arXiv:1605.08160v1 [math.

CV] 26 May 2016

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

ANDREAS HARTMANN, XAVIER MASSANEDA, ARTUR NICOLAU

A BSTRACT. In this paper we investigate finitely generated ideals in the Nevanlinna class. We
prove analogues to some known results for the algebra of bounded analytic functions H . We
also show that, in contrast to the H case, the stable rank of the Nevanlinna class is strictly bigger
than 1.

1. I NTRODUCTION
The aim of this paper is to investigate analogues for the Nevanlinna class of some known
results on finitely generated ideals of the algebra H of bounded analytic functions in the unit
disk D, equipped with the supremum norm kf k = sup |f (z)|.
zD

Let us begin by recalling these results. The first one concerns interpolating sequences. A
sequence of points = {n }nN in D is called interpolating for H if for every bounded
sequence {wn }nN of complex numbers, there exists a function f H such that f (n ) =
wn , n N. By a famous result by Carleson [2] a sequence {n }n is interpolating for H if and
only if


Y  k n 


inf
 1  > 0.
nN
n
k
k6=n
A Blaschke product with simple zeros is called an interpolating Blaschke product if its zeros are
an interpolating sequence.

The next important result in the context of this paper is Carlesons corona theorem: every
family {f1 , . . . , fm } of functions in H satisfying
inf

zD

m
X

|fi (z)| > 0

i=1

generates the whole algebra. See [5] or [21]. More generally, we denote by IH (f1 , . . . , fm ) the
ideal generated by the functions f1 , . . . , fm in H . The general structure of these ideals is not
well understood (see the references [1], [4], [7]-[11], [18], [19], [24], [25] for more information).
As it turns out, in certain situations the ideals can be characterized by growth conditions. More
Date: May 27, 2016.
2010 Mathematics Subject Classification. 30E05, 32A35.
Key words and phrases. ideals, Nevanlinna class, corona theorem, interpolation, harmonic measure, stable rank.
Second and third authors supported by the Generalitat de Catalunya (grants 2014SGR 289 and 2014SGR 75) and
the Spanish Ministerio de Ciencia e Innovacion (projects MTM2014-51834-P and MTM2014-51824-P).
1

A. HARTMANN, X. MASSANEDA, A. NICOLAU

precisely, the following ideals have been studied:

m
X


|fi (z)| , z D .
JH (f1 , . . . , fm ) = f H : c = c(f ) > 0 , |f (z)| c
i=1

It is obvious that IH (f1 , . . . , fm ) JH (f1 , . . . , fm ). This leads us to the third circle of results
we are interested in here. Tolokonnikov [24] proved that the following conditions are equivalent:
(a) JH (f1 , . . . , fm ) contains an interpolating Blaschke product,
(f1 , . . . , fm ) contains an interpolating Blaschke product,
(b) IH P
2

(c) inf m
i=1 (|fi (z)| + (1 |z| )|fi (z)|) > 0.
zD

As it turns out, in the special situation of two generators with no common zeros these conditions are equivalent to IH (f1 , f2 ) = JH (f1 , f2 ). In the case of two generators f1 and f2 with
common zeros, we have I(f1 , f2 ) = J(f1 , f2 ) if and only if I(f1 , f2 ) contains a function of the
form BC where B is an interpolating Blaschke product and C is the Blaschke product formed
with the common zeros of f1 and f2 (see [11]).
Let us now turn to the framework we want to discuss in this paper. We are interested in analogues of the above results for the Nevanlinna class N, consisting of the holomorphic functions
f on D such that log+ |f | has a positive harmonic majorant on D. Equivalently, f N if and
only if f is holomorphic on D and
Z
log+ |f (r)|d() < .
lim
r1

Here d denotes the normalized Lebesgue measure on the unit circle.

As a general rule we shall see that the results for H translate to the Nevanlinna setting
provided that the boundedness of the elements described above is replaced by a control given
by a positive harmonic majorant (or minorant). Let Har+ (D) be the cone of positive harmonic
functions in the unit disk D. Recall that any H Har+ (D) is the Poisson integral of a positive
measure on the unit circle, that is
Z
H(z) = P [](z) =
P (z, )d(),
D

where

P (z, ) = Re
is the Poisson kernel in D.

+z
z

1 |z|2
| z|2

It is a standard fact that functions f in the Nevanlinna class admit non-tangential limits f
at almost every point of the circle. It is also well-known that any f N can be factored as
f = BSE, where B is a Blaschke product containing the zeros of f , S is a singular inner
function and E is the outer function:
Z

+z

E(z) = C exp
log |f ()|d() ,
D z
where |C| = 1. In particular

log |E(z)| = P [log |f |](z),

z D.

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

A function S is singular inner if there exists a positive measure on D singular with respect to
the Lebesgue measure such that
 Z

+z
S(z) = exp
d() , z D .
D z
For the Nevanlinna class R. Mortini observed that a well known result of T. Wolff implies the
following corona theorem (see [17] or [16]).
Corona Theorem for N (R. Mortini). Let I(f1 , . . . , fm ) denote the ideal generated in N by a
given family of functions f1 , . . . , fm N. Then I(f1 , . . . , fm ) = N if and only if there exists
H Har+ (D) such that
m
X
|fi (z)| eH(z) , z D.
i=1

We need to define the ideal corresponding to JH in N. This will be done in the following
way:
m
X


H(z)
|fi (z)| , z D .
J(f1 , . . . , fm ) = f N : H = H(f ) Har+ (D) , |f (z)| e
i=1

It is clear that I(f1 , . . . , fm ) J(f1 , . . . , fm ). Let us also mention that, by the previous corona
theorem, in the case when J(f1 , . . . , fm ) = N, then I(f1 , . . . , fm ) = N.
Recall that a sequence space is called ideal if it is stable with respect to pointwise multiplication by bounded sequences. For the following definition see also [13].
Definition. A sequence of points = {n }n in D is called interpolating for N (denoted
Int N) if the trace space N| is ideal.
Equivalently, Int N if for every bounded sequence {vn }n of complex numbers there exists
f N such that
f (n ) = vn ,

n N.

Interpolating sequences for the Nevanlinna class were first investigated by Naftalevitch [20] starting from an a priori fixed target space which forces interpolating sequences to be confined in a
finite union of Stolz angles.
A rather complete study, based on the above definition, was carried out much later in [13]. In
particular, it was proved that a sequence {n }n is interpolating for N if and only if there exists a
positive harmonic function H Har+ (D) such that


Y  k n 
H(n )


(1.1)
, n N.
1  e
n k
k:k6=n
Moreover, it was also shown that if Int N, then the trace space is given by


(1.2)
N| = {wn }n : H Har+ (D) , log+ |wn | H(n ) .

A. HARTMANN, X. MASSANEDA, A. NICOLAU

It was also noticed that in the previous condition only the factors corresponding to k close to n
are relevant. More precisely, fixed any c (0, 1), the condition
Y  k n 
H(n )


(1.3)
, nN
1  e
k:k6=n

n k

(k ,n )c

is sufficient for to be interpolating (see [13, Proposition 4.1]).

A Blaschke product the zeros of which forms an interpolating sequence for the Nevanlinna
class is called a Nevanlinna interpolating Blaschke product.
The analogues of the results mentioned above in the context of H read as follows.
Theorem 1.1. Let f1 , . . . , fm be functions in N. Then the following conditions are equivalent:
(a) I(f1 , . . . , fm ) contains a Nevanlinna interpolating Blaschke product,
(b) J(f1 , . . . , fm ) contains a Nevanlinna interpolating Blaschke product,
(c) There exists a function H Har+ (D) such that
m
X

(|fi (z)| + (1 |z|2 )|fi (z)|) eH(z) , z D.

i=1

In case m = 2, if f1 and f2 have no common zeros, the above conditions are equivalent to
(d) I(f1 , f2 ) = J(f1 , f2 ).
As in H each of the conditions (a)-(c) implies I(f1 , . . . , fm ) = J(f1 , . . . , fm ). However,
when m 3, the converse fails, as will be explained after the proof of the result. Also, like in
the H -situation, if the two generators f1 and f2 have common zeros, then I(f1 , f2 ) = J(f1 , f2 )
if and only if I(f1 , f2 ) contains a function of the form BC where B is a Nevanlinna interpolating
Blaschke product and C is the Blaschke product formed with the common zeros of f1 and f2 .
Our proof of Theorem 1.1 uses some of the ideas from both [24] and [11], but also some
specific properties of the Nevanlinna class. In particular we will make use of a new description
of Nevanlinna interpolating sequences in terms of harmonic measure, which we discuss now.
Denote by


 zw 


(z, w) = 
1 zw 

the pseudohyperbolic distance in D, and by D(z, r) = {w D : (z, w) < r} the corresponding

disk of center z and radius r (0, 1). Let B denote the Blaschke product with zeros = {n }n
and let
B(z)
n n z
,
B
(z)
=
bn (z) =
.
n
nz
|n | 1
bn (z)
Q
In these terms B(z) = n bn (z) and |bn (z)| = (z, n ). Given H Har+ (D), consider the
disks DnH = D(n , eH(n ) ) and the domain
[
H
=
D
\
DkH .
n
k:k6=n

(k ,n )1/2

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

It will be clear from the proof of Theorem 1.2 below that the choice of the constant 1/2 in the
definition of H
n is of no relevance; it can be replaced by any c (0, 1). Let (z, E, ) denote
the harmonic measure at z of the set E in the domain . The following result collects
several new descriptions of Nevanlinna interpolating sequences which will be used in the proof
of Theorem 1.1.
Theorem 1.2. Let = {n }n be a Blaschke sequence of distinct points in D and let B be the
Blaschke product with zero set . The following statements are equivalent:
(a) is an interpolating sequence for N, that is, there exists H Har+ (D) such that
(1 |n |2 )|B (n )| = |Bn (n )| eH(n ) ,

n N.

