From ℋ∞ to 𝒩. Pointwise properties and algebraic structure in the Nevanlinna class

Abstract This survey shows how, for the Nevanlinna class 𝒩 of the unit disc, one can define and often characterize the analogues of well-known objects and properties related to the algebra of bounded analytic functions ℋ∞: interpolating sequences, Corona theorem, sets of determination, stable rank, as well as the more recent notions of Weak Embedding Property and threshold of invertibility for quotient algebras. The general rule we observe is that a given result for ℋ∞ can be transposed to 𝒩 by replacing uniform bounds by a suitable control by positive harmonic functions. We show several instances where this rule applies, as well as some exceptions. We also briefly discuss the situation for the related Smirnov class.


Introduction . The Nevanlinna class
The class H ∞ of bounded holomorphic functions on the unit D disc enjoys a wealth of analytic and algebraic properties, which have been explored for a long time with no end in sight, see e.g. [11], [38]. These last few years, some similar properties have been explored for the Nevanlinna class, a much larger algebra which is in some respects a natural extension of H ∞ . The goal of this paper is to survey those results.
For H ∞ , and for a function algebra in general, under the heading "pointwise properties" we mean the characterization of zero sets, interpolating sequences, and sets of determination; while the (related) aspects of algebraic structure of H ∞ we are interested in concern its ideals: the Corona theorem, which says that D is dense in the maximal ideal space of H ∞ , the computation of stable rank, and the determination of invertibility in a quotient algebra from the values of an equivalence class over the set where they coincide.
According the most common de nition, the Nevanlinna class is the algebra of analytic functions N = f ∈ Hol(D) : sup r< π log + f (re iθ ) dθ π < ∞. .
We will take a di erent -but, of course, equivalent-perspective. One way to motivate the introduction of this algebra is to consider the quotients f /g when f , g ∈ H ∞ and f /g is holomorphic. Without loss of generality we may assume f ∞ , g ∞ ≤ . Once we factor out the common zeroes of f and g using a Blaschke product (see Section 2), we may assume that g(z) ≠ for any z ∈ D. Then |g| = exp(−h), where h is a positive harmonic function on the disc, and log |f /g| ≤ h. We thus make have the following de nition, which will prove more useful in what follows. The main goal of this survey is to provide, when possible, analogues for N of several well-known results on H ∞ . We shall see that, in order to transfer the results we report on from H ∞ to N, we must apply the following general principle. A function f belongs to H ∞ if and only if log + |f | is uniformly bounded above, while f ∈ N when there exists h ∈ Har+(D) such that log + |f | ≤ h. Accordingly, in the hypotheses (and sometimes the conclusions) of the theorems regarding H ∞ we expect to replace uniform bounds by positive harmonic majorants. This turns out to yield very natural results and, sometimes, natural problems.

De nition 1.1. A function f holomorphic on D is in the
A function with a positive harmonic majorant is harder to grasp intuitively than a bounded function. Admitting a positive harmonic majorant is de nitely a restriction; for instance, Harnack's inequalities (1) show that a positive harmonic function cannot grow faster that C( − |z|) − as |z| → . But there is no easy way to recognize whether a given nonnegative function on the disc has a harmonic majorant, although some conditions are given in [22], [4].
To give a very simple example, suppose that we are given a sequence (z k ) k ⊂ D and positive numbers (v k ) k . De ne a function φ on D by φ(z k ) = v k , φ(z) = when z ∉ (z k ) k . It follows from Harnack's inequality that a necessary condition for φ to admit a harmonic majorant is sup k ( − |z k |)v k < ∞. On the other hand, a su cient condition is that k ( − |z k |)v k < ∞, and the gap between the two conditions cannot be improved without additional information about the geometry of the sequence (z k ) k (see Corollary 3.8 in Section 3).
The paper is structured as follows. In the next Section we nish describing the setup and gather several well-known properties of Nevanlinna functions. In particular, we recall that from the point of view of the topology N is far worse than H ∞ . Even though there is a well-de ned distance d that makes (N, d) a metric space, this is not even a topological vector space. This is an important restriction, since all the nice theorems about Banach spaces available for H ∞ are no longer available. The canonical factorization of Nevanlinna functions, similar to that of bounded analytic functions, makes up for these shortcomings.
Section 3 studies the analogue in N of the interpolation problem in H ∞ , which was completely solved in a famous theorem of Carleson (see e.g. [11]). The rst issue is to give the appropriate de nition of Nevanlinna interpolation. Once this is done we show that the general principle described above, applied to the various characterizations of H ∞ -interpolating sequences, provides also characterizations of Nevanlinna interpolating sequences. The same principle works well when describing nite unions of interpolating sequences.
An old theorem of Brown, Shields and Zeller [6] shows that the H ∞ -sampling sequences, i.e. sequences are precisely those for which the non-tangential accumulation set of (z k ) k on ∂D has full measure. In Section 4 we set the analogous problem for N, show the complete solution given by S. Gardiner, and provide some examples obtained by previous, more computable, conditions given by the authors. R. Mortini showed that the corona problem in N can be solved analogously to Carleson's classical result for H ∞ , provided that the substitution suggested by our guiding principle is performed. In Section 5 we show this and a result about nitely generated ideals, which is the natural counterpart of a theorem by Tolokonnikov [46]. We also show that, in contrast to the H ∞ -case, the stable rank of the algebra N has to be strictly bigger that .
Section 6 studies the Corona problem in quotient algebras. Gorkin, Mortini and Nikolski [18] showed that H ∞ quotiented by an inner function has the Corona property if and only if it has the so-called Weak Embedding Property (essentially, the inner function is uniformly big away, by a xed distance, from its zeroes; see Theorem 6.2(c)). The analogue is also true for N, but unlike in the bounded case, this is equivalent to the inner function being a Blaschke product of a nite number of Nevanlinna interpolating sequences. We are also interested in the determination of invertibility in a quotient algebra from the values of an equivalence class over the set where they coincide.
The nal Section 7 is devoted to describe the results for the Smirnov class N+, the subalgebra of N consisting of the functions f for which log + |f | has a quasi-bounded harmonic majorant, i.e. a majorant of type The general rule, with few exceptions, is that the statements about N also hold for N+ as soon as the bounded harmonic majorants are replaced by quasi-bounded harmonic majorants.
A nal word about notation. Throughout the paper A B will mean that there is an absolute constant C such that A ≤ CB, and we write A ≈ B if both A B and B A.

