Existence of a global weak solution for a reaction–diffusion problem with membrane conditions


Several problems, issued from physics, biology or the medical science, lead to parabolic equations set in two sub-domains separated by a membrane with selective permeability to specific molecules. The corresponding boundary conditions, describing the flow through the membrane, are compatible with mass conservation and energy dissipation and are called the Kedem–Katchalsky conditions. Additionally, in these models, written as reaction–diffusion systems, the reaction terms have a quadratic behaviour. M. Pierre and his collaborators have developed a complete \(L^1\) theory for reaction–diffusion systems with different diffusions. Here, we adapt this theory to the membrane boundary conditions and prove the existence of weak solutions when the initial data have only \(L^1\) regularity using the truncation method for the nonlinearities. In particular, we establish several estimates as the \(W^{1,1}\) regularity of the solutions. Also, a crucial step is to adapt the fundamental \(L^2\) (space, time) integrability lemma to our situation.

This is a preview of subscription content, access via your institution.

Fig. 1


  1. 1.

    Adams, R. A., and Fournier, J. J. Sobolev spaces. Elsevier, 2003.

  2. 2.

    Baras, P., and Pierre, M. Problemes paraboliques semi-lineaires avec donnees mesures. Applicable Analysis 18, 1-2 (1984), 111–149.

    MathSciNet  Article  Google Scholar 

  3. 3.

    Bathory, M., Bulíček, M., and Souček, O. Existence and qualitative theory for nonlinear elliptic systems with a nonlinear interface condition used in electrochemistry. Z. Angew. Math. Phys. 71, 3 (2020), Paper No. 74, 24.

    MathSciNet  Article  Google Scholar 

  4. 4.

    Bothe, D., and Pierre, M. Quasi-steady-state approximation for a reaction–diffusion system with fast intermediate. Journal of Mathematical Analysis and Applications 368, 1 (2010), 120–132.

    MathSciNet  Article  Google Scholar 

  5. 5.

    Brezis, H. Functional analysis, Sobolev spaces and partial differential equations. Springer Science & Business Media, 2010.

  6. 6.

    Brezis, H., Caffarelli, L. A., and Friedman, A. Reinforcement problems for elliptic equations and variational inequalities. Annali di Matematica Pura ed Applicata 123, 1 (1980), 219–246.

    MathSciNet  Article  Google Scholar 

  7. 7.

    Cañizo, J. A., Desvillettes, L., and Fellner, K. Improved duality estimates and applications to reaction-diffusion equations. Comm. Partial Differential Equations 39, 6 (2014), 1185–1204.

    MathSciNet  Article  Google Scholar 

  8. 8.

    Calabrò, F., and Zunino, P. Analysis of parabolic problems on partitioned domains with nonlinear conditions at the interface: application to mass transfer through semi-permeable membranes. Mathematical models and methods in applied sciences 16, 4 (2006), 479–501.

    MathSciNet  Article  Google Scholar 

  9. 9.

    Cangiani, A., and Natalini, R. A spatial model of cellular molecular trafficking including active transport along microtubules. Journal of theoretical biology 267, 4 (2010), 614–625.

    MathSciNet  Article  Google Scholar 

  10. 10.

    Chaplain, M. A., Giverso, C., Lorenzi, T., and Preziosi, L. Derivation and application of effective interface conditions for continuum mechanical models of cell invasion through thin membranes. SIAM Journal on Applied Mathematics 79, 5 (2019), 2011–2031.

    MathSciNet  Article  Google Scholar 

  11. 11.

    Demengel, F., Demengel, G., and Erné, R. Functional spaces for the theory of elliptic partial differential equations. Springer, 2012.

  12. 12.

    Dimitrio, L. Modelling nucleocytoplasmic transport with application to the intracellular dynamics of the tumor suppressor protein p53. PhD thesis, Université Pierre et Marie Curie-Paris VI and Università degli Studi di Roma La Sapienza, 2012.

  13. 13.

    Evans, L. C. Partial differential equations. American Mathematical Soc., 2010.

  14. 14.

    Gallinato, O., Colin, T., Saut, O., and Poignard, C. Tumor growth model of ductal carcinoma: from in situ phase to stroma invasion. Journal of Theoretical Biology 429 (2017), 253–266.

    Article  Google Scholar 

  15. 15.

    Kedem, O., and Katchalsky, A. A physical interpretation of the phenomenological coefficients of membrane permeability. The Journal of General Physiology 45, 1 (1961), 143–179.

    Article  Google Scholar 

  16. 16.

    Laamri, E.-H., and Perthame, B. Reaction-diffusion systems with initial data of low regularity. Journal of Differential Equations (in press).

  17. 17.

