of a solution u of (1.1), obtained as the almost everywhere limit of a sequence of weak solutions \(u_{n}\) in \(L^2(0,T;W^{1,2}_{0}(\Omega )) \cap C([0,T];L^2(\Omega ))\) of the following problems:
where \(T_n\) is the usual truncating function
and
Notice that since \(T_n(u_{n})\,E_{n}(x,t)\) belongs to \(L^{\infty }(Q_{T})\) then \(u_{n}\) also belongs \(L^{\infty }(Q_{T})\) by the results of [1]. We fix \(q > 1\), \(t > 0\), choose as test function in (2.5) \(v_n = |u_{n}|^{q-2}\,u_{n}\), and integrate on \(\Omega \). We obtain, using (1.2), and the fact that \(|T_{n}(s)| \le |s|\),
We now work with the right hand side; using Young inequality, as well as (1.5), we have
where \(C_1=\frac{\gamma ^2}{2\alpha }\). Therefore, part of the right hand side can be absorbed in the left hand one, to obtain
where \(C_2 = C_{2}(q) = C_1\,(q-1)\). We continue to work on the right hand side; we have, for some \(\rho > 0\), and thanks to Hölder inequality,
where, as usual, \(2^{*} = \frac{2N}{N-2}\) is the Sobolev embedding exponent. Since, by Sobolev embedding, we have
we therefore have that
We now choose \(\rho \) in such a way that
Such a choice is possible since B belongs to \(L^{N}(\Omega )\). Note that \(\rho \) does not depend on t. We therefore have that
where \(C_3\) is a constant depending only on \(\alpha \), q, \(\gamma \), \({\mathcal {S}}\) and B. Substituting this inequality in (2.8) we obtain, after simplifying equal terms,
Writing \(|\nabla u_{n}(t)|^{2}|u_{n}(t)|^{q-2} = \frac{4}{q^{2}}\,|\nabla [|u_{n}(t)|^{\frac{q}{2}}\mathrm{sign}(u_{n})]|^{2}\), and using Sobolev embedding, we arrive at
Since \(1< q < \frac{2^{*}}{2}\,q\), we can interpolate, to obtain
where \(\theta \) in (0, 1) is such that
We recall now (see the proof of Lemma 3.1 in [4]) that (2.1) holds true for \(u_{n}\), so that
Hence, from (2.14) it follows that
Therefore,
where we have assumed that \(u_0 \not \equiv 0\) (otherwise there is nothing to prove). Inserting (2.15) into (2.13), we obtain
where
Since
we thus have
where to simplify the notation we have set \(C=C_3\,q\). Define now
From (2.16) we have that y(t) is such that
where A and C are as above. Multiplying by \(\mathrm{e}^{-C\,t}\), we have
that is,
Define \(z(t) = \mathrm{e}^{-C\,t}\,y(t)\), so that we have
that is
Dividing by \(z(t)^{1+\delta }\), we have
which can be rewritten as
so that the function
or, simplifying \(\delta \), that the function
Thanks to (2.17), we have that, for every \(0< t < T\),
and thus,
that is,
Recalling the definition of z(t), we thus have that
that is
Hence, passing to the limit on n we deduce that, for almost every t in (0, T) we have
Recalling the value of A, we thus have
where \(C_4= \left( \frac{q{\mathcal {S}}C}{\alpha (q-1)}\right) ^{\frac{1}{\delta }}\). Since
we finally have that, for almost every t in (0, T)
Recalling that \(\mathrm{e}^{C\,\delta \,t} - 1 \ge C\,\delta \,t\) for every t (since \(t \mapsto \mathrm{e}^{C\,\delta \,t}\) is convex), we therefore have that
where
and hence (2.4) is proved.
To conclude the proof we observe that by (2.19) it follows that even if \(u_0 \in L^{1}(\Omega )\) it results
Moreover, it results
where \(c_{0}\) is a constant depending only on \(t_0\), q, N, \(\alpha \), T and B.
Finally, using the previous regularity in (2.12) we deduce that for every \(q \ge 2\)
\(\square \)
2.2 Behaviour for t large
We show here that our problem admits a global solution u (defined in all the set \(\Omega \times (0,+\infty )\)) and we study its behavior for t large. To this aim, we recall that by a global solution of (1.1) we mean a function u which solves (1.1) for every \(T>0\).
