on nevanlinna’s second theorem and extensions

47
346 ON NEVANLINNA'S SECOND THEOREM AND by W. K. H A Y M AN (Exeter, England) EXTENSIONS INTRODUCTION. 1) Suppose that f(z) is meromorphic in [z] ~R < oo. We denote by n (5 f) the number of poles of f (z) in J z [ < r counted with correct multiplicity, and by n (r, f) the corresponding number with multiple poles counted only once. We also write -{ dt (1.1) m (5 f) = n (t,f) t ' We use (1.1) instead of the more usual (i) N (r, f) = In (t,i) -- n (o, f)] t + (r,/) = n (t, D c1 t. t n (o, f) log r, to avoid possible infinities if f (z) has a pole very near the origin. The radii of all circles occurring will be bounded, so that our results will not be affected. We also define as usual and 1 (2~ ,,, (~, i) = W~; .to log+ I i (r/~ a 0 (1.2) T (r, f) = m (r, f ) -+- N (r, f). We shall be interested in the behaviour of f (z) in circles with variable cen- ires in the second section. We accordingly define (1.3) m (zo, r, f (z)) = m (r,f (z o --~ z))

Upload: w-k-hayman

Post on 14-Aug-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: On Nevanlinna’s second theorem and extensions

346

ON N E V A N L I N N A ' S SECOND THEOREM AND

by W. K. H A Y M A N (Exeter, England)

EXTENSIONS

INTRODUCTION.

1) Suppose that f ( z ) is meromorphic in [z] ~ R < oo. We denote by

n (5 f ) the number of poles of f (z) in J z [ < r counted with correct multiplicity,

and by n (r, f ) the corresponding number with multiple poles counted only once. We also write

-{ d t (1.1) m (5 f ) = n ( t , f ) t '

We use (1.1) instead of the more usual (i)

N (r, f ) = In (t , i) - - n (o, f)] t +

( r , / ) = n (t, D c1 t. t

n (o, f) log r,

to avoid possible infinities if f (z) has a pole very near the origin. The radii of

all circles occurring will be bounded, so that our results will not be affected.

We also define as usual

and

1 (2~ ,,, (~, i ) = W~; .to log+ I i (r /~ a 0

(1.2) T (r, f ) = m (r, f ) -+- N (r, f).

We shall be interested in the behaviour of f (z) in circles with variable cen-

ires in the second section. We accordingly define

(1.3) m (z o, r, f (z)) = m (r,f (z o --~ z))

Page 2: On Nevanlinna’s second theorem and extensions

ON NEYANLINNA'S SECOND THEOREM AND EXTENSIONS 347

with similar definitions for N, N, T etc. The expressions 1 / ( f - - a ) are well

defined when a is finite. When a is co we define them formally by putting f

instead of 1 / ( f - - co). Thus

(1.4) m (r, 1 / ( f - c~)) - - m (r, f) .

W e shall assume throughout, that the circle [ z - - Zo I < r moves in a fixed

boundel domain Do in which f (z) is meromorphic and that r remains greater than a

positive constant, so that zo lies well inside Do. By O (1) w e denote any term

which remains bounded as Zo, r vary subject to such conditions. The first and

second fundamental Theorems of Nevanlinna may now be written as follows

Theorem A(~). For any fixed finite a we have

(1.5) T (Zo, r, 1 / ( f - - a ) ) --- T (Zo, r, f ) - ~ - 0 (1).

Theorem B~3;. If at, a2,..., aq are q >~ 3 distinct complex numbers, (possibly infinite), then

(1.6) (q - - 2) T(zo, r, f ) ~ ~ N Zo, r , - - - -~-S (Zo, r, f) , V~I f ~ a v

where

(1.7) S (zo, r , f ) = m (zo, r, ~ ) -[- m (Zo, r, ~ ) -~-O (1),

and �9 (z) ---- II ( f (z) - - av), the product being taken over those v for which

a~ is finite. If now instead of [ z -- Zo I <~ r we take a sequence of circles] z - z. [ < r . ,

where r. increases in such a way that

T (z., r,, f ) -~ co,

then Theorem A shows that for every a

(1.8) m (z . , r., 1 ~ ( f - - a ) ) q_ N (z. , r., l / ( f - - a)) -~ 1. T (z., r., f ) T (z., r., f )

We also write following Nevanlinna(4)

- - N (Z,~, r,. 1 / ( f - a)) (1.9) 0 (a) - - 1 - . lim

T (z., r . , f )

Page 3: On Nevanlinna’s second theorem and extensions

348 w. r H A Y M A N

Then Theorem B together with (1.8), (1.9) gives the defect relation

q

(1.1o) o (a) _< 2,

provided that

(1.11) s (z,, r~,r 0. T (z,,, r~,f)

3) We shall obtain in this paper some new conditions under which (1.11)

and hence (1.10) holds. Our results will be based on a new estimate for

m(Zo, r,f--'f),(Theorem l), which allows t h e c i r c l e l z - - Z o [ < r t o m o v e r i g h t

up to the boundary of the domain of Do in which f ( z ) is meromorphic in certain cases. From this we obtain for instance that the defect relation holds for functions of unbounded characteristic in I z t < 1, which are meromorphic and of finite order in a larger domain, bounded by a finite number of analytic arcs which lie except possibly for their endpoints in [ z l > 1, (Theorem II). In particular we obtain a Theorem of Picard type for such functions.

By mapping the unit circle onto an angle we obtain

Theorem IlL Suppose that f (z) is meromorphic of finite order in the plane.

Let 1 ~_ P ~ oo and let z~ (a) - - r~ e i o, be the roots of the equation f (z) - - a, 2

lying in the angle I arg z [ ~ --.~z Then either 2p

(0 f (z) has bounded characteristic in I arg z [ < ~ In which case 2p

E COS p 0,, (2.1) r" p

converges for every a; or

(i i) f (z) has unbounded characteristic in [ arg z I < ~ in which case the 2p'

series (2.1) diverges for every a with at most 2 exceptions. In particular we see that if (2.1) diverges for a single value of a then f ( z )

takes every value with at most two exceptions infinitely often in the angle.

