Abstract
In this paper we generalise the results proved in N. Katzourakis (SIAM J. Math. Anal. 51, 1349–1370, 2019) by studying the ill-posed problem of identifying the source of a fully nonlinear elliptic equation. We assume Dirichlet data and some partial noisy information for the solution on a compact set through a fully nonlinear observation operator. We deal with the highly nonlinear nonconvex nature of the problem and the lack of weak continuity by introducing a two-parameter Tykhonov regularisation with a higher order L2 “viscosity term” for the \(L^{\infty }\) minimisation problem which allows to approximate by weakly lower semicontinuous cost functionals.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(n,k\in \mathbb {N}\) with k, n ≥ 2 and let \({\varOmega } \subseteq \mathbb {R}^{n}\) be a bounded connected domain with C1,1 regular boundary ∂Ω. Let also
be a Carathéodory function, namely x↦F(x,r,p,X) is Lebesgue measurable for all \((\mathrm {r}, \mathrm {p}, \mathrm {X}) \in \mathbb {R} \times \mathbb {R}^{n} \times \mathbb {R}^{n^{\otimes 2}}_{s}\) and (r,p,X)↦F(x,r,p,X) is continuous for a.e. x ∈Ω. In this paper, the notation \(\mathbb {R}^{n^{\otimes k}}_{s}\) stands for the vector space of fully symmetric k-th order tensors in \(\mathbb {R}^{n} \otimes {\cdots } \otimes \mathbb {R}^{n}\) (k-times). Given \(g\in W^{2,\infty }({\varOmega })\), consider the Dirichlet problem
for some appropriate source \(f :{\varOmega }\longrightarrow \mathbb {R}\). Here F[u] denotes the induced fully nonlinear 2nd order differential operator, defined on smooth functions u as
Evidently, we are employing the standard symbolisations Du = (Diu)i= 1...n, D2u = \((\mathrm {D}^{2}_{ij} u)_{i,j=1...n}\) and Di ≡ ∂/∂xi. The above direct Dirichlet problem for F asks to determine u, given a source f and boundary data g. (In fact the source f is obsolete and can be absorbed into F, but for the problem we are interested in this paper it is more convenient to write it in this separated form). This is a semi-classical problem which is essentially standard material, see e.g. [22]. In particular, it is known that under various sets of assumptions on F that (1.1) is well-posed and, given \(f\in L^{\infty }({\varOmega })\) and \(g\in W^{2,\infty }({\varOmega })\), for any p > n there exists a unique solution u in the locally convex (Fréchet) space
In general, the solution u is not in the smaller space \((W^{2, \infty }\cap W^{1,\infty }_{g})({\varOmega })\) (not even locally), due to the failure of the W2, p estimates for \(p=\infty \), which happens even in the linear case (see e.g. [21]). Additionally, (1.1) satisfies for any p > n the fully nonlinear Lp global estimate
for some constants C1, C2 > 0 depending only on the parameters and any \(v\in (W^{2,p}\cap W^{1,p}_{g})({\varOmega })\). For sufficient conditions on F which guarantee the satisfaction of solvability of (1.1) in the strong sense and of the uniform estimate (1.3) we refer to [12, 13, 16, 28, 30].
Note that the above problem contains as a special case the archetypal instance of divergence operators with C1 matrix coefficient A, as well as the non-divergence linear case with continuous coefficient:
In the above, the notations “:” and “⋅” symbolise the Euclidean inner products in the space of symmetric matrices \(\mathbb {R}^{n^{\otimes 2}}_{s}\) and in \(\mathbb {R}^{n}\) respectively. More generally, the inner product of two tensors \(T,S \in \mathbb {R}^{n^{\otimes k}}_{s}\) will also be denoted by “:”, that is
The inverse problem relating to (1.1) asks the question of perhaps determining f, given the boundary data g and some other partial information on the solution u, typically some approximate experimental measurements of some function of it known only up to some error. The inverse problem is severely ill-posed even in the linear case of the Laplacian operator F = Δ, as the noisy data measured on a subset of Ω might either not be compatible with any exact solution, or they may not suffice to determine a unique f even if compatibility holds true.
The above type of inverse problems are especially crucial for various applications, even in the model case of the Poisson equation, see e.g. [1, 8, 17, 23, 31, 32, 34, 36,37,38,39,40]. In this paper we will assume that the approximate information on u takes the form
where K is an observation operator, taken to be a first order fully nonlinear differential operator of the form
where K and its partial derivates Kr, Kp satisfy
In (1.5) and (1.7), \(\mathcal {K}\) symbolises the set on which we take measurements, which will be assumed to satisfy
In the above, \({\mathscr{H}}^{\kappa }\) denotes the Hausdorff measure of dimension κ. Our general measure and functional notation will be either standard or self-explanatory, e.g. as in [15, 18, 27]. Finally, \(k^{\gamma } \in L^{\infty }(\mathcal {K},{\mathscr{H}}^{\kappa })\) is the function of approximate (deterministic) measurements taken on \(\mathcal {K}\), at noise level at most γ > 0:
where k0 = K[u0] corresponds to ideal error-free measurements of an exact solution to (1.1) with source f = F[u0].
