Manuscriptsubmittedto doi:10.3934/xx.xx.xx.xx AIMS’Journals VolumeX,Number0X,XX200X pp. X–XX A CONTROL APPROACH TO RECOVER THE WAVE SPEED (CONFORMAL FACTOR) FROM ONE MEASUREMENT 5 1 Sebastian Acosta 0 2 DepartmentofPediatricCardiology BaylorCollegeofMedicine,Houston,TX,USA n a (Communicated by the associate editor name) J 8 Abstract. Inthispaperweconsidertheproblemofrecoveringtheconformal ] factor in a conformal class of Riemannian metrics from the boundary mea- P surementofonewavefield. Moreprecisely,usingboundarycontroloperators, A wederiveanexplicitequationsatisfiedbythecontrastbetweentwoconformal factors(orwavespeeds). ThisequationisFredholmandgenericallyinvertible . h providedthatthedomainofinterestisproperlyilluminatedataninitialtime. t WealsoshowlocallyLipschitzstabilityestimates. a m 1. Introduction. In some applications, such as imaging of soft biological tissues, [ the propagation of pressure waves can be modeled by the acoustic wave equation. 2 Then one seeks to reconstructthe wave speed within a regionof interest. However, v in some cases, it is important to account for the anisotropic properties of media 7 (muscles and bones). Although these are more appropriately modeled by elasto- 3 dynamic systems, the anisotropic scalar wave equation may serve as a first step to 7 6 incorporate such behavior. This equation can be expressed in geometric terms as . follows, 1 0 ∂t2w =Ac−2gw, (1) 4 1 where Ag is the µ-weighted Laplace-Beltrami operator defined as v: Agw =µ−1divg(µgradgw). (2) i X Hereµ(x) andc(x) aresufficiently smoothandpositive, anddiv andgrad denote g g r the divergence and the gradientwith respect to a smooth Riemannian metric g. In a this paper, we assume full knowledge of the coefficient µ and the metric g. The inverseproblemweconsideristo recoverthe conformalfactorinthe confor- malclassofmetricsrepresentedbyg,thatis,weseektorecoverthewavespeedc(x) fromknowledgeofboundarydataofw onasufficientlylargetimeinterval. Abusing terminology, we shall refer to c(x) as the wave speed even though the actual wave speed could be anisotropic and would be fully characterized by both c and g. In the literature, most works concerned with the recovery of coefficients in the wave equation from boundary measurements are limited to one of the following cases: 2010 Mathematics Subject Classification. Primary: 35L05,35R30;Secondary: 35Q93. Keywordsandphrases. waveequation,acoustics,geometricinverseproblem,imaging,control theory,singlemeasurement, anisotropicmedia. The author was partially supported by AFOSR Grant FA9550-12-1-0117 and ONR Grant N00014-12-1-0256. 1 2 SEBASTIAN ACOSTA (i) Knowledge of the hyperbolic Dirichlet-to-Neumann map. Thus, it is usually assumed that infinitely many illuminations and corresponding measurements are known. Here we find approaches based on special solutions and integral geometry (see for instance [37, 42, 43, 39, 44, 40, 6, 29] and[21, Section8.3]), and the boundary control method (see [4, 5] and [21, Section 8.4]). (ii) Knowledge of initial data and the corresponding boundary measurement. These are results inthe generalformofTheorem8.2.2in[21] originatedfrom [10]. See also [24, 36, 35, 45, 19, 20, 25, 7, 8, 41, 31, 32, 33, 30] andreferences therein. This paper falls into the secondcategory. We summarize the novelty andcontri- butions of our work in the following points: (a) Thewavespeedc(x)appearswithintheprincipalpartofthehyperbolicequa- tionwhichleadstoachallenginginterplaybetweenc(x)anditsderivatives. So for instance, our approach can be employed to determine a scalar coefficient p(x) in ∇·p∇ treated in [20, 25]. Our work can be seen as a generalization of [20, 25] and other cited references to non-Euclidean geometries. (b) Usingcontroloperators,wederiveanexplicit equationsatisfiedbythecontrast between two wave speeds. This equation has Fredholm form provided that the domainofinterestis properlyilluminatedataninitialtime. Seethe main results in Section 3 for the precise statements. In dimension n = 2, this Fredholm equation has a principal part of order zero (see Theorem 1). In dimension n ≥ 3, we can derive an analogous equation using finitely many illuminations (see Theorem 2). Otherwise, with a single illumination, we obtain a first-order PDE. We give conditions on the initial illumination for this first-order equation to be Fredholm (see Theorem 3). (c) The main restriction is that the initial profile w| is assumed to be known t=0 a-priori. However, in contrast to most previous publications, we make no assumptions about the initial velocity profile ∂ w| . t t=0 (d) Ifwell-knowngeometricalconditionsaresatisfied,thiscontrolapproachisnat- urally suited for measurements on a sub-boundary Γ ⊂ ∂Ω. See Assumption 2.1 below. We wish to acknowledge that our work is inspired by a combination of ideas developed in Puel–Yamamoto [35], Liu–Oksanen [29] and Stefanov–Uhlmann [41]. AbriefreviewofcontroltheoreticaltoolsispresentedinSection2. Ourmainresults are stated in Section 3 and the proofs are provided in Section 4. 2. BackgroundonControlTheory. Ourapproachreliesheavilyonexactbound- ary controllability for the wave equation. Hence, the purpose of this section is to reviewsome facts anddefine the notationconcerningcontroltheoreticaltools. Our guiding references are [18, 3, 14, 27]. The first item is to ensure that the media under consideration allows for exact controllability which is closely related to geo- metricnotions[18,3,14]. Weconsiderthewaveequation(1)definedinasufficiently smooth simply connected domain Ω⊂Rn with boundary ∂Ω where n≥2, and the geodesic flow associated with this wave equation is determined by the metric c−2g because, Ac−2gw =∆c−2gw + lower order terms, where ∆c−2g is the Laplace-Beltramioperator corresponding to the metric c−2g. A CONTROL APPROACH TO RECOVER THE WAVE SPEED 3 FollowingBardos,LebeauandRauch[3],weassumethatourproblemenjoysthe geometrical control condition (GCC) for the Riemannian manifold (Ω,c−2g) with only a portion Γ of the boundary ∂Ω being accessible for control and observation. WeassumethatΓisasmoothandsimply connecteddomainrelativeto ∂Ω. Inthis paper, we assume that the following condition is satisfied. Assumption 2.1 (GCC). The geodesic flow of (Ω,c−2g) reaches the accessible part of the boundary. In other words, there exists τ <∞ such that any unit-speed ray, originating from any point in Ω at t=0, reaches Γ in a nondiffractive manner (after possible geometrical reflections on ∂Ω\Γ) before time t=τ. Tofixourideas,wewillposethecontrolproblemwithaspecificlevelofregularity. However,we should keep in mind that this problem can also be formulated on any scale of Sobolev regularity [3]. In addition, we should interpret the Hilbert space H0(Ω) with the inner-product appropriately weighted by µdet(c−2g)1/2 so that Ac−2g