Page 4: On Nevanlinna’s second theorem and extensions

ON N E V A N L I N N A ' S SECOND THEOIIE/~I AND E X T E N S I O N S 349

7~

This result is not true for functions meromorphic merely in ] arg z F < - - , although 2,0

the condition that f ( z ) must be a meromorphic function of finite order could be relaxed considerably, it is also shown by an example that for any sequence z~ for which (2.1) converges and

E 1

for a finite k there exists a meromorphic function of finite order having f (z~) - - 0

and I f ( z ) I < 1 i n l a r g z [ < - - at least when p is an integer. 2p '

We next discuss the problem raised by BIoch (5) concerning the characteristic

of f (z) and f" (z). Some simple examples due to Littlewood (6) show that f (z ) may be bounded in ['z [ < 1 while f ' (z) has unbounded characteristic there. On

the other hand if f (z) is meromorphic of finite order in a domain Do satisfying the conditions of Theorem II, and has bounded characteristic in [ z [ < 1, then

all the derivatives of f ( z ) have bounded characteristic in I z i < 1 (Theorem IV).

Hence under the hypotheses of Theorem Ill, if (i) holds for f (z), then (i) holds

for all the derivatives of f (z) (Theorem V). The converse problem is left open.

3) In the second section of this paper we consider a function f ( z ) mero-

morphic of finite order in the unit circle. It is shown that if the circles

I z - - z ~ l <r,~ exhaust ] z [ < 1 properly i .e . so that

T (z,, r,, f ) (3.1) log (1/(1 - - r,~)) "~ oo,

then the defect relation (1.9) holds tl 'heorem VII). Two generalisations of Theorem

B due to Nevanlinna(:) and Milloux(s) are also discussed. If there exists a sequence of circles properly exhausting I z [ < 1, we say that the Riemann surface o f f ( z ) is properly exhaustible. This is shown to be the case (Theorem VI) whenever

(3.2) (1 - - r ) log M (r, f ) -~ c~, as r-~ 1,

where

(3.3) M (r , f ) : sup I f (z ) r. t z l = r

On the other hand to imply (3.1) with z , ~ : 0, the classical condition of 23 - R e n d , C i t e , M a t e m , P a l e r m o , - - s e r i e I I - t o m o I [ - a n n o I953

Page 5: On Nevanlinna’s second theorem and extensions

3 5 0 "W. IK, I t A Y M A N

Nevanlinna, we need the somewhat stronger condition

( 1 - - r) log 34 (r,f) ->- oo log (1/1 - - r)

on the maximum modulus. (3.2) cannot

f (z) -~- e 0+zV0-z) shows, for which I f (z)

Ahlfors(9) in a now classical paper

exhaustion from that of Nevanlinna. Let D

Riemann sphere. A region d contained with

be further weakened as the function

1 > 1 i n l z l < l . has developed a different theory of

be a simply connected domain on the

its frontier in I z - - Zo [ < r corresponds

to an island over D by f (z), if f (z) maps d p : 1 conformally onto D. We denote by

n (zo, r, D) the total number of such islands each counted with correct multiplicity p and

by n (zo, r, D) the corresponding number with each island counted once only.

Let DI, D~,..., Dq be nonoverlapping simply-connected domains on the Riemann

sphere. Then Ahlfors has proved the inequalities

Theorem A" Qo)

(3.4) n (Zo, r, Dr) < A (Zo, r) + C L (Zo, r);

Theorem B" (it)

(3.5) q

(q - 2) A (Zo, r) < ~ ~ (Zo, r, o , ) + c ' L (Zo, r);

analogous to Theorems A and B. Here ~ A (zo, r), L (Zo, r) are respectively the area and

length of the frontier of the image by f ( z ) of T z - - zr ! < r on the Riemann

sphere. The constants C, C' depend on the Dv only.

If a sequence of circlesJ z - - z , ] < r, exists exhaust ingJ z [ < 1 in such

a way that

(3.6) L (z., r.) *. O, A (zo, r.)

then the Riemann surface of f (z) is said to be regularly exhaustible (regular aussch6pJbar) by Ahlfors. in this case we can obtain the Ahlfors Defect Relation (t~)

q

(3.7) ~ o CO,) -< 2,

Page 6: On Nevanlinna’s second theorem and extensions

ON NEVANi,INNA'S SECOND THEOREM AND EXTENSIONS 3 5 ]

where

n-(z,, r., D,.) (3.8) O (O,) - - 1 - - lira

. * - - A (z . ,r . )

Suppose in particular that each island over Dv has multiplicity at least i~v.

Then (3.4), (3.0)and (3 .8)show that t-, (D,)_>_ (I P',I ) s o that (3.7)gives the

Scheibensatz (t8)

q( ,) (3.9) 1 - W -< 2.

We shall show (Theorem VIII) that (3.0) holds for a sequence of circles

containing I z - z~ , [< r~ if 1

3.10) [(1 - - r;) 2 - - [ z;[~l --" A (z; , r;)--~ co.

The condition reduces to the one obtained by Ahlfors(14) if z , - - -0 . It is always

satisfied if the Riemann surface of f (z) is properly exhaustible in the sense of

3.1), so that the Scheibensatz holds in particular if (3.2) holds.

Our method also enables us to obtain good bounds for M (r, f ) in terms

, particularly when these tend to infinity slowly of the quantities N r, f - - a

as r-~-1. Thus Theorems Vl and VII imply for instance that if (3.2) holds, the

equations f ( z ) - - a have so many roots that

n r, d r - - - + c ~

except perhaps for 2 values of a. By a refinement of Ahlfors method(15) one

could even generalise the points a to domains D, using the Ahlfors notation.

Without going into this latter generalisation we shall prove the following rather precise theorem as an example of the method.

Theorem IX. Suppose that f (z) is meromorphic in I z ] < 1 and that p (p)

is the total number o f roots of the equations f (z) ~- av, v =- 1 to q, where the a,

Page 7: On Nevanlinna’s second theorem and extensions

3 5 2 w . x . H A ~ M A ~

are q ~ 3 distinct complex numbers one of which may be infinite. Then i f

(3.11) lira (1 - -r ) p (r) ~ b ,~ o%

we have

lira (1 - - r) log M (r, f) ~ 2 9,, ( 3 . 1 2 ) u -,~. I

log (1/1 - r )

where ~ : b/(q -- 1) or b/(q -- 2) according as the a~ are all finite or not. I f

f (z) is regular, lira may be replaced by lira in (3.12). Equality is possible in (3.12)

for every set of numbers a i to a~ and b < oo, for a regular function f ~z).