(b) There exists H Har+ (D) such that |B(z)| eH(z) (z, ), z D,
(c) There exists H Har+ (D) such that |B(z)| + (1 |z|2 )|B (z)| eH(z) , z D,
(d) There exists H Har+ (D) such that the disks DkH are pairwise disjoint, and
inf (n , D, H
n ) > 0.

nN

Statement (d) and its proof are modelled after the corresponding version for H , proved in [6].
Descriptions of interpolating and sampling sequences in Bergman spaces in terms of harmonic
measure can be found in [22]. It will be clear from the proof that (d) can be replaced by a
seemingly stronger statement: for every (0, 1) there exists H Har+ (D) such that the disks
DkH are pairwise disjoint, and
inf (n , D, H
n ) 1 .

nN

The paper is organized as follows. In the next section we shall prove Theorem 1.2 and some
Corollaries which will be used later. Section 3 is devoted to the the equivalence of the statements
(a), (b) and (c) of Theorem 1.1 and Section 4 to condition (d) in the case m = 2. At the end
of Section 4 it is also explained that when m > 2 then condition (d) does not imply any of the
previous ones. The last Section is devoted to present two related open problems. The first one
concerns the stable rank of N and the second is a version of the well known f 2 problem of T.
Wolff (see [5, p. 319]), solved by S. Treil in the context of H [27].
A final 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.
It is a pleasure to thank Raymond Mortini for drawing our attention to the Corona Theorem in
the Nevanlinna class and to his paper [17].
2. I NTERPOLATING

SEQUENCES IN THE

N EVANLINNA

CLASS

We start with an elementary lemma.

Lemma 2.1. Let f H with kf k = sup |f (z)| 1.
zD

(a) For all z, D,

|f (w) f ()| 2(w, ) .
(b) Fix 0 < < 1/5. If |f (z)| 4 and (1 |z|2 )|f (z)| for a fixed z D, then
|f (w)| 4 if (z, w) = 2 .

A. HARTMANN, X. MASSANEDA, A. NICOLAU

(c) If (z, w) 1/2, then



(1 |z|2 )f (z) (1 |w|2)f (w) 6 (z, w).

Proof of the Lemma 2.1. (a) This is a direct consequence of Schwarz Lemma:
(f (w), f ()) (w, )

z, D.

(b) Assume first that z = 0 and write f (w) = f (0) + f (0)w + w 2g(w). Since kf k 1 and
|f (0)| 1, we have |g(w)| 3 for every w D, and hence,
|f (w)| |f (0)||w| |f (0)| 3|w|2 , w D.
Since 1/5, then for |w| = 2 we have |f (w)| 3 4 3 4 4 , as desired.
For arbitrary z D we apply the previous argument to the function f z , where
z (w) =

zw
1 zw

is the holomorphic automorphism of D exchanging 0 and z. Since |(f z ) (0)| = (1

|z|2 )|f (z)| and |(f z )(0)| = |f (z)| 4 , taking D such that w = z (), we
get |f (w)| = |(f z )()| 4 if (z, w) = || = 2 .
(c) Again, assume first that z = 0. If || 1/2 then


f (0) (1 ||2)f () |f (0) f ()| + ||2|f ()| |f (0) f ()| +

||2
.
1 ||2

Let g() = f () f (0). For || 1/2 we have

|g()| = |f () f (0)|

1
7
+1 .
2
1 ||
3

Applying (a) to h(z) := 3/7 g(z/2) we deduce that

|g()|

14
||,
3

|| 1/2 .

Finally, if || 1/2, from the above estimate we deduce that

as desired.

2


16
f (0) (1 ||2)f () 14 || + ||
|| 6||,
2
3
1 ||
3

(z) =
For general z D we use the case z = 0 and the invariance by automorphisms of f
2

z )() = (f
)(z ()) for any , z D. Then, for || 1/2,
(1 |z| )f (z) , that is, (f

 

(1 |z|2 )f (z) (1 ||2)(f z ) () = (f z ) (0) (1 ||2)(f z ) () 6|| .

Letting = z (w) and using the invariance we see that (1 ||2)(f z ) () = (1 |w|2)f (w)
and the result follows.


FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

In the proofs we will repeatedly use the well-known Harnack inequalities: for H Har+ (D)
and z, w D,
(2.1)

H(z)
1 + (z, w)
1 (z, w)

.
1 + (z, w)
H(w)
1 (z, w)

In certain parts of this paper, we will need to suppose that z, w are pseudohyperbolically close:
(z, w) < x for some 0 < x < 1, so that (x 1)/(x + 1) H(z)/H(w) (x + 1)/(x 1).
The constant (x + 1)/(x 1) will occasionally be called the Harnack constant.
In this section we shall always assume, without loss of generality, that positive harmonic
functions H Har+ (D) defining pseudohyperbolic neighborhoods D(, eH() ) are big enough
so that the corresponding Harnack constant is at most 2. More specifically, let H Har+ (D) be
such that H(z) ln 3 for any z D; then
(2.2)

H(w)
H(z) 2H(w) if
2

(z, w) eH(z) .

Here is another easy and useful fact.

Lemma 2.2. There exists a universal constant C > 0 such that for any f N with |f (z)|
eH(z) , z D, for some H Har+ (D), one has
(a) For every z D, (1 |z|)|f (z)| eCH(z) , (1 |z|)2 |f (z)| eCH(z) .
(b) For every z, w D with (z, w) 1/3, |f (z) f (w)| (z, w)eCH(z) .
Proof. The estimates in (a) are an easy consequence of Cauchys formula and Harnacks inequality. The estimate in (b) follows immediately from (a) integrating f from z to w and using again
Harnacks inequality.

Proof of Theorem 1.2. (a)= (b). By hypothesis there exists H0 Har+ (D) satisfying Theorem 1.2(a), and therefore the disks DnH0 = D(n , eH0 (n ) ) are pairwise disjoint. We will show
that condition (b) holds with H = CH0 , where C is an absolute constant. Consider the disks
Dn2H0 = D(n , e2H0 (n ) ).
i) Pick z Dn2H0 . By construction, n is the closest point of to z and
|B(z)| = |Bn (z)||bn (z)| = |Bn (z)|(z, )
Since Bn does not vanish in Dn2H0 , by Harnacks inequalities (2.1) and (2.2), there exists an
absolute constant C > 0 such that
|Bn (z)| |Bn (n )|C eCH0 (n ) e2CH0 (z) .
ii) Let := D \ n Dn2H0 . The function B is holomorphic and non-vanishing in . Let F be
the holomorphic function with Re F = 4CH0 on D. Then G = BeF is also holomorphic and
non-vanishing on . For z Dn2H0 , from the preceding case we know that
|G(z)| = |B(z)|e4CH0 (z) = |Bn (z)|(z, n )e4CH0 (z) e2CH0 (z)H0 (n )+4CH0 (z) 1 .
For z D we have |G(z)| = e4CH0 (z) 1. Hence throughout we have |G| 1, that is,
|B(z)| |eF (z) | = e4CH0 (z) for z .

A. HARTMANN, X. MASSANEDA, A. NICOLAU

(b)=(c). We can assume that the function H in (b) satisfies inf{H(z) : z D} ln 3.

Separate into two cases.
i) If (z, ) e10H(z) then, by hypothesis, |B(z)| + (1 |z|2 )|B (z)| |B(z)| e11H(z) .
ii) If (z, ) e10H(z) there exists a unique n such that such that (z, ) = (z, n ). Then
by hypothesis




B(z)
n z)
(1
 eH(z) , z 6= n ,

z n 

and taking the limit as z n , we deduce that (1 |n |2 )|B (n )| eH(n ) . Finally, by

Lemma 2.1(c) and by Harnacks inequality (2.2)
1
(1 |z|2 )|B (z)| eH(n ) 6(z, n ) e2H(z) e8H(z) e2H(z) e3H(z) ,
2
and therefore
|B(z)| + (1 |z|2 )|B (z)| e11H(z) ,
z D.
(c)=(a). This implication is immediate taking z = n .
(a)=(d). Let H Har+ (D) such that |Bn (n )| eH(n ) , n N, that is,
X
1
(2.3)
H(n ),
nN.
log
(
n , k )
k:k6=n

Again the disks DnH are disjoint, and so will be the smaller disks Dn4H . By definition
X
(n , D, H
)
=
1

(n , DkH , H
n
n).
k:k6=n

(n ,k )1/2

Since
(z, Dk4H , D \ Dk4H ) =

log(1/(z, k ))
,
4H(k )

estimate (2.3) is equivalent to

X
4H(k )
(n , Dk4H , D \ Dk4H )
(2.4)
sup
1.
H(n )
nN
k:k6=n

If (n , k ) 1/2 Harnacks inequalities (2.1) imply that 1/3 H(n )/H(k ) 3. Thus, by
(2.4),
X
X
(n , Dk4H , 4H
(n , Dk4H , D \ Dk4H )
n )
k:k6=n

k:k6=n

(n ,k )1/2

(n ,k )1/2

(n , Dk4H , D \ Dk4H )

k:k6=n

3
H(k )
,
H(n )
4

(n ,k )1/2

and therefore
(n , D, 4H
n )

1
.
4

H
Observe that by replacing 4H by NH in the above reasoning it is possible to get (n , D, N
n )
1 3/N.

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

(d)= (a). For simplicity we drop the superscript H in the notations DnH and H
n , and let
H(n )
n = e
. Let = inf (n , D, n ) > 0 and consider the bigger domains
nN
[
n = D \

Dk .
k:k6=n

(n ,k )1/4

n ) (n , D, n ) . Given N 1, to be determined later on,

Notice that then (n , D,
N
let n = D(n , n ) D(n , n ) and
[
Vn = D \
k .
k:k6=n

(n ,k )1/4

n Vn . Define the harmonic functions

Notice that n
Un (z) = (z, D, n )

and un (z) = (z, D, Vn ) .

