on direct critical singularities and regularity of growth
TRANSCRIPT
ON DIRECT CRITICAL SINGULARITIES
AND REGULARITY OF GROWTH
By
W. K. HAYMAN London, England
Introduct ion. It follows from a classical theorem of Ahlfors [1], that if
f(z) is an integral function having k distinct asymptotic values then
(1) ct = lim inf T(r,f.____~) > O, rk/2
r -~ oo
where T(r , f ) is the Nevanlinna characteristic* of f ( z ) . Later Heins [4, 5]
investigated the case when e < + 0% and proved that in this case
T ( r , f ) �9 (2) rX/-'---T- -
if k = 1, and that f ( z ) has exact order �89 when k > 1. The latter result was
later improved by Kennedy, [8, 9] who showed that
(3) log T(r , f ) = ~-log r + o(log r) ~',
and that this result is best possible. The theorems of Heins and Kennedy were
actually proved for subharmonic functions which become unbounded in k
disjoint domains in the open plane.
Another type of generalisation was obtained by Ahlfors [2] somewhat
earlier. According to examples of Valiron [11] a meromorphic function of
* For the definitions and results of Nevanlinna Theory, see e.g. [10].
113
114 w. K. HAYMAN
finite order can have infinitely many asymptotic values so that a direct exten-
sion of (1) is not possible. However Ahlfors [2] proved that a meromorphic
function f ( z ) whose inverse function has k distinct direct critical singularities
must satisfy (1). According to Iversen [7] the inverse function z = ~h(w) has
a direct critical singularity at a, if a branch of ~(w) can be indefinitely
analytically continued without algebraic or other singularities in some neigh-
bourhood: 0 < [ w - a l < 5, if a is finite or R < [w[ < 0% if a is infinite,
but ~(w) cannot be continued to the point a.
The assumption that Ahlfors actually used, somewhat weaker than that
of direct critical singularity, was the following. There exist k non-overlapping
simply-connected domains A v in the open plane, such that each A v is mapped
by f ( z ) into a neighbourhood Nv of a point a,, while f ( z ) ~ a, in A~ and f ( z )
is bounded away from av on the boundary of A~. Since two distinct asymptotic
paths of an integral function f ( z ) are separated by a path on which f ( z ) ~ 0%
this condition is satisfied by an integral function with k distinct asymptotic
values, if we set a~ = ~ , v = 1 to k and take for N~ a disk f w I > R, where R
is sufficiently large. These assumptions were later weakened further by Gold-
berg [3] and Heins [6], without destroying the conclusion (1).
It is natural to ask whether under these more general hypotheses the as-
sumption that a is finite in (1) leads to restrictions on the growth or 'regularity
theorems' of the type (2) or (3) and this question was in fact recently raised
to the author in an oral communication by Professor M. H. Heins. We shall
see that the answer is negative. We have in fact the following
Theorem. It is possible to choose the increasing sequences of positive
integers Pn and positive real numbers r., such that if k is a positive integer and
(4) = c o s + + r . ) - * " + ( ,- . - ], n = l
then f (z k/2) has k direct critical singularities at infinity and satisfies (1) with
a finite value of a, while at the same time m(r,f), N(r , f ) and T(r , f ) all
exceed an arbitrary preassigned positive function ~(r), for a sequence of r
tending to infinity.
ON DIRECT CRITICAL SINGULARITIES 115
2. P r o o f o f t h e T h e o r e m .
We proceed to prove our theorem. We need the following
L e m m a 1. Suppose that for n = 1,2, ... p~ is a strictly increasing
sequence of positive integers and that the r, are positive numbers, such
that r 1 = 1,
(5)
Then i f
r.+x > r . + 4, n = 1,2 ...
g(z) = ~ (z + r.) -p", n = l
we have
(6) I g(z) l < 1,
i f I z + 1".1 > 2, for every n, and
(7) Ig(~)l > [z + r . [ -~" -
I g'(~) I < 1,
1, i I l=+r. l<2.