Preliminaries
It has been known for a long time that functions in the Nevanlinna class have the same zeroes as bounded functions. These are the sequences Z := (z k ) k∈N ⊂ D satisfying the Blaschke condition k ( − |z k |) < ∞. Note that when points in the sequence (z k ) k are repeated, we understand that f must vanish with the corresponding order.
That this is necessary is an easy application of Jensen's formula. To prove the reverse implication just notice that when (z k ) k is a Blaschke sequence the the associated Blaschke product converges and yields a holomorphic function with non-tangential limits of modulus almost everywhere on the unit circle ∂D. We simply write B(z) when no confusion can arise.
Moreover, log |g| is the least harmonic majorant of log |f |.
Since N is the set of quotients f /g, where f , g ∈ H ∞ and g is zero free, an easy corollary is that F ∈ N admits an inverse in N if and only if it does not vanish on D.
Any f ∈ H ∞ veri es log |f (z)| ≤ P[log |f * |](z), where we recall that f * denotes the nontangential boundary values of f . Therefore if F = f /g ∈ N, F ≢ , F * (e iθ ) ≠ almost everywhere, it admits nite nontangential boundary values almost everywhere, again denoted F * = (f /g) * . We will see more on this in Section 2.2.
Throughout the paper we will nd it useful to consider the product with one factor removed and write We shall often need to use the pseudohyperbolic or Gleason distance between points of the disc given by It is invariant under automorphisms (holomorphic bijections) of the disc, and closely related to the Poincaré distance. We shall denote by D(a, r) the disk centered at a ∈ D and with radius r ∈ ( , ), with respect to ρ, that is, D(a, r) = {z ∈ D : ρ(z, a) < r}.
Regarding basic properties of harmonic functions, we will use repeatedly the well-known Harnack inequalities: for H ∈ Har+(D) and z, w ∈ D, In particular, taking w = , We now collect some standard facts about the Nevanlinna class. We start with the natural metric and the topology it de nes. Next we recap the canonical factorization of Nevanlinna functions.

. Topological properties of N
From the point of view of the topology, N is far worse than H ∞ . One easily sees that, given f ∈ Hol(D), log( + |f |) admits a harmonic majorant if and only if log + |f | does. Thus we may de ne a distance between f and g in the Nevanlinna class by The subharmonicity of log( + |f |) yields the pointwise estimate which shows that convergence in the distance d implies uniform convergence on compact sets [44, Proposition 1.1].
The value N(f ) can also be rewritten as an extremal solution to a harmonic majorant problem: Although (N, d) is a metric space, it is not a topological vector space: multiplication by scalars fails to be continuous, and it contains many nite-dimensional subspaces on which the induced topology is discrete. In fact, the largest topological vector space contained in N is the Smirnov class N+ (see Section 8). These facts and many others are proved in [44].
The limitations imposed by this lack of structure are important, for example, when studying interpolating or sampling sequences (see Sections 3 and 4), since such basic tools as the Open Mapping or the Closed Graph theorems are not available.

. Canonical factorization
To compensate for the lack of structure just mentioned, there is a canonical factorization of functions both in H ∞ and N (general references are e.g. [11], [38] or [42]). As a matter of fact this is at the core of our transit from H ∞ to N, although quite often in the process the existing proofs for H ∞ have to be redone.
As mentioned in the previous section, given f ∈ N with zero set Z and associated Blaschke product B Z , the function f /B Z is zero-free and belongs to N as well (Theorem 2.1).
A function f is called outer if it can be written in the form This formula allows us to freely switch between assertions about outer functions f and the associated measures wdσ.
Another important family in this context are inner functions: I ∈ H ∞ such that |I| = almost everywhere on ∂D. Any inner function I can be factorized into a Blaschke product B Z carrying the zeros Z = (z k ) k of I, and a singular inner function S de ned by for some positive Borel measure µ singular with respect to the Lebesgue measure. According to the Riesz-Smirnov factorization, any function f ∈ N is represented as where O , O are outer with O ∞ , O ∞ ≤ , S , S are singular inner, B is a Blaschke product and |α| = .
Similarly, any f ∈ H ∞ with zero set Z can be factored as f = B Z f , with f ∈ H ∞ and f ∞ = f ∞. Actually f takes the form f = αSO, where O is outer and bounded, S is singular inner and |α| = . In this sense H ∞ and N are comparable.