Using concentric circles (3.12) could only be proved with ~. -~ A replacing

~, where A is an absolute constant.(i6)

As a final application of our method we solve the following problem raised

by Littlewood in a group of research problems. Suppose that f(z), �9 (z) are

regular in I z ] < I and f (z), f (z) - �9 (z) have no zeros. Does

o [1) (3.13) log M (r, r - -

1 - - r always imply

(3.14) log MCr, f ) = 0 (1) ? l - - r

An affirmative answer was given in (4) Theorem V, which applied also to

meromorphic f (z) and �9 (z) but contained the assumption that

jl n (0, t, 1/O) d t < oo.

We can now eliminate this assumption and prove

Theorem X. Suppose c~ (z) ~ O, is regular of f inite order in I z i < 1 and

that f ( z ) is regular, and f ( z ) , f (z) - ~P (z) have no zeros there. Then we have

for 0 < r < p < 1,

log M (r, f ) < log M (p, O) -b

where K depends on f (z), ~P (z) but not on ~, r.

K p - - r '

Page 8: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA S SECOND THEOREM AND EXTENSIONS 353

1 it is now clear that (3.13) implies (3.14) taking p ~ ~2- (l + r). Our method

would also enable us to prove a modified result if poles of f(z) and O(z)

and zeros of f (z), f (z) -- �9 (z) are allowed provided that their total number n (t) in I z l < t satisfies

i n (t) d t < oo.

For simplicity we restrict ourselves to the case quoted above.

S E C T I O N 1. 4) We start by proving the following fundamental result

Theorem L Suppose that f (z) is meromorphic in a bounded domain D containing

[ z I ~ R. Let dR (0) denote the distance of z ~ Re '~ from the boundary olD, and

nR (0) the total number of roots of the equations f (z) ~ 0,1, oo distant at least

1 - - d R (0)from the boundary of D. Then we have 2

( ) l/? ,o, , m R, ]'(')f (z)(Z) ._~ A (p) log+ f (Re '~ ) d 0 -l- I -l-- log+ -~ -1- 1 )

(4.1)

where

(4.2)

[ (7) ' ] ~_.A(p) log+ m (R , f ) q-- log+ m R, 1 -q--l--l-log+ ~ + 1

I - - I ( R ) 2 = J o log+ nn (o) + l o g + an(B) d0.

Here f(~> (z) is the p th derivative of f (z), and A (p) depends only on p. We need a series of lemmas.

L e m m a l . Suppose f (z) - - a o -q- a~ z -~- . . . is regular and f (z) ~ 0 or 1

in l z l < d; then we have

l I iJa. (4.3) <A(n) 1 + log Oo ao gP i

Write

(4.4) g (z) - - log f (Z) ~-- s b. Z". o

Page 9: On Nevanlinna’s second theorem and extensions

354 w. r . HAvr~Ar~

We may without loss in generality take d : 2. Then i f l f ( 0 ) l . < 1 we have

by Schottky's Theorem I f ( z ) I < A in I z [ < 1, and if I f ( 0 ) l ~ 1 we have

I [ f (z)]-~ J < A in I z ] < 1. Thus we have either

o r

Rl g (z) < A

- - B l g (z ) < A

in I z l < 1. In either case we deduce(t7)

(4.5) ]b.l~2(IRtbol--I-A)---2 [lioglaol]+A]~A(lloglao[l-q-1),

We have next from (4.4) f ' (z) - - g ' (z) f (z) or

/ / a n Z ' ~ - i - - " ;7 bn z n - i a n z " .

1

Equating coefficients of z ~-i we deduce the recurrence relation

p--1

p a , : ~ ( p - - r) b,_~ a,, r ~ o

and using (4.5) we deduce

p - - I

}a, I<_A ( I log I ao ]l-q- 1) ~-" l ar l . ~ o

Now (4.3) follows by induction on p.

We have next

L e m m a 2. Suppose f (z) is meromorphic in [ z ] ~ t? and let 8n (0) be the radius

of the largest circle centre R e '~ in which f (z) is regular and unequal to 0 or 1.(t8) Then

m(.1"''z'~<_ t f~ i �9

Page 10: On Nevanlinna’s second theorem and extensions

ON NEVANI.INNA'S SECOND THEOREM AND EXTENSIONS 3 5 5

By applying lemma 1 to f (Re ~o + z) instead of f (z) we obtain

log+ f/~) (Re,0) < log+ A (p) 1 + log f ( R e '~ )

1 + log+ log <_A(p)+p log+ ~(0)

.--J--1 +1].

Lemma 2 now follows by integration�9

We need to estimate further the

quote the results.

two integrals occurring in lemma 2. We

L e m m a 3.(1"). We have

1 tn d ~ .< log+ m ( t ? , y O + l o g + m ( R , f ) + A.

L e m m a 4(~~ With the notation of Theorem ! we have

log+ 8\ (0-3 �9 O

( 1 ) aO<_A /+ tou+ ~ + t .

Theorem 1 now follows from lemmas 2, 3, 4.

5) Theorem I can be extended to the case when f (z) is not meromorphic

at every point of [ z ] ---R, provided only that the domain D contains almost

all points of I z ! - - R and the behaviour of 3' (z) in D is such that the integral

I (R) converges�9 We proceed next to obtain a general condition for this to occur.

We shall say that D properly contains a set of arcs z --- Re ~~ ~ < 0 < ~3,

i f these arcs lie in D and we have uniformly on all the arcs

(5.1)

where C~, C2 are positive constants. In effect this means that on each arc dn (0)

is not much smaller than the distance of z : t?e'O from the nearest endpoint

of the arc.

Page 11: On Nevanlinna’s second theorem and extensions

356 w. t~. H^VMAN

We shall also say that f ( z ) has finite order in D if for each complex a the

number n (d, a) of roots of the equation f (z) -- a which are distant at least d from

the boundary of D satisfies

(5.2) n (d, a) = O (d -c.~)

as d-t~ 0, where C.,~ is independent of d but may depend on a. This agrees

with the usual definition if D is a circle (,"t). With these definitions we can prove

Theorem I1. Suppose D is a bounded domain containing I z I ~ t? and pro-

perly containinff a set of arcs z = Re ~~ % < 0 < ~ , where

(5.3) S, ( ~ , - %) = 2 ~,

1 (5.4) ~ ('~- %)log ( L - ~,) < ~

Suppose further that f (z) is meromorphic of finite order in D. Then i.l S (r, f )

denotes the term in (1.7) with Z o - - 0 we have

S (r , f ) - - 0 flog T(r , f ) ] q- 0 (1),

as r-~ 1? from below. In particular if f ( z ) has unbounded characteristic in

z I < R, the defect relation holds as r -~ R from below.