    Laamri, E. H., and Pierre, M. Global existence for reaction–diffusion systems with nonlinear diffusion and control of mass. Annales de l’Institut Henri Poincaré C, Analyse non linéaire 34, 3 (2017), 571–591.

    MathSciNet  Article  Google Scholar 

  18. 18.

    Li, H., and Wang, X. Effective boundary conditions for the heat equation with interior inclusion. Communications in Mathematical Research 36, 3 (2020), 272–295.

    MathSciNet  Article  Google Scholar 

  19. 19.

    Li, J., Su, L., Wang, X., and Wang, Y. Bulk-Surface Coupling: Derivation of Two Models. Preprint (2019).

  20. 20.

    Morrey Jr, C. B. Multiple integrals in the calculus of variations. Springer Science & Business Media, 2009.

  21. 21.

    Pierre, M. Global existence in reaction-diffusion systems with control of mass: a survey. Milan Journal of Mathematics 78, 2 (2010), 417–455.

    MathSciNet  Article  Google Scholar 

  22. 22.

    Pierre, M., and Rolland, G. Global existence for a class of quadratic reaction-diffusion system with nonlinear diffusion and \(l^1\) initial data. Nonlinear Analysis 138 (2016), 369–387.

    MathSciNet  Article  Google Scholar 

  23. 23.

    Quarteroni, A., Veneziani, A., and Zunino, P. Mathematical and numerical modeling of solute dynamics in blood flow and arterial walls. SIAM Journal on Numerical Analysis 39, 5 (2002), 1488–1511.

    MathSciNet  Article  Google Scholar 

  24. 24.

    Quittner, P., and Souplet, P. Superlinear parabolic problems. Birkhäuser Advanced Texts: Basler Lehrbücher. [Birkhäuser Advanced Texts: Basel Textbooks]. Birkhäuser/Springer, Cham, 2019. Blow-up, global existence and steady states, Second edition of [MR2346798].

  25. 25.

    Serafini, A. Mathematical models for intracellular transport phenomena. PhD thesis, Université degli Studi di Roma La Sapienza, 2007.

Download references


The authors have received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 740623). The work of G.C. was also partially supported by GNAMPA-INdAM.

Author information



Corresponding author

Correspondence to Giorgia Ciavolella.

Ethics declarations

Conflict of interest

The authors declare that they have no competing interests.

Data availability statement

Not applicable.

Code availability

Not applicable

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.



We now analyse in detail regularity in our problem referring to Lemma 1.2 that we have rewritten here below, whereas in the next Appendix, we discuss about compactness. We extend previous results for reaction–diffusion systems without membrane [2, 4, 16, 17, 21] and we refer to [24] for the general theory of parabolic equations. We also refer to [17] for a regularity lemma.

Lemma A.1

(A priori bounds). We consider w solution of the following problem in dimension \(d\ge 2\)

$$\begin{aligned} \left\{ \begin{array}{ll} \partial _t w-D\Delta w=f,&{}\quad \text{ in }\; Q_T,\\ w=0,&{}\quad \text{ in }\; \Sigma _T,\\ \partial _{{\varvec{n}}^{1}} w^{1}=\partial _{{\varvec{n}}^{1}} w^2=k(w^2-w^{1}), &{} \quad \text{ in }\; \Sigma _{T,\Gamma },\\ w(0,x)=w_0(x)\ge 0, &{} \quad \text{ in }\; \Omega , \end{array} \right. \end{aligned}$$

with \(f\in L^1(Q_T)\) and \(w_0\in L^1(\Omega )\). Then,

  • \(w \in L^{\beta }\big (0,T;W^{1,\beta }(\Omega )\big ), \; \forall \beta \in \left[ 1,\frac{d}{d-1}\right) \) and \((1+|w|)^\alpha \in L^2\big (0,T;H^1(\Omega )\big )\; \text{ for } \alpha \in \left[ 0,\frac{1}{2}\right) \).

  • The mapping \((w_0,f)\longmapsto w\) is compact from \(L^1(\Omega )\times L^1(Q_T)\) into \(L^1\big (0,T;L^{\gamma _1} (\Omega )\big )\), for all \(\gamma _1<\frac{d}{d-2}\) and \(L^{\gamma _2}(Q_T)\) for all \({\gamma _2} < \frac{2+d}{d}\).

  • The trace mapping \((w_0,f)\longmapsto Tr_\Gamma (w)\in L^\beta \big (0,T;L^\beta (\Gamma )\big ),\; \beta \in \left[ 1,\frac{d}{d-1}\right) \) is also compact.

Notice that we do not use the information \(w\in L^2(Q_T)\) here but \(w\in L^\infty (0,T; L^1(\Omega ))\). That is used in [22] and leads to the exponent \(\beta < \frac{4}{3}\).


The proof is based on manipulating nonlinear quantities and Sobolev embeddings. We divide it in several steps.