Since p. is a strictly increasing sequence o f positive integers p . > n. Thus
if I z + r . I > 2, we have
I z + r . I -p < 2 -~, p , , I z + r . l -{p.+*' <p.2-{P-+*) < n2-{"+l ' ,
since n2 -" decreases with increasing n. Thus i f I z + r.I > 2 for every n,
we have
Ig(z)l < :~ 2 -"= 1, Ig'(z)l < ~ n2-'"+x)= 1. 1 1
The inequality [ z + r . I < 2 can hold for at most one n, in view of (5). I f it
does hold we have
116 W. K. HAYMAN
Ig(z)l [z+r.l-, - Z 2 -m, m # n m = l
which gives (7).
L e m m a 2. I f p. , r,, satisfy the hypotheses of Lemma 1, and f ( z ) is
defined by (4) then f ( z ~lz) is meromorphic in the plane and has k direct
critical singularities at ~ .
Suppose that z = x + iy, with I Y I > 2, then
(8) lf(z)[=lcosz + g(z)+ g(-z)]>=lcosz[-2 >= e I*l - e - l y l 2
- 2 > ; - 2 1 > 1 .
Thus f(z) has no zeros for l yl__> 2, and
q~(z) = logf(z)
is well defined in each of these halfplanes. Further
- - f ' (z) i e i ~ - i e - i ~ + 2 g ' ( z ) - 2 g ' ( - z ) f ( z ) e i~ + e-iz + 2g(z) + 2g( - z)
so that in view of (6) we have if y => 3
I [' q~'(z) + i'[ = 2ie'~ + 2g'(z) - 2 g ' ( - z) + 2ig(z) + 2ig( - z)
e--ZTz + < 2e-Y + 8
ey -- e - y - 4 < 1.
Thus ~b'(z) has negative imaginary part in the half-plane y > 3 and so ~b(z)
is univalent there. Again ~ = ~b(z) maps the line y = 3 onto a curve F in the
plane and if ~ = ~ + iq, then we have on F
l < e r < l c o s z 1 + 2 < e a + e -3 ~ - + 2 < e a,
so that 0 < ~ < 3. Thus qS(z) assumes equally often in y > 3 all values ( such
that ~ > 3, and since
l f (x + iy)l ~ oo as y ~ oo
ON DIRECT CRITICAL SINGULARITIES 117
uniformly in x, these values are taken exactly once. Hence ~b(z) maps a sub-
domain D + of the halfplane y > 3 (1,1) conformally onto the halfplane ~ > 3,
and so
w = f ( z ) = e ~(~
maps D + (1,1) conformally onto the infinite covering surface R, over the
annulus e3 < [ w [ < oo. Since f ( z ) is even it follows that f ( z ) also maps the
domain D - , which is obtained by rotating D § through an angle rc about the
origin, onto R. Thus f ( z ) has two direct critical singularities.
The function g ( Z ) = f ( Z ~) is clearly meromorphic in the plane, since f ( z )
is even. Also g(Z) maps D onto R, where D is the image of D + under the map
Z = z 2, which is univalent in D § Thus g(z) has one direct critical singu-
larity at oo. If k is a positive integer f ( z k/z) = g(z k) maps onto R each of
the k domains which correspond to D under the transformation Z = z k. Thus
f ( z k/2) has k direct critical singularities over infinity as required. This proves
Lemma 2.
It remains to prove that f ( z k/2) satisfies the growth conditions of our
theorem. To do this it is clearly sufficient to consider the case k = 2, since
m (r,f(z~12)} = m (r t'12 , f (z)},
N ( r , f ( ~ / 2 )} = N (r k/z, f(z)},
T (r,f(z kl2)} = T (r kiz, f(z)}.
Suppose then that [ z l = r, where rv + 2 < r <r~+, - 2. Then I z-T- r z l > 2,
for every ~ and so we have in view of (4) and (6)
1 e,: I le,, Is :)l < Icos:l + 2 =< -~-{1 + I) + 2,
(8) m ( r , f ) ~ 2re(r, e ") + 2 log 2 =--2r + 2 log 2.
Also if Pv = )_2 Pu, we have
118 w . K . HAYMAN
n(r,f) = 2P~,
f n(t,f)dt < 2P~log r. N(r,f) = t 1
Given P, , r v we chose rv+ 1 so large that
(9) 2Pv log (r~+ 1 - 2) < (r~+ 1 - 2) ~.