We need two more lemmas

L e m m a 5. Under the hypotheses of Theorem 1I, let 0 ~ r < R and let

d, (0) denote the distance of z - - r e'O from the nearest frontier point of D. Then

we have uniformly in v and r

"t~v l~ ld, (0----) ( I ) , / d o - - O ( ~ - - %) log+ - + 1 .

We shall denote by C~, C:, .... constants independent of v and r. Let z ' = R'e'0'

be the nearest point to z -~ re ~o on the frontier of D. Suppose that ~v < 0 .< ~i v

and that

1 m i n l O - - % , ~ , , - - o (5.5) 10 - 0'~ < ~- ~.

Page 12: On Nevanlinna’s second theorem and extensions

O N N E V A N L I N N A ' S S E C O N D T H E O R E M A N D E X T E N S I O N S 357

Then i t follows that

1 (0 - - ~v ) ( , 8 v - - 0) (o' - % ) ( ~ , - o') > - g

so that (5.1) gives

I R ' e ~~ - R e ~ O ' l = R " - -

Thus in this case, since

we have

(5.0)

d, tO) - - j z ' - z [ > _ R ' - r>_R'--R,

d~ (0) > C4 [(~,, ..... O) (0 - - ~,,)]c.,.

Suppose next that (5.5) is false. Then

z" - - z I - - t R " ei~ - - r e io I - - { R " ei(~176 - r l ~ R ' I sin (~' - - O) i --~> ~ [ O' - 0 I,

i f [ 0 - - 0 ' < - - . 7: Also i f - - = ~ 1 0 - - 0 ' [ ~ 7 : we have 2 2

[ z - z [ - - I R ' - - r e ~ c o - ~ ' ) [ _ > R ' - r c o s (0 - - 0') > R' _> R.

Thus in any ease

I z" - z t >- ~-to" - >- R_ rain (o - ~,, ;~, - (~) 2 n

t' (0 - =,)(}, -~)) . 4r~.

Thus in this case also (5.6) holds, possibly with different (74, C5. Thus (5.6) holds generally on all the arcs % < 0 < i3v. We have on integration

~v log+ l ~ d 0 < C:, log+ . . . . . 4- log+ �9 =~ d , ( 0 ) �9 = , L 0 - - % "r . . . . . . . .

dO

] . , a' V

= (}, - ~) [2 c~ log 1 ] ( ~ , - ~ , ) + 2 c ~ + c , .

Page 13: On Nevanlinna’s second theorem and extensions

3 5 8 W . K . H A Y M A N

This proves lemma 5. We have next

L e m m a 6, Suppose that D satisfies the conditions o[ Theorem II. Then

'/?[ '} l ( r ) : log+ n, (0) q-- log+ dO .2 .. a , io)

is uniformly bounaed for 0 .:_~,~ r < R. By condition (5.3) we have

Now

equations f (z) - - O, 1,

boundary of D we have

log 4- n, (0) = O ! log + , d , (0)

uniformly in r and e. Thus we obtain, using iemma 5,

i (r) : 2-~ : .. ~ d, (0)

since f ( z ) is of finite order and n, (0)is the number of roots of the

co within a distance not more than --1 d, (0) from the 2

+ 1!.

l ( r ) = O 1 -F log + d 0 ~V de

I ' )} = o ( ~ , - ~ , ) log , 1 = o ( t ) ,= , (~, - ~,v)

uniformly in 0 ~ . .-~ ,9 by (5.4). This proves lemma 6. We can now complete the proof of Theorem il. We have to show thai if

S (0, r , f ) is defined as in (1.7) then

(5.7) S (0, r, f ) - O 1 log T (r, f ) } --I- O t 1), as r -~ R.

To do this note that by Theorem 1 and lemma 6 we have as r-~./9

(5.8) m

Page 14: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THEOREM AND EXTENSIONS 3 5 9

Again

= < Z m r, -+-O(1), (5.9) m r, m r, = f - - a~ --- ~=l f Z a~

and the functions f ( z ) - a~ have finite order in D since f ( z ) does. We can thus apply (5.8) to f - av and obtain

(1,) _< log+ T ( r , f - - a,) -+- O (1) -~_ A log+ T(r , f ) -k- 0 (1). m r, f __ av

Thus (5.8), (5.9) yield

m + m A (q q- 1) [log+ T (r,)') -]- O (1)],

and now (5.7) follows from this and (1.7). It is also clear that if T ( r , f ) ~ a s r -~ R, then

S (0, r, f ) �9 .~ 0, T (r, y)

which is the relation (1.11), so that the defect relation (1.10) also holds. This completes the proof of Theorem 1I.

6) It may be worthwhile at this point to say a few words about the conditions on D in Theorem II. The condition (5.3) implies that D contains almost all points of [ z l = R. On the other hand (5.4) is not a metrical condition on the complement, since even the complement of a countable set need not satisfy (5.4). To see this, we first divide (0, 2 r~) into two equal arcs, then (=, 2 ~) into

( 2 2 equal arcs and more generally 2 ~ 2 n into 2 2~+1 2,~, 2~+i equal arcs, each

of length ~z/(2~"+l+"+~). Denoting these arcs by (~,, [3,) we have

1 _ _ rc 2 :~'~+~+'~+1

2'~+ ~ ~ log 2 __ - r c~.

On the other hand there exist Cantor sets of dimension 1 whose complementary arcs satisfy (5.4).

Page 15: On Nevanlinna’s second theorem and extensions

3 6 0 w . x . R A ~ M A r ~

Consider next the condition that the arcs (~. ~) are properly contained in D. This is always satisfied if D is bounded by a finite number of analytic arcs lying except for

their endpoints in [ z I > R. Let yv be such an arc having endpoints (Re i%, Rei~). Since y~ is analytic and does not coincide with an arc of ] z [ - - R, y~ must

have contact of finite order n at most with [ z l - /? at these endpoints and

so we can find a region of the type

z - r d ~ < R + . (0 - %). ( ~ - 0)',, ~ < 0 < ~ ,

contained in D. Also since there are only a finite number of arcs we may take ,, n the same for all of them. Thus condition (5.1) is satisfied. In the case of infinitely many arcs (5.1) assures that D extends at a uniform rate in some sense

across those arcs of [ z ! : R contained in it.