To recapitulate, in this paper we study the following ill-posed inverse source identification problem for fully nonlinear elliptic PDEs:
This means that we are searching for a selection process of a suitable approximation for f from the data kγ on \(\mathcal {K}\) through the observation K[u] of the solution u. To the best of our knowledge, (1.10) has not been studied before, at least in this generality. Our approach does not exclude the extreme cases of \(\mathcal {K}=\overline {{\varOmega }}\) (full information) or of \(\mathcal {K}=\emptyset \) (no information), although trivial changes are required in the proofs. Sadly, an exact solution may not exist as the constraint may be incompatible with the solution of (1.1), owing to the errors in measurements. On the other hand, it is not possible to have a uniquely determined source on the constraint-free region \({\varOmega }\setminus \mathcal {K}\). Instead, our goal is a strategy to determine an optimally fitting uγ (and respective source fγ := F[uγ]) to the ill-posed problem (1.10). A popular choice of operator K in the literature (when L = Δ) consist of some term of the separation of variables formula, as e.g. in [39].
Herein, we follow an approach based on recent advances in Calculus of Variations in the space \(L^{\infty }\) (see e.g. [26] and references therein) developed for functionals involving higher order derivatives, which has already been applied to the special case of the inverse source problem for linear PDEs (1.4) in [25]. This relatively new field was pioneered by Gunnar Aronsson in the 1960s (see e.g. [3,4,5,6,7]) and is still a very active area of research; for a review of the by-now classical theory involving scalar first order functionals we refer to [24].
Following [25], we aim at providing a regularisation strategy inspired by the classical Tykhonov regularisation strategy in L2 (see e.g. [29, 33]). As a first possible step, consider the next putative \(L^{\infty }\) “error” functional:
for some fixed parameter α > 0. The advantage of searching for a best fitting solution in \(L^{\infty }\) is evident: we can keep the error term |K[u] − kγ| uniformly small and not merely small on average, as would happen if one chose to minimise some integral of a power of the error instead. As in [25], the goal would be to minimise \(\mathrm {E}_{\infty ,\alpha }\) over \(\mathcal {W}^{2,\infty }_{g} ({\varOmega })\), and then any minimiser of (1.11) would provide a candidate solution for our problem. Then, for any fixed α, this would be the best fitting solution with the least possible uniform error, namely F[u]≅f uniformly on Ω and K[u]≅kγ uniformly on \(\mathcal {K}\). Unfortunately, even if one momentarily ignores the problem of lack of regularity for (1.11) and the fact that \(\mathcal {W}^{2,\infty }_{g} ({\varOmega })\) is not a Banach space, the main problem is that in general minimisers do not exist in the genuine fully nonlinear case of operator F (namely when X↦F(x,r,p,X) is nonlinear) as (1.11) is not weakly lower semicontinuous in the Fréchet space \(\mathcal {W}^{2,\infty }_{g} ({\varOmega })\), as the highest order term may be nonconvex/non-quasiconvex. In the special linear case of [25], this problem was not present as the linearity of the differential operator was implying the desired weak lower semi-continuity.
In this work we resolve the problem explained above by proposing a double approximation method (or rather triple, as we will see shortly) which involves an additional Tykhonov or “viscosity” term which effectively is a weakly lower-semicontinuous approximation of (1.11). Hence, we will consider instead
where \(\bar n:= [n/2]+3\). In the above β > 0 is a fixed parameter, [⋅] symbolises the integer part and \(\mathrm {D}^{\bar n} u\) is the \(\bar n\)-th order weak derivative of u.
It is well known in the Calculus of Variations in \(L^{\infty }\) that (global) minimisers of supremal functionals, although usually simple to obtain with a standard direct minimisation [15, 19], they are not genuinely minimal as they do not share the nice “local” optimisation properties of minimisers of their integral counterparts (see e.g. [10, 35]). The case of (1.12) studied herein is no exception to this rule. A relatively standard method is bypass these obstructions is to employ minimisers of Lp approximating functionals as \(p\to \infty \), establishing appropriate convergence of Lp minimisers to a limit \(L^{\infty }\) minimiser. The idea underlying this approximation technique is based on the simple measure theory fact that the Lp norm (of a function in \(L^{1}\cap L^{\infty }\)) converges to the \(L^{\infty }\) norm as \(p\to \infty \). This method is quite standard in the field and furnishes a selection principle of \(L^{\infty }\) minimisers with additional desirable properties (see e.g. [9, 11, 14, 20, 26]). In this fashion one is also able to bypass the lack of differentiability of supremal functionals and derive necessary PDE conditions satisfied by \(L^{\infty }\) extrema. This is indeed the method that is employed in this work as well, along the lines of [25].
We now present the main results to be established in this paper. As already explained, we will obtain special minimisers of (1.12) as limits of minimisers of
In (1.13) we have used the normalised Lp norms
and the integral signs with slashes symbolise the average with respect to the Hausdorff measure \({\mathscr{H}}^{\kappa }\) and the Lebesgue measure \({\mathscr{L}}^{n}\), respectively. Further, |⋅|(p) symbolises the next p-regularisation of the absolute value away from zero:
Let us also note that, due to our Lp-approximation method, as an auxiliary result we also provide an Lp regularisation strategy for finite p as well, which is of independent interest. For the proof we will need to assume that the Dirichlet problem (1.1) for the fully nonlinear operator F satisfies the W2, p elliptic estimates (1.3) for all large enough (finite) p > n, as well the following: F and its partial derivates Fr, Fp, FX satisfy
We note that sufficient general conditions of when such operators satisfy W2, p elliptic estimates can be found for instance in the papers [12, 13, 16, 28, 30].
