No label defined (Q4339650)

Consider the following notations. For a Banach space \(({\mathcal X},|\cdot|_{{\mathcal X}})\) denote by \([{\mathcal X}]\) the space of (linear continuous) operators mapping \({\mathcal X}\) into itself. If \(A: D(A)\subset{\mathcal X}\to{\mathcal X}\) is an operator, denote his range by \(R(A)\), his kernel by \(N(A)\), his resolvent set by \(\rho(A)\) and if \(\lambda\in \rho(A)\), \(R(\lambda,A):= (\lambda I-A)^{-1}\) is the resolvent. Set also \[ \widetilde{D(A)}^{\mathcal X}:= \{f\in{\mathcal X},\;\exists\{f_n\}_{n\in\mathbb{N}}\subset D(A), \text{ with }|f_n|_{D(A)}\leq M\text{ and} \lim_{n\to\infty}|f_n-f|_{\mathcal X}=0\} \] and let \(K(t,f,{\mathcal X},D(A))\) \((f\in{\mathcal X}\), \(t>0)\) be the usual \(K\)-functional. Recall that a family \({\mathcal T}= (T(t)\), \(t\in\mathbb{R})\) belonging to \([{\mathcal X}]\) is named semigroup of operators, if the following properties are satisfied: \(T(s+t)= T(t)T(s)\), \(T(0)=I\) (the identity operator) and \(\lim_{t\to 0+}|T(t)f-f |_{\mathcal X}=0\) \((f\in{\mathcal X})\). The generator \(A\) of \({\mathcal X}\) is defined by the strong limit \(Af:= \lim_{h\to 0+} h^{-1} [T(h)-I]f\) \((f\in D(A))\), where \(D(A)\) is the space of those \(f\) for which this limit exists. Also, a family \({\mathcal C}= (C(t)\), \(t\in\mathbb{R})\) belonging to \([{\mathcal X}]\) is named cosine operator function if \(C(t+s)+ C(t-s)= 2C(t)C(s)\) \((s,t\in\mathbb{R})\), \(C(0)=I\) and \(\lim_{h\to 0}|C(h)f- f|{\mathcal X}=0\). The generator \(A\) of \({\mathcal C}\) is defined by the strong limit \(Af:= \lim_{h\to 0} 2h^{-2} [C(h)f-f]\) \((f\in D(A))\). For a cosine operator function \({\mathcal C}\) as above denote by \(E^\gamma(t)f:= \gamma t^{-\gamma} \int_0^t (t-u)^{\gamma-1} C(u)f du\) \((\gamma>0\), \(t>0\), \(f\in{\mathcal X})\), the Cesàro means of order \(\gamma>0\) of \({\mathcal C}\). Also, for a semigroup \({\mathcal T}\) as above, set \(H(t)F:= t^{-1} \int_0^t T(u)f du\) \((t>0\), \(f\in{\mathcal X})\). If \({\mathcal T}\) has the generator \(A\), then \(\lambda\in \rho(A)\), the Abel mean of \({\mathcal T}\), defined by \(\lambda \int_0^\infty e^{-\lambda u}T(u)f du\) is equal to \(\lambda R(\lambda,A)f\), \((f\in N(A)\oplus R(A))\). If \({\mathcal C}\) has the generator \(A\), and if \(\lambda^2\in \rho(A)\), then the Abel mean of \({\mathcal T}: \lambda\int_0^\infty e^{-\lambda u}C(u)f du\) is equal to \(\lambda^2 R(\lambda^2,A)f\), \((f\in{\mathcal X})\). The paper contains various types of results. A first group of them refers to the convergence of the Cesàro and Abel means of cosine operator functions. One obtains counterparts of corresponding results on semigroups, established by the first author and his collaborators in the years 1956-1979. A connection between a part of the new results and the corresponding ones for semigroups, is given in Theorem 5.2: Let \({\mathcal T}\) and \({\mathcal C}\) be arbitrary semigroup respectively cosine operator function, having the common generator \(A\). Then the following seven assertions are equivalent for any given \(f\in{\mathcal X}\), for \(0<a\leq 1\) and any \(\gamma>0\). a) \(|T(t)f-f|_{\mathcal X}= \begin{cases} o(t)\\ O(t^a)\end{cases} (t\to 0+)\), b) \(|C(t)f- f|_{\mathcal X}= \begin{cases} o(t^2)\\ O(t^{2a}) \end{cases} (t\to 0)\), c) \(|H(t)f- f|_{\mathcal X}= \begin{cases} o(t)\\ O(t^a)\end{cases} (t\to 0+)\), d) \(|E^\gamma (t)f- f|_{\mathcal X} =\begin{cases} o(t^2)\\ O(t^{2a}) \end{cases} (t\to 0)\), e) \(\lambda R(\lambda,A) f-f|_{\mathcal X}= \begin{cases} o(\lambda^{-1})\\ O(\lambda^{-a}) \end{cases} (\lambda\to\infty)\), f) \(K(t,f;{\mathcal X},D(A))= \begin{cases} o(t)\\ O(t^a)\end{cases} (t\to 0+)\), g) \(\begin{cases} f\in N(A)\\ f\in \widetilde{D(A)}^{\mathcal X}\end{cases}\), provided \(a=1\). Also, from Theorem 5.1-d),e) it follows that if \(A\) is unbounded, then for