See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228580877

Normal families: New perspectives

Article  in  Bulletin of the American Mathematical Society · August 1998

DOI: 10.1090/S0273-0979-98-00755-1

CITATIONS READS
361 218

1 author:

Lawrence Zalcman
Bar Ilan University
162 PUBLICATIONS   2,906 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Lawrence Zalcman on 19 May 2015.

The user has requested enhancement of the downloaded file.

BULLETIN (New Series) OF THE
AMERICAN MATHEMATICAL SOCIETY
Volume 35, Number 3, July 1998, Pages 215–230
S 0273-0979(98)00755-1

NORMAL FAMILIES: NEW PERSPECTIVES

LAWRENCE ZALCMAN

Abstract. This paper surveys some surprising applications of a lemma char-

acterizing normal families of meromorphic functions on plane domains. These
include short and efficient proofs of generalizations of (i) the Picard Theorems,
(ii) Gol’dberg’s Theorem (a meromorphic function on C which is the solution
of a first-order algebraic differential equation has finite order), and (iii) the
Fatou-Julia Theorem (the Julia set of a rational function of degree d ≥ 2 is
the closure of the repelling periodic points). We also discuss Bloch’s Principle
and provide simple solutions to some problems of Hayman connected with this
principle.

Over twenty years ago, on the way to a partial explication of the phenomenon
known as Bloch’s Principle, I proved a little lemma characterizing normal families
of holomorphic and meromorphic functions on plane domains [68]. Over the years,
the lemma has grown and, in dextrous hands, proved amazingly versatile, with
applications to a wide variety of topics in function theory and related areas. With
the renewed interest in normal families1 (arising largely from the important role
they play in complex dynamics), it seems sensible to survey some of the most
striking of these applications to the one-variable theory, with the aim of making
this technique available to as broad an audience as possible. That is the purpose
of this report.
One pleasant aspect of the theory is that judicious application of the lemma
often leads to proofs which seem almost magical in their brevity. In such cases, we
have made no effort to resist the temptation to write out complete proofs. Hardly
anything beyond a basic knowledge of function theory is required to understand
what follows, so the reader is urged to take courage and plough on through. And
now we turn to our tale.

1. Let D be a domain in the complex plane C. We shall be concerned with analytic

maps (i.e., meromorphic functions)

f : (D, | |R2 ) → (Ĉ, χ)

Received by the editors October 15, 1997, and, in revised form, May 26, 1998.
1991 Mathematics Subject Classification. Primary 30D45; Secondary 30D35, 34A20, 58F23.
Key words and phrases. Normal families, Picard’s Theorem, algebraic differential equations,
Julia set, Bloch’s Principle.
1 Witnessed by the appearance in recent years of not one, but two treatises on the subject [15],

[59] — after a hiatus of over sixty years!

c
1998 American Mathematical Society

215
216 LAWRENCE ZALCMAN

from D (endowed with the Euclidean metric) to the extended complex plane Ĉ,
endowed with the chordal metric χ, given by
|z − z 0 |
χ(z, z 0 ) = p p z, z 0 ∈ C
1 + |z|2 1 + |z 0 |2
1
χ(z, ∞) = p .
1 + |z|2
A family F of meromorphic functions on D is said to be normal on D if each
sequence {fn } ⊂ F has a subsequence which converges χ-uniformly on compact
subsets of D. It is easy to see that in case all functions in F are holomorphic,
this condition is equivalent to the requirement that each sequence {fn } ⊂ F have
a subsequence which either converges uniformly (with respect to the Euclidean
metric) on compacta in D or diverges uniformly to ∞ on compacta in D.
Normality is, quite clearly, a compactness notion: a family F of meromorphic
functions on D is normal if and only if it is precompact in the topology of χ-uniform
convergence on compact subsets of D. Accordingly, by the Arzela-Ascoli Theorem,
normality is equivalent to equicontinuity on compacta of the functions in F . Since
these are smooth functions, this equicontinuity should be equivalent to the local
boundedness of an appropriate derivative. Such is the content of

Marty’s Theorem. [38] A family F of functions meromorphic on D is normal

on D if and only if for each compact subset K ⊂ D there exists a constant M (K)
such that
(1) f # (z) ≤ M (K)
for all z ∈ K and all f ∈ F .

Here f # denotes the spherical derivative

χ(f (z + h), f (z))
f # (z) = lim
h→0 |h|
0
|f (z)|
= (f (z) 6= ∞).
1 + |f (z)|2
Since χ(z, w) = χ(1/z, 1/w), f # = (1/f )# , which provides a convenient formula
for f # at poles of f.
Armed with this background, we can now engage the subject of our survey.

2. In its most general form, the result we shall be concerned with may be stated
as follows.

Lemma. Let F be a family of meromorphic functions on the unit disc ∆ such that
all zeros of functions in F have multiplicity greater than or equal to ` and all poles
of functions in f have multiplicity greater than or equal to j. Let α be a real number
satisfying −` < α < j. Then F is not normal in any neighborhood of z0 ∈ ∆ if and
only if there exist
(i) points zk ∈ ∆, zk → z0 ;
(ii) positive numbers ρk , ρk → 0; and
(iii) functions fk ∈ F
 NORMAL FAMILIES: NEW PERSPECTIVES 217

such that ραk fk (zk + ρk ζ) → g(ζ) spherically uniformly on compact subsets of C,

where g is a nonconstant meromorphic function. The function g may be taken to
satisfy the normalization
g # (z) ≤ g # (0) = 1 z ∈ C.
This is Pang’s generalization [48], [49] (cf. [65]) of the Main Lemma in [68] (where
α is taken to be 0), with improvements due to Schwick [60] and Chen and Gu [13];
cf. [51].
By way of illustration, consider the non-normal family F = {(2z)k } on ∆. Choos-
ing α = 0, fk (z) = (2z)k , zk = 1/2, and ρk = a/2k, we have fk (zk + ρk ζ) =
(1 + aζ/k)k → eaζ on C. For the normalization g # (z) ≤ g # (0) = 1, choose a = 2.
Clearly, with no special restrictions on the zeros and poles of functions in F ,
the Lemma holds for −1 < α < 1; on the other hand, if all functions in F are
holomorphic (so that the condition on the poles is satisfied vacuously for arbitrary
j), we may take −1 < α < ∞. Similarly, for families of meromorphic functions
which do not vanish, one may choose −∞ < α < 1. In point of fact, for most
applications the choice α = 0 is sufficient. However, for dealing with families of
functions defined by conditions on derivatives, it is essential to be able to choose α
appropriately; cf. §8.
A word is in order concerning the condition that the limit function g have
bounded spherical derivative. This requirement imposes severe restrictions on the
growth of g; specifically g must be of finite order. Recall that the order of a mero-
morphic function g on the plane is given by
log T0 (r)
(2) ρ = lim r→∞ .
log r
Here T0 (r) is the Ahlfors-Shimizu characteristic of g, defined by
Z r
S(t)
T0 (r) = dt,
0 t
where
ZZ
1
S(t) = [g # (z)]2 dxdy
π |z|≤t