Theorem 1
Let \({\varOmega } \subseteq \mathbb {R}^{n}\) be a bounded C1,1 domain and \(g \in W^{\bar n, 2}({\varOmega })\). Suppose also the operators (1.2) and (1.6) are given, satisfying the assumptions (1.3), (1.7), (1.8) and (1.14). Suppose further a function \(k^{\gamma } \in L^{\infty }(\mathcal {K},{\mathscr{H}}^{\kappa })\) is given which satisfies (1.9) for γ > 0. Let finally α, β > 0 be fixed. Then, we have the following results for the inverse problem associated to (1.10):
-
(i)
There exists a global minimiser \(u_{\infty }\equiv u_{\infty }^{\alpha ,\beta ,\gamma } \in (W^{\bar n,2} \cap W^{1,2}_{g})({\varOmega })\) of the functional \(\mathrm {E}_{\infty ,\alpha ,\beta }\) defined in (1.12). In particular, we have \(E_{\infty }(u_{\infty })\leq E_{\infty } (v)\) for all \(v\in (W^{\bar n,2} \cap W^{1,2}_{g})({\varOmega })\) and
$$ f_{\infty} \equiv f^{\alpha,\beta,\gamma}_{\infty} := \mathrm{F}[u_{\infty}^{\alpha,\beta,\gamma}] \in L^{\infty}({\varOmega}). $$Further, there exist signed Radon measures
$$ \mu_{\infty} \equiv \mu_{\infty}^{\alpha,\beta,\gamma} \in \mathcal{M}({\varOmega}), \quad \nu_{\infty} \equiv \nu_{\infty}^{\alpha,\beta,\gamma} \in \mathcal{M}(\mathcal{K}) $$such that the nonlinear divergence PDE
$$ K_{\mathrm r} [u_{\infty}]\nu_{\infty} - \text{div}\left( \mathrm{K}_{\mathrm p} [u_{\infty}]\nu_{\infty}\right) + \alpha ((\mathrm{d} \mathrm{F})_{u_{\infty}})^{\ast}[\mu_{\infty}]+\beta(-1)^{\bar n} (\mathrm{D}^{\bar n} : \mathrm{D}^{\bar n} u_{\infty}) = 0, $$(1.15)is satisfied by the triplet \((u_{\infty },\mu _{\infty },\nu _{\infty })\) in the sense of distributions (see (2.2)). In (1.15), the operator \(((\mathrm {d} \mathrm {F})_{u_{\infty }})^{\ast }\) is the formal adjoint of the linearisation of F at \(u_{\infty }\), defined via duality as
$$ ((\mathrm{d} \mathrm{F})_{u_{\infty}})^{\ast} [v] := \text{div}(\text{div} (\mathrm{F}_{\mathrm X} [u_{\infty}] v)) - \text{div}(\mathrm{F}_{\mathrm p} [u_{\infty}] v) + \mathrm{F}_{\mathrm r} [u_{\infty}]v, $$Fr, Kr, Fp, Kp, FX denote the partial derivatives of F, K with respect to the respective variables and Fr[v], Kr[v], Fp[v], Kp[v], FX[v] denote the respective differential operators \(F_{\mathrm r} (\cdot ,v,\mathrm {D}v,\mathrm {D}^{2} v), K_{\mathrm r} (\cdot ,v,\mathrm {D}v),\dots \) etc. Additionally, the error measure \(\nu _{\infty }\) is supported in the closure of the subset of \(\mathcal {K}\) of maximum error, namely
$$ \text{supp} (\nu_{\infty}) \subseteq \left\{|\mathrm{K}[u_{\infty}]-k^{\gamma}|^{\bigstar} = \|\mathrm{K}[u_{\infty}]-k^{\gamma}\|_{L^{\infty}(\mathcal{K},\mathcal{H}^{\kappa})}\right\}, $$(1.16)where \(``(\cdot )^{\bigstar }"\) denotes the “essential limsup” with respect to \({\mathscr{H}}^{\kappa } {\LARGE {\llcorner }}_{\mathcal {K}}\) on \(\mathcal {K}\) (see Definition 3 that follows). If additionally the data function kγ is continuous on \(\mathcal {K}\), (1.16) can be improved to
$$ \text{supp} (\nu_{\infty}) \subseteq \left\{|\mathrm{K}[u_{\infty}]-k^{\gamma}| = \|\mathrm{K}[u_{\infty}]-k^{\gamma}\|_{L^{\infty}(\mathcal{K},\mathcal{H}^{\kappa})} \right\}. $$(1.17) -
(ii)
For any α, β, γ > 0, the minimiser \(u_{\infty }\) can be approximated by a family of minimisers \((u_{p})_{p>n}\equiv (u_{p}^{\alpha ,\beta ,\gamma })_{p>n}\) of the respective Lp functionals (1.13) and the pair of measures \((\mu _{\infty },\nu _{\infty }) \in {\mathscr{M}}({\varOmega })\times {\mathscr{M}}(\mathcal {K})\) can be approximated by respective signed measures \((\mu _{p},\nu _{p})_{p>n}\equiv (\mu ^{\alpha ,\beta ,\gamma }_{p},\nu ^{\alpha ,\beta ,\gamma }_{p})_{p>n}\), as follows:
For any p > n, (1.13) has a global minimiser \(u_{p}\equiv u_{p}^{\alpha ,\beta ,\gamma }\) in \((W^{\bar n,2} \cap W^{1,2}_{g})({\varOmega })\) and there exists a sequence \((p_{j})_{1}^{\infty }\), such that