Nevanlinna Interpolation
The rst di culty when trying to study interpolating sequences for N is to nd out what the precise problem should be. An interpolating sequence is one over which a certain natural set of sequences of values may be realized as the restriction of functions in the class being studied. That bounded functions should be required to interpolate all bounded values seems natural.
De nition 3.1. We say that (z k ) k∈N ⊂ D is an interpolating sequence for H ∞ if for any bounded complex By an application of the Open Mapping theorem, one can show that when (z k ) k is H ∞ -interpolating, there exists M > such that f can be chosen with f ∞ ≤ M v ∞. The minimum such M is called the interpolation constant of (z k ) k .
The interpolating condition means that point evaluations at each z k are independent of each other in a strong sense, and so the sequence has to be sparse. Actually, it is easy to deduce from the Schwarz-Pick Lemma that any interpolating sequence must be separated, i.e. inf j≠ k ρ(z j , z k ) > .
Actually, a stronger uniform separation is necessary and su cient.  (Carleson, 1958, see e.g. [11]). The sequence (z k ) k is interpolating for H ∞ if and only if The H ∞ -interpolating sequences, also known as Carleson sequences, have several equivalent characterizations, as we shall see soon.
Turning to the Nevanlinna class, we see that from the usual de nition given at the beginning of the paper (in terms of limits of integrals over circles), it is not immediately obvious what the natural condition on pointwise values of f should be.
One way to circumvent this is to forego an explicit description of our target set of values, but to demand that it satis es some property that guarantees that, in a sense, values can be chosen independently over the points of the sequence. Following N. K. Nikolski and his collaborators, we adopt the following de nition.
De nition 3.3. Let X be a space of holomorphic functions in D. A sequence Z ⊂ D is called free interpolating for X if the space of restrictions X|Z of functions of X to Z, also called trace space, X|Z, is ideal. This means that if (v k ) k ∈ X|Z and (c k ) k ∈ ∞ , then necessarily (c k v k ) k ∈ X|Z.
In other words, Z being free interpolating means that whenever a sequence of values is a restriction of an element in X, any other sequence with pointwise values lesser or equal in moduli will also be a restriction. We also see that, when X is stable under multiplication by bounded functions, this immediately implies that Z is a zero set for X.
Observe also that for X a unitary algebra, X|Z is ideal if and only if ∞ ⊂ X|Z. Indeed, if X|Z is ideal, then since the constant function belongs to X, any bounded sequence must be the restriction of a function in X.
As a consequence, we remark that any H ∞ -interpolating sequence must be free interpolating for N, since any bounded sequence of values will be the restriction of a bounded function, which is in the Nevanlinna class.
An alternative approach, when trying to de ne Nevanlinna interpolation, is to think in terms of harmonic majorants and consider, for a sequence Z = (z k ) k , the subspace of values Notice that it is immediate from the rst property in De nition 1.1 that N|Z ⊂ N (Z). We then want to claim that a sequence is interpolating for N when the reverse inclusion holds. Since N (Z) is clearly ideal, this implies in particular that the sequence is free interpolating.
Conversely, assume Z is free interpolating. Given values (v k ) k and h as above, we can construct a holomorphic function H such that Re H = h; then e H ∈ N. The values (e −H(z k ) v k ) k form a bounded sequence, so by the assumption there is g ∈ N such that g(z k ) = e −H(z k ) v k , and e H g ∈ N will interpolate the given values. So the two conditions above turn out to be equivalent.

. Main result and consequences.
In this section we report mostly on results from [22].
Nevanlinna interpolating sequences are characterized by the following Carleson type condition.
Notice that (c) could be rephrased as In this form we see that this is the adaptation, according to our general principle, of Carleson's condition for H ∞ : the uniform upper bound on log |B k (z k )| − is replaced by a harmonic majorant. A rst consequence of the theorem is that Nevanlinna-interpolating sequences must satisfy a weak separation condition.

De nition 3.5.
with h the harmonic majorant in condition (c). So any Nevanlinna-interpolating sequence is weakly separated. Another proof of this is given in [21, Although it is not easy to deduce simple geometric properties from condition (c), there is a vast and interesting class of sequences which turns out to be Nevanlinna-interpolating.
In order to see a rst family of examples we point out that condition (c) is really about the "local" Blaschke product; it depends only on the behavior of the z j , j ≠ k, in a xed pseudohyperbolic neighbourhood of z k .
The idea of the proof is that for any a ∈ D and δ ∈ ( , ), there is a constant C such that for ρ(a, z) ≥ δ, where is the Privalov shadow of a (see e.g. [35, p. 124, lines 3 to 17]). One gets then the result by summing over the sequence Z, which gives the Poisson integral of an integrable function on the circle, because of the Blaschke condition. An immediate consequence is the following.