is the (normalized) area on the sphere of g({|z| ≤ t}), counting multiplicities. In

particular, if g # is bounded, then T0 (r) = O(r2 ) so that ρ ≤ 2. For holomorphic
functions, even more is true: bounded spherical derivative implies exponential type
[17], cf. [39]. In general, the limit functions which arise by the process described in
the lemma need not have bounded spherical derivative. The possibility of choosing
a function satisfying this additional condition plays an important role in certain
applications; cf. §§3,8.
Let us sketch the proof of the Lemma in the case α = 0. Suppose first that F is
normal on ∆ and that (i)–(iii) hold. Choose r such that |zk | ≤ r < 1. By Marty’s
Theorem, there exists M > 0 such that
max f # (z) ≤ M
|z|≤(1+r)/2

for all f ∈ F . Fix ζ ∈ C. For large k, |zk +ρk ζ| ≤ (1+r)/2, so that ρk fk# (zk +ρk ζ) ≤
ρk M. Thus, for all ζ ∈ C, g # (ζ) = lim ρk fk# (zk + ρk ζ) = 0. It follows that g is a
constant (possibly infinity).
218 LAWRENCE ZALCMAN

Conversely, if F is not normal on ∆, by Marty’s Theorem there exists a number

r∗ , 0 < r∗ < 1, points zk∗ in {|z| ≤ r∗ }, and functions fk ∈ F such that fk# (zk∗ ) →
∞. Fix r, r∗ < r < 1, and let
   
|z|2 # |zk |2
Mk = max 1 − 2 fk (z) = 1 − 2 fk# (zk ).
|z|≤r r r
The maximum exists since fk# is continuous for |z| ≤ r. Clearly we have Mk ≥
(1 − |zk∗ |2 /r2 )fk# (zk∗ ) → ∞. Put
 
1 |zk |2 1
ρk = 1− 2 = # .
Mk r fk (zk )
This works!
(In case z0 is the only point in ∆ at which F fails to be normal, it is automatic
that the sequence {zn } chosen above converges to z0 . Otherwise, a slightly different
argument, amounting to a diagonalization of the previous reasoning, is needed to
insure that zk → z0 ; cf. [60] pp. 242-244.)
3. A central result in the theory of normal families is Montel’s Theorem, according
to which a family of functions meromorphic on a domain D, all of which fail to take
on three fixed (and distinct) values in Ĉ, is normal on D. It is this theorem, which
Schiff calls the Fundamental Normality Test ([59] pp. 54,74), that makes available
the mechanism of normal families for proving global results in (one-dimensional)
complex dynamics [21] p. 9. Here is a simple and elementary proof of Montel’s
Theorem, based on an idea of Antonio Ros [55].
Montel’s Theorem. The collection F of all meromorphic functions which omit
three fixed values a, b, c ∈ Ĉ on a domain D ⊂ C is a normal family on D.
Proof. Since normality is a local notion, we may suppose that D = ∆, the unit
disc. Composing with a linear fractional transformation, we may also assume that
the omitted values are 0, 1, ∞. Let us denote by Fn the collection of functions on
∆ which omit the√ values 0, ∞, and all nth roots of 1, so that F = F1 . Note that
f ∈ F implies n f ∈ Fn , while if h ∈ Fn , then hn ∈ F .
Suppose now that F is not normal. Then none of the families Fn is normal, so
by the Lemma we have, for each n, a nonconstant entire function gn obtained as a
limit of functions omitting all values in Sn = {0, 1, e2πik/n , k = 0, 1, . . . , n − 1}. By
Hurwitz’s Theorem, gn also omits Sn . Moreover, gn# (z) ≤ gn# (0) = 1.
Write, for convenience, Tn = S2n , Gn = g2n , and consider the family G = {Gn }
on C. Now G# n (z) ≤ 1 for all z ∈ C, so by Marty’s Theorem G is normal on C;
hence a subsequence converges, χ-uniformly on compacta, to a limit function G.
Since G# #
n (0) = 1 for all n, G (0) = 1, so G is nonconstant. The sets Tn are nested,
so that Gm omits values in Tn as soon as m ≥ n. By Hurwitz’s Theorem, G must
omit Tn for every n. Since ∪Tn is dense in the unit circle and G(C) is an open
connected set, this implies that either G(C) ⊂ ∆ or G(C) ⊂ C \ ∆. In either case,
we have a contradiction to Liouville’s Theorem.
Immediate (and easy) corollaries of Montel’s Theorem include the theorems of
Picard, as well as the existence of a direction of Julia for entire functions [58]
p. 352. The proof just given, together with the standard deduction of Picard’s
Great Theorem from Montel’s Theorem ([58] p. 351), provides what is to my
mind the shortest and simplest route to this pinnacle of complex function theory.
 NORMAL FAMILIES: NEW PERSPECTIVES 219

Montel’s Theorem itself admits various generalizations ([59] pp. 104-105), some of
which are mentioned in §7 below.

4. According to Marty’s Theorem, a family F of meromorphic functions is normal

on D ⊂ C if and only if for each compact set K ⊂ D there exists a constant
M (K) such that f # (z) ≤ M (K) for all f ∈ F and all z ∈ K. The following lovely
application of the Lemma allows us to reduce drastically the set on which this
inequality is required to hold.
Theorem. (Lappan [36]; cf. [35], [14], [30]) A family F of meromorphic functions
is normal on D ⊂ C if and only if for each compact set K ⊂ D there exists a set
E = E(K) ⊂ Ĉ containing at least five points and a constant M = M (K) > 0 such
that
(3) f # (z) ≤ M z ∈ K, f (z) ∈ E
for all f ∈ F .

Proof. Marty’s Theorem shows that (3) is necessary with E = Ĉ. To prove suffi-
ciency, suppose that (3) holds but F is not normal. Then we can find fk ∈ F ,
zk → z0 ∈ D, and ρk → 0+ such that fk (zk + ρk ζ) = gk (ζ) → g(ζ) spherically
uniformly on compacta, where g is a nonconstant meromorphic function on C. Let
K ⊂ D be a closed disc about z0 and suppose g(ζ0 ) ∈ E. By Hurwitz’s Theorem,
there exist ζk → ζ0 such that fk (zk + ρk ζk ) = gk (ζk ) = g(ζ0 ) for large k. Now by
(3), fk# (zk + ρk ζk ) ≤ M for k sufficiently large, so that

g # (ζ0 ) = lim gk# (ζk ) = lim ρk fk# (zk + ρk ζk ) ≤ lim ρk M = 0.

k→∞ k→∞ k→∞

Thus g # (ζ0 ) = 0 whenever g(ζ0 ) ∈ E, i.e., each value in E is totally ramified.

