ON THE IDEAL STRUCTURE OF THE RING

OF ENTIRE FUNCTIONS

MELVIN HENRIKSEN

1. Introduction. Let R be the ring of entire functions, and let K be the com-
plex field. The ring R consists of all functions from K to K differentiate every-
where (in the usual sense).
The algebraic structure of the ring of entire functions seems to have been
investigated extensively first by 0. Helmer [l].
The ideals of R are herein classified as in [2]: an ideal / is called fixed if
every function in it vanishes at at least one common point; otherwise, / is called
free. The structure of the fixed ideals was determined in [1]. The structure of
the free ideals is determined below.
While examples of free ideals are easily given, transfinite methods seem to
be needed to construct maximal free ideals. The latter are characterized below,
and it is shown that the residue class field of a maximal free ideal is always
isomorphic to K; the field theory of E. Steinitz [5] is used.

2. Elementary properties. Many expositions of the elementary properties of

entire functions are available; see [3]. Some of these properties will be repeated
below for the sake of completeness.

DEFINITION 1 . I f f C R , letΛ(f) ^ [z C K \ f(z) =o ] .

NOTE. AS in [l], we take A (/) as an "algebraic set." That is, if z is a

zero of multiplicity m of /, the z appears m times in A (/). The union and inter-
section of two such sets is taken in the same sense.

DEFINITION 2. If / is any subset of R, let A{1) ~ [A(f) \ f G / ] .

(2.1) If /is any nonzero element of R, then A(f) is a closed, discrete (and
hence finite or countably infinite) subset of K in the natural topology of K.
Conversely, given any closed, discrete subset D of K, there is an / C R such
that D ~ A (f). Note that any nonempty closed discrete subset of K either is

Received September 18, 1951; this paper is a part of the author's doctoral thesis
at the University of Wisconsin under the direction of R. H. Bruck and W. F. Eberlein.
Pacific J. Math. 2 (1952), 179- 184
179
180 MELVIN HENRΪKSEN

finite or is a sequence { zn \°^-ι such that lim^-^oo zn = oo.

(2.2) R is an integral domain,

(2.3) If D = { zn i^ = 1 is any closed, discrete subset of K, and if { wn }^=1 is

any sequence of complex numbers, then there is an / £ R such that f{zι)-
Wj ; i = 1, 2, (see [3] p. 33, Exercise 3).

(2.4) If A (/) = A ( g ) , then / = gh, where A is a unit (element with inverse)

of/?.

3. Ideals of R. It will be seen below that the nature of the ideals / of R is

completely determined by the sets A (I). Since an element of R is a unit if and
only if it vanishes nowhere, every function in an ideal vanishes somewhere.
Hence the ideals of R are classified as in [2].

DEFINITION 3. An ideal / of R is called fixed if Πy £ / A (/) is nonempty.

Otherwise / is called a free ideal.

As the definition indicates, the structure of the fixed ideals is very simple.
Helmer determined their structure in the course of his investigation of the
arithmetic properties of entire functions. He noted that if S is any subset of
R9 then any function d such that

A(d)= n Λ(/),

is a greatest common divisor (unique to within a unit factor) of the functions

of S. Moreover, Helmer showed that if the set S is finite, its elements being
t n e n
A* ••* 5 fni there exist elements ei9 , en of R such that d= eι fx +
e
' + n fn Hence we have:

THEOREM 1 (Helmer). // fl9 ••• , fn is any finite set of elements of R,

there exists a function d and elements e t , , en of R such that

and ^ e J j + . . + e J n . Hence any ideal I of R with finite basis is principal,

and so is fixed.

Proof. The proof of the first part is given in [1, Theorem 9]. Clearly if the
ideal / is generated by fl9 , fny then f)f £ / A(f) = A(d)9 where i is a
greatest common divisor of fl9 . • , fn.
 ON THE IDEAL STRUCTURE OF THE RING OF ENTIRE FUNCTIONS 181

COROLLARY 1. No free ideal of R has a finite basis.

COROLLARY 2. No polynomial can belong to a free ideal.

