A general inverse problem for a memory kernel in one-dimensional viscoelasticity (Q699317)

The authors are concerned with the problem of recovering \(N\) functions \(m_1,\dots,m_N:[0,+\infty)\to {\mathbb R}\) in the degenerate relaxation kernel \(m(x,t) = \sum_{k=1}^n m_k(t)\mu_k(x)\), related to a viscoelastic one-dimensional system, by performing \(N\) additional measurements with (possibly) different force terms. The explicit problem consists in determining a vector displacement \(u=(u_1,\dots,u_l):(0,1)\times [0,+\infty)\to {\mathbb R}\) and \(N\) unknown functions \(m_1,\dots,m_N:[0,+\infty)\to {\mathbb R}\) satisfying the following integro-differential equations in the half-strip \(\Omega=(0,1)\times {\mathbb R}_+\): \[ \rho(x)D_t^2u_j(x,t) = D_x(\beta(x)D_xu_j)(x,t) \] \[ - D_x\Big( \int_0^t m(t-s,x)D_xu_j(x,s) ds\Big) + r_j(x,t),\qquad (x,t)\in \Omega,\;j=1,\dots,l, \eqno(1) \] when \(u_j\) is subject to initial and boundary conditions \[ u_j(x,0) = \varphi_j(x),\qquad D_tu_j(x,0) = \psi_j(x),\qquad \forall x\in [0,1],\;j=1,\dots,l, \eqno(2) \] \[ u_j(0,t) = f_j(t),\qquad u_j(1,t) = g_j(t),\qquad \forall t\in [0,+\infty),\;j=1,\dots,l. \eqno(3) \] Functions \(\rho\), \(\beta\), \(\mu_j\), \(j=1,\dots,N\), \(\varphi_j\) and \(\psi_j\), \(j=1,\dots,l\), are smooth and positive on \([0,1]\), while \(f_j\) and \(g_j\), \(j=1,\dots,l\), are smooth and bounded on \([0,+\infty)\). The \(N\) additional measurements involving the displacements \(u_j\), \(j=1,\dots,l\), needed to recover functions \(m_1,\dots,m_N\) are taken in inner points \(x_{(i,j)}\in (0,1)\): \[ u(x_{_(i,j_)},t)=h_{(i,j)}(t),\quad t\in [0,+\infty),\;i=1,\dots,N_j,\;j=1,\dots,l,\quad \Big(\sum_{j=1}^l N_j = N\Big), \eqno(4) \] the \(h_{(i,j)}\)'s being smooth bounded functions on \([0,+\infty)\). Using Laplace transform, the authors reduce problem (1)--(4) to a fixed-point system for \((M_1,\dots M_N)\), the Laplace transform of \((m_1,\dots, m_N)\). Taking advantage of specific regularity properties of one-dimensional Green function related to the transformed problem, the authors can solve the previous system in a specific space of analytic functions, and prove that problem (1)--(4) has a unique solution \((u,m)\) for all positive \(t\), provided an explicit operator of the data admits a suitable representation, and the determinant of the algebraic system \[ \sum_{k=1}^N \gamma_{(i,j),k}y_k=z_{(i,j)},\quad i=1,\dots,N_j,\;j=1,\dots,l \] does not vanish.