Abstract
We introduce a notion of duality solution for a single or a system of transport equations in spaces of probability measures reminiscent of the viscosity solution notion for nonlinear parabolic equations. Our notion of solution by duality is, under suitable assumptions, equivalent to gradient flow solutions in case the single/system of equations has this structure. In contrast, we can deal with a quite general system of nonlinear non-local, diffusive or not, system of PDEs without any variational structure.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Many PDE modelling instances of applied analysis lead to transport equations for a density function \(\rho \) of the form \(\partial _t \rho = \nabla \cdot (\rho \varvec{v})\), where \(t> 0\), \(x \in {\mathbb {R}^d}\), and \(\varvec{v}\) is the velocity field. In many of these applications, there is a ubiquitous choice of convolutional-type velocities depending on the density well motivated by applications in mathematical biology, social sciences and neural networks, see for instance [4, 18, 21, 36, 43, 44, 48]. Moreover, in many of these applications, \(\varvec{v}\) is given by a function of \(\rho \) itself, and it may happen that the density \(\rho \) is not an integrable function, but rather a measure. Since the mathematical treatment of these problems is rather tricky, and indeed many “non-physical” solutions may appear, the natural way to approach these problems is the so-called vanishing-viscosity method. With this method linear diffusion \(D \Delta \rho \) is included to the right-hand side, leading to the problem
The “physical” or entropic solution is the one obtained in the limit \(D \searrow 0\). In this work, we deal with systems of such equations, where several species inhabit a domain. We will consider that the evolution of these species is coupled only by the velocity field. We are interested in systems of n aggregation–diffusion equations of the form
where \(D^i \ge 0\) and \(\varvec{K}^i_t\) are velocity-fields depending non-locally on the full vector of densities \(\varvec{\mu } = (\mu ^1, \cdots , \mu ^n)\) and possibly in the spatial or time variables, under some hypotheses described below. In fact, we are to work with a generalisation allowing \(K_t\) to depend on the previous time states see (P). Notice that we cover the cases with and without diffusion (\(D_i>0\) and \(D_i =0\)) in a unified framework. We will introduce a new notion of solution, which we call dual-viscosity solution, with well-posedness, and that is able to detect this vanishing-viscosity limit.
A particularly interesting case is when the problem (1.1) has a 2-Wasserstein gradient-flow structure. For example, in the scalar setting (\(n = 1\)) without diffusion (\(D=0\)) and with an interaction potential energy functional of the form
the classical gradient flow theory developed in [2, 3] applies when W is \(\lambda \)-convex and smooth, and this later generalised to different settings [6, 8, 13, 14, 26, 27, 29, 32, 33, 38] for instance. This can also be applied to (1.1) with \(D > 0\) by including the Boltzmann entropy in the free energy functional, see [7, 15, 20, 22, 23, 42, 49, 51]. We refer to [19] for a recent survey of results in aggregation–diffusion equations. We show in Sect. 6 that our new notion of dual viscosity solutions is equivalent, in some examples, to the notion of gradient flow solutions.
In many situations, a theory of entropy solutions in the sense of Kruzkov (see [16, 40]) is possible, when we have initial data and solutions in \(L^1 ({\mathbb {R}^d}) \cap L^\infty ({\mathbb {R}^d}) \cap BV ({\mathbb {R}^d})\). For example, the work [30] deals with \(d = n = 1\), \(D= 0\) and \(K =\theta (\mu ) W' * \mu \) where \(\theta (M) = 0\) for some \(M > 0\), and W very regular. Then, clearly initial data \(0 \le \rho _0 \le M \) lead to solutions \(0 \le \rho _t \le M\). The authors prove uniqueness of entropy solutions in the sense of Kruzkov in the space \(L^\infty (0,T; L^1 ({\mathbb {R}^d}) \cap L^\infty ({\mathbb {R}^d}) \cap BV({\mathbb {R}^d}))\), by using the continuous dependence results in [39] and a fixed point argument. This requires fairly strict hypothesis on the regularity of the initial data (namely bounded variation). They also introduce an interesting constructive method based on sums of characteristics of domains evolving in time. A recent extension of [30] to related problems can be seen in [35].
However, in the diffusion-less regime \(D = 0\), these models based on transport equations can describe particle systems described by Dirac deltas, which interact through an ODE. The idea of working with Dirac deltas leads to numerical methods for general integrable data by approximating the initial datum by a sum of Diracs and regularizing the entropy [17, 24, 25]. Kruzkov’s notion of entropy solutions cannot be extended to measures. Furthermore, for less regular W (or \(\varvec{K} [ \mu ] = \nabla V(x)\) known), the finite-time formation of Dirac deltas even for smooth initial data is not known. The paper [32] studies uniqueness of distributional solutions for \(n = 2\), when \(D^i = 0\),
where \(\mathcal {H}_i, \mathcal {K}_i \in \mathcal {W}^{2,\infty }\). The authors indicate that any distributional solution is a push-forward solution (pointing to [2, Chapter 8]) of the flow that solves
This push-forward structure is not available for problems with diffusion. The main goal of this work is to construct a notion of well-posed solutions that works for the case
\(d, n \ge 1\), \(D \ge 0\), and measure initial data. We will work in 1-Wasserstein space, and recover continuous dependence estimates by discussing the “dual” problem, for which we use viscosity solutions by studying the corresponding Lipschitz estimates. We show also how our results can be adapted to different spaces (e.g., \(\dot{H}^{-1}\) and \(H^1\)).
So far, the velocity field depends only on t and \(\varvec{\mu }_t\). We can consider a generalisation of K, denoted \(\mathfrak {K}\), that depends also on the past. We denote this extended notation by \(\mathfrak {K} [{\overline{\mu }}]_t\). By saying that \(\mathfrak {K}\) depends only on the past we mean that \(0 \le t \le s \le T\) and \(\varvec{\mu } \in C([0,T]; \mathcal {P}_1 ({\mathbb {R}^d})^n)\) then
To simplify the notation, when it will not lead to confusion, we simply write \(\mathfrak {K}[\mu ]_t\). For example, we cover the case
where \(W_t\) are a family of convolution kernels. Hence, we extend the (1.1) to the more general
This covers the previous setting by defining
The structure of the paper is as follows. Section 2 is devoted to the main results and methods. We begin in Sect. 2.1 by motivating and introducing a new notion of solution in 1-Wasserstein space, motivated by the duality with viscosity solution of the dual problem, which we call dual viscosity solution. In Sect. 2.2 we present well-posedness results for \(\varvec{K}\) regular enough. In Sect. 2.3 we present a result of stability under passage to the limit. In Sect. 2.4 we discuss a limit case of the regularity assumptions of the well-posedness theory, given by the Newtonian potential. We close the presentation of the main results with Sect. 2.5 on open problems our current results do not cover, mainly non-linear diffusion and non-linear mobility. The full proofs are postponed to the next sections. First, in Sect. 3 we prove well-posedness and estimates for the dual problem of (1.1), where we assume \(\varvec{K}[\mu ]\) is replaced by a known field. Then, in Sect. 4 we prove well-posedness of (1.1) when \(\varvec{K}[\mu ]\) is again replaced by a known field. In Sect. 5 we prove the results stated in Sect. 2.2 by a fixed-point argument based on the previous estimates.
We conclude the paper with two sections on comments and extensions of our main results. First, in Sect. 6, we show that, where a gradient-flow structure is available, our solutions coincide with the steepest descent solutions, under some assumptions. In Sect. 7 we extend our main results from the 1-Wasserstein framework to the negative homogeneous Sobolev space \(\dot{H}^{-1}\). We recall that \(\dot{H}^{-1}\) is related to the 2-Wasserstein space (see below).
2 Main results and methods
2.1 A new notion of solution via duality
Considering first the case \(n = 1\), let \(\varvec{E}_t = - \varvec{{\mathfrak {K}}}[\mu ]_t\), so that we are looking at the problem
Furthermore, if we naively assume the equation to be satisfied pointwise, we can multiply by a test function \(\varphi \) and assume we can integrate by parts integrate by parts, we recover
To cancel the double integral, we take \(\varphi _t = \psi _{T-t}\) and \(\psi _s\) a solution of
and \(\psi _s\) Lipschitz in x. Using \(\varphi _t = \psi _{T-t}\) we reduce (2.1) to the equivalent formulation
This is an interesting formulation because it will allow us to exploit the duality properties of the spaces for \(\mu \) and \(\psi \).
Hence, we are exploiting the well-known duality between the continuity problem (\(\hbox {P}_{\varvec{E}}\)) and its dual (\(\hbox {P}_{\varvec{E}}^*\)).
When \(D= 0\), this connects with the push-forward formulation for first order problems. Problem (\(\hbox {P}_{\varvec{E}}^*\)) can be solved by a generalised characteristics field \(X_t\) (e.g., [34, Sect. 3.2, p.97]), and (\(\hbox {D}_{\varvec{E}}\)) means precisely that \(\mu _T \) is the push-forward by the same field (see, e.g., [2, Sect. 5.2, p.118]).
Notice that in the general setting we have n equations, and \(\psi \) depends on T. Hence, we denote our test functions \(\psi ^{T,i}\). We consider the system of n dual problems
We also recover n duality conditions
Remark 2.1
Notice that we have not introduced a condition as \(|x| \rightarrow \infty \) for \(\psi \). In the formal derivation of the notion of solution, we have required that we can formally integrate by parts. This is also the standard argument to define distributional solutions. As with distributional solutions, we set (\(\hbox {D}_i\)) as our definition and prove that this is mathematically sound in terms well-posedness. We do not claim the PDE is satisfied in any stronger sense. We will make a comment on gradient flow solutions. Furthermore, we will not rigorously integrate by parts at any point in this manuscript. We will actually discuss in detail the precise admissible initial data \(\psi _0\) for (\(\hbox {P}_{\varvec{E}}^*\)). When we study solutions in 1-Wasserstein space we will only assume that \(\psi \) are Lipschitz (but not bounded). The uniqueness of solutions of (\(\hbox {P}_{\varvec{E}}^*\)) in that setting is discussed in Sect. 3. In Sect. 7 we extend the results to negative Sobolev spaces, and thus we will use some decay (see Remark 7.3).