$$ \left\{ \begin{array}{ll} u_{p} \longrightarrow u_{\infty}&\quad \text{ in }C^{2}(\overline{{\varOmega}}),\\ \mathrm{D}^{k} u_{p} \longrightarrow \mathrm{D}^{k} u_{\infty}&\quad \text{ in }L^{2}\big({\varOmega},\mathbb{R}^{n^{\otimes k}}_{s}\big),~\text{ for all }k\in\{3,\dots,\bar n-1\},\\ \mathrm{D}^{\bar n} u_{p} \rightharpoonup \mathrm{D}^{\bar n} u_{\infty} &\quad \text{ in }L^{2}\big({\varOmega},\mathbb{R}^{n^{\otimes \bar n}}_{s}\big), \end{array} \right. $$(1.18)as \(p_{j}\to \infty \). Moreover, we have
$$ \left\{ \begin{array}{ll} \nu_{p} := \frac{\left|\mathrm{K}[u_{p}]-k^{\gamma}\right|^{p-2}_{(p)} \left( \mathrm{K}[u_{p}]-k^{\gamma} \right)}{\mathcal{H}^{\kappa}(\mathcal{K})\left\||\mathrm{K}[u_{p}]-k^{\gamma}|_{(p)} \right\|^{p-1}_{L^{p}(\mathcal{K},\mathcal{H}^{\kappa})}}\mathcal{H}^{\kappa} {\LARGE{\llcorner}}_{\mathcal{K}} \overset{\ast}{\rightharpoonup} \nu_{\infty} &\quad \text{ in }\mathcal{M}(\mathcal{K}),\\ \mu_{p} := \frac{|\mathrm{F}[u_{p}] |^{p-2}_{(p)} \mathrm{F}[u_{p}] }{ \mathcal{L}^{n}({\varOmega}) \left\||\mathrm{F}[u_{p}] |_{(p)} \right\|^{p-1}_{L^{p}({\varOmega})}}\mathcal{L}^{n} {\LARGE\llcorner}_{{\varOmega}} \overset{\ast}{\rightharpoonup} \mu_{\infty}&\quad \text{ in }\mathcal{M}({\varOmega}), \end{array} \right. $$(1.19)as \(p_{j}\to \infty \). Further, for each p > n, the triplet (up, μp, νp) solves the PDE
$$ K_{\mathrm r} [u_{p}]\nu_{p} - \text{div}(\mathrm{K}_{\mathrm p} [u_{p}]\nu_{p}) + \alpha((\mathrm{d} \mathrm{F})_{u_{p}})^{\ast}[\mu_{p}] + \beta (-1)^{\bar n} (\mathrm{D}^{\bar n} : \mathrm{D}^{\bar n} u_{p}) = 0, $$(1.20)in the sense of distributions (see (2.1)).
-
(iii)
For any exact solution \(u^{0} \in (W^{\bar n,2} \cap W^{1,2}_{g})({\varOmega })\) of (1.10) (with f = F[u0] and K[u0] = k0) corresponding to measurements with zero error, we have the estimate:
$$ \left\|\mathrm{K}[u^{\alpha,\beta,\gamma}_{\infty}] - \mathrm{K}[u^{0}] \right\|_{L^{\infty}(\mathcal{K},\mathcal{H}^{\kappa})} \leq 2\gamma + \alpha \|\mathrm{F}[u^{0}]\|_{L^{\infty}({\varOmega})} + \frac{\beta}{2} \|\mathrm{D}^{\bar n} u^{0}\|^{2}_{L^{2}({\varOmega})}, $$(1.21)for any α, β, γ > 0.
-
(iv)
For any exact solution \(u^{0} \in (W^{\bar n,2} \cap W^{1,2}_{g})({\varOmega })\) of (1.10) (with f = F[u0] and F[u0] = k0) corresponding to measurements with zero error and for p > n, we have the estimate:
$$ \left\|\mathrm{K}[u^{\alpha,\beta,\gamma}_{p}] - \mathrm{K}[u^{0}] \right\|_{L^{p}(\mathcal{K},\mathcal{H}^{\kappa})} \leq 2\gamma + \alpha \|\mathrm{F}[u^{0}]\|_{L^{p}({\varOmega})} + \frac{\beta}{2} \|\mathrm{D}^{\bar n} u^{0}\|^{2}_{L^{2}({\varOmega})}, $$(1.22)for any α, β, γ > 0.
We note that the estimate in part (iv) is useful if we have merely that F[u0] ∈ Lp(Ω) for \(p<\infty \) (namely, when \(\mathrm {F}[u^{0}] \not \in L^{\infty }({\varOmega })\)). We close this section by noting that the reader may find in [25] various comments and counter-examples regarding the optimality of Theorem 1 (therein stated for the case of a linear differential operator F).
2 Discussion, Auxiliary Results and Proofs
We begin with some clarifications on Theorem 1.
Remark 2
In index form, the definition of the formal adjoint \(((\mathrm {d} \mathrm {F})_{u_{\infty }})^{\ast }\) of the linearisation of F at \(u_{\infty }\) can be written as
and its distributional interpretation via duality reads
for all \(\phi \in C^{\bar n}_{c}({\varOmega })\). Hence, by taking into account the definitions of the measures μp, νp in (1.19), the distributional interpretation of (1.20) is
for all \(\phi \in C^{\bar n}_{c}({\varOmega })\). Similarly, the distributional interpretation of (1.15) is
for all \(\phi \in C^{\bar n}_{c}({\varOmega })\).
Further, we note that the physical interpretation of the measures \(\mu _{\infty }\) and \(\nu _{\infty }\) arising in the PDE is that they essentially “charge” the sets whereon \(|\mathrm {F}[u_{\infty }]|\) and \(|\mathrm {K}[u_{\infty }]-k^{\gamma }|\) are maximised over Ω and \(\mathcal K\), respectively.
