Analytic solutions of Volterra equations via semigroups (Q2480769)

This is a study about the analyticity of the solutions of the Volterra integrodifferential equation \[ \dot u(t) = A u(t) + \int_0^t B(t-s) u(s) ds + f(t), \quad t \geq 0;\quad u(0) = x. \leqno{\text{(IDE)}} \] The so called resolvent operator (or solution operator) of this equation is analytic whenever \(A\) is the generator of an analytic semigroup on a Banach space \(X\) and \(B\) is an analytic operator-valued kernel satisfying appropriate assumptions. This statement is typically proved with the help of Laplace transforms. Here a different approach is used: the equation (IDE) is rewritten so that the dynamics is described by a \(C_0\)-semigroup generated by \[ {\mathcal A} = \left ( \begin{matrix} A & \delta_0 \cr B(\cdot) & d/ds \end{matrix}\right ), \quad \mathcal{D}(\mathcal{A}) = \mathcal{D}(A) \times \mathcal{D}(d/ds). \] The main obstacle with this semigroup approach is that the shift semigroup generated by \(d/ds\) is not analytic in the standard \(L^q({\mathbb R}_+;X)\)-setting, and this prevents the semigroup generated by \(\mathcal{A}\) from being analytic. Here this problem is solved in the following way: the shift on \(L^q({\mathbb R}_+;X)\) is replaced by a shift on the space of bounded, uniformly continuous holomorphic functions in a sector around the real axis, or a Bergman space of analytic functions in a similar sector. This makes both the shift-semigroup and the semigroup generated by \(\mathcal{A}\) analytic, and this in turn implies that the resolvent operator of (IDE) is analytic. It is also mentioned that maximal \(L_p\)-regularity can be achieved when \(X\) is an analytic UMD space.