But it is a well-known consequence of Nevanlinna’s Second Fundamental Theorem
([44] p. 279) that a nonconstant meromorphic function on C can have at most four
totally ramified values. It follows that g is constant, a contradiction.

As a corollary, we have the following sharpening of Schwick’s extension [61] of a

theorem of Royden [56].
Theorem. Let F be a family of functions meromorphic on D ⊂ C with the property
that for each compact set K ⊂ D there is a function hK : [0, ∞] → [0, ∞], which is
finite somewhere on (0, ∞), such that
|f 0 (z)| ≤ hK (|f (z)|)
for all f ∈ F and z ∈ K. Then F is normal on D.
Proof. Fix K ⊂ D and take x0 > 0 such that hK (x0 ) < ∞. Putting M (K) =
hK (x0 ) and E(K) = {w ∈ C : |w| = x0 }, we have, for each f ∈ F ,
f # (z) ≤ |f 0 (z)| ≤ hK (|f (z)|) = hK (x0 ) = M (K)
whenever z ∈ K and f (z) ∈ E(K). By the previous theorem, F is normal on D.

This last result yields the normality of the family of solutions (on some common
domain) of a differential equation of the form f 0 = F (f ), where |F (w)| ≤ h(|w|) for
220 LAWRENCE ZALCMAN

2
some function h as above. For example, the solutions to w0 = e−w /(w + 1) defined
on a common domain form a normal family there since
 
 e−w2  e|w|
2
 
 ≤ .
 w + 1  | |w| − 1|
In the next section, we shall encounter another application to solutions of first-order
differential equations.

5. According to a theorem of A.A. Gol’dberg [26], a meromorphic function on the

plane which satisfies a first order algebraic differential equation must have finite
order. Recently, Bergweiler [5] observed that the Lemma can be exploited to yield
a short proof of a significantly more general result, due to Barsegian [4].
Let f be a meromorphic function on C. For an m-tuple r = (r1 , r2 , . . . , rm ) of
nonnegative integers, define
Mr [f ](z) = f 0 (z)r1 f 00 (z)r2 · · · f (m) (z)rm
and set M(0) [f ](z) ≡ 1. A differential polynomial P [f ] is an expression of the form
X
P [f ](z) = ar (z, f (z))Mr [f ](z),
r∈I

where the coefficients ar are rational in both variables and the index set is finite.
Define the weight of P [f ] by
w(P ) = max w(r),
r∈I

where w(r) = r1 + 2r2 + · · · + mrm .

Theorem. A meromorphic function f on C which satisfies a differential equation
of the form (f 0 )n = P [f ], where P [f ] is a differential polynomial such that n >
w(P ), has finite order.
Proof. Assume, to the contrary, that f has infinite order. From the definition of
order (cf. §2), one sees easily that there exists a sequence {wk }, tending to infinity,
such that
log f # (wk )
(4) → ∞.
log |wk |
Write fk (z) = f (z + wk ) and consider the family F = {fk } on ∆. Since fk# (0) =
f # (wk ) → ∞, F is not normal on ∆. So, by the Lemma, there exist zk ∈ ∆,
ρk → 0+, and a nonconstant meromorphic function g on C such that
f (wk + zk + ρk ζ) = fk (zk + ρk ζ) = gk (ζ) → g(ζ).
Writing τk = wk + zk , we see from the proof of the Lemma (cf. §2) that one may
take ρk = 1/fk# (zk ) = 1/f # (τk ) and f # (τk ) = fk# (zk ) ≥ fk# (0) = f # (wk ). This
implies, via (4), that τk` ρk → 0 as k → ∞ for each fixed `. Substituting τk + ρk ζ for
z into the differential equation, we have
X −w(r)
ρ−n 0
k gk (ζ) =
n
ar (τk + ρk ζ, gk (ζ))ρk Mr [gk ](ζ).
r∈I

Multiply both sides by ρnk and let k → ∞; the left hand side clearly tends to
g 0 (ζ)n , while the right hand side vanishes in the limit since n − w(r) ≥ 1 for each
 NORMAL FAMILIES: NEW PERSPECTIVES 221

r ∈ I and τk` ρk → 0 for each fixed `. Thus g 0 (ζ)n ≡ 0, a contradiction since g is

nonconstant.

Taking somewhat greater care in the organization of the argument, Frank and
Wang [23] have sharpened this result and obtained an explicit bound for the order
of f in terms of the differential equation satisfied by f. An example of an equation
of the form (f 0 )n = P [f ] with solutions of infinite order is
(f 0 )2 = f f 000 − f f 00 − f 0 f 00 ,
z
which is satisfied by f (z) = ee . In this case, n = 2 < 3 = w(P ).

6. We have already alluded to the historically important role played by normal

family techniques in the development of complex dynamics. Even today, judicious
application of the Lemma can result in significant simplifications. We illustrate this
point with a streamlined proof of one of the principal results of the classical theory.
Let f be a rational function of degree d ≥ 2 (or an entire function) and consider
the collection of iterates fn , n ∈ N, where f1 = f and fn = f ◦ fn−1 . A point z is
called periodic if fn (z) = z for some n ∈ N; it is repelling if |fn0 (z)| > 1. The Julia
set J(f ) is the complement of the maximal open set on which {fn } is normal. It
is well-known, and easy to prove ([10] pp. 56-57), that J(f ) is a nonempty perfect
set and that J(f ) = J(fm ) for each m ∈ N. We have the following result of Fatou
and Julia.
Theorem. Let f be a rational function of degree d ≥ 2. Then J(f ) is the closure
of the repelling periodic points.
Proof. (Schwick [62]) We may assume that d ≥ 5; otherwise, consider f3 in place
of f. Let A be the finite set in Ĉ consisting of ∞, f (∞), and all critical values of
f (i.e., values of f at points where f 0 vanishes). We shall prove that the repelling
periodic points are dense in J(f ) \ A; this is sufficient to prove the theorem, as J(f )
is a perfect set.
For w0 ∈ J(f ) \ A, apply the Lemma (with α = 0) to the family F = {fn }.
We obtain an increasing sequence of positive integers {nk } and zk → w0 , ρk →
0+ such that fnk (zk + ρk ζ) = gk (ζ) → g(ζ), χ-uniformly on compacta, where g
is a nonconstant meromorphic function on C. By the choice of w0 , the equation
f (z) = w0 has at least five simple roots z1 , . . . , z5 ∈ C. Since g can have at most
four totally ramified values, for at least one of the zj , say z1 , there exists ζ0 ∈ C
with g(ζ0 ) = z1 and g 0 (ζ0 ) 6= 0.
Consider now
fnk +1 (zk + ρk ζ) − (zk + ρk ζ) → (f ◦ g)(ζ) − w0 .
The right hand side is nonconstant and vanishes for ζ = ζ0 . It follows by Hurwitz’s
Theorem that, for k sufficiently large, fnk +1 (zk + ρk ζ) = zk + ρk ζ has a solution
ζk with ζk → ζ0 . Thus zk + ρk ζk is a fixed point of fnk +1 , i.e., a periodic point
of f, and zk + ρk ζk → w0 . It is repelling for large k since ρk fn0 k +1 (zk + ρk ζk ) =
(f ◦ gk )0 (ζk ) → (f ◦ g)0 (ζ0 ) = f 0 (g(ζ0 ))g 0 (ζ0 ) = f 0 (z1 )g 0 (ζ0 ) 6= 0.