We now state a definition and a result taken from [25] which are required for our proofs.
Definition 3
(The essential limsup, [25]) Let \(X \subseteq \mathbb {R}^{n}\) be a Borel set and let \(\nu \in {\mathscr{M}}(X)\) be a finite positive Radon measure on X. Given \(f\in L^{\infty }(X,\nu )\), we define \(f^{\bigstar } \in L^{\infty }(X,\nu )\) by
and call f⋆ the ν-essential limsup of f. Here, \(\mathbb {B}_{\varepsilon }(x)\) denotes the open ball of radius ε centred at x ∈ X.
The following result studies what we call “concentration measures” of the approximate Lp minimisation problems as \(p\to \infty \).
Proposition 4
(Lp concentration measures as \(p\to \infty \)) Let X be a compact metric space, endowed with a positive finite Borel measure ν which gives positive values to any open subset of X except ∅. Consider \((f_{p})_{1}^{\infty } \subseteq L^{\infty }(X,\nu )\) and the sequence of signed Radon measures \((\nu _{p})_{1}^{\infty } \subseteq {\mathscr{M}}(X)\), given by:
where |⋅|(p) = (|⋅|2 + p− 2)1/2. Then:
-
(i)
There exists a subsequence \((p_{i})_{1}^{\infty }\) and a limit measure \(\nu _{\infty } \in {\mathscr{M}}(X)\) such that
$$ \nu_{p} ~\overset{\ast}\rightharpoonup~ \nu_{\infty} \quad\text{ in }~\mathcal{M}(X), $$as \(p_{i}\to \infty \).
-
(ii)
If there exists \(f_{\infty } \in L^{\infty }(X,\nu )\setminus \{0\}\) such that
$$ \underset{X}{\sup} |f_{p} -f_{\infty}| \longrightarrow 0\quad\text{ as }p\to\infty, $$then \(\nu _{\infty }\) is supported in the set where \(|f_{\infty }|\) is maximised:
$$ \text{supp}(\nu_{\infty}) \subseteq \left\{|f_{\infty}|^{\bigstar}= \|f_{\infty}\|_{L^{\infty}(X,\nu)}\right\}. $$ -
(iii)
If additionally to (ii) the modulus \(|f_{\infty }|\) of \(f_{\infty }\) is continuous on X, then the following stronger assertion holds:
$$ \text{supp}(\nu_{\infty}) \subseteq \left\{|f_{\infty}|= \|f_{\infty}\|_{L^{\infty}(X,\nu)} \right\}. $$
Now we establish Theorem 1. The proof consists of several lemmas. We note that some details might be quite well known to the experts of Calculus of Variations, but we chose to give most of the niceties for the convenience of the readers and for the sake of completeness of the exposition.
Lemma 5
For any p > n and α, β, γ > 0, the functional (1.13) has a minimiser \(u_{p} \in (W^{\bar n, 2}\cap W^{1,2}_{g} )({\varOmega })\):
Proof
Since \(g\in W^{2,\infty }({\varOmega })\)
Hence,
Further, Ep, α, β is coercive in the space \((W^{\bar n, 2}\cap W^{1,2}_{g} )({\varOmega })\). Indeed, by our assumption (1.3), Hölder’s inequality and that p > n ≥ 2, for any \(v\in (W^{\bar n, 2}\cap W^{1,2}_{g} )({\varOmega })\) we have
which implies
for any \(v\in (W^{\bar n, 2}\cap W^{1,2}_{g} )({\varOmega })\). Now, by Poincaré inequality in \(W^{1,2}_{g}({\varOmega })\) we have
for some C > 0, and by the interpolation inequalities in the Sobolev space \(W^{\bar n,2}({\varOmega })\), we have
for some C > 0 and any \(k\in \{1,\dots ,\bar n\}\). By putting the last three estimates together, we conclude that
where the constant C > 0 in general depends on p but is uniform for \(v \in (W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\).
Let now \(({u_{p}^{m}})_{1}^{\infty }\) be a minimising sequence of Ep, α, β:
as \(m\to \infty \). Then, by the coercivity estimate, we have the uniform bound
for some C > 0 independent of \(m\in \mathbb {N}\). By standard weak and strong compactness arguments in Sobolev and Hölder spaces, together with the Morrey estimate
applied to \(k=\bar n\), there exists a subsequence \((u_{p}^{m_{k}})_{1}^{\infty }\) and \(u_{p}\in (W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\) such that, along this subsequence we have
as \(m_{k}\to \infty \). The above modes of convergence and the continuity of the function K defining the operator K imply that \(\mathrm {K}[{u^{m}_{p}}] \longrightarrow \mathrm {K}[u_{p}]\) uniformly on \(\mathcal {K}\) as \(m_{k} \to \infty \). Therefore,
as \(m_{k} \to \infty \). Additionally, by the continuity of the function F defining the operator F and the uniform convergence of the minimising sequence up to second order derivatives, we have
as \(m_{k} \to \infty \). Finally, by weak lower semi-continuity in L2 we have
By putting all the above together, we infer that
which concludes the proof. □
Note that the proof above reveals the fact that Ep, α, β is weakly lower semi-continuous on the space \((W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\), even though it is not explicitly stated.
Lemma 6
For any α, β, γ > 0, there exists a (global) minimiser \(u_{\infty }\) of \(\mathrm {E}_{\infty ,\alpha ,\beta }\) in the space \((W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\), as well as a sequence of minimisers \((u_{p_{i}})_{1}^{\infty }\) of the respective Ep, α, β-functionals from Lemma 5, such that (1.18) holds true.
