On a Problem of Bremermann Concerning Runge Domains This work was supported by a Marie Curie International Reintegration Grant , ,

1 Introduction

In [1] Bremermann asked the following question:

”Suppose that

D

is a Stein domain in

C^{n}

such that for every complex line

l

C^{n}

l \ D

is connected. Is it true that

D

is Runge in

C^{n}

?” The question remained open and was mentioned again in a recent book by T. Ohsawa ([4] , page 81). In this paper we will give a negative answer to Bremermann's question. Namely, we will give an example of a bounded, strictly pseudoconvex domain in

C^{n}

with real analytic boundary which is not Runge in

C^{n}

but whose intersection with every complex line is simply connected.

Note that if

D

is bounded the hypothesis of the problem means simply that for every complex line

l

l \cap D

is Runge in

l

. If, in addition, one requires that

l \cap D

is connected as well then it does follow that

D

is Runge. See for example [2] , page 309, Theorem 4.7.8.

For simplicity our construction will be done in

C^{2}

but it can be easily adapted to

C^{n}

for

n \geq 2

. To produce our example we will construct first a bounded, strictly pseudoconvex domain

W \subset C^{2}

with smooth, real analytic boundary which is Runge but its closure is not holomorphically convex. (Note that this is not possible in

C

.) Next we show that, in fact, we can construct

W

as above and moreover it has the following geometric property: for every complex line

l

the set of points where

l

is tangent to

\partial W

is at most finite. If this is the case, then one can show that

l \cap \bar{W}

is polynomially convex, again for every complex line

l

. Finally, we show that an appropriate neighborhood of

\bar{W}

is a counterexample to Bremermann's problem.

2 The Example

The construction will be done in several steps.

First we prove that there exists a bounded domain in

C^{2}

with smooth, real analytic boundary which is strictly pseudoconvex, Runge in

C^{2}

, and its closure is not polynomially convex.

J. Wermer [5] proved that there exists a biholomorphic map

F

from a polydisc

P = {(z_{1}, z_{2}) \in C^{2} : | z_{1} | < a, | z_{2} | < b}

C^{2}

onto an open set

F (P)

C^{2}

such that

F (P)

is not polynomially convex. (Wermer's original result was in

C^{3}

but it can be modified to hold in

C^{2}

as well; see [3] or [4] .) We start with such a map and let

U_{n} : = {z \in C^{2} : | \frac{z_{1}}{a} |^{n} + | \frac{z_{2}}{b} |^{n} < 1}

. Since

U_{n} \subset U_{n + 1} \subset P

and

\cup U_{n} = P

, it follows that there exists

m \in N

such that

F (U_{m})

is not polynomially convex. Set

U : = U_{m}

and

V = F (U)

. If we define

φ : U \to R

φ (z) = \frac{1}{1 - | z_{1} / a |^{2 m} - | z_{2} / b |^{2 m}} + | \frac{z_{1}}{a} |^{2} + | \frac{z_{2}}{b} |^{2} - 1

then

φ

is a strictly plurisubharmonic real analytic function and has only one critical point. Since

F

is a biholomorphism,

φ \circ F^{- 1} : V \to R

has the same properties and it is an exhaustion function for

V

. For

α > 0

let

V_{α} = {z \in V : φ \circ F^{- 1} (z) < α}

. It follows that there exists

α > 0

such that

V_{α}

is not polynomially convex. On the other hand, if

z_{0} = F (0)

(this is the minimum point and the only critical point of

φ \circ F^{- 1}

and

φ \circ F^{- 1} (z_{0}) = 0

) and we choose

B \subset V

a ball centered at

z_{0}

, then there exists

α > 0

such that

V_{α} \subset B

. It follows that

V_{α}

is Runge in

B

(because

φ \circ F^{- 1}

is defined on

B

) and therefore is polynomially convex.