Then un (z) Un (z) for n . In particular un (n ) > 0, n N. We apply Greens formula

to the functions (z) = log(1/(z, n )) and un on the domain Vn :
Z
1
un dm
un (n ) =
2 Vn
Z 2
Z
X
1
1
1
un
i
=
P (n , e )d
() d(),
log
2 0
2
(
,
)
n
n

k
k:k6=n
(n ,k )1/4

where /n indicates the outer normal derivative. Using the hypothesis and the fact that for
k one has


1
1

log
,
log
(n , )
(n , k )
we deduce that
Z
X

1
1
un
log
() d()
(n , k ) 2 k n
k:k6=n
(n ,k )1/4

k:k6=n

(n ,k )1/4

1
2

log
k

1
un
() d() 1 .
(n , ) n

n
() 0, k and
Taking into account (1.3) we will be done as soon as we prove that u
n
Z
un
1
(2.5)
() d() &
, k 6= n .
H(n )
k n

Define for k 6= n,

un,k (z) = (z, D, Vn k ),

vk (z) = (z, k , D \ k ) =

log(1/(z, k ))
log(1/nN )

and notice that, again by the maximum principle,

(2.6)

un un,k vk

on Vn .

10

A. HARTMANN, X. MASSANEDA, A. NICOLAU

For k , j such that (n , k ), (k , j ) 1/4 we have (n , j ) 1/2 and therefore



[
un,k (k ) k , D, D \
j (k , D, k ) > 0 .
j6=k,n

(j ,k )1/2

By Harnacks inequalities there exists = () > 0 such that un,k (z) > 0 for z Dk .
Also, for z Dk ,
1
log(1/k )
=
vk (z) =
N
N
log(1/k )
and inequality (2.6) yields
un (z) un,k (z) vk (z)

1
,
N

z Dk .

Choose N so that 1/N < /2. Then un (z) /2 for z Dk , k 6= n, and by the maximum
principle

un (z) k (z), z Dk \ k ,
2
where
log((z, k )/kN )
.
k (z) = (z, Dk , Dk \ k ) =
log(1/kN 1 )
Since log(1/k ) = H(k ), this inequality implies that for k
k
1
un

()
() &
log (, k )
n
2 n
H(k ) n
and therefore
Z

un
1
() d() &
n
H(k )

log (z, k ) d() .

n

Finally, we use Greens formula with u 1, v() = log (, k ) and the domain D \ k :
Z
Z
Z

log (, k ) d() =
log (, k ) d() =
P (k , ) d() = 1 .
k n
D n
D

We end this section with two easy consequences which will be used later. The first one says
that Nevanlinna interpolating sequences are stable under convenient pseudohyperbolic perturbations, and will be deduced from Theorem 1.2(d).
Corollary 2.3. Let = {n } be a Nevanlinna interpolating sequence and let H Har+ (D),
satisfying Theorem 1.2(a). If = {n }n D satisfies
1
(n , n ) eH(n ) ,
4
then is also a Nevanlinna interpolating sequence.

n N,

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

11

Proof. We shall use the characterization of Nevanlinna interpolating sequences given in Theorem 1.2(d). Consider the domains
[
[

n = D \
D(k , eH(k ) ) , n = D \
D(k , e2H(k ) )
k:k6=n

k:k6=n

(k ,n )1/4

(k ,n )1/2

Then n n , and by Harnacks inequality there exists c > 0 such that

(n , D, n ) c (n , D, n ) c (n, D, n ).


The result follows then from the hypothesis.

Corollary 2.4. Let be a Nevanlinna interpolating sequence and let H Har+ (D) be such that
inf{H(z) : z D} ln 3 and |B(z)| eH(z) (z, ), z D. Then for every H1 Har+ (D)
with inf{H1 (z) : z D} ln 3, we have
|B(z)| e(2H(z)+2H1 (z)) whenever z
/ n DnH1 .
Proof of Corollary 2.4. Suppose first z
/ n D(n , 1/2). Then (z, ) 1/2 and
1
|B(z)| eH(z) (z, ) eH(z) e2H(z) .
2
Next, if z n D(n , 1/2) picking the closest point 0 with (z, ) = (z, 0 ) eH1 (0 ) ,
Harnacks inequality (2.2) gives
|B(z)| eH(z) (z, ) = eH(z) (z, 0 ) eH(z)H1 (0 ) eH(z)2H1 (z) .

3. P ROOF

OF

T HEOREM 1.1

Notice first that we can assume throughout the proof that the functions fi are Blaschke products. For conditions (a), (b) and (d) this is easily seen by considering the Nevanlinna factorization fi = Bi egi , where Bi is the Blaschke product with the zeros of fi and gi is such that
Re(gi ) = Hi+ Hi , for some Hi+ , Hi Har+ (D). Then, since egi , i = 1, . . . , m, are invertible
functions in N, we have I(f1 , . . . , fm ) = I(B1 , . . . , Bm ) and J(f1 , . . . , fm ) = J(B1 , . . . , Bm ).
As for condition (c), let us now see that there exists H Har+ (D) such that
(3.1)

m
X

(|Bi (z)| + (1 |z|2 )|Bi (z)|) eH(z) ,

zD

i=1

if and only if (c) holds with a suitable, possibly different, H Har+ (D).
Let us first suppose that (3.1) holds. Let Ei = egi , i = 1, . . . , m, and take H1 Har+ (D) such
that


log |Ei (z)| = | Re(gi (z))| H1 (z), z D,
i = 1, . . . , m .

Recall from Lemma 2.2 that

(3.2)

(1 |z|2 )|Ei (z)| eCH1 (z) ,

z D,

where C > 0 is an absolute constant. Fix z D. We shall distinguish two cases.

12

A. HARTMANN, X. MASSANEDA, A. NICOLAU

(i) Assume first that z D is such that

m
X
i=1

Then

m
X

|fi (z)| =

i=1

and (c) holds.

1
|Bi (z)| eH(z)(1+C)H1 (z) .
4
m
X
i=1

1
|Bi (z)||Ei (z)| eH(z)(2+C)H1 (z)
4

(ii) Assume now that

m
X
i=1

1
|Bi (z)| eH(z)(1+C)H1 (z) ,
4

which is in particular bounded by 14 eH(z) . Then by (3.1) we have

m
X

3
(1 |z|2 )|Bi (z)| eH(z) .
4
i=1

Therefore

m
X

3
(1 |z|2 )|Bi (z)||Ei (z)| eH1 (z) eH(z) ,
4
i=1

and by (3.2)

m
X

1
(1 |z|2 )|Bi (z)||Ei (z)| eH(z)H1 (z)
4
i=1

Thus
m
X

(1 |z|

i=1

)|fi (z)|

m
X

(1 |z|

)|Bi (z)||Ei (z)|

i=1

m
X

(1 |z|2 )|Bi (z)||Ei (z)|

i=1

3
1
1
eH(z)H1 (z) eH(z)H1 (z) = e(H(z)+H1 (z))
4
4
2

and so (c) holds.

The converse is based on exactly the same argument. Observe that we can write Bi = fi /Ei =
fi Ei where Ei is an invertible function in N for which we get similar estimates as for Ei . Now,
replacing in the arguments above Bi by fi and Ei by Ei , we will reach (3.1) when starting from
(c).
Before giving the proof of Theorem 1.1 we shall see that (a) implies
I(f1 , . . . , fm ) = J(f1 , . . . , Jm ).
We only have to show the reverse inclusion. For this, let g J(f1 , . . . , fm ) and let H Har+ (D)
be such that
m
X
H(z)
|fi (z)|, z D.
|g(z)| e
i=1

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

13

Let B be a Nevanlinna interpolating Blaschke product in I(f1 , . . . , fm ) and denote by = {n }n

its zero set. Since for any i = 1, . . . , m, we have
P
2
|g(n )fi (n )|
eH(n ) ( m
i=1 |fi (n )|)
Pm
P

meH(n ) ,
n N,
m
2
2
|f
(
)|
|f
(
)|
i=1 i n
i=1 i n

using the description of the trace space N| in (1.2) we see that there exist hi N such that

g(n )fi (n )
hi (n ) = Pm
,
n N.
2
i=1 |fi (n )|
P
Consequently, the function m
i=1 fi hi g vanishes on , and therefore there exists G N such
that
m
X
fi hi g = BG .
i=1

Since BG I(f1 , . . . , fm ), this shows that g I(f1 , . . . , fm ) as well.

Let us now move to the proof of Theorem 1.1.
(a) = (b) is obvious because I(f1 , . . . , fm ) J(f1 , . . . , fm ).

(b) = (c). Assume that B J(f1 , . . . , fm ) is a Nevanlinna interpolating Blaschke product

and let = {n }n denote its zero set. By definition and by Theorem 1.2(b) there exist H, H1
Har+ (D) such that
m
X
H1 (z)
H(z)
(z, )e
|B(z)| e
(3.3)
|fi (z)|, z D.
i=1

Recall from Lemma 2.2 that there exists H2 Har+ (D) such that
|fi (z)| + (1 |z|)|fi(z)| + (1 |z|)2 |fi(z)| eH2 (z) ,

z D, i = 1, . . . , m.

Now let H3 Har+ (D), H3 H + H1 + H2 + ln 3 to be chosen later. Observe that the disks
Dn = DnH3 = D(n , eH3 (n ) ) are disjoint. Observe also that (2.2) holds. By (3.3) and Corollary
2.4, we have
m
X
[
|fi (z)| e2(H(z)+H1 (z)+H3 (z)) , z
/
Dn .
i=1

n1

So, it only remains to discuss the estimate on Dn . We will prove that

m
X
(3.4)
|fi (z)| + (1 |z|)|fi (z)| e6H3 (z) , z Dn .
i=1

We argue by contradiction. Suppose there is a z Dn where this estimate does not hold. Let u
be the closest point of Dn = DnH3 to z, that is, u Dn and (z, u) = (z, Dn ). Then using
a Taylor expansion at z, as Tolokonnikov did in the H -case, for every i = 1, . . . , m, one has
Z u




|fi (u)| = fi (z) + fi (z)(u z) +
(u t)fi (t)dt
z

. |fi (z)| + (1 |z|)|fi(z)|(z, u) + (1 |z|)2 sup |fi (v)|(z, u)2

v[z,u]

6H3 (z)

. e

6H3 (z)H3 (n )

+e

H2 (v)2H3 (n )

+e

14

A. HARTMANN, X. MASSANEDA, A. NICOLAU

where v is a suitable point in Dn . Since (u, ) = eH3 (n ) , using (3.3) we deduce

m
X


(H(u)+H1 (u)) H3 (n )
e
e

|fi (u)| . m 2e6H3 (z) + eH2 (v)2H3 (n ) .

i=1

Harnacks inequality (2.2) gives H3 (z) H3 (n )/2 and we deduce

m
X


(H(u)+H1 (u)) H3 (n )
e
e

|fi (u)| . m 2e3H3 (n ) + eH2 (v)2H3 (n ) .

i=1

Since the functions H, H1 and H2 are fixed and H3 can be taken arbitrarily large, we obtain a
contradiction. Hence (3.4) holds and the statement (c) follows.
(c) = (a). First of all recall that in condition (c) we can assume that the functions fi are
Blaschke products. We can also assume that the positive harmonic function H appearing in
condition (c) satisfies inf{H(z) : z D} > ln(3m). Then Harnacks inequality (2.1) gives that
for any h Har+ (D) one has

4
h(z)
5

if (w, z) < e2H(z) .