Proof
For each p > n, let \(u_{p} \in (W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\) be the minimiser of Ep, α, β given by Lemma 5. (We will follow a similar method and utilise the estimates appearing therein). For any fixed \(q \in (n,\infty )\) and p ≥ q, Hölder’s inequality and minimality yield
By employing again the coercivity of Eq, α, β, we have the estimate
which by Poincaré’s inequality and the interpolation inequalities in \(W^{\bar n,2}({\varOmega })\), yield
for some C > 0 depending on q and all the parameters, but independent of p. By a standard diagonal argument, for any sequence \((p_{i})_{1}^{\infty }\) with \(p_{i} \longrightarrow \infty \) as \(i\to \infty \), there exists a function
and a subsequence (denoted again by \((p_{i})_{1}^{\infty }\)) along which (1.18) holds true. It remains to show that \(u_{\infty }\) is in fact a minimiser of \(\mathrm {E}_{\infty }\) over the same space. To this end, note that for any fixed \(q\in (n,\infty )\) and p ≥ q, we have
for any \(v\in (W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\). By the weak lower semi-continuity of Eq, α, β in the space \((W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\) demonstrated in Lemma 5, we have
for any \(v\in (W^{\bar n, 2}\cap W^{1,2}_{g})({\varOmega })\). By letting \(q\to \infty \) in the estimate above, we deduce that
as desired. □
Lemma 7
For any α, β, γ > 0 and p > n, consider the minimiser up of the functional Ep, α, β over \((W^{2,p}\cap W^{1,p}_{g})({\varOmega })\) constructed in Lemma 5. Consider also the signed Radon measures \(\mu _{p} \in {\mathscr{M}}({\varOmega })\) and \(\nu _{p} \in {\mathscr{M}}(\mathcal {K})\), defined in (1.19). Then, the triplet (up, μp, νp) satisfies the PDE (1.20) in the distributional sense, namely (2.1) holds true for all test functions \(\phi \in C^{\bar n}_{c}({\varOmega })\).
Proof
We involve a standard Gateaux differentiability argument. Let us begin by checking that μp, νp as defined in (1.19) are uniformly bounded Radon measures when \(u_{p} \in W^{\bar n,2}({\varOmega })\). Since by the regularity of F, K, up they obviously define absolutely continuous measures, it suffices to check that by Hölder inequality’s, we have the total variation estimates
and similarly
Next, fix \(\phi \in C^{\bar n}_{c}({\varOmega })\). By using the regularity assumptions on F, K, we formally compute, recalling the abbreviations F[v] = F(⋅, v,Dv,D2v) and K[v] = K(⋅, v,Dv):
By invoking the definitions of μp, νp, the above yields that
Since up is the minimiser of Ep, α, β, we have that Ep, α, β(up) ≤Ep, α, β(up + εϕ) for all \(\varepsilon \in \mathbb {R}\) and all \(\phi \in C^{\bar n}_{c}({\varOmega })\). Hence, our computation implies that the PDE (1.20) is indeed satisfied as claimed, once we confirm that the formal computation is rigorous, and that therefore Ep is Gateaux differentiable at the minimiser up for any direction \(\phi \in C^{\bar n}_{c}({\varOmega })\) because by the continuity of F, K and the fact that \(u_{p} \in (C^{2} \cap W^{\bar n,2})({\varOmega })\), \(\mathrm {F}[u_{p}] \in C(\overline {{\varOmega }})\) and \(\mathrm {K}[u_{p}]-k^{\gamma } \in L^{\infty }(\mathcal {K},{\mathscr{H}}^{\kappa })\), Hölder’s inequality implies that
and
for any \(\phi \in C^{\bar n}_{c}({\varOmega })\). Finally, \(\mathrm {D}^{\bar n} u_{p} : \mathrm {D}^{\bar n} \phi \in L^{1}({\varOmega })\) since \(\mathrm {D}^{\bar n} u_{p} \in L^{2}\big ({\varOmega },\mathbb {R}^{n^{\otimes \bar n}}_{s}\big )\). □
Lemma 8
For any α, β, γ > 0, consider the minimiser \(u_{\infty }\) of \(\mathrm {E}_{\infty ,\alpha ,\beta }\) constructed in Lemma 6 and the minimisers (up)p>n of the functionals (Ep, α, β)p>n. Then, there exist signed Radon measures \(\mu _{\infty } \in {\mathscr{M}}({\varOmega })\) and \(\nu _{\infty } \in {\mathscr{M}}(\mathcal {K})\) such that the triplet \((u_{\infty },\mu _{\infty },\nu _{\infty })\) satisfies the PDE (1.15) in the distributional sense, that is (2.2) holds true. Additionally, there exists a further subsequence along which the weak* modes of convergence of (1.19) hold true as \(p\to \infty \).