With minor modifications, the argument above also establishes

Baker’s Theorem. [2] The Julia set of a nonlinear entire function is the closure
of the repelling period points.
222 LAWRENCE ZALCMAN

Indeed, one may assume that f is transcendental. It follows from Nevanlinna’s

Second Fundamental Theorem that there exists a (possibly empty) set A ⊂ C
consisting of at most two points such that for w ∈ / A the equation f 0 (z) = w has
infinitely many simple roots. (Specifically, one takes A = {w ∈ C : Θ(w) ≥ 1/2};
cf. [44] p. 280.) The proof then proceeds exactly as before. For the extension of
this argument to meromorphic functions (and beyond), see [7].
Baker’s original proof made essential use of Ahlfors’ Five Islands Theorem, one of
the deepest results in all of function theory. The proof given above, depending only
on elementary reasoning and a standard result of Nevanlinna Theory, thus repre-
sents a significant simplification. A proof that avoids Nevanlinna Theory altogether
is in [3].

7. Marty’s Theorem provides a complete and satisfying answer to the question of

when a family of functions is normal. Unfortunately, in practice it is almost useless,
as verification of the condition (1) in cases when normality is not already evident is
generally extremely difficult. In many cases, the results in §4 provide an effective
substitute for Marty’s Theorem. Long before these results had been obtained,
however, the search for more useful sufficient conditions for normality had given
rise to the following heuristic principle: “A family of holomorphic (meromorphic)
functions which have a property P in common in a domain D is (apt to be) a normal
family in D if P cannot be possessed by nonconstant entire (meromorphic) functions
in the finite plane” [29] p. 250. This principle is often attributed to André Bloch;
however, I have been unable to find any mention of it in his published writings.
In the examples which follow we shall make repeated use of the family of functions
F = {nz : n = 1, 2, 3, . . . }, which fails to be normal on the unit disc ∆. Indeed,
set fn (z) = nz and let K be any compact subset of ∆ which contains the origin
and at least one other point z1 . No subsequence of {fn } can converge uniformly
on K (since fn (z1 ) → ∞), nor can any subsequence of {fn } diverge to infinity (as
fn (0) = 0 for all n).
Examples. 1. Say f has property P on D if |f (z)| ≤ 17 for z ∈ D. By Liouville’s
Theorem, any entire function with P is constant. That a family of (holomorphic)
functions having P on a domain D is normal follows at once from Montel’s Theorem.
2. Say f has P on D if f (z) 6= a, b, c on D, where a, b, c are (distinct) fixed
values in Ĉ. Picard’s Little Theorem asserts that a meromorphic function on C
with this property must be constant, and Montel’s Theorem states that a family of
meromorphic functions having P on a domain D is normal.
3. Say f has P on D if f is univalent on D, and f (z) 6= a, b on D, where a and
b are (distinct) fixed values in Ĉ. We leave the details to the reader.
The heuristic principle has proved itself extremely effective in the identification of
criteria which insure normality. However, it must be used with care. Consider, for
example, the property “f is not entire”. Obviously, no nonconstant entire function
has this property (nor, for that matter, do the constants). On the other hand, it is
clear that the property does not imply normality. Indeed, fix an analytic function g
having the unit circle as a natural boundary and consider the collection of functions
F = {gn } on ∆, where gn (z) = nz +g(z). Clearly, no function in F is entire; equally
clearly, F is not a normal family.
If the (counter)example of the previous paragraph seems frivolous (and I do not
think it is), more serious examples are near at hand. For example, the property
 NORMAL FAMILIES: NEW PERSPECTIVES 223

“f is bounded” forces an entire function to be constant; however, every function in

our paradigm non-normal family {nz} has this property on ∆. A similar comment
applies to the property “f omits three values”. Examples like these point up the
need for a rigorous version of the heuristic principle. In his retiring presidential
address to the Association for Symbolic Logic [54], Abraham Robinson listed this
as one of twelve problems worthy of attention of logicians (and, by extension, of
mathematicians in general).
It turns out to be possible to answer Robinson’s question in a stronger form than
he had expected and without recourse to nonstandard analysis (which, naturally
enough, he had hoped might provide the key to the solution). However, before
turning to this, we need to dispose of the question: what is a property? Answer: a
set (viz., the set of all objects having the property!). As we shall be concerned with
properties of functions on domains in the plane, it will be convenient to display
the domain of the function explicitly together with the function. Thus, following
Robinson, we write hf, Di to denote the function f defined on the domain D ⊂ C;
and we distinguish between the functions hf, Di and hf, D0 i if D 6= D0 . If the
function f has property P on D, we write hf, Di ∈ P. It is now possible to state
the

Theorem. [68] (cf. [69]) Let P be a property of meromorphic (holomorphic) func-

tions which satisfies the following three conditions:

(i) If hf, Di ∈ P and D0 ⊂ D, then hf, D0 i ∈ P.

(ii) If hf, Di ∈ P and ϕ(z) = ρz + b is nonconstant, then hf ◦ ϕ, ϕ−1 (D)i ∈ P.
S
∞
(iii) Let hfn , Dn i ∈ P, where D1 ⊂ D2 ⊂ D3 ⊂ . . . and Dn = C. If fn → f
n=1
χ-uniformly on compact subsets of C, then hf, Ci ∈ P.

Suppose (a) hf, Ci ∈ P only if f is constant.

Then (b) {f : hf, Di ∈ P } is normal on D for each D ⊂ C.
Conversely, if (i) and (ii) hold, then (b) implies (a).

This result is a more or less immediate consequence of the Lemma, with α = 0.