With such a choice o f the quantities p~, r,, we have for r = r '+x = r~+l - 2,
r t ~ (10) N(r,+ l , f ) < r, + 1,
and in view o f this and (8) we have (1) with k = 2 and e < 2/re.
On the other hand suppose that r = r;' = r , + �89 Then
( 1 1 ) " = N(r~,f) >
r'tv
f n(t,f)dt t
r v
- - > p~ l o g - ~ = p, log 1 + .
~p iO Also if z = rye , with t 01 < 1/(4r~) then
1 r,[ 0 3 [ z - r~[<[ ( r ; ' - rOe i~176 [ < ~---.
Thus in view o f (7) we have
I g ( - z)l > - 1, [g(z)l < 1.
Hence
Hence
( 4 r ~ ) - 1
m[r~, g(z) + g( - z)] =>
- ( 4 r ~ ) - 1
[ 4 ] > (4~zr,) -1 p ~ l o g ~ - - - 21og2 .
(12) m(r",f)
ON DIRECT CRITICAL SINGULARITIES
> m[r",g(z) + g ( - z)] - m(r",cosz) - log2
>= --lP~A _ A 2 r ~ _ A3, rv
119
where A 1, a2, A 3 are absolute constants.
Suppose now that r is an arbitrary positive function of r and that
rl,pl,r2,P2,...pv_l,r~ have been assigned to satisfy (5). We then chose Pv
so large that
Alp~
P~l~ ( l+2--~f) > r (r~ + 1 ) ,
- - - A2rv- A3 > r (r~ + 1 )
and pv > P~-l. Thus in view of (11) and (12) we have with r~= r~ + �89
/ ! tv m(r~,f) > r N(r , , f ) > r ,
and so
t ! t t r(rv,f) > 2 r
Thus m(r,f), N(r,f) and T(r,f) grow faster than the preassigned function
r on the sequence r = r'v'.
On the other hand having chosen Pv, we now choose r,+l to satisfy (9)
so that in view of (8) and (10) we have (1) with k -- 2 and a __< 2/re. This proves
our theorem.
BIBLIOGRAPHY.
1. L. V. Ahlfors, Untersuchungen zur Theorie der konformen Abbildungen und der ganzen Funktionen, Acta Soc. ScL Fenn. nova Set. 1 (1930), No. 9.
2. L. V. Ahlfors, Ober die asymptotischen Werte der meromorphen Funktionen endli- cher Ordnung, Acta Acad. Aboensis Math. et phys. 6 (1932), No. 9.
3. A. A. Gol'dberg, On the influence of algebraic branch points of a Riemann surface on the order of growth of a meromorphic mapping function, Dokl. Akad. Nauk. S S R (N.S.) 98 (1954), 709-711, correction 101 (1955) 4. (Russian).
120 W.K. I-IAYMAN
4. M. H. Heins, Entire functions with bounded minimum modulus; subharmonic function analogues, Ann. of Math. 49 (1948), 200-213.
5. M. H. Heins, On the Denjoy-Carleman-Ahlfors theorem, ibid. 533-537. 6. M. H. Heins, Asymptotic spots of entire and meromorphic functions, Ann. of Math.
66 (1957), 430--439. 7. F. Iversen, Recherches sur les fonctions inverses des fonctions m6romorphes,
(Th~se, Helsingfors 1914). 8. P. B. Kennedy, On a conjecture of Heins, Proc. London Math. Soc. (3) V (1955),
22-47. 9. P. B. Kennedy, A class of integral functions bounded on certain curves, Proc.
London Math. Soc. (3) VI (1956), 518-547. I0. R. Nevanlinna, Eindeutige analytische Funktionen, Berlin (1936). 11. T. Valiron, Sur les valeurs asymptotiques de quelques fonctions m6romorphes,
Rendiconti del Circ. Math. di Palermo 49 (1925), 415-421.
IMPERIAL COLLEGE, UNIVERSITY OF LONDON, LONDON, ENGLAND
(Received February 28, 1966)