A stabilized finite element method for inverse problems subject to the convection-diffusion equation. I: diffusion-dominated regime

Abstract

The numerical approximation of an inverse problem subject to the convection–diffusion equation when diffusion dominates is studied. We derive Carleman estimates that are on a form suitable for use in numerical analysis and with explicit dependence on the Peclet number. A stabilized finite element method is then proposed and analysed. An upper bound on the condition number is first derived. Combining the stability estimates on the continuous problem with the numerical stability of the method, we then obtain error estimates in local H1- or L2-norms that are optimal with respect to the approximation order, the problem’s stability and perturbations in data. The convergence order is the same for both norms, but the H1-estimate requires an additional divergence assumption for theconvective field. The theory is illustrated in some computational examples

Authors

Erik Burman
Department of Mathematics, University College London, Gower Street, London UK, WC1E 6BT

Mihai Nechita
Department of Mathematics, University College London, Gower Street, London UK, WC1E 6BT

Lauri Oksanen
Department of Mathematics, University College London, Gower Street, London UK, WC1E 6BT

Keywords

convection-diffusion equation; FEM (Finite Element Method);

Paper coordinates

E. Burman, M. Nechita, L. Oksanen, A stabilized finite element method for inverse problems subject to the convection-diffusion equation. I: diffusion-dominated regime, Numer. Math. 144 (2020), pp. 451-477.
https://doi.org/10.1007/s00211-019-01087-x

PDF

About this paper

Journal

Numerische Mathematik

Publisher Name

Springer

Print ISSN

0029-599X

Online ISSN

0945-3245

[1] R. Becker and B. Vexler. Optimal control of the convection-diffusion equation using stabilized finite element methods. Numer. Math., 106(3):349–367, 2007.
[2] S. C. Brenner. Poincare-Friedrichs inequalities for piecewise H1 functions. SIAM J. Numer. Anal., 41(1):306–324, 2003.

[3] E. Burman. A unified analysis for conforming and nonconforming stabilized finite element methods using interior penalty. SIAM J. Numer. Anal., 43(5):2012–2033, 2005.

[4] E. Burman. Stabilized finite element methods for nonsymmetric, noncoercive, and ill-posed problems. Part I: elliptic equations. SIAM J. Sci. Comput., 35(6):A2752 A2780, 2013.

[5] E. Burman. Error estimates for stabilized finite element methods applied to ill-posed problems. C. R. Math. Acad. Sci. Paris, 352(7-8):655–659, 2014.

[6] E. Burman, P. Hansbo, and M. G. Larson. Solving ill-posed control problems by stabilized finite  element methods: an alternative to Tikhonov regularization. Inverse Problems, 34:035004, 2018.

[7] E. Burman, M. Nechita, and L. Oksanen. A stabilized finite element method for inverse problems subject to the convection–diffusion equation. II: convection dominated regime. in preparation, 2018.

[8] E. Burman, M. Nechita, and L. Oksanen. Unique continuation for the Helmholtz equation using stabilized finite element methods. J. Math. Pures Appl., 2018.

[9] E. Burman and L. Oksanen. Data assimilation for the heat equation using stabilized finite element methods. Numer. Math., 139(3):505–528, 2018.

[10] L. Dede’ and A. Quarteroni. Optimal control and numerical adaptivity for advection-diffusion equations. M2AN Math. Model. Numer. Anal., 39(5):1019–1040, 2005.

[11] D. Dos Santos Ferreira, C. E. Kenig, M. Salo, and G. Uhlmann. Limiting Carleman weights and anisotropic inverse problems. Invent. Math., 178(1):119–171, 2009.

[12] A. Ern and J.-L. Guermond. Theory and practice of finite elements, volume 159 of Applied Mathematical Sciences. Springer-Verlag, New York, 2004.

[13] A. Ern and J.-L. Guermond. Evaluation of the condition number in linear systems arising in finite element approximations. M2AN Math. Model. Numer. Anal., 40(1):29 48, 2006.

[14] F. Hecht. New development in FreeFem++. J. Numer. Math., 20(3-4):251–265, 2012.

[15] M. Hinze, N. Yan, and Z. Zhou. Variational discretization for optimal control governed by convection dominated diffusion equations. J. Comput. Math., 27(2-3):237–253, 2009.

[16] J. Le Rousseau and G. Lebeau. On Carleman estimates for elliptic and parabolic operators. Applications to unique continuation and control of parabolic equations. ESAIM Control Optim. Calc. Var., 18(3):712–747, 2012.

[17] E. Malinnikova and S. Vessella. Quantitative uniqueness for elliptic equations with singular lower order terms. Math. Ann., 353(4):1157–1181, 2012.

[18] P. Monk and E. Suli. The adaptive computation of far-field patterns by a posteriori error estimation of linear functionals. SIAM J. Numer. Anal., 36(1):251–274, 1999.

[19] N. Yan and Z. Zhou. A priori and a posteriori error analysis of edge stabilization Galerkin method for the optimal control problem governed by convection-dominated diffusion equation. J. Comput. Appl. Math., 223(1):198–217, 2009.

[20] M. Zworski. Semiclassical analysis, volume 138 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2012.[32] M.D. Zeiler, ADADELTA : An adaptive learning rate method, 2012,http://arxiv.org/abs/1212.5701.

1811.00431v2

A STABILIZED FINITE ELEMENT METHOD FOR INVERSE PROBLEMS SUBJECT TO THE CONVECTION-DIFFUSION EQUATION. I: DIFFUSION-DOMINATED REGIME

ERIK BURMAN, MIHAI NECHITA, AND LAURI OKSANEN

Abstract

The numerical approximation of an inverse problem subject to the convectiondiffusion equation when diffusion dominates is studied. We derive Carleman estimates that are of a form suitable for use in numerical analysis and with explicit dependence on the Péclet number. A stabilized finite element method is then proposed and analysed. An upper bound on the condition number is first derived. Combining the stability estimates on the continuous problem with the numerical stability of the method, we then obtain error estimates in localH1H1H^(1)H^{1}H1- orL2L2L^(2)L^{2}L2-norms that are optimal with respect to the approximation order, the problem's stability and perturbations in data. The convergence order is the same for both norms, but theH1H1H^(1)H^{1}H1-estimate requires an additional divergence assumption for the convective field. The theory is illustrated in some computational examples.

1. Introduction

We consider the convection-diffusion equation
(1)Lu:=μΔu+βu=finΩ,(1)Lu:=μΔu+βu=f in Ω,{:(1)Lu:=-mu Delta u+beta*grad u=f quad" in "Omega",":}\begin{equation*} \mathcal{L} u:=-\mu \Delta u+\beta \cdot \nabla u=f \quad \text { in } \Omega, \tag{1} \end{equation*}(1)Lu:=μΔu+βu=f in Ω,
whereΩRnΩRnOmega subR^(n)\Omega \subset \mathbb{R}^{n}ΩRnis open, bounded and connected,μ>0μ>0mu > 0\mu>0μ>0is the diffusion coefficient andβ[W1,(Ω)]nβW1,(Ω)nbeta in[W^(1,oo)(Omega)]^(n)\beta \in \left[W^{1, \infty}(\Omega)\right]^{n}β[W1,(Ω)]nis the convective velocity field. We assume that no information is given on the boundaryΩΩdel Omega\partial \OmegaΩand that there exists a solutionuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)satisfying (1). For an open and connected subsetωΩωΩomega sub Omega\omega \subset \OmegaωΩ, define the perturbed restrictionU~ω:=uω+δU~ω:=uω+δtilde(U)_(omega):=u∣_(omega)+delta\tilde{U}_{\omega}:=u \mid{ }_{\omega}+\deltaU~ω:=uω+δ, whereδL2(ω)δL2(ω)delta inL^(2)(omega)\delta \in L^{2}(\omega)δL2(ω)is an unknown function modelling measurement noise. The data assimilation (or unique continuation) problem consists in findinguuuuugivenfffffandU~ωU~ωtilde(U)_(omega)\tilde{U}_{\omega}U~ω. Here the coefficientsμμmu\muμandββbeta\betaβ, and the source termfffffare assumed to be known. This linear problem is ill-posed and it is closely related to the elliptic Cauchy problem, see e.g. [ARRV09]. Potential applications include for example flow problems for which full boundary data are not accessible, but where local measurements (in a subset of the domain or on a part of the boundary) can be obtained.
The aim is to design a finite element method for data assimilation with weakly consistent regularization applied to the convection-diffusion equation (1). In the present analysis we consider the regime where diffusion dominates and in the companion paper [BNO19a] we treat the one with dominating convective transport. To make this more precise we introduce the Péclet number associated to a given length scalelllllby
Pe(l):=|β|lμ,Pe(l):=|β|lμ,Pe(l):=(|beta|l)/(mu),P e(l):=\frac{|\beta| l}{\mu},Pe(l):=|β|lμ,
for a suitable norm|||||*||\cdot|||forββbeta\betaβ. Ifhhhhhdenotes the characteristic length scale of the computation, we define the diffusive regime byPe(h)<1Pe(h)<1Pe(h) < 1P e(h)<1Pe(h)<1and the convective regime byPe(h)>1Pe(h)>1Pe(h) > 1P e(h)>1Pe(h)>1. It is known that the character of the system changes drastically in the two regimes and we therefore need to apply different concepts of stability in the two cases. In the present paper we assume that the Péclet number is small and we use an approach similar to that employed for the Laplace equation in [Bur14], for the Helmholtz equation in [BNO19b] and for the heat equation in [BO18], that is we combine conditional stability estimates for the physical problem with optimal numerical stability obtained using a bespoke weakly consistent stabilizing term. For high Péclet numbers on the other hand, we prove in [BNO19a] weighted estimates directly on the discrete solution, that reflect the anisotropic character of the convection-diffusion problem.
In the case of optimal control problems subject to convection-diffusion problems that are well-posed, there are several works in the literature on stabilized finite element methods. In [DQ05] the authors considered stabilization using a Galerkin least squares approach in the Lagrangian. Symmetric stabilization in the form of local projection stabilization was proposed in [BV07] and using penalty on the gradient jumps in [YZ09, HYZ09]. The key difference between the well-posed case and the ill-posed case that we consider herein is that we can not use stability of neither the forward nor the backward equations. Crucial instead is the convergence of the weakly consistent stabilizing terms and the matching of the quantities in the discrete method and the available (best) stability of the continuous problem. Such considerations lead to results both in the case of high and low Péclet numbers, but the different stability properties in the two regimes lead to a different analysis for each case that will be considered in the two parts of this paper.
The main results of this current work are the convergence estimates with explicit dependence on the Péclet number in Theorem 1 and Theorem 2, that rely on the continuous three-ball inequalities in Lemma 2 and Corollary 2.