P r o o f o f T h e o r e m ill 7) We proceed to prove Theorem Iii, quoted in the

introduction. Put

(7.1) w z P -t- i \1 - - w/

so that the angle l a r g z J < ,-c corresponds t o ] ~v [ < 1. Then 2p

(7.2) / (z) = / [z (w)] -- g (w)

is meromorphic in the w plane cut from - - 1 to - - co and + 1 to -[- co, along the

real axis. We take for D the part of this cut plane lying in ] w[ < 2. Then D

contains the whole of ] ~v ] ~ 1 except w - - + 1 and if w -- e/~ we have

2 d ( w ) ~ ] s in0 ] > - - r a i n t ] 0 ] , [ ~ - - 0 ] , I ~ + 0 l t , i n - - ~ < 0 < %

~z

so that D properly contains the arcs - - ~ < 0 < 0 and 0 < 0 <~z, and

(5.3), (5.4) are satisfied. Suppose next that n (d, a) is the number of roots of the equation g (w) - - a

for which w is distant at least d from the boundary of D. The corresponding

w must then satisfy l w - - l [ ~ d

and since I w l < 2 w e deduce from (7.1) that the corresponding z satisfies ]

Page 16: On Nevanlinna’s second theorem and extensions

ON NEVAm.~r~;S SF.CO~. THEOn~M ~Nt, EXa~SmNS 361

Now the number of roots of f (z) = a in the circle (7.3) is O [d -(k+')/p] if f (z) has

finite order k in the plane and since each z corresponds to at most l/p -l- 1 values

of w in D we obtain n (d, a) ~--- O [d-(k+s//P], so that g (w) has finite order in D.

Thus g(w) satisfies the conditions of Theorem II and we deduce from that

Theorem that if w, (a) are the roots of g (w) = a in [ w [ < 1 then

~ ( 1 - - I w,,(a) I)

either converges for every a, if g (w) has finite characteristic in [ w] ,< 1

diverges for every a with at most two exceptions.

Put now z = r r Then

o r

so that

I w I~ -- ] zP-l_zp@. 1 ~---- r ~ p - 2 r p cos pO-q- 1

r 2P + 2 r P c o s p 0 4 - 1

1 - I w I ~ = 4 rP cos p 0

r ~p -t- 2 rP cos p 0 -+- 1

Also w,, runs over the roots of g (w) = a in [ w I < 1, while z runs over the

' ~ The convergence of ~ (1 - - I w~ [) roots r,, e 0f, of f (z) =: a in I arg z I < 2--~o"

4 r~ cos p 0 is equivalent to that of ~ ( , 1 - - i w , = [ ~ ) i .e . ~ 1 + r: ~ + 2 r P c o s p O '

cos p 0 since r,, ~- co . ]h i s proves Theorem I11. and hence of ~ r~

We next show thai if p is a positive integer and the points z~ = r e~~

2-,~,are such that (2.1) converges, then there exists a function f ( z )

meromorphic in the plane and bounded in I arg z I < =- such that f (z,,) - - 0. 2 ~

Consider in fact

z : - z , ~ n=, z~ + z~ \ z n / "

Page 17: On Nevanlinna’s second theorem and extensions

362 w. l~.. I-I A Y ~r A N

It is easily seen that the product is convergent in the plane to a meromorphic

function of order k at most provided that (2.1) converges and ~ I z, I -k converges.

Also if all the z,, and z lie in } arg z ] < ~ each factor of the product satisfies 2 ~

z~ + z~ ~.

Thus I / (z) l < 1, i. e. f(z) takes no value of modulus greater than 1 in

I arg z I ~ ~ . Thus for a meromorphic function of finite order in the plane, the 2 p

condition that (2.1) taken over the zeros of f (z) in 1 arg z [ < -~-- should diverge, 2t~

is the weakest condition on the zeros which will insure that f ( z ) satisfies a

Picard Theorem in the angle.

C h a r a c t e r i s t i c of f (z) a n d f ' (z). 8) A problem mentioned by Nevanlinna

as due to Bloch is the following (~). Suppose f ( z ) has bounded characteristic

in I z l < 1, is the same always true of f" (z)? Without further restrictions the

answer is no. Consider in fact

f (z) ~ a . . . . " Z ,

where a is a positive integer. These functions have been investigated by Little-

wood (~) and he has shown that i f - 1 ( c < 0, then f(z) is continuous and

s o bounded and of bounded characteristic in I z l --< I. On the other hand.

f ' (z) - - ~ a (i+~ z ~'. t t = l

Here 1 - ] - c > 0 by hypothesis, and in this case Littlewood showed that if a is large enough there exists a sequence ~,, < ~,,§ "~ 1 such that

I f ' (z) I > A (a, c) (1 - - p,,~-<'+~ I z I = ~,.

Page 18: On Nevanlinna’s second theorem and extensions

ON N E V A N I , I N N A ' S SECOND T H E O R E M AND EXTENSIONS 303

Thus

, I f " ~ [ , T (p,,, [ (z)) "-- - - - log + I (P- e '~ I a o 2 ~ J o

> tl + c) log | m [~n . . . . . . -~ O (1), as n-)- o%

and so / ' (z) does not have bounded characteristic in I z I < 1. We can however prove

Theorem IV. Suppose that f (z) satisfies the hypotheses of Theorem I!

and has bounded characteristic in [ z [ ~ R. Then all the derivatives of f (z)

also have bounded characteristic in [ z I < R. We note that if f ( z ) has a pole of order q at a point, /(~)(z) has a pole

of order p -~-q ~ (p -Jr- 1) q. Thus

N (r, I<~' (z)) ~ (p -F 1) N (r, f (z) - - O (l), as r-~./?,

by hypothesis. Again

f(') (z) m(r,f<'>(z))_<m(r,f)+m r, f i - ~ ]"

Here the first term on the right hand side is bounded by hypothesis as r - ~ ,q

from below. Also for the second term we have by Theorem I

( ,,<z> I ( ' ) f m r,-7~zj .)~- .4 (p) log + m (r, f ) - ] - l o g + m r , ~ - ~ - / ( r ) - - ~ O ( l ) ,

and here the right hand side remains as bounded as r -~ R by our hypotheses and

lemma 6. This proves Theorem IV.

We can deduce a corresponding result under the hypotheses of Theorem 11I. We have

Theorem V. Suppose that f (z) is meromorphic of finite order in the plane

and has bounded characteristic in I arg z I < - - . Then the same is true of 2 ~

all the derivatives of f (z).