is formally self-adjoint with respect to the duality pairing of H0(Ω). The same weight is incorporated in the inner product for H0(∂Ω). Now we consider the following auxiliary problem. Given ζ ∈ H1((0,τ) ×Γ), 0 find the generalized solution ξ ∈ Ck([0,τ];H1−k(Ω)) for k = 0,1 of the following time-reversedproblem ∂t2ξ−Ac−2gξ =0 in (0,τ)×Ω (3a) ξ =0 and ∂ ξ =0 on {t=τ}×Ω (3b) t ξ =ζ on (0,τ)×Γ (3c) ξ =0 on (0,τ)×(∂Ω\Γ) (3d) This is a well-posed problem satisfying a stability estimate of the following form, kξkCk([0,τ];H1−k(Ω))+k∇c−2gξkCk([0,τ];H−k(Ω)) .kζkH01((0,τ)×Γ). for k = 0,1. Here and in the rest of the paper, the symbol . means inequality up to a positive constant. Given arbitraryφ∈H0(Ω), the goalofthe controlproblemis to find a Dirichlet boundary conditionζ ∈H1((0,τ)×Γ) to drive the solutionξ of (3) fromvanishing 0 Cauchy data at t=τ to the desired Cauchy data (ξ,∂ ξ)=(0,φ) at time t=0. t The well-posedness of this control problem under the geometrical control con- dition was obtained in [3] for smooth data, and extended in [11] for less regular domains and coefficients. For our reference, we state it now as a theorem. Theorem 2.2 (Controllability). Let the geometrical control condition 2.1 hold. Then for any function φ ∈ H0(Ω), there exists Dirichlet boundary control ζ ∈ H1((0,τ)×Γ) so that the solution ξ of (3) satisfies 0 (ξ,∂ ξ)=(0,φ) at time t=0. t Among all such boundary controls there exists ζ which is uniquely determined min by φ as the minimum norm control and satisfies the following stability condition kζminkH01((0,τ)×Γ) ≤CkφkH0(Ω) for some positive constant C =C(Ω,Γ,c,g,τ). Notice that we drive the wavefield from vanishing Cauchy data at time t=τ to Cauchy data at time t=0 of the form (ξ,∂ ξ)=(0,φ). Now we define a particular t waytodrivethesolutiontoaCauchydataattimet=0oftheform(ξ,∂ ξ)=(φ,0). t 4 SEBASTIAN ACOSTA Definition 2.3 (Control Operators). Let the control operator C :H0(Ω)→H0((0,τ)×Γ), be given by the map φ 7→ ∂ ζ where ζ is defined in Theorem 2.2. We also t min min define the corresponding solution operator S :H0(Ω)→H0((0,τ)×Ω), mapping φ7→∂ ξ where ξ is the solution of (3) with ζ =ζ . t min The control andsolution operatorsgivenby Definition2.3 arepurposelydefined as the time derivative of more regular functions so that they enjoy a gain of space regularity with integration in time. Denote by ∂−1 the following operation t t (∂−1v)(t)= v(s)ds. t Z 0 Usingtheabovenotation,weseethat∂−1∂ ξ =ξbecauseξ| =0,and∂−1∂ ζ =ζ t t t=0 t t because ζ| = 0 since ζ ∈ H1((0,τ)×Γ). This leads to the following properties t=0 0 for the control and solution operators in the form of a lemma, whose proof follows directly from the conclusions in Theorem 2.2 and well-known results [28, 26]. Lemma 2.4. The control and solution operators from Definition 2.3 are bounded. Moreover, for given φ ∈ H0(Ω) we have that Dirichet control η = Cφ drives the generalized solution ψ = Sφ of (3) from vanishing Cauchy data at time t = τ to Cauchy data (ψ,∂ ψ) = (φ,0) at time t = 0. Also, the wave field ψ = Sφ satisfies t the following estimate, kψkCk([0,τ];H−k(Ω))+k∇c−2gψkC([0,τ];H−k−1(Ω)) .kηkH0((0,τ)×Γ), (4) for k =0,1, and k∂t−1ψkCk([0,τ];H1−k(Ω))+k∇c−2g∂t−1ψkCk([0,τ];H−k(Ω)) .k∂t−1ηkH01((0,τ)×Γ) (5) for k =0,1. 