2. Stability estimates

We prove conditional stability estimates for the unique continuation problem subject to the convection-diffusion equation (1) in the form of three-ball inequalities, see e.g. [MV12] and the references therein. The novelty here is that we keep track of explicit dependence on the diffusion coefficientμμmu\muμand the convective vector fieldββbeta\betaβ. The first such inequality is proven in Corollary 1, followed by Lemma 2 and Corollary 2, where the norms for measuring the size of the data are weakened to serve the purpose of devising a finite element method in Section 3.
First we prove an auxiliary logarithmic convexity inequality, which is a more explicit version of [LRL12, Lemma 5.2].
Lemma 1. Suppose thata,b,c0a,b,c0a,b,c >= 0a, b, c \geq 0a,b,c0andp,q>0p,q>0p,q > 0p, q>0p,q>0satisfycbcbc <= bc \leq bcbandcepλa+eqλbcepλa+eqλbc <= e^(p lambda)a+e^(-q lambda)bc \leq e^{p \lambda} a+e^{-q \lambda} bcepλa+eqλbfor allλ>λ00λ>λ00lambda > lambda_(0) >= 0\lambda>\lambda_{0} \geq 0λ>λ00. Then there areC>0C>0C > 0C>0C>0andκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)(depending only onpppppandqqqqq) such that
cCeqλ0aκb1κcCeqλ0aκb1κc <= Ce^(qlambda_(0))a^(kappa)b^(1-kappa)c \leq C e^{q \lambda_{0}} a^{\kappa} b^{1-\kappa}cCeqλ0aκb1κ
Proof. We may assume thata,b>0a,b>0a,b > 0a, b>0a,b>0, sincec=0c=0c=0c=0c=0ifa=0a=0a=0a=0a=0orb=0b=0b=0b=0b=0. The minimizerλλlambda_(**)\lambda_{*}λof the functionf(λ)=epλa+eqλbf(λ)=epλa+eqλbf(lambda)=e^(p lambda)a+e^(-q lambda)bf(\lambda)=e^{p \lambda} a+e^{-q \lambda} bf(λ)=epλa+eqλbis given by
λ=1p+qlogqbpaλ=1p+qlogqbpalambda_(**)=(1)/(p+q)log((qb)/(pa))\lambda_{*}=\frac{1}{p+q} \log \frac{q b}{p a}λ=1p+qlogqbpa
and writingr=q/pr=q/pr=q//pr=q / pr=q/p, the minimum value is
f(λ)=a(qbpa)p/(p+q)+b(qbpa)q/(p+q)=(rp/(p+q)+rq/(p+q))aq/(p+q)bp/(p+q)fλ=aqbpap/(p+q)+bqbpaq/(p+q)=rp/(p+q)+rq/(p+q)aq/(p+q)bp/(p+q)f(lambda_(**))=a((qb)/(pa))^(p//(p+q))+b((qb)/(pa))^(-q//(p+q))=(r^(p//(p+q))+r^(-q//(p+q)))a^(q//(p+q))b^(p//(p+q))f\left(\lambda_{*}\right)=a\left(\frac{q b}{p a}\right)^{p /(p+q)}+b\left(\frac{q b}{p a}\right)^{-q /(p+q)}=\left(r^{p /(p+q)}+r^{-q /(p+q)}\right) a^{q /(p+q)} b^{p /(p+q)}f(λ)=a(qbpa)p/(p+q)+b(qbpa)q/(p+q)=(rp/(p+q)+rq/(p+q))aq/(p+q)bp/(p+q)
This shows that ifλ>λ0λ>λ0lambda_(**) > lambda_(0)\lambda_{*}>\lambda_{0}λ>λ0then
cC1aκb1κcC1aκb1κc <= C_(1)a^(kappa)b^(1-kappa)c \leq C_{1} a^{\kappa} b^{1-\kappa}cC1aκb1κ
whereκ=q/(p+q)κ=q/(p+q)kappa=q//(p+q)\kappa=q /(p+q)κ=q/(p+q)andC1=rp/(p+q)+rq/(p+q)C1=rp/(p+q)+rq/(p+q)C_(1)=r^(p//(p+q))+r^(-q//(p+q))C_{1}=r^{p /(p+q)}+r^{-q /(p+q)}C1=rp/(p+q)+rq/(p+q). On the other hand, ifλλ0λλ0lambda_(**) <= lambda_(0)\lambda_{*} \leq \lambda_{0}λλ0then it holds thateqλ0eqλ=aq/(p+q)(rb)q/(p+q)eqλ0eqλ=aq/(p+q)(rb)q/(p+q)e^(-qlambda_(0)) <= e^(-qlambda_(**))=a^(q//(p+q))(rb)^(-q//(p+q))e^{-q \lambda_{0}} \leq e^{-q \lambda_{*}}=a^{q /(p+q)}(r b)^{-q /(p+q)}eqλ0eqλ=aq/(p+q)(rb)q/(p+q), or equivalently,
bq/(p+q)eqλ0aq/(p+q)rq/(p+q)bq/(p+q)eqλ0aq/(p+q)rq/(p+q)b^(q//(p+q)) <= e^(qlambda_(0))a^(q//(p+q))r^(-q//(p+q))b^{q /(p+q)} \leq e^{q \lambda_{0}} a^{q /(p+q)} r^{-q /(p+q)}bq/(p+q)eqλ0aq/(p+q)rq/(p+q)
Therefore
cb=bq/(p+q)bp/(p+q)eqλ0rq/(p+q)aq/(p+q)bp/(p+q)cb=bq/(p+q)bp/(p+q)eqλ0rq/(p+q)aq/(p+q)bp/(p+q)c <= b=b^(q//(p+q))b^(p//(p+q)) <= e^(qlambda_(0))r^(-q//(p+q))a^(q//(p+q))b^(p//(p+q))c \leq b=b^{q /(p+q)} b^{p /(p+q)} \leq e^{q \lambda_{0}} r^{-q /(p+q)} a^{q /(p+q)} b^{p /(p+q)}cb=bq/(p+q)bp/(p+q)eqλ0rq/(p+q)aq/(p+q)bp/(p+q)
That is, ifλλ0λλ0lambda_(**) <= lambda_(0)\lambda_{*} \leq \lambda_{0}λλ0then
cC2eqλ0aκb1κcC2eqλ0aκb1κc <= C_(2)e^(qlambda_(0))a^(kappa)b^(1-kappa)c \leq C_{2} e^{q \lambda_{0}} a^{\kappa} b^{1-\kappa}cC2eqλ0aκb1κ
whereC2=rq/(p+q)C2=rq/(p+q)C_(2)=r^(-q//(p+q))C_{2}=r^{-q /(p+q)}C2=rq/(p+q). Aseqλ01eqλ01e^(qlambda_(0)) >= 1e^{q \lambda_{0}} \geq 1eqλ01andC1>C2C1>C2C_(1) > C_(2)C_{1}>C_{2}C1>C2, the claim follows by takingC=C1C=C1C=C_(1)C=C_{1}C=C1.
The following Carleman inequality is well-known, see e.g. [LRL12]. For the convenience of the reader we have included an elementary proof in Appendix A.
Proposition 1. LetρC3(Ω)ρC3(Ω)rho inC^(3)(Omega)\rho \in C^{3}(\Omega)ρC3(Ω)andKΩKΩK sub OmegaK \subset \OmegaKΩbe a compact set that does not contain critical points ofρρrho\rhoρ. Letα,τ>0α,τ>0alpha,tau > 0\alpha, \tau>0α,τ>0andϕ=eαρϕ=eαρphi=e^(alpha rho)\phi=e^{\alpha \rho}ϕ=eαρ. LetwC02(K)wC02(K)w inC_(0)^(2)(K)w \in C_{0}^{2}(K)wC02(K)andv=eτϕwv=eτϕwv=e^(tau phi)wv=e^{\tau \phi} wv=eτϕw. Then there isC>0C>0C > 0C>0C>0such that
Ke2τϕ(τ3w2+τ|w|2)dxCKe2τϕ|Δw|2dxKe2τϕτ3w2+τ|w|2dxCKe2τϕ|Δw|2dxint_(K)e^(2tau phi)(tau^(3)w^(2)+tau|grad w|^(2))dx <= Cint_(K)e^(2tau phi)|Delta w|^(2)dx\int_{K} e^{2 \tau \phi}\left(\tau^{3} w^{2}+\tau|\nabla w|^{2}\right) \mathrm{d} x \leq C \int_{K} e^{2 \tau \phi}|\Delta w|^{2} \mathrm{~d} xKe2τϕ(τ3w2+τ|w|2)dxCKe2τϕ|Δw|2 dx
forααalpha\alphaαlarge enough andττ0ττ0tau >= tau_(0)\tau \geq \tau_{0}ττ0, whereτ0>1τ0>1tau_(0) > 1\tau_{0}>1τ0>1depends only onααalpha\alphaαandρρrho\rhoρ.
Using the above Carleman estimate we prove a three-ball inequality that is explicit with respect toμμmu\muμandββbeta\betaβ, i.e. the constants in the inequality are independent of the Péclet number. The corresponding inequality with constant depending implicitly on the Péclet number is proven for instance in [MV12]. We denote byB(x,r)B(x,r)B(x,r)B(x, r)B(x,r)the open ball of radiusrrrrrcentred atxxxxx, and byd(x,Ω)d(x,Ω)d(x,del Omega)d(x, \partial \Omega)d(x,Ω)the distance fromxxxxxto the boundary ofΩΩOmega\OmegaΩ.
Corollary 1. Letx0Ωx0Ωx_(0)in Omegax_{0} \in \Omegax0Ωand0<r1<r2<d(x0,Ω)0<r1<r2<dx0,Ω0 < r_(1) < r_(2) < d(x_(0),del Omega)0<r_{1}<r_{2}<d\left(x_{0}, \partial \Omega\right)0<r1<r2<d(x0,Ω). DefineBj=B(x0,rj),j=1,2Bj=Bx0,rj,j=1,2B_(j)=B(x_(0),r_(j)),j=1,2B_{j}=B\left(x_{0}, r_{j}\right), j=1,2Bj=B(x0,rj),j=1,2. Then there areC>0C>0C > 0C>0C>0andκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)such that forμ>0,β[L(Ω)]nμ>0,βL(Ω)nmu > 0,beta in[L^(oo)(Omega)]^(n)\mu>0, \beta \in\left[L^{\infty}(\Omega)\right]^{n}μ>0,β[L(Ω)]nanduH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)it holds that
uH1(B2)CeCP~e2(uH1(B1)+1μLuL2(Ω))κuH1(Ω)1κuH1B2CeCP~e2uH1B1+1μLuL2(Ω)κuH1(Ω)1κ||u||_(H^(1)(B_(2))) <= Ce^(C tilde(P)e^(2))(||u||_(H^(1)(B_(1)))+(1)/(mu)||Lu||_(L^(2)(Omega)))^(kappa)||u||_(H^(1)(Omega))^(1-kappa)\|u\|_{H^{1}\left(B_{2}\right)} \leq C e^{C \tilde{P} e^{2}}\left(\|u\|_{H^{1}\left(B_{1}\right)}+\frac{1}{\mu}\|\mathcal{L} u\|_{L^{2}(\Omega)}\right)^{\kappa}\|u\|_{H^{1}(\Omega)}^{1-\kappa}uH1(B2)CeCP~e2(uH1(B1)+1μLuL2(Ω))κuH1(Ω)1κ
whereP~e=1+|β|/μP~e=1+|β|/μtilde(P)e=1+|beta|//mu\tilde{P} e=1+|\beta| / \muP~e=1+|β|/μand|β|=β[L(Ω)]n|β|=βL(Ω)n|beta|=||beta||_([L^(oo)(Omega)]^(n))|\beta|=\|\beta\|_{\left[L^{\infty}(\Omega)\right]^{n}}|β|=β[L(Ω)]n.
Proof. Due to the density ofC2(Ω)C2(Ω)C^(2)(Omega)C^{2}(\Omega)C2(Ω)inH2(Ω)H2(Ω)H^(2)(Omega)H^{2}(\Omega)H2(Ω), it is enough to consideruC2(Ω)uC2(Ω)u inC^(2)(Omega)u \in C^{2}(\Omega)uC2(Ω). Let now0<r0<r10<r0<r10 < r_(0) < r_(1)0<r_{0}<r_{1}0<r0<r1andr2<r3<r4<d(x0,Ω)r2<r3<r4<dx0,Ωr_(2) < r_(3) < r_(4) < d(x_(0),del Omega)r_{2}<r_{3}<r_{4}<d\left(x_{0}, \partial \Omega\right)r2<r3<r4<d(x0,Ω). We choose non-positiveρC(Ω)ρC(Ω)rho inC^(oo)(Omega)\rho \in C^{\infty}(\Omega)ρC(Ω)such thatρ(x)=d(x,x0)ρ(x)=dx,x0rho(x)=-d(x,x_(0))\rho(x)=-d\left(x, x_{0}\right)ρ(x)=d(x,x0)outsideB0B0B_(0)B_{0}B0. Since|ρ|=1|ρ|=1|grad rho|=1|\nabla \rho|=1|ρ|=1outsideB0,ρB0,ρB_(0),rhoB_{0}, \rhoB0,ρdoes not have critical points inB4B0B4B0B_(4)\\B_(0)B_{4} \backslash B_{0}B4B0. LetχC0(B4B0)χC0B4B0chi inC_(0)^(oo)(B_(4)\\B_(0))\chi \in C_{0}^{\infty}\left(B_{4} \backslash B_{0}\right)χC0(B4B0)satisfyχ=1χ=1chi=1\chi=1χ=1inB3B1B3B1B_(3)\\B_(1)B_{3} \backslash B_{1}B3B1, and setw=χuw=χuw=chi uw=\chi uw=χu. We apply Proposition 1 withK=B4B0K=B4¯B0K= bar(B_(4))\\B_(0)K=\overline{B_{4}} \backslash B_{0}K=B4B0to get
(2)μ2B4B0(τ3|w|2+τ|w|2)e2τϕdxCB4B0|μΔw|2e2τϕdx(2)μ2B4B0τ3|w|2+τ|w|2e2τϕdxCB4B0|μΔw|2e2τϕdx{:(2)mu^(2)int_(B_(4)\\B_(0))(tau^(3)|w|^(2)+tau|grad w|^(2))e^(2tau phi)dx <= Cint_(B_(4)\\B_(0))|mu Delta w|^(2)e^(2tau phi)dx:}\begin{equation*} \mu^{2} \int_{B_{4} \backslash B_{0}}\left(\tau^{3}|w|^{2}+\tau|\nabla w|^{2}\right) e^{2 \tau \phi} \mathrm{~d} x \leq C \int_{B_{4} \backslash B_{0}}|\mu \Delta w|^{2} e^{2 \tau \phi} \mathrm{~d} x \tag{2} \end{equation*}(2)μ2B4B0(τ3|w|2+τ|w|2)e2τϕ dxCB4B0|μΔw|2e2τϕ dx
forϕ=eαρϕ=eαρphi=e^(alpha rho)\phi=e^{\alpha \rho}ϕ=eαρ, with large enoughα>0α>0alpha > 0\alpha>0α>0, andττ0ττ0tau >= tau_(0)\tau \geq \tau_{0}ττ0(whereτ0>1τ0>1tau_(0) > 1\tau_{0}>1τ0>1depends only onααalpha\alphaαandρρrho\rhoρ). We bound from above the right-hand side by a constant times
B4B0|μΔwβw|2e2τϕdx+|β|2B4B0|w|2e2τϕdxB4B0|μΔwβw|2e2τϕdx+|β|2B4B0|w|2e2τϕdxint_(B_(4)\\B_(0))|mu Delta w-beta*grad w|^(2)e^(2tau phi)dx+|beta|^(2)int_(B_(4)\\B_(0))|grad w|^(2)e^(2tau phi)dx\int_{B_{4} \backslash B_{0}}|\mu \Delta w-\beta \cdot \nabla w|^{2} e^{2 \tau \phi} \mathrm{~d} x+|\beta|^{2} \int_{B_{4} \backslash B_{0}}|\nabla w|^{2} e^{2 \tau \phi} \mathrm{~d} xB4B0|μΔwβw|2e2τϕ dx+|β|2B4B0|w|2e2τϕ dx
Takingτ2|β|2/μ2τ2|β|2/μ2tau >= 2|beta|^(2)//mu^(2)\tau \geq 2|\beta|^{2} / \mu^{2}τ2|β|2/μ2, the second term above is absorbed by the left-hand side of (2) to give
(3)μ2B4B0(τ3|w|2+τ2|w|2)e2τϕdxCB4B0|μΔwβw|2e2τϕdx(3)μ2B4B0τ3|w|2+τ2|w|2e2τϕdxCB4B0|μΔwβw|2e2τϕdx{:(3)mu^(2)int_(B_(4)\\B_(0))(tau^(3)|w|^(2)+(tau)/(2)|grad w|^(2))e^(2tau phi)dx <= Cint_(B_(4)\\B_(0))|mu Delta w-beta*grad w|^(2)e^(2tau phi)dx:}\begin{equation*} \mu^{2} \int_{B_{4} \backslash B_{0}}\left(\tau^{3}|w|^{2}+\frac{\tau}{2}|\nabla w|^{2}\right) e^{2 \tau \phi} \mathrm{~d} x \leq C \int_{B_{4} \backslash B_{0}}|\mu \Delta w-\beta \cdot \nabla w|^{2} e^{2 \tau \phi} \mathrm{~d} x \tag{3} \end{equation*}(3)μ2B4B0(τ3|w|2+τ2|w|2)e2τϕ dxCB4B0|μΔwβw|2e2τϕ dx
Sinceϕ1ϕ1phi <= 1\phi \leq 1ϕ1everywhere, by definingΦ(r)=eαrΦ(r)=eαrPhi(r)=e^(-alpha r)\Phi(r)=e^{-\alpha r}Φ(r)=eαrwe now bound from below the left-hand side in (3) by
μ2B2B1(τ3|w|2+τ|w|2)e2τϕdxμ2τe2τΦ(r2)uH1(B2)2μ2τe2τuH1(B1)2μ2B2B1τ3|w|2+τ|w|2e2τϕdxμ2τe2τΦr2uH1B22μ2τe2τuH1B12mu^(2)int_(B_(2)\\B_(1))(tau^(3)|w|^(2)+tau|grad w|^(2))e^(2tau phi)dx >= mu^(2)taue^(2tau Phi(r_(2)))||u||_(H^(1)(B_(2)))^(2)-mu^(2)taue^(2tau)||u||_(H^(1)(B_(1)))^(2)\mu^{2} \int_{B_{2} \backslash B_{1}}\left(\tau^{3}|w|^{2}+\tau|\nabla w|^{2}\right) e^{2 \tau \phi} \mathrm{~d} x \geq \mu^{2} \tau e^{2 \tau \Phi\left(r_{2}\right)}\|u\|_{H^{1}\left(B_{2}\right)}^{2}-\mu^{2} \tau e^{2 \tau}\|u\|_{H^{1}\left(B_{1}\right)}^{2}μ2B2B1(τ3|w|2+τ|w|2)e2τϕ dxμ2τe2τΦ(r2)uH1(B2)2μ2τe2τuH1(B1)2
An upper bound for the right-hand side in (3) is given by
CB4|μΔuβu|2e2τϕdx+C(B4B3)B1|(μ[Δ,χ]βχ)u|2e2τϕdxCe2τμΔuβuL2(B4)2+Ce2τΦ(r3)(μ2+|β|2)uH1(B4B3)2+Ce2τ(μ2+|β|2)uH1(B1)2CB4|μΔuβu|2e2τϕdx+CB4B3B1|(μ[Δ,χ]βχ)u|2e2τϕdxCe2τμΔuβuL2B42+Ce2τΦr3μ2+|β|2uH1B4B32+Ce2τμ2+|β|2uH1B12{:[Cint_(B_(4))|mu Delta u-beta*grad u|^(2)e^(2tau phi)dx+Cint_((B_(4)\\B_(3))uuB_(1))|(mu[Delta","chi]-beta*grad chi)u|^(2)e^(2tau phi)dx],[quad <= Ce^(2tau)||mu Delta u-beta*grad u||_(L^(2)(B_(4)))^(2)+Ce^(2tau Phi(r_(3)))(mu^(2)+|beta|^(2))||u||_(H^(1)(B_(4)\\B_(3)))^(2)+Ce^(2tau)(mu^(2)+|beta|^(2))||u||_(H^(1)(B_(1)))^(2)]:}\begin{aligned} & C \int_{B_{4}}|\mu \Delta u-\beta \cdot \nabla u|^{2} e^{2 \tau \phi} \mathrm{~d} x+C \int_{\left(B_{4} \backslash B_{3}\right) \cup B_{1}}|(\mu[\Delta, \chi]-\beta \cdot \nabla \chi) u|^{2} e^{2 \tau \phi} \mathrm{~d} x \\ & \quad \leq C e^{2 \tau}\|\mu \Delta u-\beta \cdot \nabla u\|_{L^{2}\left(B_{4}\right)}^{2}+C e^{2 \tau \Phi\left(r_{3}\right)}\left(\mu^{2}+|\beta|^{2}\right)\|u\|_{H^{1}\left(B_{4} \backslash B_{3}\right)}^{2}+C e^{2 \tau}\left(\mu^{2}+|\beta|^{2}\right)\|u\|_{H^{1}\left(B_{1}\right)}^{2} \end{aligned}CB4|μΔuβu|2e2τϕ dx+C(B4B3)B1|(μ[Δ,χ]βχ)u|2e2τϕ dxCe2τμΔuβuL2(B4)2+Ce2τΦ(r3)(μ2+|β|2)uH1(B4B3)2+Ce2τ(μ2+|β|2)uH1(B1)2
Combining the last two inequalities we thus obtain that
μ2e2τΦ(r2)uH1(B2)2Ce2τ((μ2+|β|2)uH1(B1)2+μΔuβuL2(B4)2)+Ce2τΦ(r3)(μ2+|β|2)uH1(B4)2μ2e2τΦr2uH1B22Ce2τμ2+|β|2uH1B12+μΔuβuL2B42+Ce2τΦr3μ2+|β|2uH1B42{:[mu^(2)e^(2tau Phi(r_(2)))||u||_(H^(1)(B_(2)))^(2) <= Ce^(2tau)((mu^(2)+|beta|^(2))||u||_(H^(1)(B_(1)))^(2)+||mu Delta u-beta*grad u||_(L^(2)(B_(4)))^(2))],[+Ce^(2tau Phi(r_(3)))(mu^(2)+|beta|^(2))||u||_(H^(1)(B_(4)))^(2)]:}\begin{aligned} \mu^{2} e^{2 \tau \Phi\left(r_{2}\right)}\|u\|_{H^{1}\left(B_{2}\right)}^{2} & \leq C e^{2 \tau}\left(\left(\mu^{2}+|\beta|^{2}\right)\|u\|_{H^{1}\left(B_{1}\right)}^{2}+\|\mu \Delta u-\beta \cdot \nabla u\|_{L^{2}\left(B_{4}\right)}^{2}\right) \\ & +C e^{2 \tau \Phi\left(r_{3}\right)}\left(\mu^{2}+|\beta|^{2}\right)\|u\|_{H^{1}\left(B_{4}\right)}^{2} \end{aligned}μ2e2τΦ(r2)uH1(B2)2Ce2τ((μ2+|β|2)uH1(B1)2+μΔuβuL2(B4)2)+Ce2τΦ(r3)(μ2+|β|2)uH1(B4)2
forττ0+2|β|2/μ2ττ0+2|β|2/μ2tau >= tau_(0)+2|beta|^(2)//mu^(2)\tau \geq \tau_{0}+2|\beta|^{2} / \mu^{2}ττ0+2|β|2/μ2. We divide byμ2μ2mu^(2)\mu^{2}μ2and conclude by Lemma 1 withp=1Φ(r2)>0p=1Φr2>0p=1-Phi(r_(2)) > 0p=1-\Phi\left(r_{2}\right)>0p=1Φ(r2)>0andq=Φ(r2)Φ(r3)>0q=Φr2Φr3>0q=Phi(r_(2))-Phi(r_(3)) > 0q=\Phi\left(r_{2}\right)-\Phi\left(r_{3}\right)>0q=Φ(r2)Φ(r3)>0, followed by absorbing thePe~=1+|β|/μPe~=1+|β|/μtilde(Pe)=1+|beta|//mu\tilde{P e}=1+|\beta| / \muPe~=1+|β|/μfactor into the exponential factoreCP~e2eCP~e2e^(C tilde(P)e^(2))e^{C \tilde{P} e^{2}}eCP~e2.
We now shift down the Sobolev indices in Corollary 1 by making a similar argument to that in Section 4 of [DSFKSU09] or Section 2.2 of [BNO19b], based on semiclassical pseudodifferential calculus.
Lemma 2. Letx0Ωx0Ωx_(0)in Omegax_{0} \in \Omegax0Ωand0<r1<r2<d(x0,Ω)0<r1<r2<dx0,Ω0 < r_(1) < r_(2) < d(x_(0),del Omega)0<r_{1}<r_{2}<d\left(x_{0}, \partial \Omega\right)0<r1<r2<d(x0,Ω). DefineBj=B(x0,rj),j=1,2Bj=Bx0,rj,j=1,2B_(j)=B(x_(0),r_(j)),j=1,2B_{j}=B\left(x_{0}, r_{j}\right), j=1,2Bj=B(x0,rj),j=1,2. Then there areC>0C>0C > 0C>0C>0andκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)such that forμ>0,β[L(Ω)]nμ>0,βL(Ω)nmu > 0,beta in[L^(oo)(Omega)]^(n)\mu>0, \beta \in\left[L^{\infty}(\Omega)\right]^{n}μ>0,β[L(Ω)]nanduH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)it holds that
uL2(B2)CeCP~e2(uL2(B1)+1μLuH1(Ω))κuL2(Ω)1κ,uL2B2CeCP~e2uL2B1+1μLuH1(Ω)κuL2(Ω)1κ,||u||_(L^(2)(B_(2))) <= Ce^(C tilde(P)e^(2))(||u||_(L^(2)(B_(1)))+(1)/(mu)||Lu||_(H^(-1)(Omega)))^(kappa)||u||_(L^(2)(Omega))^(1-kappa),\|u\|_{L^{2}\left(B_{2}\right)} \leq C e^{C \tilde{P} e^{2}}\left(\|u\|_{L^{2}\left(B_{1}\right)}+\frac{1}{\mu}\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right)^{\kappa}\|u\|_{L^{2}(\Omega)}^{1-\kappa},uL2(B2)CeCP~e2(uL2(B1)+1μLuH1(Ω))κuL2(Ω)1κ,
wherePe~=1+|β|/μPe~=1+|β|/μtilde(Pe)=1+|beta|//mu\tilde{P e}=1+|\beta| / \muPe~=1+|β|/μand|β|=β[L(Ω)]nn|β|=βL(Ω)nn|beta|=||beta||_([L^(oo)(Omega)]^(n)^(n))|\beta|=\|\beta\|_{{\left[L^{\infty}(\Omega)\right]^{n}}^{n}}|β|=β[L(Ω)]nn.
Proof. Let>0>0ℏ > 0\hbar>0>0be the semiclassical parameter that satisfies=1/τ=1/τℏ=1//tau\hbar=1 / \tau=1/τ, whereττtau\tauτis the parameter previously introduced in Proposition 1. We will make use of the theory of semiclassical pseudodifferential operators, which we briefly recall in Appendix B for the convenience of the reader. In particular we will use semiclassical Sobolev spaces with norms given by
uHscls(Rn)=JsuL2(Rn)uHsclsRn=JsuL2Rn||u||_(H_(scl)^(s)(R^(n)))=||J^(s)u||_(L^(2)(R^(n)))\|u\|_{H_{\mathrm{scl}}^{s}\left(\mathbb{R}^{n}\right)}=\left\|J^{s} u\right\|_{L^{2}\left(\mathbb{R}^{n}\right)}uHscls(Rn)=JsuL2(Rn)
where the scale of the semiclassical Bessel potentials is defined by
Js=(12Δ)s/2,sRJs=12Δs/2,sRJ^(s)=(1-ℏ^(2)Delta)^(s//2),quad s inRJ^{s}=\left(1-\hbar^{2} \Delta\right)^{s / 2}, \quad s \in \mathbb{R}Js=(12Δ)s/2,sR
We will also use the following commutator and pseudolocal estimates, see Appendix B. Suppose thatη,ϑC0(Rn)η,ϑC0Rneta,vartheta inC_(0)^(oo)(R^(n))\eta, \vartheta \in C_{0}^{\infty}\left(\mathbb{R}^{n}\right)η,ϑC0(Rn)and thatη=1η=1eta=1\eta=1η=1nearsupp(ϑ)supp(ϑ)supp(vartheta)\operatorname{supp}(\vartheta)supp(ϑ), and letAψ,BψAψ,BψA_(psi),B_(psi)A_{\psi}, B_{\psi}Aψ,Bψbe two semiclassical pseudodifferential operators of orderss,ms,ms,ms, ms,m, respectively. Then for allp,q,NRp,q,NRp,q,N inRp, q, N \in \mathbb{R}p,q,NR, there isC>0C>0C > 0C>0C>0,
(4)[Aψ,Bψ]uHsclp(Rn)CuHsclp+s+m1(Rn)(5)(1η)AψϑuHsclp(Rn)CNuHsclq(Rn)(4)Aψ,BψuHsclpRnCuHsclp+s+m1Rn(5)(1η)AψϑuHsclpRnCNuHsclqRn{:[(4)||[A_(psi),B_(psi)]u||_(H_(scl)^(p)(R^(n))) <= Cℏ||u||_(H_(scl)^(p+s+m-1)(R^(n)))],[(5)||(1-eta)A_(psi)vartheta u||_(H_(scl)^(p)(R^(n))) <= Cℏ^(N)||u||_(H_(scl)^(q)(R^(n)))]:}\begin{align*} \left\|\left[A_{\psi}, B_{\psi}\right] u\right\|_{H_{\mathrm{scl}}^{p}\left(\mathbb{R}^{n}\right)} & \leq C \hbar\|u\|_{H_{\mathrm{scl}}^{p+s+m-1}\left(\mathbb{R}^{n}\right)} \tag{4}\\ \left\|(1-\eta) A_{\psi} \vartheta u\right\|_{H_{\mathrm{scl}}^{p}\left(\mathbb{R}^{n}\right)} & \leq C \hbar^{N}\|u\|_{H_{\mathrm{scl}}^{q}\left(\mathbb{R}^{n}\right)} \tag{5} \end{align*}(4)[Aψ,Bψ]uHsclp(Rn)CuHsclp+s+m1(Rn)(5)(1η)AψϑuHsclp(Rn)CNuHsclq(Rn)
Let0<rj<rj+1<d(x0,Ω),j=0,,40<rj<rj+1<dx0,Ω,j=0,,40 < r_(j) < r_(j+1) < d(x_(0),del Omega),j=0,dots,40<r_{j}<r_{j+1}<d\left(x_{0}, \partial \Omega\right), j=0, \ldots, 40<rj<rj+1<d(x0,Ω),j=0,,4andBj=B(x0,rj)Bj=Bx0,rjB_(j)=B(x_(0),r_(j))B_{j}=B\left(x_{0}, r_{j}\right)Bj=B(x0,rj), keepingB1,B2B1,B2B_(1),B_(2)B_{1}, B_{2}B1,B2unchanged. Letr~j(rj1,rj)r~jrj1,rjtilde(r)_(j)in(r_(j-1),r_(j))\tilde{r}_{j} \in\left(r_{j-1}, r_{j}\right)r~j(rj1,rj)andB~j=B(x0,r~j),j=0,,3B~j=Bx0,r~j,j=0,,3tilde(B)_(j)=B(x_(0), tilde(r)_(j)),j=0,dots,3\tilde{B}_{j}=B\left(x_{0}, \tilde{r}_{j}\right), j=0, \ldots, 3B~j=B(x0,r~j),j=0,,3, wherer1=0r1=0r_(-1)=0r_{-1}=0r1=0. ChooseρC(Ω)ρC(Ω)rho inC^(oo)(Omega)\rho \in C^{\infty}(\Omega)ρC(Ω)such thatρ(x)=d(x,x0)ρ(x)=dx,x0rho(x)=-d(x,x_(0))\rho(x)=-d\left(x, x_{0}\right)ρ(x)=d(x,x0)outsideB~0B~0tilde(B)_(0)\tilde{B}_{0}B~0, and defineϕ=eαρϕ=eαρphi=e^(alpha rho)\phi=e^{\alpha \rho}ϕ=eαρfor large enoughααalpha\alphaα. ConsidervC0(B5B~0)vC0B5B~0v inC_(0)^(oo)(B_(5)\\ tilde(B)_(0))v \in C_{0}^{\infty}\left(B_{5} \backslash \tilde{B}_{0}\right)vC0(B5B~0). As in Appendix A, by taking=ϕ/=ϕ/ℓ=phi//ℏ\ell=\phi / \hbar=ϕ/andσ=Δ+3αλϕ/σ=Δ+3αλϕ/sigma=Deltaℓ+3alpha lambda phi//ℏ\sigma=\Delta \ell+3 \alpha \lambda \phi / \hbarσ=Δ+3αλϕ/, we obtain