Theorem 2.2
Assume (1.3) and that (1.2) and (1.5) hold true in \(\Omega \times (0,+\infty )\). Let u be the solution of (1.1) given by Theorem 2.1. Then u can be extended to a global solution defined in \(\Omega \times (0,+\infty )\) (that we denote again u) satisfying
where \(c_0\) is as in (2.3) a constant depending only on \(t_0\), q, N, \(\alpha \) and B, and such that for every \(q \ge 2\) and \(t_0 >0\) we have
Moreover, if for some \(t_0 >0\) it results
or more generally, if
then
Proof
Let \(T>0\) be arbitrarily fixed and let u be the solution of (1.1) given by Theorem 2.1. Hence u is obtained as the almost everywhere limit in \(Q_T\) of a sequence \(u_{n}\) in \(L^{2}(0,T;W^{1,2}_{0}(\Omega )) \cap C([0,T];L^{2}(\Omega ))\) of weak solutions of (2.5). Notice that each \(u_{n}\) can be extended to a global solution of (2.5), that we denote again \(u_{n}\). Hence \(u_{n}\) belongs to \(L^{2}_{\mathrm{loc}}([0,+\infty );W^{1,2}_{0}(\Omega )) \cap C_{\mathrm{loc}}([0,+\infty );L^{2}(\Omega ))\), and solves (2.5) in every set \(Q_{\texttt {T}}\) for every \( \texttt {T}>0\) arbitrarily fixed. We show now that we can extend the solution u of (1.1) in \(Q_T\) to a global solution. To this aim let us denote with \(u_{n}^{(1)}\) the subsequence of \(u_{n}\) that converges almost everywhere in \(Q_{2T}\) to a weak solution \(u^{(2)}\) of (1.1) in \(Q_{2T}\). By construction \(u = u^{(2)}\) in \(Q_T\). Now, let us denote with \(u_{n}^{(2)}\) the subsequence of \(u_{n}^{(1)}\) that converges almost everywhere in \(Q_{3T}\) to a weak solution \(u^{(3)}\) of (1.1) in \(Q_{3T}\). By construction \( u^{(2)} = u^{(3)}\) in \(Q_{2T}\). Iterating this procedure, the function \(u(x,t) \equiv u^{(n)}(x,t)\) in \(Q_{nT}\) (for every integer n) is well defined and is a global solution of (1.1) which, by construction and thanks to Theorem 2.1, satisfies (2.24)–(2.26).
We show now that the global solution u constructed above, under further assumptions on E, satisfies the other estimates enounced in Theorem 2.2.
To this aim, proceeding as in the proof of (2.8), and integrating in time, we deduce that for every \(0<t_0<t_1<t_2\)
Now, thanks to the Poincaré inequality we deduce
where \(C_P\) is the Poincaré constant. We can rewrite the previous estimate in the following way
where we have set
Applying Proposition 3.2 in [11] we obtain
and if \(g_{n} \in L^1(t_0,+\infty )\) (again by Proposition 3.2) we get for every \(t>t_0\)
where
Notice that thanks to (2.25) it results
Suppose now that
Then
and
If instead one assumes that
then it results
and
Under either assumption (2.35) or (2.36) one therefore has that
Thus,
so that (recalling that the solution u of (1.1) is the limit of \(u_{n}\))
\(\square \)
3 \(L^{\infty }\)-regularity
In this section we prove the following result.
Theorem 3.1
Assume (1.2), (1.3), (1.5) and
Then there exists a solution u of (1.1) satisfying
Moreover,
where c depends only on \(\alpha \), \(\beta \), N, T and E.
Finally, if (1.2) and (1.5) hold true in \(\Omega \times (0,+\infty )\) then u can be extended to a global solution (that we denote again u) defined in all \(\Omega \times (0,+\infty )\); this extension coincides with the one given by Theorem 2.2; if |E| belongs to \(L^s_{\mathrm{loc}}(0,+\infty ;L^{r}(\Omega ))\) and (2.37) holds true (for example if (2.35) or (2.36) holds true) then it results
Proof
We recall that under the assumptions of Theorem 2.1 (see Sect. 2.1) there exists a solution u of (1.1) satisfying the regularity property (2.2) and estimate (2.3). Hence, if we assume that
then there exist \({\overline{s}} > 1\) and \({\overline{r}}\) (depending on r and s) such that it results
Notice that (3.5) is satisfied if for example \(|E| \in L^{\infty }(0,T;L^{r}(\Omega ))\) with \(r >N\).
Thus, if we assume (3.5) we can apply Theorem 8.1 at page 192 of [10] to the solution u of (1.1) constructed in Theorem 2.1 to conclude that
Moreover it is also possible to estimate the \(L^\infty \)-norm of u in dependence of the data and of the \(L^2\)-norm of u.