3. Inverse Problem and Recovery Equation. In this section we state the in- verse problem for the recovery of the wave speed, and we also present our main results. We start by defining the forward problem. Let w be the strong solution of the following problem, ∂t2w−Ac−2gw =0 in (0,τ)×Ω, (6a) w =α and ∂ w=β on {t=0}×Ω, (6b) t w =γ on (0,τ)×Γ, (6c) w =0 on (0,τ)×(∂Ω\Γ). (6d) We also define w˜ to solve a wave problem with potentially different speed c˜, and initial Cauchy data (w˜,∂ w˜)| = (α,β˜) and the same Dirichlet boundary values t t=0 w˜ = γ on (0,τ)×Γ. Notice that β and β˜ do not have to coincide. We define the Neumann traces λ and λ˜ of w and w˜, respectively, as follows λ=ν·gradc−2gw and λ˜ =ν·gradc−2gw˜ on ((0,τ)×Γ), where ν denotes the outward normal on ∂Ω. Now we realize that the governing operator in (1) can be expressed as follows, 2−n Ac−2gw=c2Agw+ 2 ∇c2·∇gw (7) A CONTROL APPROACH TO RECOVER THE WAVE SPEED 5 where ∇ is the Euclidean gradient. In dimension n = 2, we have the convenient identity Ac−2gw = c2Agw. However, we consider the general case n ≥ 2. In view of the identity (7) and in order to obtain a linear inversesource problem, we define u=w−w˜, f =c2−c˜2, (8a) 2−n σ =A w˜, and Σ= ∇ w˜. (8b) g g 2 Hence we have that u is the strong solution of the following problem, ∂t2u−Ac−2gu=σf +Σ·∇f in (0,τ)×Ω (9a) u=0 and ∂ u=β−β˜ on {t=0}×Ω (9b) t u=0 on (0,τ)×∂Ω. (9c) Recall that we measure the Neumann data m=λ−λ˜ =ν·grad u, on ((0,τ)×Γ) (10) c−2g Withthisnotationwecanpreciselystatethe inverseproblemforthedetermination of f which is the contrast between the two conformal factors. Problem (Inverse Problem). For an appropriate initial illumination α, show that if the Neumann traces λ and λ˜ coincide, then c = c˜. Or by conversion into a linearized inverse source problem, given the Neumann boundary measurement m in (10) and the source factors σ and Σ in (9), determine the source term f. In order to state our main results, we define the followingtime integraloperator P :H0((0,τ)×Ω)→H0(Ω) given by τ (Pv)(x)= v(t,x)dt. (11) Z 0 Now notice that u˙ solves a wave problem (in the standard weak sense) with forcing term σ˙f +Σ˙ ·∇f and initial Cauchy data (u˙,∂ u˙)| =(β−β˜,σ| f +Σ| ·∇f). t t=0 t=0 t=0 In what follows,we will evaluate the duality pairing betweenthe time derivative ofthe terms in equation(9a) againstψ, the solutionof (3). Let m be givenby (10) and consider, hσ˙f +Σ˙ ·∇f,ψi(0,τ)×Ω =h∂t2u˙ −Ac−2gu˙,ψi(0,τ)×Ω =−hσ0f +Σ0·∇f,φiΩ−hm˙ ,ηi(0,τ)×Γ whereψ =Sφ andη =Cφ, andu possessesvanishingtrace on(0,τ)×∂Ω. Here we have also used the fact that ψ has zero Cauchy data at time t=τ, and ∂ ψ =0 at t time t = 0. This last fact is very important since it eliminates the unknown term u˙| = β −β˜ from the above equation. Now, since φ ∈ H0(Ω) is arbitrary, then t=0 we obtain the following, ∗ ∗ Σ +(PΣ˙S) ∇f + σ +(Pσ˙S) f =−C∗m˙ , (12) 0 0 h i h i where σ = σ| : H0(Ω) → H0(Ω) is understood as a multiplicative operator 0 t=0 mapping φ 7→ σ φ, which makes it self-adjoint. Similarly, Σ = Σ| : H0(Ω) → 0 0 t=0 H0(Ω)n maps φ7→Σ φ and its adjoint Σ∗ :H0(Ω)n →H0(Ω) maps Φ7→Σ ·Φ. 0 0 0 This last equation is suited for our purposes and it represents the main result of this paper along with statements of its solvability provided by the following theorems. 