Remark 2.2
Notice that the problems for the different \(\psi ^i\) are de-coupled from each other.
It is well-known that classical solutions of (\(\hbox {P}_{\varvec{E}}^*\)) may not exist, specially in the case \(D= 0\). The approach developed by Crandall, Ishii and Lions to deal with limit is the notion of viscosity solution (see, e.g., [28]), which we introduce below. Since the theory of viscosity solutions is usually well-posed for Lipschitz functions \(\psi \), a natural metric for \(\mu \) is the 1-Wasserstein distance (see, e.g., [50, p. 207]). We recall that the space \(\mathcal {P}_1 ({\mathbb {R}^d})\) is the space of the probability measures \(\mu \) such that
The distance \(d_1\) that makes it a complete metric space is usually constructed by a Kantorovich optimal transport problem. The reason we can exploit the duality between (\(\hbox {P}_{\varvec{E}}\)) and (\(\hbox {P}_{\varvec{E}}^*\)) is the following well-known duality characterisation (see [50, Theorem 1.14]): if \(\mu , {\widehat{\mu }} \in \mathcal {P}_1 ({\mathbb {R}^d})\) then
where, here and below
We recall the notion of viscosity solution.
Definition 2.3
We say that \(\psi \in C([0,T] \times {\mathbb {R}^d})\) is a viscosity sub-solution of
if, for every \(z_0 = (s_0, x_0) \in [0,T] \times {\mathbb {R}^d}\) and U neighbourhood of \(z_0\) and \(\varphi \in C^2(U)\) touching \(\psi \) from above (i.e., \(\varphi \ge \psi \) on U and \(\varphi (z_0) = \psi (z_0)\)) we have that
Conversely, we say that \(\psi \in C([0,T] \times {\mathbb {R}^d})\) is a super-solution if the inequalities above are reversed for functions touching from below. We say that \(\psi \in C([0,T] \times {\mathbb {R}^d})\) is a viscosity solution if it is both a viscosity sub and super solution.
Thus, we introduce the following notion of solution of (1.1):
Definition 2.4
We say that \((\varvec{\mu }, \{\varvec{\Psi }^T\}_{T\ge 0})\) is an entropy pair if:
-
1.
For every \(T \ge 0\),
$$\begin{aligned}{} & {} \varvec{\Psi }^T: \{ \psi _0: \nabla \psi _0 \in L^\infty ({\mathbb {R}^d}) \}\\{} & {} \qquad \longrightarrow \left\{ (\psi ^1, \cdots , \psi ^n) \in C([0,T] \times {\mathbb {R}^d})^n: \frac{\psi ^i}{1+|x|} \in L^\infty ((0,T) \times {\mathbb {R}^d}) \right\} \end{aligned}$$is a linear map with the following property: for every \(\psi _0\) and \(i = 1, \cdots , n\) we have \(\psi ^{T,i} = \Psi ^{T,i}[\psi _0]\) is a viscosity solution of (\(\hbox {P}^*_i\)).
-
2.
For each \(i = 1, \cdots , n\) and \(T \ge 0\), \(\mu _T^i \in \mathcal {P}_1 ({\mathbb {R}^d})\) and satisfies the duality condition (\(\hbox {D}_i\)).
Remark 2.5
Notice that in our definition we are not assuming that \(\psi _0\) are bounded or decay at infinity, only that they are Lipschitz continuous. This means that they are bounded by \(C(1+|x|)\), so they can be integrated against functions in \(\mathcal {P}_1({\mathbb {R}^d})\).
For convenience, we will simply denote \(\varvec{\Psi } = \{ \varvec{\Psi }^T \}_{T\ge 0} \).
Definition 2.6
We say that \(\varvec{\mu }\) is a dual viscosity solution if there exists \(\varvec{\Psi }\) so that \((\varvec{\mu }, \varvec{\Psi })\) is an entropy pair.
These definitions can be extended to other norms where a duality characterisation exists. The 2-Wasserstein distance, which is natural for many problems because of the gradient-flow structure, does not have any known such characterisation. However, other spaces like \(\dot{H}^{-1}\) do. We present extension of our results to this setting in Sect. 7.
Remark 2.7
Notice that this duality characterisation is somewhat in the spirit of the Benamou–Brenier formula for the 2-Wasserstein distance between two measures, where the optimality conditions involve viscosity solutions of a Hamilton–Jacobi equation.
Remark 2.8
Notice that we have not requested the time continuity of \(\mu \), so the notion of initial trace is not clear. However, it will follow from the continuity of \(\psi ^T\). Under some assumptions on \(\varvec{{\mathfrak {K}}}\) we will show that \(\mu \in C([0,T]; \mathcal {P}_1 ({\mathbb {R}^d}))\). However, in more general settings, the definition guarantees that the initial trace is satisfied in the weak-\(\star \) sense. Notice that if \(\psi _0 \in C_c\) and \(\psi _t\) is continuous in time then
Remark 2.9
(General linear diffusions). Our definitions, methods, and results can be extended to the case where \(D\Delta u\) is replaced by any other linear operator that commutes with \(\partial _x\), and that satisfies the maximum principle, e.g., the fractional Laplacian \(D(-\Delta )^s\).
2.2 Well-posedness when \(\varvec{K}\) is regularising
So far, \(\varvec{E}_t = -\varvec{K}_t [\varvec{\mu }_t]\), i.e., we assume that at every time t the convection comes from a function \(\varvec{K}_t : \varvec{\mu }\rightarrow \varvec{E}\). This covers, for example, non-local operators in space. However, our arguments extend to much broader arguments, where \(\varvec{K}\) could be non-local in time as well. We will consider the more general case of
The results in this section correspond to \(X = \mathcal {P}_1({\mathbb {R}^d})^n\) and Y the space
However, the scheme of the proof is general and can be extended to other X, Y. In Sect. 7 we work on the negative homogeneous Sobolev space.
Notice that \({Lip}_0 ({\mathbb {R}^d}; {\mathbb {R}^d})\) is very similar \(\dot{W}^{1,\infty }\), but we avoid this terminology to endow it with the following special norm
The reason for the \({Lip}_0\) nomenclature is that
so it guarantees the bound of \(\varvec{E}(0)\). It is not difficult to see that this is a Banach space. In this setting, we prove the following well-posedness result
Theorem 2.10
Let \(D\ge 0\), \(T > 0\), and assume that for all \(t > 0\)
with the property (H), maps bounded sets into bounded sets, and satisfies the following Lipschitz condition
Then, for each \(\varvec{\mu }_0 \in \mathcal {P}_1 ({\mathbb {R}^d})^n\) there exists a unique dual viscosity solution of (P), \(\mu \in C([0,T]; \mathcal {P}_1({\mathbb {R}^d})^n)\), and it depends continuously on the initial datum with respect to the \(d_1\) distance. This solution \(\mu \) also satisfies (P) in distributional sense. Furthermore, if \(\varvec{\mathfrak {K}}^i[\varvec{\mu }]_t = K^i[\varvec{\mu }_t]\), then the map \(S_T : \varvec{\mu }_0 \in \mathcal {P}_1 ({\mathbb {R}^d})^n \mapsto \varvec{\mu }_T \in \mathcal {P}_1({\mathbb {R}^d})^n\) is a continuous semigroup.
Going back to (2.2) we have that
Thus, we can use the duality relation (\(\hbox {D}_i\)) to get the following upper bound
To exploit this fact we take advantage of the following, rather obvious, estimate
In fact, we show that, on the second supremum in (2.7), we can focus on functions such that \(\psi _0(0) = 0\). The reason for this election will be made clearer later.
Lemma 2.11
Let \(\mu \) and \({\widehat{\mu }}\) two dual viscosity solutions of (\(\hbox {P}_{\varvec{E}}\)) such that \(\varvec{E}, \widehat{\varvec{E}} \in C([0,T];{{\,\textrm{Lip}\,}}_0 ({\mathbb {R}^d}, {\mathbb {R}^d}))\). Then, letting \(\psi ^T\) and \({\widehat{\psi }}^T\) be the corresponding solutions of the dual problems (\(\hbox {P}_{\varvec{E}}^*\)), we have that
Proof
Notice that \(\psi ^T = \Psi ^T[\psi _0]\), by linearity
By the uniqueness in Proposition 3.1, which we will prove below, since the problem does not contain a zero-order term \(\Psi ^T [1] = 1 ={\widehat{\Psi }}^T[1]\). Therefore, we can write
Thus, we can take the second supremum of (2.7) exclusively on functions such that \(\psi _0(0) = 0\). \(\square \)
Remark 2.12
We recall a basic property of the 1-Wasserstein distance. It turns out that the first moment is precisely the distance to the Dirac delta, i.e.,
Hence, our fixed point argument will be performed in the balls
This explains that it is natural to work with the metric of \((1 + |x|) L^\infty \).
We will devote Sect. 3 to the estimates of \(\psi \) that allow us to control each term of (2.8). The estimates are recovered by standard, albeit involved, arguments.
The first term in (2.8), which corresponds to continuous dependence respect to the initial datum, requires only the estimation of a coefficient. If \(\varvec{\mathfrak {K}}\) is well-behaved, we expect to obtain an estimate of the type
For the well-posedness theorem \({\widehat{\mu }}_0 = \mu _0\), we can avoid the first term in (2.8) and this constant is not relevant.