Now to complete the proof of the theorem we need to show the behavior of the \(L^\infty \)-norm of u near zero and for t large.
To this aim, we recall that applying Theorem 2 at page 18 of [1] we obtain the following estimate
where C is a constant depending only on the data (i.e. on \(\alpha \), \(\beta \), \(\theta \) and N) \(Q(\rho ) = R(\rho ) \times (t_0-\rho ^2,t_0)\), \(R(\rho ) \) is the open cube in \(R^N\) of edge lenght \(\rho \) centered in \(x_0 \in \Omega \), \(\rho >0\) such that \(Q(3 \rho ) \subset Q_T\), \(\Vert \cdot \Vert _{s_0,r_0,Q(3 \rho )}\) denotes the norm \( \Vert \cdot \Vert _{ L^{s_0}(t_0-(3 \rho )^2,t_0;L^{r_0}(K(3\rho )))}\) and \(\theta \in (0,1)\) is defined as follows
We conclude the proof distinguishing the two cases of t near zero and t large.
Case 1: t near zero. We observe that it results
which implies
Hence, choosing \((3 \rho )^2 = \frac{t_0}{2}\) and t arbitrarily in \((\frac{17}{18}t_0,t_0)\) (hence \(t = \theta _0 t_0\) with \(\theta _0 \in (\frac{17}{18},1)\)) and using estimate (2.4) we deduce
Here with C or c we denote positive constants, depending only on the data in the structure conditions and on E, which can vary from line to line. Notice that with such a choice of \(\rho \) the left hand side of (3.8) is greater then \(\Vert u(t)\Vert _{L^{\infty }(K(\rho ))}\). Hence, it remains to estimate the last term in the right hand side of (3.8).
Let us consider the particular case when \(|E| \in L^{\infty }(0,T;L^r(\Omega ))\) with \(r >N\) (i.e. (3.5) with \(s=+\infty \)). Then \(|uE| \in L^{\infty }(t_0,T;L^{{\overline{r}}}(\Omega ))\) for every \(N< {\overline{r}} < r\) for every \(t_{0}>0\), and we have
Hence, by (3.13) and (2.4) we deduce that
which implies, since \(\theta >0\) and being possible any choice of \({\overline{r}}\) satisfying \(N< {\overline{r}} < r\), the following estimate
Now, by (3.15), (3.12) and (3.8) we can conclude that for t in (0, T) it results
which implies, thanks to the arbitrariness of \(K(\rho )\),
We conclude the proof considering the general case when |E| satisfies (3.5) with \(s \not =+\infty \). In this case we estimate the last term in (3.8) as follows
for every \({\overline{s}}<s\) and \({\overline{r}} < r\) satisfying (3.6). Hence, by the previous estimate and (2.4) we deduce, for every t in (0, T), that
which implies (thanks to the arbitrariness of \({\overline{s}}<s\) and \({\overline{r}} < r\)) that (3.15) holds true and hence estimate (3.16) is true also in this more general case.
Case 2: t large. To study the solution for t large we need to assume further structure assumptions to guarantee the existence of a solution defined for every value of t. Hence, let us assume that (1.2) and (1.5) hold true in \(\Omega \times (0,+\infty )\). Thus, by Theorem 2.2u can be extended to a global solution (that we denote again u) defined in all \(\Omega \times (0,+\infty )\). Moreover, assuming also that \(|E| \in L^s_{loc}(0,+\infty ;L^r(\Omega ))\) we obtain that estimate (3.8) holds true for every \(Q(3\rho ) \subset \Omega \times (0,+\infty )\).
Hence, if (2.37) holds true (for example if (2.35) or (2.36) holds true) then by (3.8) we deduce that also
and hence, thanks to the arbitrariness of \(K(\rho )\), we conclude that
\(\square \)
4 E does not belong to \(L^{\infty }(0,T;L^{N}(\Omega ))\)
We consider here a particular case when \(E \not \in L^{\infty }(0,T;L^N(\Omega ))\). We have the following result.
Theorem 4.1
Assume that 0 belongs to \(\Omega \), that (1.2), (1.3) hold, and that
Then there exists a solution u of (1.1) satisfying the following estimate for every \(1 \le q <(N-2)\frac{\alpha }{\gamma }\) and for a.e. \(t \in (0,T)\)
where \(c = c(N,q,\alpha ,\gamma )\).
Proof
Proceeding as in the proof of Theorem 2.1 we deduce that (2.8) holds true with \(B(x)=\frac{\gamma }{|x|}\), i.e.