each \(0<a\leq 1\) there are elements \(f_a,f_a^*\in{\mathcal X}\) such that (i) \(|E^\gamma(t) f_a-f_a |_{\mathcal X}=O(t^{2a})\) and \(\neq o(t^{2a})\), \((t\to 0+)\); and (ii) \(|\lambda R(\lambda,A) f_a^*-f_a^* |_{\mathcal X}= O(\lambda^{-a})\) and \(\neq o(\lambda^{-a})\), \(\lambda\to\infty)\). Moreover, if \(a=1\), these are true even if \(A\) is bounded, but \(A\neq\theta\). A second group of results deals with the convergence of the Cesàro and Abel means of semigroups of operators. If \(A\) is the generator of the semigroup \({\mathcal T}\), set \({\mathcal X}_0:= R(A)\oplus N(A)\) and \(A_0:= A|_{{\mathcal X}_0}\). Denote by \(P\) the projection of \({\mathcal X}_0\) onto \(N(A)\) and by \(B\) of operator mapping \(f= Ag+Pf\in R(A_0)\oplus N(A)\) onto \(g\in D(A_0)\cap N(P)\). For \(M>0\) and \(0\leq\beta<1\), set \({\mathcal T}_M^1:= \{{\mathcal T}\); \(|T(t)|_{[{\mathcal X}]}= o(t)\); \(t\to\infty\), \(|H(t)|_{[{\mathcal X}]}\leq M\}\), \({\mathcal T}_M:= \{{\mathcal T}\); \((0,\infty)\subset \rho(A)\), \(|\lambda R(\lambda,A) |_{[{\mathcal X}]}\leq M\}\), and \({\mathcal T}_M^\beta:= \{{\mathcal T}\); \(|T(t)|_{[{\mathcal X}]}= O(t^\beta)\), \(t\to\infty\), \(|H(t) |_{[{\mathcal X}]}\leq M\}\). We have \({\mathcal T}_M^\beta\subset {\mathcal T}_M^1\subset {\mathcal T}_M\). The main results for a semigroup \({\mathcal T}\in{\mathcal T}_M\) with the generator \(A\) are given in Theorem 5.3 a) The following three assertions are equivalent to \(f\in{\mathcal X}_0\), (i) \(|H(t)f- Pf|_{\mathcal X}= o(t^{-1}\) \((t\to\infty)\), if in addition \({\mathcal T}\in{\mathcal T}_M^1\); (ii) \(|\lambda R(\lambda,A)f- Pf|_{\mathcal X}= o(\lambda)\), \((\lambda\to 0+)\) and (iii) \(Pf=f\), i.e. \(f\in N(A)\). b) The following assertions are equivalent for \(0<a\leq 1\) and any \(f\in{\mathcal X}_0\). (i) \({|H(t)f-Pf|_{\mathcal X}}=O(t^{-1})\) if in addition \({\mathcal T}\in{\mathcal T}_M^0\); (ii) \(|\lambda R(\lambda,A)f- Pf|_{\mathcal X}= O(\lambda^a)\) \((\lambda\to 0+)\) and (iii) \(K(t^{-1},f; {\mathcal X}_0,D(B))= O(t^{-a})\) \((t\to\infty)\). c) If \(a=1\) then the assertions (i) to (iii) of b) are also equivalent to \(f\in \widetilde{D(B)}^{{\mathcal X}_0}\) (or \(f\in B(D)\), provided \({\mathcal X}\) is reflexive). The points d) and e) of this theorem show the sharpness of conditions (i) and (ii) of b). A different result appears in Theorem 4.6: If \({\mathcal T}\in{\mathcal T}_M^\beta\), \(\beta\in [0,1)\), then there exists a constant \(c_1\in \mathbb{R}\) such that \(|H(t)f- Pf|_{\mathcal X}\leq c_1 K(t^{\beta-1}, f,{\mathcal X}_0, D(B))\) \((f\in{\mathcal X}_0\), \(t\geq 1)\). A third major aim of the paper is to give a general method for obtaining examples of cosine operator functions. In Theorem 6.3 the following construction is given. Let \(I\) be an arbitrary interval, let \(J:=\mathbb{R}\) or \(J:= [a,b]\) and let \(g:I\to J\) be a bijection, possessing a continuous continuation of \(g^{-1}\) to the whole axis, denoted by \(g^*\). Suppose that if \(J= [a,\infty)\) or \(J=[a,b]\) \(g^*\) satisfies \(g^*(a+x)= g^*(a-x)\) \((x\in \mathbb{R})\), and if \(J=[a,b]\), then \(g^*\) has period \(2(b-a)\). Set \(G(x,t):= g^*(g(x)+t)\) and let \(p\in C(I)\). Under some additional conditions it is shown that one can obtain a cosine operator function by the formula: \[ [C(t)f](x):= \frac{1} {2p(x)} [p(G(x,t)) f(G(x,t))+ p(G(x,-t)) f(G(x,-t))], \] where \(f\) belongs to a Banach space \(X(I)\) of functions defined on \(I\). As applications, there are discussed the Chebyshev translation given by \[ C_{\text{Cheb}} (t)f(x)= {\textstyle \frac12} \bigl[f(x\cos t+ \sqrt{1-x^2} \sin t)-f(x\cos t- \sqrt{1-x^2} \sin t) \bigr], \] the exponential translation given by \(C_{\text{Exp}} (t) f(x)=\frac12 [f(xe^t)+ f(xe^{-t})]\), as well as the ``classical'' translation operator given by \(C(t) f(x):= \frac12 [f(x+t)+ f(x-t)]\).