Put

r : = sup {α \in R : V_{α} is polynomially convex}

. From the above observations we deduce that

0 < r < \infty

We claim that

V_{r}

is the example that we are looking for. Indeed

V_{r}

is Runge in

C^{2}

as an increasing union of Runge domains and it has smooth, real analytic boundary because

φ \circ F^{- 1}

has no critical point on the

\partial V_{r}

. We only need to convince ourselves that

{\bar{V}}_{r} = {z \in V : φ \circ F^{- 1} (z) \leq r}

is not polynomially convex. If

{\bar{V}}_{r}

was polynomially convex then it would have a Runge (in

C^{2}

) neighborhood

W

with

W \subset V

. If this was the case then for

ε > 0

small enough

V_{r + ε} \subset W

and

V_{r + ε}

would be Runge in

W

and therefore in

C^{2}

. This would contradict the choice of

r

Let us refrase what we have done so far. We proved that if

V

is a domain in

C^{2}

and

φ : V \to R

is a strictly plurisubharmonic function such that there exist

a_{0} < a_{1}

real numbers with the following properties:

\begin{matrix} {x \in V : φ (x) < a_{1}} \subset \subset V, & {x \in V : φ (x) < a_{0}} is connected and contains C (φ) : = the set of & critical points of φ, & {x \in V : φ (x) < a_{0}} is Runge in C^{2}, & {x \in V : φ (x) < a_{1}} is not Runge in C^{2} \end{matrix}} (*)

Then there exists a unique real number

r = r (φ) \in [a_{0}, a_{1})

such that

V_{r (φ)} : = {x \in V : φ (x) < r (φ)}

is Runge and

{\bar{V}}_{r (φ)} = {x \in V : φ (x) \leq r (φ)}

is not holomorphically convex. Note that

V_{r (φ)}

must be connected since each of its components contains a (minimum) critical point,

V_{r (φ)}

contains

{x \in V : φ (x) < a_{0}}

which is connected and

{x \in V : φ (x) < a_{0}} \supset C (φ)

We also proved that there exists a real analytic function

φ

satisfying (*). We fix such a

φ

. Shrinking

V

we can assume that

\bar{V}

is compact and that

φ

is defined on a neighborhood of

\bar{V}

Next we want to show that there exists

ψ

, a small perturbation of

φ

, which satisfies (*) and in addition it has the following geometric property: for every complex line

l

the set

T (ψ, l) : = {x \in V_{r (ψ)} \cap l : l is tangent to \partial V_{r (ψ)} at x}

is finite.

Indeed: let

U

be an open and connected set such that

C (φ) \subset U \subset \subset V_{r (φ)}

and let

W

be an open and relatively compact neighborhood of

\partial V_{r (φ)}

and

0 < δ < δ^{'} < a_{1} - r (φ)

two real numbers such that

U \subset \subset {x \in V : φ (x) < r (φ) - δ}

and

{x \in V : r (φ) - δ < φ (x) < r (φ) + δ} \subset \subset W \subset \subset {x \in V : φ (x) < r (φ) + δ^{'}}

ε > 0

is small enough then, for every

ψ : V \to R

, a

C^{\infty}

function, if the

sup

norms on

\bar{V}

ψ - φ

\frac{\partial (ψ - φ)}{\partial x_{j}}

\frac{\partial^{2} (ψ - φ)}{\partial x_{j} \partial x_{k}}

j, k = 1, . . ., n

(here we denote

z_{j} = x_{2 j + 1} + i x_{2 j}

) are less than

ε

then

ψ

is strictly plurisubharmonic and satisfies (*). Moreover

C (ψ) \subset U

r (ψ) \in [r (φ) - δ, r (φ) + δ]

and

\partial {x \in V : ψ < s} \subset W

for every

s \in [r (φ) - δ, r (φ) + δ]

