A new approach to abstract linear viscoelastic equation in Hilbert space (Q6123315): Difference between revisions

The authors consider the abstract linear integro-differential or Volterra equation: \(\overset{.}{x}(t)=Ax(t)+\int_{0}^{t}a(t-s)Ax(s)ds\), \(t\geq 0\), with the initial condition \(x(0)=x_{0}\). Here \(x(t)\) belongs to a real Hilbert space \(X\), \(a(\cdot )\in BV_{\mathrm{loc}}(\mathbb{R}_{+})\) is a real function, \(A\) is a linear self-adjoint operator on \(X\), such that \(-A\) is positive definite: \(\left\langle Ax,x\right\rangle _{X}\leq -m_{0}\left\Vert x\right\Vert _{X}^{2}\), \(\forall x\in D(A)=X_{1}\), and \(A\) generates an exponentially stable contraction semigroup on \(X\). A classical solution to the above problem is a function \(x(\cdot )\in C^{1}(\mathbb{R}_{+};X)\cap C( \mathbb{R}_{+};X_{1})\) which satisfies the equation and the initial condition. A function \(x(\cdot ):\mathbb{R}_{+}\rightarrow X\) is a strong solution to the above problem if \(x(\cdot )\in C(\mathbb{R}_{+};X)\cap L_{\mathrm{loc}}^{\infty }(\mathbb{R}_{+};X_{1})\) and \(x(t)=x_{0}+A(t)\ast x(t)\) holds with \(A(t)=A+1\ast a(t)A\). A family of bounded operators \(\{S(t):t\geq 0\}\) on \(X\) is the resolvent operator family of the preceding equality if \( S(0)=I\), \(S(\cdot )\) is strongly continuous in \(X\) on \(\mathbb{R}_{+}\), \( S(t)D(A)\subset D(A)\) for all \(t\geq 0\), \(S(t)Ax_{0}=AS(t)x_{0}\) for all \( t\geq 0\) and \(x_{0}\in D(A)\), and for each \(x_{0}\in D(A)\) the function \( S(\cdot )x_{0}\) is the unique strong solution to this equality with \( x(0)=x_{0}\). The purpose of the paper is to build a new semigroup approach for the above Volterra problem. The authors introduce the product Hilbert space \(\mathcal{X}=X\times L_{\nu }^{2}(\mathbb{R}_{+};X_{1/2})\), \(\nu (t)=-1/a(t)\), and the operator \(\mathcal{A}\left[ \begin{array}{c} x \\ f \end{array} \right] =\left[ \begin{array}{c} A(x+\int_{0}^{\infty }f(\tau )d\tau ) \\ f^{\prime }+\overset{.}{\mu }x \end{array} \right] \), with \(D(\mathcal{A})=\{\left[ \begin{array}{c} x \\ f \end{array} \right] \in D((-A)^{1/2})\times \mathcal{H}_{\nu }^{1}(\mathbb{R} _{+};X_{1/2}):x+\int_{0}^{\infty }f(\tau )d\tau \in D(A)\}\), where \(X_{1/2}\) is the Hilbert space \(D((-A)^{1/2})\). The main result of the paper assumes that the function \(\mu (t)=-a(t):\mathbb{R}_{+}\rightarrow (0,\infty )\) is non-increasing, \(\mu \in L^{1}(\mathbb{R}_{+})\cap W_{\mathrm{loc}}^{2,1}(\mathbb{R} _{+})\), \(\overset{.}{\mu },\frac{\overset{.}{\mu }}{\sqrt{\mu }},\frac{ \overset{..}{\mu }}{\sqrt{\mu }}\in L^{2}(\mathbb{R}_{+})\), for \(0<\kappa =\int_{0}^{\infty }\mu (t)dt\) and \(m=\left\Vert \overset{..}{\mu }/\sqrt{\mu }\right\Vert _{L^{2}(\mathbb{R}_{+})}\), it holds that \(\gamma =\sqrt{\kappa }+m\leq 2\), \(\overset{.}{\mu }(t)+\gamma \mu (t)\leq 0\), \(\forall t\geq 0\). It proves that: \(\mathcal{A}\) generates a contraction semigroup \( \{T(t):t\geq 0\}\) on \(\mathcal{X}\); for \(\phi _{0}\in L_{\nu }^{2}(\mathbb{R} _{+};X_{1/2})\), let \(\xi _{0}(\tau )=\int_{0}^{\infty }\overset{.}{\mu } (\tau +s)\phi _{0}(s)ds\in X_{1/2}\) for all \(\tau \geq 0\).\ If \(\left[ \begin{array}{c} x_{0} \\ \xi _{0} \end{array} \right] \in D(A)\), the first component of \(\left[ \begin{array}{c} x(t) \\ \xi ^{t} \end{array} \right] =\mathcal{T}(t)\left[ \begin{array}{c} x_{0} \\ \xi _{0} \end{array} \right] \) gives the unique classical solution to the last problem and the second one is given by \(\xi ^{t}(\tau )=\xi _{0}(t+\tau )+\int_{0}^{t} \overset{.}{\mu }(t+\tau -s)x(s)ds=\int_{t}^{\infty }\overset{.}{\mu }(\tau +s)\phi _{0}(s-t)ds+\int_{0}^{t}\overset{.}{\mu }(\tau +s)x(t-s)ds\), for all \(\tau ,t\geq 0\). In particular, for each \(x_{0}\in D(A)\) and every \(t\geq 0\) , the first coordinate of \(\mathcal{T}(t)\left[ \begin{array}{c} x_{0} \\ 0 \end{array} \right] \) gives the unique classical solution to the original problem; for each \(\left[ \begin{array}{c} x_{0} \\ \xi _{0} \end{array} \right] \in D(A)\) and every \(t\geq 0\), an upper bound is given for \(\frac{d}{ dt}\left\Vert \mathcal{T}(t)\left[ \begin{array}{c} x_{0} \\ \phi _{0} \end{array} \right] \right\Vert _{X}^{2}\). For the proof, the authors use the Lumer-Phillips theorem, first proving that \(\mathcal{A}\) is dissipative and that \(Ran(I-\mathcal{A})=\mathcal{X}\). They then use the general theory of operator semigroups. In the last part of their paper, the authors prove an exponential stability result, that is there exist two positive constants \(K, \sigma \) such that for each initial value \(x_{0}\in D(A)\), the corresponding unique classical solution \(x(\cdot )\) exists and satisfies \( \left\Vert x(t)\right\Vert _{X}\leq Ke^{-\sigma t}\left\Vert x_{0}\right\Vert _{X}\) \(\forall t\geq 0\). The paper ends with an example for which numerical computations confirm the exponential stability.