The set of monomial convergence of the bounded holomophic functions on B_{c0} and of m-homogeneous polynomials on c0 was studied in Chapter 10. Here the space c0 is replaced by some other l_p spaces, or even by polynomials on an arbitrary Banach sequence space and holomorphic functions on Reinhardt domains. The only complete case is p=1, where the set of monomial convergence of the m-homogeneous polynomials is exactly l_1, and the set of monomial convergence of the bounded holomorphic functions on the open unit ball of l_1 is again the ball. For other p’s upper and lower bounds are presented that give a pretty tight description.

Given a family of formal power series, its set of monomial convergence is defined as those z’s for which the series converges. The main focus is given to the sets of monomial convergence of the m-homogeneous polynomials on c0 and of the bounded holomorphic functions on B_{c0}. The first one is completely described in terms of the Marcinkiewicz space l_{(2m)/(m-1), ∞}. For the second one there is no complete description. If z is such that limsup (log n)^(1/2) ∑_j^n (z*_j)^{2} < 1 (where z* is the decreasing rearrangement of z), then z is in the set of monomial convergence of the bounded holomorphic functions. Also, if z belongs to the set of monomial convergence, then the limit superior is ≤ 1. This is related to Bohr’s problem (see Chapter 1). First of all, if M denotes the supremum over all q so that l_q is contained in the set of monomial convergence of the bounded holomorphic functions on Bc0, then S=1/M. But this can be more precise: S is the infimum over all σ >0 so that the sequence (p_n^(-σ))_n (being p_n the n-th prime number) belongs to the set of monomial convergence of the bounded holomorphic functions on Bc0.

A classical result of Fatou gives that every bounded holomorphic function on the disc has radial limits for almost every point in the torus, and the limit function belongs to the Hardy space H_\infty of the torus. This property is no longer true when we consider vector-valued functions. The Banach spaces X for which this property is satisfied are said to have the analytic Radon-Nikodym property (ARNP). Some important equivalent reformulations of ARNP are studied in this chapter. Among others, X has ARNP if and only if each X-valued H_p- function f on the disc has radial limits almost everywhere on the torus (and not only H_\infty-functions). Even more, in this case each such f has non-tangential limits within any Stolz region. The basic tools are subharmonic functions and certain maximal inequalities. Finally, it is shown that if X has the ARNP, then every L_p of functions taking values in X with a finite measure also has ARNP.

This is a short introduction to the theory of holomorphic functions in finitely and infinitely many variables. We begin with functions in finitely many variables, giving the definition of holomorphic function. Every such function has a monomial series expansion, where the coefficients are given by a Cauchy integral formula. Then we move to infinitely many variables, considering functions defined on B_{c0}, the open unit ball of the space of null sequences. Holomorphic functions are defined by means of Fréchet differentiability. We have versions of Weierstrass and Montel theorems in this setting. Every holomorphic function on B_{c0} defines a family of coefficients through a Cauchy integral formula and a (formal) monomial series expansion. Every bounded analytic (represented by a convergent power series) function is holomorphic. Hilbert’s criterion, that gives conditions on a family of scalars so that it is attached to a bounded holomorphic function on B_{c0}. Homogeneous polynomials are those entire functions having non-zero coefficients only for multi-indices of a given order. We show how these are related to multilinear forms on c0 through the polarization formulas.

The text is closed by coming back to Bohr’s absolute convergence problem, this time for vector-valued Dirichlet series. For a Banach space X abscissas and strips S(X) and S_p(X), analogous to those defined in Chapters 1 and 12 are considered. It is shown that all these strips equal 1-1/cot(X), where cot(X) is the optimal cotype of X.

We study the relationship between Hardy spaces of functions on the polytorus and certain spaces of holomorphic functions. We deal first with functions in finitely many variables, and later we jump to the infinite dimensional setting. For each N we consider the space of holomorphic functions g on the N-dimensional polydisc for which the L_p norms of g(rz) for 0<r<1 are bounded (known as the Hardy space of holomorphic functions). For each p these two Hardy spaces (of integrable functions on the N-dimensional polytorus and the N-dimensional polydisc) are isometrically isomorphic. The main tool in the proof is the Poisson operator (defined in Chapter 5). For the infinite dimensional case, we define the space of holomorphic functions g on l_2 ∩ Bc0 whose restrictions to the first N variables all belong to the corresponding Hardy space, and the norms are uniformly bounded (in N). These Hardy spaces of holomorphic functions on l_2 ∩ Bc0 and the Hardy spaces of integrable functions on the infinite-dimensional polytorus are isometrically isomorphic. The jump is given using a Hilbert criterion for Hardy spaces.

We continue the study initiated in Chapter 7 of polynomials with small norms. This time the norm of the polynomial is not taken as the supremum on the n-dimensional polydisc, we take it on B_X, the unit ball of some Banach space. The goal is to show that, given a polynomial, signs can be found in such a way that the norm of the new polynomial, whose coefficients are the original ones multiplied by the signs, has small norm. We do this with three different approaches. The first two approaches use Rademacher random variables as the main probabilistic tools. The first one is based on finding out how many balls of a fixed radius are needed to cover B_X while the second one uses entropy integrals and a good estimate for the entropy numbers of the inclusions between l_p spaces. The third approach is different, and relies on Gaussian random variables, Slepian’s lemma and the fact that Rademacher averages are dominated by Gaussian averages. This approach also allows to get estimates for vector-valued polynomials.