The second term in (2.8), which represents continuous dependence on \(\varvec{E}\) for the dual problem, is harder to estimate. We prove first that even if \(\psi _0 \notin L^\infty \) but is Lipschitz and \(\varvec{E}, \varvec{\widehat{E}}\) are bounded, then \(\psi _s^T - {\widehat{\psi }}_s^T \in L^\infty ({\mathbb {R}^d})\). A second, more elaborate, argument will allow us to work of the case where \(\varvec{E}\) grows no-more-than linearly at infinity. If (\(\hbox {P}_{\varvec{E}}^*\)) has good continuous dependence on \(\varvec{E}\), we can expect
where we leave the second norm unspecified for now. So, if \(\varvec{\mathfrak {K}}\) is Lipschitz in this suitable norm, then we will be able to use Banach’s fixed point theorem.
Remark 2.13
Karlsen and Risebro [39] prove that if \(\rho _0, {\widehat{\rho }}_0 \in L^1 \cap L^\infty \cap BV\) and \(\rho , {\widehat{\rho }} \in L^\infty ([0,T]; L^1)\) are entropy solutions of
where \(\varvec{E}, \varvec{\widehat{E}}: {\mathbb {R}^d}\rightarrow {\mathbb {R}^d}\), and we have an a priori estimates \(\rho , {\widehat{\rho }}: {\mathbb {R}^d}\rightarrow I \subset \mathbb {R}\) with I bounded and
Then,
where \(\rho , {\widehat{\rho }} : {\mathbb {R}^d}\rightarrow I \subset \mathbb {R}\) and C depends on \(C_\rho \) and the norms of \(f,{\widehat{f}} \in W^{1,\infty }\), \(\varvec{E}, \varvec{\widehat{E}} \in L^\infty \cap BV\). In [30] the authors use the fixed-point argument of entropy solutions only to prove uniqueness. Existence is done by proving the convergence of the particle systems, which is their main aim. They use [39] which deals directly with the conservation law in \(L^1\), we obtain simple estimates for the Hamilton–Jacobi dual problem in \(L^\infty \) using the maximum principle. Our duality argument is not directly extensible to \(L^1 ({\mathbb {R}^d})\). Applying the duality characterisation of \(L^1\) and (\(\hbox {D}_i\)) we can compute
Due to the maximum principle we will show that \(|\psi _s^T| \le 1\) and hence
There is no apparent way to bound the second supremum with a quantity related only to \(|\varvec{E} - \varvec{\widehat{E}}|\). A quantity depending on \(|\nabla \psi _0|\) appears, and there is no available bound of its supremum over functions such that \(|\psi _0| \le 1\).
2.3 Stability theorem
We conclude the main results by stating a stability theorem. This allows to prove existence of solutions by approximation, in more general settings that the well-posedness theory. We will use it below for several applications.
Theorem 2.14
Let \((\varvec{\mu }^k, \varvec{\Psi }^k) \) be a sequence of entropy pairs in the sense of Definition 2.4 corresponding to some operators \(\varvec{{\mathfrak {K}}}^k\) under the assumptions of Theorem 2.10. Assume that
-
1.
\(D^k \rightarrow D^\infty \).
-
2.
For every \(t> 0\), \(\varvec{\mu }_t^k \overset{\star }{\rightharpoonup }\varvec{\mu }_t^\infty \) in \(\mathcal {M} ({\mathbb {R}^d})\).
-
3.
For every \(\psi _0 \in C_c^\infty ({\mathbb {R}^d})\) and \(T > 0\) we have \(\varvec{\Psi }^{T,k} [\psi _0] \rightarrow \varvec{\Psi }^{T,\infty }[ \psi _0 ]\) uniformly over compacts of \([0,T] \times {\mathbb {R}^d}\). This allows to pass to the limit in estimate (3.3) below.
-
4.
\(\varvec{{\mathfrak {K}}}^k [\varvec{\mu }^k] \rightarrow \varvec{{\mathfrak {K}}}^\infty [\varvec{\mu }^\infty ]\) uniformly over compacts of \([0,T] \times {\mathbb {R}^d}\).
-
5.
For every \(\psi _0 \in C_c^\infty ({\mathbb {R}^d})\) we have that
$$\begin{aligned} \int _{\mathbb {R}^d}\varvec{\Psi }^{T,k}[\psi _0]_T \text {d}\varvec{\mu }_0^k \rightarrow \int _{\mathbb {R}^d}\varvec{\Psi }^{T,\infty }[\psi _0]_T \text {d}\varvec{\mu }_0^\infty . \end{aligned}$$
Then \((\varvec{\mu }^\infty , \varvec{\Psi }^\infty )\) is an entropy pair of the limit problem.
The result follows directly from the definition and the following classical result of stability of viscosity solutions (see, e.g., [28]).
Theorem 2.15
Let \(D^k \ge 0\), \(\psi ^k \in C([0,T] \times {\mathbb {R}^d})\), \(\varvec{E}^k \in C([0,T]\times {\mathbb {R}^d})^d\) be viscosity solutions of
and assume that \(D^k \rightarrow D^\infty \) and \(\psi ^k \rightarrow \psi ^\infty \), \(\varvec{E}^k \rightarrow \varvec{E}^\infty \) uniformly over compacts of \([0,T] \times {\mathbb {R}^d}\). Then, \(\psi ^\infty \) is a viscosity solution of
2.4 The limits of the theory: the diffusive Newtonian potential in \(d=1\)
Let us consider the case \(n = d = 1\) with K coming from the gradient-flow structure (1.4) and the Newtonian potential \(W(x) = -|x|\). Then \(\nabla W(x) = {{\,\textrm{sign}\,}}(x)\). This W falls outside the theory developed in Theorem 2.10, but can be approximated by solutions in this framework. It was proved in [11] that for \(\mu _0 = \delta _0\) then the gradient flow solution is not given by particles but rather
Naturally, any smooth approximation of W (for example \(W_k = -(x^2 + \frac{1}{k})^{\frac{1}{2}}\)) yields \(\mu ^k_t = \delta _0\). However, solving with \(\mu _0^m = \frac{1}{2m} \chi _{[-m,m]}\) gives the expected reasonable solution. We therefore have the following diagram
Clearly, \(\mu _t^k \rightarrow \delta _0\) in any conceivable topology. Checking whether \(\delta _0\) is or is not a distributional solution for W is not totally trivial. Notice that \({{\,\textrm{sign}\,}}\) (as a distributional derivative of |x|) is not pointwise defined at 0, so the meaning of \(\int \varphi {{\,\textrm{sign}\,}}* \delta _0 \text {d}\delta _0\) is not completely clear. With the choice \({{\,\textrm{sign}\,}}(0) = 0\) one could be convinced that \(\mu _t = \delta _0\) is a distributional solution. But this seems arbitrary.
In fact, if one takes the approximating entropy pairs \((\mu ^k, \Psi ^k)\), it is not difficult to show that a discontinuity appears in \(\psi _s^{k,T}\) as \(k \rightarrow \infty \). Hence, we do not have uniform convergence of \(\psi ^{k,T}\). To show this fact, let \(E^k = -\partial _x (W_k * \mu ^k)\). Since \(\mu ^k = \delta _0\), we have that
Since \(E^k\) does not depend on t, \(\psi ^{k,T}\) does not depend on T. Therefore, we drop the T to simplify the notation. We have to solve the equation
This equation can be solved by characteristics \(\psi ^k_t (Y^k_t(y_0)) = \psi _0 (y_0)\) given by
As \(k \rightarrow \infty \) we recover \( Y^{k} _t (y_0) \rightarrow Y_t(y_0) = y_0 - {{\,\textrm{sign}\,}}(y_0) t \) or, equivalently, that
Therefore, \(\psi _t(0^-) = \psi _0 (-t)\) and \(\psi _t(0^+) = \psi _0 (t)\). Thus, \(\psi _t\) is not continuous in general. So we cannot pass to the limit \(\mu ^k \rightarrow \mu \) in terms of entropy pairs.
The gradient flow solution is recovered by the entropy pairs if one approximates \(\delta _0\) by \(\mu _0^m = \frac{1}{2m} \chi _{[-m,m]}\) and passes to the limit. It is also not difficult to show that the entropy pairs of twice regularised problem, \((\mu ^{k,m}, \Psi ^{k,m})\) do converge in the sense of entropy pairs as \(k \rightarrow \infty \) to \((\mu ^m, \Psi ^m)\). And then we can also pass to the limit in the sense of entropy pairs as \(m \rightarrow \infty \).
Thus, the stable semigroup solutions are entropy pairs, and this notion of solution is powerful enough to discard the wrong order of the limits.
2.5 Open problems
We finally briefly discuss the key difficulties for some related problems that our duality theory cannot cover.
2.5.1 Nonlinear diffusion
We could deal, in general, with problems of the form
Going back to (2.10), the dual problem can be written in the form
We have not been able to estimates of \(\Vert \nabla ( \psi - {\overline{\psi }}) \Vert _{L^\infty }\) of \(\Vert \psi - {\overline{\psi }} \Vert _{H^1}\) with respect to \(E - {\overline{E}}\) and \(\varvec{A} - \overline{\varvec{A}}\) using the techniques below. The main difficulty is that letting \(v = \frac{\partial \psi }{\partial x_i}\) and \({\overline{v}} = \frac{\partial {\overline{\psi }}}{\partial x_i}\), the reasonable extension of (3.11) contains a term \(-\nabla \cdot ( (\varvec{A} - \overline{\varvec{A}}) \nabla {\overline{v}}) \). Controlling these terms requires estimates of \(D^2 \psi \), which are not in principle present for the initial data we discuss. However, smarter estimates and choices of space for \(\mu \) and \(\psi \) may lead to a successful extension of our results.
2.5.2 Coupling in the dual problem
Another possible extension is the analysis of more general systems of the form
The dual problem is now not decoupled, and we have to solve it a system
This kind of problems does not have a comparison principle in general, for example
leads to the wave equation \(\partial _{tt} u =\partial _{t} (\partial _x v) = \partial _x \partial _t v = \partial _{xx} u\). Some examples of equations resembling this structure appear in mathematical biology models, e.g., [47].