Corollary 3.7. Any separated Blaschke sequence is Nevanlinna-interpolating.
There are other cases, when the geometry of Z is especially regular, where it is possible to characterize Nevanlinna interpolating sequences. This is the case for well concentrated sequences, in the sense that they are contained in a nite union of Stolz angles It is also the case for for well spread sequences, in the sense that the measure µ Z = k ( − |z k |)δz k has bounded Poisson balayage, i.e.,

has bounded Poisson balayage. Then Z is Nevanlinna interpolating if and only if
Condition (5) is always necessary for Nevanlinna interpolation. This is just a consequence of Theorem 3.4 (c) and Harnack's inequalities (see (1)).
In order to see the converse we can assume that the sequence is contained in just one Stolz angle: if In this situation the proof is immediate. Assume Z is contained in the Stolz angle of vertex ∈ ∂D. Let C denote the supremum in (5) and de ne the positive harmonic (singular) function where δ indicates the Dirac mass on . From the hypothesis and the result follows from Theorem 3.4. On the other hand, condition (6) is always su cient, since then In order to see that in case the Poisson balayage is nite (6) is also necessary, take h ∈ Har+(D) satisfying Theorem 3.4(c) and let ν be a nite positive measure with h = P[ν]. Then, by Fubini's theorem and the result follows from the previous estimate.

. Comparison with previous results.
There had been previous works on the question of interpolation in the Nevanlinna class. As early as 1956, Naftalevič [34] described the sequences Λ for which the trace N|Λ coincides with the sequence space

Theorem 3.9. [34] N|Z = Na (Z) if and only if Z is contained in a nite union of Stolz angles and (5) holds.
On the other hand, in a paper about the Smirnov class, Yanagihara [53] had introduced the sequence space As we have just seen in the previous section, for any The target space Na seems "too big", since the growth condition it imposes forces the sequences to be con ned in a nite union of Stolz angles. Consequently a big class of H ∞ -interpolating sequences, namely those containing a subsequence tending tangentially to the boundary, cannot be interpolating in the sense of Naftalevič. This does not seem natural, for H ∞ is in the multiplier space of N.
On the other hand, the target space Ya (Z) seems "too small": there are H ∞ -interpolating sequences such that N|Z (or even N+|Z) does not embed into l Ya (Z) [53,Theorem 3].
If one requires that N|Z ⊃ Ya (Z), this implies that all bounded values can be interpolated, and the sequence is Nevanlinna-interpolating. But then the natural space of restrictions of functions in N is the a priori larger N (Z). As seen in Corollary 3.8, the target space Ya (Z) is only natural when considering sequences for which µ Z := k ( − |z k |)δz k has bounded Poisson balayage.

. Equivalent conditions for Nevanlinna interpolation
The following result collects several alternative descriptions of Nevanlinna interpolating sequences. All of them have their corresponding analogues in H ∞ .
Given H ∈ Har+(D), consider the disks D H k = D(z k , e −H(z k ) ) and the domain The proof of Theorem 3.10 below shows clearly that the choice of the constant 1/2 in the de nition of Ω H k is of no relevance; it can be replaced by any c ∈ ( , ). Let ω(z, E, Ω) denote the harmonic measure at z ∈ Ω of the set E ⊂ ∂Ω in the domain Ω.
The proof of (d) shows that it can be replaced by an a priori stronger statement: for every ϵ ∈ ( , ) there exists H ∈ Har+(D) such that the disks D H k are pairwise disjoint, and Vasyunin proved in [50] (see also [25]) that B Z is an H ∞ interpolating Blaschke product if and only if there exists δ > such that Therefore, condition (b) is, again, the natural counterpart of the existing condition for H ∞ . Similarly, statement (d) and its proof are modelled after the corresponding version for H ∞ , proved by J.B. Garnett, F.W. Gehring and P.W. Jones in [12]. In that case the pseudohyperbolic discs D H k = D(z k , e −H(z k ) ) have to be replaced by uniform discs D(z k , δ), δ > .
A useful, and natural, consequence of Theorem 3.10(d) is that Nevanlinna interpolating sequences are stable under small pseudohyperbolic perturbations. Corollary 3.11. Let Z = (z k ) k be a Nevanlinna interpolating sequence and let H ∈ Har+(D), satisfying Theo- then Z is also a Nevanlinna interpolating sequence.

. Finite unions of interpolating sequences
Interpolation can be considered also with multiplicities, or more generally, with divided di erences. We show next that a discrete sequence Z = (z k ) k of the unit disk is the union of n interpolating sequences for the Nevanlinna class N if and only if the trace N|Z coincides with the space of functions on Z for which the pseudohyperbolic divided di erences of order n − are uniformly controlled by a positive harmonic function.
In Section 6 we will see other characterizations of nite unions of Nevanlinna interpolating sequences.
De nition 3.12. Let Z = (z k ) k be a discrete sequence in D and let ω be a function given on Z. The pseudohyperbolic divided di erences of ω are de ned by induction as follows It is not di cult to see that X n (Z) ⊂ X n− (Z) ⊂ · · · ⊂ X (Z) = N(Z). For example, if ω ∈ X (Z), we can take a xed z k ∈ Z and write Using that there exists H ∈ Har+(D) with we readily see that there is H ∈ Har+(D), depending on H and z k , such that The following result is the analogue of Vasyunin's description of the sequences Z in D such that the trace of the algebra H ∞ on Z equals the space of pseudohyperbolic divided di erences of order n (see [51], [52]). Similar results hold also for Hardy spaces (see [7] and [19]) and the Hörmander algebras, both in C and in D [28]. One may consider the dual problem to interpolation: which sequences are "thick" enough so that the norm of a function can be computed from its values on the sequence? Here there is no reason to restrict ourselves to sequences.