5
h(w)
4
Now take C > 1 big enough to be determined later on, and let
m
X


E= zD:
|fi (z)| eCH(z) = n En ,
(3.5)

i=1

where En are the connected components of E. For every n N choose n En , if any, such
that
m
X
|fi (n )| e2CH(n ) ,
i=1

and let = {n }n (we discard those En for which such a n does not exist and keep the
indexation with N). Observe that the sum above is trivially bounded by e2H(n ) .
Claim 1. Assume C 24. Then for every n , one has
D(n , e2CH(n ) ) En D(n , e6H(n ) ) .

The first inclusion is an immediate consequence of Lemma 2.1(a) and Harnacks inequality
(3.5),
|fi (z)| |fi (n )| + 2(z, n ) e2CH(n ) + 2e2CH(n ) = 3e2CH(n )
3e(8/5)CH(z) eCH(z) .
In order to see the second inclusion notice that, by hypothesis, on the set E, and so on En , the
following estimate holds
m
X
(1 |z|2 )
|fi(z)| e2H(z) ,
i=1

and in particular there exists i such that (1 |n|2 )|fi(n )| e2H(n ) /m e3H(n ) = . Thus
by Lemma 2.1(b), for every z with (z, n ) = e6H(n ) = 2 we have |fi (z)| e12H(n ) . By
Harnacks inequality (2.2) we get
(3.6)

|fi (z)| e24H(z)

if (z, n ) = e6H(n ) .

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

15

Thus, taking C 24, we get the desired inclusion.

Observe also that D(n , e6H(n ) ) E = and in particular
(n , k ) max(e6H(n ) , e6H(k ) ),

k 6= n.

Lemma 3.1. The sequence constructed above is interpolating for N.

Proof. We shall use the characterization given in Theorem 1.2(d). Consider the disks DnC =
D(n , e2CH(n ) ) and the domains
[
C
DkC .
n = D\
k:k6=n

(n ,k )1/2

Since DnC En , Harnacks inequality (3.5) and the fact that kfi k 1 give that, for every
i = 1, . . . , m we have
C
log |fi ()| CH() H(k ) if DkC ,
2
log |fi ()| 0
if D.
Hence, by the maximum principle
C
log |fi (z)|
2

H(k )(z, DkC , C

n) ,

z C
n.

k:k6=n

(k ,n )1/2

Notice that, by the separation above, the disk D(n , e6H(n ) ) is contained in C
n . Then, as
established in (3.6) there is i such that
|fi ()| e24H()

if D(n , e6H(n ) ) C
n,

whence
C
2

H(k )(, DkH , H

n ) 24H(),

D(n , eCH(n ) )

k:k6=n

(k ,n )1/2

By Harnacks inequality applied to both H and (, DkC , C

n ), we deduce that
X
192
.
(n , DkH , H
n)
C
k6=n
(k ,n )1/2

Choosing C big enough we finally have

(n , D, H
n) = 1

k:k6=n

1
(n , DkH , H
n) .
2

(n ,k )1/2


Notice that, by Theorem 1.2(a) and the proof above, there exists C0 > 0 such that
Y
(n , k ) eC0 H(n ) , n N .
k:k6=n

16

A. HARTMANN, X. MASSANEDA, A. NICOLAU

Although our choice of {n }n depends on C, the constant C0 is uniform. We indicate to the

reader that the Nevanlinna interpolating Blaschke product we are heading for is not constructed
with the zero-set but with a sequence close to . This, in view of Lemma 2.3, will guarantee
that the new sequence is still interpolating. In the sequel we will need to introduce a new constant
D C C0 , where C 24 is the constant fixed in the preceding discussions. Given a
its harmonic conjugate.
harmonic function G, we denote by G

Claim 2. For every n N there exists i {1, . . . , m} such that gi := fi e12D(H+iH) has a
(i)
unique zero an in Dn6H .
By condition (c) we can assume that for some i {1, . . . , m} (not necessarily unique) we
have (1 |n |2 )|fi (n )| e3H(n ) . Since |fi (n )| e2CH(n ) and C 6, applying again
Lemma 2.1(b) we obtain
|fi (z)| e12H(z)

(3.7)

for z Dn6H .

We use this and Rouches theorem to compare the number of zeros of gi and the function hi =
fi fi (n ) in Dn6H . Observe that hi vanishes at n and (1 |n |)|hi (n )| e3H(n ) , so that
with Lemma 2.1(b), applied to any < e6H(n ) , it can be shown that hi does not vanish at any
other point of Dn6H . Now, for z Dn6H , Harnacks inequality (3.5), D C 24 and (3.7) give

|gi (z) (fi (z) fi (n )) | = |fi (n ) e12D (H+iH)(z) | e2CH(n ) + e12DH(z)

eCH(z) < |fi (z) fi (n )| ,
as desired. This proves the Claim. 
The argument works for every i with (1 |n |2 )|fi (n )| e3H(n ) , but we will pick ain for
only one i. We will denote by i(n) the index in {1, . . . , m} satisfying Claim 2. The previous
(i)
argument with Rouches theorem also allows to show that (an , n ) eCH(n ) . Since C
(i)
C0 , we deduce from Lemma 2.3 that the sequence Ai := {an }n is also interpolating for N. Let
Ii denote the Nevanlinna interpolating Blaschke product with zero set Ai .
P
4CH(z)
for any z D.
Claim 3. Assume C 24. Then m
j=1 |gj (z)|/|Ij (z)| e
P
CH(z)
. Hence, there exists fi
To see this consider first z
/ n Dn6H , so that m
i=1 |fi (z)| > e
2CH(z)
such that |fi (z)| e
, and therefore


m
X  gj (z) 
12DH(z)


(3.8)
e2CH(z) e12DH(z) e3CH(z) .
 Ij (z)  |gi (z)| |fi (z)| e
j=1

Consider now z Dn6H . Notice first that for Dn6H and for i = i(n) by (3.7), we have
|gi ()| e3CH() . Applying the minimum modulus principle to gi /Ii we deduce that


 gi (z) 


inf |gi ()| inf e3CH() e4CH(z) .
 Ii (z)  D
6H
D 6H
n

This finishes the proof of the Claim. 

Since (g1 /I1 , . . . , gm /Im ) is unimodular, by the Corona Theorem for N (see Introduction),
there exist hi N such that
m
X
gi
hi 1
I
i=1 i

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

and

m
X

17

|hi (z)| eM0 H(z) , z D .

i=1

Here M0 = M0 (C) > 0 is a constant which may depend on C but not D, since the estimate in

Claim 3 only depends on C. Since gi = fi e12D(H+iH) , we have

Qm
Qm
m
m
m
X
Y
X
Ik

12D (H+iH)
k=1 Ik
Ik + e
hi k=1 .
)=
(3.9)
F :=
fi (hi
Ii
Ii
i=1
i=1
k=1

Since the function F is obviously in I(f1 , . . . , fm ), we will be done as soon as we show that the
zero set of this function is an interpolating sequence for N. In order to consider the zeros of F
we will again distinguish two cases.
(i)

Observe first that since (an , n ) eCH(n ) , choosing C C0 , and observing that Ai are
Nevanlinna interpolating sequences, we will have
m
Y
(3.10)
Ik (z)| e2C0 H(z) for z D \ n Dn6H .
|
k=1

Since

12D (H+iH)

|e

m
X
i=1

hi

Qm

k=1 Ik

Ii

| e(12D+M0 )H ,

which, choosing D large enough, can be assumed neglectible with respect to eC0 H , we see that
F cannot vanish outside the disks Dn6H .
To consider the disks Dn6H , we again use Rouches theorem to see that F has exactly one
zero in such a disk. Since Ai is Nevanlinna interpolating we can then conclude by applying the
stability
result Lemma 2.3. To apply Rouches theorem we shall compare the function (3.9) with
Qm
I
.
In view of (3.10), for every n N and z Dn6H
k
k=1
Qm
m
m
m
Y

X
Y



12D (H+iH)(z)
k=1 Ik (z)
hi (z)

Ik (z)
 Ik (z) + e
Ii (z)
i=1
k=1
k=1


m
m
Y

X


(12DM0 )H(z)
2C0 H(z)
12DH(z)
|hi (z)| e
e
<  Ik (z) ,
e


i=1

k=1

as desired.

4. T HE

CASE OF TWO GENERATORS

In this section we shall assume m = 2 and prove the equivalence between condition (d)
and (a),(b), (c) in Theorem 1.1. We have already proved that (a) implies that I(f1 , . . . , fm ) =
J(f1 , . . . , fm ) for any m 2. Hence we only need to prove the sufficiency of condition (d) when
m = 2. We start with an auxiliary result which allows to reduce the situation to the case where
B1 and B2 have no common zeros.
be the Blaschke product formed with the common zeros of f1 and f2 . Then
Lemma 4.1. Let B
f2 /B)
= J(f1 /B,
f2 /B).