CRn|eϕ/Δ(eϕ/v)|2dxRn(1|v|2+3v2|v|22v2)dx.CRneϕ/Δeϕ/v2dxRn1|v|2+3v2|v|22v2dx.Cint_(R^(n))|e^(phi//ℏ)Delta(e^(-phi//ℏ)v)|^(2)dx >= int_(R^(n))(ℏ^(-1)|grad v|^(2)+ℏ^(-3)v^(2)-|grad v|^(2)-ℏ^(-2)v^(2))dx.C \int_{\mathbb{R}^{n}}\left|e^{\phi / \hbar} \Delta\left(e^{-\phi / \hbar} v\right)\right|^{2} \mathrm{~d} x \geq \int_{\mathbb{R}^{n}}\left(\hbar^{-1}|\nabla v|^{2}+\hbar^{-3} v^{2}-|\nabla v|^{2}-\hbar^{-2} v^{2}\right) \mathrm{d} x .CRn|eϕ/Δ(eϕ/v)|2 dxRn(1|v|2+3v2|v|22v2)dx.
Scaling this withμ24μ24mu^(2)ℏ^(4)\mu^{2} \hbar^{4}μ24, we insert the convective term and obtain that
CRn(μeϕ/2Δ(eϕ/v)eϕ/2β(eϕ/v))2dxCRnμeϕ/2Δeϕ/veϕ/2βeϕ/v2dxCint_(R^(n))(mue^(phi//ℏ)ℏ^(2)Delta(e^(-phi//ℏ)v)-e^(phi//ℏ)ℏ^(2)beta*grad(e^(-phi//ℏ)v))^(2)dxC \int_{\mathbb{R}^{n}}\left(\mu e^{\phi / \hbar} \hbar^{2} \Delta\left(e^{-\phi / \hbar} v\right)-e^{\phi / \hbar} \hbar^{2} \beta \cdot \nabla\left(e^{-\phi / \hbar} v\right)\right)^{2} \mathrm{~d} xCRn(μeϕ/2Δ(eϕ/v)eϕ/2β(eϕ/v))2 dx
can be bounded from below by
Rnμ2(2|v|2+v2)dxRn2μ2(2|v|2+v2)dxRn(eϕ/2β(eϕ/v))2dx.Rnμ22|v|2+v2dxRn2μ22|v|2+v2dxRneϕ/2βeϕ/v2dx.int_(R^(n))ℏmu^(2)(ℏ^(2)|grad v|^(2)+v^(2))dx-int_(R^(n))ℏ^(2)mu^(2)(ℏ^(2)|grad v|^(2)+v^(2))dx-int_(R^(n))(e^(phi//ℏ)ℏ^(2)beta*grad(e^(-phi//ℏ)v))^(2)dx.\int_{\mathbb{R}^{n}} \hbar \mu^{2}\left(\hbar^{2}|\nabla v|^{2}+v^{2}\right) \mathrm{d} x-\int_{\mathbb{R}^{n}} \hbar^{2} \mu^{2}\left(\hbar^{2}|\nabla v|^{2}+v^{2}\right) \mathrm{d} x-\int_{\mathbb{R}^{n}}\left(e^{\phi / \hbar} \hbar^{2} \beta \cdot \nabla\left(e^{-\phi / \hbar} v\right)\right)^{2} \mathrm{~d} x .Rnμ2(2|v|2+v2)dxRn2μ2(2|v|2+v2)dxRn(eϕ/2β(eϕ/v))2 dx.
Since
eϕ/2β(eϕ/v)=(βϕ)v+2βveϕ/2βeϕ/v=(βϕ)v+2βve^(phi//ℏ)ℏ^(2)beta*grad(e^(-phi//ℏ)v)=-ℏ(beta*grad phi)v+ℏ^(2)beta*grad ve^{\phi / \hbar} \hbar^{2} \beta \cdot \nabla\left(e^{-\phi / \hbar} v\right)=-\hbar(\beta \cdot \nabla \phi) v+\hbar^{2} \beta \cdot \nabla veϕ/2β(eϕ/v)=(βϕ)v+2βv
introducing the conjugated operatorPv=2eϕ/L(eϕ/v)Pv=2eϕ/Leϕ/vPv=-ℏ^(2)e^(phi//ℏ)L(e^(-phi//ℏ)v)P v=-\hbar^{2} e^{\phi / \hbar} \mathcal{L}\left(e^{-\phi / \hbar} v\right)Pv=2eϕ/L(eϕ/v), the previous bound implies
CPvL2(Rn)2μ2vHscl1(Rn)22μ2vHscl1(Rn)22|β|2vHscl1(Rn)2.CPvL2Rn2μ2vHscl1Rn22μ2vHscl1Rn22|β|2vHscl1Rn2.C||Pv||_(L^(2)(R^(n)))^(2) >= ℏmu^(2)||v||_(H_(scl)^(1)(R^(n)))^(2)-ℏ^(2)mu^(2)||v||_(H_(scl)^(1)(R^(n)))^(2)-ℏ^(2)|beta|^(2)||v||_(H_(scl)^(1)(R^(n)))^(2).C\|P v\|_{L^{2}\left(\mathbb{R}^{n}\right)}^{2} \geq \hbar \mu^{2}\|v\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)}^{2}-\hbar^{2} \mu^{2}\|v\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)}^{2}-\hbar^{2}|\beta|^{2}\|v\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)}^{2} .CPvL2(Rn)2μ2vHscl1(Rn)22μ2vHscl1(Rn)22|β|2vHscl1(Rn)2.
The last two terms in the right-hand side can be absorbed by the first one when
(6)12and12μ2|β|2,(6)12 and 12μ2|β|2,{:(6)ℏ <= (1)/(2)" and "ℏ <= (1)/(2)(mu^(2))/(|beta|^(2))",":}\begin{equation*} \hbar \leq \frac{1}{2} \text { and } \hbar \leq \frac{1}{2} \frac{\mu^{2}}{|\beta|^{2}}, \tag{6} \end{equation*}(6)12 and 12μ2|β|2,
thus obtaining
(7)μvHscl1(Rn)CPvL2(Rn).(7)μvHscl1RnCPvL2Rn.{:(7)sqrtℏmu||v||_(H_(scl)^(1)(R^(n))) <= C||Pv||_(L^(2)(R^(n))).:}\begin{equation*} \sqrt{\hbar} \mu\|v\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)} \leq C\|P v\|_{L^{2}\left(\mathbb{R}^{n}\right)} . \tag{7} \end{equation*}(7)μvHscl1(Rn)CPvL2(Rn).
Let nowη,ϑC0(B5B~0)η,ϑC0B5B~0eta,vartheta inC_(0)^(oo)(B_(5)\\ tilde(B)_(0))\eta, \vartheta \in C_{0}^{\infty}\left(B_{5} \backslash \tilde{B}_{0}\right)η,ϑC0(B5B~0)and suppose thatϑ=1ϑ=1vartheta=1\vartheta=1ϑ=1nearB4B0B4B0B_(4)\\B_(0)B_{4} \backslash B_{0}B4B0andη=1η=1eta=1\eta=1η=1nearsupp(ϑ)supp(ϑ)supp(vartheta)\operatorname{supp}(\vartheta)supp(ϑ). Let alsoχC0(B4B0)χC0B4B0chi inC_(0)^(oo)(B_(4)\\B_(0))\chi \in C_{0}^{\infty}\left(B_{4} \backslash B_{0}\right)χC0(B4B0)satisfyχ=1χ=1chi=1\chi=1χ=1inB3B~1B3B~1B_(3)\\ tilde(B)_(1)B_{3} \backslash \tilde{B}_{1}B3B~1. Then there is0>00>0ℏ_(0) > 0\hbar_{0}>00>0such that forv=χw,wC(Ω)v=χw,wC(Ω)v=chi w,w inC^(oo)(Omega)v=\chi w, w \in C^{\infty}(\Omega)v=χw,wC(Ω), and<0<0ℏ < ℏ_(0)\hbar<\hbar_{0}<0,
(8)vL2(Rn)ηJ1vHscl1(Rn)+(1η)J1ϑvHscl1(Rn)CηJ1vHscl1(Rn)(8)vL2RnηJ1vHscl1Rn+(1η)J1ϑvHscl1RnCηJ1vHscl1Rn{:(8)||v||_(L^(2)(R^(n))) <= ||etaJ^(-1)v||_(H_(scl)^(1)(R^(n)))+||(1-eta)J^(-1)vartheta v||_(H_(scl)^(1)(R^(n))) <= C||etaJ^(-1)v||_(H_(scl)^(1)(R^(n))):}\begin{equation*} \|v\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq\left\|\eta J^{-1} v\right\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)}+\left\|(1-\eta) J^{-1} \vartheta v\right\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)} \leq C\left\|\eta J^{-1} v\right\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)} \tag{8} \end{equation*}(8)vL2(Rn)ηJ1vHscl1(Rn)+(1η)J1ϑvHscl1(Rn)CηJ1vHscl1(Rn)
where we used (5) to absorb one term by the left-hand side. From (8) and (7) we have
(9)μvL2(Rn)CμηJ1vHscl1(Rn)CP(ηJ1v)L2(Rn),(9)μvL2RnCμηJ1vHscl1RnCPηJ1vL2Rn,{:(9)sqrtℏmu||v||_(L^(2)(R^(n))) <= Csqrtℏmu||etaJ^(-1)v||_(H_(scl)^(1)(R^(n))) <= C||P(etaJ^(-1)v)||_(L^(2)(R^(n)))",":}\begin{equation*} \sqrt{\hbar} \mu\|v\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq C \sqrt{\hbar} \mu\left\|\eta J^{-1} v\right\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)} \leq C\left\|P\left(\eta J^{-1} v\right)\right\|_{L^{2}\left(\mathbb{R}^{n}\right)}, \tag{9} \end{equation*}(9)μvL2(Rn)CμηJ1vHscl1(Rn)CP(ηJ1v)L2(Rn),
and the commutator estimate (4) gives
[P,ηJ1]vL2(Rn)CμvL2(Rn)+C2|β|vHscl1(Rn).P,ηJ1vL2RnCμvL2Rn+C2|β|vHscl1Rn.||[P,etaJ^(-1)]v||_(L^(2)(R^(n))) <= Cℏmu||v||_(L^(2)(R^(n)))+Cℏ^(2)|beta|||v||_(H_(scl)^(-1)(R^(n))).\left\|\left[P, \eta J^{-1}\right] v\right\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq C \hbar \mu\|v\|_{L^{2}\left(\mathbb{R}^{n}\right)}+C \hbar^{2}|\beta|\|v\|_{H_{\mathrm{scl}}^{-1}\left(\mathbb{R}^{n}\right)} .[P,ηJ1]vL2(Rn)CμvL2(Rn)+C2|β|vHscl1(Rn).
Recalling the assumption (6), these terms can be absorbed by the left-hand side of (9), obtaining
(10)μvL2(Rn)CηJ1(Pv)L2(Rn)CPvHscl1(Rn)(10)μvL2RnCηJ1(Pv)L2RnCPvHscl1Rn{:(10)sqrtℏmu||v||_(L^(2)(R^(n))) <= C||etaJ^(-1)(Pv)||_(L^(2)(R^(n))) <= C||Pv||_(H_(scl)^(-1)(R^(n))):}\begin{equation*} \sqrt{\hbar} \mu\|v\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq C\left\|\eta J^{-1}(P v)\right\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq C\|P v\|_{H_{\mathrm{scl}}^{-1}\left(\mathbb{R}^{n}\right)} \tag{10} \end{equation*}(10)μvL2(Rn)CηJ1(Pv)L2(Rn)CPvHscl1(Rn)
We now combine this estimate with the technique used to prove Corollary 1. ConsideruC(Rn)uCRnu inC^(oo)(R^(n))u \in C^{\infty}\left(\mathbb{R}^{n}\right)uC(Rn)and setw=eϕ/uw=eϕ/uw=e^(phi//ℏ)uw=e^{\phi / \hbar} uw=eϕ/u. TakeψC0(Ω)ψC0(Ω)psi inC_(0)^(oo)(Omega)\psi \in C_{0}^{\infty}(\Omega)ψC0(Ω)supported inB1(B5B~3)B1B5B~3B_(1)uu(B_(5)\\ tilde(B)_(3))B_{1} \cup\left(B_{5} \backslash \tilde{B}_{3}\right)B1(B5B~3)withψ=1ψ=1psi=1\psi=1ψ=1in(B~1B0)(B4B3)B~1B0B4B3( tilde(B)_(1)\\B_(0))uu(B_(4)\\B_(3))\left(\tilde{B}_{1} \backslash B_{0}\right) \cup\left(B_{4} \backslash B_{3}\right)(B~1B0)(B4B3). Recall thatχC0(B4B0)χC0B4B0chi inC_(0)^(oo)(B_(4)\\B_(0))\chi \in C_{0}^{\infty}\left(B_{4} \backslash B_{0}\right)χC0(B4B0)satisfiesχ=1χ=1chi=1\chi=1χ=1inB3B~1B3B~1B_(3)\\ tilde(B)_(1)B_{3} \backslash \tilde{B}_{1}B3B~1. Using (4) to bound the commutator
[P,χ]wHscl1(Rn)[P,χ]ψwHscl1(Rn)C(μ+|β|)ψwL2(Rn)[P,χ]wHscl1Rn[P,χ]ψwHscl1RnC(μ+|β|)ψwL2Rn||[P,chi]w||_(H_(scl)^(-1)(R^(n))) <= ||[P,chi]psi w||_(H_(scl)^(-1)(R^(n))) <= Cℏ(mu+|beta|)||psi w||_(L^(2)(R^(n)))\|[P, \chi] w\|_{H_{\mathrm{scl}}^{-1}\left(\mathbb{R}^{n}\right)} \leq\|[P, \chi] \psi w\|_{H_{\mathrm{scl}}^{-1}\left(\mathbb{R}^{n}\right)} \leq C \hbar(\mu+|\beta|)\|\psi w\|_{L^{2}\left(\mathbb{R}^{n}\right)}[P,χ]wHscl1(Rn)[P,χ]ψwHscl1(Rn)C(μ+|β|)ψwL2(Rn)
we obtain from (10) that
μχwL2(Rn)CχPwHscl1(Rn)+C(μ+|β|)ψwL2(Rn).μχwL2RnCχPwHscl1Rn+C(μ+|β|)ψwL2Rn.sqrtℏmu||chi w||_(L^(2)(R^(n))) <= C||chi Pw||_(H_(scl)^(-1)(R^(n)))+Cℏ(mu+|beta|)||psi w||_(L^(2)(R^(n))).\sqrt{\hbar} \mu\|\chi w\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq C\|\chi P w\|_{H_{\mathrm{scl}}^{-1}\left(\mathbb{R}^{n}\right)}+C \hbar(\mu+|\beta|)\|\psi w\|_{L^{2}\left(\mathbb{R}^{n}\right)} .μχwL2(Rn)CχPwHscl1(Rn)+C(μ+|β|)ψwL2(Rn).
This leads to
μχeϕ/uL2(Rn)Cχeϕ/(μΔuβu)H1(Rn)+C(μ+|β|)ψeϕ/uL2(Rn),μχeϕ/uL2RnCχeϕ/(μΔuβu)H1Rn+C(μ+|β|)ψeϕ/uL2Rn,sqrtℏmu||chie^(phi//ℏ)u||_(L^(2)(R^(n))) <= C||chie^(phi//ℏ)(mu Delta u-beta*grad u)||_(H^(-1)(R^(n)))+Cℏ(mu+|beta|)||psie^(phi//ℏ)u||_(L^(2)(R^(n))),\sqrt{\hbar} \mu\left\|\chi e^{\phi / \hbar} u\right\|_{L^{2}\left(\mathbb{R}^{n}\right)} \leq C\left\|\chi e^{\phi / \hbar}(\mu \Delta u-\beta \cdot \nabla u)\right\|_{H^{-1}\left(\mathbb{R}^{n}\right)}+C \hbar(\mu+|\beta|)\left\|\psi e^{\phi / \hbar} u\right\|_{L^{2}\left(\mathbb{R}^{n}\right)},μχeϕ/uL2(Rn)Cχeϕ/(μΔuβu)H1(Rn)+C(μ+|β|)ψeϕ/uL2(Rn),
where we used the norm inequalityHscl1(Rn)C2H1(Rn)Hscl1RnC2H1Rn||*||_(H_(scl)^(-1)(R^(n))) <= Cℏ^(-2)||*||_(H^(-1)(R^(n)))\|\cdot\|_{H_{\mathrm{scl}}^{-1}\left(\mathbb{R}^{n}\right)} \leq C \hbar^{-2}\|\cdot\|_{H^{-1}\left(\mathbb{R}^{n}\right)}Hscl1(Rn)C2H1(Rn). LettingΦ(r)=eαrΦ(r)=eαrPhi(r)=e^(-alpha r)\Phi(r)=e^{-\alpha r}Φ(r)=eαrand using a similar argument as in the proof of Corollary 1, we find that
μeΦ(r2)/uL2(B2)Ce1/((μ+|β|)uL2(B1)+32(μΔuβu)H1(Ω))+CeΦ(r~3)/12(μ+|β|)uL2(Ω)μeΦr2/uL2B2Ce1/(μ+|β|)uL2B1+32(μΔuβu)H1(Ω)+CeΦr~3/12(μ+|β|)uL2(Ω){:[mue^(Phi(r_(2))//ℏ)||u||_(L^(2)(B_(2))) <= Ce^(1//ℏ)((mu+|beta|)||u||_(L^(2)(B_(1)))+ℏ^(-(3)/(2))||(mu Delta u-beta*grad u)||_(H^(-1)(Omega)))],[+Ce^(Phi( tilde(r)_(3))//ℏ)ℏ^((1)/(2))(mu+|beta|)||u||_(L^(2)(Omega))]:}\begin{aligned} \mu e^{\Phi\left(r_{2}\right) / \hbar}\|u\|_{L^{2}\left(B_{2}\right)} & \leq C e^{1 / \hbar}\left((\mu+|\beta|)\|u\|_{L^{2}\left(B_{1}\right)}+\hbar^{-\frac{3}{2}}\|(\mu \Delta u-\beta \cdot \nabla u)\|_{H^{-1}(\Omega)}\right) \\ & +C e^{\Phi\left(\tilde{r}_{3}\right) / \hbar} \hbar^{\frac{1}{2}}(\mu+|\beta|)\|u\|_{L^{2}(\Omega)} \end{aligned}μeΦ(r2)/uL2(B2)Ce1/((μ+|β|)uL2(B1)+32(μΔuβu)H1(Ω))+CeΦ(r~3)/12(μ+|β|)uL2(Ω)
when\hbarsatisfies (6) and is small enough. Absorbing the negative power of\hbarin the exponential, we then use Lemma 1 and conclude by absorbing theP~e=1+|β|/μP~e=1+|β|/μtilde(P)e=1+|beta|//mu\tilde{P} e=1+|\beta| / \muP~e=1+|β|/μfactor into the exponential factoreCP~e2eCP~e2e^(C tilde(P)e^(2))e^{C \tilde{P} e^{2}}eCP~e2.
Making the additional coercivity assumptionβ0β0grad*beta <= 0\nabla \cdot \beta \leq 0β0, we can weaken the norms just in the right-hand side of Corollary 1 by using the stability estimate for a well-posed convection-diffusion problem with homogeneous Dirichlet boundary conditions.
Corollary 2. Letx0Ωx0Ωx_(0)in Omegax_{0} \in \Omegax0Ωand0<r1<r2<d(x0,Ω)0<r1<r2<dx0,Ω0 < r_(1) < r_(2) < d(x_(0),del Omega)0<r_{1}<r_{2}<d\left(x_{0}, \partial \Omega\right)0<r1<r2<d(x0,Ω). DefineBj=B(x0,rj),j=1,2Bj=Bx0,rj,j=1,2B_(j)=B(x_(0),r_(j)),j=1,2B_{j}=B\left(x_{0}, r_{j}\right), j=1,2Bj=B(x0,rj),j=1,2. Then there areC>0C>0C > 0C>0C>0andκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)such that forμ>0,β[W1,(Ω)]nμ>0,βW1,(Ω)nmu > 0,beta in[W^(1,oo)(Omega)]^(n)\mu>0, \beta \in\left[W^{1, \infty}(\Omega)\right]^{n}μ>0,β[W1,(Ω)]nhavingesssupΩβ0esssupΩβ0esssup_(Omega)grad*beta <= 0\operatorname{esssup}_{\Omega} \nabla \cdot \beta \leq 0esssupΩβ0, anduH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)it holds that
uH1(B2)CeCP~e2(uL2(B1)+1μLuH1(Ω))κ(uL2(Ω)+1μLuH1(Ω))1κ,uH1B2CeCP~e2uL2B1+1μLuH1(Ω)κuL2(Ω)+1μLuH1(Ω)1κ,||u||_(H^(1)(B_(2))) <= Ce^(C tilde(P)e^(2))(||u||_(L^(2)(B_(1)))+(1)/(mu)||Lu||_(H^(-1)(Omega)))^(kappa)(||u||_(L^(2)(Omega))+(1)/(mu)||Lu||_(H^(-1)(Omega)))^(1-kappa),\|u\|_{H^{1}\left(B_{2}\right)} \leq C e^{C \tilde{P} e^{2}}\left(\|u\|_{L^{2}\left(B_{1}\right)}+\frac{1}{\mu}\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right)^{\kappa}\left(\|u\|_{L^{2}(\Omega)}+\frac{1}{\mu}\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right)^{1-\kappa},uH1(B2)CeCP~e2(uL2(B1)+1μLuH1(Ω))κ(uL2(Ω)+1μLuH1(Ω))1κ,
wherePe~=1+|β|/μPe~=1+|β|/μtilde(Pe)=1+|beta|//mu\tilde{P e}=1+|\beta| / \muPe~=1+|β|/μand|β|=β[L(Ω)]n|β|=βL(Ω)n|beta|=||beta||_([^(L^(oo)(Omega)]^(n)):})|\beta|=\|\beta\|_{\left[^{\left.L^{\infty}(\Omega)\right]^{n}}\right.}|β|=β[L(Ω)]n.
Proof. Let the ballsB0,B3ΩB0,B3ΩB_(0),B_(3)sub OmegaB_{0}, B_{3} \subset \OmegaB0,B3Ωsuch thatBjBj+1BjBj+1B_(j)subB_(j+1)B_{j} \subset B_{j+1}BjBj+1, forj=0,2j=0,2j=0,2j=0,2j=0,2. Consider the wellposed problem
Lw=LuinB3,w=0onB3.Lw=Lu in B3,w=0 on B3.Lw=Lu" in "B_(3),quad w=0" on "delB_(3).\mathcal{L} w=\mathcal{L} u \text { in } B_{3}, \quad w=0 \text { on } \partial B_{3} .Lw=Lu in B3,w=0 on B3.
SinceesssupΩβ0esssupΩβ0esss u p_(Omega)grad*beta <= 0\operatorname{ess} \sup _{\Omega} \nabla \cdot \beta \leq 0esssupΩβ0, as a consequence of the divergence theorem we have
wH1(B3)C1μLuH1(B3)wH1B3C1μLuH1B3||w||_(H^(1)(B_(3))) <= C(1)/(mu)||Lu||_(H^(-1)(B_(3)))\|w\|_{H^{1}\left(B_{3}\right)} \leq C \frac{1}{\mu}\|\mathcal{L} u\|_{H^{-1}\left(B_{3}\right)}wH1(B3)C1μLuH1(B3)
Takingv=uwv=uwv=u-wv=u-wv=uw, we haveLv=0Lv=0Lv=0\mathcal{L} v=0Lv=0inB3B3B_(3)B_{3}B3. The stability estimate in Corollary 1 used forB0,B2,B3B0,B2,B3B_(0),B_(2),B_(3)B_{0}, B_{2}, B_{3}B0,B2,B3reads as
vH1(B2)CeCP~e2vH1(B0)κvH1(B3)1κ,vH1B2CeCP~e2vH1B0κvH1B31κ,||v||_(H^(1)(B_(2))) <= Ce^(C tilde(P)e^(2))||v||_(H^(1)(B_(0)))^(kappa)||v||_(H^(1)(B_(3)))^(1-kappa),\|v\|_{H^{1}\left(B_{2}\right)} \leq C e^{C \tilde{P} e^{2}}\|v\|_{H^{1}\left(B_{0}\right)}^{\kappa}\|v\|_{H^{1}\left(B_{3}\right)}^{1-\kappa},vH1(B2)CeCP~e2vH1(B0)κvH1(B3)1κ,
and the following estimates hold
uH1(B2)vH1(B2)+wH1(B2)CeCP~e2(uH1(B0)+1μLuH1(Ω))κ(uH1(B3)+1μLuH1(Ω))1κuH1B2vH1B2+wH1B2CeCP~e2uH1B0+1μLuH1(Ω)κuH1B3+1μLuH1(Ω)1κ{:[||u||_(H^(1)(B_(2))) <= ||v||_(H^(1)(B_(2)))+||w||_(H^(1)(B_(2)))],[ <= Ce^(C tilde(P)e^(2))(||u||_(H^(1)(B_(0)))+(1)/(mu)||Lu||_(H^(-1)(Omega)))^(kappa)(||u||_(H^(1)(B_(3)))+(1)/(mu)||Lu||_(H^(-1)(Omega)))^(1-kappa)]:}\begin{aligned} \|u\|_{H^{1}\left(B_{2}\right)} & \leq\|v\|_{H^{1}\left(B_{2}\right)}+\|w\|_{H^{1}\left(B_{2}\right)} \\ & \leq C e^{C \tilde{P} e^{2}}\left(\|u\|_{H^{1}\left(B_{0}\right)}+\frac{1}{\mu}\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right)^{\kappa}\left(\|u\|_{H^{1}\left(B_{3}\right)}+\frac{1}{\mu}\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right)^{1-\kappa} \end{aligned}uH1(B2)vH1(B2)+wH1(B2)CeCP~e2(uH1(B0)+1μLuH1(Ω))κ(uH1(B3)+1μLuH1(Ω))1κ
Now we choose a cutoff functionχC0(B1)χC0B1chi inC_(0)^(oo)(B_(1))\chi \in C_{0}^{\infty}\left(B_{1}\right)χC0(B1)such thatχ=1χ=1chi=1\chi=1χ=1inB0B0B_(0)B_{0}B0. Thenχuχuchi u\chi uχusatisfies
L(χu)=χLu+[L,χ]u,χu=0onB1,L(χu)=χLu+[L,χ]u,χu=0 on B1,L(chi u)=chiLu+[L,chi]u,quad chi u=0" on "delB_(1),\mathcal{L}(\chi u)=\chi \mathcal{L} u+[\mathcal{L}, \chi] u, \quad \chi u=0 \text { on } \partial B_{1},L(χu)=χLu+[L,χ]u,χu=0 on B1,
and we obtain
uH1(B0)χuH1(B1)C1μ([L,χ]uH1(B1)+χLuH1(B1))C1μ((μ+|β|)uL2(B1)+LuH1(Ω))uH1B0χuH1B1C1μ[L,χ]uH1B1+χLuH1B1C1μ(μ+|β|)uL2B1+LuH1(Ω){:[||u||_(H^(1)(B_(0))) <= ||chi u||_(H^(1)(B_(1))) <= C(1)/(mu)(||[L,chi]u||_(H^(-1)(B_(1)))+||chiLu||_(H^(-1)(B_(1))))],[ <= C(1)/(mu)((mu+|beta|)||u||_(L^(2)(B_(1)))+||Lu||_(H^(-1)(Omega)))]:}\begin{aligned} \|u\|_{H^{1}\left(B_{0}\right)} \leq\|\chi u\|_{H^{1}\left(B_{1}\right)} & \leq C \frac{1}{\mu}\left(\|[\mathcal{L}, \chi] u\|_{H^{-1}\left(B_{1}\right)}+\|\chi \mathcal{L} u\|_{H^{-1}\left(B_{1}\right)}\right) \\ & \leq C \frac{1}{\mu}\left((\mu+|\beta|)\|u\|_{L^{2}\left(B_{1}\right)}+\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right) \end{aligned}uH1(B0)χuH1(B1)C1μ([L,χ]uH1(B1)+χLuH1(B1))C1μ((μ+|β|)uL2(B1)+LuH1(Ω))
The same argument forB3ΩB3ΩB_(3)sub OmegaB_{3} \subset \OmegaB3Ωgives
uH1(B3)C1μ((μ+|β|)uL2(Ω)+LuH1(Ω))uH1B3C1μ(μ+|β|)uL2(Ω)+LuH1(Ω)||u||_(H^(1)(B_(3))) <= C(1)/(mu)((mu+|beta|)||u||_(L^(2)(Omega))+||Lu||_(H^(-1)(Omega)))\|u\|_{H^{1}\left(B_{3}\right)} \leq C \frac{1}{\mu}\left((\mu+|\beta|)\|u\|_{L^{2}(\Omega)}+\|\mathcal{L} u\|_{H^{-1}(\Omega)}\right)uH1(B3)C1μ((μ+|β|)uL2(Ω)+LuH1(Ω))
thus leading to the conclusion after absorbing theP~e=1+|β|/μP~e=1+|β|/μtilde(P)e=1+|beta|//mu\tilde{P} e=1+|\beta| / \muP~e=1+|β|/μfactor into the exponential factoreCP~e2eCP~e2e^(C tilde(P)e^(2))e^{C \tilde{P} e^{2}}eCP~e2.
Remark 1. In the geometric setting of this section one can be more precise about the Hölder exponentκκkappa\kappaκin the conditional stability estimates. For this we recall some known results for second-order elliptic equations: we refer to [ARRV09, Theorem 2.1] for the Laplace equation, and for the case including lower-order terms to [MV12, Theorem 3]. Letuuuuube a homogeneous solution of (1) withf0f0f-=0f \equiv 0f0. For a constantCstCst C_("st ")C_{\text {st }}Cst depending implicitly on the coefficientsμμmu\muμandββbeta\betaβ, the following three-ball inequality holds
uL2(B2)CstuL2(B1)κuL2(B3)1κuL2B2CstuL2B1κuL2B31κ||u||_(L^(2)(B_(2))) <= C_(st)||u||_(L^(2)(B_(1)))^(kappa)||u||_(L^(2)(B_(3)))^(1-kappa)\|u\|_{L^{2}\left(B_{2}\right)} \leq C_{s t}\|u\|_{L^{2}\left(B_{1}\right)}^{\kappa}\|u\|_{L^{2}\left(B_{3}\right)}^{1-\kappa}uL2(B2)CstuL2(B1)κuL2(B3)1κ
whereBjBjB_(j)B_{j}Bjare concentric balls inΩΩOmega\OmegaΩwith increasing radiirjrjr_(j)r_{j}rj. The constantCstCstC_(st)C_{s t}Cstdoes not depend on the radiir1,r2r1,r2r_(1),r_(2)r_{1}, r_{2}r1,r2, but it does depend onr3r3r_(3)r_{3}r3. The exponentκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)is given by
κ=logr3r2C3logr2r1+logr3r2κ=logr3r2C3logr2r1+logr3r2kappa=(log((r_(3))/(r_(2))))/(C_(3)log((r_(2))/(r_(1)))+log((r_(3))/(r_(2))))\kappa=\frac{\log \frac{r_{3}}{r_{2}}}{C_{3} \log \frac{r_{2}}{r_{1}}+\log \frac{r_{3}}{r_{2}}}κ=logr3r2C3logr2r1+logr3r2
whereC3>0C3>0C_(3) > 0C_{3}>0C3>0is a constant depending onr3r3r_(3)r_{3}r3.