De nition 4.1. We say that
Recall that we say that a sequence (z k ) k converges to z * ∈ ∂D non-tangentially if lim k→∞ z k = z * and if there exists A > such that for all k, |z * − z k | ≤ ( + A)( − |z k |). We write NT lim z→e iθ f (z) = λ if the limit is achieved over all sequences tending to e iθ non-tangentially.
Determination sets for H ∞ are characterized by a simple geometric condition. . De ning the question for N.
In general a sequence Z = (z k ) k is called "sampling" for a space of holomorphic functions X when any function f ∈ X is determined by its restriction f |Z, with control of norms. For for the Nevanlinna class, which has nothing like a norm, the situation is not so obvious. We start from the notion of set of determination for H ∞ . Instead of requiring that the least upper bound obtained from the values of f |Z be the same as sup D |f |, we will consider the set of harmonic majorants of (log + |f |) |Z compared to the set of harmonic majorants of log + |f |.
Recall that the topology on N is de ned with the help of the functional N, de ned in (2) and (3). We give a variant.
Notice that the expression in (3) and the second expression in the equation above make sense for any measurable function on the disk, and we will apply them to f |Z. We say that Z is of determination (or sampling) for N if the above properties are satis ed.

. Main result.
If property (b) above is satis ed, then in particular it must hold for zero-free functions in the Nevanlinna class, and passing to log |f |, the set Z must be a set of determination for the class Har±(D) of harmonic functions which are the di erence of two positive harmonic functions.
To state subsequent results, we need a variant of the decomposition of the disc into Whitney squares, and points in them. Given n ∈ N and k ∈ { , . . . , n+ − }, let  Note that in contrast with the condition in Theorem 4.2, the condition of accumulation to the boundary must be met at every point with no exception. In particular, a set Z such that every boundary point is a nontangential limit of points of Z will satisfy the above property. The sets of determination for N have been characterized by S. Gardiner [10]. We need an auxiliary quantity depending on a set A ⊂ D and t ≥ . If either A = ∅ or t = , we set Q(A, t) = ; otherwise, Finally, for any set A and λ > , λA := {λz, z ∈ A}. . Regular discrete sets.
In this section we give some computable conditions for regular sets.
De nition 4.6. Let g : ( , ] −→ ( , ] be a non-decreasing continuous function with g( ) = . A sequence Z = (z k ) k is called a g-net if and only if (i) The disks D(z k , g( − |z k |)), k ∈ N, are mutually disjoint, (ii) There exists C > such that k∈N D(z k , Cg( − |z k |)) = D.
Another way to think of it is that a g-net is a maximal g-separated sequence. For n large enough, up to multiplicative constants, a g-net will have g( −n ) − points in each domain S n,k .

. Uniformly dense disks
The sampling sets we consider here are no longer discrete. Recall a class of sequences considered by Ortega-Cerdà and Seip in [40].

De nition 4.9. A sequence
In the terminology of the previous subsection, those are g-nets with a constant g. Let φ be a non-decreasing continuous function, bounded by some constant less than . Given a uniformly dense Λ, de ne D φ k = D(λ k , φ( − |λ k |)) and

Theorem 4.10. [29, Theorem 4.5] The set Λ(φ) is sampling for N if and only if
This condition actually also characterizes determination sets for the space of subharmonic functions in the disk having the characteristic growth of the Nevanlinna class [29, Section 5]. Condition (9) is equivalent to the fact that the harmonic measure of the exterior boundary ∂D of D \ Λ(φ) is zero, see [40,Theorem 1]. Notice also that for any xed K > , condition (9) is equivalent to n log( /φ(K −n )) = ∞ .
The above family of examples allows us to see that there is no general relationship between A −α -sampling sets and Nevanlinna sampling sets. We recall this former, and better known, notion of sampling. Let Λg be a g-net associated to a function g with dt t / g(t) < ∞, for instance g(t) = t / . According to Theorem 4.7, Λg is not a Nevanlinna sampling set. On the other hand, for any given α > , we can extract a maximal separated sequence Λ with a separation constant small enough so that D−(Λ ) > α, hence Λg is A −α -sampling for all α > .
Alternatively, take a set Λ as in Theorem 4.8, sampling for N, with lim n→∞ −rn+ −rn = . Then D−(Λ) = , so it cannot be A −α -sampling for any α > .

Finitely Generated Ideals
Here we report mostly on [20].
In the study of the uniform algebra H ∞ it is important to know its maximal ideals, and thus to know whether an ideal is or is not the whole of the algebra. When considering the ideal generated by functions f , . . . , fn ∈ H ∞ , an easy necessary condition for it to be the whole of the algebra is inf z∈D |f (z)| + · · · + |fn(z)| > . This turns out to be su cient: this is the content of the Corona Theorem.
More generally, we denote by I H ∞ (f , . . . , fn) the ideal generated by the functions f , . . . , fn in H ∞ . The general structure of these ideals is not well understood (see the references in [20]).
In certain situations the ideals can be characterized by growth conditions. In this context, the following ideals have been studied: It is obvious that . . . , fn). This leads us to the type of results we are interested in here. Tolokonnikov [46] proved that the following conditions are equivalent:  [18]).
It will be enough to take a holomorphic g such that |g| = e −H , and conclude using the fact that g is invertible in N.
The ideal corresponding to J H ∞ in N is de ned as: It is clear that I  . . . , fn). On the other hand, an example similar to that given for the H ∞ case shows that when n ≥ , the converse fails. This example is given by two Nevanlinna interpolating Blaschke products B , B with respective zero-sets Z , Z . In this situation I (B , B , B B ) = J(B , B , B B ).