I(f1 , f2 ) = J(f1 , f2 ) if and only if I(f1 /B,

18

A. HARTMANN, X. MASSANEDA, A. NICOLAU

f2 /B),
then |f | eH (|f1 /B|
+ |f2 /B|)
for some H Har+ (D), and
Proof. If f J(f1 /B,
J(f1 , f2 ) = I(f1 , f2 ) giving f I(f1 /B,
f2 /B).
Conversely, if f J(f1 , f2 ), then
so f B
H
divides f . Hence f /B

|f | e (|f1 | + |f2 |), for some H Har+ (D). In particular B
f2 B)
= I(f1 /B,
f2 /B)
giving f I(f1 , f2 ).
J(f1 /B,

In order to prove the sufficiency of condition (d) when m = 2 we need some more auxiliary
results.
Q
Lemma 4.2. Let 0 < m1 m2 mN 1, with N 2. Assume N
j=1 mj < 1 and
Q
QN
1/2
. Then there exists an integer k with 1 k < N such that kj=1 mj 1/4 and
j=2 mj
QN
1/2
( 1/4 ).
j=k+1 mj

Q
Proof. Let k be the smallest positive integer such that kj=1 mj 1/4 . Observe that k < N,
QN 1
because otherwise j=1
mj > 1/4 and it would follow that mN < 3/4 , and then
N
1
Y

mj < m1 mN < 3/4 < 1/4 ,

j=1

which is a contradiction.
Hence k < N. If k = 1, the conclusion
follows immediately
from the
Qk
Qk1
QN
1/4
1/2
assumption j=2 mj . Next, if k > 1, we have j=1 mj > and j=1 mj 1/4 .
The first estimate gives m1 > 1/4 and hence mj > 1/4 for any j = 1, . . . , N. Then
QN
N
Y

j=1 mj
1/4 1/4 = 1/2 .
mj =
Qk1

mk j=1 mj
j=k+1

Lemma 4.3. Let fi , gi N, i = 1, 2, such that f1 g1 and f2 g2 have no common zeros. If

I(f1 g1 , f2 g2 ) = J(f1 g1 , f2 g2 ), then I(f1 , f2 ) = J(f1 , f2 ).
Proof. We need to show that J(f1 , f2 ) I(f1 , f2 ). Let f J(f1 , f2 ), that is |f | eH (|f1 | +
|f2 |), for some H Har+ (D). Then there exists another H1 Har+ (D) such that |f g1 g2 |
eH1 (|f1 g1 | + |f2 g2 |). By assumption, there exist h1 , h2 N, such that f g1 g2 = f1 g1 h1 + f2 g2 h2 .
Thus f1 g1 h1 vanishes at the zeros of g2 , and since f1 g1 and f2 g2 have no common zeros, so that
f1 g1 and g2 have no common zeros, it is h1 vanishing at the zeros of g2 . We thus may write
h1 = g2 h1 for a suitable h1 N. A similar argument leads to h2 = g1 h2 for some h2 N. Thus

f = f1 h1 + f2 h2 .
Lemma 4.4. Let B be a Blaschke product with zero sequence . Let z D be such that
D(z, ) = and let denote the pseudohyperbolic distance in = D(z, ). Then
1+ (z,w)

1 (z,w)

(a) |B(z)| 1 (z,w) |B(w)| |B(z)| 1+ (z,w) ,

(b) (1 |z|2 )|B (z)|

|B(z)|
1
log
.

|B(z)|2

w ,

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

19

Proof. The estimates in (a) are just Harnacks inequalities rescaled to and applied to the positive harmonic function u = log |B|. To prove (b) let = {n }n . A direct computation shows
that

X
B(z) n 1 |n |2
B (z) =
n z)2 .
(z)
b
|
|
(1

n
n
n=1
Hence

(1 |z|2 )|B (z)|

X
|B(z)| (1 |z|2 )(1 |n |2 )
n z|2

|1
n=1

and we finish by using the estimate log(1/x) 1 x, x > 0, since then

X
X
1
(1 |z|2 )(1 |n |2 ) X
1
2
=
(1 (n , z)) 2
log
= log
.
2

(n , z)
|B(z)|2
|1 n z|
n=1

n=1

n=1

Lemma 4.5. Let = {n }n be a sequence of distinct points in D which is the union of two
Nevanlinna interpolating sequences. Then the trace of N on is


|wk wn | H(n )H(k )
e
< .
N| = {wn }n : H Har+ (D) : sup
k:k6=n (k , n )
It is also true in general that when is the union of n Nevanlinna interpolating sequences then
the trace coincides with the set of sequences such that the pseudohyperbolic divided differences
of order n 1 have a positive harmonic majorant (see [12]).
Proof. Let {wn }n N| and let f N with f (n ) = wn , n N. Let H be a positive
harmonic majorant of log |f |. Given n , k , k 6= n. Define
f (n , k ) =

f (k ) f (n )
.
bn (k )

If (n , k ) 1/2 we get


 f (k ) f (n ) 


1

|f (n , k )| = 
eH(k ) + eH(n ) 2eH(k )+H(n ) .

bn (k )
(n , k )

If (n , k ) < 1/2 apply the maximum principle to the holomorphic function z 7 f (n , z)

and use Harnacks inequalities (2.2) to get
|f (n , k )|

sup

|f (n , )|

:(k ,)=1/2

sup

2eH(n )+H() e3H(n )+3H(k ) .

:(k ,)=1/2
(i)

Let = 1 2 , where i = {n }n are Nevanlinna interpolating sequences, i = 1, 2,

and denote by Bi the corresponding Blaschke products. We will also denote wki = wn when
(i)
k = n . A usual technique to interpolate on finite unions of interpolating sequences is to
look for an interpolating function of the form h0 + B1 h1 , where h0 interpolates on 1 and h1
interpolates suitable values controlled by the divided differences on 2 . Since by assumption

20

A. HARTMANN, X. MASSANEDA, A. NICOLAU

(1)

(1)

(1)

(1)

{wk }k has a majorant eH(k ) , there exists h0 N with h0 (k ) = wk , k N. If we want an

(2)
(2)
interpolating function of the form h = h0 + B1 h1 , with h1 N, then, h(k ) = wk reduces to
(2)

(4.1)

(2)
h1 (k )

(2)

wk h0 (k )
(2)

B1 (k )

k N.

Since 2 Int N we only need to see that the values on the right hand side have a suitable
(2)
(1)
(2)
(2)
(1)
majorant. Given k take k such that (k , 1 ) = (k , k ). There is no restriction in
(2)
(1)
assuming that (k , k ) 1/2, since otherwise the estimate below is immediate. Since 2 is
a Nevanlinna interpolating sequence, by Theorem 1.2(b), there exists H1 Har+ (D) such that
(2)

(2)

(1)

(2)

|B1 (k )| eH1 (k ) (k , k ),

k N,

and therefore
 
 


 w (2) h ((2) )   w (2) w (1)   h ((1) ) h ((2) ) 
 0 k
  k

 k
0
0
k 
k
k

+


(2)

 B1 ((2) )   B1 ((2) )  
B
(
)
1
k
k
k
(2)

(1)

|wk wk |
(1)
(2)
(k , k )

(2)

(1)

|h0 (k ) h0 (k )|
(1)
(2)
(k , k )

(2)

eH1 (k

(1)

.
(2)

By hypothesis the first term between parentheses has a majorant of the form eH(k )+H(k ) . The
second term can be assumed to satisfy the same estimate because of the first inclusion and the
fact that h0 N. Thus, there exists H2 Har+ (D) such that


 w (2) h ((2) ) 
(1)
(2)
(2)
 k

0
H2 (k )+H2 (k ) H1 (k )
k
.

2e
e


 B1 ((2) ) 
k
(2)

(2)

By Harnacks inequality this is bounded by 2e2H2 (k ) eH1 (k ) . Then (1.2) yields the existence
of h1 such that (4.1) holds.


Lemma 4.6. Let = {n }n be a separated Blaschke sequence and let := inf k6=n (k , n ) >
0. Given 0 < n < /2 consider the disks Dn = D(n , n ). Let B1 and B2 be two Blaschke products without common zeros, having each exactly two zeros in each disk Dn . Assume I(B1 , B2 ) =
J(B1 , B2 ). Then there exists H Har+ (D) such that
n > eH(n ) ,

n N.

Proof. The assumptions a priori allow B1 and B2 to have zeros outside n Dn . In order to get
rid of these, let hi be the Blaschke product vanishing on the zeros of Bi which are not in n Dn .
Setting Bi0 = Bi /hi , Lemma 4.3 shows that I(B10 , B20 ) = J(B10 , B20 ) (note that B1 and B2
are assumed to have no common zeros). Thus we can henceforth assume that the zeros of Bi ,
i = 1, 2, are contained in n Dn .
Let cin , din denote the zeros of Bi in Dn , n N, i = 1, 2. Pick the largest of the mutual
distances (c1n , c2n ), (c1n , d2n ), (d1n , c2n ), (d1n , d2n ), say (d1n , d2n ). Then we have
(4.2)

2n (d1n , d2n ) max{(c1n , c2n ), (c1n , d2n ), (d1n , c2n )}.