It provides a highly satisfactory explication of the heuristic principle so far as
properties formulated in terms of the values taken on or omitted by functions is
concerned. In such cases, conditions (i) and (ii) will generally be satisfied trivially,
while (iii) follows more or less routinely from Hurwitz’s Theorem.
Actually, much stronger versions of the Theorem hold. For instance, in (a) one
can assume that the function f has order ≤ 2 (or is of exponential type, in case f
is entire). This follows from the fact that the limit function g in the Lemma can
be chosen to be such a function. Similarly, condition (ii) can be replaced [48], [49]
by the more general

(ii0 ) There exists α, −1 < α < 1, such that if hf, Di ∈ P and ϕ(z) = ρz + b is
nonconstant, then hρα (f ◦ ϕ), ϕ−1 (D)i ∈ P.

This condition is particularly well adapted for dealing with properties formulated
in terms of values omitted by derivatives of a function. In the discussion below, we
shall restrict ourselves to applications of the Theorem as stated in its pristine form
above.
224 LAWRENCE ZALCMAN

Examples. (continued)
4. Fix ε (small) and let hf, Di ∈ P if f (z) 6= a, b, c on D, where now a, b, c ∈ Ĉ
are allowed to vary with f but χ(a, b)χ(b, c)χ(c, a) ≥ ε. By Picard’s Little Theorem,
(a) holds; hence we obtain a sharpening of the classical version of Montel’s Theorem
due to Carathéodory (cf. [59] pp. 104-105).
5. Fix a, b, c ∈ Ĉ (distinct) and natural numbers `, m, n such that 1/` + 1/m +
1/n < 1. Let hf, Di ∈ P if all a-points of f in D have multiplicity at least `, all
b-points multiplicity at least m, and all c-points multiplicity at least n. It is an easy
consequence of Nevanlinna’s Second Main Theorem that if hf, Ci ∈ P, then f must
be constant. Thus (a) holds, and we obtain a generalization of a result of Montel
([42] pp. 125-126) due to Drasin ([18] pp. 238-239).
6. Say hf, Di ∈ P if f = g 0 , where g is (analytic and) univalent on D or f ≡ 0
on D. (This last possibility is required if (iii) is to hold, as the limit of univalent
functions could be constant.) Since the only univalent entire functions are linear, it
is clear that hf, Ci ∈ P implies f is constant. Thus the (non-normalized) family of
derivatives of all univalent functions on D ⊂ C is normal on D. By way of contrast,
the family of all univalent functions on a domain is not a normal family (consider
{nz} on ∆), nor is the collection of second derivatives of univalent functions (cf.
Example 9 below).

One attractive aspect of the Theorem is that it explains the failure of the heuristic
principle in those cases where it does not give correct results.

Examples. (continued)
7. Say hf, Di ∈ P if f is bounded on D. Clearly (i), (ii), and (a) hold in this
case; however, as we have seen, (b) does not follow. This is because (iii) does not
obtain. Indeed, fix any nonconstant entire function f and let Dn = {|z| < n}.
Then hf, Dn i ∈ P for each n, but clearly hf, Ci ∈ / P. A similar discussion applies
to the properties “f omits 3 (distinct) values on D” and “f is not entire” (i.e.,
hf, Di ∈ P ⇔ D 6= C).
8. Say hf, Di ∈ P if f is analytic on D and satisfies |f (z)| ≤ |f 0 (z)| and
0 ∈ f (D). Suppose hf, Ci ∈ P. Then |f (z)| ≤ |f 0 (z)| on C, so f /f 0 is constant, and
(log f )0 = f 0 /f is also. Hence f (z) = Keaz and, since 0 ∈ f (D), K = 0. Thus
the only entire function having P is f (z) ≡ 0. That P does not force normality is
evident from the family {nz} on ∆. In this example, P fails to satisfy any of the
conditions (i), (ii), and (iii).
9. Define hf, Di ∈ P if f = g 00 , where g is analytic and univalent on D. The only
entire function with this property is f (z) ≡ 0. Setting gn (z) = n(z + z 2 /10 + z 3 /10),
we have Re gn0 (z) > 0 on ∆, so gn is univalent there. Since fn (z) = gn00 (z) =
n(1/5 + 3z/5) vanishes at z = −1/3 for each n, {fn } does not form a normal family
on ∆. Clearly (i) and (ii) hold, so it must be (iii) that fails. Verifying this is an
amusing exercise, which we leave to the interested reader.
10. Say hf, Di ∈ P if f is analytic on D and f 0 (z) 6= −1, f 0 (z) 6= −2, f 0 (z) 6= f (z)
for z ∈ D or f ≡ 0 on D. Suppose hf, Ci ∈ P ; then f 0 is entire, hence constant (since
f 0 6= −1, −2 on C). Thus f (z) = az + b. But then f (z) − f 0(z) = az + (b − a) 6= 0, so
that a = 0 and f is constant. However, {nz} has P on ∆. Here it is condition (ii)
that fails. It is obvious that (i) holds. To verify (iii), suppose fn → f uniformly
on compacta, where fn0 6= −1, −2 and fn0 − fn 6= 0 on Dn . Then, by Hurwitz’s
 NORMAL FAMILIES: NEW PERSPECTIVES 225

Theorem, f must satisfy

A B
0 0
1. f 6= −1 or f ≡ −1 (i.e., f (z) = −z + b)
0 0
2. f 6= −2 or f ≡ −2 (i.e., f (z) = −2z + c)
3. f − f 6= 0 or f − f ≡ 0 (i.e., f (z) = Kez ).
0 0

1B and 2B contradict 3, and 3B contradicts 1 and 2 unless K = 0. Thus hf, Ci ∈ P.

These last two examples of the failure of the heuristic principle are due to Rubel
[57]. They show clearly that neither (ii) nor (iii) may be dispensed with in the
formulation of the Theorem.