2.5.3 Problems with saturation
Some authors have studied the case \(\varvec{{\mathfrak {K}}}_i [\varvec{\mu }] = \frac{\theta (\mu )}{\mu } H_i [\varvec{\mu }] \) where \(\theta (0) = \theta (M) = 0\). This problem was studied for \(n = 1\) and \(D = 0\) by [30] using a fixed point argument with the estimates by [39]. In this setting the solutions with initial datum \(0 \le \rho _0 \le M\) remain bounded. Therefore, we expect bounded (even continuous) solutions \(\rho \). So the natural duality would be \(\psi \) integrable. We point the reader to [1], where the authors study the duality between \(L^\infty \) estimates for entropy solutions of conservation laws and \(L^1\) bounds for viscosity solutions of Hamilton–Jacobi equations.
3 Viscosity solutions of (\(\hbox {P}_{\varvec{E}}^*\))
The aim of this section is to prove the following existence, uniqueness, regularity and continuous dependence result for (\(\hbox {P}_{\varvec{E}}^*\)) which is one of the key tools in this paper. The usefulness of the estimates obtained is clear going back to Lemma 2.11.
Proposition 3.1
Let \(D \ge 0\), \(\varvec{E} \in C([0,T]; {{\,\textrm{Lip}\,}}_0 ({\mathbb {R}^d})) \) and \(\psi _0 \) be such that \( \nabla \psi _0 \in L^\infty ({\mathbb {R}^d}) \). Then, there exists a unique viscosity solution of (\(\hbox {P}_{\varvec{E}}^*\)) such that \(\frac{ \psi }{ 1 + |x| } \in L^\infty ((0,T)\times {\mathbb {R}^d})\). Furthermore, it satisfies the following estimates
where \(C = C(d)\) depends only on the dimension, and \(\left\| \!\left| \nabla \right\| \!\right| \psi _s \left\| \!\left| _{L^\infty }\right\| \!\right| = \sup _i \Vert \tfrac{\partial \psi }{\partial x_i}\Vert _{L^\infty }\). If \(\psi _0 \ge 0\) then \(\psi \ge 0\). Moreover, we obtain the following time-regularity estimate
where \( C = C(T,D, \left\| \psi _0 \right\| _{Lip_0} , \sup _\sigma \Vert \varvec{E}_\sigma \Vert _{Lip_0} )\). Lastly, given \({\widehat{\psi }}_0 = \psi _0\) and \(\varvec{\widehat{E}}\) in the same hypotheses above, there exists a corresponding solution of (2.3), denoted by \({\widehat{\psi }}\), and we have the continuous dependence estimate
where \(C = C\left( d, D, \Vert \psi _0 \Vert _{Lip_0}, \int _0^T \left\| E_{T-\sigma } \right\| _{Lip_0} \right) \) is monotone non-decreasing in each entry.
We will prove this result in Sect. 3.2. The uniqueness of viscosity solutions leads to the following
Corollary 3.2
Assume that \(\varvec{{\mathfrak {K}}}: C([0,T]; \mathcal {P}_1 ({\mathbb {R}^d})) \rightarrow \mathcal {C}([0,T]; {Lip}_0({\mathbb {R}^d}; {\mathbb {R}^d}))\), and let \((\mu , \Psi )\) and \((\mu , {\widehat{\Psi }})\) be entropy pairs of (1.1), then \(\Psi = {\widehat{\Psi }}\).
There are some immediate consequences that come from the linearity of equation.
Remark 3.3
Notice that the properties above imply a comparison principle. Given two solutions, we have that \(\psi _0 \le {\widehat{\psi }}_0\), then \({\widetilde{\psi }}_0 = {\widehat{\psi }}_0 - \psi _0 \ge 0\). Then, by linearity, uniqueness, and preservation of positivity \({\widehat{\psi }} - \psi = {\widetilde{\psi }} \ge 0\).
Notice that the generic initial datum \({{\,\textrm{Lip}\,}}(\psi _0) \le 1\), need not be bounded. This can initially seem like a problem for uniqueness, since we cannot prescribe conditions at infinity. However, following Aronson [5, Theorem 2 and 3], existence and uniqueness are obtained under the assumption that \(e^{-\lambda |x|^2} u_t \in L^2 ((0,T) \times {\mathbb {R}^d})\) if the coefficients are bounded. We tackle this issue by studying weighted versions of our solutions \( v_s(x) = \psi _s(x) / \eta (x) \), which solve the following problem
Notice that, if \(\eta (x) = (1 + |x|^2)^{k/2}\) with \(k \ge 1\), then \(\nabla \eta / \eta \sim |x|^{-1}\) at infinity. If E is Lipschitz, then all the coefficients of the equation above are bounded. And, if \(k > 1\), then \(|v_0(x)|\le C(1+|x|)^{1-k} \rightarrow 0 \) as \(|x|\rightarrow \infty \).
Remark 3.4
A similar argument can be adapted to initial data which are weighted with respect to \(1 + |x|^p\) for any \(p \ge 1\), if this is satisfied by the initial datum. However, this escapes the interest of this work.
3.1 A priori estimates in \(L^\infty \)
3.1.1 General linear problem
For the dual problem we are interested in the existence and uniqueness of the linear parabolic problem in non-divergence form
The theory of existence and uniqueness of classical solutions dates back to [41]. When the coefficients are smooth, the linear problem can be rewritten in divergence form as
We focus on obtaining suitable a priori estimates assuming that the data and solutions are smooth enough, and these estimates pass to the limit to the unique viscosity solution by approximation of the coefficients.
By studying (3.7) at the point of maximum/minimum, we formally have that
This intuition can be made precise by the following result
Lemma 3.5
If \(a,f \in C({\mathbb {R}^d})\) and bounded, \(\varvec{b} \in C({\mathbb {R}^d})^d\), and u is a classical solution of (3.7) such that \(\sup _{(0,T) \times \partial B_R} |u| \rightarrow 0\) as \(R \rightarrow \infty \) for each T , then
Proof
Define
This is a classical super-solution of (3.7) and \(u_0 \le {\overline{u}}_0\). Using the weak maximum principle on a ball \(B_R\) and the initial condition we get
As \(R \rightarrow \infty \) this tends to 0 by hypothesis. We conclude that \(u \le {\overline{u}}\). Similarly, we prove that \(-u \le {\overline{u}}\). \(\square \)
Now we compute estimates on the solution formally assuming that \(\psi \in C ([0,T]; C_0^1( {\mathbb {R}^d}) )\), and that \(\varvec{E}\) is smooth and satisfies all necessary bounds. Later, we will justify this formal computations.
3.1.2 \(L^\infty \) estimates of \(\psi ^T\), \(\nabla \psi ^T\) and \(\psi ^T/(1+|x|)\)
Going to (\(\hbox {P}_{\varvec{E}}^*\)) when we have suitable decay of \(\psi \) we deduce (3.1) by Lemma 3.5. When we define \( U^i_s = \frac{\partial \psi _s^T}{\partial x_i }, \) we recover that
Since this is a system, we cannot directly apply Lemma 3.5. Nevertheless, we compute
Using the norm equivalence of norms of \({\mathbb {R}^d}\), we have
Eventually, we have that
As \(p \rightarrow \infty \), we obtain (3.2). Take \(\eta (x) = (1 + |x|^2)^{\frac{1}{2}}\). Recalling (3.6) and applying Lemma 3.5, we deduce
Since \(\eta \) is regular we eventually deduce (3.3) where C is such that \(C^{-1} \le \frac{\eta (x)}{1+|x|} \le C\), and it does not depend even on the dimension. For the continuous dependence, we write (dropping s and \(T-s\) from the subindex for convenience)
Then we have
and
Notice that
Hence, we can deduce using Lemma 3.5 that
where
which can be estimated using (3.2) and (3.3).
Eventually, since \(\psi _0^T - {\widehat{\psi }}_0^T = 0\), we recover (3.5). Notice that \(C_1\) cannot be uniformly bounded over the set \({{\,\textrm{Lip}\,}}(\psi _0) \le 1\), where we can bound \(C_2\). Here is where the assumptions that \(\varvec{{\mathfrak {K}}}_i\) satisfies (2.5) and (2.6) will come into play in Sect. 5.
3.1.3 Time continuity
Taking \(v_s = \psi _{s} / \eta ^{(1)}\) and \({\widehat{v}}_s = \psi _{s+h} / \eta ^{(1)}\) where \(\eta ^{(1)} = (1 + |x|^2)^{\frac{1}{2}}\), we similarly deduce going back to (3.12) that
Therefore, \(\psi \) inherits the time continuity of E. Then the only remaining difficulty is the time continuity at 0.
For \(D > 0\) we use Duhamel’s formula for the heat equation \(u_t - D \Delta u = f\), where we denote the heat kernel \(K_D\). Notice that \(K_D(t,z) = K_1 (Dt,z) \). For the first term we have that
We estimate as follows
Now since \(E(y)/(1+|y|)\) and \(\nabla \psi \) are bounded, it leaves to integrate
Eventually, we recover using (3.2) that
This result can also be deduced for \(D = 0\) without involving any convolution. Also, it is recovered as a limit \(D \searrow 0\). Joining (3.14) and (3.15) we recover (3.4).
Remark 3.6
(Time continuity at \(s=0\) if \(\Delta \psi _0\) is bounded or \(D = 0\)). To estimate the time derivative at time 0, we consider the candidate sub and super-solutions
We have that
So we need
Then
As \(s \rightarrow 0\), we get an estimate of the time derivative at \(s = 0\). A similar computation can be done for \(\psi / (1 + |x|)\). This kind of result is useful for the case \(D = 0\), since even as \(D_k \searrow 0\), one can take approximate initial data \(\psi _0^{(k)}\) so that the constant is uniformly bounded.