Q
For i = 1, 2 let Di be the Blaschke product with zeros {din }n and let Ci = Bi /Di =
bcin .
Since is separated and Bi has exactly two zeros on each Dn we deduce from [13, Corollary

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

21

1.9] that Ci and Di are Nevanlinna interpolating Blaschke products. Hence, taking into account
(4.2) there exists H Har+ (D) such that the values






 C1 (d2n )  (c1n , d2n )  (C1 /bc1n )(d2n ) 
(c2n , d1n )  (C2 /bc2n )(d1n ) 
C2 (d1n )
 ,
=


=
 D1 (d2 )  (d1 , d2 )  (D1 /b 1 )(d2 ) 
D2 (d1n )
(d2n , d1n )  (D2 /bd2n )(d1n ) 
dn
n
n
n n
2

are bounded by eH(dn ) and eH(dn ) , respectively. Consequently, there exist h1 , h2 N such that
h1 (d2n )

C1 (d2n )
,
=
D1 (d2n )

h2 (d1n )

C2 (d1n )
=
.
D2 (d1n )

Hence, there are g1 , g2 N with C1 = D1 h1 + D2 g1 and C2 = D2 h2 + D1 g2 . Next we show that

C1 C2 J(B1 , B2 ). Indeed, assume (without loss of generality) that |C2 (z)| |C1 (z)|. Then
|C1 (z)C2 (z)| |(D1 h1 )(z) + (D2 g1 )(z)||C2 (z)| |h1 (z)||B1 (z)| + |g1 (z)||B2 (z)| .
Hence C1 C2 J(B1 , B2 ) = I(B1 , B2 ) so that there exist f1 , f2 N with
C1 C2 = B1 f1 + B2 f2 = C1 D1 f1 + C2 D2 f2 .
Therefore, f1 vanishes at the zeros of C2 and f2 vanishes at the zeros of C1 , and there exist
f1 , f2 N with f2 = C1 f2 and f1 = C2 f1 . Hence
C1 C2 = C1 D1 C2 f1 + C2 D2 C1 f2
and we deduce that 1 = D1 f1 + D2 f2 . Then there exists H1 Har+ (D) such that |D1 | + |D2|
eH1 . Consequently, and since (d1n , d2n ) we can use Harnacks inequalities to deduce that
2

n (d1n , d2n ) |D1 (d2n )| eH(dn ) e2H(n ) .


Let us now move to the proof of (d)=(c) in Theorem 1.1 in the case m = 2. Recall that we
can assume that fi = Bi are Blaschke products. Let i be the zero set of Bi and denote
k(z) =

2
X

(|Bi (z)| + (1 |z|2 )|Bi (z)|) ,

z D.

i=1

In view of [13, Proposition 4.1], for any > 0 there exists H Har+ (D) such that
|Bi (z)| eH (z)

for z with (z, i ) , i = 1, 2 .

Hence, to prove estimate (c) we can assume that z belongs to a Whitney box T (I) = {z = rei
D : ei I, |I|/2 1 r |I|} such that (T (I), i ) 1/2, i = 1, 2. Here I indicates an
arc in D. Let {T (Ij )}j be the collection of Whitney boxes satisfying this condition and pick
j T (Ij ) such that
k(j ) = min k(z) .
zT (Ij )

To prove (c) we need to construct H Har+ (D) such that

(4.3)

k(j ) eH(j ) ,

j N,

since then, by Harnacks inequalities, the inequality propagates to the whole T (Ij ), that is, if
z T (Ij ), we have k(z) k(j ) eH(j ) eCH(z) .

22

A. HARTMANN, X. MASSANEDA, A. NICOLAU

Splitting {j }j into finitely many subsequences if necessary, one can assume that the pseudohyperbolic disks Dj = D(j , 1/2) are pairwise disjoint.
For i = 1, 2 and j N let Bi (j) be the subproduct of Bi formed with the zeros of Bi placed
outside Dj . Then (again using [13, Proposition 4.1], see also (1.3)) there exists H0 Har+ (D)
independent of i and j such that
|Bi (j)(j )| eH0 (j ) ,

(4.4)

j N.

We can also assume that each Dj contains at least two zeros of B1 and two zeros of B2 . Indeed, suppose is the only zero of B1 in Dj (if there is none, then B1 (j) = B1 and k(j )
|B1 (j )| eH(j ) so that there is nothing to do). If (j , ) eH1 (j ) for a suitable fixed
H1 , then since |B1 | = |B1 (j)||b | we get (4.3). If (j , ) eH1 (j ) , first observe that
(1 ||2 )|B1 ()| = |B1 (j)()| e2H0 () . Then, by Lemma 2.2(b) we deduce that, for a
sufficiently big H1 (depending on H0 only), (1 |j |)|B1 (j )|| e3H0 (j ) , which again yields
(4.3).
We can also assume that
k(j ) e100H0 (j ) ,

(4.5)
since otherwise (4.3) holds.
(i)

(i)

For i = 1, 2 and j N let j be a zero of Bi such that (j , j ) = (j , i ). Denote

Bi,j = Bi /b(i) . We claim that there exists a universal constant C > 0 such that
j

|Bi,j (j )| Ck(j )1/2 ,

(4.6)

j N, i = 1, 2 .
(i)

To see this notice first that we have |Bi (j )| k(j ). If (j , j ) k(j )1/2 we obtain (4.6)
from
|Bi (j )|
|Bi,j (j )| =
k(j )1/2 .
(i)
(j , j )
(i)

If (j , j ) k(j )1/2 we use Lemma 2.1(c) to see that




(i)
(i)

(i) 
(1 |j |)|Bi (j )| (1 |j |)|Bi (j )| 6(j , j ) .
(i)

(i)

Since (1 |j |)|Bi (j )| k(j ) we deduce that (1 |j |)|Bi (j )| 7k(j )1/2 , that is,
(i)
(i)
|Bi,j (j )| 7k(j )1/2 . Since (j , j ) k(j )1/2 , by Schwarzs lemma we deduce that
|Bi,j (j )| C1 k(j )1/2 for some C1 > 0 and (4.6) holds also in this case.
For i = 1, 2 and j N let
Ei,j = {z D : Bi (z) = 0 and (z, j ) < 1/2} ,
and let bi,j be the Blaschke product with zeros in Ei,j so that Bi = bi,j Bi (j). Since |Bi (j )|
k(j ), estimates (4.4), (4.5) and (4.6) give
(4.7)
(4.8)

|bi,j (j )| k(j )eH0 (j ) k(j )99/100 ,

Y
|Bi,j (j )|
Ck(j )1/2 eH0 (j ) Ck(j )1/21/100 .
(z, j ) =
|Bi (j)(j )|
zE
i,j
(i)

z6=j

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

23

In order to prove (4.3) we will now split {j }j into different pieces and consider different
cases according to the number of zeros of B1 and B2 in the following neighborhoods of j :
Uj = D(j , k(j )1/10 ) and Uj = D(j , k(j )1/100 ) Uj . Here are the cases we are going to
discuss now:
(i) At least one Blaschke product has at least two zeros in Uj . The set of these j will be
denoted by A1 . Splitting possibly A1 into two subsequences we can assume that B1 has
at least two zeros in Uj (in case B2 has at least two zeros in Uj while B1 has not, inversing
the roles of B1 and B2 yields the exact same estimate). In this case we will distinguish
three subcases.
(i)-a. B2 has at least two zeros in Uj . The set of these j will be denoted by A11 .
(i)-b. B2 has no zero in Uj . The set of these j will be denoted by A12 .
(i)-c. B2 has exactly one zero in Uj . The set of these j will be denoted by A13 .
(ii) Both Blaschke products have at most one zero in Uj . The set of these j will be denoted
by A2 .
We will establish (4.3) in each of these cases.
Case (i)-a. We will start with j A11 . For i = 1, 2 pick two zeros of bi,j in Uj and let bi,j
Q
i =
be the corresponding Blaschke product of degree 2. Consider B
j bi,j where the product

is taken over all j such that j A11 . Since Bi is a subproduct of Bi , the assumption (d) and
1 , B
2 ) = J(B
1 , B
2 ). Applying Lemma 4.6 with j = k(j )1/100 we obtain
Lemma 4.3 give I(B
H Har+ (D) such that
k(j )1/100 eH(j ) , j A11 .
This gives the required estimate (4.3) for the points in A11 .
1 ,
Case (i)-b. The idea in this case is to replace B2 by an appropriate perturbation B2 GB

where B1 is a sub-product of B1 vanishing exactly twice in each Uj , in order to generate two

zeros (controlled by Rouchee theorem) and then conclude as in Case (ii)-a.
j . For each j A12 pick two zeros of B1 in Uj
For j A12 the function b2,j has no zero in U

and let B1 be the Blaschke product formed with these zeros as in case (i)-a. Since Uj Dj and
1 is a Blaschke product whose zeros form a union of two Nevanlinna
the disks Dj are disjoint, B
interpolating sequences [13, Corollary 1.9]. Hence there exists H Har+ (D) such that for every
1 , and z with (z, j ) = k(j )1/30 ,
zero Uj of B
1 (z)| eH(z) (z, ) eH(z) dist(z, Uj ) k(j )1/15 eH(z) .
(4.9)
|B

is the harmonic conjugate of H. By Lemma 4.3, I(B

1 , B2 ) =
Let G = eH+iH , where H
1 , B2 ). Then, observing that G is invertible in N, one has
J(B

1 , B2 ) = I(B
1 , GB
1 B2 ) J(B
1 , GB
1 B2 ) J(B
1 , B2 ) = I(B
1 , B2 ) ,
I(B
hence

1 , GB
1 B2 ) = J(B
1 , GB
1 B2 ) .
I(B
Now, for points z D such that (z, j ) = k(j )1/30 we have, by Lemma 4.4 and the assumption:


1j
1j


1+j

(4.10)
k(j ) 1+j ,
B1 (z)G(z) (B1 (z)G(z) B2 (z)) = |B2 (z)| |B2 (j )|

24

A. HARTMANN, X. MASSANEDA, A. NICOLAU

j . Since (z, j ) =
where j = Uj (z, j ) and Uj indicates the pseudohyperbolic distance in U
k(j )1/30 and Uj = D(j , k(j )1/100 ) Uj , we have
Uj (z, j ) k(j )1/301/100 .
Indeed we can assume j = 0 and let : D Uj be given by (w) = k(j )1/100 w; then
Uj (z, j ) = (

z
|z|
, 0) =
k(j )1/301/100 .
1/100
k(j )
k(j )1/100

Since we can assume that k(j ) is small, say k(j )1/301/100 < , we deduce from (4.10) and
(4.9) that


1


1 (z)G(z)| .