6 SEBASTIAN ACOSTA Main Theorem 1. Let the geometrical control condition 2.1 hold. If (i) there exists a constant δ >0 such that |σ (x)|≥δ for a.a. x∈Ω, and 0 (ii) σ ∈C1([0,τ];C(Ω)), then the operator σ +(Pσ˙S) :H0(Ω)→H0(Ω) (13) 0 h i is Fredholm. Moreover, there exists an open and dense subset U of (i)∩(ii) such thatfor eachσ inthis set, theoperator (13)andits adjoint areboundedly invertible. Concerning the Inverse Problem 3, in dimension n = 2 where Σ ≡ 0, for each σ ∈ U, the inverse problem is uniquely solvable, and the following locally Lipschitz stability estimate, kc2−c˜2kH0(Ω) ≤Ckλ−λ˜kH1([0,τ];H0(Γ)), holds for a positive constant C that remains uniformly bounded for c and c˜in small bounded sets of C2(Ω). We findappropriatetomakethefollowingremarksconcerningglobaluniqueness in the recovery of the wave speed (conformal factor) from a single measurement. Remark 1. In dimension n=2 we have that Σ≡0. Thus, under the geometrical control condition 2.1, if the initial state α of the forward problem (6) satisfies |A α(x)|≥δ >0, then for a sufficiently regular andgeneric term A α, the Cauchy g g data of w on the sub-boundary ((0,τ)×Γ) uniquely identifies the wavespeed c(x). The positivity condition |A α(x)| ≥ δ > 0 indicates that the region of interest g shouldbeproperlyilluminatedattheinitialtimet=0. Similarpositivityconditions are also found in studies of related inverse problems. See [21, Thm. 8.2.2] and [24, 36, 35, 45, 19, 20, 7, 8, 41, 31, 32, 33, 30] and references therein. Remark 2. Recall the definition of σ in (8b). The regularity condition (ii) of Theorem 1 requires the initial Cauchy data (w˜,∂ w˜)| =(α,β˜) and the Dirichlet t t=0 boundary data γ to be sufficiently regular. We note that if w˜ ∈ C2([0,τ]×Ω) (a classicalsolutionofthe waveequation)thenthis regularityrequirementissatisfied. Remark 3. Also notice that, aside from regularity, Theorem 1 does not require any a-priori knowledge of the initial velocity β in the forward problem (6). Now, for the case n ≥ 3 we can obtain a similar result using more than one illumination. More precisely, consider the forward problem (6) with n+1 triples of input data (α ,β ,γ ) for i = 1,...,n+1. We do not impose conditions on the i i i initialvelocitiesβ butwedoassumeknowledgeofα andγ . Eachtriple(α ,β ,γ ) i i i i i i induces a corresponding wave field w with wave speed c, and w˜ with wave speed i i c˜. As in (8a), we define 2−n σ =A w˜ , and Σ = ∇ w˜ for i=1,...,n+1. (14) i g i i g i 2 We set up the following operator-valuedmatrices Σ Σ ··· Σ σ 1,1 1,2 1,n 1  Σ Σ ··· Σ σ  A= 2...,1 2...,2 ... 1...,n ...2  and Ki,j =(PA˙i,jS)∗    Σ Σ ··· Σ σ   n+1,1 n+1,2 n+1,n n+1  A CONTROL APPROACH TO RECOVER THE WAVE SPEED 7 whereΣ isthejthentryofΣ . Noticethatthemainequation(12)canbeexpressed i,j i as [M0+K]F =M where F =(∇f,f) and Mi =−C∗mi, mi =ν·∇c−2g(wi−w˜i) and M =M| . 0 t=0 With this notation we state the second main result. Main Theorem 2. Let the geometrical control condition 2.1 hold. If (i) there exists a constant δ >0 such that |detM (x)|≥δ for a.a. x∈Ω, and 0 (ii) σ ∈C1([0,τ];C(Ω)) and Σ∈C1([0,τ];C(Ω))n, then the operator [M +K]:H0(Ω)n+1 →H0(Ω)n+1 (15) 0 is Fredholm. Moreover, there exists an open and dense subset U of (i)∩(ii) such that for each pair (σ,Σ) in this set, the operator (15) is boundedly invertible. Concerning the Inverse Problem 3, for each pair (σ,Σ)∈U, the inverse problem is uniquely solvable and the following stability estimate, kc2−c˜2kH0(Ω) ≤Ckλ−λ˜kH1([0,τ];H0(Γ)), (16) holds for a positive constant C that remains uniformly bounded for c,c˜ in small bounded sets of C2(Ω). Remark 4. With obvious modifications, Remarks 2 and 3 made