3.2 Proof of Proposition 3.1
First we prove uniqueness. Take \(v^{(2)} = \psi / \eta ^{(2)}\) where \(\eta ^{(k)}(x) = (1 + |x|^2)^{k/2}\). We observe that if \(\frac{\psi _s}{1+|x|} \in L^\infty \) then \( v_0^{(2)} \in L^\infty ({\mathbb {R}^d})\), with decay 1/|x| at infinity. By dividing viscosity test functions by \(\eta ^{(2)}\), we observe that \(v^{(2)}\) is a viscosity solution of (3.6). Notice since \(v^{(2)}\) remains bounded, \(\sup _{(0,T) \times \mathbb R^d {\setminus } B_R } \psi / (1 + |x|)\) is bounded by \(R^{-\frac{1}{2}}\). The uniqueness of \(\psi \) follows from the uniqueness of \(v^{(2)}\) (see, e.g., [45, Theorem 3.1]).
Let us now construct the solution and prove its properties. If \(D> 0\), \(\psi _0 \in W_c^{2,\infty } ({\mathbb {R}^d})\), and \(\varvec{E}\) is very regular, then existence is simple by standard arguments. For the problem in \({\mathbb {R}^d}\), regularity and decays as \(|x| \rightarrow \infty \) for compactly supported initial data follow as in [34] and hence all the estimates above are justified. If \(\psi _0 \ge 0\) we can proceed like in Lemma 3.5 to prove \(\psi \ge 0\).
Now we consider the general setting, and we argue by approximation. Consider an approximating sequence \(0 < D^{(k)} \rightarrow D\), satisfying
We have the uniform continuity estimates (3.2) and (3.4). Due to the Ascoli–Arzelá theorem and a diagonal argument, \(\psi ^{(k)} \rightarrow \psi \) uniformly over compacts of \([0,T] \times {\mathbb {R}^d}\). Applying Lemma 2.15, the limit is a viscosity solution with the limit coefficients. By the uniqueness above this is the solution we are studying. All estimates are stable by convergence uniformly over compact sets. \(\square \)
4 Dual-viscosity solutions of problem (\(\hbox {P}_{\varvec{E}}\))
Now we focus on showing well-posedness and estimates for (\(\hbox {P}_{\varvec{E}}\)). We construct the solutions as the duals of those in Sect. 3.
Proposition 4.1
For every \(D, T \ge 0\) and \(\varvec{E} \in C([0,T] ; {{\,\textrm{Lip}\,}}_0 ({\mathbb {R}^d}, {\mathbb {R}^d})\), and \(\mu _0 \in \mathcal {P}_1 ({\mathbb {R}^d})\) there exists exactly one dual viscosity solution \(\mu \in C([0,T]; \mathcal {P}_1 ({\mathbb {R}^d}))\) of (\(\hbox {P}_{\varvec{E}}\)). It is also a distributional solution. Furthermore, the map
is continuous with the following estimate
depends monotonically on each entry. The semigroup property holds in the sense that
Therefore, we have constructed a continuous flow
Proof of Proposition 4.1
We begin by proving uniqueness and continuous dependence. For \(D \ge 0\) and any two weak dual viscosity solutions \(\mu \) and \({\widehat{\mu }}\) corresponding to \((\mu _0, \varvec{E})\) and \(({\widehat{\mu }}_0, \varvec{\widehat{E}})\) we have, applying Lemma 2.11, (3.3), and (3.5), that
The second constant obtained from (3.13) by taking the supremum when \({{\,\textrm{Lip}\,}}(\psi _0) \le 1\). Eventually, we can simplify this expression to (4.1). Therefore, we have uniqueness of dual viscosity solutions for any \(D \ge 0\).
For \(D> 0, \mu _0 \in H^1({\mathbb {R}^d})\) and \(\varvec{E}_t \in C([0,T]; W^{2,\infty } ({\mathbb {R}^d}; {\mathbb {R}^d}))\), the theory of existence and regularity of weak solutions is nowadays well known (see, e.g., [9, 10] and the references therein). The dual problem is decoupled from (\(\hbox {P}_{\varvec{E}}\)), and we have already constructed the unique viscosity solutions of the dual problem (see Proposition 3.1), that are also weak solutions when \(\varvec{E}\) is regular. For regular enough datum \(\psi _0\), we can use \(\psi \) as test function in the weak formulation of (\(\hbox {P}_{\varvec{E}}\)) to deduce (\(\hbox {D}_i\)). Hence, any weak solution of (\(\hbox {P}_{\varvec{E}}\)) is a dual viscosity solution.
Let us now show the time continuity with respect to the \(\mathcal {P}_1 ({\mathbb {R}^d})\) distance in space. We take \(\psi _0\) such that \({{\,\textrm{Lip}\,}}(\psi _0) \le 1\), and we estimate
This last quantity is controlled by continuous dependence on \(\varvec{E}\) and time continuity of (\(\hbox {P}_{\varvec{E}}^*\)). First, letting \({\widehat{\psi }}_s = \psi ^{t+h}_s\) with \(\psi ^{t+h}_0 = \psi _0\), which corresponds to \(\varvec{\widehat{E}}_s = \varvec{E}_{t+h-s}\), we have that
The right-hand side of this equation is a modulus of continuity, which we denote \(\omega _E\). Now we use the time continuity of (\(\hbox {P}_{\varvec{E}}^*\)) given by (3.4) to deduce that
where C and \(\omega _D\) are given by the right-hand side of (3.4). The constants are uniform for \(\psi _0\) with \({{\,\textrm{Lip}\,}}(\psi _0) \le 1\). Eventually, taking the supremum on \(\psi _0\) and applying (2.2) we deduce
Let us now consider the general case for \(D\ge 0\), \(\mu _0\) and \(\varvec{E}_0\). The uniform estimates (4.1) shows that, if \(0 < D^n \rightarrow D, H^1({\mathbb {R}^d}) \ni \mu _0^n \rightarrow \mu _0\) in \(\mathcal {P}_1 ({\mathbb {R}^d})\) and \(C([0,T]; W^{2,\infty } ({\mathbb {R}^d}; {\mathbb {R}^d})) \ni \varvec{E}^n \rightarrow \varvec{E}\) in \(C([0,T]; {Lip}_0 ({\mathbb {R}^d},{\mathbb {R}^d}))\), then the sequence \(\mu ^n\) is Cauchy in the metric space \(C([0,T]; \mathcal {P}_1 ({\mathbb {R}^d}))\). Since this is a Banach space, the sequence \(\mu ^n\) converges to a unique limit \(\mu \). Due to the stability of the dual problem, we can pass to the limit in (\(\hbox {D}_i\)) to check that \(\mu \) is a dual viscosity solution. We have already shown the uniqueness at the beginning of the proof. Due to the approximation, the solution constructed is also a distributional solution.
The semigroup property holds in the regular setting, so we can pass to the limit. This completes the proof. \(\square \)
5 Existence and uniqueness for (1.1). Proof of Theorem 2.10
We prove the existence by using Banach’s fixed-point contraction theorem. We construct the map
by
We must work on the bounded cubes \(Q(R) = B_{\mathcal {P}_1} (R)^n\) where we recall the definition of ball in Wasserstein space given by (2.9). Hence, we take
To get a very rough estimate of \(d_1 (\mathcal {T}_T[\varvec{\mu }], \delta _0)\) we first indicate that when \(\varvec{E} = 0\) we recover the heat kernel at time \(Dt\)
Hence, using the triangle inequality
For the first term, we simply write
To get a uniform constant in (4.1) define
by the assumptions. As a consequence, for any \(\mathbf \mu \) such that \(d_1 (\mu _t^i, \delta _0) \le R\) for all \(i = 1, \cdots , n \) and \(t \le T_1\), then \( d_1 (\delta _0, \mathcal {T}_t[\varvec{\mu }]^i) \le R \) for all \(i = 1, \cdots , n \) and \(t \le T_1\), i.e.,
Notice that, since we are constructing solutions using \(S_T\), the constructed solution is also a distributional solution.
Applying again (4.1) we have that for every \(i = 1, \cdots , n\)
where \(C(T_1,R)\) is recovered again through (4.1). Given that R is fixed, we can find \(T_2 < T_1\) so that the map \(\mathcal {T}_{T_2}\) is contracting with the norm
Therefore, we can apply Banach’s fixed-point theorem to proof existence and uniqueness for short times. Since \(\mathfrak {K}\) is uniformly Lipschitz, we can extend the existence time to infinity applying the classical argument. To prove the continuous dependence on \(\mu _0\), we apply (2.8), (3.5) and (2.6). This completes the proof of existence and uniqueness. \(\square \)
Remark 5.1
Following the non-explicit constant in (3.5), it would be possible to recover quantitative dependence estimates.
Remark 5.2
(Numerical analysis when \(D= 0\): the particle method). Our notion of solution justifies the convergence of the particle method when \(D= 0\). The aim of the particle method is to consider an approximation of the initial datum given by finitely many isolated particles
Then, it is not difficult to see that, for \(D= 0\) the solution is given by particles
The evolution of these particles is given by a system of ODEs for the particles
This system is well posed. Due to the continuous dependence
It is easy to see that the finite combinations of Dirac deltas is dense in 1-Wasserstein distance, and estimates for convergence are well known (see, e.g., [37]).
6 Gradient flows of convex interaction potentials and \(D= 0\)
The aim of this section is to give an application of our results to the classical aggregation equation
This problem falls into the well-posedness theory developed in Theorem 2.10 provided that \( D^2 W \in L^\infty ({\mathbb {R}^d})^{d\times d}\) for the \(\mathcal {P}_1 ({\mathbb {R}^d})\) theory. In the context of gradient-flow solutions (see, e.g., [2]) we are able to weaken the hypothesis on \(\mathfrak {K}\).