Proof
By the proof of Lemma 7, we have the uniform in p total variation bounds ∥μp∥(Ω) ≤ 1 and \(\|\nu _{p}\|(\mathcal {K})\leq 1\). Hence, by the sequential weak* compactness of the spaces of Radon measures
there exists a further subsequence denoted again by \((p_{i})_{1}^{\infty }\) such that \(\mu _{p} \overset {\ast }{\rightharpoonup } \mu _{\infty }\) in \({\mathscr{M}}({\varOmega })\) and \(\nu _{p} \overset {\ast }{\rightharpoonup } \nu _{\infty }\) in \({\mathscr{M}}(\mathcal {K})\), as \(p_{i} \to \infty \). Fix now \(\phi \in C^{\bar n}_{c}({\varOmega })\). By Lemma 7, we have that the triplet (up, μp, νp) satisfies (1.20), that is (2.1) holds true for any fixed test function. Since \(u_{p} \longrightarrow u_{\infty }\) in \(C^{2}(\overline {{\varOmega }})\) as \(p_{i} \to \infty \), F, Fr, Fp, FX, Kr, Kp are all continuous up to the boundary, we have as \(p_{i} \to \infty \) that
uniformly on \(\mathcal K\) and on \(\overline {{\varOmega }}\) respectively, for any fixed \(\phi \in C^{\bar n}_{c}({\varOmega })\). Further, we have \(\mu _{p} \overset {\ast }{\rightharpoonup } \mu _{\infty }\) and \(\nu _{p} \overset {\ast }{\rightharpoonup } \nu _{\infty }\) in \({\mathscr{M}}({\varOmega })\) and in \({\mathscr{M}}(\mathcal K)\) respectively as \(p_{j}\to \infty \). By standard properties of the weak*-strong continuity of the duality pairings (2.3), we have that the first two terms of (2.2) converge to the respective first two terms of (2.1), as \(p_{i} \to \infty \). Finally, since \(\mathrm {D}^{\bar n}u_{p} \rightharpoonup \mathrm {D}^{\bar n}u_{p}\) in \(L^{2}\big ({\varOmega },\mathbb {R}^{n^{\otimes \bar n}}_{s}\big )\) and \(\mathrm {D}^{\bar n}\phi \in L^{2}\big ({\varOmega },\mathbb {R}^{n^{\otimes \bar n}}_{s}\big )\), we also have that the last term of (2.2) converges to the respective last term of (2.1), as \(p_{i} \to \infty \). Hence, we have indeed obtained (1.15) by passing to the limit as \(p_{i} \to \infty \) in (1.20). □
Lemma 9
For any α, β, γ > 0, p > n and \(u^{0} \in (W^{2,p}\cap W^{1,p}_{g})({\varOmega })\) such that
the ((α, β, γ)-dependent) minimiser up of Ep (constructed in Lemmas 5–8), satisfies the error bounds (1.22). If additionally \(u^{0} \in (W^{\bar n,2}\cap W^{1,2}_{g})({\varOmega })\), then the ((α, β, γ)-dependent) minimiser \(u_{\infty }\) of \(\mathrm {E}_{\infty }\) (constructed in Lemmas 5–8), satisfies the error bounds (1.21).
Proof
Let us denote k0 := K[u0], noting that \(k^{0} \in C(\mathcal {K})\) and that
Recall that for any p > n, up is a global minimiser of Ep, α, β in \((W^{\bar n,2}\cap W^{1,2}_{g})({\varOmega })\). Therefore,
This implies
The above together with Minkowski’s and Hölder’s inequalities yields
as claimed. To obtain the corresponding estimate for \(u_{\infty }\) in the case that additionally \(u^{0} \in (W^{\bar n,2}\cap W^{1,2}_{g})({\varOmega })\), we pass to the limit as \(p_{i}\to \infty \) in the estimate above and the conclusion follows by letting \(p_{i}\to \infty \) due to the strong convergence \(u_{p} \longrightarrow u_{\infty }\) in \(C^{2}(\overline {{\varOmega }})\) as \(p_{i}\to \infty \). The lemma ensues. □
The proof is now complete by noting that the statements (1.16)–(1.17) in Theorem 1 follow from Proposition 4 and the established modes of convergence.
References
Alves, C. J. S., Abdallah, J. B., Jaoua, M.: Recovery of cracks using a point-source reciprocity gap function. Inverse Probl. Sci. Eng. 12, 519–574 (2004)
Ambrosio, L., Fusco, N., Pallara, D.: Functions of Bounded Variation and Free Discontinuity Problems. Oxford University Press, Oxford (2000)
Aronsson, G.: Minimization problems for the functional \(\sup _{x} \mathcal {F}(x,f(x), f^{\prime }(x))\). Arkiv Mat. 6, 33–53 (1965)
Aronsson, G.: Minimization problems for the functional \(\sup _{x} \mathcal {F}(x, f(x), f^{\prime }(x))\) II. Arkiv Mat. 6, 409–431 (1966)
Aronsson, G.: Extension of functions satisfying Lipschitz conditions. Arkiv Mat. 6, 551–561 (1967)
Aronsson, G.: On certain minimax problems and Pontryagin’s maximum principle. Calc. Var. PDE 37, 99–109 (2010)
Aronsson, G., Barron, E.N.: \(L^{\infty },\) variational problems with running costs and constraints. Appl. Math. Optim. 65, 53–90 (2012)
El Badia, A., Ha Duong, T.: Some remarks on the problem of source identification from boundary measurements. Inverse Probl. 14, 883 (1998)
Barron, E. N., Jensen, R. R., Wang, C. Y.: The Euler equation and absolute minimizers of \(L^{\infty }\) functionals. Arch. Rational Mech. Anal. 157, 255–283 (2001)
Bocea, M., Nesi, V.: Γ-convergence of power-law functionals, variational principles in \(L^{\infty }\), and applications. SIAM J. Math. Anal. 39, 1550–1576 (2008)
Bocea, M., Popovici, C.: Variational principles in \(L^{\infty }\) with applications to antiplane shear and plane stress plasticity. J. Convex Anal. 18, 403–416 (2011)
Caffarelli, L. A.: Interior a priori estimates for solutions of fully non-linear equations. Ann. Math. 130, 189–213 (1989)