8. In his little problem book [28], Hayman posed a number of open problems on
normal families. Hayman’s questions on normal families are all of similar shape:
in each case, a property involving the values of a function and its derivatives is
known to imply that an entire or globally defined meromorphic function must be
constant. Does the same property imply normality for a family of holomorphic
or meromorphic functions? (Of course, this is just a special case of the “Bloch
Principle” discussed above.2) Over the years, most of these questions have been
answered (affirmatively); let us indicate how the Lemma can be used to provide
simple and uniform solutions to these problems.
We begin with the first, and easiest, of these problems, 5.11 of [28]. Fix n ≥ 1. It
is classical that an entire function which satisfies f 6= 0, f (n) 6= 1 must be constant;
Hayman [27] proved that this remains true for meromorphic functions on C. That a
family of analytic functions on a domain D with this property is normal on D goes
back to Miranda [41]. It was considered a great advance when Gu showed that the
family of meromorphic functions on D such that f 6= 0, f (n) 6= 1 on D is normal
there [33].
The gap of almost half a century that separates Gu’s theorem from Miranda’s
testifies to the formidable technical difficulties involved in extending a result for
holomorphic functions to meromorphic functions. Remarkably, the Lemma does
not distinguish between these cases. Accordingly, when it can be applied (and
when appropriate results for globally defined individual functions are available), it
yields normality results for families of meromorphic functions. Let us illustrate this
point in the current situation.
Denote the family in question by F and suppose that F is not normal. As usual,
one may assume D = ∆. Choose α = −n (as we may since f 6= 0 for f ∈ F ). By the
Lemma, there exist fk ∈ F and zk , ρk such that ρ−n k fk (zk + ρk ζ) = gk (ζ) → g(ζ)
locally χ-uniformly, where g is a nonconstant meromorphic function on C. Since
(n)
gk 6= 0 and g is nonconstant, g 6= 0 by Hurwitz’s Theorem. Also, fk (zk + ρk ζ) =
(n) (n)
gk (ζ) → g (n) (ζ); hence, since fk 6= 1, either g (n) 6= 1 or g (n) ≡ 1. The latter
is impossible, for then g would be a nonconstant polynomial and g 6= 0 would
contradict the Fundamental Theorem of Algebra. Thus g 6= 0, g (n) 6= 1. But then
g is constant by Hayman’s theorem, a contradiction.

2 And, in point of fact, many (though not all) of Hayman’s problems can be settled by invoking

the Theorem of the previous section (in its generalized form) with an appropriate choice of α; cf.
[70]. Here we prefer the slightly more direct (and more general) approach to these problems via
the Lemma.
226 LAWRENCE ZALCMAN

The above method is easily adapted to handle the extension of Gu’s theorem in
which the condition f (n) 6= 1 is replaced by
Xn
h(z) ≡ f (n) (z) + an−j (z)f (n−j) (z) 6= 1,
j=1

where a0 (z), a1 (z), . . . , an−1 (z) are fixed holomorphic functions. For holomorphic
functions, this is due to Chuang and is one of the main results in [18]. The reader
is invited to try his hand at proving this result for meromorphic functions (which
do not vanish) using the Lemma.
Problems 5.12 and 5.13 of [28] deal with the condition f 0 f n 6= 1 where n ≥ 1 is
fixed. This condition is known to force an entire function (Hayman [27] for n ≥ 2,
Clunie [16] for n = 1) or a meromorphic function on C (Hayman [27] for n ≥ 3,
Mues [43] for n = 2) to be constant. The corresponding normality results are due
to Yang and Zhang [66] (for analytic functions, n ≥ 2) and [67] (for meromorphic
functions, n ≥ 5), Gu [32] (for meromorphic functions, n = 3, 4), Oshkin [45] (for
analytic functions, n = 1; cf. Li and Xie [37]), and Pang [48] (for meromorphic
functions, n = 2).
Here is a simple proof, based on the Lemma, that the family F of functions
meromorphic on ∆ which satisfy f 0 f n 6= 1 (n ≥ 2 fixed) is a normal family on ∆.
1
Suppose that F is not normal. Taking α = − , we may choose fk ∈ F , zk ∈ ∆
n+1
and ρk → 0+ such that
1
− n+1
ρk fk (zk + ρk ζ) = gk (ζ) → g(ζ)
locally χ-uniformly, where g is a nonconstant meromorphic function on C. Now
fk0 (zk + ρk ζ)fkn (zk + ρk ζ) = gk0 (ζ)gkn (ζ) → g 0 (ζ)g n (ζ);
since fk0 fkn 6= 1, either g 0 g n 6= 1 or g 0 g n ≡ 1. But if g 0 g n ≡ 1, then g n+1 (ζ) =
(n + 1)ζ + C, an impossibility (since g is single-valued). Thus g 0 g n 6= 1. But then
g must be constant, a contradiction.
All that prevents the above proof from settling the case n = 1 is the absence
of the corresponding theorem for global meromorphic functions. However, we can
actually get by with much less. In fact, Bergweiler and Eremenko [6] have proved
that a function of finite order which satisfies f 0 f 6= 1 must be constant. Since the
limit function g may be taken to have finite order, we see that the above proof
establishes that the condition f 0 f 6= 1 also implies normality.
Once the normality criterion is established, it is easy to conclude that the only
meromorphic functions on C for which f 0 f 6= 1 are constant; this settles problem
1.19 in [28]. Indeed, let f be such a function, and suppose that f is not constant.
Then we can find z0 ∈ C such that f # (z0 ) 6= 0. Let fk (z) = k −1/2 f (z0 + kz)
for z ∈ ∆. Clearly, fk0 (z)fk (z) 6= 1 on ∆, so that F = {fk } is a normal family.
By Marty’s Theorem, there exists C > 0 such that fk# (0) ≤ C for all k. But
√
fk# (0) ≥ kf # (z0 ), which tends to infinity, a contradiction.
The results discussed above concerning the condition f 0 f 6= 1 are due, indepen-
dently and simultaneously, to Bergweiler and Eremenko [6], Chen and Fang [12],
and Zalcman [71].
It seems worth commenting on the “philosophy” of the proof of this result, which
flies in the face of the conventional wisdom concerning proving theorems on entire
and meromorphic functions. It has generally been felt that proofs involving special
 NORMAL FAMILIES: NEW PERSPECTIVES 227