Theorem 6.1
Let \(\mu _0 \in \mathcal {P}_2({\mathbb {R}^d}) \) be compactly supported. Assume W is convex and that \(W \in C^{1,s}_{loc} ({\mathbb {R}^d})\). Then, the gradient flow solution \(\mu \in C([0,T]; \mathcal {P}_2 ({\mathbb {R}^d}))\) is a dual-viscosity solution.
We can approximate W by a sequence of convex functions \(W^k\), so that \(\nabla W^k \rightarrow \nabla W\) uniformly over compacts but satisfies \(D^2 W^k \in L^\infty ({\mathbb {R}^d}) \). We construct \(\mu ^k_t\) through Theorem 2.10. Since \(\mu _0\) is compactly supported and \(W^k\) is convex, if we pick \(A_0\) the convex envelope of \({{\,\textrm{supp}\,}}\mu _0\), then
This is easy to see, for example, looking at the solution by characteristics. We set ourselves in the hypothesis of Theorem 2.14 by proving below that
-
1.
\(\psi ^{k,T}\) for every \(\psi _0 \in C_c^\infty ({\mathbb {R}^d})\) are uniformly equicontinuous.
-
2.
\(\mu ^k \rightarrow \mu \) in \(C([0,T]; \mathcal {P}_2 ( {\mathbb {R}^d}))\).
-
3.
\(W^k * \mu ^k \rightarrow W*\mu \) uniformly over compacts.
Remark 6.2
We point that the result in Theorem 6.1 can be extended from (6.1) to a system of equations coming from the gradient flow of interaction potentials between the different components. To fix ideas, we can treat systems of two species like
See, e.g., [31].
6.1 Regularity in space
Constructing solutions by characteristics it is well known that
then characteristics grow apart. Then, the solution of (2.3) satisfies
We check that this holds for \(\varvec{E} = \nabla W * \mu \). Since W is convex, then
Then, the convolution with a non-negative measure is also convex
and so (6.2) holds.
6.2 Regularity in time
First, we look at the evolution of the support. When \(\varvec{E}_{T-s}\) is locally bounded, we can construct a super-solutions by characteristics. We define
If we assume that \(A(s,0) \subset B(x_0,R_s)\) then
So we end up with an estimate
Assume that \(\psi _0 \in C_c^\infty ({\mathbb {R}^d})\) with \(\Vert \nabla \psi _0\Vert _{L^\infty ({\mathbb {R}^d})} \le 1\). Take \(s_0 < s\). Consider
We have that
They are a sub and super-solution in \([s_0,s_1]\) if
Hence, we deduce that
This implies that a uniform bound on the time continuity based on T, local bounds of \(\varvec{E}\) and the support of \(\psi _0\).
6.3 Convergence of the convolution
Following [2, Theorem 11.2.1] the \(\Gamma \)-convergence of uniformly \(\lambda \)-convex interaction free energy functionals is sufficient so that
In particular, if \(W \in C^s\) is convex and \(W^k (x) (1 +|x|)^{-2} \rightarrow W(x) (1 + |x|)^{-2}\) uniformly, (6.5) holds. It follows that the sequence \(\mu ^k\) is uniformly continuous, i.e., the function
is a modulus of continuity.
On the other hand, for the convergence of \(\nabla W^k * \mu ^k\) we prove the following result.
Lemma 6.3
Assume that
-
1.
\(\nabla W \in C^s_{loc}\)
-
2.
\(\nabla W^k \rightarrow \nabla W\) uniformly over compacts of \({\mathbb {R}^d}\).
-
3.
\(\mu _t^k, \mu _t \in \mathcal {P} ({\mathbb {R}^d})\), and, for every \(t > 0\), \(\mu ^k_t \rightharpoonup \mu _t\) weak-\(\star \) in \(\mathcal {M} ({\mathbb {R}^d})\)
-
4.
There exists \(A_0 \subset {\mathbb {R}^d}\) convex bounded such that for all \(t > 0\), \({{\,\textrm{supp}\,}}\mu _t^k \subset A_0\)
-
5.
\(\mu ^k \in C([0,T]; \mathcal {P}_2 ({\mathbb {R}^d}))\) are uniformly continuous.
Then
Proof
We use the intermediate element \(\nabla W * \mu ^k\). First, for \(A \subset {\mathbb {R}^d}\) compact
Hence, we have that
Due to weak-\(\star \) convergence, if \(\nabla W \in C^s_{loc}\) then
Now we prove uniform continuity. First in space. Let \(A \subset {\mathbb {R}^d}\) be compact. Take
Now, for \(x,y \in A\) we can compute
So \(\nabla W * \mu ^k\) is uniformly continuous in x over compacts of \([0,T] \times {\mathbb {R}^d}\).
Lastly, let \(\pi \) be optimal plan between \(\mu ^k_t\) and \(\mu ^k_{\tau }\). Due to assumption 4, the optimal plan can be selected so that \({{\,\textrm{supp}\,}}\pi \subset A_0 \times A_0\). Let \(z \in A\). We have that
Hence, we have the uniform estimate of continuity in time
And we finally deduce that
In particular, by the Ascoli–Arzelá theorem, there is a subsequence converging uniformly in [0, T]. Since we have characterised the point-wise limit, every convergent subsequence does so to \(\nabla W * \mu \). Hence, the whole sequence converges uniformly over compacts, i.e.,
Using the triangular inequality, (6.6), and (6.7) the result is proven. \(\square \)
Proof of Theorem 6.1
When \(W^k\) is \(C^2\), we can construct a unique classical solution as the push-forward of regular characteristics. This solution is well-known to be the gradient flow solution. By construction, it coincides with the unique dual viscosity solution that exists by Theorem 2.10.
First, we showed in (6.5) that \(\mu ^k\) converges to the gradient flow solution in \(C([0,T]; \mathcal {P}_2 ({\mathbb {R}^d}))\), and let us denote it by \({\overline{\mu }}\). Now we apply Theorem 2.14 where the hypothesis have been check in (6.4) (using (6.3)), and Lemma 6.3 to show that \({\overline{\mu }}\) is a dual viscosity solution. This completes the proof. \(\square \)
7 A \(\dot{H}^{-1}\) and 2-Wasserstein theory when \(D> 0\)
7.1 Notion of solution and well-posedness theorem
Many of cases of (1.1) studied in the literature are 2-Wasserstein gradient flow. Our situation is more general. Unfortunately, the 2-Wasserstein distance, \(d_2\), does not have a duality characterisation similar to \(d_1\). It is known (see, e.g., [46]) that it can be one-side compared with the \(\dot{H}^{-1}({\mathbb {R}^d})\)
where, for \(\mu \in D' ({\mathbb {R}^d})\) we define the norm
The converse inequality to (7.1) only holds for absolutely continuous measures, and the constant depends strongly on the uniform continuity.
We consider the Sobolev semi-norm \( [f]_{H^1} = \Vert \nabla f\Vert _{L^2}. \) The space \((C_c^\infty ({\mathbb {R}^d}) , [ \cdot ]_{H^1})\) is a normed space. Notice that, if \([f]_{H^1} = 0\) then f is constant. But since it has compact supported, the value of the constant is 0. This allows to define the dual space
Since it is the dual of a normed space, \(\dot{H}^{-1} ({\mathbb {R}^d})\) is a Banach space.
Remark 7.1
The space \(\dot{H}^1({\mathbb {R}^d})\) is defined as the completion of \((C_c({\mathbb {R}^d}), [ \cdot ]_{H^1({\mathbb {R}^d})})\), which is easy to see is not complete itself. This completion can be complicated (see, e.g., [12]). Hence, with our construction \(\dot{H}^{-1} ({\mathbb {R}^d})\) is not the dual of \(\dot{H}^{1} ({\mathbb {R}^d})\).
Similarly to above, we define
Definition 7.2
We say that \((\varvec{\mu }, \{\varvec{\Psi }^T\}_{T\ge 0})\) is an \(\dot{H}^{-1}\) entropy pair if:
-
1.
For every \(T \ge 0\),
$$\begin{aligned} \varvec{\Psi }^T: X = \{ \psi _0 \in C_c({\mathbb {R}^d}): \nabla \psi _0 \in L^2 ({\mathbb {R}^d}) \} \longrightarrow C([0,T]; X )^n \end{aligned}$$is a linear map with the following property: for every \(\psi _0\) and \(i = 1, \cdots , n\) we have \(\Psi ^{T,i}[\psi _0]\) is a viscosity solution of (\(\hbox {P}^*_i\)).
-
2.
For each i and \(T \ge 0\), \(\mu _T^i \in \dot{H}^{-1} ({\mathbb {R}^d})\) and satisfies the duality condition (\(\hbox {D}_i\))
Remark 7.3
Notice that in this section we require that \(\psi _0\) is compactly supported, and thus the unique viscosity solutions will satisfy \(\psi (x) \rightarrow 0\) as \(|x| \rightarrow \infty \).
The main result of this section is
Theorem 7.4
Let \(D> 0\), \(\mu _0 \in \dot{H}^{-1} ({\mathbb {R}^d})\) and assume that
is Lipschitz with data in \( \dot{H}^{-1} ({\mathbb {R}^d})\) in the sense that, for any \(\varvec{\mu }\) and \(\widehat{\varvec{\mu }}\) in \(C([0,T]; \dot{H}^{-1} ({\mathbb {R}^d}))\)
Then there exists exactly one dual viscosity solution \(\varvec{\mu } \in C([0,T], \dot{H}^{-1} ({\mathbb {R}^d})^n)\).