Some \(L^2\) regularity of \(\nabla w\). Multiplying the equation of w in (30) by \(\frac{w}{(1+|w|^\frac{1}{\mu })^\mu }\) and integrating on \(\Omega \), we obtain three terms which we estimate separately.

We begin with the Laplacian term. Recalling the membrane conditions and applying the Leibniz rule and the divergence theorem, arguing by a regularization and a limit technique, we gain, since \(\frac{w}{(1+|w|^\frac{1}{\mu })^\mu }\) is an increasing function,

$$\begin{aligned} \int _{\Omega } \frac{w}{(1+|w|^\frac{1}{\mu })^{\mu }}\; \Delta w&=\int _{\Gamma }\frac{w^1}{(1+|w^1|^{\frac{1}{\mu }})^{\mu }} \partial _{n_1} w^1\;+\;\int _{\Gamma }\frac{w^2}{(1+|w^2|^{\frac{1}{\mu }})^{\mu }} \partial _{n_2} w^2\nonumber \\&\quad -\int _{\Omega } \frac{|\nabla w|^2}{(1+|w|^{\frac{1}{\mu }})^{\mu +1}}\nonumber \\&=\int _{\Gamma }\left( \frac{w^1}{(1+|w^1|{^\frac{1}{\mu })^{\mu }}} - \frac{w^2}{(1+|w^2|{^\frac{1}{\mu }})^{\mu }}\right) k(w^2-w^1)\nonumber \\&\quad -\int _{\Omega } \frac{|\nabla w|^2}{(1+|w|^{\frac{1}{\mu }})^{\mu +1}}\\&\le - \int _{\Omega } \frac{|\nabla w|^2}{(1+|w|^{\frac{1}{\mu }})^{\mu +1}}. \end{aligned}$$

We analyse now the reaction term. We remark that \(0\le \frac{w}{(1+|w|^\frac{1}{\mu })^\mu }\le 1\) and, using that \(f\in L^1(Q_T)\), we conclude

$$\begin{aligned} \int _{\Omega } \left| \frac{w}{(1+|w|^\frac{1}{\mu })^{\mu }}\;f\right| \le \int _{\Omega } |f| = \Vert f\Vert _{L^1(\Omega )}. \end{aligned}$$

Next, for the time derivative, we define the anti-derivative \(0 \le \psi _\mu (w)= \int _0^w \frac{v\, dv}{(1+|v| ^\frac{1}{\mu })^{\mu }} \le w\), then

$$\begin{aligned} \frac{w}{(1+|w|^\frac{1}{\mu })^{\mu }}\;\partial _t w\;=:\; \partial _t \psi _{\mu }(w). \end{aligned}$$

Therefore, combining the previous equality and inequalities, we find

$$\begin{aligned} \int _{\Omega } \partial _t \psi _{\mu }(w)\;+\;D\int _{\Omega } \frac{|\nabla w|^2}{(1+|w|^{\frac{1}{\mu }})^{\mu +1}}\le \Vert f\Vert _{L^1(\Omega )}. \end{aligned}$$

At this point, we can integrate in time and obtain

$$\begin{aligned} D\int _{Q_T} \frac{|\nabla w|^2}{(1+|w|^{\frac{1}{\mu }})^{\mu +1}} \le \int _{\Omega } \psi _{\mu }\big (w_0(x)\big ) + \Vert f\Vert _{L^1(Q_T)} \le \Vert w_0\Vert _{L^1(\Omega )} + \Vert f\Vert _{L^1(Q_T)}. \end{aligned}$$

Since, for all \(\mu > 1\) there is a \(C_\mu \) such that

$$\begin{aligned} (1+|w|^{\frac{1}{\mu }})^{\mu +1} \le C_\mu (1+|w|)^{2(1-\alpha )}, \qquad \alpha = \frac{1}{2}\left( 1-\frac{1}{\mu }\right) , \end{aligned}$$

we conclude that

$$\begin{aligned} \int _{Q_T} (1+|w|)^{2(\alpha -1)} |\nabla w|^2 \le \frac{C_\mu }{D}\left[ \Vert w_0\Vert _{L^1(\Omega )} + \Vert f\Vert _{L^1(Q_T)}\right] , \qquad 0< \alpha < \frac{1}{2}. \end{aligned}$$

And thus, there is a constant \(C_\alpha \) which also depends on \(\Vert w_0\Vert _{L^1(\Omega )} + \Vert f\Vert _{L^1(Q_T)}\) such that

$$\begin{aligned} \int _{Q_T} |\nabla (1+|w|)^\alpha |^2 \le C_\alpha , \qquad 0< \alpha < \frac{1}{2}. \end{aligned}$$