properties of functions of finite order should be avoided whenever possible and,

where order considerations are unavoidable, that the lower order (obtained by re-
placing lim sup in (2) by lim inf) should be used in preference to the order. While
this has often led to proofs of stronger results than would otherwise have been ob-
tained, the results of this section suggest adopting a less doctrinaire stance. Indeed,
what is remarkable about the solution for the case n = 1 is that a result established
initially only for functions of finite order is shown by very general considerations to
be valid for arbitrary meromorphic functions. It is to be expected that this same
approach will prove successful in dealing with a variety of other problems. For a
first step in this direction, see [11], [50].
For complete solutions of all of Hayman’s problems on normal families via the
Lemma, see [72].
9. In this report, we have focused on applications to one-dimensional complex
analysis. It turns out that the ideas involved in the Lemma have much wider
applicability. In this final section we survey briefly three such directions.
a. Quasimeromorphic mappings. The theory of quasiregular and quasimero-
morphic mappings [53] is a generalization to Rn of the theory of holomorphic and
meromorphic functions on the plane, bearing much the same relation to that the-
ory that the theory of quasiconformal mappings in space [63] bears classical con-
formal mapping. Ruth Miniowitz [40] has generalized our Lemma to families of
K-quasimeromorphic mappings on the unit ball B n . As a consequence, she obtains
analogues of Bloch’s Principle, Montel’s Theorem, and the Big Picard Theorem
for quasimeromorphic mappings. It should be mentioned that these last two re-
sults depend on Rickman’s extension of the Little Picard Theorem to quasiregular
functions [52]. It would be interesting to adapt the argument of §3 to provide an
independent proof of this important result, but I have not seen how to do it. Quite
recently, Miniowitz’s result has been used by Eremenko [19] to extend the classical
covering theorem of Bloch ([29] pp. 385-388) to quasimeromorphic mappings on
Rn .
b. Hyperbolicity. Recall that a complex manifold M is hyperbolic if the Kobayashi
pseudodistance dM (p, q) is actually a metric [31], Chapter 4. It is easy to see that if
M contains a complex line (i.e., if there exists a nonconstant holomorphic mapping
g : C → M ), then M is not hyperbolic, since then dM (g(z), g(w)) = 0 for all
z, w ∈ C. The converse fails, as one sees by taking
M = {(z, w) : |z| < 1, |zw| < 1 and |w| < 1 if z = 0}.
(This example is due to D. Pelles (Eisenman) and L. Taylor [31] p. 130.) However,
for compact manifolds one has
Brody’s Theorem. [9] (cf. [34] p. 68) A compact complex manifold is hyperbolic
if and only if it contains no complex lines.
The ideas used in the proof of this result are closely related to the Lemma;
indeed, as pointed out by Wu [64] p. 95, Brody’s Theorem follows at once from the
argument used to prove the Lemma. For further developments in this direction, see
[34] Chapter 3.
c. Minimal surfaces. Let M be a complete minimal surface in R3 . The Gauss
map assigns to each point p ∈ M the unit normal to M at p, regarded as a point in
the unit sphere S 2 ⊂ R3 . In 1959, Osserman [46] showed that the Gauss map of a
complete minimal surface in R3 must be dense in S 2 unless M is a plane (in which
228 LAWRENCE ZALCMAN

case it is a constant), thereby initiating a sequence of developments (described in

[47]; cf. [25]) which culminated in
Fujimoto’s Theorem. [24] The Gauss map of a complete minimal surface M in
R3 omits at most four points unless M is a plane.
This is optimal, as it is not difficult to construct minimal surfaces whose Gauss
maps omit any four (or fewer) prescribed points in S 2 . In a very interesting paper
[55], Antonio Ros has shown how the ideas around the Lemma can be adapted to
deal with minimal surfaces and, in particular, to simplify the proof of Fujimoto’s
Theorem. This appears to be a most promising direction of future research.

10. For a systematic account of the theory of normal families, the reader is directed
to [42], [59], and [15]. The monograph [59], in particular, provides an admirable
introit to the modern theory.

References
1. Gerardo Aladro and Steven G. Krantz, A criterion for normality in Cn , J. Math. Anal. Appl.
161 (1991), 1-8. MR 92j:32004
2. I.N. Baker, Repulsive fixpoints of entire functions, Math. Z. 104 (1968), 252-256. MR 37:1599
3. Detlef Bargmann, Simple proofs of some fundamental properties of the Julia set, preprint.
4. G. Barsegian, Geometrical theory of meromorphic functions, manuscript.
5. Walter Bergweiler, On a theorem of Gol’dberg concerning meromorphic solutions of algebraic
differential equations, Complex Variables Theory Appl. (to appear).
6. Walter Bergweiler and Alexandre Eremenko, On the singularities of the inverse to a meromor-
phic function of finite order, Rev. Mat. Iberoamericana 11 (1995), 355-373. MR 96h:30055
7. Andreas Bolsch, Repulsive periodic points of meromorphic functions, Complex Variables The-
ory Appl. 31 (1996), 75-79. MR 98c:30033
8. Mario Bonk and Alexandre Eremenko, Schlicht regions for entire and meromorphic functions,
preprint.
9. Robert Brody, Compact manifolds and hyperbolicity, Trans. Amer. Math. Soc. 235 (1978),
213-219. MR 57:10010
10. Lennart Carleson and Theodore W. Gamelin, Complex Dynamics, Springer, 1993. MR
94h:30033
11. Chen Huaihui, Yosida functions and Picard values of integral functions and their derivatives,
Bull. Austral. Math. Soc. 54 (1996), 373-381. MR 97k:30035
12. Chen Huaihui and Fang Mingliang, On the value distribution of f n f 0 , Sci. China Ser. A 38
(1995), 789-798. MR 97a:30035
13. Chen Huaihui and Gu Yongxing, An improvement of Marty’s criterion and its applications,
Sci. China Ser. A 36 (1993), 674-681. MR 94j:30031
14. Huai-hui Chen and Xin-hou Hua, Normality criterion and singular directions, Proceedings of
the Conference on Complex Analysis, (Tianjin, 1992), International Press, 1994, pp. 34-40.
MR 96d:30039
15. Chi-Tai Chuang, Normal Families of Meromorphic Functions, World Scientific, 1993. MR
95d:30065
16. J. Clunie, On a result of Hayman, J. London Math. Soc. 42 (1967), 389-392. MR 35:5618
17. J. Clunie and W.K. Hayman, The spherical derivative of integral and meromorphic functions,
Comment. Math. Helv 40 (1966), 117-148. MR 33:282
18. David Drasin, Normal families and the Nevanlinna theory, Acta Math. 122 (1969), 231-263.
MR 40:2835
19. Alexandre Eremenko, Bloch radius, normal families, and quasiregular mappings, preprint.
20. , Normal holomorphic curves from parabolic regions to projective spaces, preprint.
21. Jan-Erik Fornaess, Dynamics in Several Complex Variables, CBMS 87, Amer. Math. Soc.,
Providence, 1996. MR 96j:32033
22. Günter Frank and Wilhelm Schwick, A counterexample to the generalized Bloch principle,
New Zealand J. Math. 23 (1994), 121-123. MR 95m:30044
 NORMAL FAMILIES: NEW PERSPECTIVES 229