3. Finite element method

LetVhVhV_(h)V_{h}Vhdenote the space of piecewise affine finite element functions defined on a conforming computational meshTh={K}.ThTh={K}.ThT_(h)={K}.T_(h)\mathcal{T}_{h}=\{K\} . \mathcal{T}_{h}Th={K}.Thconsists of shape regular triangular elementsKKKKKwith diameterhKhKh_(K)h_{K}hKand is quasi-uniform. We define the global mesh size byh=maxKThhKh=maxKThhKh=max_(K inT_(h))h_(K)h=\max _{K \in \mathcal{T}_{h}} h_{K}h=maxKThhK. The interior faces of the triangulation will be denoted byFiFiF_(i)\mathcal{F}_{i}Fi, the jump of a quantity across a faceFFFFFby[[]]F[[]]F[[*]]_(F)\llbracket \cdot \rrbracket_{F}[[]]F, and the outward unit normal bynnnnn.
Letβ[W1,(Ω)]nβW1,(Ω)nbeta in[W^(1,oo)(Omega)]^(n)\beta \in\left[W^{1, \infty}(\Omega)\right]^{n}β[W1,(Ω)]nand adopt the shorthand notation|β|:=β[L(Ω)]n|β|:=βL(Ω)n|beta|:=||beta||_([L^(oo)(Omega)]^(n))|\beta|:=\|\beta\|_{\left[L^{\infty}(\Omega)\right]^{n}}|β|:=β[L(Ω)]n. As already stated in Section 1, we consider the diffusion-dominated regime given by the low Péclet number
(11)Pe(h):=|β|hμ<1.(11)Pe(h):=|β|hμ<1.{:(11)Pe(h):=(|beta|h)/(mu) < 1.:}\begin{equation*} P e(h):=\frac{|\beta| h}{\mu}<1 . \tag{11} \end{equation*}(11)Pe(h):=|β|hμ<1.
We will denote byCCCCCa generic positive constant independent of the mesh size and the Péclet number. Letπh:L2(Ω)Vhπh:L2(Ω)Vhpi_(h):L^(2)(Omega)|->V_(h)\pi_{h}: L^{2}(\Omega) \mapsto V_{h}πh:L2(Ω)Vhdenote the standardL2L2L^(2)L^{2}L2-projection onVhVhV_(h)V_{h}Vh, which fork=1,2k=1,2k=1,2k=1,2k=1,2andm=0,k1m=0,k1m=0,k-1m=0, k-1m=0,k1satisfies
πhuHm(Ω)CuHm(Ω),uHm(Ω)uπhuHm(Ω)ChkmuHk(Ω),uHk(Ω)πhuHm(Ω)CuHm(Ω),uHm(Ω)uπhuHm(Ω)ChkmuHk(Ω),uHk(Ω){:[||pi_(h)u||_(H^(m)(Omega)) <= C||u||_(H^(m)(Omega))","quad u inH^(m)(Omega)],[||u-pi_(h)u||_(H^(m)(Omega)) <= Ch^(k-m)||u||_(H^(k)(Omega))","quad u inH^(k)(Omega)]:}\begin{aligned} \left\|\pi_{h} u\right\|_{H^{m}(\Omega)} & \leq C\|u\|_{H^{m}(\Omega)}, \quad u \in H^{m}(\Omega) \\ \left\|u-\pi_{h} u\right\|_{H^{m}(\Omega)} & \leq C h^{k-m}\|u\|_{H^{k}(\Omega)}, \quad u \in H^{k}(\Omega) \end{aligned}πhuHm(Ω)CuHm(Ω),uHm(Ω)uπhuHm(Ω)ChkmuHk(Ω),uHk(Ω)
We introduce the standard inner products with the induced norms
(vh,wh)Ω:=Ωvhwhdxvh,whΩ:=Ωvhwhdsvh,whΩ:=Ωvhwhdxvh,whΩ:=Ωvhwhds{:[(v_(h),w_(h))_(Omega):=int_(Omega)v_(h)w_(h)dx],[(:v_(h),w_(h):)_(del Omega):=int_(del Omega)v_(h)w_(h)ds]:}\begin{aligned} \left(v_{h}, w_{h}\right)_{\Omega} & :=\int_{\Omega} v_{h} w_{h} \mathrm{~d} x \\ \left\langle v_{h}, w_{h}\right\rangle_{\partial \Omega} & :=\int_{\partial \Omega} v_{h} w_{h} \mathrm{~d} s \end{aligned}(vh,wh)Ω:=Ωvhwh dxvh,whΩ:=Ωvhwh ds
and the following bilinear forms
ah(vh,wh):=(βvh,wh)Ω+(μvh,wh)Ωμvhn,whΩsΩ(vh,wh):=γFFiFh(μ+|β|h)[[vhn]]F[[whn]]Fdssω(vh,wh):=((μ+|β|h)vh,wh)ωs(vh,wh):=sΩ(vh,wh)+sω(vh,wh)ahvh,wh:=βvh,whΩ+μvh,whΩμvhn,whΩsΩvh,wh:=γFFiFh(μ+|β|h)[[vhn]]F[[whn]]Fdssωvh,wh:=(μ+|β|h)vh,whωsvh,wh:=sΩvh,wh+sωvh,wh{:[a_(h)(v_(h),w_(h)):=(beta*gradv_(h),w_(h))_(Omega)+(mu gradv_(h),gradw_(h))_(Omega)-(:mu gradv_(h)*n,w_(h):)_(del Omega)],[s_(Omega)(v_(h),w_(h)):=gammasum_(F inF_(i))int_(F)h(mu+|beta|h)[[gradv_(h)*n]]_(F)[[gradw_(h)*n]]_(F)ds],[s_(omega)(v_(h),w_(h)):=((mu+|beta|h)v_(h),w_(h))_(omega)],[s(v_(h),w_(h)):=s_(Omega)(v_(h),w_(h))+s_(omega)(v_(h),w_(h))]:}\begin{gathered} a_{h}\left(v_{h}, w_{h}\right):=\left(\beta \cdot \nabla v_{h}, w_{h}\right)_{\Omega}+\left(\mu \nabla v_{h}, \nabla w_{h}\right)_{\Omega}-\left\langle\mu \nabla v_{h} \cdot n, w_{h}\right\rangle_{\partial \Omega} \\ s_{\Omega}\left(v_{h}, w_{h}\right):=\gamma \sum_{F \in \mathcal{F}_{i}} \int_{F} h(\mu+|\beta| h) \llbracket \nabla v_{h} \cdot n \rrbracket_{F} \llbracket \nabla w_{h} \cdot n \rrbracket_{F} \mathrm{~d} s \\ s_{\omega}\left(v_{h}, w_{h}\right):=\left((\mu+|\beta| h) v_{h}, w_{h}\right)_{\omega} \\ s\left(v_{h}, w_{h}\right):=s_{\Omega}\left(v_{h}, w_{h}\right)+s_{\omega}\left(v_{h}, w_{h}\right) \end{gathered}ah(vh,wh):=(βvh,wh)Ω+(μvh,wh)Ωμvhn,whΩsΩ(vh,wh):=γFFiFh(μ+|β|h)[[vhn]]F[[whn]]F dssω(vh,wh):=((μ+|β|h)vh,wh)ωs(vh,wh):=sΩ(vh,wh)+sω(vh,wh)
and
s(vh,wh):=γ((μh1+|β|)vh,whΩ+(μvh,wh)Ω+sΩ(vh,wh))svh,wh:=γμh1+|β|vh,whΩ+μvh,whΩ+sΩvh,whs_(**)(v_(h),w_(h)):=gamma_(**)((:(muh^(-1)+|beta|)v_(h),w_(h):)_(del Omega)+(mu gradv_(h),gradw_(h))_(Omega)+s_(Omega)(v_(h),w_(h)))s_{*}\left(v_{h}, w_{h}\right):=\gamma_{*}\left(\left\langle\left(\mu h^{-1}+|\beta|\right) v_{h}, w_{h}\right\rangle_{\partial \Omega}+\left(\mu \nabla v_{h}, \nabla w_{h}\right)_{\Omega}+s_{\Omega}\left(v_{h}, w_{h}\right)\right)s(vh,wh):=γ((μh1+|β|)vh,whΩ+(μvh,wh)Ω+sΩ(vh,wh))
The termssΩsΩs_(Omega)s_{\Omega}sΩandsss_(**)s_{*}sare stabilizing terms, while the termsωsωs_(omega)s_{\omega}sωis aimed for data assimilation. After scaling with the coefficients in the above forms, Lemma 2 in [BHL18] writes as
(12)(μ12h+|β|12h32)vhH1(Ω)Cγ12s(vh,vh)12,vhVh(12)μ12h+|β|12h32vhH1(Ω)Cγ12svh,vh12,vhVh{:(12)||(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))v_(h)||_(H^(1)(Omega)) <= Cgamma^(-(1)/(2))s(v_(h),v_(h))^((1)/(2))","quad AAv_(h)inV_(h):}\begin{equation*} \left\|\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right) v_{h}\right\|_{H^{1}(\Omega)} \leq C \gamma^{-\frac{1}{2}} s\left(v_{h}, v_{h}\right)^{\frac{1}{2}}, \quad \forall v_{h} \in V_{h} \tag{12} \end{equation*}(12)(μ12h+|β|12h32)vhH1(Ω)Cγ12s(vh,vh)12,vhVh
and Lemma 2 in [BO18] gives the jump inequality
(13)sΩ(πhu,πhu)Cγ(μ+|β|h)h2|u|H2(Ω)2,uH2(Ω).(13)sΩπhu,πhuCγ(μ+|β|h)h2|u|H2(Ω)2,uH2(Ω).{:(13)s_(Omega)(pi_(h)u,pi_(h)u) <= C gamma(mu+|beta|h)h^(2)|u|_(H^(2)(Omega))^(2)","quad AA u inH^(2)(Omega).:}\begin{equation*} s_{\Omega}\left(\pi_{h} u, \pi_{h} u\right) \leq C \gamma(\mu+|\beta| h) h^{2}|u|_{H^{2}(\Omega)}^{2}, \quad \forall u \in H^{2}(\Omega) . \tag{13} \end{equation*}(13)sΩ(πhu,πhu)Cγ(μ+|β|h)h2|u|H2(Ω)2,uH2(Ω).
The parametersγγgamma\gammaγandγγgamma_(**)\gamma_{*}γinsΩsΩs_(Omega)s_{\Omega}sΩandsss_(**)s_{*}s, respectively, are fixed at the implementation level and, to alleviate notation, our analysis covers the choiceγ=1=γγ=1=γgamma=1=gamma_(**)\gamma=1=\gamma_{*}γ=1=γ.
We can then use the general framework in [Bur13] to write the finite element method for unique continuation subject to (1) as follows. Consider a discrete Lagrange multiplierzhVhzhVhz_(h)inV_(h)z_{h} \in V_{h}zhVhand aim to find the saddle points of the functional
Lh(uh,zh):=12sω(uhU~ω,uhU~ω)+ah(uh,zh)(f,zh)Ω+12sΩ(uh,uh)12s(zh,zh)Lhuh,zh:=12sωuhU~ω,uhU~ω+ahuh,zhf,zhΩ+12sΩuh,uh12szh,zh{:[L_(h)(u_(h),z_(h)):=(1)/(2)s_(omega)(u_(h)- tilde(U)_(omega),u_(h)- tilde(U)_(omega))+a_(h)(u_(h),z_(h))-(f,z_(h))_(Omega)],[+(1)/(2)s_(Omega)(u_(h),u_(h))-(1)/(2)s_(**)(z_(h),z_(h))]:}\begin{aligned} L_{h}\left(u_{h}, z_{h}\right): & =\frac{1}{2} s_{\omega}\left(u_{h}-\tilde{U}_{\omega}, u_{h}-\tilde{U}_{\omega}\right)+a_{h}\left(u_{h}, z_{h}\right)-\left(f, z_{h}\right)_{\Omega} \\ & +\frac{1}{2} s_{\Omega}\left(u_{h}, u_{h}\right)-\frac{1}{2} s_{*}\left(z_{h}, z_{h}\right) \end{aligned}Lh(uh,zh):=12sω(uhU~ω,uhU~ω)+ah(uh,zh)(f,zh)Ω+12sΩ(uh,uh)12s(zh,zh)
where we recall thatU~ω=u|ω+δU~ω=uω+δtilde(U)_(omega)=u|_(omega)+delta\tilde{U}_{\omega}=\left.u\right|_{\omega}+\deltaU~ω=u|ω+δanduH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)is a solution to (1). The Euler-Lagrange equations forLhLhL_(h)L_{h}Lhlead to the following discrete problem: find(uh,zh)[Vh]2uh,zhVh2(u_(h),z_(h))in[V_(h)]^(2)\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}(uh,zh)[Vh]2such that
(14){ah(uh,wh)s(zh,wh)=(f,wh)Ωah(vh,zh)+s(uh,vh)=sω(U~ω,vh),(vh,wh)[Vh]2(14)ahuh,whszh,wh=f,whΩahvh,zh+suh,vh=sωU~ω,vh,vh,whVh2{:(14){[a_(h)(u_(h),w_(h))-s_(**)(z_(h),w_(h)),=(f,w_(h))_(Omega)],[a_(h)(v_(h),z_(h))+s(u_(h),v_(h)),=s_(omega)( tilde(U)_(omega),v_(h))],quad AA(v_(h),w_(h))in[V_(h)]^(2):}:}\left\{\begin{array}{rl} a_{h}\left(u_{h}, w_{h}\right)-s_{*}\left(z_{h}, w_{h}\right) & =\left(f, w_{h}\right)_{\Omega} \tag{14}\\ a_{h}\left(v_{h}, z_{h}\right)+s\left(u_{h}, v_{h}\right) & =s_{\omega}\left(\tilde{U}_{\omega}, v_{h}\right) \end{array}, \quad \forall\left(v_{h}, w_{h}\right) \in\left[V_{h}\right]^{2}\right.(14){ah(uh,wh)s(zh,wh)=(f,wh)Ωah(vh,zh)+s(uh,vh)=sω(U~ω,vh),(vh,wh)[Vh]2
We observe that by the ill-posed character of the problem, only the stabilization operatorssΩsΩs_(Omega)s_{\Omega}sΩandsss_(**)s_{*}sprovide some stability to the discrete system, and the corresponding system matrix is expected to be ill-conditioned. To quantify this effect we first prove an upper bound on the condition number.
Proposition 2. The finite element formulation (14) has a unique solution(uh,zh)[Vh]2uh,zhVh2(u_(h),z_(h))in[V_(h)]^(2)\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}(uh,zh)[Vh]2and the Euclidean condition numberK2K2K_(2)\mathcal{K}_{2}K2of the system matrix satisfies
K2Ch4.K2Ch4.K_(2) <= Ch^(-4).\mathcal{K}_{2} \leq C h^{-4} .K2Ch4.
Proof. We write (14) as the linear systemA[(uh,zh),(vh,wh)]=(f,wh)Ω+sω(U~ω,vh)Auh,zh,vh,wh=f,whΩ+sωU~ω,vhA[(u_(h),z_(h)),(v_(h),w_(h))]=(f,w_(h))_(Omega)+s_(omega)( tilde(U)_(omega),v_(h))A\left[\left(u_{h}, z_{h}\right),\left(v_{h}, w_{h}\right)\right]=\left(f, w_{h}\right)_{\Omega}+s_{\omega}\left(\tilde{U}_{\omega}, v_{h}\right)A[(uh,zh),(vh,wh)]=(f,wh)Ω+sω(U~ω,vh), for all(vh,wh)[Vh]2vh,whVh2(v_(h),w_(h))in[V_(h)]^(2)\left(v_{h}, w_{h}\right) \in\left[V_{h}\right]^{2}(vh,wh)[Vh]2, where
A[(uh,zh),(vh,wh)]:=ah(uh,wh)s(zh,wh)+ah(vh,zh)+s(uh,vh).Auh,zh,vh,wh:=ahuh,whszh,wh+ahvh,zh+suh,vh.A[(u_(h),z_(h)),(v_(h),w_(h))]:=a_(h)(u_(h),w_(h))-s_(**)(z_(h),w_(h))+a_(h)(v_(h),z_(h))+s(u_(h),v_(h)).A\left[\left(u_{h}, z_{h}\right),\left(v_{h}, w_{h}\right)\right]:=a_{h}\left(u_{h}, w_{h}\right)-s_{*}\left(z_{h}, w_{h}\right)+a_{h}\left(v_{h}, z_{h}\right)+s\left(u_{h}, v_{h}\right) .A[(uh,zh),(vh,wh)]:=ah(uh,wh)s(zh,wh)+ah(vh,zh)+s(uh,vh).
SinceA[(uh,zh),(uh,zh)]=s(uh,uh)+s(zh,zh)Auh,zh,uh,zh=suh,uh+szh,zhA[(u_(h),z_(h)),(u_(h),-z_(h))]=s(u_(h),u_(h))+s_(**)(z_(h),z_(h))A\left[\left(u_{h}, z_{h}\right),\left(u_{h},-z_{h}\right)\right]=s\left(u_{h}, u_{h}\right)+s_{*}\left(z_{h}, z_{h}\right)A[(uh,zh),(uh,zh)]=s(uh,uh)+s(zh,zh), using (12) the following inf-sup condition holds
Ψh:=inf(uh,zh)[Vh]2sup(vh,wh)[Vh]2A[(uh,zh),(vh,wh)](uh,zh)L2(Ω)(vh,wh)L2(Ω)Cμ(1+Pe(h))h2.Ψh:=infuh,zhVh2supvh,whVh2Auh,zh,vh,whuh,zhL2(Ω)vh,whL2(Ω)Cμ(1+Pe(h))h2.Psi_(h):=i n f_((u_(h),z_(h))in[V_(h)]^(2))s u p_((v_(h),w_(h))in[V_(h)]^(2))(A[(u_(h),z_(h)),(v_(h),w_(h))])/(||(u_(h),z_(h))||_(L^(2)(Omega))||(v_(h),w_(h))||_(L^(2)(Omega))) >= C mu(1+Pe(h))h^(2).\Psi_{h}:=\inf _{\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}} \sup _{\left(v_{h}, w_{h}\right) \in\left[V_{h}\right]^{2}} \frac{A\left[\left(u_{h}, z_{h}\right),\left(v_{h}, w_{h}\right)\right]}{\left\|\left(u_{h}, z_{h}\right)\right\|_{L^{2}(\Omega)}\left\|\left(v_{h}, w_{h}\right)\right\|_{L^{2}(\Omega)}} \geq C \mu(1+P e(h)) h^{2} .Ψh:=inf(uh,zh)[Vh]2sup(vh,wh)[Vh]2A[(uh,zh),(vh,wh)](uh,zh)L2(Ω)(vh,wh)L2(Ω)Cμ(1+Pe(h))h2.
This provides the existence of a unique solution for the linear system. We use [EG06, Theorem 3.1] to estimate the condition number by
(15)K2CΥhΨh,(15)K2CΥhΨh,{:(15)K_(2) <= C(Υ_(h))/(Psi_(h))",":}\begin{equation*} \mathcal{K}_{2} \leq C \frac{\Upsilon_{h}}{\Psi_{h}}, \tag{15} \end{equation*}(15)K2CΥhΨh,
where
Υh:=sup(uh,zh)[Vh]2sup(vh,wh)[Vh]2A[(uh,zh),(vh,wh)](uh,zh)L2(Ω)(vh,wh)L2(Ω).Υh:=supuh,zhVh2supvh,whVh2Auh,zh,vh,whuh,zhL2(Ω)vh,whL2(Ω).Υ_(h):=s u p_((u_(h),z_(h))in[V_(h)]^(2))s u p_((v_(h),w_(h))in[V_(h)]^(2))(A[(u_(h),z_(h)),(v_(h),w_(h))])/(||(u_(h),z_(h))||_(L^(2)(Omega))||(v_(h),w_(h))||_(L^(2)(Omega))).\Upsilon_{h}:=\sup _{\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}} \sup _{\left(v_{h}, w_{h}\right) \in\left[V_{h}\right]^{2}} \frac{A\left[\left(u_{h}, z_{h}\right),\left(v_{h}, w_{h}\right)\right]}{\left\|\left(u_{h}, z_{h}\right)\right\|_{L^{2}(\Omega)}\left\|\left(v_{h}, w_{h}\right)\right\|_{L^{2}(\Omega)}} .Υh:=sup(uh,zh)[Vh]2sup(vh,wh)[Vh]2A[(uh,zh),(vh,wh)](uh,zh)L2(Ω)(vh,wh)L2(Ω).
We recall the following discrete inverse inequality, see for instance [EG04, Lemma 1.138],
(16)vhL2(K)Ch1vhL2(K),vhP1(K).(16)vhL2(K)Ch1vhL2(K),vhP1(K).{:(16)||gradv_(h)||_(L^(2)(K)) <= Ch^(-1)||v_(h)||_(L^(2)(K))","quad AAv_(h)inP_(1)(K).:}\begin{equation*} \left\|\nabla v_{h}\right\|_{L^{2}(K)} \leq C h^{-1}\left\|v_{h}\right\|_{L^{2}(K)}, \quad \forall v_{h} \in \mathbb{P}_{1}(K) . \tag{16} \end{equation*}(16)vhL2(K)Ch1vhL2(K),vhP1(K).
We also recall the following continuous trace inequality, see for instance [MS99],
(17)vL2(K)C(h12vL2(K)+h12vL2(K)),vH1(K),(17)vL2(K)Ch12vL2(K)+h12vL2(K),vH1(K),{:(17)||v||_(L^(2)(del K)) <= C(h^(-(1)/(2))||v||_(L^(2)(K))+h^((1)/(2))||grad v||_(L^(2)(K)))","quad AA v inH^(1)(K)",":}\begin{equation*} \|v\|_{L^{2}(\partial K)} \leq C\left(h^{-\frac{1}{2}}\|v\|_{L^{2}(K)}+h^{\frac{1}{2}}\|\nabla v\|_{L^{2}(K)}\right), \quad \forall v \in H^{1}(K), \tag{17} \end{equation*}(17)vL2(K)C(h12vL2(K)+h12vL2(K)),vH1(K),
and the discrete one
(18)vhnL2(K)Ch12vhL2(K),vhP1(K).(18)vhnL2(K)Ch12vhL2(K),vhP1(K).{:(18)||gradv_(h)*n||_(L^(2)(del K)) <= Ch^(-(1)/(2))||gradv_(h)||_(L^(2)(K))","quad AAv_(h)inP_(1)(K).:}\begin{equation*} \left\|\nabla v_{h} \cdot n\right\|_{L^{2}(\partial K)} \leq C h^{-\frac{1}{2}}\left\|\nabla v_{h}\right\|_{L^{2}(K)}, \quad \forall v_{h} \in \mathbb{P}_{1}(K) . \tag{18} \end{equation*}(18)vhnL2(K)Ch12vhL2(K),vhP1(K).
Using the Cauchy-Schwarz inequality together with (18) and (16) we get
sΩ(uh,vh)=γμ(1+Pe(h))FFiFh[[uhn]]F[[vhn]]FdsCμ(1+Pe(h))h2uhL2(Ω)vhL2(Ω),sΩuh,vh=γμ(1+Pe(h))FFiFh[[uhn]]F[[vhn]]FdsCμ(1+Pe(h))h2uhL2(Ω)vhL2(Ω),{:[s_(Omega)(u_(h),v_(h))=gamma mu(1+Pe(h))sum_(F inF_(i))int_(F)h[[gradu_(h)*n]]_(F)[[gradv_(h)*n]]_(F)ds],[ <= C mu(1+Pe(h))h^(-2)||u_(h)||_(L^(2)(Omega))||v_(h)||_(L^(2)(Omega))","]:}\begin{aligned} s_{\Omega}\left(u_{h}, v_{h}\right) & =\gamma \mu(1+P e(h)) \sum_{F \in \mathcal{F}_{i}} \int_{F} h \llbracket \nabla u_{h} \cdot n \rrbracket_{F} \llbracket \nabla v_{h} \cdot n \rrbracket_{F} \mathrm{~d} s \\ & \leq C \mu(1+P e(h)) h^{-2}\left\|u_{h}\right\|_{L^{2}(\Omega)}\left\|v_{h}\right\|_{L^{2}(\Omega)}, \end{aligned}sΩ(uh,vh)=γμ(1+Pe(h))FFiFh[[uhn]]F[[vhn]]F dsCμ(1+Pe(h))h2uhL2(Ω)vhL2(Ω),
hence
s(uh,vh)Cμ(1+Pe(h))h2uhL2(Ω)vhL2(Ω).suh,vhCμ(1+Pe(h))h2uhL2(Ω)vhL2(Ω).s(u_(h),v_(h)) <= C mu(1+Pe(h))h^(-2)||u_(h)||_(L^(2)(Omega))||v_(h)||_(L^(2)(Omega)).s\left(u_{h}, v_{h}\right) \leq C \mu(1+P e(h)) h^{-2}\left\|u_{h}\right\|_{L^{2}(\Omega)}\left\|v_{h}\right\|_{L^{2}(\Omega)} .s(uh,vh)Cμ(1+Pe(h))h2uhL2(Ω)vhL2(Ω).
Combining this with the Cauchy-Schwarz inequality and the inequalities (16) and (17), we obtain
s(zh,wh)Cμ(1+Pe(h))h2zhL2(Ω)whL2(Ω).szh,whCμ(1+Pe(h))h2zhL2(Ω)whL2(Ω).-s_(**)(z_(h),w_(h)) <= C mu(1+Pe(h))h^(-2)||z_(h)||_(L^(2)(Omega))||w_(h)||_(L^(2)(Omega)).-s_{*}\left(z_{h}, w_{h}\right) \leq C \mu(1+P e(h)) h^{-2}\left\|z_{h}\right\|_{L^{2}(\Omega)}\left\|w_{h}\right\|_{L^{2}(\Omega)} .s(zh,wh)Cμ(1+Pe(h))h2zhL2(Ω)whL2(Ω).
Again due to the Cauchy-Schwarz inequality, and trace and inverse inequalities, we have
ah(uh,wh)=(βuh,wh)Ω+μFFiFh[[uhn]]FwhdsCμ(1+Pe(h))h2uhL2(Ω)whL2(Ω)ahuh,wh=βuh,whΩ+μFFiFh[[uhn]]FwhdsCμ(1+Pe(h))h2uhL2(Ω)whL2(Ω){:[a_(h)(u_(h),w_(h))=(beta*gradu_(h),w_(h))_(Omega)+musum_(F inF_(i))int_(F)h[[gradu_(h)*n]]_(F)w_(h)ds],[ <= C mu(1+Pe(h))h^(-2)||u_(h)||_(L^(2)(Omega))||w_(h)||_(L^(2)(Omega))]:}\begin{aligned} a_{h}\left(u_{h}, w_{h}\right) & =\left(\beta \cdot \nabla u_{h}, w_{h}\right)_{\Omega}+\mu \sum_{F \in \mathcal{F}_{i}} \int_{F} h \llbracket \nabla u_{h} \cdot n \rrbracket_{F} w_{h} \mathrm{~d} s \\ & \leq C \mu(1+P e(h)) h^{-2}\left\|u_{h}\right\|_{L^{2}(\Omega)}\left\|w_{h}\right\|_{L^{2}(\Omega)} \end{aligned}ah(uh,wh)=(βuh,wh)Ω+μFFiFh[[uhn]]Fwh dsCμ(1+Pe(h))h2uhL2(Ω)whL2(Ω)
Collecting the above estimates we haveΥhCμ(1+Pe(h))h2ΥhCμ(1+Pe(h))h2Υ_(h) <= C mu(1+Pe(h))h^(-2)\Upsilon_{h} \leq C \mu(1+P e(h)) h^{-2}ΥhCμ(1+Pe(h))h2, and we conclude by (15).
3.1. Error estimates for the weakly consistent regularization. The error analysis proceeds in two main steps:
(i) First we prove that the stabilizing terms and the data fitting term must vanish at an optimal rate for smooth solutions, with constant independent of the physical stability (Proposition 3).
(ii) Then we show that the residual of the PDE is bounded by the stabilizing terms and the data fitting term. Using this result together with the first step and the continuous stability estimates in Section 2, we proveL2L2L^(2)L^{2}L2- andH1H1H^(1)H^{1}H1-convergence results (Theorems 1 and 2).
To quantify stabilization and data fitting for(vh,wh)[Vh]2vh,whVh2(v_(h),w_(h))in[V_(h)]^(2)\left(v_{h}, w_{h}\right) \in\left[V_{h}\right]^{2}(vh,wh)[Vh]2we introduce the norm
(vh,wh)s2:=s(vh,vh)+s(wh,wh)vh,whs2:=svh,vh+swh,wh||(v_(h),w_(h))||_(s)^(2):=s(v_(h),v_(h))+s_(**)(w_(h),w_(h))\left\|\left(v_{h}, w_{h}\right)\right\|_{s}^{2}:=s\left(v_{h}, v_{h}\right)+s_{*}\left(w_{h}, w_{h}\right)(vh,wh)s2:=s(vh,vh)+s(wh,wh)
We also define the "continuity norm" onH32+ϵ(Ω)H32+ϵ(Ω)H^((3)/(2)+epsilon)(Omega)H^{\frac{3}{2}+\epsilon}(\Omega)H32+ϵ(Ω), for anyϵ>0ϵ>0epsilon > 0\epsilon>0ϵ>0,
v:=|β|12h12vΩ+μ12vΩ+μ12h12vnΩ.v:=|β|12h12vΩ+μ12vΩ+μ12h12vnΩ.||v||_(♯):=|||beta|^((1)/(2))h^(-(1)/(2))v||_(Omega)+||mu^((1)/(2))grad v||_(Omega)+||mu^((1)/(2))h^((1)/(2))grad v*n||_(del Omega).\|v\|_{\sharp}:=\left\||\beta|^{\frac{1}{2}} h^{-\frac{1}{2}} v\right\|_{\Omega}+\left\|\mu^{\frac{1}{2}} \nabla v\right\|_{\Omega}+\left\|\mu^{\frac{1}{2}} h^{\frac{1}{2}} \nabla v \cdot n\right\|_{\partial \Omega} .v:=|β|12h12vΩ+μ12vΩ+μ12h12vnΩ.