Page 19: On Nevanlinna’s second theorem and extensions

364 w. K.. HA'gMAN

We again make use of the substitutions (7.1) and (7.2). Then since f (z) has bounded characteristic in j z l < 1, g I w) has bounded characteristic in

w I "< 1, and satisfies the conditions of Theorem IV. Thus g" (w) has bounded

characteristic in ] w l < 1. Now 1 + l

f ' (z) = g ' [w (z)] a w = g, (w) P (l - w) ~ d z •

2 (1 + w) P

Since g" (w), (1 - w ) t / 0 + ~, (1-~-w)~-I/0 each have bounded characteristic in

] w l G 1, so has f ' [z (w)], and so [ ' (z) has bounded characteristic in

By repeating the argument the same is true of the higher deri- [arg z I < 2 ~"

vatives of f(z). This completes the proof of Theorem V.

S E C T I O N !I . 9 ) I n this section we consider functions [(z) mero- morphic and of finite order in 1 z S < 1, and obtain conditions under which the second fundamental Theorem holds effectively for f (z) as [ z ! < 1 is exhausted by a sequence of circles ] z - - z,~ ] < r,. In addition to (1.1) to (1.4) and (3.3) we define

(9.1) ['* (z) - - f (z) 2- ~ [ ~ "l - ~ z '

where bl , . . ,b~,r are the poles of J(z)lying in [ b ~ - - z 1 1 �9 ~ - - . An empty I 1 - - b ~ z [ 2

product is taken to be 1. This definition, which is convenient for our purposes here differs slightly from one given in M. M. p. 128, where [(z) is also divided by a corresponding product taken over the zeros.

Circles j z - - zo I < r will be supposed contained with their circumference in [ z l < 1 unless the contrary is stated, i. e.

(0.2) IZo I < I - r < 1.

A sequence of circles I z -- z, I < r,, will be said to exhaust I z I < 1 if

(0.3) r. -~ 1 as n -~ c~,

Page 20: On Nevanlinna’s second theorem and extensions

) O N N E V A N L I N N A S S E C O N D T H E O n E M A N D E X T E N S I O N S 365

and to exhaust ] z i < 1 properly (w. r. t. a function f (z)) if in addition

T (z., r., [)

1 (9.4) log 1 - - r .

if there exists a sequence of circles exhaus t ing [ z [ < 1 properly w. r. t. f (z) , we shall say that the Riemann surface of f ( z ) is properly exhaustible. We shall prove the following results.

Theorem VL Suppose that f (z) is meromorphic in l z I < 1 and that

(9.5) lim (I -- I) log M (r, f , (z)) - - + c~.

Then T (Zo, r , f ) is unbounded as Zo varies subject to !Zo ! < 1 - r for every f ixed r in 0 < r < 1. Further i f

(9.6) lim (1 r) log M (r, f ) : - ~ - oo,

then either (9.5) holds, or

(9.7) T (r, f) . . . . . . . . . . . . . . - ~ r

1 ,Og 1 - r

, a s r ~ l .

In either case the Riemann surface of f (z) is properly exhaustible.

Note that if f (z) is regular (9.5) reduces to (9.6) with lira instead of lira. Also the conclusions of Theorem VI are not altered if f ( z ) is replaced by 1/( f ~ a ) for a finite a in (9.5) or (9.6), since by Theorem A the condition (9.4) is unaltered by such a transformation. [f the sequence of circles [ z -z,~ I < r,, e x h a u s t s [ z [ < 1 properly, the second fundamental Theo rem and its usual extensions become effective. We have.

Theorem VIA Suppose that f (z) is meromorphic and o f finite order in z : < 1 and that the circles [ z - - z . ] < r. exhaust [ z ] < 1 properly w. r. t. f (z). Then

(0 I f ai, a~, .. . , aq denote q ~_ 3 distinct complex numbers, finite or inlinite, and ~ > O, we have for n > nl

(9.8) ( q - - 2) T(z. ,r , , , f ) < (l 4- ~) ~ N z . , r . , - -

~4 " R e n d . C i r c . ; ' ) l a t c )B . P a l $ r ) ) l o , - - s e r i e I I . t o m o I I - a n n o 1953

1

I - a,)

Page 21: On Nevanlinna’s second theorem and extensions

3 6 0 w . ~ . H A Y M A ~

(i i) I f q - - 3 in (i) we may replace the av by distinct meromorphic functions

av (z) having bounded characteristic in [ z [ < 1.

(i i i) Suppose that I is a positive intes av (z), v - - 0 to l are meromorphic

functions having bounded characteristic in [ z [ < I and

(9.9) ~b (z) = ao (z) f (z) -}- ai (z) f ' (z) -}-- �9 �9 �9 --}- aL (z) fc,) (z).

Then if n > n2 we have

(9.10)

T ( z . , r . , f ) < (l q-r [ N ( z . , r , . f ) + N ( z . , )) ( ' ) ] r., + N z., ,r., b ( z ~ l .

Here (i) is just Theorem B. in the case of concentr ic circles, the extension

(i i) is due to Nevanlinna(7) and (i i i) as well as s o m e more general results

is due to Milloux(S).

P r o o f o f T h e o r e m Vi. 10) W e base our proof of Theorem VI on a

number of lemmas. r - - p 1

L e m m a 7. Suppose that 0 < ~ < r < 1 a n d - - _~_ -- . Then 1 - - o r 2

log i _ - . ~ r < A o 1_+_ ~ log ~, r ~ p l ~ , a r

where Ao - - 3 log 2 - - 1.039 .... Asymptotic equality holds 2

r - - p 1 while . . . . . . . . . . = - - .

1 - - ~ r 2

W e put

and

so that

1 - - 0 1 - - x 1 + ~ ' l + x

r - - ~ 1 - - h

1 - - p r 1 f h

1 - - r 1-~-~

1-+- r 1 - - ~

1 - - h x

l + h x

as ~,r tend to 1,

Page 22: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA S SECOND THEOREM AND EXTENSIONS 3o7

We also put

~ (p, r):

log 1 ~ p r log 1 4-____hh 1 - - 9 r - - p - - x 1 - - h

1 4- p log _1 log 1 -[- h x r 1 - - h x

1 h ~ 4 - 1 h4_~ - .. ~ + ~ - ~- .