after Theorem 1 are also valid for Theorem 2. Wenotethatinthismulti-measurementapproach,wearetreatingtheunknowns f and ∇f as independent from each other. In view of results such as [20, 25], we consider appropriate to attempt to recover f with a single properly chosen initial illumination. In particular, equation (12) can be treated as a first-order partial differential equation with operator-valued coefficients. Formally speaking, the principal part of this equation is given by (Σ ·∇) and the other terms can 0 be seen as compact perturbations. So one would expect (12) to (generally) have a unique solution if Σ is a vector field satisfying conditions to guarantee a unique 0 global solution of Σ ·∇f =h for arbitrary h. Some of these admissible conditions 0 are reviewed in [16, Ch. 3] and [2]. To be precise, we will make the following assumption. Assumption 3.1. Recall that Γ is the accessible portion of the boundary ∂Ω. AssumethatthevectorfieldΣ andthedomainΩsatisfythefollowingcompatibility 0 conditions: (i) Non-characteristic boundary, that is, Σ (x)·ν(x) < 0 for x ∈ Γ and Σ (x)· 0 0 ν(x)>0 for x∈∂Ω\Γ where ν denotes the outward normal on ∂Ω. (ii) Characteristic flow across the domain, that is, we assume that the family of characteristic trajectories, given by the solutions of the ODE x˙(s)=Σ (x(s)), s>0, 0 issued from Γ, covers the domain Ω and exits through ∂Ω\Γ in finite length and there is δ >0 such that |Σ (x)|≥δ for all x∈Ω. 0 These assumptions simply say that the flow of characteristic curves induced by Σ sweeps the domain Ω in a way that is consistent with the accessible portion of 0 the boundary. Inpreparationtostateourthirdmainresult,weneedtomakesomemoredefini- tions andprovesome lemmas. The firstitem to coveris a Poincar´e–Friedrichstype 8 SEBASTIAN ACOSTA inequality which can be proven using the method of characteristics [16, Ch. 3] and elementary calculus. Lemma 3.2. Let Assumption 3.1 hold. If v ∈ C∞(Ω) such that v| = 0, then the Γ following estimate kvkH0(Ω) ≤CkΣ0·∇vkH0(Ω) holds for some positive constant C =C(Ω,Γ,Σ ) independent of v. 0 Let H denote the completion of the set {v ∈ C∞(Ω) : v| = 0} with respect to Γ the norm associated with the following inner product hu,viH =hΣ0·∇u, Σ0·∇viH0(Ω). FromLemma 3.2, we see that this is a well-defined inner product which makesH a Hilbert space. Now we are ready to state our final result. Main Theorem 3. Let Assumptions 2.1 and 3.1 hold. If σ ∈C1([0,τ];C(Ω)) and Σ∈C2([0,τ];C1(Ω))n, then the operator Σ ·∇+σ +(∇·PΣ˙S)∗+(Pσ˙S)∗ :H→H0(Ω) (17) 0 0 h i isFredholm. Moreover, thereexistsanopenanddensesubsetU ⊂C1([0,τ];C(Ω))× C2([0,τ];C1(Ω))n (satisfying Assumption 3.1) such that for each pair (σ,Σ) in this set, the operator (17) is boundedly invertible. Concerning the Inverse Problem 3, provided that c = c˜ on Γ, then for each pair (σ,Σ) ∈ U, the inverse problem is uniquely solvable and the following locally Lipschitz stability estimate, kc2−c˜2kH ≤Ckλ−λ˜kH1([0,τ];H0(Γ)), (18) holds for a positive constant C that remains uniformly bounded for c and c˜in small bounded sets of C2(Ω). We wish make the following remarks. Remark 5. Assumption 3.1 cannot hold in dimension n = 2 since we have that Σ≡0 which follows from (8b). Theorem 3 is intended for dimension n≥3. Remark 6. Recall the definitions of σ and Σ found in (8b). The regularity con- ditions of Theorem 3 require that the initial Cauchy data (w˜,∂ w˜)| = (α,β˜) t t=0 and the Dirichlet boundary data γ to be sufficiently smooth. We note that if w˜ ∈C2([0,τ]×Ω) (a classical solution of the wave equation) then this smoothness requirement is satisfied. Also notice that one of the regularity conditions implies thatΣ isLipschitzcontinuousonΩwhichguaranteestheexistenceanduniqueness 0 of the characteristics curves discussed in Assumption 3.1. Remark7. Asidefromsufficientregularity,noticethatTheorem3doesnotrequire any a-priori knowledge of the initial velocity β in the forward problem (6). Remark 8. There is a similitude between (the principal part of) the operator (17) and a fundamental first-order equation found in [20, Eqn. 2.6], and also in the analysis of stationary inverse problems with internal measurements, such as [1, Eqn. 7], [12, Eqn. 34], and [13, Eqn. 59]. This is due to the fact that the unknown function is found within the principal part of the elliptic operator. A CONTROL APPROACH TO RECOVER THE WAVE SPEED 9 Before going into the proofs, we would like to provide some examples for initial conditionsαintheforwardproblem(6)sothatσ =A αandΣ = 2−n∇ αsatisfy 0 g 0 2 g the positivity and flow conditions needed for Theorems 1 and 3, respectively. Example 3.3. In Theorem 1, it is required that σ ≥δ >0. Then, we can simply 0 chooseαtobethesolutionofthePoissonequationA α=δwithvanishingDirichlet g values on ∂Ω where δ = const. > 0. This solution α is guaranteed to exist for any Riemannian metric g since A is coercive on H1(Ω). g 0 Example 3.4. In Theorem 3, it is required that Σ satisfies the flow conditions 0 of Assumption 3.1. This is intended for dimension n ≥ 3. Let α be chosen as the solution of the following Laplace boundary value problem, A α=0 in Ω, g α=−1 on Γ, α=0 on (∂Ω\Γ). Then, by the maximum principle or Hopf lemma for elliptic equations (see for instance [17, Lemma 3.4]), we have that ∂ α(x) < 0 for x ∈ Γ and ∂ α(x) > 0 ν,g ν,g for x ∈ ∂Ω\Γ as required by condition (i) of Assumption 3.1. Next we consider the contour surfaces, Γ ={x∈Ω : α(x)=s} s∈[−1,0]. s AslongasthesecontoursurfacesaresmoothandmapΓonto∂Ω\Γhomotopically, then the Hopf lemma applies at each contour surface implying that ∇ α cannot g vanish in Ω. This renders the positivity condition in part (ii) of Assumption 3.1. For sake of simplicity, we have ignored the regularity conditions in this example. Specifically, there is a singularity at the interface between Γ and (∂Ω\Γ) where the Dirichlet boundary condition for α has a jump discontinuity. A mollification argumentcould be employedto overcomethis limitation, but we do not purse that route any further. 4. Proof of Main Results. In this section we proceed to prove the main results of our paper. Proof of Main Theorem 1. Notice that the first term of the governing operator in (13)isboundedly invertibleonH0(Ω)providedthat|σ (x)|≥δ >0fora.a. x∈Ω. 0 The second term of the operator is a compact operator on H0(Ω) as asserted by Lemma 4.2 below. Hence, we obtain a Fredholm equation. Now we prove the existence of an open and dense subset (i)∩(ii) on which (13) is boundedly invertible. First, standardperturbationshows that the set of σ’s over which (13) is invertible in H0(Ω) is open. To show denseness, consider replacing σ with ρ(λ)=λσ+(1−λ)σ , with λ∈C. 0 Notice now that the first term in (13) remains unchanged for any choice of λ∈ C, and the second term remains compact and analytic with respect to λ∈C because ρ˙(λ)=λσ˙ for all λ∈C. If we set λ=0, then the operator in (13) simply becomes σ whichis boundedly invertible providedthat (i)holds. Bythe analyticFredholm 0 theorem [38], then the system is boundedly invertible for all but a discrete set of λ’s. In particular, this holds for values arbitrarily close to λ = 1. This shows the desired denseness. 