B1 (z)G(z) (B1 (z)G(z) B2 (z)) k(j ) 1+ < k(j )1/15 < |B

1 G B2 has two zeros in D(j , k(j )1/30 ). Observe that we

Then, by Rouches theorem B
1 G B2 by the Blaschke product vanishing on the zeros of B
1 G B2 , and we
can replace B
can thus argue as we have done for A11 (note that k(j ) now only gives the size of Dj , Uj and
Uj , and it only depends on the fact that the Blaschke products under consideration have zeros in
these neighborhoods, but not on the explicit form of these products).
Case (i)-c. Recall that A13 is the set of j A1 such that b2,j has one zero in Uj . If j A13 ,
j must be (2) . Recall from (4.6) that
the zero set of b2,j in U
j

|B2,j (j )| Ck(j )1/2

and B2,j has no zeros in Uj . Hence by (4.4) and (4.5), we deduce that

  B ( ) 
 QB2 (j )
  Q 2,j j


=
 Ck(j )1/2 eH0 (j ) Ck(j )0.49 .
b(2) (j )
b(2) (j )
k

k A13

k6=j

k A13

Thus, replacing B2 by B2 /

k A13

argue as in the previous case.

b(2) we can assume that B2 has no zeros in Uj and we can

j

Case (ii). For j A2 and i = 1, 2 the function bi,j has at most one zero in Uj . If it has one
(i)
zero, this must actually be j . In this case, from (4.4) and (4.6),


 b ( )   B ( ) 
 i,j j   i,j j 
(4.11)
Ck(j )1/2 eH0 (j ) Ck(j )0.49 .
=


 b(i) (j ) 
Bi (j)(j )
j

Hence, replacing bi,j by bi,j /b(i) we can assume that bi,j has no zeros in Uj and satisfies the
j
above estimate (4.11). Observe from this estimate that the initial zero-set Ei,j cannot be reduced
(i)
to the sole point j . We will henceforth assume that Ei,j does not contain any point in Uj . In
order to apply Lemma 4.2 write Ei,j = {ak (i, j) : k = 1, . . . , N}, where the points ak (i, j) are
taken so that the corresponding distances mk = mk (i, j) = (ak (i, j), j ) satisfy m1 m2
mN . In particular a1 (i, j) is the closest point of Ei,j to j , and it is outside Uj , that is,

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

25

(a1 (i, j), j ) k(j )1/10 . According to (4.11), and setting = Ck(j )0.49 , we have
N
Y

mk = |bi,j (j )| Ck(j )0.49 = .

k=1

Moreover
N
Y

mk =

k=2

|bi,j (j )|
Ck(j )0.39 1/2 ,
(a1 (i, j), j )

when k(j ) is sufficiently small (which we can assume). We are thus in the conditions of
Lemma 4.2, which allows to split the product bi,j into two sub-products, denoted by bi,j , b
i,j
each of which is controlled by 1/4 = C 1/4 k(j )0.49/4 . More concretely
bi,j = bi,j b
i,j ,

j N , i = 1, 2 ,

and
(4.12)

0.1225
|bi,j (j )| |b
i,j (j )| C1 k(j )

(if the first inequality does not hold interchange the roles of bi,j and b
i,j ). Let
Bi

bi,j ,

Bi

b
i,j ,

j=1

j=1

where the product is taken over the indices j such that j A2 . For j N and i = 1, 2 we have
(4.13)

|Bi (j )| + |Bi (j )| 2C1 k(j )0.1225 .

Moreover, taking into account (4.4), there exists H0 Har+ (D) such that
(4.14)

|Bi (j )| eH0 (j ) |Bi (j )| .

Split A2 into two sequences A2 = A21 A22 , where

A21 = {j : |B1 (j )| |B2 (j )|} ,

A22 = {j : |B2 (j )| < |B1 (j )|} .

For i = 1, 2 we will construct H2i Har+ (D) such that k(j ) eH2i (j ) for any j A2i .
This will give (4.3) also in this case and finish the proof. Let us explain how to construct H21 .
The same argument applies to H22 . For j A21 pick j D with (j , j ) = |B1 (j )|/4.
Observe that (4.14) yields
(4.15)

|B1 (j )|
|B1 (j )|

|B
(
)|
|B1 (j )| eH0 (j ) 4(j , j ) .
j
1
|B2 (j )|
|B2 (j )|

Since |B1 (j )| |B2 (j )| and (j , j ) = |B1 (j )|/4, Schwarz Lemma (see Lemma
2.1(a)) gives |B2 (j )| |B2 (j )|/2 and |B1 (j )| 3|B1 (j )|/2. Hence, using again
Lemma 2.1(a) and (4.14),
|B1 (j )|
|B1 (j )|

|B
(
)|
3|B1 (j )| 3 |B1 (j )| + |B1 (j ) B1 (j )|)
1
j
|B2 (j )|
|B2 (j )|
|B (j )|

3 |B1 (j )| + 2(j , j )|) = 3 |B1 (j )| + 1
2
(4.16)
C2 eH0 (j ) |B1 (j )| = 4C2 eH0 (j ) (j , j ) ,

26

A. HARTMANN, X. MASSANEDA, A. NICOLAU

where C2 > 0 is an absolute constant. From (4.15) and (4.16) we get



 B1 (j )
|B1 (j )|
B1 (j ) 
|B1 (j )|
H0 (j )


(j , j )
 B ( ) B (j )  |B ( )| + |B (j )| 4(C2 + 1)e
2
2
2
2
j
j

Hence the sequence defined by w(j ) = B1 (j )/B2 (j ) and w(j) = B1 (j )/B2 (j ) is in

the trace space defined on the sequence {j , j }j A21 , and according to Lemma 4.5 we find
h N such that
B1 (j )
,
h(j ) =
B2 (j )

h(j )

B1 (j )
= ,
B2 (j )

j A21 .

Setting b the Blaschke product with zeros j A21 and b the Blaschke product with zeros j ,
we thus get g N such that
B1 = B2 h + bb g .

(4.17)

Since I(B1 , B2 ) = J(B1 , B2 ), Lemma 4.3 yields I(B1 , B2 ) = J(B1 , B2 ). Now (4.17) gives
also I(B1 , B2 ) = I(bb g, B2) and J(B1 , B2 ) = J(bb g, B2 ). Hence, I(bb g, B2 ) = J(bb g, B2),
and again by Lemma 4.3, I(bb , B2 ) = J(bb , B2 ) (observe that bb g and B2 which is a
subproduct of B2 have no common zeros, since by (4.17) those common zeros would be in
common with B1 , which we excluded).
Now notice that bb has two zeros in D(j , |B2 (j )|/2). Also, from (4.12) we can deduce
that D(j , |B2 (j )|/2) D(j , C1 k(j )0.1 ). Hence we are in the same situation as we were
discussing for A1 , now applied to (B2 , bb ), and therefore there exists H21 Har+ (D) such that
h
i
(4.18) |B2 (j )| + (1 |j |) |(B2 ) (j )| + |(bb ) (j )| eH21 (k ) , j A21 .
Now, by (4.12), |B2 (j )| 2C1 k(j )0.1225 . Also B2 has no zeros in Uj , and so Lemma 4.4(b)
gives that
(1 |j |)|(B2 ) (j )|

|B2 (j )|
log |B2 (j )|2 . k(j )0.02 , j A21
k(j )1/10

Moreover, since (bb ) (j ) = b (j )b (j ) and (1 |j |)|b (j )| 2, we get

(1 |j |)|(bb ) (j )| 2(j , j) |B1 (j )|/2
and, again by (4.12), this expression is controlled by C1 k(j )0.1225 . As a result, there exists an
absolute constant C3 > 0 such that the left hand side of (4.18) is upper bounded by C3 k(j )0.02 ,
and we deduce that
C3 k(j )0.02 eH(j ) , j A21 ,


as desired.

Finally let us show that when m 3, condition (d) does not imply the equivalent conditions
(a), (b) or (c) in Theorem 1.1. The example is analogous to the one given in the context of H
in [11]. Let B1 , B2 be Nevanlinna interpolating Blaschke products with zero sets 1 and 2 . We
first claim that
(4.19)

I(B12 , B22 , B1 B2 ) = J(B12 , B22 , B1 B2 ) .

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

27

To prove this we can assume that B1 and B2 have no common zeros. Let f J(B12 , B22 , B1 B2 ).
Then there exists H Har+ (D) such that
(4.20)

|f (z)| eH(z) (|B1 (z)|2 + |B2 (z)|2 + |B1 (z)B2 (z)|),

z D.

Then |f ()| eH() |B1 ()|2 for 2 so that there exists g1 N with g1 () = f ()/B12 (),
2 . This implies that there is g2 N such that f = g1 B12 + B2 g2 . Observe that for
every 1 we have |g2 ()|/|B2 ()| = |f ()|/|B2()|2 which by (4.20) is bounded by eH() .
Hence there exists g3 N with g3 () = g2 ()/B2 (), 1 . Hence, there exists g4 N
with g2 = B2 g3 + B1 g4 . Finally, f = B12 g1 + B22 g3 + B1 B2 g4 and f I(B12 , B22 , B1 B2 ).
Hence (4.19) holds. However, if the sequences 1 and 2 are too close, then using condition
(c) of Theorem 1.1 it can be seen that the ideal I(B12 , B22 , B1 B2 ) cannot contain a Nevanlinna
interpolating Blaschke product.
5. T WO

OPEN PROBLEMS

5.1. The stable rank of the Nevanlinna class. The first open problem we discuss concerns
the stable rank of the Nevanlinna algebra. Recall that an m-tuple (a1 , . . . , am ) of elements of a
commutative unital algebra A is called unimodular if the ideal it P
generates is the whole algebra,
m
that is, there exists an m-tuple (b1 , . . . , bm ) in A such that m
i=1 ai bi = 1. The m-tuple
(a1 , . . . , am ) is called reducible if there exists an (m 1)-tuple (x1 , . . . , xm1 ) in Am1 such
that (a1 + x1 am , . . . , am1 + xm1 am ) is unimodular (so, the ideal generated by (a1 , . . . , am )
contains a specific (m 1)-tuple that already generates A). The stable rank of the algebra is the
least m for which every unimodular m + 1-tuple is reducible.
It is known that the stable rank of the disk algebra and of H is equal to one (see [3] or [14]
for the disk algebra and [26] for H ). The stable rank for the Nevanlinna class is unknown, but
the following result shows that it is at least two.
Proposition 5.1. The stable rank of the Nevanlinna class is at least 2.
It is worth mentioning that any triple (f1 , f2 , f3 ) N 3 such that for some i the zeros of fi
form a Nevanlinna interpolating sequence, can be reduced. The argument uses Theorem 1.2, but
it is lenghty and we do not include the details here.
Open problem: Is the stable rank of N equal to 2?
Proof of Proposition 5.1. Suppose to the contrary that the stable rank of N is one and let us reach
a contradiction. For any unimodular pair of Blaschke products, there will then exist 1 N such
that B1 + 1 B2 is invertible in N, i.e.
(5.1)

B1 + 1 B2 = ef ,

where Re(f ) = H+ H , for some H+ , H+ Har+ (D). We will show that this is not possible
in general. To this end, let 1 = {n }n := {1 2n }n and B1 the associated Blaschke product.
The sequence 1 is H -interpolating. Take now {n }n (0, 1) with (n , n ) small enough so
that
(
1
e 1|n | if n even
|B1 (n )| =
(5.2)
2
e 1|n | if n odd.