. We claim that there exists a real analytic

ψ

such that for every complex line

l

and for every

s \in [r (φ) - δ, r (φ) + δ]

the following set

T (ψ, l, s) : = {x \in \partial {x \in V : ψ (x) < s} \cap l : l is tangent to \partial {x \in V : ψ (x) < s} at x}

is finite. Indeed, if

x_{0}

is not isolated in

T (ψ, l, s)

and if we denote by

u : = ψ_{| l}

then

x_{0}

is not isolated in

{z \in l : u (z) = u (x_{0}), \nabla u (z) = 0}

. On the other hand

u

is real analytic and strictly subharmonic. It follows that around

x_{0}

at least one of the sets

{z \in l \cap V : \frac{\partial u}{\partial x} (z) = 0}

{z \in l \cap V : \frac{\partial u}{\partial y} (z) = 0}

is smooth and then the smooth one is contained in the other one. Hence there exists around

x_{0}

a smooth real analytic curve

C

such that

u_{| C} = u (x_{0})

and

\nabla u_{| C} = 0

. If

{f = 0}

is a local equation for

C

it follows that

ψ_{| C} = u = u (x_{0}) + f^{2} g

. However it is not difficult to see that this condition is not satisfied by a generic real analytic function

ψ

. For example one notices that

det H e s s (u) (x_{0}) = 0

and after a linear change of coordinates we can assume that

f (x) = x +

higher order terms and

g (x_{0}) = 1

. Then a straightforward computation shows that

u

must satisfy the following four conditions at

x_{0}

\frac{\partial^{3} u}{\partial y^{3}} = 0

\frac{1}{4!} \frac{\partial^{4} u}{\partial y^{4}} = [\frac{1}{4} \frac{\partial^{3} u}{\partial x \partial y^{2}}]^{2}

\frac{1}{5!} \frac{\partial^{5} u}{\partial y^{5}} = [\frac{1}{3!} \frac{\partial^{4} u}{\partial x \partial y^{3}} - \frac{1}{8} \frac{\partial^{3} u}{\partial x \partial y^{2}} \frac{\partial^{3} u}{\partial x^{2} \partial y}] \frac{1}{4} \frac{\partial^{3} u}{\partial x \partial y^{2}}

We fix now a

ψ

which satisfies (*) and the geometrical property mentioned above.

Our next goal will be to show that, for every complex line

l

C^{2}

l \cap {\bar{V}}_{r (ψ)}

is polynomially convex (although

{\bar{V}}_{r (ψ)}

is not). Note that

(l \cap {\bar{V}}_{r (ψ)}) \ \bar{l \cap V_{r (ψ)}}

is a finite set (as a subset of

T (ψ, l)

). Hence it suffices to show that

\bar{l \cap V_{r (ψ)}}

is polynomially convex. Let's assume that it is not. Note that

l \cap V_{r (ψ)}

is Runge in

l

(since

V_{r (ψ)}

is Runge in

C^{2}

) and that it has a smooth boundary except at a finite set of points (the set of points of non-smoothness is also a subset of

T (ψ, l)

). As we assumed that

\bar{l \cap V_{r (ψ)}}

is not polynomially convex it follows that there exists a rectifiable loop

γ

l

such that

γ \ (l \cap V_{r (ψ)})

contains only points where the boundary of

l \cap V_{r (ψ)}

l

is not smooth and therefore is finite and

\hat{γ} \cap (l \ (\bar{l \cap V_{r (ψ)}})) \neq \emptyset

(in fact it has a nonempty interior). Using again the finiteness of

T (ψ, l)

it follows that

\hat{γ} \cap (C^{2} \ {\bar{V}}_{r (ψ)}) \neq \emptyset

. We claim that there exists a

C^{\infty}

family of biholomorphisms

{f_{ε} : C^{2} \to C^{2}}_{ε \in R}

such that

f_{0}

is the identity and for

ε > 0

small enough

f_{ε} (γ) \subset V_{r (ψ)}

. Without loss of generality we can assume that

l = {z = (z_{1}, z_{2}) \in C^{2} : z_{2} = 0}

. We write

γ \ (l \cap V_{r (ψ)}) = : {(p_{1}, 0), . . ., (p_{s}, 0)}

and we denote by

(0, q_{1}), . . . (0, q_{s})

the unit inner normals to

\partial V_{r (ψ)}

. We choose

h : C \to C

a holomorphic function such that

h (p_{j}) = q_{j}

and we define

f_{ε} (z) = (z_{1}, z_{2} + ε h (z_{1}))