10 SEBASTIAN ACOSTA The local uniformity of the constant C in Theorem 1 does not follow directly from the arguments so far. So we proceed to prove this claim for a σ ∈ U. Notice that the operator (13) and its inverse depend on σ and S which in turn depend on c˜and c, respectively. First, we address perturbations of σ within the space C1([0,τ];C(Ω)). Fromthe definition in (8b), we see that σ depends on the evolution operator (c -semigroup) 0 forthewaveequationwithspeedc˜. Thisisaproblemofregularperturbationinthe framework of classical solutions. In the appropriate norms, the evolution operator for a well-posed hyperbolic PDE is locally Lipschitz continuous with respect to the coefficients of the PDE. This follows from perturbation theory of semigroups in Banach spaces. See [15, Ch. 3] and [23, Ch. 9]. In particular, we have local Lipschitz stability of σ ∈ C1([0,τ];C(Ω)) with respect to c˜ ∈ C2(Ω). In fact, using the stability in classical Ho¨lder spaces [17] of elliptic operators (semigroup generators),we obtain that c˜∈C1,s(Ω) for s>0 suffices for our purposes. Similarly,theoperatorS :H0(Ω)→H0((0,τ)×Ω)islocallyLipschitzstablewith respect to c ∈ C2(Ω). This follows from the manner (HUM method [18, 3, 27]) in whichthecontroloperatorC :H0(Ω)→H0((0,τ)×Γ)isconstructed. Theoperator C is given by the (bounded) inverse of a certain composition of evolution operators (semigroups). In turn, these evolution operators are locally Lipschitz stable with respect to c ∈ C2(Ω). In fact, in this generalized framework, perturbation of c ∈ L∞(Ω) suffices [9]. Hence, by well-known perturbation arguments [23], we obtain both the locally Lipschitz continuity of C and the invariance of the control time τ under small perturbations of the speed c∈C2(Ω). Finally,wehavetheoperator(13)beinglocallyLipschitzcontinuouswithrespect toperturbationsofcandc˜inC2(Ω). Thereforethe sameis trueforits inverse[23]. This provides the local uniformity of the stability constant C which concludes the proof. Proof of Main Theorem 2. Theproofofthistheoremiscompletelyanalogoustothe proof of Theorem 1. Every statement holds if we make the obvious modifications to handle the operator-valuedmatrices in (15). The proof of Theorem 3 is a little more involved. The first item to cover is the well-posedness of the principal part of the operator (17). We establish this as a lemma. Lemma 4.1. Let Assumption 3.1 hold. Then the operator (Σ ·∇+σ ):H→H0(Ω) 0 0 is boundedly invertible. Proof. Theoperatorisclearlybounded. SinceAssumption3.1holds,bythemethod ofcharacteristicswe obtainthat the operatoris surjective. Againusing the formof the solutions from the method of characteristics, we find that for all v ∈H kvkH =kΣ0·∇vkH0(Ω) ≤ 1+CλeCλ kΣ0·∇v+σ0vkH0(Ω) (cid:16) (cid:17) where λ = kσ0kL∞(Ω) and C = C(Ω,Γ,Σ0) is the constant appearing in Lemma 3.2. This implies the operator is injective and bounded from below. Therefore its inverse is also bounded. With this lemma we can proceed to the proof of the third main theorem of this paper.