Integrability of w. The Sobolev embedding (see Appendix C) gives

$$\begin{aligned} \left( \int _\Omega (1+|w|)^{\alpha 2^*} \right) ^{\frac{2}{2^*}} \le C\int _{\Omega } |\nabla (1+|w|)^\alpha |^2, \quad \qquad {2^*}= \frac{2d}{d-2}. \end{aligned}$$

which is only useful when \(\alpha 2^* >1\), i.e. \( \frac{d-2}{2d} <\alpha \). Then, we can interpolate between \(L^1\) and \(L^{\alpha 2^*}\) and find

$$\begin{aligned}&\left( \int _\Omega (1+|w|)^{\gamma } \right) ^{\frac{1}{\gamma }} \le C \left( \int _\Omega (1+|w|)\right) ^{\theta } \left( \int _{\Omega } |\nabla (1+|w|)^\alpha |^2\right) ^{\frac{1-\theta }{2\alpha }}, \\&\quad \frac{1}{\gamma }= \theta +\frac{1-\theta }{\alpha 2^*} . \end{aligned}$$

We may choose \(\frac{1-\theta }{2\alpha }=1\), and, recalling that \(\alpha < \frac{1}{2}\), we find the integrability

$$\begin{aligned} w\in L^1\big (0,T; L^{\gamma _1}(\Omega )\big ) \qquad \text {with} \qquad {\gamma _1} = \frac{d}{2\big ( d(1-\alpha ) -1\big ) } < \frac{d}{d-2} . \end{aligned}$$

We may also choose \(\frac{\gamma (1-\theta )}{2\alpha }=1\), \(\alpha < \frac{1}{2}\) and find the integrability

$$\begin{aligned} w\in L^{\gamma _2}(Q_T) \qquad \text {with} \qquad {\gamma _2} = \frac{2\big (1+\alpha d\big )}{d} < \frac{2+d}{d} . \end{aligned}$$

Regularity of \(\nabla w\). On the other hand, Hölder inequality gives

$$\begin{aligned} \begin{aligned} \displaystyle \int _{\Omega } |\nabla w|^\beta&=\int _{\Omega } \frac{|\nabla w|^\beta }{(1+|w|)^\eta }(1+|w|)^\eta \displaystyle \le \left( \int _{\Omega } \frac{|\nabla w|^{\beta r}}{(1+|w|)^{\eta r}}\right) ^\frac{1}{r}\left( \int _{\Omega }(1+|w|)^{\eta p}\right) ^\frac{1}{p}\\&\displaystyle \le C\left( \int _{\Omega } |\nabla (1+|w|)^\alpha |^2 \right) ^{\frac{1}{r}} \left( \int _{\Omega }(1+|w|)^{\eta p}\right) ^\frac{1}{p} \end{aligned} \end{aligned}$$


$$\begin{aligned} \frac{1}{r}+\frac{1}{p}=1, \qquad \beta =\frac{2}{r} \le 2, \quad \eta r=2 (1-\alpha ). \end{aligned}$$

We can choose \(\eta p = \gamma _1\) from above, which requires \(\eta \left( \frac{1}{2(1-\alpha )} + \frac{1}{\gamma _1} \right) = 1\), \(\beta = \frac{\eta }{1-\alpha }= \frac{2 \gamma _1}{\gamma _1+2(1-\alpha )}\) and we find, thanks to the estimate (31),

$$\begin{aligned} \int _{\Omega } |\nabla w|^\beta \in L^1(0,T) \qquad \text {with} \qquad \beta < \frac{d}{d-1}. \end{aligned}$$

This concludes the proof of the gradient estimate. Moreover, considering that \(\beta <\gamma _2\), thanks to Sobolev embeddings, we can infer that \(w \in L^\beta (0,T;L^\beta (\Omega ))\).

The trace. The regularity of the trace derives from its continuity property [5] (p. 315), i.e.

$$\begin{aligned} \int _0^T\Vert \text{ Tr }(w)\Vert ^\beta _{W^{1-\frac{1}{\beta },\beta }(\Gamma )}\le \int _0^T\Vert w\Vert ^\beta _{W^{1,\beta }(\Omega )}, \quad 1\le \beta <\frac{d}{d-1}. \end{aligned}$$


In order to conclude the proof of Lemma A.1, it remains to adapt compactness arguments to the case of the membrane problem. A proof based on a dual approach, see [2, 4], could be used. We rather go to a direct proof.

Compactness in space. It can be obtained using the Rellich–Kondrachov theorem [1], since we know the approximate family is bounded in the spaces \(W^{1,\beta }(\Omega ^\lambda )\), \(\lambda =1, \; 2\) which are compactly embedded in \(L^{\gamma _1}(\Omega ^\lambda )\), with \(\gamma _1<\frac{d}{d-2}\).