Using standard approximation properties and the trace inequality (17), we have
uπhuC(μ12h+|β|12h32)|u|H2(Ω).uπhuCμ12h+|β|12h32|u|H2(Ω).||u-pi_(h)u||_(♯) <= C(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))|u|_(H^(2)(Omega)).\left\|u-\pi_{h} u\right\|_{\sharp} \leq C\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right)|u|_{H^{2}(\Omega)} .uπhuC(μ12h+|β|12h32)|u|H2(Ω).
Using (13) and interpolation
(uπhu,0)s2=s(uπhu,uπhu)=sΩ(πhu,πhu)+sω(uπhu,uπhu)C(μh2+|β|h3)|u|H2(Ω)2uπhu,0s2=suπhu,uπhu=sΩπhu,πhu+sωuπhu,uπhuCμh2+|β|h3|u|H2(Ω)2{:[||(u-pi_(h)u,0)||_(s)^(2)=s(u-pi_(h)u,u-pi_(h)u)=s_(Omega)(pi_(h)u,pi_(h)u)+s_(omega)(u-pi_(h)u,u-pi_(h)u)],[ <= C(muh^(2)+|beta|h^(3))|u|_(H^(2)(Omega))^(2)]:}\begin{aligned} \left\|\left(u-\pi_{h} u, 0\right)\right\|_{s}^{2} & =s\left(u-\pi_{h} u, u-\pi_{h} u\right)=s_{\Omega}\left(\pi_{h} u, \pi_{h} u\right)+s_{\omega}\left(u-\pi_{h} u, u-\pi_{h} u\right) \\ & \leq C\left(\mu h^{2}+|\beta| h^{3}\right)|u|_{H^{2}(\Omega)}^{2} \end{aligned}(uπhu,0)s2=s(uπhu,uπhu)=sΩ(πhu,πhu)+sω(uπhu,uπhu)C(μh2+|β|h3)|u|H2(Ω)2
where we used thatsΩ(u,vh)=0sΩu,vh=0s_(Omega)(u,v_(h))=0s_{\Omega}\left(u, v_{h}\right)=0sΩ(u,vh)=0, sinceuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω). Hence it follows that foruH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)
(19)(uπhu,0)s+uπhuC(μ12h+|β|12h32)|u|H2(Ω).(19)uπhu,0s+uπhuCμ12h+|β|12h32|u|H2(Ω).{:(19)||(u-pi_(h)u,0)||_(s)+||u-pi_(h)u||_(♯) <= C(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))|u|_(H^(2)(Omega)).:}\begin{equation*} \left\|\left(u-\pi_{h} u, 0\right)\right\|_{s}+\left\|u-\pi_{h} u\right\|_{\sharp} \leq C\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right)|u|_{H^{2}(\Omega)} . \tag{19} \end{equation*}(19)(uπhu,0)s+uπhuC(μ12h+|β|12h32)|u|H2(Ω).
Observe that, whenPe(h)<1Pe(h)<1Pe(h) < 1P e(h)<1Pe(h)<1, the first term dominates and the estimate isO(h)O(h)O(h)O(h)O(h), whereas whenPe(h)>1Pe(h)>1Pe(h) > 1P e(h)>1Pe(h)>1the bound isO(h32)Oh32O(h^((3)/(2)))O\left(h^{\frac{3}{2}}\right)O(h32). We note in passing that the same estimates hold for the nodal interpolant.
Lemma 3 (Consistency). LetuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)be a solution to (1) and (uh,zh)[Vh]2uh,zhVh2{:u_(h),z_(h))in[V_(h)]^(2)\left.u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}uh,zh)[Vh]2the solution to (14), then
ah(πhuuh,wh)+s(zh,wh)=ah(πhuu,wh)ahπhuuh,wh+szh,wh=ahπhuu,wha_(h)(pi_(h)u-u_(h),w_(h))+s_(**)(z_(h),w_(h))=a_(h)(pi_(h)u-u,w_(h))a_{h}\left(\pi_{h} u-u_{h}, w_{h}\right)+s_{*}\left(z_{h}, w_{h}\right)=a_{h}\left(\pi_{h} u-u, w_{h}\right)ah(πhuuh,wh)+s(zh,wh)=ah(πhuu,wh)
and
ah(vh,zh)+s(πhuuh,vh)=sΩ(πhuu,vh)+sω(πhuU~ω,vh),ahvh,zh+sπhuuh,vh=sΩπhuu,vh+sωπhuU~ω,vh,-a_(h)(v_(h),z_(h))+s(pi_(h)u-u_(h),v_(h))=s_(Omega)(pi_(h)u-u,v_(h))+s_(omega)(pi_(h)u- tilde(U)_(omega),v_(h)),-a_{h}\left(v_{h}, z_{h}\right)+s\left(\pi_{h} u-u_{h}, v_{h}\right)=s_{\Omega}\left(\pi_{h} u-u, v_{h}\right)+s_{\omega}\left(\pi_{h} u-\tilde{U}_{\omega}, v_{h}\right),ah(vh,zh)+s(πhuuh,vh)=sΩ(πhuu,vh)+sω(πhuU~ω,vh),
for all(vh,wh)[Vh]2vh,whVh2(v_(h),w_(h))in[V_(h)]^(2)\left(v_{h}, w_{h}\right) \in\left[V_{h}\right]^{2}(vh,wh)[Vh]2.
Proof. The first claim follows from the definition ofahaha_(h)a_{h}ah, since
ah(uh,wh)s(zh,wh)=(f,wh)Ω=(βuμΔu,wh)Ω=ah(u,wh)ahuh,whszh,wh=f,whΩ=βuμΔu,whΩ=ahu,wha_(h)(u_(h),w_(h))-s_(**)(z_(h),w_(h))=(f,w_(h))_(Omega)=(beta*grad u-mu Delta u,w_(h))_(Omega)=a_(h)(u,w_(h))a_{h}\left(u_{h}, w_{h}\right)-s_{*}\left(z_{h}, w_{h}\right)=\left(f, w_{h}\right)_{\Omega}=\left(\beta \cdot \nabla u-\mu \Delta u, w_{h}\right)_{\Omega}=a_{h}\left(u, w_{h}\right)ah(uh,wh)s(zh,wh)=(f,wh)Ω=(βuμΔu,wh)Ω=ah(u,wh)
where in the last equality we integrated by parts. The second claim follows similarly from
ah(vh,zh)+s(uh,vh)=sω(U~ω,vh)ahvh,zh+suh,vh=sωU~ω,vha_(h)(v_(h),z_(h))+s(u_(h),v_(h))=s_(omega)( tilde(U)_(omega),v_(h))a_{h}\left(v_{h}, z_{h}\right)+s\left(u_{h}, v_{h}\right)=s_{\omega}\left(\tilde{U}_{\omega}, v_{h}\right)ah(vh,zh)+s(uh,vh)=sω(U~ω,vh)
leading to
ah(vh,zh)+s(πhuuh,vh)=s(πhu,vh)sω(U~ω,vh)=sΩ(πhuu,vh)+sω(πhuU~ω,vh)ahvh,zh+sπhuuh,vh=sπhu,vhsωU~ω,vh=sΩπhuu,vh+sωπhuU~ω,vh{:[-a_(h)(v_(h),z_(h))+s(pi_(h)u-u_(h),v_(h))=s(pi_(h)u,v_(h))-s_(omega)( tilde(U)_(omega),v_(h))],[=s_(Omega)(pi_(h)u-u,v_(h))+s_(omega)(pi_(h)u- tilde(U)_(omega),v_(h))]:}\begin{aligned} -a_{h}\left(v_{h}, z_{h}\right)+s\left(\pi_{h} u-u_{h}, v_{h}\right) & =s\left(\pi_{h} u, v_{h}\right)-s_{\omega}\left(\tilde{U}_{\omega}, v_{h}\right) \\ & =s_{\Omega}\left(\pi_{h} u-u, v_{h}\right)+s_{\omega}\left(\pi_{h} u-\tilde{U}_{\omega}, v_{h}\right) \end{aligned}ah(vh,zh)+s(πhuuh,vh)=s(πhu,vh)sω(U~ω,vh)=sΩ(πhuu,vh)+sω(πhuU~ω,vh)
Lemma 4 (Continuity). Assume the low Péclet regime (11) and that|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|. LetvH2(Ω)vH2(Ω)v inH^(2)(Omega)v \in H^{2}(\Omega)vH2(Ω)andwhVhwhVhw_(h)inV_(h)w_{h} \in V_{h}whVh, then
ah(v,wh)Cv(0,wh)s.ahv,whCv0,whs.a_(h)(v,w_(h)) <= C||v||_(♯)||(0,w_(h))||_(s).a_{h}\left(v, w_{h}\right) \leq C\|v\|_{\sharp}\left\|\left(0, w_{h}\right)\right\|_{s} .ah(v,wh)Cv(0,wh)s.
Proof. Writing out the terms ofahaha_(h)a_{h}ahand integrating by parts in the advective term leads to
ah(v,wh)=(v,βwh)Ω(vβ,wh)Ω+vβn,whΩ+(μv,wh)Ωμvn,whΩahv,wh=v,βwhΩvβ,whΩ+vβn,whΩ+μv,whΩμvn,whΩa_(h)(v,w_(h))=-(v,beta*gradw_(h))_(Omega)-(v grad*beta,w_(h))_(Omega)+(:v beta*n,w_(h):)_(del Omega)+(mu grad v,gradw_(h))_(Omega)-(:mu grad v*n,w_(h):)_(del Omega)a_{h}\left(v, w_{h}\right)=-\left(v, \beta \cdot \nabla w_{h}\right)_{\Omega}-\left(v \nabla \cdot \beta, w_{h}\right)_{\Omega}+\left\langle v \beta \cdot n, w_{h}\right\rangle_{\partial \Omega}+\left(\mu \nabla v, \nabla w_{h}\right)_{\Omega}-\left\langle\mu \nabla v \cdot n, w_{h}\right\rangle_{\partial \Omega}ah(v,wh)=(v,βwh)Ω(vβ,wh)Ω+vβn,whΩ+(μv,wh)Ωμvn,whΩ.
Using the Cauchy-Schwarz inequality and the trace inequality (17) forvvvvv, we see that
vβn,whΩ+(μv,wh)Ωμvn,whΩCv(0,wh)s.vβn,whΩ+μv,whΩμvn,whΩCv0,whs.(:v beta*n,w_(h):)_(del Omega)+(mu grad v,gradw_(h))_(Omega)-(:mu grad v*n,w_(h):)_(del Omega) <= C||v||_(♯)||(0,w_(h))||_(s).\left\langle v \beta \cdot n, w_{h}\right\rangle_{\partial \Omega}+\left(\mu \nabla v, \nabla w_{h}\right)_{\Omega}-\left\langle\mu \nabla v \cdot n, w_{h}\right\rangle_{\partial \Omega} \leq C\|v\|_{\sharp}\left\|\left(0, w_{h}\right)\right\|_{s} .vβn,whΩ+(μv,wh)Ωμvn,whΩCv(0,wh)s.
By the Cauchy-Schwarz inequality and a discrete Poincaré inequality forwhwhw_(h)w_{h}wh, see e.g. [Bre03], we bound
(vβ,wh)ΩC|β|1,vΩwhΩC|β|1,|β|Pe(h)12v(0,wh)svβ,whΩC|β|1,vΩwhΩC|β|1,|β|Pe(h)12v0,whs-(v grad*beta,w_(h))_(Omega) <= C|beta|_(1,oo)||v||_(Omega)||w_(h)||_(Omega) <= C(|beta|_(1,oo))/(|beta|)Pe(h)^((1)/(2))||v||_(♯)||(0,w_(h))||_(s)-\left(v \nabla \cdot \beta, w_{h}\right)_{\Omega} \leq C|\beta|_{1, \infty}\|v\|_{\Omega}\left\|w_{h}\right\|_{\Omega} \leq C \frac{|\beta|_{1, \infty}}{|\beta|} P e(h)^{\frac{1}{2}}\|v\|_{\sharp}\left\|\left(0, w_{h}\right)\right\|_{s}(vβ,wh)ΩC|β|1,vΩwhΩC|β|1,|β|Pe(h)12v(0,wh)s
Under the assumption|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|, we get
(vβ,wh)ΩCPe(h)12v(0,wh)s.vβ,whΩCPe(h)12v0,whs.-(v grad*beta,w_(h))_(Omega) <= CPe(h)^((1)/(2))||v||_(♯)||(0,w_(h))||_(s).-\left(v \nabla \cdot \beta, w_{h}\right)_{\Omega} \leq C P e(h)^{\frac{1}{2}}\|v\|_{\sharp}\left\|\left(0, w_{h}\right)\right\|_{s} .(vβ,wh)ΩCPe(h)12v(0,wh)s.
We bound the remaining term by
(v,βwh)Ω|β|12h12vwhΩCPe(h)12v(0,wh)s.v,βwhΩ|β|12h12vwhΩCPe(h)12v0,whs.-(v,beta*gradw_(h))_(Omega) <= |beta|^((1)/(2))h^((1)/(2))||v||_(♯)||gradw_(h)||_(Omega) <= CPe(h)^((1)/(2))||v||_(♯)||(0,w_(h))||_(s).-\left(v, \beta \cdot \nabla w_{h}\right)_{\Omega} \leq|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\|v\|_{\sharp}\left\|\nabla w_{h}\right\|_{\Omega} \leq C P e(h)^{\frac{1}{2}}\|v\|_{\sharp}\left\|\left(0, w_{h}\right)\right\|_{s} .(v,βwh)Ω|β|12h12vwhΩCPe(h)12v(0,wh)s.
Finally, exploiting the low Péclet regimePe(h)<1Pe(h)<1Pe(h) < 1P e(h)<1Pe(h)<1, we obtain the conclusion.
Proposition 3 (Convergence of regularization). Assume the low Péclet regime (11) and that|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|. LetuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)be a solution to (1) and (uh,zh)[Vh]2uh,zhVh2{:u_(h),z_(h))in[V_(h)]^(2)\left.u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}uh,zh)[Vh]2the solution to (14), then
(πhuuh,zh)sC(μ12h+|β|12h32)(|u|H2(Ω)+h1δω).πhuuh,zhsCμ12h+|β|12h32|u|H2(Ω)+h1δω.||(pi_(h)u-u_(h),z_(h))||_(s) <= C(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))(|u|_(H^(2)(Omega))+h^(-1)||delta||_(omega)).\left\|\left(\pi_{h} u-u_{h}, z_{h}\right)\right\|_{s} \leq C\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right)\left(|u|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right) .(πhuuh,zh)sC(μ12h+|β|12h32)(|u|H2(Ω)+h1δω).
Proof. Denotingeh=πhuuheh=πhuuhe_(h)=pi_(h)u-u_(h)e_{h}=\pi_{h} u-u_{h}eh=πhuuh, it holds by definition that
(eh,zh)s2=ah(eh,zh)+s(zh,zh)ah(eh,zh)+s(eh,eh)eh,zhs2=aheh,zh+szh,zhaheh,zh+seh,eh||(e_(h),z_(h))||_(s)^(2)=a_(h)(e_(h),z_(h))+s_(**)(z_(h),z_(h))-a_(h)(e_(h),z_(h))+s(e_(h),e_(h))\left\|\left(e_{h}, z_{h}\right)\right\|_{s}^{2}=a_{h}\left(e_{h}, z_{h}\right)+s_{*}\left(z_{h}, z_{h}\right)-a_{h}\left(e_{h}, z_{h}\right)+s\left(e_{h}, e_{h}\right)(eh,zh)s2=ah(eh,zh)+s(zh,zh)ah(eh,zh)+s(eh,eh)
Using both claims in Lemma 3 we may write
(eh,zh)s2=ah(πhuu,zh)+sΩ(πhuu,eh)+sω(πhuU~ω,eh).eh,zhs2=ahπhuu,zh+sΩπhuu,eh+sωπhuU~ω,eh.||(e_(h),z_(h))||_(s)^(2)=a_(h)(pi_(h)u-u,z_(h))+s_(Omega)(pi_(h)u-u,e_(h))+s_(omega)(pi_(h)u- tilde(U)_(omega),e_(h)).\left\|\left(e_{h}, z_{h}\right)\right\|_{s}^{2}=a_{h}\left(\pi_{h} u-u, z_{h}\right)+s_{\Omega}\left(\pi_{h} u-u, e_{h}\right)+s_{\omega}\left(\pi_{h} u-\tilde{U}_{\omega}, e_{h}\right) .(eh,zh)s2=ah(πhuu,zh)+sΩ(πhuu,eh)+sω(πhuU~ω,eh).
Lemma 4 gives the bound
ah(πhuu,zh)Cπhuu(0,zh)s.ahπhuu,zhCπhuu0,zhs.a_(h)(pi_(h)u-u,z_(h)) <= C||pi_(h)u-u||_(♯)||(0,z_(h))||_(s).a_{h}\left(\pi_{h} u-u, z_{h}\right) \leq C\left\|\pi_{h} u-u\right\|_{\sharp}\left\|\left(0, z_{h}\right)\right\|_{s} .ah(πhuu,zh)Cπhuu(0,zh)s.
The other terms are simply bounded using the Cauchy-Schwarz inequality as follows
sΩ(πhuu,eh)+sω(πhuU~ω,eh)((πhuu,0)s+(μ12+|β|12h12)δω)(eh,0)s.sΩπhuu,eh+sωπhuU~ω,ehπhuu,0s+μ12+|β|12h12δωeh,0s.s_(Omega)(pi_(h)u-u,e_(h))+s_(omega)(pi_(h)u- tilde(U)_(omega),e_(h)) <= (||(pi_(h)u-u,0)||_(s)+(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))||delta||_(omega))||(e_(h),0)||_(s).s_{\Omega}\left(\pi_{h} u-u, e_{h}\right)+s_{\omega}\left(\pi_{h} u-\tilde{U}_{\omega}, e_{h}\right) \leq\left(\left\|\left(\pi_{h} u-u, 0\right)\right\|_{s}+\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)\|\delta\|_{\omega}\right)\left\|\left(e_{h}, 0\right)\right\|_{s} .sΩ(πhuu,eh)+sω(πhuU~ω,eh)((πhuu,0)s+(μ12+|β|12h12)δω)(eh,0)s.
Collecting the above bounds we have
(eh,zh)s2C(πhuu+(πhuu,0)s+(μ12+|β|12h12)δω)(eh,zh)seh,zhs2Cπhuu+πhuu,0s+μ12+|β|12h12δωeh,zhs||(e_(h),z_(h))||_(s)^(2) <= C(||pi_(h)u-u||_(♯)+||(pi_(h)u-u,0)||_(s)+(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))||delta||_(omega))||(e_(h),z_(h))||_(s)\left\|\left(e_{h}, z_{h}\right)\right\|_{s}^{2} \leq C\left(\left\|\pi_{h} u-u\right\|_{\sharp}+\left\|\left(\pi_{h} u-u, 0\right)\right\|_{s}+\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)\|\delta\|_{\omega}\right)\left\|\left(e_{h}, z_{h}\right)\right\|_{s}(eh,zh)s2C(πhuu+(πhuu,0)s+(μ12+|β|12h12)δω)(eh,zh)s
and the claim follows by applying the approximation (19).
Lemma 5 (Covergence of the convective term). Assume the low Péclet regime (11) and that|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|. LetuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)be a solution to (1),(uh,zh)[Vh]2uh,zhVh2(u_(h),z_(h))in[V_(h)]^(2)\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}(uh,zh)[Vh]2the solution to (14) andwH01(Ω)wH01(Ω)w inH_(0)^(1)(Omega)w \in H_{0}^{1}(\Omega)wH01(Ω), then
(βuh,wπhw)ΩC(μ+|β|)(huH2(Ω)+δω)wH1(Ω)βuh,wπhwΩC(μ+|β|)huH2(Ω)+δωwH1(Ω)(beta*gradu_(h),w-pi_(h)w)_(Omega) <= C(mu+|beta|)(h||u||_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega))\left(\beta \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega} \leq C(\mu+|\beta|)\left(h\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)}(βuh,wπhw)ΩC(μ+|β|)(huH2(Ω)+δω)wH1(Ω)
Proof. Denote byβh[Vh]nβhVhnbeta_(h)in[V_(h)]^(n)\beta_{h} \in\left[V_{h}\right]^{n}βh[Vh]na piecewise linear approximation ofββbeta\betaβthat isLLL^(oo)L^{\infty}L-stable and for which
ββh0,Ch|β|1,ββh0,Ch|β|1,||beta-beta_(h)||_(0,oo) <= Ch|beta|_(1,oo)\left\|\beta-\beta_{h}\right\|_{0, \infty} \leq C h|\beta|_{1, \infty}ββh0,Ch|β|1,
and recall the approximation estimate in [Bur05, Theorem 2.2]
(20)infxhVhh12(βhuhxh)ΩC(FFih[[βhuh]]F2)12C|β|12sΩ(uh,uh)12(20)infxhVhh12βhuhxhΩCFFih[[βhuh]]F212C|β|12sΩuh,uh12{:(20)i n f_(x_(h)inV_(h))||h^((1)/(2))(beta_(h)*gradu_(h)-x_(h))||_(Omega) <= C(sum_(F inF_(i))||h([[)beta_(h)*gradu_(h)(]])||_(F)^(2))^((1)/(2)) <= C|beta|^((1)/(2))s_(Omega)(u_(h),u_(h))^((1)/(2)):}\begin{equation*} \inf _{x_{h} \in V_{h}}\left\|h^{\frac{1}{2}}\left(\beta_{h} \cdot \nabla u_{h}-x_{h}\right)\right\|_{\Omega} \leq C\left(\sum_{F \in \mathcal{F}_{i}}\left\|h \llbracket \beta_{h} \cdot \nabla u_{h} \rrbracket\right\|_{F}^{2}\right)^{\frac{1}{2}} \leq C|\beta|^{\frac{1}{2}} s_{\Omega}\left(u_{h}, u_{h}\right)^{\frac{1}{2}} \tag{20} \end{equation*}(20)infxhVhh12(βhuhxh)ΩC(FFih[[βhuh]]F2)12C|β|12sΩ(uh,uh)12
We also use Proposition 3 and the jump inequality (13) to estimate
sΩ(uh,uh)12sΩ(uhπhu,uhπhu)12+sΩ(πhu,πhu)12C(μ12h+|β|12h32)(|u|H2(Ω)+h1δω)+C(μ12+|β|12h12)h|u|H2(Ω),sΩuh,uh12sΩuhπhu,uhπhu12+sΩπhu,πhu12Cμ12h+|β|12h32|u|H2(Ω)+h1δω+Cμ12+|β|12h12h|u|H2(Ω),{:[s_(Omega)(u_(h),u_(h))^((1)/(2)) <= s_(Omega)(u_(h)-pi_(h)u,u_(h)-pi_(h)u)^((1)/(2))+s_(Omega)(pi_(h)u,pi_(h)u)^((1)/(2))],[ <= C(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))(|u|_(H^(2)(Omega))+h^(-1)||delta||_(omega))+C(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))h|u|_(H^(2)(Omega))","]:}\begin{aligned} s_{\Omega}\left(u_{h}, u_{h}\right)^{\frac{1}{2}} & \leq s_{\Omega}\left(u_{h}-\pi_{h} u, u_{h}-\pi_{h} u\right)^{\frac{1}{2}}+s_{\Omega}\left(\pi_{h} u, \pi_{h} u\right)^{\frac{1}{2}} \\ & \leq C\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right)\left(|u|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right)+C\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right) h|u|_{H^{2}(\Omega)}, \end{aligned}sΩ(uh,uh)12sΩ(uhπhu,uhπhu)12+sΩ(πhu,πhu)12C(μ12h+|β|12h32)(|u|H2(Ω)+h1δω)+C(μ12+|β|12h12)h|u|H2(Ω),
obtaining
(21)sΩ(uh,uh)12C(μ12h+|β|12h32)(|u|H2(Ω)+h1δω)(21)sΩuh,uh12Cμ12h+|β|12h32|u|H2(Ω)+h1δω{:(21)s_(Omega)(u_(h),u_(h))^((1)/(2)) <= C(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))(|u|_(H^(2)(Omega))+h^(-1)||delta||_(omega)):}\begin{equation*} s_{\Omega}\left(u_{h}, u_{h}\right)^{\frac{1}{2}} \leq C\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right)\left(|u|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right) \tag{21} \end{equation*}(21)sΩ(uh,uh)12C(μ12h+|β|12h32)(|u|H2(Ω)+h1δω)
We now write
(βuh,wπhw)Ω=(βhuh,wπhw)Ω+((ββh)uh,wπhw)Ωβuh,wπhwΩ=βhuh,wπhwΩ+ββhuh,wπhwΩ(beta*gradu_(h),w-pi_(h)w)_(Omega)=(beta_(h)*gradu_(h),w-pi_(h)w)_(Omega)+((beta-beta_(h))*gradu_(h),w-pi_(h)w)_(Omega)\left(\beta \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega}=\left(\beta_{h} \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega}+\left(\left(\beta-\beta_{h}\right) \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega}(βuh,wπhw)Ω=(βhuh,wπhw)Ω+((ββh)uh,wπhw)Ω
and using orthogonality, (20), (21), interpolation and (11), we bound the first term by
(βhuh,wπhw)ΩC|β|12h12sΩ(uh,uh)12hwH1(Ω)C|β|12h12(μ12+|β|12h12)(h|u|H2(Ω)+δω)wH1(Ω)C(μ+|β|h)(h|u|H2(Ω)+δω)wH1(Ω)βhuh,wπhwΩC|β|12h12sΩuh,uh12hwH1(Ω)C|β|12h12μ12+|β|12h12h|u|H2(Ω)+δωwH1(Ω)C(μ+|β|h)h|u|H2(Ω)+δωwH1(Ω){:[(beta_(h)*gradu_(h),w-pi_(h)w)_(Omega) <= C|beta|^((1)/(2))h^(-(1)/(2))s_(Omega)(u_(h),u_(h))^((1)/(2))h||w||_(H^(1)(Omega))],[ <= C|beta|^((1)/(2))h^((1)/(2))(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))(h|u|_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega))],[ <= C(mu+|beta|h)(h|u|_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega))]:}\begin{aligned} \left(\beta_{h} \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega} & \leq C|\beta|^{\frac{1}{2}} h^{-\frac{1}{2}} s_{\Omega}\left(u_{h}, u_{h}\right)^{\frac{1}{2}} h\|w\|_{H^{1}(\Omega)} \\ & \leq C|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)\left(h|u|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} \\ & \leq C(\mu+|\beta| h)\left(h|u|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} \end{aligned}(βhuh,wπhw)ΩC|β|12h12sΩ(uh,uh)12hwH1(Ω)C|β|12h12(μ12+|β|12h12)(h|u|H2(Ω)+δω)wH1(Ω)C(μ+|β|h)(h|u|H2(Ω)+δω)wH1(Ω)
We now use the Poincaré-type inequality (12) and interpolation to bound the second term