28

A. HARTMANN, X. MASSANEDA, A. NICOLAU

Set 2 = {n }n and B2 its Blaschke product.

We shall see first that (B1 , B2 ) is unimodular, i.e, that there exists H Har+ (D) such that
|B1 | + |B2 | eH . Fix a > 0 such that the regions n = D(n , ) D(n , ) are mutually
disjoint. Since 1 and 2 are H -interpolating sequences, there exists > 0 such that
|Bi (z)| , z D \ n n ,

i = 1, 2.

Thus we only need to care about the estimate on D(, ), for 1 2 . So suppose = n or
= n . Since |B1 /bn | and |B2 /bn | are bounded below on D(, ) (by Carlesons condition),
we only need to take care of |bn (z)| + |bn (z)|. By (5.2)
2

(n , n ) = |bn (n )| e 1|n | .
By the triangular inequality |(n , z) (z, n )| (n , n ) , thus either (n , z) or (n , z)
2
are greater than (1/2)e 1|n | . Take now c (independent of n) such that for z D(, ),
2

(1/2)e 1|n | e 1|z| .

With this

|B1 (z)| + |B2 (z)| e 1|z| ,

Since
H0 (z) = Re(

z D(, ) .

1 |z|2
1+z
)=
Har+ (D)
1z
|1 z|2

and

1
,
1 |z|
this finally implies that (B1 , B2 ) is unimodular.
H0 (z)

z D(, )

Let us now show that the pair (B1 , B2 ) cannot be reduced. Equation (5.1) on n yields
log |B1 (n )| = H+ (n ) H (n ) = P [](n ) ,

n N,

where is a finite measure on D such that Re(f ) = P []. Then, since {n }n tends radially
towards 1,
lim (1 |n |2 )P [](n ) = ({1}) .
n

But from (5.2) we see that {(1 |n |2 ) log |B1 (n )|}n has no limit, so we have reached a contradiction.

5.2. The f 2 problem. In the late seventies T. Wolff presented a problem on ideals of H ,
known now as the f 2 problem, which was finally solved by S. Treil in [27]. We now discuss an
analogous problem in the Nevanlinna class. Let f1 , . . . , fn be functions in the Nevanlinna class,
and let f N be such that there exists H Har+ (D) with
(5.3)

|f (z)| eH(z) (|f1 (z)| + |fn (z)|)p ,

z D,

for some p 1. Does it follow that f I(f1 , . . . , fn ) ?

As in the H case, when p > 2, the estimates by T. Wolff show that the answer is affirmative.
When p < 2 the answer is in general negative, as the following example shows. Let N be an
integer such that N + 1 > 2Np, f = B1N B2N , f1 = B1N +1 and f2 = B2N +1 . Then (5.3) holds but
f
/ I(f1 , f2 ) if (B1 , B2 ) is not unimodular in N.

FINITELY GENERATED IDEALS IN THE NEVANLINNA CLASS

29

Open problem: What happens in the case p = 2?

R EFERENCES
[1] Bourgain, J. On finitely generated closed ideals in H (D). Ann. Inst. Fourier (35), n 4 (1985), 163
174.
[2] Carleson, L. An interpolation problem for bounded analytic functions. Amer. J. Math. 80 (1958), 921
930.
[3] Corach, G., Suarez, D. On the stable range of uniform algebras and H . Proc. Amer. Math. Soc. 98
(1986), no. 4, 607610.
[4] Daepp, U., Gorkin, P., Mortini, R. Finitely generated radical ideals in H . Proc. Amer. Math. Soc.
116 (1992), no. 2, 483488.
[5] Garnett, John B. Bounded analytic functions. Revised first edition. Graduate Texts in Mathematics, 236.
Springer, New York, 2007. xiv+459 pp.
[6] Garnett, J. B., Gehring, F. W. , Jones, P. W. Conformally invariant length sums. Indiana Univ. Math. J.
32 (1983), no. 6, 809829.
[7] Gorkin, P., Izuchi, K., Mortini, R. Higher order hulls in H II. J. Funct. Anal. 177 (2000), no. 1,
107129.
[8] Gorkin, P., Mortini, R. Countably generated prime ideals in H . Math. Z. 251 (2005), no. 3, 523533.
[9] Gorkin, P., Mortini, R. Hulls of closed prime ideals in H . Illinois J. Math. 46 (2002), no. 2, 519532.
[10] Gorkin, P., Mortini, R. Allings conjecture on closed prime ideals in H . J. Funct. Anal. 148 (1997),
no. 1, 185-190.
[11] Gorkin, P., Mortini, R. , Nicolau, A. The generalized corona theorem. Math. Ann. 301 (1995), no. 1,
135154.
[12] Hartmann, A., Massaneda, X., Nicolau, A. Traces of the Nevanlinna class on discrete sequences. In
preparation.
[13] Hartmann, A., Massaneda, X., Nicolau, A., Thomas, P.J. Interpolation in the Nevanlinna and Smirnov
classes and harmonic majorants. J. Funct. Anal. 217 (2004), no. 1, 137.
[14] Jones, P. W.; Marshall, D.; Wolff, T. Stable rank of the disc algebra. Proc. Amer. Math. Soc. 96 (1986),
no. 4, 603604.
[15] Kerr-Lawson A., Some lemmas on interpolating Blaschke products and correction, Canad. J. Math., 21
(1969), pp. 531534.
[16] Martin, R. On the ideal structure of the Nevanlinna class. Proc. Amer. Math. Soc. 114 (1992), no. 1,
135143.
[17] Mortini, R. Zur Idealstruktur von Unterringen der Nevanlinna-Klasse N. (German) [On the ideal structure of subrings of Nevanlinna class N] . Publ. du CUL, Sminaire de Math. de Luxembourg, Travaux
mathmatiques I (1989), 8191
[18] Mortini, R. Generating sets for ideals of finite type in H . Bull. Sci. Math. 136 (2012), no. 6, 687708.
[19] Mortini, R. Finitely generated closed ideals in H . Arch. Math. (Basel) 45 (1985), no. 6, 546548.
[20] Naftalevic, A.G., On interpolation by functions of bounded characteristic (Russian), Vilniaus Valst.
Univ. Mokslu Darbai. Mat. Fiz. Chem. Mokslu Ser. 5 (1956), 527.
[21] Nikolski, N.K., Treatise on the shift operator. Spectral function theory. With an appendix by S. V. Hruev
[S. V. Khrushchv] and V. V. Peller. Translated from the Russian by Jaak Peetre. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], 273. Springer-Verlag,
Berlin, 1986
[22] Ortega-Cerd`a, J., Seip, K. Harmonic measure and uniform densities. Indiana Univ. Math. J. 53 (2004),
no. 3, 905923.
[23] Shapiro, J. H.; Shields, A. L. Unusual topological properties of the Nevanlinna class. Amer. J. Math.
97 (1975), no. 4, 915936.
[24] Tolokonnikov, V. A. Interpolating Blaschke products and ideals of the algebra H . (Russian) Investigations on linear operators and the theory of functions, XII. Zap. Nauchn. Sem. Leningrad. Otdel. Mat.
Inst. Steklov. (LOMI) 126 (1983), 196201.

30

A. HARTMANN, X. MASSANEDA, A. NICOLAU

[25] Tolokonnikov, V. A. Blaschke products with the Carleson-Newman condition, and ideals of the algebra
H . (Russian) Zap. Nauchn. Sem. Leningrad. Otdel. Mat. Inst. Steklov. (LOMI) 149 (1986), Issled.
Linein. Teor. Funktsii. XV, 93102, 188; translation in J. Soviet Math. 42 (1988), no. 2, 16031610
[26] Treil, S. The stable rank of the algebra H equals 1. J. Funct. Anal. 109 (1992), no. 1, 130154.
[27] Treil, S. Estimates in the corona theorem and ideals of H : a problem of T. Wolff. Dedicated to the
memory of Thomas H. Wolff. J. Anal. Math. 87 (2002), 481495.
A.H ARTMANN : U NIVERSIT E
F RANCE

DE

B ORDEAUX , IMB, 351

COURS DE LA

L IB E RATION , 33405 TALENCE ,

E-mail address: Andreas.Hartmann@math.u-bordeaux1.fr

X.M ASSANEDA : U NIVERSITAT DE BARCELONA , D EPARTAMENT
BGS-M ATH , G RAN V IA 585, 08007-BARCELONA , S PAIN

DE

`
M ATEM ATIQUES

`
I I NFORM ATICA

&

E-mail address: xavier.massaneda@ub.edu

`
A.N ICOLAU : U NIVERSITAT AUT ONOMA
08193-B ELLATERRA , S PAIN
E-mail address: artur@mat.uab.cat

DE

`
BARCELONA , D EPARTAMENT DE M ATEM ATIQUES
, E DIFICI C,