Littlewood’s and Bohnenblust-Hille’s inequalities (recall Chapter 6) bound certain sequence norms of the coefficients of a polynomial by a constant (not depending on the number of variables) times the supremum of the polynomial on the polydisc. A similar problem is handled here, replacing the polydisc by the unit ball of C^n with some p-norm. Optimal exponents (that depend on the degree of the polynomial and on p) are given. The proof relies on the interplay between homogeneous polynomials and multilinear mappings and an analogous inequality for multilinear mappings. This one is proved by giving a generalized mixed inequality that bounds a mixed norm of the coefficients of a matrix by the supremum on the p-balls of the associated multilinear mapping.

Basic topics on Banach space theory needed for the text are reviewed. Hahn-Banach theorem, Baire’s theorem, uniform boundedness principle, closed graph theorem, weak topologies, Banach-Alaoglu theorem, unconditional basis, Banach sequence spaces, summing operators, factorable operators, cotype, Kahane inequality.

We give the solution of Bohr’s problem, showing that in fact S=1/2. This is done by considering an analogous problem where only m-homogeneous Dirichlet series are taken into account (defining, then, S^m). Using the isometry between homogeneous Dirichlet series and polynomials, the problem is translated into a problem for these. For each m we produce an m-homogeneous polynomial P such that for every q > (2m)/(m-1) there is a point z in l_q for which the monomial series expansion of P does not converge at z. This shows that, contrary to what happens for finitely many variables, holomorphic functions in infinitely many variables may not be analytic. This also shows that (2m)/(m-1) ≤ S^m for every m and then gives the result. There is more. For each fixed 0 ≤ σ ≤ 1/2 there is a Dirichlet series whose abscissas of uniform and absolute convergence are at distance exactly σ.

The solution of Bohr’s problem (see Chapter 4) implies that for every Dirichlet series in \mathcal{H}_\infty, the sum ∑ |a_n| n^(-s) is finite for every Re s > 1/2, and we ask if we can in fact get to Re s=1/2. This is addressed by considering, for Dirichlet polynomials, the quotient between ∑ | a_n | and the norm (in \mathcal{H}_\infty) of the polynomial. We define S(x) as the supremum over all Dirichlet polynomials of length x ≥ 2 of these quotients. It is shown that S(x)=exp(- (1/\sqrt{2} + o(1)) (log n loglog n)^(1/2)) as x goes to ∞. This is reformulated in terms of the Sidon constant of the monomials as characters of the infinite-dimensional polydisc. The proof uses the hypercontractive Bohnenblust-Hille inequality and a fine decomposition of the natural numbers as those having ‘big’ and ‘small’ prime factors. Also, a version for homegeneous Dirichlet series is given.

This is a short introduction to the basics of the theory of normed tensor products. The m-fold tensor product of linear spaces is defined through the universal property. If the involved spaces are normed, then the projective and injective norms on the tensor product are. Basic properties are given: the metric mapping property and their relationship with continuous linear mappings. The symmetric m-fold tensor product and the symmetric projective and injective norms are defined analogously. These are related to the m-homogeneous polynomials.

A holomorphic function f on the disc has a Taylor expansion with coefficients c_k. Bohr asked about the maximal 0<r<1 so that the supremum for |z|<r of ∑ | c_k z^k | is less than or equal to the supremum for |z|<1 of |f(z)|. Bohr’s power series theorem answers this question showing that r=1/3 is best possible. The n-th Bohr radius K_n is defined as the best r for which an analogous question holds for holomorphic functions on the n-dimensional polydisc. The sequence (K_n) is decreasing and tends to 0 as n goes to ∞ asymptotically like (\log n/n)^(1/2). The proof os this relies on an improved version of the polynomial Bohnenblust-Hille inequality (see Chapter 6), where the constant grows at most exponentially, and to get this a Khinchin-Steinhaus inequality for polynomials is needed, showing that all L_p norms of polynomials in n variables are equivalent.

Each Hardy space of Dirichlet series \mathcal{H}_p has an associated abscissa, and the analogue to Bohr’s problem arises in a natural way: to determine the maximal distance S_p between this abscissa and the abscissa of absolute convergence. If a Dirichlet series with coefficients (a_n) belongs to \mathcal{H}_p, then the series with coefficients (a_n/n^{ε}) belongs to \mathcal{H}_q for all q>p and ε >0. It is shown that S_p=1/2, and that, if we only consider m-homogeneous Dirichlet series, S_p^m=1/2. For every 1 ≤ p < ∞ the set of monomial convergence of the Hardy space H_p of functions on the infinite dimensional polytorus (hence also of the Hardy space H_2 on the infinite-dimensional polytorus) is l_2 ∩ Bc0. The space of all multipliers on the Hardy space of Dirichlet series \mathcal{H}_p coincides with \mathcal{H}_\infty.