. Stable rank and two open problems
The ideal generated by an n-tuple of functions f , . . . , fn ∈ H ∞ satisfying the conclusion of the Corona Theorem 5.1 is the whole of H ∞ . In general, in an algebra A (or even a unitary ring) an n-tuple which is not contained in any ideal smaller than A is called unimodular. An unimodular n + -tuple f , . . . , fn , f n+ ∈ A such that there exist h , . . . , hn ∈ A with f + h f n+ , . . . , fn + hn f n+ unimodular is called reducible.
De nition 5.5. The stable rank of an algebra A is the smallest integer n such that any n + -tuple in A is reducible.
It is known, for example, that the disc algebra (consisting of the holomorphic functions countinuous up to the boundary) has stable rank 1 (see [8], [24]). The following result by Treil goes in the same direction. Theorem 5.6 (Treil, 1992, [48]). The stable rank of H ∞ is equal to ; more explicitly, given any f , As far as we know, the stable rank for the Nevanlinna class is unknown, but the following result shows that it is at least two.

Proposition 5.7. ([20, Proposition 5.1]) The stable rank of the Nevanlinna class is at least 2.
Open problem: What is the stable rank of N? Our guess is that it is either two or in nity. In support of the rst option is that any triple (f , f , f ) ∈ N such that for some i the zeros of f i form a Nevanlinna interpolating sequence, can be reduced.
In order to see that the stable rank of N cannot be 1 we construct a pair of Blaschke products B , B such that (B , B ) is unimodular but for which there are no ϕ ∈ N and no H harmonic in D such that This prevents the existence of e f ∈ N, invertible in N, with Let Z = (z k ) k := ( − −k ) k and B its associated Blaschke product. Take now a point µ k ∈ ( , ) close enough to z k so that Set Z = (µ k ) k and B its Blaschke product.
Since both Z and Z are H ∞ -interpolating and on the segment ( , ), is not di cult to see that for some c > |B (z)| + |B (z)| ≥ e − Re c +z −z , z ∈ D, On the other hand, for any ϕ ∈ N, This prevents (11) from holding for any H = P[ν] harmonic, with ν nite measure on ∂D, since .

. The f problem
In the late seventies T. Wol presented a problem on ideals of H ∞ , known now as the f problem, which was nally solved by S. Treil in [49]. The analogue for the Nevanlinna class is the following: let f , . . . , fn be functions in the Nevanlinna class, and let f ∈ N be such that there exists H ∈ Har+(D) with |f (z)| ≤ e H(z) (|f (z)| + · · · |fn(z)|) p , z ∈ D, for some p ≥ . Does it follow that f ∈ I(f , . . . , fn) ? As in the H ∞ case, when p > , the ∂ estimates by T. Wol show that the answer is a rmative. When p < the answer is in general negative, as the following example shows. Let N be an integer such that (N + )p > N, f = B N B N , f = B N+ and f = B N+ . Then (12) Open problem: What happens in the case p = ?
Weak embedding property and invertibility threshold . Weak embedding property in H ∞ .
The quotient algebras of H ∞ arise naturally in several questions, notably problems of invertibility of operators. The paper [17] gives more detail about this, in the more general framework of uniform algebras; we shall concentrate on the question of invertibility within a quotient algebra of H ∞ .
Recall that a function I ∈ H ∞ is called inner if lim r→ |I(rξ )| = for almost every ξ ∈ ∂D. It is a consequence of Beurling's theorem about cyclic functions in the Hardy classes that any closed principal ideal of H ∞ is of the form IH ∞ . Given two functions f , g in the same equivalence class [f ] ∈ H ∞ /IH ∞ , they always coincide on I − { } =: Z. If [f ] is invertible in H ∞ /IH ∞ , then inf Z |f | > . One may ask whether the converse holds. This is the n = case of the following property (see [18]).
De nition 6.1. Let I be an inner function with zero set Z. We say that H ∞ /IH ∞ has the Corona Property if: for some δ > , then there exist g , . . . , gn , h ∈ H ∞ such that f g + · · · + fn gn = + hI. Among inner functions with the same zero set, Blaschke products are the largest, and thus most likely to enjoy the WEP. Vasyunin's condition (7) (see Section 3.3) shows that H ∞ -interpolating Blaschke products satisfy the weak embedding property with η = ϵ/C. Thus for any Carleson-Newman Blaschke product, i.e. a nite product of interpolating Blaschke products, we will have the WEP with η = ϵ N /C, and if it is satis ed for I with that value of η, then I is a Carleson-Newman Blaschke product [14], [2]. But there are other functions with the WEP [18]. No geometric characterization is known for the WEP, but further examples and results can be found in [2], [4].