Compactness in time. We use the Fréchet–Kolmogorov criteria, see [5] for instance. Let \(\varphi (x)\) be a non-negative, radially symmetric, \(C^\infty _c(\mathbb {R}^d)\) standard mollifier with mass 1. We define the family \((\varphi _\delta )_{\delta >0}\) by

$$\begin{aligned} \varphi _\delta (x)=\frac{1}{\delta ^d}\,\varphi \left( \frac{x}{\delta }\right) , \qquad \Vert \varphi _\delta \Vert _{L^1(\Omega )}=1. \end{aligned}$$

Moreover, we have

$$\begin{aligned} \Vert g*\varphi _\delta \Vert _{L^p(\Omega )}\le \Vert \varphi _\delta \Vert _{L^1(\Omega )}\Vert g\Vert _{L^p(\Omega )} , \end{aligned}$$

and it holds ([13], p. 273) that for any function \(g\in W^{1,p}(\Omega )\),

$$\begin{aligned} \Vert g*\varphi _\delta -g\Vert _{L^p(\Omega )}\le \delta \Vert \nabla g\Vert _{L^p(\Omega )}. \end{aligned}$$

About the derivative of order k of \(\varphi _\delta \), we know that

$$\begin{aligned} \nabla ^k \varphi _\delta (x)=\frac{1}{\delta ^{d+k}}\nabla ^k \varphi \left( \frac{x}{\delta }\right) , \qquad \Vert \nabla ^k \varphi _\delta \Vert _{L^1(\Omega )}\le \frac{C}{\delta ^{k}}. \end{aligned}$$


To complete the proof of time compactness, we shall prove that, as \(h \rightarrow 0\),

$$\begin{aligned} \int _{0}^{T-h}\int _{\Omega }|w(t+h,x)-w(t,x)|\hbox {d}x \hbox {d}t \rightarrow 0. \end{aligned}$$

By comparison with the mollified versions, the triangular equality yields

$$\begin{aligned}&\int _{0}^{T-h}\int _{\Omega }|w(t+h,x)-w(t,x)|\hbox {d}x\hbox {d}t \nonumber \\&\quad \le \int _{0}^{T-h}\int _{\Omega }|w(t,x)- w(t,\cdot )*\varphi _\delta (x)|\hbox {d}x\hbox {d}t\nonumber \\&\qquad + \int _{0}^{T-h}\int _{\Omega }|w(t+h,x)- w(t+h,\cdot )*\varphi _\delta (x)|\hbox {d}x\hbox {d}t \nonumber \\&\qquad +\int _{0}^{T-h}\int _{\Omega }|w(t+h,\cdot )*\varphi _\delta (x)-w(t,\cdot )*\varphi _\delta (x)|\hbox {d}x \hbox {d}t \end{aligned}$$

Here, \(\delta \) depends on h (to be specified later on) and converges to zero. It suffices to prove that each integral converges to zero as \(h\rightarrow 0\).

First term. We analyse the first term in the right-hand side. It holds that

$$\begin{aligned} \int _{0}^{T-h}\int _{\Omega }|w(t,x)- w(t,\cdot )*\varphi _\delta (x)|\hbox {d}x \hbox {d}t\le \delta \int _0^{T-h} \Vert \nabla w(t,x)\Vert _{L^1(\Omega )} \hbox {d}t \le C\delta (h),\nonumber \\ \end{aligned}$$

thanks to w regularity and to (36), which proves that it converges to zero as \(h\rightarrow 0\).

Second term. For the second integral, we can proceed as for the first one obtaining

$$\begin{aligned} \int _{0}^{T-h}\int _{\Omega }|w(t+h,x)- w(t+h,\cdot )*\varphi _\delta (x)|\hbox {d}x \hbox {d}t \le C \delta (h). \end{aligned}$$

Third term. Remembering (30), the last term can be written as

$$\begin{aligned}&\int _{0}^{T-h}\int _{\Omega }|w(t+h,\cdot )*\varphi _\delta (x)-w(t,\cdot )*\varphi _\delta (x)|\hbox {d}x\, \hbox {d}t\\&\quad =\int _{0}^{T-h}\int _{\Omega }\left| \int _{t}^{t+h} \frac{\partial w}{\partial s}(s,x)*\varphi _\delta (x)\hbox { d}s\right| \hbox {d}x\,\hbox {d}t\\&\quad =\int _{0}^{T-h}\int _{\Omega }\left| \int _{t}^{t+h} \left[ D\Delta w + f\right] *\varphi _\delta \hbox {d}s\right| \hbox {d}x\,\hbox {d}t \\&\quad = \int _{0}^{T-h}\int _{\Omega }\left| \int _{t}^{t+h} D w *\Delta \varphi _\delta + f*\varphi _\delta \, \hbox {d}s \right| \hbox {d}x\,\hbox {d}t \end{aligned}$$