Furthermore, if \(\varvec{\mathfrak {K}}^i\) is autonomous (i.e., \(\varvec{\mathfrak {K}}^i[\varvec{\mu }]_t = \varvec{K}^i[\varvec{\mu }_t]\)), then the map \(S_T : \varvec{\mu }_0 \in \dot{H}^{-1} ({\mathbb {R}^d})^n \mapsto \varvec{\mu }_T \in \dot{H}^{-1} ({\mathbb {R}^d}) ^n\) is a continuous semigroup.
We point that we can only get the \(\dot{H}^{-1}({\mathbb {R}^d})\) theory when diffusion is present (i.e., \(D> 0\)). When \(D = 0\) we cannot get an estimate of \(\Vert \nabla \psi _s\Vert _{L^2}\).
Proof of Theorem 7.4
The proof follows the blueprint of the proof of Theorem 2.10 using a fixed-point argument. We need an adapted version of (2.8) given by
This is shown by a suitable modification of the argument in (2.7). Lastly, we use the bounds and continuous dependence proved below in Propositions 7.6 and 7.9. \(\square \)
Lastly, due to (7.1) we have the following partial result in 2-Wasserstein space.
Corollary 7.5
Let \(D> 0\), \(\varvec{\mu }_0 \in \dot{H}^{-1} ({\mathbb {R}^d})^n \cap \mathcal {P}({\mathbb {R}^d})^n \), then the unique solution constructed in Theorem 7.4 is \(C([0,T]; \mathcal {P}_2 ({\mathbb {R}^d})^n)\).
Proof
Due to the previous theorem, we only need to show that probability distributions stay probability distributions. The non-negativity is trivial, since by Proposition 3.1 the test functions preserve the non-negativity. Hence, we get
Thus \(\mu _T \ge 0\) for all \(T \ge 0\).
We now prove the conservation of total mass. Consider a sequence of \(0 \le \psi _{0}^{(k)} \in X\) point-wise increasing k and converging uniformly over compacts to 1. Then, by the comparison principle, \(0 \le \psi ^{(k)} \le 1\) are point-wise increasing in k. By Dini’s theorem, functions \(\psi ^{(k)}\) converges uniformly over compacts. Let \(\psi ^{(\infty )}\) be its limit. By stability of viscosity solutions \(\psi ^{(\infty )}\) solves the same equation. By the uniform convergence \(\psi ^{(\infty )}_0 = \lim _n \psi _0^{(k)} = 1\). Then \(\psi ^{(\infty )} = 1\). By definition
We can apply the monotone convergence theorem on both sides to deduce
\(\square \)
7.2 Study of (\(\hbox {P}_{\varvec{E}}^*\))
To deal with the \(\dot{H}^{-1}({\mathbb {R}^d})\) estimates, we must get \(L^2({\mathbb {R}^d})\) of \(\nabla \psi \).
Proposition 7.6
Under the hypothesis of Proposition 3.1 together with \(\psi _0 \in C_c ({\mathbb {R}^d})\), then the unique viscosity solution of (2.3) constructed in Proposition 3.1 also satisfies
Lastly, given \({\widehat{\psi }}_0 = \psi _0\) and \(\varvec{\widehat{E}}\) in the same hypotheses above, there exists a corresponding solution of (2.3), denoted by \({\widehat{\psi }}\), and we have the continuous dependence estimate
Remark 7.7
Notice that for \(\nabla \psi _s \in L^2({\mathbb {R}^d})\) we do not use that \(\varvec{E} \in L^\infty ({\mathbb {R}^d})\), or the ellipticity constant.
Proof
We begin with the \(L^2({\mathbb {R}^d})\) estimates for smooth \(\psi _0 \in C_c^\infty ({\mathbb {R}^d})\) and \(\varvec{E} \in C_c^\infty ([0,T] \times {\mathbb {R}^d})\), where the solutions are classical and differentiable. For our duality characterisation in \(\dot{H}^{-1}({\mathbb {R}^d})\) we do not need \(L^2({\mathbb {R}^d})\) estimates on \(\psi \), only on \(\nabla \psi \). Notice that (7.5) is simply (3.10) when \(p = 2\).
For the continuous dependence, we note that
Multiplying by \(U^i_s - {\widehat{U}}^i_s\) and integrating, we get
We estimate term by term the integrals \(I_i\), \(i=1,\dots ,5\), as
and
There is only one problematic term that requires the use of the ellipticity condition
Notice that \(I_5\) cancels out the last term in \(|I_3|\). Arguing as above we recover
Eventually, since \(\psi _0^T - {\widehat{\psi }}_0^T = 0\) we recover (7.6) where the constants depend on d, \(\Vert \nabla \psi _0 \Vert _{L^2}\), \(\Vert \nabla {\widehat{\psi }}_0 \Vert _{L^2}\) and \(\int _0^T\Vert \varvec{E}_{T-\sigma }\Vert _{W^{1,\infty }} \text {d}\sigma \). This completes the \(L^2\) estimates.
When \(\psi _0\) is a general initial datum in \(C_c ({\mathbb {R}^d})\) we proceed by an approximation argument as in Proposition 3.1. As for the \(L^\infty ({\mathbb {R}^d})\) estimates, we can first assume that \(\varvec{E} \in C^\infty _c ([0,T] \times {\mathbb {R}^d})\) and \(\psi _0 \in C^\infty _c ({\mathbb {R}^d})\) are smooth, recover the \(L^2\) estimates for the gradient, and then pass to the limit. \(\square \)
Remark 7.8
Notice that in (7.7) we use strongly the fact that \(D > 0\), and the estimates are not uniform as \(D \searrow 0\).
7.3 Study of (\(\hbox {P}_{\varvec{E}}\))
Similarly, applying (7.4), (7.5) and (7.6), we deduce that
Proposition 7.9
For every \(D, T > 0\), \(\varvec{E} \in C([0,T]; W^{1,\infty } ({\mathbb {R}^d}; {\mathbb {R}^d}))\) and \(\mu _0 \in \dot{H}^{-1} ({\mathbb {R}^d})\) there exists exactly one \(\dot{H}^{-1} ({\mathbb {R}^d})\) dual viscosity solution \(\mu \in C([0,T]; \dot{H}^{-1} ({\mathbb {R}^d}))\) of (\(\hbox {P}_{\varvec{E}}\)). Furthermore, the map
is continuous with the following estimate
where C depends monotonically on each entry. Lastly, the semigroup property holds, i.e.,
Remark 7.10
(Solutions by characteristics when \(D= 0\)). For \(D= 0\) our notion of solution is \(\mu _t = X_t \# \mu _0\) where \(X_t\) is the unique solution of the flow equation
A unique solution of this pointwise-decoupled problem exists via the Picard-Lindelöf theorem. Since we have forwards and backwards uniqueness, for each \(t > 0\) we know that \(X_t\) is a bijection. It is easy to show that the unique viscosity solution is given by
And the duality condition (\(\hbox {D}_i\)) is trivially satisfied. The problem is that we do not have suitable \(H^{1}\) in this theory.
Data availability
The manuscript has no associated data.
References
Alibaud, N., Endal, J., Jakobsen, E.R.: Optimal and Dual Stability Results for \(L^1\) Viscosity and \(L^\infty \) Entropy Solutions. (2018). arXiv:1812.02058v1
Ambrosio, L., Gigli, N., Savare, G.: Gradient Flows. In: Lectures in Mathematics ETH Zürich. pp. 1–27. Birkhäuser-Verlag, Basel. (2005). https://doi.org/10.1007/b137080
Ambrosio, L., Savaré, G.: Gradient flows of probability measures. In: Handbook of differential equations: evolutionary equations. Vol. III. Handb. Differ. Equ. Elsevier/North-Holland, Amsterdam, pp. 1–136. (2007) https://doi.org/10.1016/S1874-5717(07)80004-1
Armstrong, N.J., Painter, K.J., Sherratt, J.A.: A continuum approach to modelling cell–cell adhesion. J. Theoret. Biol. 243(1), 98–113 (2006). https://doi.org/10.1016/j.jtbi.2006.05.030
Aronson, D.: Non-negative solutions of linear parabolic equations. Ann. Della Sci. Norm. Super. di Pisa 22.4, 607–694 (1968)
Balagué, D., Carrillo, J.A., Laurent, T., Raoul, G.: Dimensionality of local minimizers of the interaction energy. Arch. Ration. Mech. Anal. 209(3), 1055–1088 (2013). https://doi.org/10.1007/s00205-013-0644-6
Barré, J., Degond, P., Zatorska, E.: Kinetic theory of particle interactions mediated by dynamical networks. Multisc. Model. Simul. 15(3), 1294–1323 (2017). https://doi.org/10.1137/16M1085310
Bertozzi, A.L., Laurent, T., Léger, F.: Aggregation and spreading via the Newtonian potential: the dynamics of patch solutions. Math. Models Methods Appl. Sci. 22(supp. 1), 1140005 (2012). https://doi.org/10.1142/S0218202511400057
Boccardo, L.: On the regularizing effect of strongly increasing lower order terms. J. Evol. Equ. 3, 225–236 (2003). https://doi.org/10.1007/s00028-003-0089-8
Boccardo, L., Orsina, L., Porzio, M.M.: Regularity results and asymptotic behavior for a noncoercive parabolic problem. J. Evol. Equ. (2021). https://doi.org/10.1007/s00028-021-00678-2
Bonaschi, G.A., Carrillo, J.A., Di Francesco, M., Peletier, M.A.: Equivalence of gradient flows and entropy solutions for singular nonlocal interaction equations in 1D. ESAIM Control. Optim. Calc. Var. 21(2), 414–441 (2015). https://doi.org/10.1051/cocv/2014032. arXiv:1310.4110
Brasco, L., Gómez-Castro, D., Vázquez, J.L.: Characterisation of homogeneous fractional Sobolev spaces. Calc. Var. Partial Differ. Equ. 60(2), 60 (2021). https://doi.org/10.1007/s00526-021-01934-6. arXiv:2007.08000