. It is obvious that

f_{ε}

are biholomorphisms and since

\frac{d f_{ε}}{d ε} (p_{j}, 0) = (0, q_{j})

it follows that

f_{ε}

have the sought properties. Because

\hat{γ} \cap (C^{2} \ {\bar{V}}_{r (ψ)}) \neq \emptyset

and

{f_{ε}}

is a continuous family we deduce that for

ε

small enough

f_{ε} (\hat{γ}) ⊄ {\bar{V}}_{r (ψ)}

. On the other hand

f_{ε} (\hat{γ}) = \hat{f_{ε} (γ)}

and

f_{ε} (γ) \subset V_{r (ψ)} \cap f_{ε} (l)

. It follows from here that

V_{r (ψ)} \cap f_{ε} (l)

is not Runge in

f_{ε} (l)

which is a contradiction since

V_{r (ψ)}

is Runge in

C^{2}

and

f_{ε} (l)

is a closed analytic submanifold in

C^{2}

We are now ready to produce our example. For

ε > 0

we set

W_{ε} : = {x \in V : ψ (x) < r (ψ) + ε}

. It follows from the definition of

r (ψ)

that

W_{ε}

is not Runge in

C^{2}

. We wish to prove that there exists

ε > 0

such that for every complex line

l

W_{ε} \cap l

is Runge in

l

Suppose that this is not the case. Then for

n \in N

large enough there exists a complex line

l_{n}

such that

W_{\frac{1}{n}} \cap l_{n}

is not Runge in

l_{n}

.Note that

{l_{n}}

is a sequence of lines that intersect a given compact subset of

C^{2}

. It contains then a convergent subsequence. By passing to this subsequence we can assume that

{l_{n}}

converges to a line

l

We already proved that

l \cap {\bar{V}}_{r (ψ)}

is holomorphically convex and this implies that there exists

Ω

a Runge open subset of

C^{2}

such that

l \cap {\bar{V}}_{r (ψ)} \subset Ω \subset V

\cap W_{\frac{1}{n}} = {\bar{V}}_{r (ψ)}

and

l_{n}

converges to

l

we deduce that there exists

n_{0} \in N

such that for every

n \geq n_{0}

W_{\frac{1}{n}} \cap l_{n} \subset Ω

. Hence

W_{\frac{1}{n}} \cap l_{n} = (W_{\frac{1}{n}} \cap Ω) \cap l_{n}

On the other hand,

ψ

is a plurisubharmonic function defined on the whole

Ω

which is Stein and therefore

W_{\frac{1}{n}} \cap Ω

, which is a level set for

ψ_{| Ω}

, is Runge in

Ω

. Since

Ω

is Runge in

C^{2}

it follows that

W_{\frac{1}{n}} \cap Ω

is also Runge in

C^{2}

and from here we obtain that

W_{\frac{1}{n}} \cap l_{n}

is Runge in

l_{n}

. This contradicts our assumption.

In conclusion, we proved that for

ε > 0

small enough

W_{ε}

is bounded, strictly pseudoconvex, is not Runge in

C^{2}

and for every complex line

l

C^{2}

W_{ε} \cap l

is Runge in

l

. In the same way as before

W_{ε}

must be connected since each of its components contains a critical point of

ψ

A c k n o w l e d g m e n t s :

I am very grateful to Professor Mihnea Colţ oiu for bringing Bremermann's problem to my attention and to Professor Terrence Napier for very useful discussions.

References

Bremermann, H.J.: Die Charakterisierung Rungescher Gebiete durch plurisubharmonische Funktionen. Math. Ann. 136, 173–186, 1958.
Hörmander, L.: Notions of convexity. Progress in Mathematics, 127. Birkhäuser, 1994.
Fornæss, J.E.; Stensønes, B.: Lectures on Counterexamples in Several Complex Variables. Mathematical Notes, 33. Princeton University Press, 1987.
Ohsawa, T.: Analysis of several complex variables. Translations of Mathematical Monographs, 211. American Mathematical Society, 2002.
Wermer, J.: An example concerning polynomial convexity. Math. Ann. 139, 147–150, 1959.

Cezar Joiţ a Institute of Mathematics of the Romanian Academy P.O. Box 1-764, Bucharest 014700 ROMANIA E-mail address : Cezar.Joita@imar.ro