1 1 1 4- ~- (h x)~ + -~ (h x)' + . . .

so that

(10.1) 1 h e j v 1 h4 1 l o g - - - - c D ( p , r ) < 1 -~ -~ - -5 4- . . . . . 2 h

l + h 1 - - h

It is clear that right hand side of (10.1) increases with h and by the hypotheses

of lemma 7 we have 1 - - h ~ 1 1 - - - - so that h ~__ -- . Thus (10.1) gives (b (p, r) < ~ + k 2 3

3 -2- log 2, which is the desired inequality. Also equality holds in (10.1) asymptotically

l __3 log 2 if p, r tend to 1 while h - - - as x - ~ 0 i. e. ~-~ 1 and s o ~ (p, r)-~ 2 3

i . e . r - - ~ _ 1 This completes the proof of lemma 7. 1 - - p r 2

We have next

L e m m a 8 . Suppose that O ~ l z o [ ~ 1 - - r < l , [ Z - Z o l ~ r , l a [ < l. Let

r 2 - ( z - z o ) ( a - - z o ) 1 - - a z (10.2) g (Zo, r, z, a) - - log+ -- log+ 2 (z a) " r (a - - z)

Then we have

00.3) g(Zo, r, z,a) ~Ao r4- Iz-- zol log+ r r-- lZ--Zo[ t(a)'

where

(lO.4) ( , ) t ( a ) : m a x [Z - -Zo l ,~ r , l a - Z o l .

Page 23: On Nevanlinna’s second theorem and extensions

3 6 8 w . K . I.-i A u M ,A IN

C l e a r l y g ( z o , r, z , a ) can be positive only i f l a - z o l ~ r , so that we make

this assumpt ion in wha t follows. We have in this case,

(lO.5) - / - ( z - Zo) (a - Z o ) 1 a z I l o g + ~ . O. log 2 (z - - a) i - - 2 r (a - - z ) I

For the left hand side of (10.5) is a harmonic function of a in the circle

I a - - zo i ~ r so that by the maximum principle we may suppose l a - - zo I = r.

In this case (10.5) reduces to

log 1 -- a z ~ 0 , z - - a

which is true since the c i r c l e l a - - Z o l = = r lies by hypothes is i n l a I " ~ 1 .

It follows also that i l i a - z o I ~ r we may write log+ instead of log in

(10.5) so that (10.2) gives

(1o.6)

r ' - (z - Zo) (~ - Z o ) g (zo, r, z, a) ~- log+ - - , r ~ a - - z )

r 2 - ( z - z o ) (a - - Z o) l o g + 2 r (a - - z) --

We next put

(10.7) z -- zo a - - Zo : Z t , - - - 121.

r r

Then (10.6) becomes

(10.8) g(Zo, r , z , a ) ~ log+ 1 - - -z , a t _ log+ a t - - Z t I

1 - - g I Qt .

2 (0 t --. z I )

Suppose first that ?, ~ I at I ~ 1, where ~, is given by

~ - I z , I 1 1 .... ;~ l z , I - - 2"

Page 24: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THEOREM AND EXTENSIONS 36~

In this case we have clearly l at l ~ max (l z~ l, l ) ,

in (10.4). W e have also

so t h a t t ( a ) : l a - - Z o ]

t - a ~ z ~ 1 - 1 a ~ l l z ~ l 2

Thus in this case (10.8) and lemma 7 give

g (Zo, r, z, a) ~ log 1 - - i zi l l a~ [ < Ao l -+- ! Zx l log 1 , l a, I - I z, l 1 - 1 z ~ l l a~

which reduces to (10.3), using (10.7). Suppose next that I a~ I ~ ~.. Then we

have trivially from (10.8) and lemma 7

1 - ~ I z, j / < 1o 1 + t z~ l log 1 g (Zo, r, z, a) ~ log 2 ~- log ~. -~ i zi [ ] 1 - - [ z~ [ )-f'

Also by hypothes is I <_~ 1 _ __~ r and it follows from (10.9) that z la, f l a - - z o l

) ~ m a x , I z ~ l so that

r ) -~- ~ rain ' - I' - -

r I z zo l a - - Z o l t (a)"

Thus (10.3) holds in this case also and so holds generally.

11) W e are now able to prove the following fundamental

L e m m a 9. Suppose that O ~ l zo [ < 1 - - r < 1 and that l z --zo [ < r.

Then i f f (z) is meromorphic in l z l < 1 we have

log I f , (z) [ < r -~- I z - - z . t m (Zo, r, f ) -[- Ao n (zo, t, f ) r - - l Z - - Z o l o

where to = max ( 2 ) r, l z - z o l .

Page 25: On Nevanlinna’s second theorem and extensions

370 w. x. riAyraAr~

The poisson-Jensen formula applied to f(Zo -q- P e ~~ gives for 0 ~ p < r

( r ~ - - p~) d �9 1 ]"~ log I f (Zo -~ r ~'o) [,~ log I f (Zo -t-- [9 e~0)[ --- 2 r, g o - - - 2 r p cos (0 o) +

r2 (b; - - Zo) [9 e iO I r ' - (a~ Zo) p e'~ ~- ~ log+ - - - - E log+ --

r ( b y - - Z~ - - [9 e iO) I r (a~ -- Zo -- p e '~

where at, are the zeros and b, the poles of f (z). We write Zo ~ [9 e t~ z,

ignore the terms over the zeros, which are nonpositive, and replace the inte-

grand on the right hand side by the larger

r + [9 l o g + [ f (Zo -~- r eiq')] . r - - [ 9

In this way we obtain

r ps--L- log [ f (z) [ ~ ---------~ m (Zo, r, f) -f- ~ log~-

r - - p

r 2 - ( b ~ - - z o ) ( z - - z o ) .

Using (9.1), (10.2) we deduce

log I f , (z) I - log I f (z} I -- ~ Jog+ ~ ( : - t , 0

r§ r - - [ 9

m(Zo, r , f ) - [ - ~ - ' g ( : o , r , z , b ~ ) , v

and s ince [ z - - Zo ) [9, (10.3) g ives

I r] (11.1) I o g l f , ( z } l <r--[gr~-~ m(Zo, r,f)--[-Am ~ log+ t(b,~) "

Now by (10.4) we have, using the notation of lemma 9,

( , ) t ( b v ) : m a x t z - z o l , ~ r , [ b v - z o l - -max( to , lb , - -zo]) .

t. d t Also the contribution of each pole b, to the integral r n (Zo, t, f ) ~ -

Page 26: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THEOREM AND EXTENSIONS 371

r is just log+ t (b,)" Thus

/ '~ d t r n (z o, t, f ) - - - "-- ~ log +

,o t t(b,)

Now lemma 9 follows from (11.1) and the fact that I z - - Zo I -" P. We quote one final result

L e m m a l o . ( ~ ) Given ~,0 < ~ < 1, we can f ind t such that ~ ~ t 1

- - (1 -~- ~) and 2

l o g , f ( z ) [ < iog l f . (z) l + A n ( 3 - - + ~ ,f), l z l - - t,

where A is an absolute constant.