Carrillo, J.A., DiFrancesco, M., Figalli, A., Laurent, T., Slepčev, D.: Global-in-time weak measure solutions and finite-time aggregation for nonlocal interaction equations. Duke Math. J. 156(2), 229–271 (2011). https://doi.org/10.1215/00127094-2010-211
Carrillo, J.A., Francesco, M.D., Esposito, A., Fagioli, S., Schmidtchen, M.: Measure Solutions to a System of Continuity Equations Driven by Newtonian Nonlocal Interactions. (2018)
Carrillo, J.A., Gvalani, R.S., Pavliotis, G.A., Schlichting, A.: Long-time behaviour and phase transitions for the McKean–Vlasov equation on the torus. Arch. Ration. Mech. Anal. 235(1), 635–690 (2020). https://doi.org/10.1007/s00205-019-01430-4
Carrillo, J.: Entropy solutions for nonlinear degenerate problems. Arch. Ration. Mech. Anal. 147(4), 269–361 (1999). https://doi.org/10.1007/s002050050152
Carrillo, J.A., Hu, J., Wang, L., Wu, J.: A particle method for the homogeneous Landau equation. J. Comput. Phys. X 7, 100066 (2020). https://doi.org/10.1016/j.jcpx.2020.100066
Carrillo, J.A., Murakawa, H., Sato, M., Togashi, H., Trush, O.: A population dynamics model of cell-cell adhesion incorporating population pressure and density saturation. J. Theoret. Biol. 474, 14–24 (2019). https://doi.org/10.1016/j.jtbi.2019.04.023
Carrillo, J.A., Craig, K., Yao, Y.: Aggregation-diffusion equations: dynamics, asymptotics, and singular limits. In: Active particles, Vol. 2. Model. Simul. Sci. Eng. Technol. pp. 65–108. Birkhäuser/Springer, Cham. (2019)
Carrillo, J.A., Gómez-Castro, D., Yao, Y., Zeng, C.: Asymptotic simplification of Aggregation–Diffusion equations towards the heat kernel. (2021). arXiv: 2105.13323
Carrillo, J.A., Huang, Y., Schmidtchen, M.: Zoology of a nonlocal cross-diffusion model for two species. SIAM J. Appl. Math. 78(2), 1078–1104 (2018). https://doi.org/10.1137/17M1128782
Carrillo, J.A., McCann, R.J., Villani, C.: Kinetic equilibration rates for granular media and related equations: Entropy dissipation and mass transportation estimates. Rev. Mat. Iberoam. 19(3), 971–1018 (2003). https://doi.org/10.4171/RMI/376
Carrillo, J.A., McCann, R.J., Villani, C.: Contractions in the 2-Wasserstein length space and thermalization of granular media. Arch. Ration. Mech. Anal. 179(2), 217–263 (2006). https://doi.org/10.1007/s00205-005-0386-1
Carrillo, J.A., Craig, K., Patacchini, F.S.: A blob method for diffusion. Calc. Var. Partial Differ. Equ. 58(2), 53 (2019). https://doi.org/10.1007/s00526-019-1486-3
Craig, K., Bertozzi, A.L.: A blob method for the aggregation equation. Math. Comp. 85(300), 1681–1717 (2016). https://doi.org/10.1090/mcom3033
Craig, K., Kim, I., Yao, Y.: Congested aggregation via Newtonian interaction. Arch. Ration. Mech. Anal. 227(1), 1–67 (2018). https://doi.org/10.1007/s00205-017-1156-6
Craig, K., Topaloglu, I.: Convergence of regularized nonlocal interaction energies. SIAM J. Math. Anal. 48(1), 34–60 (2016). https://doi.org/10.1137/15M1013882
Crandall, M.G., Ishii, H., Lions, P.-L.: User’s guide to viscosity solutions of second order partial differential equation. Bull. Am. Math. Soc. 27(1), 1–67 (1992)
Di Francesco, M., Esposito, A., Schmidtchen, M.: Many-particle limit for a system of interaction equations driven by Newtonian potentials. Calc. Var. Partial Differ. Equ. 60.2, 68 (2021). https://doi.org/10.1007/s00526-021-01960-4
Di Francesco, M., Fagioli, S., Radici, E.: Deterministic particle approximation for nonlocal transport equations with nonlinear mobility. J. Differ. Equ. 266(5), 2830–2868 (2019). https://doi.org/10.1016/j.jde.2018.08.047. arXiv:1801.08770
Di Francesco, M., Esposito, A., Fagioli, S.: Nonlinear degenerate cross-diffusion systems with nonlocal interaction. Nonlinear Anal. Theory Methods Appl. 169, 94–117 (2018). https://doi.org/10.1016/j.na.2017.12.003
Di Francesco, M., Fagioli, S.: Measure solutions for non-local interaction PDEs with two species. Nonlinearity 26(10), 2777–2808 (2013). https://doi.org/10.1088/0951-7715/26/10/2777
Esposito, A., Patacchini, F.S., Schlichting, A., Slepcev, D.: Nonlocal-interaction equation on graphs: gradient flow structure and continuum limit. Arch. Ration. Mech. Anal. 240(2), 699–760 (2021). https://doi.org/10.1007/s00205-021-01631-w
Evans, L.C.: Partial Differential Equations. American Mathematical Society (1998)
Fagioli, S., Tse, O.: On gradient flow and entropy solutions for nonlocal transport equations with nonlinear mobility. Nonlinear Anal. 221, 112904 (2022). https://doi.org/10.1016/j.na.2022.112904. arXiv:2105.11389
Fernández-Real, X., Figalli, A.: The Continuous Formulation of Shallow Neural Networks as Wasserstein-Type Gradient Flows. (2020). URL: https://sma.epfl.ch/~xafernan/Documents/Fernandez-Real_Figalli_Flows_Survey.pdf
Jabin, P.E.: A review of the mean field limits for vlasov equations. Kinet. Relat. Model. 7(4), 661–711 (2014). https://doi.org/10.3934/krm.2014.7.661
James, F., Vauchelet, N.: Equivalence between duality and gradient flow solutions for one-dimensional aggregation equations. Discrete Contin. Dyn. Syst. 36(3), 1355–1382 (2016). https://doi.org/10.3934/dcds.2016.36.1355
Karlsen, K.H., Risebro, N.H.: On the uniqueness and stability of entropy solutions of nonlinear degenerate parabolic equations with rough coefficients. Discret. Contin. Dyn. Syst. A 9(5), 1081–1104 (2003)
Kružkov, S.N.: First order quasilinear equations in several independent variables. Math. USSR-Sbornik 10(2), 217–243 (1970)
Ladyzhenskaia, O.A., Solonnikov, V.A., Ural’tseva, N.N.: Linear and quasi-linear equations of parabolic type. In: vol. 23. American Mathematical Soc. (1968)
Messenger, D., Fetecau, R.C.: Equilibria of an aggregation model with linear diffusion in domains with boundaries. Math. Models Methods Appl. Sci. 30(4), 805–845 (2020). https://doi.org/10.1142/S0218202520400059
Mogilner, A., Edelstein-Keshet, L.: A non-local model for a swarm. J. Math. Biol. 38(6), 534–570 (1999). https://doi.org/10.1007/s002850050158
Murakawa, H., Togashi, H.: Continuous models for cell-cell adhesion. J. Theor. Biol. 374, 1–12 (2015). https://doi.org/10.1016/j.jtbi.2015.03.002
Nunziante, D.: Existence and uniqueness of unbounded viscosity solutions of parabolic equations with discontinuous time-dependence. Nonlinear Anal. 18(11), 1033–1062 (1992). https://doi.org/10.1016/0362-546X(92)90194-J
Peyre, R.: Comparison between W2 distance and H-1 norm, and Localization of Wasserstein distance. ESAIM Control Optim. Cal. Var. 24, 1489–1501 (2018). https://doi.org/10.1051/cocv/2017050
Shigesada, N., Kawasaki, K., Teramoto, E.: Spatial segregation of interacting species. J. Theor. Biol. 79.1, 83–99 (1979)
Topaz, C.M., Bertozzi, A.L., Lewis, M.A.: A nonlocal continuum model for biological aggregation. Bull. Math. Biol. 68(7), 1601–1623 (2006). https://doi.org/10.1007/s11538-006-9088-6
Tugaut, J.: Convergence to the equilibria for self-stabilizing processes in double-well landscape. Ann. Probab. 41(3A), 1427–1460 (2013). https://doi.org/10.1214/12-AOP749
Villani, C.: Topics in Optimal Transportation. In: American Mathematical Society (2003)
Zinsl, J., Matthes, D.: Exponential convergence to equilibrium in a coupled gradient flow system modeling chemotaxis. Anal. PDE 8(2), 425–466 (2015). https://doi.org/10.2140/apde.2015.8.425
Acknowledgements
The research of JAC and DGC was supported by the Advanced Grant Nonlocal-CPD (Nonlocal PDEs for Complex Particle Dynamics: Phase Transitions, Patterns and Synchronization) of the European Research Council Executive Agency (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 883363). JAC was partially supported by EPSRC Grant Nos. 878 EP/T022132/1 and EP/V051121/1 and the “Maria de Maeztu” Excellence Unit IMAG, reference CEX2020-001105-M, funded by MCIN/AEI/10.13039/501100011033/. DGC was partially supported by Grant Nos. RYC2022-037317-I, PID2021-127105NB-I00, PGC2018-098440-B-I00 from the Ministerio de Ciencia, Innovación y Universidades of the Spanish Government.
Funding
Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Carrillo, J.A., Gómez-Castro, D. Interpreting systems of continuity equations in spaces of probability measures through PDE duality. Rev. Real Acad. Cienc. Exactas Fis. Nat. Ser. A-Mat. 118, 127 (2024). https://doi.org/10.1007/s13398-024-01628-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s13398-024-01628-6