. Weak embedding property in the Nevanlinna Class.
In the algebra N, any nonvanishing function is invertible and so any principal ideal is generated by some Blaschke product B with zero set Z. The elements of the quotient algebra N B = N/BN are in one-to-one correspondence with their traces over Z.
De nition 6.3. Let B be a Blaschke product with zero set Z. We say that the Corona Property holds for N B if for any positive integer n and any f , . . . , fn ∈ N for which there exists H ∈ Har+(D) such that there exist g , . . . , gn , h ∈ N such that f g + · · · + fn gn = + Bh, that is, there exist g , . . . , gn ∈ N such that f (z)g (z) + · · · + fn(z)gn(z) = , z ∈ Z.
Observe that condition (13) is necessary, and that the case n = simply expresses invertibility in N B . Observe also that when Z is an interpolating sequence and n = , condition (13) means that log /|f (z k )| ≤ H(z k ), and so the values /f (z k ) can be interpolated by g ∈ N. So in the Nevanlinna case we immediately see that interpolating sequences are related to the Corona property for the quotient algebra. As in the case of H ∞ , we can see that the Corona property for N B can be reduced to bounding |B| from below, except that in this case both the distance condition and the bounds have to be expressed in terms of positive harmonic functions rather than constants. Blaschke products satisfying those equivalent properties are said to have the Nevalinna WEP. Once this is known, it becomes clear that a nite product of Nevanlinna-interpolating Blaschke products will have the Nevalinna WEP. Assume that Z = ∪ N j= Z j , with Z j Nevanlinna interpolating, and let H j ∈ Har+(D) be such that the estimate in Theorem 3.10(b) holds for Z j and its associated Blaschke product B j . Now, given z ∈ D with ρ(z, Z) ≥ e −H (z) , so Theorem 3.10(b) holds for the whole Z.
In contrast with the H ∞ case, the converse to this holds. . Some elements of proof.
The class of Nevanlinna WEP sequences is thus quite di erent, and much simpler, than that of WEP sequences. For instance, in the setting of H ∞ , any WEP Blaschke product which is not a nite union of interpolating Blaschke products admits a subproduct which fails to be WEP [27,Lemma 1.3]. The following result follows immediately from Theorem 6.5, but is actually a step in its proof.

Lemma 6.6. Any subproduct of a Nevanlinna WEP Blaschke product is also a Nevanlinna WEP Blaschke product.
To give a avor of how the Nevanlinna version of the WEP di ers from the H ∞ version, we will sketch the proof of this. First we observe that points which are "far away" from the zero set, in the sense of distances measured with positive harmonic functions, can be found in a neighborhood of any point. We will need an auxiliary function.
De nition 6.7. Given a Blaschke sequence Z = (z k ) k , let H Z denote the positive harmonic function de ned by where I k := {ξ ∈ ∂D : |ξ − z k /|z k || ≤ − |z k |} denotes the Privalov shadow of z k .
Since B has no zeros in D(z, e − H (z) ), Harnack's inequalities applied in that disc give |B (z)| ≥ e − H (z) .
In the H ∞ framework, one easily sees from Carleson's Theorem 3.2 that any Blaschke product with separated zero set enjoying the WEP is actually an interpolating Blaschke product. The Nevanlinna analogue holds. In particular, |B λ (z)| ≥ e −H (z) for z ∈ ∂D(λ, e −H (λ) ).
Since B λ has no zeros in D(λ, e −H (λ) ) we can apply the maximum principle to the harmonic function log |B λ | − to deduce that |B λ (λ)| ≥ min The bulk of the proof of Theorem 6.5 is spent on showing that any Nevanlinna WEP Blaschke product must have a zero set which is a nite union of weakly separated subsequences [27,Theorem B and Section 3]. Then Lemma 6.6 shows that each of those subsequences generates a Nevanlinna WEP Blaschke product, which then must be Nevanlinna interpolating by Lemma 6.9. Our last result collects several di erent descriptions of products of exactly N Nevanlinna interpolating Blaschke products, which are then equivalent to the property about the trace space on the zero set given in Theorem 3.13. Analogous results for interpolating Blaschke products were proved by Kerr-Lawson [25], Gorkin and Mortini [14] and Borichev [2].
Given a Blaschke product B and z ∈ D, let |B(N)(z)| denote the value at the point z ∈ D of the modulus of the Blaschke product obtained from B after deleting the N zeros of B closest to z (in the pseudo-hyperbolic metric).
The equivalence between (a), (b) and (d) for N = is stated in Theorem 3.10 (see [20,Theorem 1.2]). A consequence of these characterizations is a kind of stability of the property of being a nite product of Nevanlinna Interpolating Blaschke products.

. Invertibility threshold in the Nevanlinna Class.
Along with the de nition of the WEP, the even more subtle question of the invertibility threshold was raised in [18]: is there c ∈ [ , ) such that given any The case of the WEP corresponds to c = . It was shown in [39] that for any value of c ∈ ( , ), there is some Blaschke product with zero set Z so that under the condition inf Z |f | > c, then f is invertible in H ∞ /BH ∞ , but no c < c will work.
In the case of the Nevanlinna Class, |f | being bounded from below by a small constant will be replaced by |f | being bounded from below by e −H , where H is a large positive harmonic function. To study the question, we need a quantitative version of Theorem 6.4. Since any f ∈ N can be written f = g /g with g , g ∈ H ∞ , invertibility of f is equivalent to that of g . We thus state the result for bounded functions, which is a way of normalizing the functions we are considering.
Observe that a restriction arises which does not occur in the H ∞ case: the two properties below are equivalent only when considering positive harmonic functions which are larger than H Z , the function in De nition 6.7.  So we get a su cient condition (a) for a solution to the invertibility problem in the quotient algebra N/BN, and a necessary condition when the harmonic function under consideration is large enough. This condition is in terms of the following class. The Blaschke products which we exhibit in Theorem 6.14 have a distribution of zeroes tailored to the variation of the positive harmonic functions involved. When the harmonic functions we consider are "too big" compared to the density of zeroes of B, the delicate phenomenon involved in Theorem 6.14(b) disappears, and multiplying by a constant does not a ect membership in H(B).
Recall that the disc is partitioned in the Whitney squares S n,k de ned in (8).
for any dyadic Whitney square Q. Assume that H ∈ H(B); then, for any C > , CH ∈ H(B).
Of course, the result is non-trivial for C > only. Note that the harmonic functions that verify the hypothesis of Theorem 6.12(b) will verify (16). It would be nice, given any Blaschke product, to determine the class H(B). Although this seems out of reach in general, a su cient condition is known. Given a dyadic Whitney square Q, let z(Q) denote its center. Notice that we impose no direct restriction on the values of H in the dyadic squares where no zero of B is present. Moreover, there is a class of Blaschke products for which this su cient condition is also necessary [36, Section 2].
However, as everywhere else, we take the equivalent de nition in terms of harmonic majorants.