((ββh)uh,wπhw)ΩCh2|β|1,uhΩwH1(Ω)Ch|β|1,(μ12+|β|12h12)1s(uh,uh)12wH1(Ω)Ch|β|1,(h|u|H2(Ω)+uΩ+δω)wH1(Ω)Ch|β|1,(uH2(Ω)+δω)wH1(Ω)ββhuh,wπhwΩCh2|β|1,uhΩwH1(Ω)Ch|β|1,μ12+|β|12h121suh,uh12wH1(Ω)Ch|β|1,h|u|H2(Ω)+uΩ+δωwH1(Ω)Ch|β|1,uH2(Ω)+δωwH1(Ω){:[((beta-beta_(h))*gradu_(h),w-pi_(h)w)_(Omega) <= Ch^(2)|beta|_(1,oo)||gradu_(h)||_(Omega)||w||_(H^(1)(Omega))],[ <= Ch|beta|_(1,oo)(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))^(-1)s(u_(h),u_(h))^((1)/(2))||w||_(H^(1)(Omega))],[ <= Ch|beta|_(1,oo)(h|u|_(H^(2)(Omega))+||u||_(Omega)+||delta||_(omega))||w||_(H^(1)(Omega))],[ <= Ch|beta|_(1,oo)(||u||_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega))]:}\begin{aligned} \left(\left(\beta-\beta_{h}\right) \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega} & \leq C h^{2}|\beta|_{1, \infty}\left\|\nabla u_{h}\right\|_{\Omega}\|w\|_{H^{1}(\Omega)} \\ & \leq C h|\beta|_{1, \infty}\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)^{-1} s\left(u_{h}, u_{h}\right)^{\frac{1}{2}}\|w\|_{H^{1}(\Omega)} \\ & \leq C h|\beta|_{1, \infty}\left(h|u|_{H^{2}(\Omega)}+\|u\|_{\Omega}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} \\ & \leq C h|\beta|_{1, \infty}\left(\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} \end{aligned}((ββh)uh,wπhw)ΩCh2|β|1,uhΩwH1(Ω)Ch|β|1,(μ12+|β|12h12)1s(uh,uh)12wH1(Ω)Ch|β|1,(h|u|H2(Ω)+uΩ+δω)wH1(Ω)Ch|β|1,(uH2(Ω)+δω)wH1(Ω)
since due to Proposition 3 and inequality (13)
s(uh,uh)12s(uhπhu,uhπhu)12+sΩ(πhu,πhu)12+sω(πhu,πhu)12C(μ12+|β|12h12)(h|u|H2(Ω)+δω+uΩ)suh,uh12suhπhu,uhπhu12+sΩπhu,πhu12+sωπhu,πhu12Cμ12+|β|12h12h|u|H2(Ω)+δω+uΩ{:[s(u_(h),u_(h))^((1)/(2)) <= s(u_(h)-pi_(h)u,u_(h)-pi_(h)u)^((1)/(2))+s_(Omega)(pi_(h)u,pi_(h)u)^((1)/(2))+s_(omega)(pi_(h)u,pi_(h)u)^((1)/(2))],[ <= C(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))(h|u|_(H^(2)(Omega))+||delta||_(omega)+||u||_(Omega))]:}\begin{aligned} s\left(u_{h}, u_{h}\right)^{\frac{1}{2}} & \leq s\left(u_{h}-\pi_{h} u, u_{h}-\pi_{h} u\right)^{\frac{1}{2}}+s_{\Omega}\left(\pi_{h} u, \pi_{h} u\right)^{\frac{1}{2}}+s_{\omega}\left(\pi_{h} u, \pi_{h} u\right)^{\frac{1}{2}} \\ & \leq C\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)\left(h|u|_{H^{2}(\Omega)}+\|\delta\|_{\omega}+\|u\|_{\Omega}\right) \end{aligned}s(uh,uh)12s(uhπhu,uhπhu)12+sΩ(πhu,πhu)12+sω(πhu,πhu)12C(μ12+|β|12h12)(h|u|H2(Ω)+δω+uΩ)
Under the assumption|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|, we collect the above bounds to get
(βuh,wπhw)ΩC(μ+|β|)(huH2(Ω)+δω)wH1(Ω)βuh,wπhwΩC(μ+|β|)huH2(Ω)+δωwH1(Ω)(beta*gradu_(h),w-pi_(h)w)_(Omega) <= C(mu+|beta|)(h||u||_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega))\left(\beta \cdot \nabla u_{h}, w-\pi_{h} w\right)_{\Omega} \leq C(\mu+|\beta|)\left(h\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)}(βuh,wπhw)ΩC(μ+|β|)(huH2(Ω)+δω)wH1(Ω)
We now combine these results with the conditional stability estimates from Section 2 to obtain error bounds for the discrete solution. For this purpose, we consider an open bounded setBΩBΩB sub OmegaB \subset \OmegaBΩthat contains the data regionωωomega\omegaωsuch thatBωBωB\\omegaB \backslash \omegaBωdoes not touch the boundary ofΩΩOmega\OmegaΩ. Then the estimates in Lemma 2 and Corollary 2 hold true by a covering argument, see e.g. [MV12], and we obtain local error estimates inBBBBB. For global unique continuation fromωωomega\omegaωto the entireΩΩOmega\OmegaΩ, however, the stability deteriorates and it is of a different nature: the modulus of continuity for the given data is not of Hölder type||κ||κ|*|^(kappa)|\cdot|^{\kappa}||κany more, but of a logarithmic kind|log()|κ|log()|κ|log(*)|^(-kappa)|\log (\cdot)|^{-\kappa}|log()|κ.
Theorem 1 (L2L2L^(2)L^{2}L2-error estimate). Assume the low Péclet regime (11) and that|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|. ConsiderωBΩωBΩomega sub B sub Omega\omega \subset B \subset \OmegaωBΩsuch thatBωΩBω¯Ωbar(B\\omega)sub Omega\overline{B \backslash \omega} \subset \OmegaBωΩ. LetuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)be a solution to (1) and(uh,zh)[Vh]2uh,zhVh2(u_(h),z_(h))in[V_(h)]^(2)\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}(uh,zh)[Vh]2the solution to (14), then there isκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)such that
uuhL2(B)ChκeCP~e2(uH2(Ω)+h1δω),uuhL2(B)ChκeCP~e2uH2(Ω)+h1δω,||u-u_(h)||_(L^(2)(B)) <= Ch^(kappa)e^(C tilde(P)e^(2))(||u||_(H^(2)(Omega))+h^(-1)||delta||_(omega)),\left\|u-u_{h}\right\|_{L^{2}(B)} \leq C h^{\kappa} e^{C \tilde{P} e^{2}}\left(\|u\|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right),uuhL2(B)ChκeCP~e2(uH2(Ω)+h1δω),
wherePe~=1+|β|/μPe~=1+|β|/μtilde(Pe)=1+|beta|//mu\tilde{P e}=1+|\beta| / \muPe~=1+|β|/μ.
Proof. Let us consider the residual defined byr,w=ah(uh,w)f,wr,w=ahuh,wf,w(:r,w:)=a_(h)(u_(h),w)-(:f,w:)\langle r, w\rangle=a_{h}\left(u_{h}, w\right)-\langle f, w\rangler,w=ah(uh,w)f,w, forwH01(Ω)wH01(Ω)w inH_(0)^(1)(Omega)w \in H_{0}^{1}(\Omega)wH01(Ω). Using (14) we obtain
r,w=ah(uh,wπhw)f,wπhw+ah(uh,πhw)f,πhw=ah(uh,wπhw)f,wπhw+s(zh,πhw)r,w=ahuh,wπhwf,wπhw+ahuh,πhwf,πhw=ahuh,wπhwf,wπhw+szh,πhw{:[(:r","w:)=a_(h)(u_(h),w-pi_(h)w)-(:f,w-pi_(h)w:)+a_(h)(u_(h),pi_(h)w)-(:f,pi_(h)w:)],[=a_(h)(u_(h),w-pi_(h)w)-(:f,w-pi_(h)w:)+s_(**)(z_(h),pi_(h)w)]:}\begin{aligned} \langle r, w\rangle & =a_{h}\left(u_{h}, w-\pi_{h} w\right)-\left\langle f, w-\pi_{h} w\right\rangle+a_{h}\left(u_{h}, \pi_{h} w\right)-\left\langle f, \pi_{h} w\right\rangle \\ & =a_{h}\left(u_{h}, w-\pi_{h} w\right)-\left\langle f, w-\pi_{h} w\right\rangle+s_{*}\left(z_{h}, \pi_{h} w\right) \end{aligned}r,w=ah(uh,wπhw)f,wπhw+ah(uh,πhw)f,πhw=ah(uh,wπhw)f,wπhw+s(zh,πhw)
We split the first term in the right-hand side into convective and non-convective terms, and for the latter we integrate by parts on each elementKKKKKand use Cauchy-Schwarz followed by the trace inequality (17) to get
(μuh,(wπhw))Ωμuhn,wπhwΩ=FFiFμ[[uhn]]F(wπhw)dsCμ(μ+|β|h)12sΩ(uh,uh)12(h1wπhwL2(Ω)+wπhwH1(Ω))μuh,wπhwΩμuhn,wπhwΩ=FFiFμ[[uhn]]FwπhwdsCμ(μ+|β|h)12sΩuh,uh12h1wπhwL2(Ω)+wπhwH1(Ω){:[(mu gradu_(h),grad(w-pi_(h)w))_(Omega)-(:mu gradu_(h)*n,w-pi_(h)w:)_(del Omega)=sum_(F inF_(i))int_(F)mu[[gradu_(h)*n]]_(F)(w-pi_(h)w)ds],[ <= C mu(mu+|beta|h)^(-(1)/(2))s_(Omega)(u_(h),u_(h))^((1)/(2))(h^(-1)||w-pi_(h)w||_(L^(2)(Omega))+||w-pi_(h)w||_(H^(1)(Omega)))]:}\begin{aligned} & \left(\mu \nabla u_{h}, \nabla\left(w-\pi_{h} w\right)\right)_{\Omega}-\left\langle\mu \nabla u_{h} \cdot n, w-\pi_{h} w\right\rangle_{\partial \Omega}=\sum_{F \in \mathcal{F}_{i}} \int_{F} \mu \llbracket \nabla u_{h} \cdot n \rrbracket_{F}\left(w-\pi_{h} w\right) \mathrm{d} s \\ & \leq C \mu(\mu+|\beta| h)^{-\frac{1}{2}} s_{\Omega}\left(u_{h}, u_{h}\right)^{\frac{1}{2}}\left(h^{-1}\left\|w-\pi_{h} w\right\|_{L^{2}(\Omega)}+\left\|w-\pi_{h} w\right\|_{H^{1}(\Omega)}\right) \end{aligned}(μuh,(wπhw))Ωμuhn,wπhwΩ=FFiFμ[[uhn]]F(wπhw)dsCμ(μ+|β|h)12sΩ(uh,uh)12(h1wπhwL2(Ω)+wπhwH1(Ω))
Using (21) and interpolation we obtain
(μuh,(wπhw))Ωμuhn,wπhwΩCμ(h|u|H2(Ω)+δω)wH1(Ω).μuh,wπhwΩμuhn,wπhwΩCμh|u|H2(Ω)+δωwH1(Ω).(mu gradu_(h),grad(w-pi_(h)w))_(Omega)-(:mu gradu_(h)*n,w-pi_(h)w:)_(del Omega) <= C mu(h|u|_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega)).\left(\mu \nabla u_{h}, \nabla\left(w-\pi_{h} w\right)\right)_{\Omega}-\left\langle\mu \nabla u_{h} \cdot n, w-\pi_{h} w\right\rangle_{\partial \Omega} \leq C \mu\left(h|u|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} .(μuh,(wπhw))Ωμuhn,wπhwΩCμ(h|u|H2(Ω)+δω)wH1(Ω).
We bound the convective term inah(uh,wπhw)ahuh,wπhwa_(h)(u_(h),w-pi_(h)w)a_{h}\left(u_{h}, w-\pi_{h} w\right)ah(uh,wπhw)by Lemma 5, hence obtaining
ah(uh,wπhw)C(μ+|β|)(huH2(Ω)+δω)wH1(Ω).ahuh,wπhwC(μ+|β|)huH2(Ω)+δωwH1(Ω).a_(h)(u_(h),w-pi_(h)w) <= C(mu+|beta|)(h||u||_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega)).a_{h}\left(u_{h}, w-\pi_{h} w\right) \leq C(\mu+|\beta|)\left(h\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} .ah(uh,wπhw)C(μ+|β|)(huH2(Ω)+δω)wH1(Ω).
The next term in the residual is bounded by
f,wπhwfL2(Ω)wπhwL2(Ω)ChfL2(Ω)wH1(Ω).f,wπhwfL2(Ω)wπhwL2(Ω)ChfL2(Ω)wH1(Ω).(:f,w-pi_(h)w:) <= ||f||_(L^(2)(Omega))||w-pi_(h)w||_(L^(2)(Omega)) <= Ch||f||_(L^(2)(Omega))||w||_(H^(1)(Omega)).\left\langle f, w-\pi_{h} w\right\rangle \leq\|f\|_{L^{2}(\Omega)}\left\|w-\pi_{h} w\right\|_{L^{2}(\Omega)} \leq C h\|f\|_{L^{2}(\Omega)}\|w\|_{H^{1}(\Omega)} .f,wπhwfL2(Ω)wπhwL2(Ω)ChfL2(Ω)wH1(Ω).
The last term left to bound from the residual is
s(zh,πhw)(0,zh)s(0,πhw)s,szh,πhw0,zhs0,πhws,s_(**)(z_(h),pi_(h)w) <= ||(0,z_(h))||_(s)||(0,pi_(h)w)||_(s),s_{*}\left(z_{h}, \pi_{h} w\right) \leq\left\|\left(0, z_{h}\right)\right\|_{s}\left\|\left(0, \pi_{h} w\right)\right\|_{s},s(zh,πhw)(0,zh)s(0,πhw)s,
and using (18) for the jump term, together with theH1H1H^(1)H^{1}H1-stability ofπhπhpi_(h)\pi_{h}πh, we see that
(0,πhw)sC(μ12(πhw)Ω+(μ12+|β|12h12)(πhw)Ω+(μh1+|β|)12πhwΩ)C(μ12+|β|12h12)wH1(Ω),0,πhwsCμ12πhwΩ+μ12+|β|12h12πhwΩ+μh1+|β|12πhwΩCμ12+|β|12h12wH1(Ω),{:[||(0,pi_(h)w)||_(s) <= C(mu^((1)/(2))||grad(pi_(h)w)||_(Omega)+(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))||grad(pi_(h)w)||_(Omega)+(muh^(-1)+|beta|)^((1)/(2))||pi_(h)w||_(del Omega))],[ <= C(mu^((1)/(2))+|beta|^((1)/(2))h^((1)/(2)))||w||_(H^(1)(Omega))","]:}\begin{aligned} \left\|\left(0, \pi_{h} w\right)\right\|_{s} & \leq C\left(\mu^{\frac{1}{2}}\left\|\nabla\left(\pi_{h} w\right)\right\|_{\Omega}+\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)\left\|\nabla\left(\pi_{h} w\right)\right\|_{\Omega}+\left(\mu h^{-1}+|\beta|\right)^{\frac{1}{2}}\left\|\pi_{h} w\right\|_{\partial \Omega}\right) \\ & \leq C\left(\mu^{\frac{1}{2}}+|\beta|^{\frac{1}{2}} h^{\frac{1}{2}}\right)\|w\|_{H^{1}(\Omega)}, \end{aligned}(0,πhw)sC(μ12(πhw)Ω+(μ12+|β|12h12)(πhw)Ω+(μh1+|β|)12πhwΩ)C(μ12+|β|12h12)wH1(Ω),
where for the boundary term we used thatw|Ω=0wΩ=0w|_(del Omega)=0\left.w\right|_{\partial \Omega}=0w|Ω=0together with interpolation and (17). Bounding(0,zh)s0,zhs||(0,z_(h))||_(s)\left\|\left(0, z_{h}\right)\right\|_{s}(0,zh)sby Proposition 3, we get
s(zh,πhw)C(μ+|β|h)(h|u|H2(Ω)+δω)wH1(Ω).szh,πhwC(μ+|β|h)h|u|H2(Ω)+δωwH1(Ω).s_(**)(z_(h),pi_(h)w) <= C(mu+|beta|h)(h|u|_(H^(2)(Omega))+||delta||_(omega))||w||_(H^(1)(Omega)).s_{*}\left(z_{h}, \pi_{h} w\right) \leq C(\mu+|\beta| h)\left(h|u|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)\|w\|_{H^{1}(\Omega)} .s(zh,πhw)C(μ+|β|h)(h|u|H2(Ω)+δω)wH1(Ω).
Collecting the above estimates we bound the residual norm by
rH1(Ω)C(μ+|β|)(huH2(Ω)+δω)+ChfL2(Ω)C(μ+|β|)(huH2(Ω)+δω).rH1(Ω)C(μ+|β|)huH2(Ω)+δω+ChfL2(Ω)C(μ+|β|)huH2(Ω)+δω.{:[||r||_(H^(-1)(Omega)) <= C(mu+|beta|)(h||u||_(H^(2)(Omega))+||delta||_(omega))+Ch||f||_(L^(2)(Omega))],[ <= C(mu+|beta|)(h||u||_(H^(2)(Omega))+||delta||_(omega)).]:}\begin{aligned} \|r\|_{H^{-1}(\Omega)} & \leq C(\mu+|\beta|)\left(h\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)+C h\|f\|_{L^{2}(\Omega)} \\ & \leq C(\mu+|\beta|)\left(h\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right) . \end{aligned}rH1(Ω)C(μ+|β|)(huH2(Ω)+δω)+ChfL2(Ω)C(μ+|β|)(huH2(Ω)+δω).
We now use the stability estimate in Lemma 2 to write
uuhL2(B)CeCP~e2(uuhL2(ω)+1μrH1(Ω))κuuhL2(Ω)1κ.uuhL2(B)CeCP~e2uuhL2(ω)+1μrH1(Ω)κuuhL2(Ω)1κ.||u-u_(h)||_(L^(2)(B)) <= Ce^(C tilde(P)e^(2))(||u-u_(h)||_(L^(2)(omega))+(1)/(mu)||r||_(H^(-1)(Omega)))^(kappa)||u-u_(h)||_(L^(2)(Omega))^(1-kappa).\left\|u-u_{h}\right\|_{L^{2}(B)} \leq C e^{C \tilde{P} e^{2}}\left(\left\|u-u_{h}\right\|_{L^{2}(\omega)}+\frac{1}{\mu}\|r\|_{H^{-1}(\Omega)}\right)^{\kappa}\left\|u-u_{h}\right\|_{L^{2}(\Omega)}^{1-\kappa} .uuhL2(B)CeCP~e2(uuhL2(ω)+1μrH1(Ω))κuuhL2(Ω)1κ.
By Proposition 3 we have
uuhL2(ω)uπhuL2(ω)+uhπhuL2(ω)Ch2|u|H2(Ω)+Ch|u|H2(Ω)+Cδω.C(h|u|H2(Ω)+δω)uuhL2(ω)uπhuL2(ω)+uhπhuL2(ω)Ch2|u|H2(Ω)+Ch|u|H2(Ω)+Cδω.Ch|u|H2(Ω)+δω{:[||u-u_(h)||_(L^(2)(omega)) <= ||u-pi_(h)u||_(L^(2)(omega))+||u_(h)-pi_(h)u||_(L^(2)(omega))],[ <= Ch^(2)|u|_(H^(2)(Omega))+Ch|u|_(H^(2)(Omega))+C||delta||_(omega).],[ <= C(h|u|_(H^(2)(Omega))+||delta||_(omega))]:}\begin{aligned} \left\|u-u_{h}\right\|_{L^{2}(\omega)} & \leq\left\|u-\pi_{h} u\right\|_{L^{2}(\omega)}+\left\|u_{h}-\pi_{h} u\right\|_{L^{2}(\omega)} \\ & \leq C h^{2}|u|_{H^{2}(\Omega)}+C h|u|_{H^{2}(\Omega)}+C\|\delta\|_{\omega} . \\ & \leq C\left(h|u|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right) \end{aligned}uuhL2(ω)uπhuL2(ω)+uhπhuL2(ω)Ch2|u|H2(Ω)+Ch|u|H2(Ω)+Cδω.C(h|u|H2(Ω)+δω)
Using (12) and Proposition 3 again, we bound
uuhL2(Ω)uπhuL2(Ω)+uhπhuL2(Ω)Ch2|u|H2(Ω)+C(μ12h+|β|12h32)1s(uhπhu,uhπhu)12C(|u|H2(Ω)+h1δω)uuhL2(Ω)uπhuL2(Ω)+uhπhuL2(Ω)Ch2|u|H2(Ω)+Cμ12h+|β|12h321suhπhu,uhπhu12C|u|H2(Ω)+h1δω{:[||u-u_(h)||_(L^(2)(Omega)) <= ||u-pi_(h)u||_(L^(2)(Omega))+||u_(h)-pi_(h)u||_(L^(2)(Omega))],[ <= Ch^(2)|u|_(H^(2)(Omega))+C(mu^((1)/(2))h+|beta|^((1)/(2))h^((3)/(2)))^(-1)s(u_(h)-pi_(h)u,u_(h)-pi_(h)u)^((1)/(2))],[ <= C(|u|_(H^(2)(Omega))+h^(-1)||delta||_(omega))]:}\begin{aligned} \left\|u-u_{h}\right\|_{L^{2}(\Omega)} & \leq\left\|u-\pi_{h} u\right\|_{L^{2}(\Omega)}+\left\|u_{h}-\pi_{h} u\right\|_{L^{2}(\Omega)} \\ & \leq C h^{2}|u|_{H^{2}(\Omega)}+C\left(\mu^{\frac{1}{2}} h+|\beta|^{\frac{1}{2}} h^{\frac{3}{2}}\right)^{-1} s\left(u_{h}-\pi_{h} u, u_{h}-\pi_{h} u\right)^{\frac{1}{2}} \\ & \leq C\left(|u|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right) \end{aligned}uuhL2(Ω)uπhuL2(Ω)+uhπhuL2(Ω)Ch2|u|H2(Ω)+C(μ12h+|β|12h32)1s(uhπhu,uhπhu)12C(|u|H2(Ω)+h1δω)
Hence we conclude by
uuhL2(B)CeCP~e2(huH2(Ω)+δω)κ(|u|H2(Ω)+h1δω)1κCeCP~e2hκ(uH2(Ω)+h1δω),uuhL2(B)CeCP~e2huH2(Ω)+δωκ|u|H2(Ω)+h1δω1κCeCP~e2hκuH2(Ω)+h1δω,{:[||u-u_(h)||_(L^(2)(B)) <= Ce^(C tilde(P)e^(2))(h||u||_(H^(2)(Omega))+||delta||_(omega))^(kappa)(|u|_(H^(2)(Omega))+h^(-1)||delta||_(omega))^(1-kappa)],[ <= Ce^(C tilde(P)e^(2))h^(kappa)(||u||_(H^(2)(Omega))+h^(-1)||delta||_(omega))","]:}\begin{aligned} \left\|u-u_{h}\right\|_{L^{2}(B)} & \leq C e^{C \tilde{P} e^{2}}\left(h\|u\|_{H^{2}(\Omega)}+\|\delta\|_{\omega}\right)^{\kappa}\left(|u|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right)^{1-\kappa} \\ & \leq C e^{C \tilde{P} e^{2}} h^{\kappa}\left(\|u\|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right), \end{aligned}uuhL2(B)CeCP~e2(huH2(Ω)+δω)κ(|u|H2(Ω)+h1δω)1κCeCP~e2hκ(uH2(Ω)+h1δω),
where we have absorbed theP~e=1+|β|/μP~e=1+|β|/μtilde(P)e=1+|beta|//mu\tilde{P} e=1+|\beta| / \muP~e=1+|β|/μfactor into the exponential factoreCP~e2eCP~e2e^(C tilde(P)e^(2))e^{C \tilde{P} e^{2}}eCP~e2.
Remark 2. Let us briefly discuss the effect of decreasing the size of the data regionωωomega\omegaωby considering the case of balls, that isω=B(x0,r1)ω=Bx0,r1omega=B(x_(0),r_(1))\omega=B\left(x_{0}, r_{1}\right)ω=B(x0,r1)andB=B(x0,r2)B=Bx0,r2B=B(x_(0),r_(2))B=B\left(x_{0}, r_{2}\right)B=B(x0,r2), withx0Ωx0Ωx_(0)in Omegax_{0} \in \Omegax0Ωandr1<r2r1<r2r_(1) < r_(2)r_{1}<r_{2}r1<r2. Notice from Remark 1 that the exponentκκkappa\kappaκis an increasing function of the radiusr1r1r_(1)r_{1}r1and that decreasing the size of the data regionωωomega\omegaωimplies that the convergence ratehκhκh^(kappa)h^{\kappa}hκdecreases as well. Bounding the radiusr2r2r_(2)r_{2}r2away from zero and lettingr10r10r_(1)rarr0r_{1} \rightarrow 0r10implies that the exponentκ0κ0kappa rarr0\kappa \rightarrow 0κ0. The continuum three-ball inequality then becomes the trivial inequalityuL2(B)uL2(Ω)uL2(B)uL2(Ω)||u||_(L^(2)(B)) <= ||u||_(L^(2)(Omega))\|u\|_{L^{2}(B)} \leq\|u\|_{L^{2}(\Omega)}uL2(B)uL2(Ω)and the method does not converge any more.
Theorem 2 (H1H1H^(1)H^{1}H1-error estimate). Assume the low Péclet regime (11) and that|β|1,C|β||β|1,C|β||beta|_(1,oo) <= C|beta||\beta|_{1, \infty} \leq C|\beta||β|1,C|β|andesssupΩβ0esssupΩβ0esss u p_(Omega)grad*beta <= 0\operatorname{ess} \sup _{\Omega} \nabla \cdot \beta \leq 0esssupΩβ0. ConsiderωBΩωBΩomega sub B sub Omega\omega \subset B \subset \OmegaωBΩsuch thatBωΩBω¯Ωbar(B\\omega)sub Omega\overline{B \backslash \omega} \subset \OmegaBωΩ. LetuH2(Ω)uH2(Ω)u inH^(2)(Omega)u \in H^{2}(\Omega)uH2(Ω)be a solution to (1), and(uh,zh)[Vh]2uh,zhVh2(u_(h),z_(h))in[V_(h)]^(2)\left(u_{h}, z_{h}\right) \in\left[V_{h}\right]^{2}(uh,zh)[Vh]2the solution to (14), then there isκ(0,1)κ(0,1)kappa in(0,1)\kappa \in(0,1)κ(0,1)such that
uuhH1(B)ChκeCPe~2(uH2(Ω)+h1δω),uuhH1(B)ChκeCPe~2uH2(Ω)+h1δω,||u-u_(h)||_(H^(1)(B)) <= Ch^(kappa)e^(C tilde(Pe)^(2))(||u||_(H^(2)(Omega))+h^(-1)||delta||_(omega)),\left\|u-u_{h}\right\|_{H^{1}(B)} \leq C h^{\kappa} e^{C \tilde{P e}^{2}}\left(\|u\|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right),uuhH1(B)ChκeCPe~2(uH2(Ω)+h1δω),
wherePe~=1+|β|/μPe~=1+|β|/μtilde(Pe)=1+|beta|//mu\tilde{P e}=1+|\beta| / \muPe~=1+|β|/μ.
Proof. Lettingeh=uuheh=uuhe_(h)=u-u_(h)e_{h}=u-u_{h}eh=uuh, we combine the proof of Theorem 1 with the stability estimate in Corollary 2 to obtain
ehH1(B)CeCP~e2(ehL2(ω)+1μrH1(Ω))κ(ehL2(Ω)+1μrH1(Ω))1κCeCP~e2hκ(uH2(Ω)+h1δω)ehH1(B)CeCP~e2ehL2(ω)+1μrH1(Ω)κehL2(Ω)+1μrH1(Ω)1κCeCP~e2hκuH2(Ω)+h1δω{:[||e_(h)||_(H^(1)(B)) <= Ce^(C tilde(P)e^(2))(||e_(h)||_(L^(2)(omega))+(1)/(mu)||r||_(H^(-1)(Omega)))^(kappa)(||e_(h)||_(L^(2)(Omega))+(1)/(mu)||r||_(H^(-1)(Omega)))^(1-kappa)],[ <= Ce^(C tilde(P)e^(2))h^(kappa)(||u||_(H^(2)(Omega))+h^(-1)||delta||_(omega))]:}\begin{aligned} \left\|e_{h}\right\|_{H^{1}(B)} & \leq C e^{C \tilde{P} e^{2}}\left(\left\|e_{h}\right\|_{L^{2}(\omega)}+\frac{1}{\mu}\|r\|_{H^{-1}(\Omega)}\right)^{\kappa}\left(\left\|e_{h}\right\|_{L^{2}(\Omega)}+\frac{1}{\mu}\|r\|_{H^{-1}(\Omega)}\right)^{1-\kappa} \\ & \leq C e^{C \tilde{P} e^{2}} h^{\kappa}\left(\|u\|_{H^{2}(\Omega)}+h^{-1}\|\delta\|_{\omega}\right) \end{aligned}ehH1(B)CeCP~e2(ehL2(ω)+1μrH1(Ω))κ(ehL2(Ω)+1μrH1(Ω))1κCeCP~e2hκ(uH2(Ω)+h1δω)