Proof. If / is an ideal of R and if f, g £ /, then by Theorem 1 we have

d £ / where A (d) = A(f) n A{g). Hence any finite number of elements of /
have common zeros. But a polynomial p has only a finite number of zeros. Hence
if p £ /, then all the functions of / would have a common zero, whence / would
be a fixed ideal.
Helmer gave an example of a fixed ideal without finite basis (see [ l , proof
of Theorem 8]).
The example below shows that free ideals exist.

EXAMPLE. Let \zn\<ζ:=.i be any sequence of complex numbers such that

limπ_»oo zn-co. Let {S/v ! = { zyy, zyv+i, ••• !• By (2.1), we can construct for
each natural number N an entire function F^ vanishing precisely on Syy. Let / be
the ideal generated by the FN. It is easily seen that / is a free ideal.

Free ideals are characterized in terms of the families A (/) as in [2, Theorem
36]:

THEOREM 2. Let \Aa}a^A {where A is an index set) be a family of se-

quences of complex numbers \z^ ! such that l i m ^ . ^ z^a' = oo , for all QkζiA.
Moreover, suppose that:
(i) The family \Aa}ar-/ι is closed under finite set intersection.
(ii) Γ\aζ^A Aa is empty.
If Fa= [f £ R I f(za) = 0, all za £ Aa], then the ideal I generated by the families
Fa is free. Conversely, if 1 - { fa\ is any free ideal of R, and / ί α = [z £ K\ fa(z)^
0 ] , then the family { Aa\ satisfies ( i ) and ( i i ) above.

Proof. Suppose / is defined a s above. Since ( i ) holds, / is closed under

subtraction. If / £ /, g £ R, then / is in some Fa9 whence fg i s in the same
Fa. So / i s an ideal, and ( i i ) ensures that it is free.
Conversely, if / is any free ideal, Theorem 1 ensures that ( i ) holds, and ( i i )
follows from the definition of a free ideal.

4. Maximal ideals and their residue class fields. If z0 is any fixed element
of K, l e t / U o ) s [ / £ R | / U o ) = 0 ] . Clearly I{z0) is a fixed ideal of/?. More-
over, the mapping f(z)—*f(z0) i s clearly a homomorphism of R upon K whose
l s a
kernel is l(z0). Hence /(<z 0 ) maximal fixed ideal. Conversely, if / is a
fixed ideal and if Πy^£/ A (/) contains two points zl9 z2 (not necessarily d i s -
tinct), then / i s properly contained in I(zx ) or I(z2 ). Hence we have proved:
182 MELVIN HENRIKSEN

THEOREM 3. Every maximal fixed ideal of R is of the form

= ίfCR\f(zo) = 0]

for some z0 £ K. Moreover, the residue class field of every maximal fixed ideal
is the complex field K.

The maximal free ideals are not so simple in structure. They may, however,
be characterized as follows:

THEOREM 4. A free ideal M of R is maximal if and only if A(M) satisfies:

(iii) // D — \ zn }°°= is any infinite, closed, discrete set of K such that
D n A (f) is nonempty for every f £ M, then D ζi A{M).

Proof. Suppose ill is a free ideal and ( i i i ) holds. If M is not maximal, then
there is an ideal N properly containing M. Suppose g £ N and apply ( i ) of
Theorem 2 to A (/V). Then A(g) n A (f) is nonempty for every/ £ N, and hence
for every / £ M. Hence g £ M by ( i i i ) . Hence M is a maximal free ideal.
Conversely, suppose M is a maximal free ideal. If there were an infinite,
closed, discrete set D violating (iii), then anyg £ R such that A (g) = D would
together with M generate an ideal /V properly containing Λf. Hence ( i i i ) must
hold.

NOTE. This result is similar to a theorem of Hewitt on maximal ideals of

rings of real valued continuous functions; see [2, Theorem 36].

Since maximal free ideals are complicated in structure, it is natural to expect

the same of their residue class fields. First we show:

THEOREM 5. // M is a maximal free ideal, then R/M contains a sub field

isomorphic to the field R(z) of all rational functions of a complex variable z.