after exchanging derivatives in the convolution. From (35) we deduce

$$\begin{aligned}&\int _{0}^{T-h}\int _{\Omega }|w(t+h,\cdot )*\varphi _\delta (x)-w(t,\cdot )*\varphi _\delta (x)|\hbox {d}x\, \hbox {d}t \\&\quad \le \int _{0}^{T-h}\int _t^{t+h} D\Vert w\Vert _{L^1(\Omega )} \Vert \Delta \varphi _\delta \Vert _{L^1(\Omega )}\nonumber \\&\qquad + \int _{0}^{T-h}\int _t^{t+h}\Vert f\Vert _{L^1(\Omega )} \Vert \varphi _\delta \Vert _{L^1(\Omega )}. \end{aligned}$$

Finally, thanks to (34) and (37), we obtain choosing \(\delta =h^{1/4}\)

$$\begin{aligned} \int _{0}^{T-h}\int _{\Omega }|w(t+h,\cdot )*\varphi _\delta (x)-w(t,\cdot )*\varphi _\delta (x)|\hbox {d}x\, \hbox {d}t \le C\left[ \frac{h}{\delta ^2} +h\right] \le C \sqrt{h} \end{aligned}$$

and (38) follows combining this estimate with (39) and (40). \(\square \)

Applying the Fréchet–Kolmogorov theorem [5], we conclude that the set of functions \(w \in L^1(Q_T)\) under consideration is compact in \(L^1(Q_T)\). Consequently, we claim compactness in \(L^1\big (0,T; L^{\gamma _1}(\Omega )\big )\) with \({\gamma _1}< \frac{d}{d-2}\) and in \(L^{\gamma _2}(Q_T)\) with \({\gamma _2}< \frac{2+d}{d} \). In fact, since we have \(L^1\)-convergence of \(L^p\)-functions, we deduce convergence in the space \(L^q\), for \(q<p\).

Compactness of traces in \(L^\beta \big (0,T;L^\beta (\Gamma )\big )\). Space compactness can be deduced, in each \(\Omega ^\lambda \), from trace continuity and a compactness result for the boundary ([11], p. 167) such that \(W^{1-\frac{1}{\beta },\beta }(\Gamma ) \subset \subset L^{\beta }(\Gamma )\). Time compactness is again achieved through the Fréchet–Kolmogorov theorem. Following the same proof as before and W changing the order of the time integrals, we need to recall Kedem–Katchalsky membrane conditions from which we can infer that \(\partial _tTr_\Gamma (w) \in L^1(0,T;L^1(\Gamma ))\) and so we can conclude the proof. \(\square \)

Sobolev and Poincaré inequalities with membrane

For completeness, we explain why the Sobolev embeddings can be extended to the membrane problem, leading to (31) and (32). More precisely, we explain how to arrive to

$$\begin{aligned} \Vert \phi _{\alpha }(w^1)\Vert ^2_{L^{2^*}(\Omega ^1)} + \Vert \phi _{\alpha }(w^2)\Vert ^2_{L^{2^*}(\Omega ^2)}\; \le C \left( \Vert \nabla \phi _{\alpha }(w^1)\Vert ^2_{L^2(\Omega ^1)} + \Vert \nabla \phi _{\alpha }(w^2)\Vert ^2_{L^2(\Omega ^2)} \right) . \end{aligned}$$

There are two difficulties. First, the boundary condition is not Dirichlet everywhere. Second, we are dealing with a singular domain \(\Omega \) and so we cannot use directly the Sobolev or Poincaré inequalities in \(\Omega \), but only some easy generalizations that we explain now.

We are going to prove the

Theorem C.1

(Gagliardo–Nirenberg–Sobolev inequality with membrane). We consider the bounded domain \(\Omega = \Omega ^1\, \cup \,\Omega ^2 \subset \mathbb {R}^d, \; d\ge 2\), with piecewise \(C^1\) sub-domains \(\Omega ^1\) and \(\Omega ^2\) and a \(C^1\) membrane \(\Gamma =\partial \Omega ^1\, \cap \,\partial \Omega ^2\) which decomposes \(\Omega \) in the two parts. We take the function \(v=(v^1,v^2) \in {\mathbf{H}^\mathbf{1}}\) (see Definition 1.3), then, for \(\lambda =1,2\),

$$\begin{aligned} \Vert v^\lambda \Vert _{L^{2^*}(\Omega ^\lambda )}\;\le \;C(\Omega ^\lambda )\;\;\Vert \nabla v^\lambda \Vert _{{L^{2}(\Omega ^\lambda )}^d}, \end{aligned}$$

and consequently

$$\begin{aligned} {[}\;\Vert v^1\Vert _{L^{2^*}(\Omega ^1)}\;+\;\Vert v^2\Vert _{L^{2^*}(\Omega ^2)} \;]\;\le \;C(\Omega ^1,\Omega ^2)\;[\;\Vert \nabla v^1\Vert _{{L^{2}(\Omega ^1)}^d}\;+\;\Vert \nabla v^2\Vert _{{L^{2}(\Omega ^2)}^d}\;].\nonumber \\ \end{aligned}$$