4. Numerical experiments

We illustrate the theoretical results with some numerical examples. The implementation of the stabilized FEM (14) has been carried out in FreeFem++ [Hec12] on uniform triangulations with alternating left and right diagonals. The mesh size is taken as the inverse square root of the number of nodes. The parameters insΩsΩs_(Omega)s_{\Omega}sΩandsss_(**)s_{*}sare set toγ=105γ=105gamma=10^(-5)\gamma=10^{-5}γ=105andγ=1γ=1gamma_(**)=1\gamma_{*}=1γ=1. We also rescale the boundary term insss_(**)s_{*}sby the factor 50 , drawing on results from different numerical experiments. In this section we denoteeh=πhuuheh=πhuuhe_(h)=pi_(h)u-u_(h)e_{h}=\pi_{h} u-u_{h}eh=πhuuh.
We considerΩΩOmega\OmegaΩto be the unit square and the exact solution with global unitL2L2L^(2)L^{2}L2-norm
u(x,y)=30x(1x)y(1y).u(x,y)=30x(1x)y(1y).u(x,y)=30 x(1-x)y(1-y).u(x, y)=30 x(1-x) y(1-y) .u(x,y)=30x(1x)y(1y).
We take the diffusion coefficientμ=1μ=1mu=1\mu=1μ=1and investigate two cases for the convection field: the coercive case of the constant field
βc=(1,0),βc=(1,0),beta_(c)=(1,0),\beta_{c}=(1,0),βc=(1,0),
and the case
βnc=100(x+y,yx),βnc=100(x+y,yx),beta_(nc)=100(x+y,y-x),\beta_{n c}=100(x+y, y-x),βnc=100(x+y,yx),
plotted in Figure 2, for whichβ=200β=200grad*beta=200\nabla \cdot \beta=200β=200andβ0,=200β0,=200||beta||_(0,oo)=200\|\beta\|_{0, \infty}=200β0,=200. This makes the (well-posed) problem strongly non-coercive with a medium high Péclet number. The latter example was also considered in [Bur13] for numerical experiments on a non-coercive convectiondiffusion equation with Cauchy data.
We consider the following domains for data assimilation, shown in Figure 1,
(22)ω=(0.2,0.45)×(0.2,0.45),B=(0.2,0.45)×(0.55,0.8),(22)ω=(0.2,0.45)×(0.2,0.45),B=(0.2,0.45)×(0.55,0.8),{:(22)omega=(0.2","0.45)xx(0.2","0.45)","quad B=(0.2","0.45)xx(0.55","0.8)",":}\begin{equation*} \omega=(0.2,0.45) \times(0.2,0.45), \quad B=(0.2,0.45) \times(0.55,0.8), \tag{22} \end{equation*}(22)ω=(0.2,0.45)×(0.2,0.45),B=(0.2,0.45)×(0.55,0.8),
(23)ω=(0,0.125)×(0.4,0.6)(0.875,1)×(0.4,0.6),B=(0.25,0.75)×(0.4,0.6),(23)ω=(0,0.125)×(0.4,0.6)(0.875,1)×(0.4,0.6),B=(0.25,0.75)×(0.4,0.6),{:(23)omega=(0","0.125)xx(0.4","0.6)uu(0.875","1)xx(0.4","0.6)","quad B=(0.25","0.75)xx(0.4","0.6)",":}\begin{equation*} \omega=(0,0.125) \times(0.4,0.6) \cup(0.875,1) \times(0.4,0.6), \quad B=(0.25,0.75) \times(0.4,0.6), \tag{23} \end{equation*}(23)ω=(0,0.125)×(0.4,0.6)(0.875,1)×(0.4,0.6),B=(0.25,0.75)×(0.4,0.6),
(24)ω=Ω[0,0.875]×[0.125,0.875],B=Ω[0,0.125]×[0.125,0.875].(24)ω=Ω[0,0.875]×[0.125,0.875],B=Ω[0,0.125]×[0.125,0.875].{:(24)omega=Omega\\[0","0.875]xx[0.125","0.875]","quad B=Omega\\[0","0.125]xx[0.125","0.875].:}\begin{equation*} \omega=\Omega \backslash[0,0.875] \times[0.125,0.875], \quad B=\Omega \backslash[0,0.125] \times[0.125,0.875] . \tag{24} \end{equation*}(24)ω=Ω[0,0.875]×[0.125,0.875],B=Ω[0,0.125]×[0.125,0.875].
(a) Boundaries for (22).
(b) Boundaries for (23).
(c) Boundaries for (24).
Figure 1. Computational domains.
The condition number upper bound in Proposition 2 is illustrated for a particular case in Figure 2, where we plot the condition numberK2K2K_(2)\mathcal{K}_{2}K2versus the mesh sizehhhhh, together with reference dotted lines proportional toh3h3h^(-3)h^{-3}h3andh4h4h^(-4)h^{-4}h4. For five meshes with2N2N2^(N)2^{N}2Nelements on each side,N=3,,7N=3,,7N=3,dots,7N=3, \ldots, 7N=3,,7, the approximate rates forK2K2K_(2)\mathcal{K}_{2}K2are3.03,3.16,3.2,3.343.03,3.16,3.2,3.34-3.03,-3.16,-3.2,-3.34-3.03,-3.16,-3.2,-3.343.03,3.16,3.2,3.34.
The results in Figure 3 for the domains (22) strongly agree with the convergence rates expected from Theorem 1 and Theorem 2 for the relative errors inBBBBBcomputed in theL2L2L^(2)L^{2}L2- andH1H1H^(1)H^{1}H1-norms, and with the rates for(eh,zh)seh,zhs||(e_(h),z_(h))||_(s)\left\|\left(e_{h}, z_{h}\right)\right\|_{s}(eh,zh)sgiven in Proposition 3.
The numerical approximation improves when considering the setting in (23), in which data is given both downstream and upstream, as reported in Figure 4. The convergence is almost linear and the size of the errors is considerably reduced in the non-coercive case.
The resolution increases all the more when data is given near a big part of the boundaryΩΩdel Omega\partial \OmegaΩ, as for the computational domains (24) considered in Figure 5. In this configuration of the setωωomega\omegaω, for both convective fieldsβcβcbeta_(c)\beta_{c}βcandβncβncbeta_(nc)\beta_{n c}βnc, theL2L2L^(2)L^{2}L2-errors decrease below10410410^(-4)10^{-4}104with superlinear rates on the same meshes considered in Figure 3 and Figure 4.
Comparing the geometries in (22) and (23) we also expect to see different effects of the two convective fieldsβcβcbeta_(c)\beta_{c}βcandβncβncbeta_(nc)\beta_{n c}βnc. Notice that for both geometries the horizontal magnitude ofβncβncbeta_(nc)\beta_{n c}βncis greater than that ofβcβcbeta_(c)\beta_{c}βc. In (22) the solution is continued in the
Figure 2. Left: convection fieldβncβncbeta_(nc)\beta_{n c}βnc. Right: condition numberK2K2K_(2)\mathcal{K}_{2}K2for domains (22),β=βcβ=βcbeta=beta_(c)\beta=\beta_{c}β=βc; the dotted lines are proportional toh3h3h^(-3)h^{-3}h3andh4h4h^(-4)h^{-4}h4.
Figure 3. Convergence for domains (22). Left:β=βcβ=βcbeta=beta_(c)\beta=\beta_{c}β=βc. Right:β=βncβ=βncbeta=beta_(nc)\beta=\beta_{n c}β=βnc.
crosswind direction for bothβcβcbeta_(c)\beta_{c}βcandβncβncbeta_(nc)\beta_{n c}βnc, and a stronger convective field is not expected to improve the reconstruction. On the other side, in (23) information is propagated both downstream and upstream, and a stronger convective field can improve the resolution, despite the increase in the Péclet number. Indeed, we can see in Figure 3 that for the geometry in (22) the numerical approximation is better forβcβcbeta_(c)\beta_{c}βcthan forβncβncbeta_(nc)\beta_{n c}βnc, while Figure 4 shows better results forβncβncbeta_(nc)\beta_{n c}βncthan forβcβcbeta_(c)\beta_{c}βcin the case of (23), especially for theL2L2L^(2)L^{2}L2-error.
To exemplify the noisy dataU~ω=u|ω+δU~ω=uω+δtilde(U)_(omega)=u|_(omega)+delta\tilde{U}_{\omega}=\left.u\right|_{\omega}+\deltaU~ω=u|ω+δ, we perturb the restriction ofuuuuutoωωomega\omegaωon every node of the mesh with uniformly distributed values in[h12,h12]h12,h12[-h^((1)/(2)),h^((1)/(2))]\left[-h^{\frac{1}{2}}, h^{\frac{1}{2}}\right][h12,h12], respectively[h,h][h,h][-h,h][-h, h][h,h]. Recall that by the error estimates in Section 3 the contribution of the perturbationδδdelta\deltaδ
Figure 4. Convergence for domains (23). Left:β=βcβ=βcbeta=beta_(c)\beta=\beta_{c}β=βc. Right:β=βncβ=βncbeta=beta_(nc)\beta=\beta_{n c}β=βnc.
Figure 5. Convergence for domains (24). Left:β=βcβ=βcbeta=beta_(c)\beta=\beta_{c}β=βc. Right:β=βncβ=βncbeta=beta_(nc)\beta=\beta_{n c}β=βnc.
is bounded byh1δωh1δωh^(-1)||delta||_(omega)h^{-1}\|\delta\|_{\omega}h1δω. It can be seen in Figure 6 that the perturbations are strongly visible for anO(h12)Oh12O(h^((1)/(2)))O\left(h^{\frac{1}{2}}\right)O(h12)amplitude, but not for anO(h)O(h)O(h)O(h)O(h)one.

Appendix A.

Denote by(,),||(,),||(*,*),|*|(\cdot, \cdot),|\cdot|(,),||, div,grad\nablaandD2D2D^(2)D^{2}D2the inner product, norm, divergence, gradient and Hessian in the Euclidean setting ofΩRnΩRnOmega subR^(n)\Omega \subset \mathbb{R}^{n}ΩRn. We recall the following identity [BNO19b, Lemma 1].
Lemma 6. Let,wC2(Ω),wC2(Ω)ℓ,w inC^(2)(Omega)\ell, w \in C^{2}(\Omega),wC2(Ω)andσC1(Ω)σC1(Ω)sigma inC^(1)(Omega)\sigma \in C^{1}(\Omega)σC1(Ω). We definev=ewv=ewv=e^(ℓ)wv=e^{\ell} wv=ewand
a=σΔ,q=a+||2,b=σv2(v,),B=(|v|2qv2).a=σΔ,q=a+||2,b=σv2(v,),B=|v|2qv2.a=sigma-Deltaℓ,quad q=a+|gradℓ|^(2),quad b=-sigma v-2(grad v,gradℓ),quad B=(|grad v|^(2)-qv^(2))gradℓ.a=\sigma-\Delta \ell, \quad q=a+|\nabla \ell|^{2}, \quad b=-\sigma v-2(\nabla v, \nabla \ell), \quad B=\left(|\nabla v|^{2}-q v^{2}\right) \nabla \ell .a=σΔ,q=a+||2,b=σv2(v,),B=(|v|2qv2).
Figure 6. Convergence for perturbedU~ωU~ωtilde(U)_(omega)\tilde{U}_{\omega}U~ωin domains (22),β=βcβ=βcbeta=beta_(c)\beta=\beta_{c}β=βc.
Then
e2(Δw)2/2=(Δv+qv)2/2+b2/2+a|v|2+2D2(v,v)+(a||2+2D2(,))v2+div(bv+B)+Re2(Δw)2/2=(Δv+qv)2/2+b2/2+a|v|2+2D2(v,v)+a||2+2D2(,)v2+div(bv+B)+R{:[e^(2ℓ)(Delta w)^(2)//2=(Delta v+qv)^(2)//2+b^(2)//2],[+a|grad v|^(2)+2D^(2)ℓ(grad v","grad v)+(-a|gradℓ|^(2)+2D^(2)ℓ(gradℓ,gradℓ))v^(2)],[+div(b grad v+B)+R]:}\begin{aligned} e^{2 \ell}(\Delta w)^{2} / 2 & =(\Delta v+q v)^{2} / 2+b^{2} / 2 \\ & +a|\nabla v|^{2}+2 D^{2} \ell(\nabla v, \nabla v)+\left(-a|\nabla \ell|^{2}+2 D^{2} \ell(\nabla \ell, \nabla \ell)\right) v^{2} \\ & +\operatorname{div}(b \nabla v+B)+R \end{aligned}e2(Δw)2/2=(Δv+qv)2/2+b2/2+a|v|2+2D2(v,v)+(a||2+2D2(,))v2+div(bv+B)+R
whereR=(σ,v)v+(div(a)aσ)v2R=(σ,v)v+(div(a)aσ)v2R=(grad sigma,grad v)v+(div(a gradℓ)-a sigma)v^(2)R=(\nabla \sigma, \nabla v) v+(\operatorname{div}(a \nabla \ell)-a \sigma) v^{2}R=(σ,v)v+(div(a)aσ)v2.
Proof of Proposition 1. Let=τϕ=τϕℓ=tau phi\ell=\tau \phi=τϕand letλ>0λ>0lambda > 0\lambda>0λ>0such that|D2ρ(X,X)|λ|X|2D2ρ(X,X)λ|X|2|D^(2)rho(X,X)| <= lambda|X|^(2)\left|D^{2} \rho(X, X)\right| \leq \lambda|X|^{2}|D2ρ(X,X)|λ|X|2. Recalling thatϕ=eαρϕ=eαρphi=e^(alpha rho)\phi=e^{\alpha \rho}ϕ=eαρand using the product rule we have that
D2ϕ(X,X)=αϕ(α(ρ,X)2+D2ρ(X,X)),D2ϕ(X,X)=αϕα(ρ,X)2+D2ρ(X,X),D^(2)phi(X,X)=alpha phi(alpha(grad rho,X)^(2)+D^(2)rho(X,X)),D^{2} \phi(X, X)=\alpha \phi\left(\alpha(\nabla \rho, X)^{2}+D^{2} \rho(X, X)\right),D2ϕ(X,X)=αϕ(α(ρ,X)2+D2ρ(X,X)),
hence
D2ϕ(X,X)αϕD2ρ(X,X)αλϕ|X|2.D2ϕ(X,X)αϕD2ρ(X,X)αλϕ|X|2.D^(2)phi(X,X) >= alpha phiD^(2)rho(X,X) >= -alpha lambda phi|X|^(2).D^{2} \phi(X, X) \geq \alpha \phi D^{2} \rho(X, X) \geq-\alpha \lambda \phi|X|^{2} .D2ϕ(X,X)αϕD2ρ(X,X)αλϕ|X|2.
Combining this with the previous equality, we obtain
D2ϕ(ϕ,ϕ)α3ϕ3(α|ρ|4λ|ρ|2).D2ϕ(ϕ,ϕ)α3ϕ3α|ρ|4λ|ρ|2.D^(2)phi(grad phi,grad phi) >= alpha^(3)phi^(3)(alpha|grad rho|^(4)-lambda|grad rho|^(2)).D^{2} \phi(\nabla \phi, \nabla \phi) \geq \alpha^{3} \phi^{3}\left(\alpha|\nabla \rho|^{4}-\lambda|\nabla \rho|^{2}\right) .D2ϕ(ϕ,ϕ)α3ϕ3(α|ρ|4λ|ρ|2).
Choosingϵ>0ϵ>0epsilon > 0\epsilon>0ϵ>0such thatϵ|ρ|2ϵ1ϵ|ρ|2ϵ1epsilon <= |grad rho|^(2) <= epsilon^(-1)\epsilon \leq|\nabla \rho|^{2} \leq \epsilon^{-1}ϵ|ρ|2ϵ1it holds
D2ϕ(ϕ,ϕ)α3ϕ3(αϵ2λϵ1).D2ϕ(ϕ,ϕ)α3ϕ3αϵ2λϵ1.D^(2)phi(grad phi,grad phi) >= alpha^(3)phi^(3)(alphaepsilon^(2)-lambdaepsilon^(-1)).D^{2} \phi(\nabla \phi, \nabla \phi) \geq \alpha^{3} \phi^{3}\left(\alpha \epsilon^{2}-\lambda \epsilon^{-1}\right) .D2ϕ(ϕ,ϕ)α3ϕ3(αϵ2λϵ1).
Since
2D2(v,v)2αλϕτ|v|2,2D2(v,v)2αλϕτ|v|2,2D^(2)ℓ(grad v,grad v) >= -2alpha lambda phi tau|grad v|^(2),2 D^{2} \ell(\nabla v, \nabla v) \geq-2 \alpha \lambda \phi \tau|\nabla v|^{2},2D2(v,v)2αλϕτ|v|2,
by choosingσ=Δ+3αλϕτσ=Δ+3αλϕτsigma=Deltaℓ+3alpha lambda phi tau\sigma=\Delta \ell+3 \alpha \lambda \phi \tauσ=Δ+3αλϕτ, i.e.a=3αλϕτa=3αλϕτa=3alpha lambda phi taua=3 \alpha \lambda \phi \taua=3αλϕτin Lemma 6 we obtain the bounds
a|v|2+2D2(v,v)αλϕτ|v|2,(a||2+2D2(,))v2(2αϵ2(3+2λ)ϵ1)(αϕτ)3v2.a|v|2+2D2(v,v)αλϕτ|v|2,a||2+2D2(,)v22αϵ2(3+2λ)ϵ1(αϕτ)3v2.{:[a|grad v|^(2)+2D^(2)ℓ(grad v","grad v) >= alpha lambda phi tau|grad v|^(2)","],[(-a|gradℓ|^(2)+2D^(2)ℓ(gradℓ,gradℓ))v^(2) >= (2alphaepsilon^(2)-(3+2lambda)epsilon^(-1))(alpha phi tau)^(3)v^(2).]:}\begin{aligned} a|\nabla v|^{2}+2 D^{2} \ell(\nabla v, \nabla v) & \geq \alpha \lambda \phi \tau|\nabla v|^{2}, \\ \left(-a|\nabla \ell|^{2}+2 D^{2} \ell(\nabla \ell, \nabla \ell)\right) v^{2} & \geq\left(2 \alpha \epsilon^{2}-(3+2 \lambda) \epsilon^{-1}\right)(\alpha \phi \tau)^{3} v^{2} . \end{aligned}a|v|2+2D2(v,v)αλϕτ|v|2,(a||2+2D2(,))v2(2αϵ2(3+2λ)ϵ1)(αϕτ)3v2.
We now bound
(σ,v)v=((Δ),v)v+3αλ(,v)v(|(Δϕ)|+3αλ|ϕ|)τ|v||v|(σ,v)v=((Δ),v)v+3αλ(,v)v(|(Δϕ)|+3αλ|ϕ|)τ|v||v|(grad sigma,grad v)v=(grad(Deltaℓ),grad v)v+3alpha lambda(gradℓ,grad v)v >= -(|grad(Delta phi)|+3alpha lambda|grad phi|)tau|grad v||v|(\nabla \sigma, \nabla v) v=(\nabla(\Delta \ell), \nabla v) v+3 \alpha \lambda(\nabla \ell, \nabla v) v \geq-(|\nabla(\Delta \phi)|+3 \alpha \lambda|\nabla \phi|) \tau|\nabla v||v|(σ,v)v=((Δ),v)v+3αλ(,v)v(|(Δϕ)|+3αλ|ϕ|)τ|v||v|
and
(div(a)aσ)v2=((a,)a2)v2(3αλ|ϕ|29α2λ2ϕ2)τ2v2.(div(a)aσ)v2=(a,)a2v23αλ|ϕ|29α2λ2ϕ2τ2v2.(div(a gradℓ)-a sigma)v^(2)=((grad a,gradℓ)-a^(2))v^(2) >= (3alpha lambda|grad phi|^(2)-9alpha^(2)lambda^(2)phi^(2))tau^(2)v^(2).(\operatorname{div}(a \nabla \ell)-a \sigma) v^{2}=\left((\nabla a, \nabla \ell)-a^{2}\right) v^{2} \geq\left(3 \alpha \lambda|\nabla \phi|^{2}-9 \alpha^{2} \lambda^{2} \phi^{2}\right) \tau^{2} v^{2} .(div(a)aσ)v2=((a,)a2)v2(3αλ|ϕ|29α2λ2ϕ2)τ2v2.
Combining these lower bounds with
τ|v||v|12(|v|2+τ2|v|2),τ|v||v|12|v|2+τ2|v|2,tau|grad v||v| <= (1)/(2)(|grad v|^(2)+tau^(2)|v|^(2)),\tau|\nabla v||v| \leq \frac{1}{2}\left(|\nabla v|^{2}+\tau^{2}|v|^{2}\right),τ|v||v|12(|v|2+τ2|v|2),
and takingααalpha\alphaαlarge enough, we obtain from Lemma 6 that
(25)Ce2τϕ(Δw)2(a1τ3a2τ2)v2+(b1τb0)|v|2+div(bv+B)(25)Ce2τϕ(Δw)2a1τ3a2τ2v2+b1τb0|v|2+div(bv+B){:(25)Ce^(2tau phi)(Delta w)^(2) >= (a_(1)tau^(3)-a_(2)tau^(2))v^(2)+(b_(1)tau-b_(0))|grad v|^(2)+div(b grad v+B):}\begin{equation*} C e^{2 \tau \phi}(\Delta w)^{2} \geq\left(a_{1} \tau^{3}-a_{2} \tau^{2}\right) v^{2}+\left(b_{1} \tau-b_{0}\right)|\nabla v|^{2}+\operatorname{div}(b \nabla v+B) \tag{25} \end{equation*}(25)Ce2τϕ(Δw)2(a1τ3a2τ2)v2+(b1τb0)|v|2+div(bv+B)
withaj,bj>0aj,bj>0a_(j),b_(j) > 0a_{j}, b_{j}>0aj,bj>0depending only onα,ϕα,ϕalpha,phi\alpha, \phiα,ϕandλλlambda\lambdaλ. Takingττtau\tauτlarge enough and using the elementary inequality
|v|2=e2τϕ|τwϕ+w|2e2τϕ12|w|2e2τϕ|ϕ|2τ2w2,|v|2=e2τϕ|τwϕ+w|2e2τϕ12|w|2e2τϕ|ϕ|2τ2w2,|grad v|^(2)=e^(2tau phi)|tau w grad phi+grad w|^(2) >= e^(2tau phi)(1)/(2)|grad w|^(2)-e^(2tau phi)|grad phi|^(2)tau^(2)w^(2),|\nabla v|^{2}=e^{2 \tau \phi}|\tau w \nabla \phi+\nabla w|^{2} \geq e^{2 \tau \phi} \frac{1}{2}|\nabla w|^{2}-e^{2 \tau \phi}|\nabla \phi|^{2} \tau^{2} w^{2},|v|2=e2τϕ|τwϕ+w|2e2τϕ12|w|2e2τϕ|ϕ|2τ2w2,
we conclude by integrating overKKKKKand using the divergence theorem.

Appendix B.

