On functions with differences from \(L_p(0,1)\) (Q492718)

If \((X,{\mathcal X},\mu_X)\) is a \(\sigma\)-finite measure space, \(1\leq p<\infty\), and \([f]\) denotes the equivalence class of functions, which differ from \(f\) on sets of measure \(0\), then \(L^p(X)\) denotes \(\{[f]:\|[f]\|_p<\infty\}\), where \[ \| f\|_p= \Biggl(\int_X|f(x)|^p\, d\mu_X(x)\Biggr)^{1/p}. \] In addition, \(\| f\|_\infty= \text{ess} \sup\{|f(x)|: x\in X\}= \inf \{a> 0: \mu_X(\{x:|f(x)|> a\})= 0\}\). Let \(I\) be the real line interval \((0,1)\) and \({\mathcal M}(I)\) the class of Lebesgue measurable functions on \(I\). The difference \(\Delta_h(f)(x)\) of \(f\in{\mathcal M}(I)\) at a point \(x\in(0, 1-h)\) is defined for \(h>0\) by \(\Delta_h(f)(x)= f(x+h)- f(x)\), and the higher differences \(\Delta_h^k(f)(x)\), \(k=2,3,\dots\), are defined by \[ \Delta^k_h(f)(x)= \Delta_h(\Delta^{k-1}_h(f)(x))= \Delta^{k-1}_h(\Delta_h(f)(x)). \] If \(\{\beta\}= \{\beta_j:j= 0,1,2,\dots, k\}\) is a set of nonzero complex numbers and \(\{\tau_j: j=0,1,2,\dots, k\}\) is a set of real numbers, let \({\mathcal D}^k_h={\mathcal D}_{h,\{\beta\}, \{\tau\}}^k:{\mathcal M}(I)\to{\mathcal M}(I_h)\), \(I_h= (-\tau_0 h,1-\tau_k h)\), be defined as \[ {\mathcal D}^k_h(f)(x)= \sum^k_{j=0} \beta_j f(x+\tau_j h). \] The main theorem of this paper states that, if \(f\in{\mathcal M}(I)\) and for some number \(p>0\), \({\mathcal D}^k_h(f)\in L^p(I_h)\), for \(h\in{\mathcal H}\subset (0,{1\over 2}H)\), \(H= H(\{\tau\})= (\tau_k- \tau_0)^{-1}\), where \({\mathcal H}\) has positive measure, then \(f\in L^p(I)\). The conclusions are reported to solve a problem considered by O. V. Besov, as stated by \textit{S. M. Nikol'skij} [Approximations of functions of several variables and imbeddings. Berlin-Heidelberg-New York: Springer-Verlag (1975; Zbl 0307.46024)].