The reason why we want to prove this theorem is that the domain \(\Omega \) described above is not enough regular to use the usual Gagliardo–Nirenberg–Sobolev inequality ([5], p. 284). Consequently, we need to build smoother domains containing each \(\Omega ^i\) in which we can apply known results and then, with a restriction to \(\Omega \), we can find (41) and (42). The construction is made considering an extension of \(\Gamma \) and a domain with the same internal structure as \(\Omega \) such that it contains \(\Omega \) and each extension of the \(\Omega ^i\) is of class \(C^1\).

We first recall the standard Sobolev inequality ([5], p. 284) in a bounded open set.

Theorem C.2

(Sobolev embedding). Let Q be a bounded open subset of class \(C^1\) in \(\mathbb {R}^d\). There is a constant \( C_Q\) such that for all \(v\in H^1(Q),\) we have

$$\begin{aligned} v\in L^{2^*}(Q) \quad \text{ and } \quad \Vert v\Vert _{L^{2^*}(Q)}\;\le \;C_Q\;\left[ \;\Vert v\Vert _{L^{2}(Q)}\;+\;\Vert \nabla v\Vert _{{L^{2}(Q)}^d}\;\right] . \end{aligned}$$


We recall how to prove Theorem C.2 departing from the case of the full space. We use the regularity of the domain which assures us the existence of a linear and continuous extension operator \(T: H^1(Q) \rightarrow H^1(R^d)\), which is also the extension from \(L^2(Q)\) into \(L^2(\mathbb {R}^d)\) ([5], p. 272). So, we obtain that:

$$\begin{aligned}&\bullet \; \text{ taken } v\in H^1(Q), \quad T(v)\in H^1(\mathbb {R}^d) \text{ and } T(v)=v \text{ on } Q; \end{aligned}$$
$$\begin{aligned}&\bullet \; \Vert T(v)\Vert ^2_{L^2(\mathbb {R}^d)} \le C^2_{\small {\text{ exten }}L^2}(Q)\;\Vert v\Vert ^2_{L^2(Q)};\end{aligned}$$
$$\begin{aligned}&\bullet \; \Vert \nabla T(v)\Vert ^2_{{L^2(\mathbb {R}^d)}^d} \le C^2_{\small {\text{ exten }}H^1}(Q)\;\Vert v\Vert ^2_{H^1(Q)}. \end{aligned}$$

Moreover, for construction (see the proof of the extension theorem [5], p. 272), this operator is in \(H^1_0(R^d)\). Consequently, using a corollary of the Sobolev inequality ([13], p. 265), we get that

$$\begin{aligned} T(v)\in L^{2^*}(\mathbb {R}^d) \text{ and } \Vert T(v)\Vert _{L^{2^*}(\mathbb {R}^d)}\le C_{\small {\text{ sob }}}(d,2)\; \Vert \nabla T(v)\Vert _{{L^2(\mathbb {R}^d)}^d}. \end{aligned}$$

We proceed with some estimates due to the application of (43), (44), (45). First of all, we deduce

$$\begin{aligned} \Vert \nabla v\Vert ^2_{{L^2(Q)}^d}= & {} \Vert \nabla T(v)\Vert _{{L^2(Q)}^d}\le \Vert \nabla T(v)\Vert _{{L^2(\mathbb {R}^d)}^d}\le C_{\small {\text{ exten }}H^1}(Q)\;\Vert v\Vert _{H^1(Q)}^2\\= & {} C_{\small {\text{ exten }}H^1}(Q)\;\left[ \Vert v\Vert _{L^2(Q)}+\Vert \nabla v\Vert _{{L^2(Q)}^d}\right] . \end{aligned}$$

Since \(T(v)\in L^{2^*}(\mathbb {R}^d)\) and \(T(v)=v\) on Q, we get \(v\in L^{2^*}(Q)\) and

$$\begin{aligned} \Vert v\Vert _{L^{2^{*}}(Q)}^2= & {} \Vert T(v)\Vert _{L^{2^{*}}(Q)}^2\le \Vert T(v)\Vert _{L^{2^{*}}(\mathbb {R}^d)}^2\le (C_{\small {\text{ sob }}}(d,2))^2 \;\Vert \nabla T(v)\Vert _{{L^2(\mathbb {R}^d)}^d}^2\\\le & {} (C_{\small {\text{ sob }}}(d,2))^2\; C^2_{\small {\text{ exten }}H^1}(Q)\;\left[ \Vert v\Vert _{L^2(Q)}^2+\Vert \nabla v\Vert _{{L^2(U)}^d}^2\right] . \end{aligned}$$

The proof of Theorem C.2 is complete. \(\square \)

Since we do not impose Dirichlet conditions on the full boundary, we need the following generalized Poincaré inequality ([20] p. 82).