De nition 7.1. A harmonic function h is called quasi-bounded
if it is the Poisson integral of a measure absolutely continuous with respect to the Lebesgue measure on the circle, i. e. h = P[w] for some w ∈ L (∂D). Let QB(D) denote the set of quasi-bounded harmonic functions, and QB+(D) the cone of those which are nonnegative.

De nition 7.2.
The Smirnov class N+ is the set of f ∈ N such that log |f | has a quasi bounded harmonic majorant. Equivalently, it is the set of f ∈ N such that log |f (z)| ≤ P(log |f * |)(z).
In terms of the factorization (4), Smirnov functions are the Nevanlinna functions with singular factor S ≡ .
Then, any f ∈ N+ has a factorization of the form where O , O are outer with O ∞ , O ∞ ≤ , S is singular inner, B is a Blaschke product and |α| = .
In particular, unlike Nevanlinna functions, the inverse of a nonvanishing Smirnov function is not necessarily a Smirnov function. Actually, the class N + is the subalgebra of N where the invertible functions are exactly the outer functions.
Other aspects of N+ are better than in N. For instance, (N+, d) is a topological vector space (see [44]). It can also be represented as the union of certain weighted Hardy spaces H (w), and hence given the inductive limit topology [30].
According to the above de nition, we expect the results explained in the previous sections to hold as well for N+ as soon as H ∈ Har+(D) is replaced by H ∈ QB+(D). This is indeed the case most of the times, but not always.
Let us brie y comment how the previous results are modi ed or adapted to N+. This is in accordance with the fact that for h ∈ QB(D) NT lim z→e iθ ( − |z|)h(z) = , for all θ ∈ [ , π).
As for (b), when µ Z = k ( − |z k |)δz k has bounded Poisson balayage, the same argument as before shows that Z is Smirnov interpolating if and only if (6) holds. In particular, in this situation Nevanlinna and Smirnov interpolating sequences are the same. Theorem 3.10 and Corollary 3.11 work according to the general substitution explained above. Same thing with Theorem 3.13.
. Sampling sets for N + Sampling (or determination) sets for the Smirnov class can be de ned as in Section 4, with the standard replacement of N by N+ and Har+(D) by QB+(D). Then the analogue of Theorem 4.3 holds, but much more can be said.
By Theorem 4.2, for Λ ⊂ D to be sampling for N+ it is necessary that |NT(Λ)| = π. But Smirnov functions are determined by their boundary values, so this condition is also su cient (see [29,Theorem 2.3]).

. Finitely generated ideals
Mortini's solution to the corona problem given in Theorem 5.2 works also for N+, with the usual substitution explained above. The same thing happens with Theorem 5.4.
The situation changes for the stable rank. Notice that the example showing that the stable rank of N has to be at least two does not work for N+, because the function Re( +z −z ) is not quasi-bounded (it is the inverse of the Poisson integral of a delta measure on 1). As far as we know the study of the stable rank of N+ is wide open.

. Corona problem in quotient algebras
As mentioned before, the class N+ is an algebra where the invertible functions are exactly the outer functions. So any quotient of the Smirnov class by a principal ideal (which are the only closed ideals [41, Theorem 2]) can be represented by N + I := N+/IN+, where I is an inner function. Carefully following the proof of Theorems 6.4 and 6.5 above, we see that the natural analogues hold for N+, with the usual replacement of Har+(D) by QB(D).
This explains the situation for quotients of type N + B , with B Blaschke product, but not the general situation of quotients by inner functions. Here we say that an inner function I with zero set Z satis es the Smirnov WEP if and only if for any H ∈ QB+(D) there exists H ∈ QB+(D) such that |B(z)| ≥ e −H (z) for any z ∈ D such that ρ(z, Z) ≥ e −H (z) .
Notice that if it does and I = BS, where B is a Blaschke product and S a singular inner function, then since |I| ≤ |B| and I − { } = B − { }, it is clear that B must be Smirnov WEP as well. One may wonder which singular inner functions are admissible as divisors of Smirnov WEP functions, in analogy to the study begun in [3], [4]. It turns out that there aren't any besides the constants. To see this one has to check that the WEP condition forces − log |S| to have a quasi-bounded harmonic majorant, which is only possible when S is a constant.