We briefly recall herein the definition of semiclassical pseudodifferential operators and semiclassical Sobolev spaces. We then discuss the composition rule of two such operators, which is also called symbol calculus, and some estimates that are used in the proof of Lemma 2. This presentation is based on [Zwo12, Chapter 4] and [LRL12, Section 2], to which we refer the reader for more details.
We shall use the following standard notation. ForξRnξRnxi inR^(n)\xi \in \mathbb{R}^{n}ξRnwe setξ=(1+|ξ|2)12ξ=1+|ξ|212(:xi:)=(1+|xi|^(2))^((1)/(2))\langle\xi\rangle=\left(1+|\xi|^{2}\right)^{\frac{1}{2}}ξ=(1+|ξ|2)12, and for a multi-indexα=(α1,,αn)Nnα=α1,,αnNnalpha=(alpha_(1),dots,alpha_(n))inN^(n)\alpha=\left(\alpha_{1}, \ldots, \alpha_{n}\right) \in \mathbb{N}^{n}α=(α1,,αn)Nnlet|α|=α1+αn,α!=α1!αn!|α|=α1+αn,α!=α1!αn!|alpha|=alpha_(1)+dotsalpha_(n),alpha!=alpha_(1)!cdotsalpha_(n)!|\alpha|=\alpha_{1}+\ldots \alpha_{n}, \alpha!=\alpha_{1}!\cdots \alpha_{n}!|α|=α1+αn,α!=α1!αn!,ξα=ξ1α1ξnαnξα=ξ1α1ξnαnxi^(alpha)=xi_(1)^(alpha_(1))cdotsxi_(n)^(alpha_(n))\xi^{\alpha}=\xi_{1}^{\alpha_{1}} \cdots \xi_{n}^{\alpha_{n}}ξα=ξ1α1ξnαn. Let alsoα=x1α1xnαn,D=1iα=x1α1xnαn,D=1idel^(alpha)=del_(x_(1))^(alpha_(1))cdotsdel_(x_(n))^(alpha_(n)),D=(1)/(i)del\partial^{\alpha}=\partial_{x_{1}}^{\alpha_{1}} \cdots \partial_{x_{n}}^{\alpha_{n}}, D=\frac{1}{i} \partialα=x1α1xnαn,D=1iandDα=1i|ααDα=1iααD^(alpha)=(1)/(i|^(alpha∣))del^(alpha)D^{\alpha}=\frac{1}{\left.i\right|^{\alpha \mid}} \partial^{\alpha}Dα=1i|αα. The Schwartz spaceS(Rn)SRnS(R^(n))\mathcal{S}\left(\mathbb{R}^{n}\right)S(Rn)is the set of rapidly decreasingCCC^(oo)C^{\infty}Cfunctions and its dualS(Rn)SRnS^(')(R^(n))\mathcal{S}^{\prime}\left(\mathbb{R}^{n}\right)S(Rn)is the set of tempered distributions. The semiclassical parameter\hbaris assumed to be small:(0,0)0,0ℏin(0,ℏ_(0))\hbar \in\left(0, \hbar_{0}\right)(0,0)with0101ℏ_(0)≪1\hbar_{0} \ll 101.
The semiclassical Fourier transform is a rescaled version of the standard Fourier transform. It is given by
Fφ(ξ):=Rneixξφ(x)dxFφ(ξ):=Rneixξφ(x)dxF_(ℏ)varphi(xi):=int_(R^(n))e^(-(i)/(ℏ)x*xi)varphi(x)dx\mathcal{F}_{\hbar} \varphi(\xi):=\int_{\mathbb{R}^{n}} e^{-\frac{i}{\hbar} x \cdot \xi} \varphi(x) \mathrm{d} xFφ(ξ):=Rneixξφ(x)dx
and its inverse is
F1ψ(x):=1(2π)nRneixξψ(ξ)dξF1ψ(x):=1(2π)nRneixξψ(ξ)dξF_(ℏ)^(-1)psi(x):=(1)/((2piℏ)^(n))int_(R^(n))e^((i)/(ℏ)x*xi)psi(xi)dxi\mathcal{F}_{\hbar}^{-1} \psi(x):=\frac{1}{(2 \pi \hbar)^{n}} \int_{\mathbb{R}^{n}} e^{\frac{i}{\hbar} x \cdot \xi} \psi(\xi) \mathrm{d} \xiF1ψ(x):=1(2π)nRneixξψ(ξ)dξ
The following properties hold:F((Dx)αφ)=ξαFφFDxαφ=ξαFφF_(ℏ)((ℏD_(x))^(alpha)varphi)=xi^(alpha)F_(ℏ)varphi\mathcal{F}_{\hbar}\left(\left(\hbar D_{x}\right)^{\alpha} \varphi\right)=\xi^{\alpha} \mathcal{F}_{\hbar} \varphiF((Dx)αφ)=ξαFφand(Dξ)αFφ(ξ)=F((x)αφ)DξαFφ(ξ)=F(x)αφ(ℏD_(xi))^(alpha)F_(ℏ)varphi(xi)=F_(ℏ)((-x)^(alpha)varphi)\left(\hbar D_{\xi}\right)^{\alpha} \mathcal{F}_{\hbar} \varphi(\xi)=\mathcal{F}_{\hbar}\left((-x)^{\alpha} \varphi\right)(Dξ)αFφ(ξ)=F((x)αφ).
B.1. Symbol classes. FormRmRm inRm \in \mathbb{R}mRthe symbol classSmSmS^(m)S^{m}Smconsists of functionsa(x,ξ,)C(Rn×Rn)a(x,ξ,)CRn×Rna(x,xi,ℏ)inC^(oo)(R^(n)xxR^(n))a(x, \xi, \hbar) \in C^{\infty}\left(\mathbb{R}^{n} \times \mathbb{R}^{n}\right)a(x,ξ,)C(Rn×Rn)such that for all multi-indicesα,α~Nnα,α~Nnalpha, tilde(alpha)inN^(n)\alpha, \tilde{\alpha} \in \mathbb{N}^{n}α,α~Nnthere exists a constantCα,α~>0Cα,α~>0C_(alpha, tilde(alpha)) > 0C_{\alpha, \tilde{\alpha}}>0Cα,α~>0uniform in(0,0)0,0ℏin(0,ℏ_(0))\hbar \in\left(0, \hbar_{0}\right)(0,0)such that
|xαξα~a(x,ξ,)|Cα,α~ξm|α~|,xRn,ξRnxαξα~a(x,ξ,)Cα,α~ξm|α~|,xRn,ξRn|del_(x)^(alpha)del_(xi)^( tilde(alpha))a(x,xi,ℏ)| <= C_(alpha, tilde(alpha))(:xi:)^(m-| tilde(alpha)|),quad x inR^(n),xi inR^(n)\left|\partial_{x}^{\alpha} \partial_{\xi}^{\tilde{\alpha}} a(x, \xi, \hbar)\right| \leq C_{\alpha, \tilde{\alpha}}\langle\xi\rangle^{m-|\tilde{\alpha}|}, \quad x \in \mathbb{R}^{n}, \xi \in \mathbb{R}^{n}|xαξα~a(x,ξ,)|Cα,α~ξm|α~|,xRn,ξRn
Symbols inSmSmS^(m)S^{m}Smthus behave roughly as polynomials of degreemmmmm. We write thataNSmaNSma inℏ^(N)S^(m)a \in \hbar^{N} S^{m}aNSmif
|xαξα~a(x,ξ,)|Cα,α~Nξm|α~|,xRn,ξRnxαξα~a(x,ξ,)Cα,α~Nξm|α~|,xRn,ξRn|del_(x)^(alpha)del_(xi)^( tilde(alpha))a(x,xi,ℏ)| <= C_(alpha, tilde(alpha))ℏ^(N)(:xi:)^(m-| tilde(alpha)|),quad x inR^(n),xi inR^(n)\left|\partial_{x}^{\alpha} \partial_{\xi}^{\tilde{\alpha}} a(x, \xi, \hbar)\right| \leq C_{\alpha, \tilde{\alpha}} \hbar^{N}\langle\xi\rangle^{m-|\tilde{\alpha}|}, \quad x \in \mathbb{R}^{n}, \xi \in \mathbb{R}^{n}|xαξα~a(x,ξ,)|Cα,α~Nξm|α~|,xRn,ξRn
Lemma (Asymptotic series). LetmRmRm inRm \in \mathbb{R}mRand the symbolsajSmjajSmja_(j)inS^(m-j)a_{j} \in S^{m-j}ajSmjforj=0,1,j=0,1,j=0,1,dotsj=0,1, \ldotsj=0,1,. Then there exists a symbolaSmaSma inS^(m)a \in S^{m}aSmsuch thataj=0jajaj=0jaja∼sum_(j=0)^(oo)ℏ^(j)a_(j)a \sim \sum_{j=0}^{\infty} \hbar^{j} a_{j}aj=0jaj, that is for everyNNNNN inNN \in \mathbb{N}NN,
aj=0NjajN+1SmN1aj=0NjajN+1SmN1a-sum_(j=0)^(N)ℏ^(j)a_(j)inℏ^(N+1)S^(m-N-1)a-\sum_{j=0}^{N} \hbar^{j} a_{j} \in \hbar^{N+1} S^{m-N-1}aj=0NjajN+1SmN1
The symbol a is unique up toSSℏ^(oo)S^(-oo)\hbar^{\infty} S^{-\infty}S, in the sense that the difference of two such symbols is inNSMNSMℏ^(N)S^(-M)\hbar^{N} S^{-M}NSMfor allN,MRN,MRN,M inRN, M \in \mathbb{R}N,MR. The principal symbol ofaaaaais given bya0a0a_(0)a_{0}a0.
B.2. Pseudodifferential operators. Using these symbol classes we can define semiclassical pseudodifferential operators (ψDOsψDOspsiDOs\psi \mathrm{DOs}ψDOs). For a symbolaSmaSma inS^(m)a \in S^{m}aSmwe define the corresponding semiclassicalψDOψDOpsiDO\psi \mathrm{DO}ψDOof orderm,Op(a):S(Rn)S(Rn)m,Op(a):SRnSRnm,Op(a):S(R^(n))rarrS(R^(n))m, O p(a): \mathcal{S}\left(\mathbb{R}^{n}\right) \rightarrow \mathcal{S}\left(\mathbb{R}^{n}\right)m,Op(a):S(Rn)S(Rn),
Op(a)u(x):=1(2π)nRnRnei(xy)ξa(x,ξ,)u(y)dydξOp(a)u(x):=1(2π)nRnRnei(xy)ξa(x,ξ,)u(y)dydξOp(a)u(x):=(1)/((2piℏ)^(n))int_(R^(n))int_(R^(n))e^((i)/(ℏ)(x-y)*xi)a(x,xi,ℏ)u(y)dydxiO p(a) u(x):=\frac{1}{(2 \pi \hbar)^{n}} \int_{\mathbb{R}^{n}} \int_{\mathbb{R}^{n}} e^{\frac{i}{\hbar}(x-y) \cdot \xi} a(x, \xi, \hbar) u(y) \mathrm{d} y \mathrm{~d} \xiOp(a)u(x):=1(2π)nRnRnei(xy)ξa(x,ξ,)u(y)dy dξ
This is also called quantization of the symbol.Op(a)Op(a)Op(a)O p(a)Op(a)can be extended toS(Rn)SRnS^(')(R^(n))\mathcal{S}^{\prime}\left(\mathbb{R}^{n}\right)S(Rn)andOp(a):S(Rn)S(Rn)Op(a):SRnSRnOp(a):S^(')(R^(n))rarrS^(')(R^(n))O p(a): \mathcal{S}^{\prime}\left(\mathbb{R}^{n}\right) \rightarrow \mathcal{S}^{\prime}\left(\mathbb{R}^{n}\right)Op(a):S(Rn)S(Rn)continuously. Note thatOp(a)u(x)=F1(a(x,)Fu())Op(a)u(x)=F1a(x,)Fu()Op(a)u(x)=F_(ℏ)^(-1)(a(x,*)F_(ℏ)u(*))O p(a) u(x)=\mathcal{F}_{\hbar}^{-1}\left(a(x, \cdot) \mathcal{F}_{\hbar} u(\cdot)\right)Op(a)u(x)=F1(a(x,)Fu())and that the operator corresponding to the symbola(x,ξ)=|α|Naα(x)ξαa(x,ξ)=|α|Naα(x)ξαa(x,xi)=sum_(|alpha| <= N)a_(alpha)(x)xi^(alpha)a(x, \xi)=\sum_{|\alpha| \leq N} a_{\alpha}(x) \xi^{\alpha}a(x,ξ)=|α|Naα(x)ξαisOp(a)u=|α|Naα(x)(D)αuOp(a)u=|α|Naα(x)(D)αuOp(a)u=sum_(|alpha| <= N)a_(alpha)(x)(ℏD)^(alpha)uO p(a) u= \sum_{|\alpha| \leq N} a_{\alpha}(x)(\hbar D)^{\alpha} uOp(a)u=|α|Naα(x)(D)αu. Notice that each derivative of this operator scales with\hbar.
For the present paper the most important example is the second order differential operatorA=2Δ+2j=1nβj(x)jA=2Δ+2j=1nβj(x)jA=-ℏ^(2)Delta+ℏ^(2)sum_(j=1)^(n)beta_(j)(x)del_(j)A=-\hbar^{2} \Delta+\hbar^{2} \sum_{j=1}^{n} \beta_{j}(x) \partial_{j}A=2Δ+2j=1nβj(x)j. Its symbol is given bya(x,ξ,)=|ξ|2+ij=1nβj(x)ξja(x,ξ,)=|ξ|2+ij=1nβj(x)ξja(x,xi,ℏ)=|xi|^(2)+iℏsum_(j=1)^(n)beta_(j)(x)xi_(j)a(x, \xi, \hbar)=|\xi|^{2}+i \hbar \sum_{j=1}^{n} \beta_{j}(x) \xi_{j}a(x,ξ,)=|ξ|2+ij=1nβj(x)ξj, and its principal symbol isa0(x,ξ,)=|ξ|2a0(x,ξ,)=|ξ|2a_(0)(x,xi,ℏ)=|xi|^(2)a_{0}(x, \xi, \hbar)=|\xi|^{2}a0(x,ξ,)=|ξ|2.
B.3. Semiclassical Sobolev spaces. ForsRsRs inRs \in \mathbb{R}sRthe semiclassical Sobolev spacesHscls(Rn)HsclsRnH_(scl)^(s)(R^(n))H_{\mathrm{scl}}^{s}\left(\mathbb{R}^{n}\right)Hscls(Rn)are algebraically equal to the standard Sobolev spacesHs(Rn)HsRnH^(s)(R^(n))H^{s}\left(\mathbb{R}^{n}\right)Hs(Rn)but are endowed with different norms
uHscls(Rn)=JsuL2(Rn)uHsclsRn=JsuL2Rn||u||_(H_(scl)^(s)(R^(n)))=||J^(s)u||_(L^(2)(R^(n)))\|u\|_{H_{\mathrm{scl}}^{s}\left(\mathbb{R}^{n}\right)}=\left\|J^{s} u\right\|_{L^{2}\left(\mathbb{R}^{n}\right)}uHscls(Rn)=JsuL2(Rn)
where the semiclassical Bessel potential is defined byJs=Op(ξs)Js=OpξsJ^(s)=Op((:xi:)^(s))J^{s}=O p\left(\langle\xi\rangle^{s}\right)Js=Op(ξs). Informally,
Js=(12Δ)s/2,sRJs=12Δs/2,sRJ^(s)=(1-ℏ^(2)Delta)^(s//2),quad s inRJ^{s}=\left(1-\hbar^{2} \Delta\right)^{s / 2}, \quad s \in \mathbb{R}Js=(12Δ)s/2,sR
For example,uHscl1(Rn)2=uL2(Rn)2+uL2(Rn)2uHscl1Rn2=uL2Rn2+uL2Rn2||u||_(H_(scl)^(1)(R^(n)))^(2)=||u||_(L^(2)(R^(n)))^(2)+||ℏgrad u||_(L^(2)(R^(n)))^(2)\|u\|_{H_{\mathrm{scl}}^{1}\left(\mathbb{R}^{n}\right)}^{2}=\|u\|_{L^{2}\left(\mathbb{R}^{n}\right)}^{2}+\|\hbar \nabla u\|_{L^{2}\left(\mathbb{R}^{n}\right)}^{2}uHscl1(Rn)2=uL2(Rn)2+uL2(Rn)2. A semiclassicalψDOψDOpsiDO\psi \mathrm{DO}ψDOof ordermmmmmis continuous fromHscls(Rn)HsclsRnH_(scl)^(s)(R^(n))H_{\mathrm{scl}}^{s}\left(\mathbb{R}^{n}\right)Hscls(Rn)toHsclsm(Rn)HsclsmRnH_(scl)^(s-m)(R^(n))H_{\mathrm{scl}}^{s-m}\left(\mathbb{R}^{n}\right)Hsclsm(Rn).
B.4. Composition. Composition of semiclassicalψDOψDOpsiDO\psi \mathrm{DO}ψDOs can be analysed using the following calculus.
Theorem (Symbol calculus). LetaSmaSma inS^(m)a \in S^{m}aSmandbSmbSmb inS^(m^('))b \in S^{m^{\prime}}bSm. ThenOp(a)Op(b)=Op(a#b)Op(a)Op(b)=Op(a#b)Op(a)@Op(b)=Op(a#b)O p(a) \circ O p(b)=O p(a \# b)Op(a)Op(b)=Op(a#b)for a certaina#bSm+ma#bSm+ma#b inS^(m+m^('))a \# b \in S^{m+m^{\prime}}a#bSm+mthat admits the following asymptotic series
a#b(x,ξ,)α|α|i|α|α!Dξαa(x,ξ,)Dxαb(x,ξ,)a#b(x,ξ,)α|α|i|α|α!Dξαa(x,ξ,)Dxαb(x,ξ,)a#b(x,xi,ℏ)∼sum_(alpha)(ℏ^(|alpha|)i^(|alpha|))/(alpha!)D_(xi)^(alpha)a(x,xi,ℏ)D_(x)^(alpha)b(x,xi,ℏ)a \# b(x, \xi, \hbar) \sim \sum_{\alpha} \frac{\hbar^{|\alpha|} i^{|\alpha|}}{\alpha!} D_{\xi}^{\alpha} a(x, \xi, \hbar) D_{x}^{\alpha} b(x, \xi, \hbar)a#b(x,ξ,)α|α|i|α|α!Dξαa(x,ξ,)Dxαb(x,ξ,)
The commutator and disjoint support estimates (4) and (5) follow, respectively, from the following.
Corollary (Commutator and disjoint support). LetaSmaSma inS^(m)a \in S^{m}aSmandbSmbSmb inS^(m^('))b \in S^{m^{\prime}}bSm. Then
(i)a#bb#aSm+m1a#bb#aSm+m1a#b-b#a inℏS^(m+m^(')-1)a \# b-b \# a \in \hbar S^{m+m^{\prime}-1}a#bb#aSm+m1.
(ii) Ifsupp(a)supp(b)=supp(a)supp(b)=supp(a)nn supp(b)=O/\operatorname{supp}(a) \cap \operatorname{supp}(b)=\emptysetsupp(a)supp(b)=, thena#bSa#bSa#b inℏ^(oo)S^(-oo)a \# b \in \hbar^{\infty} S^{-\infty}a#bS, i.e.a#bNSMa#bNSMa#b inℏ^(N)S^(-M)a \# b \in \hbar^{N} S^{-M}a#bNSMfor allN,MRN,MRN,M inRN, M \in \mathbb{R}N,MR.
Proof. (i) The principal symbol ofa#ba#ba#ba \# ba#b, that is the first term in its asymptotic series, isabababa bab. The second term isij=1nξja(x,ξ,)xjb(x,ξ,)ij=1nξja(x,ξ,)xjb(x,ξ,)(ℏ)/(i)sum_(j=1)^(n)del_(xi_(j))a(x,xi,ℏ)del_(x_(j))b(x,xi,ℏ)\frac{\hbar}{i} \sum_{j=1}^{n} \partial_{\xi_{j}} a(x, \xi, \hbar) \partial_{x_{j}} b(x, \xi, \hbar)ij=1nξja(x,ξ,)xjb(x,ξ,). We thus have that the principal symbol of the commutator[Op(a),Op(b)]=Op(a#bb#a)[Op(a),Op(b)]=Op(a#bb#a)[Op(a),Op(b)]=Op(a#b-b#a)[O p(a), O p(b)]=O p(a \# b-b \# a)[Op(a),Op(b)]=Op(a#bb#a)is given by
ij=1n(ξjaxjbxjaξjb)Sm+m1ij=1nξjaxjbxjaξjbSm+m1(ℏ)/(i)sum_(j=1)^(n)(del_(xi_(j))adel_(x_(j))b-del_(x_(j))adel_(xi_(j))b)inℏS^(m+m^(')-1)\frac{\hbar}{i} \sum_{j=1}^{n}\left(\partial_{\xi_{j}} a \partial_{x_{j}} b-\partial_{x_{j}} a \partial_{\xi_{j}} b\right) \in \hbar S^{m+m^{\prime}-1}ij=1n(ξjaxjbxjaξjb)Sm+m1
(ii) Ifsupp(a)supp(b)=supp(a)supp(b)=supp(a)nn supp(b)=O/\operatorname{supp}(a) \cap \operatorname{supp}(b)=\emptysetsupp(a)supp(b)=, then each term in the asymptotic series ofa#ba#ba#ba \# ba#bvanishes.

Acknowledgements

E.B. was supported by EPSRC grants EP/P01576X/1 and EP/P012434/1, and L.O. by EPSRC grants EP/P01593X/1 and EP/R002207/1.

References

[ARRV09]G. Alessandrini, L. Rondi, E. Rosset, and S. Vessella. The stability for the Cauchy problem for elliptic equations. Inverse Problems, 25:123004, 2009.
[BHL18]E. Burman, P. Hansbo, and M. G. Larson. Solving ill-posed control problems by stabilized finite element methods: an alternative to Tikhonov regularization. Inverse Problems, 34:035004, 2018.
[BNO19a]E. Burman, M. Nechita, and L. Oksanen. A stabilized finite element method for inverse problems subject to the convection-diffusion equation. II: convection-dominated regime. in preparation, 2019.
[BNO19b]E. Burman, M. Nechita, and L. Oksanen. Unique continuation for the Helmholtz equation using stabilized finite element methods. J. Math. Pures Appl., 129:1-24, 2019.
[BO18]E. Burman and L. Oksanen. Data assimilation for the heat equation using stabilized finite element methods. Numer. Math., 139(3):505-528, 2018.
[Bre03]S. C. Brenner. Poincaré-Friedrichs inequalities for piecewiseH1H1H^(1)H^{1}H1functions. SIAM J. Numer. Anal., 41(1):306-324, 2003.
[Bur05]E. Burman. A unified analysis for conforming and nonconforming stabilized finite element methods using interior penalty. SIAM J. Numer. Anal., 43(5):2012-2033, 2005.
[Bur13]E. Burman. Stabilized finite element methods for nonsymmetric, noncoercive, and ill-posed problems. Part I: elliptic equations. SIAM J. Sci. Comput., 35(6):A2752-A2780, 2013.
[Bur14]E. Burman. Error estimates for stabilized finite element methods applied to ill-posed problems. C. R. Math. Acad. Sci. Paris, 352(7-8):655-659, 2014.
[BV07]R. Becker and B. Vexler. Optimal control of the convection-diffusion equation using stabilized finite element methods. Numer. Math., 106(3):349-367, 2007.
[DQ05]L. Dede' and A. Quarteroni. Optimal control and numerical adaptivity for advectiondiffusion equations. M2AN Math. Model. Numer. Anal., 39(5):1019-1040, 2005.
[DSFKSU09]D. Dos Santos Ferreira, C. E. Kenig, M. Salo, and G. Uhlmann. Limiting Carleman weights and anisotropic inverse problems. Invent. Math., 178(1):119-171, 2009.
[EG04]A. Ern and J.-L. Guermond. Theory and practice of finite elements, volume 159 of Applied Mathematical Sciences. Springer-Verlag, New York, 2004.
[EG06]A. Ern and J.-L. Guermond. Evaluation of the condition number in linear systems arising in finite element approximations. M2AN Math. Model. Numer. Anal., 40(1):29-48, 2006.
[Hec12]F. Hecht. New development in FreeFem++. J. Numer. Math., 20(3-4):251-265, 2012.
[HYZ09]M. Hinze, N. Yan, and Z. Zhou. Variational discretization for optimal control governed by convection dominated diffusion equations. J. Comput. Math., 27(2-3):237-253, 2009.
[LRL12]
J. Le Rousseau and G. Lebeau. On Carleman estimates for elliptic and parabolic operators. Applications to unique continuation and control of parabolic equations. ESAIM Control Optim. Calc. Var., 18(3):712-747, 2012.
[MS99]P. Monk and E. Süli. The adaptive computation of far-field patterns by a posteriori error estimation of linear functionals. SIAM J. Numer. Anal., 36(1):251-274, 1999.
[MV12]E. Malinnikova and S. Vessella. Quantitative uniqueness for elliptic equations with singular lower order terms. Math. Ann., 353(4):1157-1181, 2012.
[YZ09]N. Yan and Z. Zhou. A priori and a posteriori error analysis of edge stabilization Galerkin method for the optimal control problem governed by convection-dominated diffusion equation. J. Comput. Appl. Math., 223(1):198-217, 2009.
[Zwo12]M. Zworski. Semiclassical analysis, volume 138 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2012.
[ARRV09] G. Alessandrini, L. Rondi, E. Rosset, and S. Vessella. The stability for the Cauchy problem for elliptic equations. Inverse Problems, 25:123004, 2009. [BHL18] E. Burman, P. Hansbo, and M. G. Larson. Solving ill-posed control problems by stabilized finite element methods: an alternative to Tikhonov regularization. Inverse Problems, 34:035004, 2018. [BNO19a] E. Burman, M. Nechita, and L. Oksanen. A stabilized finite element method for inverse problems subject to the convection-diffusion equation. II: convection-dominated regime. in preparation, 2019. [BNO19b] E. Burman, M. Nechita, and L. Oksanen. Unique continuation for the Helmholtz equation using stabilized finite element methods. J. Math. Pures Appl., 129:1-24, 2019. [BO18] E. Burman and L. Oksanen. Data assimilation for the heat equation using stabilized finite element methods. Numer. Math., 139(3):505-528, 2018. [Bre03] S. C. Brenner. Poincaré-Friedrichs inequalities for piecewise H^(1) functions. SIAM J. Numer. Anal., 41(1):306-324, 2003. [Bur05] E. Burman. A unified analysis for conforming and nonconforming stabilized finite element methods using interior penalty. SIAM J. Numer. Anal., 43(5):2012-2033, 2005. [Bur13] E. Burman. Stabilized finite element methods for nonsymmetric, noncoercive, and ill-posed problems. Part I: elliptic equations. SIAM J. Sci. Comput., 35(6):A2752-A2780, 2013. [Bur14] E. Burman. Error estimates for stabilized finite element methods applied to ill-posed problems. C. R. Math. Acad. Sci. Paris, 352(7-8):655-659, 2014. [BV07] R. Becker and B. Vexler. Optimal control of the convection-diffusion equation using stabilized finite element methods. Numer. Math., 106(3):349-367, 2007. [DQ05] L. Dede' and A. Quarteroni. Optimal control and numerical adaptivity for advectiondiffusion equations. M2AN Math. Model. Numer. Anal., 39(5):1019-1040, 2005. [DSFKSU09] D. Dos Santos Ferreira, C. E. Kenig, M. Salo, and G. Uhlmann. Limiting Carleman weights and anisotropic inverse problems. Invent. Math., 178(1):119-171, 2009. [EG04] A. Ern and J.-L. Guermond. Theory and practice of finite elements, volume 159 of Applied Mathematical Sciences. Springer-Verlag, New York, 2004. [EG06] A. Ern and J.-L. Guermond. Evaluation of the condition number in linear systems arising in finite element approximations. M2AN Math. Model. Numer. Anal., 40(1):29-48, 2006. [Hec12] F. Hecht. New development in FreeFem++. J. Numer. Math., 20(3-4):251-265, 2012. [HYZ09] M. Hinze, N. Yan, and Z. Zhou. Variational discretization for optimal control governed by convection dominated diffusion equations. J. Comput. Math., 27(2-3):237-253, 2009. [LRL12] J. Le Rousseau and G. Lebeau. On Carleman estimates for elliptic and parabolic operators. Applications to unique continuation and control of parabolic equations. ESAIM Control Optim. Calc. Var., 18(3):712-747, 2012. [MS99] P. Monk and E. Süli. The adaptive computation of far-field patterns by a posteriori error estimation of linear functionals. SIAM J. Numer. Anal., 36(1):251-274, 1999. [MV12] E. Malinnikova and S. Vessella. Quantitative uniqueness for elliptic equations with singular lower order terms. Math. Ann., 353(4):1157-1181, 2012. [YZ09] N. Yan and Z. Zhou. A priori and a posteriori error analysis of edge stabilization Galerkin method for the optimal control problem governed by convection-dominated diffusion equation. J. Comput. Appl. Math., 223(1):198-217, 2009. [Zwo12] M. Zworski. Semiclassical analysis, volume 138 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2012.| [ARRV09] | G. Alessandrini, L. Rondi, E. Rosset, and S. Vessella. The stability for the Cauchy problem for elliptic equations. Inverse Problems, 25:123004, 2009. | | :--- | :--- | | [BHL18] | E. Burman, P. Hansbo, and M. G. Larson. Solving ill-posed control problems by stabilized finite element methods: an alternative to Tikhonov regularization. Inverse Problems, 34:035004, 2018. | | [BNO19a] | E. Burman, M. Nechita, and L. Oksanen. A stabilized finite element method for inverse problems subject to the convection-diffusion equation. II: convection-dominated regime. in preparation, 2019. | | [BNO19b] | E. Burman, M. Nechita, and L. Oksanen. Unique continuation for the Helmholtz equation using stabilized finite element methods. J. Math. Pures Appl., 129:1-24, 2019. | | [BO18] | E. Burman and L. Oksanen. Data assimilation for the heat equation using stabilized finite element methods. Numer. Math., 139(3):505-528, 2018. | | [Bre03] | S. C. Brenner. Poincaré-Friedrichs inequalities for piecewise $H^{1}$ functions. SIAM J. Numer. Anal., 41(1):306-324, 2003. | | [Bur05] | E. Burman. A unified analysis for conforming and nonconforming stabilized finite element methods using interior penalty. SIAM J. Numer. Anal., 43(5):2012-2033, 2005. | | [Bur13] | E. Burman. Stabilized finite element methods for nonsymmetric, noncoercive, and ill-posed problems. Part I: elliptic equations. SIAM J. Sci. Comput., 35(6):A2752-A2780, 2013. | | [Bur14] | E. Burman. Error estimates for stabilized finite element methods applied to ill-posed problems. C. R. Math. Acad. Sci. Paris, 352(7-8):655-659, 2014. | | [BV07] | R. Becker and B. Vexler. Optimal control of the convection-diffusion equation using stabilized finite element methods. Numer. Math., 106(3):349-367, 2007. | | [DQ05] | L. Dede' and A. Quarteroni. Optimal control and numerical adaptivity for advectiondiffusion equations. M2AN Math. Model. Numer. Anal., 39(5):1019-1040, 2005. | | [DSFKSU09] | D. Dos Santos Ferreira, C. E. Kenig, M. Salo, and G. Uhlmann. Limiting Carleman weights and anisotropic inverse problems. Invent. Math., 178(1):119-171, 2009. | | [EG04] | A. Ern and J.-L. Guermond. Theory and practice of finite elements, volume 159 of Applied Mathematical Sciences. Springer-Verlag, New York, 2004. | | [EG06] | A. Ern and J.-L. Guermond. Evaluation of the condition number in linear systems arising in finite element approximations. M2AN Math. Model. Numer. Anal., 40(1):29-48, 2006. | | [Hec12] | F. Hecht. New development in FreeFem++. J. Numer. Math., 20(3-4):251-265, 2012. | | [HYZ09] | M. Hinze, N. Yan, and Z. Zhou. Variational discretization for optimal control governed by convection dominated diffusion equations. J. Comput. Math., 27(2-3):237-253, 2009. | | [LRL12] | | | | J. Le Rousseau and G. Lebeau. On Carleman estimates for elliptic and parabolic operators. Applications to unique continuation and control of parabolic equations. ESAIM Control Optim. Calc. Var., 18(3):712-747, 2012. | | [MS99] | P. Monk and E. Süli. The adaptive computation of far-field patterns by a posteriori error estimation of linear functionals. SIAM J. Numer. Anal., 36(1):251-274, 1999. | | [MV12] | E. Malinnikova and S. Vessella. Quantitative uniqueness for elliptic equations with singular lower order terms. Math. Ann., 353(4):1157-1181, 2012. | | [YZ09] | N. Yan and Z. Zhou. A priori and a posteriori error analysis of edge stabilization Galerkin method for the optimal control problem governed by convection-dominated diffusion equation. J. Comput. Appl. Math., 223(1):198-217, 2009. | | [Zwo12] | M. Zworski. Semiclassical analysis, volume 138 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2012. |

  1. Department of Mathematics, University College London, Gower Street, London UK, WC1E 6BT.
    E-mail addresses: {e.burman, mihai.nechita.16, l.oksanen}@ucl.ac.uk.
    Date: October 8, 2019.
2020

Related Posts