Interpolation of Hardy spaces of Dirichlet series (Q2420444)

This interesting article is a very careful study of interpolation of Hardy spaces \(\mathcal{H}_p\), \(1 \leq p \leq \infty \), of ordinary Dirichlet series \(\sum a_n n^{-s}\). Due to the definition of this scale of Banach spaces (we repeat the definition below), this study is carried out within the setting of Hardy spaces \(H_p(\mathbb{T}^\infty)\) of \(p\)-integrable functions defined on the compact abelian group \(\mathbb{T}^\infty\) (the countable product of the torus \(\mathbb{T}\) in the complex plane). More precisely, \(H_p(\mathbb{T}^\infty)\), \(1 \leq p \leq \infty\), by definition is the closed subspace of \(L_p(\mathbb{T}^\infty)\) which consists of all functions \(f\) on the infinite-dimensional torus \(\mathbb{T}^\infty\) for which the Fourier coefficient \(\hat{f}(\alpha)=0\) whenever the finite sequence \(\alpha\) of integers has at least one negative entry. It is well known that the classical Hardy spaces \(H_p(\mathbb{T})\) (each a natural subspace of \( H_p(\mathbb{T}^\infty)\)) behave perfectly under the methods of complex and real interpolation. But the essential message of the article under review is that this changes drastically whenever we switch from functions in one variable to functions in infinitely many variables: For each \(0 < \theta <1\), \(1 \leq r \leq \infty\), and \(1 \leq p_0 < p_1 \leq \infty\) we have that \([H_{p_0}(\mathbb{T}^\infty),H_{p_1}(\mathbb{T}^\infty)]_{\theta} \neq H_{p_\theta}(\mathbb{T}^\infty)\) and \([H_{p_0}(\mathbb{T}^\infty),H_{p_1}(\mathbb{T}^\infty)]_{(\theta,r)} \neq H_{p_\theta}(\mathbb{T}^\infty)\), where as usual \(1/p_\theta = (1-\theta)/p_0 + \theta/p_1\). In fact, it is proved that similar results hold for large classes of exact interpolation functors. The proofs are based on a clever construction of a scale of finite-dimensional subspaces of polynomials on \(\mathbb{T}^\infty\) with only finitely many variables, which are (in a certain sense) complemented in the couples \((H_{p_0}(\mathbb{T}^\infty),H_{p_1}(\mathbb{T}^\infty))\). How is all this related to the interpolation of ordinary Dirichlet series? Recall that the Hardy space \(\mathcal{H}_p\), \(1 \leq p \leq \infty\), of Dirichlet series is the Banach space of all series \(\sum a_n n^{-s}\) which have coefficients \((a_n)\) generated by a function \(f \in H_p(\mathbb{T}^\infty)\) in the sense that \(a_n = \hat{f}(\alpha)\) whenever \(n\) has the prime number decomposition \(2^{\alpha_1} \times 3^{\alpha_2} \times 5^{\alpha_3} \times \dots\). Hence, by definition, there is an isometry \(\mathcal{H}_p = H_p(\mathbb{T}^\infty)\) which preserves Dirichlet and Fourier coefficients. Then it is obvious that the above (negative) results on interpolation immediately transfer to results on interpolation of Hardy spaces of Dirichlet series. \(\mathcal{H}_p\)'s of ordinary Dirichlet series form an important part of the renaissance of the structure theory of ordinary Dirichlet series (see, e.g., the recent monographs [the reviewer et al., Dirichlet series and holomorphic functions in high dimensions. Cambridge: Cambridge University Press (2019; Zbl 1460.30004); \textit{H. Queffélec} and \textit{M. Queffélec}, Diophantine approximation and Dirichlet series. New Delhi: Hindustan Book Agency (2013; Zbl 1317.11001)]). Several independently interesting applications are given.