Recalling Hardy inequality, we thus have that
that is
We observe that, since \({\mathcal {H}} = \frac{N-2}{2}\), it results
Since we need to have \(q \ge 1\), by (4.4) we deduce that under the following assumption
estimate (4.3) holds true for every \(1 \le q <(N-2)\frac{\alpha }{\gamma }\). By Sobolev inequality and (4.3) it follows for every \(0<t_1 < t_2 \le T\)
where \(c_1 = \frac{1}{2}\alpha (q-1)\bigg (1 -\frac{1}{\alpha ^2}\frac{\gamma ^2\,q^2}{4{\mathcal {H}}^2}\bigg )\frac{4}{q^2S}\). We recall that since \(E\in L^2(Q_T)\) it results
(see the proof of Lemma 3.1 in [4], or (2.1)). Hence, we can apply Theorem 2.1 in [13] (since the exponents in the integral inequality (4.6) satisfy all the needed requirements) and we can conclude that if (4.5) is satisfied then for every \(1 \le q <(N-2)\frac{\alpha }{\gamma }\) it results
where \(c_3 = \left( \frac{N(q-1)}{2c_1}\right) ^{\frac{N}{2}\frac{1}{q'}}\), from which the result follows (since the solution u of (1.1) is the limit of \(u_{n}\)). \(\square \)
References
D. G. Aronson, J. Serrin. Local behavior of solutions of quasilinear parabolic equations. Arch. Rational Mech. Anal., 25 (1967), 81–122.
L. Boccardo. Some developments on Dirichlet problems with discontinuous coefficients. Boll. Unione Mat. Ital. (9), 2 (2009), 285–297.
L. Boccardo, S. Buccheri, G. R. Cirmi. Two linear noncoercive Dirichlet problems in duality. Milan J. Math., 86 (2018), 97–104.
L. Boccardo, L. Orsina, A. Porretta. Some noncoercive parabolic equations with lower order terms in divergence form. J. Evol. Equ., 3 (2003), 407–418.
F. Cipriani, G. Grillo. Uniform bounds for solutions to quasilinear parabolic equations. J. Differential Equations, 177 (2001), 209–234.
R.G. Cirmi, S. D’Asero, S. Leonardi, M.M. Porzio. Local regularity results for solutions of linear elliptic equations with drift term. Adv. Calc. Var. https://doi.org/10.1515/acv-2019-0048(to appear).
T. Coulhon, D. Hauer. Functional inequalities and regularizing effect of nonlinear semigroups—theory and application. In: SMAI—Mathématiques et Applications (2020), pp. 1–195.
T. Del Vecchio, M. R. Posteraro. Existence and regularity results for nonlinear elliptic equations with measure data. Adv. Differential Equations, 1 (1996), 899–917.
F. Farroni, G. Moscariello. A nonlinear parabolic equation with drift term. Nonlinear Anal., 177 (2018), 397–412.
O. A. Ladyženskaja, V. A. Solonnikov, N. N. Ural’ceva. Linear and quasilinear equations of parabolic type. Translations of the American Mathematical Society, American Mathematical Society, Providence (1968), xi+648.
G. Moscariello, M. M. Porzio. Quantitative asymptotic estimates for evolution problems. Nonlinear Anal., 154 (2017), 225–240.
M. M. Porzio. On decay estimates. J. Evol. Equ., 9 (2009), 561–591.
M. M. Porzio. On uniform and decay estimates for unbounded solutions of partial differential equations. J. Differential Equations, 259 (2015), 6960–7011.
G. Stampacchia. Le problème de Dirichlet pour les équations elliptiques du second ordre à coefficients discontinus. Ann. Inst. Fourier (Grenoble), 15 (1965), 189–258.
G. Stampacchia. Équations elliptiques du second ordre à coefficients discontinus. Les Presses de l’Université de Montréal, Montreal (1966), 326.
L. Véron. Effets régularisants de semi-groupes non linéaires dans des espaces de Banach. Ann. Fac. Sci. Toulouse Math. (5), 1 (1979), 171–200.
Acknowledgements
The authors would like to thank Google and its Google Meet software for supporting all the discussions (both preliminary and final) which generated the present paper during the COVID-19 lockdown in Italy.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Boccardo, L., Orsina, L. & Porzio, M.M. Regularity results and asymptotic behavior for a noncoercive parabolic problem. J. Evol. Equ. 21, 2195–2211 (2021). https://doi.org/10.1007/s00028-021-00678-2
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00028-021-00678-2