Theorem C.3

(Poincaré inequality). Suppose Q a bounded and connected open subset of \(\mathbb {R}^d\) of class \(C^1\) and consider a portion of its boundary \(\Sigma _0 \subset \partial Q \) such that \(|\Sigma _0|>0\). Then, there exists a constant \(C(Q,\Sigma _0)\) such that

$$\begin{aligned} \forall v\in H^1(Q) \ \mathrm{such \ that}\ Tr_{\Sigma _0}(v)=0, \quad \Vert v\Vert _{L^2(Q)}^2\le C(Q,\Sigma _0) \Vert \nabla v\Vert _{{L^2(Q)}^d}^2. \end{aligned}$$


If the statement is not true, we can find a sequence \({v_n}\) such that each \(v_n\in H^1(Q)\) and

$$\begin{aligned} \Vert v_n\Vert _{L^2(Q)}^2>n\;\left[ \Vert \nabla v_n\Vert _{{L^2(Q)}^d}^2\;+\;\left( \int _{\Sigma _0}|v_n|\hbox {d}S\right) ^2\right] . \end{aligned}$$

On account of the homogeneity (normalizing), we may assume that \(\Vert v_n\Vert _{L^2(Q)}=1\), for each n. So, we infer that

$$\begin{aligned} n\;\left[ \Vert \nabla v_n\Vert _{{L^2(Q)}^d}^2\;+\;\left( \int _{\Sigma _0}|v_n|\hbox {d}S\right) ^2\right] <1, \end{aligned}$$

which implies that

$$\begin{aligned} \Vert \nabla v_n\Vert _{{L^2(Q)}^d}^2<\frac{1}{n}. \end{aligned}$$

Therefore, \(\nabla v_n \rightarrow 0\) in \(L^2(Q)\). Moreover, \(v_n\) is bounded in \(H^1(Q)\), so, up to a sub-sequence, it converges weakly in \(H^1(Q)\) to some v. So, \(\nabla v_n \rightharpoonup \nabla v\), that means \(\nabla v=0\). This shows that v is a constant (since Q is connected). For the continuity of the trace operator and (47), we deduce

$$\begin{aligned} 0=\lim _{n\rightarrow +\infty }\int _{\Sigma _0} |v_n| \hbox {d}S=\int _{\Gamma _0} |v| \hbox {d}S=|c||\Gamma _0|, \end{aligned}$$

and so \(v=0\).

At the same time, thanks to the Rellich–Kondrachov compactness theorem [1, 5, 13], up to a sub-sequence, \(v_n\) converges strongly in \(L^2(Q)\) to \(v=0\). Hence, since \(\Vert v_n\Vert _{L^2(Q)}=1\), we arrive to a contradiction. \(\square \)

At this point we are able to give the proof of Theorem C.1.


We apply Theorems C.2 and C.3. First of all, we consider the extension of \(\Gamma \) into the space \(\mathbb {R}^d\) such that now \(\Gamma \) separates the space into two pieces \(P^\lambda \) with \(\lambda =1,2\). Since we have Dirichlet boundary conditions on \(\Gamma ^\lambda \), we can extend the function to zero in the whole \(P^\lambda \). So now, considering \(Q^\lambda \) a domain of class \(C^1\) such that \(\Omega ^\lambda \subset Q^\lambda \subset P^\lambda \) and for \(\lambda , \sigma =1,2\), \(\;Q^\lambda \cap P^\sigma \) is a portion of \(\Gamma \), we can apply Theorems C.2 and C.3 to

$$\begin{aligned} {\tilde{v}}^\lambda = \left\{ \begin{array}{ll} v^\lambda , &{} \quad \text{ in } \Omega ^\lambda , \\ 0, &{} \quad \text{ in } \Gamma ^\lambda \cup \{Q^\lambda \setminus \Omega ^\lambda \}. \end{array}\right. \end{aligned}$$

This proves Theorem C.1 in \(Q^\lambda \) and, so, in \(\Omega ^\lambda \). \(\square \)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Ciavolella, G., Perthame, B. Existence of a global weak solution for a reaction–diffusion problem with membrane conditions. J. Evol. Equ. (2020). https://doi.org/10.1007/s00028-020-00633-7

Download citation


  • Kedem–Katchalsky conditions
  • Membrane boundary conditions
  • Reaction–diffusion equations
  • Mathematical biology

Mathematics Subject Classification

  • 35K57
  • 35D30
  • 35Q92