Proof. By Corollary 2 of Theorem 1, M can contain no polynomial. Hence

a r e t w o
if Pi9 Pi distinct polynomials, then p t φ p2 (mod M). Hence R/M con-
tains as a subring all polynomials in z. So R/M contains R(z) as a subfield.

COROLLARY. The field K is subfield of R/M. If R is considered as an

algebra over K, then the residue class field R/M may be considered as a divi-
sion algebra containing K as a proper subfield.

Proof. If R is considered to be an algebra over K, the homomorphism of R

upon the quotient algebra R/M is assumed to keep complex numbers fixed. Hence
K is a proper division subalgebra of R/M.
If one does not insist that the complex numbers stay fixed under a homo-
 ON THE IDEAL STRUCTURE OF THE RING OF ENTIRE FUNCTIONS 183

morphism of R upon R/M, we have:

THEOREM 6. // M is a maximal free ideal, then R/M is isomorphic (as a

ring) to K.

We shall establish two lemmas before proving the theorem.

LEMMA 1. The field R/M is algebraically closed.

Proof. Note first that if / C M, then M contains all functions vanishing on

the distinct points of A (f). This is true since the maximality of M ensures that
M contains a function g with simple zeros at the distinct points of A (f); and by
(2.4) it contains all such functions. Now let

Φ(z,X) = / o U ) + Λ U ) X+...+ fniz) Xn

be any polynomial with coefficients / 0 , fi9 ••• , fn C R, where fn is not in M,

n > 0. Choose any sequence ί z^ \ €1 A (M). Now for any fixed k, the function
Φ(z&, X) is a polynomial with coefficients in K and hence h a s n complex roots.
Choose any such root and call it r^. Then construct, by (2.3), a function g C R
such that g(zk) ~ rjς ( k = 1, 2, 3, ). Clearly, by the above, Φ (z, g(z)) =0
(mod M). Hence R/M is algebraically closed.

LEMMA 2. The field R/M contains c elements, where c is the cardinal

number of the continuum.

Proof. Since K contains a countable dense subset, there are only c con-
tinuous functions, and hence only c entire functions. Hence R/M has at most c
elements. But all complex numbers are incongruent (mod M), so R/M has at
least c elements. Hence R/M has precisely c elements.

Proof of Theorem 6. Steinitz has shown [5, p. 125, Section 6] that any alge-
braically closed field of characteristic 0 and of degree of transcendency c over
its prime field is isomorphic to K. Since R/M contains K, it has degree of tran-
scendency at least c. By Lemma 2, R/M can have degree of transcendency at
most c. Hence R/M has degree of transcendency c. By Lemma 1, R/M is alge-
braically closed. Hence R/M is isomorphic to K.

5. Topological considerations. In [4], Schilling investigated the ring R of

entire functions as a topological ring under the topology of uniform convergence
on compact sets. He showed [4, p. 949, Lemma 3] that any closed ideal of R is
principal. Hence, in particular, no free ideal is closed. He also stated [4, p. 952,
Lemma 6] that a maximal ideal M of R is nonclosed if and only if R/M is a
proper extension of K. This is in apparent contradiction with our Theorem 6.
184 MELVIN HENRIKSEN

This apparent discrepancy is easily resolved. Although it is nowhere ex-

plicitly stated, Schilling considered R, R/M, and K as algebras over K. He then
proved the equivalent of our Theorem 5 Thus R/M is a commutative division
algebra containing K properly, which is isomorphic as a ring to K.

REFERENCES

1. O. Helmer, Divisibility properties of integral functions, Duke Math. J. 6(1940),

345-356.
2. E. Hewitt, Rings of real valued continuous functions, I, Trans. Amer. Math. Soc.
64(1948), 45-99.
3. K. Knopp, Theory of Functions, vol. 2, Dover Publications, New York, 1947.
4. 0. F. G. Schilling, Ideal theory on open Riemann surfaces, Bull. Amer. Math.
Soc. 52(1946), 945-963c
5. E. Steinitz, Algebraische Theorie der Kδrper, Walter de Gruyter and Co., Berlin,
1930.

UNIVERSITY OF WISCONSIN AND

UNIVERSITY OF ALABAMA