23. Günter Frank and Yufei Wang, On the meromorphic solutions of algebraic differential equa-
tions, preprint.
24. Hirotaka Fujimoto, On the number of exceptional values of the Gauss map of minimal sur-
faces, J. Math. Soc. Japan 40 (1988), 237-249. MR 89b:53013
25. , Value Distribution Theory of the Gauss Map of Minimal Surfaces in Rm , Vieweg,
Braunschweig, 1993. MR 95d:32029
26. A.A. Gol’dberg, On single-valued solutions of first-order differential equations, Ukrain. Math.
Zh. 8 (1956), 254-261.
27. W.K. Hayman, Picard values of meromorphic functions and their derivatives, Ann. of Math.
(2) 70 (1959), 9-42. MR 22:1675
28. , Research Problems in Function Theory, Athlone Press, London, 1967. MR 36:359
29. Einar Hille, Analytic Function Theory, Vol. 2, Ginn, Boston, 1962. MR 34:1490
30. A. Hinkkanen, Normal families and Ahlfors’s five islands theorem, New Zealand J. Math. 22
(1993), 39-41. MR 95b:30049
31. Shoshichi Kobayashi, Hyperbolic Manifolds and Holomorphic Mappings, Marcel Dekker, New
York, 1970. MR 43:3503
32. Ku Yung-hsing, Sur les familles normales de fonctions méromorphes, Sci. Sinica 21 (1978),
431-445. MR 80a:30033
33. Ku Yongxing, Un critère de normalité des familles de fonctions méromorphes, Sci. Sinica
Special Issue 1 (1979), 267-274 (Chinese).
34. Serge Lang, Introduction to Complex Hyperbolic Spaces, Springer, 1987. MR 88f:32065
35. Peter Lappan, A criterion for a meromorphic function to be normal, Comment. Math. Helv.
49 (1974), 492-495. MR 52:755
36. , A uniform approach to normal families, Rev. Roumaine Math. Pures Appl. 39 (1994),
691-702. MR 96b:30077
37. Li Song-ying and Xie Hui-chun, On normal families of meromorphic functions, Acta Math.
Sinica 29 (1986), 468-476 (Chinese). MR 88f:30048
38. F. Marty, Recherches sur le répartition des valeurs d’une fonction méromorphe, Ann. Fac.
Sci. Univ. Toulouse (3) 23 (1931), 183-261.
39. David Minda, Yosida functions, Lectures on Complex Analysis (Xian, 1987) (Chi-Tai Chuang,
ed.), World Scientific Pub. Co, Singapore, 1988, pp. 197-213. MR 90d:30097
40. Ruth Miniowitz, Normal families of quasimeromorphic mappings, Proc. Amer. Math. Soc. 84
(1982), 35-43. MR 83c:30026
41. Carlo Miranda, Sur un nouveau critère de normalité pour les familles de fonctions holomor-
phes, Bull. Soc. Math. France 63 (1935), 185-196.
42. Paul Montel, Leçons sur les familles normales des fonctions analytiques et leurs applications,
Gauthier-Villars, Paris, 1927.
43. Erwin Mues, Über ein Problem von Hayman, Math. Z. 164 (1979), 239-259. MR 80d:30026
44. Rolf Nevanlinna, Analytic Functions, Springer, 1970. MR 43:5003
45. I.B. Oshkin, On a test for the normality of families of holomorphic functions, Uspehi Mat.
Nauk 37 (2) (1982), 221-222; Russian Math. Surveys 37 (2) (1982), 237-238. MR 83f:30026
46. Robert Osserman, Proof of a conjecture of Nirenberg, Comm. Pure Appl. Math. 12 (1959),
229-232. MR 21:4436
47. , Minimal surfaces in R3 , in Global Differential Geometry (S.S. Chern, ed.), Math.
Assoc. Amer., 1989, pp. 73-98. MR 90g:53012
48. Pang Xue-cheng, Bloch’s principle and normal criterion, Sci. China Ser. A 32 (1989), 782-791.
MR 91i:30031
49. , On normal criterion of meromorphic functions, Sci. China Ser. A 33 (1990), 521-527.
MR 92b:30041
50. Pang Xue-cheng and Lawrence Zalcman, On theorems of Hayman and Clunie, New Zealand
J. Math. (to appear).
51. , Normal families and shared values, preprint.
52. Seppo Rickman, On the number of omitted values of entire quasiregular mappings, J. Analyse
Math. 37 (1980), 100-117. MR 81m:30030
53. , Quasiregular Mappings, Springer, 1993. MR 95g:30026
54. Abraham Robinson, Metamathematical problems, J. Symbolic Logic 38 (1973), 500-516. MR
49:2240
55. Antonio Ros, The Gauss map of minimal surfaces, preprint.
 230 LAWRENCE ZALCMAN

56. H.L. Royden, A criterion for the normality of a family of meromorphic functions, Ann. Acad.
Sci. Fenn. Ser. A. I 10 (1985), 499-500. MR 86i:30040
57. Lee A. Rubel, Four counterexamples to Bloch’s principle, Proc. Amer. Math. Soc. 98 (1986),
257-260. MR 87i:30064
58. Stanislaw Saks and Antoni Zymund, Analytic Functions, 3rd ed., Elsevier, 1971.
59. Joel L. Schiff, Normal Families, Springer, 1993. MR 94f:30046
60. Wilhelm Schwick, Normality criteria for families of meromorphic functions, J. Analyse Math.
52 (1989), 241-289. MR 90k:30061
61. , On a normality criterion of H.L. Royden, New Zealand J. Math. 23 (1994), 91-92.
MR 95g:30047
62. , Repelling periodic points in the Julia set, Bull. London Math. Soc. 29 (1997), 314-316.
MR 97m:30029
63. Jussi Väisälä, Lectures on n-Dimensional Quasiconformal Mappings, Lecture Notes in Math-
ematics 229, Springer, 1971. MR 56:12260
64. H. Wu, Some theorems on projective hyperbolicity, J. Math. Soc. Japan 33 (1981), 79-104.
MR 82j:53061
65. Xue Guo-fen and Pang Xue-cheng, A criterion for normality of a family of meromorphic
functions, J. East China Norm. Univ. Natur. Sci. Ed. (1988), no. 2, 15-22 (Chinese). MR
90i:30051
66. Yang Le and Chang Kuang-hou, Recherches sur la normalité des familles de fonctions analy-
tiques à des valeurs multiples, I. Un nouveau critère et quelques applications, Sci. Sinica 14
(1965), 1258-1271.
67. , Recherches sur la normalité des familles de fonctions analytiques à des valeurs mul-
tiples, II. Généralisations, Sci. Sinica 15 (1966), 433-453.
68. Lawrence Zalcman, A heuristic principle in complex function theory, Amer. Math. Monthly
82 (1975), 813-817. MR 52:757
69. , Modern perspectives on classical function theory, Rocky Mountain J. Math. 12
(1982), 75-92. MR 83d:30003
70. , Normal families revisited, Complex Analysis and Related Topics (J.J.O.O.
Wiegerinck, ed.), University of Amsterdam, 1993, pp. 149-164.
71. , On some questions of Hayman, unpublished manuscript, 5pp., 1994.
72. , New light on normal families, Proceedings of the Ashkelon Workshop on Complex
Function Theory (May, 1996) (L. Zalcman, ed.), Bar-Ilan Univ., 1997, pp. 237-245. CMP
98:03

Department of Mathematics and Computer Science, Bar-Ilan University, Ramat-Gan

52900, Israel
E-mail address: [email protected]

View publication stats