Caffarelli, L. A., Cabré, X.: Fully Nonlinear Elliptic Equations. American Mathematical Society, Providence (1995)
Champion, T., De Pascale, L., Prinari, F.: Γ-convergence and absolute minimizers for supremal functionals. ESAIM: COCV 10, 14–27 (2004)
Dacorogna, B.: Direct Methods in the Calculus of Variations, 2nd edn. Applied Mathematical Sciences, vol. 78. Springer, Berlin (2008)
Dong, H., Krylov, N. V., Li, X.: On fully nonlinear elliptic and parabolic equations with VMO coefficients in domains. St. Petersburg Math. J. 24, 39–69 (2013)
Engl, H. W., Hanke, M., Neubauer, A.: Regularization of Inverse Problems. Springer, Netherlands (1996)
Evans, L. C.: Weak Convergence Methods for Nonlinear Partial Differential Equations. Regional Conference Series in Mathematics, vol. 74. American Mathematical Society, Providence (1990)
Fonseca, I., Leoni, G.: Modern Methods in the Calculus of Variations: Lp Spaces. Springer Monographs in Mathematics. Springer, New York (2007)
Garroni, A., Nesi, V., Ponsiglione, M.: Dielectric breakdown: optimal bounds. Proc. R. Soc. A 457, 2317–2335 (2001)
Giaquinta, M., Martinazzi, L.: An Introduction to the Regularity Theory for Elliptic Systems, Harmonic Maps and Minimal Graphs. Publications of the Scuola Normale Superiore. Edizioni della Normale Pisa (2012)
Gilbarg, D., Trudinger, N.: Elliptic Partial Differential Equations of Second Order. Classics in Mathematics, vol. 224 (Reprint of the 1998 edition). Springer, Berlin (2001)
Isakov, V.: Inverse Source Problems. Mathematical Surveys and Monographs, vol. 34. American Mathematical Society, Providence (1990)
Katzourakis, N.: An Introduction to Viscosity Solutions for Fully Nonlinear PDE with Applications to Calculus of Variations in \(L^{\infty }\). Springer Briefs in Mathematics. Springer, Cham (2015)
Katzourakis, N.: An \(L^{\infty }\) regularisation strategy to the inverse source identification problem for elliptic equations. SIAM J. Math. Anal. 51, 1349–1370 (2019)
Katzourakis, N., Moser, R.: Existence, uniqueness and structure of second order absolute minimisers. Arch. Rational Mech. Anal. 231, 1615–1634 (2019)
Katzourakis, N., Vǎrvǎrucǎ, E.: An Illustrative Introduction to Modern Analysis. CRC Press/Taylor & Francis, Boca Raton (2017)
Krylov, N.V.: On the existence of \({W^{2}_{p}}\) solutions for fully nonlinear elliptic equations under either relaxed or no convexity assumptions. Commun. Contemp. Math. 19(06), 1750009 (2017)
Kirsch, A.: An Introduction to the Mathematical Theory of Inverse Problems, 2nd edn. Applied Mathematical Sciences, vol. 120. Springer, New York (2011)
Li, D., Zhang, K.: W2, p interior estimates of fully nonlinear elliptic equations. Bull. Lond. Math. Soc. 47, 301–314 (2015)
Ling, L., Hon, Y.C., Yamamoto, M.: Inverse source identification for Poisson equation. Inverse Probl. Sci. Eng. 13, 433–447 (2005)
Magnoli, N., Viano, G.A.: The source identification problem in electromagnetic theory. J. Math. Phys. 38, 2366–2388 (1997)
Neubauer, A.: An a posteriori parameter choice for Tikhonov regularization in Hilbert scales leading to optimal convergence rates. SIAM J. Numer. Anal. 25, 1313–1326 (1988)
Nara, T., Ando, S.: A projective method for an inverse source problem of the Poisson equation. Inverse Probl. 19, 355 (2003)
Ribeiro, A.N., Zappale, E.: Existence of minimisers for nonlevel convex supremal functionals. SIAM J. Control Optim. 52, 3341–3370 (2014)
Shigeta, T., Hon, Y.-C.: Numerical source identification for Poisson equation. In: Tanaka, M. (ed.) Inverse Problems in Engineering Mechanics IV (Nagano, Japan), pp 137–175. Elsevier (2003)
Xie, O., Zhao, Z.: Identifying an unknown source in the Poisson equation by a modified Tikhonov regularization method. Int. J. Math. Comput. Sci. 6, 86–90 (2012)
Yang, F.: The truncation method for identifying an unknown source in the Poisson equation. Appl. Math. Comput. 22, 9334–9339 (2011)
Yang, F., Fu, C.-L.: The modified regularization method for identifying the unknown source on Poisson equation. Appl. Math. Model. 36, 756–763 (2012)
Zhao, Z., Meng, Z., You, L., Xie, O.: Identifying an unknown source in the Poisson equation by the method of Tikhonov regularization in Hilbert scales. Appl. Math. Model. 38, 4686–4693 (2014)
Author information
Authors and Affiliations
Corresponding author
Additional information
Dedicated to Enrique Zuazua on the occasion of his 60th birthday.
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
Ayanbayev, B., Katzourakis, N. On the Inverse Source Identification Problem in \(L^{\infty }\) for Fully Nonlinear Elliptic PDE. Vietnam J. Math. 49, 815–829 (2021). https://doi.org/10.1007/s10013-021-00515-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10013-021-00515-6
Keywords
- Regularisation strategy
- Tykhonov regularisation
- Inverse source identification problem
- Fully nonlinear elliptic equations
- Calculus of Variations in \(L^{\infty }\)