Identification of memory kernels depending on time and on an angular variable (Q2492099)

Let \(\Omega=\{x\in {\mathbb R}^3: (| x'| -\delta)^2+x_3^2<R^2\}\), \(| x'| =(x_1^2+x_2^2)^{1/2}\), be a torus in \({\mathbb R}^3\) whose boundary admits the usual representation \(x_1=(\delta+R\cos\theta)\cos\varphi\), \(x_2=(\delta+R\cos\theta)\sin\varphi\), \(x_3=R\sin\varphi\), \(\varphi,\theta\in [-\pi/2,3\pi/2)\). Moreover, let \(A\), \(B\), \(C\) the linear differential operators (with smooth coefficients) defined by \[ A=\sum_{j,l=1}^3\,D_{x_j}[a_{j,l}(x)D_{x_j}],\quad B=\sum_{j,l=1}^3\,D_{x_j}[b_{j,l}(x)D_{x_j}],\quad C=\sum_{j=1}^3\, c_{j}(x)D_{x_j}. \] The coefficients \(a_{j,l}\), \(j,l=1,\ldots,n\) define a symmetric uniformly elliptic quadratic form and are assumed to satisfy specific additional conditions. The author is concerned with recovering the unknown kernel \(k(t,\varphi)\), with \[ k(t,-\pi/2)=k(t,3\pi/2),\quad t\in [0,T], \tag{1} \] in the linear parabolic integrodifferential equation \[ \begin{multlined} D_tu(x,t) = Au(t,x) + [k(\cdot,\varphi(x))*Bu(\cdot,x)](t)\\+ [D_\varphi k(\cdot,\varphi(x))*Cu(\cdot,x)](t)+f(t,x),\quad (t,x)\in [0,T]\times\Omega, \end{multlined} \] where \(*\) denotes time convolution, i.e., \(p*q(t)=\int_0^t p(t-s)q(s)\,ds\), and \(\varphi(x)\) denotes the angle \(\varphi\) related to \(x\in \Omega\). Such an equation is endowed with initial and homogeneous conormal boundary conditions, i.e., \(u(0,\cdot)=u_0\) in \(\Omega\) and \(D_\nu u=0\) on \(\partial \Omega\), as well as with the following additional conditions \[ \Phi[u(t,\cdot)](\varphi)=g_1(t,\varphi),\quad \Psi[u(t,\cdot)]=g_2(t),\quad (t,\varphi)\in [0,T]\times [-\pi/2,3\pi/2). \] The functions \(g_1\) and \(g_2\) are prescribed, while the linear operators \(\Phi\) and \(\Psi\) act, respectively, on the variables \((r,\theta)\) and \((r,\varphi,\theta)\) and have specific integral representations with smooth (given) kernels. Under suitable assumptions on the data \((f,u_0,g_1,g_2)\) the author shows that the present identification problem admits, at most, a unique solution \((u,k)\). He observes that the periodicity condition (1) prevents from proving the existence result. However, it can be proved, for small \(T\)'s, if condition (1) is removed.