12) We can n o w prove Theorem VI. We s h o w first that if (9.5) ho lds

T (z,,, r, f ) cannot be uniformly bounded in [ z o I < 1 - - r for any r. Suppose contrary to this that we have for some fixed r, where 0 < r < 1

(12.1) T ( z ~ , r , f ) < K, I Zo I < 1 -- r.

Then if z

This gives

1 is real, 1 - - r < z < I choose zo so that z, Jr- r : - - (1-[- z}.

2

1 (l - z ) . r-(Z-Zo)--r- Iz-zol:-~

Also lemma 9 and (1.1), (1.2) s h o w that

loglf,(z) l<Ao r-]-IZ--Zol T(zo, r , / )< 2 A ... . . . . . K = . . . . r--tz---zol 1(1 --z)

2

4 A o K l w z

Similarly by considering f (z e i~ instead of f (z) we have generally

i~ e ~

4 A ~ l - - r , log I f . (z) l < 1 - - I z l

log 34 (~, f , Cz)) < - - 4 A ~ - - r < ~ < 1. l - - p

Page 27: On Nevanlinna’s second theorem and extensions

372 w. x. uA~rra.~r

This contradicts (9.5) and so (12.1) must be false.

Suppose next that (9.6) holds but (9.7) is false. Then we cannot have

(1 - - p) n (P,3') -~ 0% a s p -)- 1,

since this would imply (9.7). We can thus find ~ arbitrarily near 1 such that

3 @ p , f~ < K 4 K / 4 1 3 --I- p 1 - - p '

4

1 ( l § where K i s a constant. Hence by lemma 10 we can find t in p ~ t <~ ~-

such that

4 A K log I f , (=) ! --> log I f (z) i , t z ] = t,

l - - p

and so since t > p we deduce for some t arbitrarily near 1

log M (t, f . (z)) log M (t,f) 4 A K 1 - - t

Now (9.6) implies (9.5). Thus with the hypotheses of Theorem VI either (9.7)

holds, which gives (9.4) with z~- - -0 provided (9.3) holds; or T (zo, r, f ) is un-

bounded for every fixed r, in which case given any sequence r~ satisfying (9.3)

and 0 <: r~ < 1, we can find z~ such that I z, ! < 1 -- r~ and

1 )2, T ( z ~ , r , , f } > log 1 -- r,~

and this again implies (9.4). Thus we can always tind a sequence of circles

[z - - z~ ] < r~ exhausting ] z I < 1 properly w. r. t. f (z) (even with assigned rfl)

and so Theorem V! is proved.

P r o o f of T h e o r e m VII. 13) To prove Theorem VII, we again need a subsidiary result

L e m m a I[. Suppose that f (z) is rneromorphic of finite order in [ z 1 < 1.

Then given a positive integer p, there exists a constant K depending only on f (z)

Page 28: On Nevanlinna’s second theorem and extensions

ON NFVANLINNK'$ SECOND THEOREM &ND EXTENSIONS 373

and p such that i f O ~ ! Zo I < l ~ r < 1 we have

m Zo, r, f ~ - / < K i O g r ( 1 - - r ) '

We shall denote by K any constant independent of zo and r, not necessarily the same each time. Since f (z) has finite order in I z I < I we have

(13.1) T (r, f ) < K (1 - - r)-% 0 < r < 1.

l Taking r - - - - (1 q- I z I) this gives, using lemma 9

2

4 (13.2) log I .t, (z) l < A o - - - - 1-1~!

T ( r , f ) < K ( I - - l z l ) K, I z l < 1.

Again (0.1) gives

(13.3) log I f (z) I = log I f . (z) 4 - ~ l o g ' ; 1 - b ~ z 2-iz---~,-;~"

where b, are the poles of f (z) satisfying b v -- z 1

1 - -b~ z ': <- 2 ' and this latter implies

�9 _ i ~ l - - i z l l - - l z l . Ib~ ~ 1 + 2 1 z l 1 , b , ; > 2-q--i z I ' 2 + ] e l 3

Thus by (13.1) the number of terms in the sum in (13.3) is at most K (1 - I z [)-K,

and if ~ (z) denotes the distance of z from the nearest pole b~, then (13.2), 13.3) give

(13.4) l i ) ( n - I z I) -~c. IoglfCz) l < K 1 -t- log+ ~ (z

We now use Theorem 1 applied to f ( z o - t - z ) i n s t e ad off (z ) with r instead of R, and we take for D the domain I Zo + z I < 1. Then

dr (O) = l - - I Zo -~- r e iO I,

and since f (z) has finite order in J z I < 1, nr (0), the total number of roots of

the equations f (z) = 0, 1, oo in I z I < 1 - - 1 dr (0), satisfies 2

,,r (0) < K [dr (0)l -~,

Page 29: On Nevanlinna’s second theorem and extensions

374 w. K. .AvM,r~

so that

(13.0) 1 = I (r) = ~ log+ n, (0) -t- log+ dr (0)

< K (1-J- ffr~ log I dO) 1 - - I Zo - } - r e iO [

Also applying (13.4) to lff(z), which also has finite order in z] < 1, as well as f (z), we deduce that

( ! log If(Zo + r e'~ ! I _< K 1 + ~og+ ~

where ~,. (0) denotes, as in lemma 2, the radius of the largest circle centre z, -~- r e i~ in which [ (z) is regular and unequal to 0 or 1. Thus we have

(13.7) log+ t log[ f (Zo + r e'~ [ : d 0

< K 1 log 1 - - I Zo q- r e iO [ ~- I~ I~ ~r- d 0 . o

Again it fo l lows from lemma 4 that

f z~ 1 f ~ 1 log+ log+ d 0 < log< o ~ (0) ~ (0)

- - d O < K ( l ~ - I ~ ' r

where i is defined as in (13.0). Combining this with (13.0), (13.7) and Theorem l we deduce finally that

(13.8) m Zo, r, ~ : ~ ] < K l ~ l o g - - + r o log l _ ] z o _ J _ r e , 8 I . d O .

We proceed to estimate the integral in (13.8). We have to show that

f ~ 1 1 (13.9) log .-- d 0 < A log 1 - - Zo-~-re'~ r(1 -- r)'

where A is an absolute constant. Lemma 11 will follow at once from this and

Page 30: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THF~OREM AND EXTENSIONS 375

(13.8). To prove (13.9) it is clearly sufficient to take z o real and positive,

0 < Z o < 1 - - r . Then

�9 __ 2 r 2 (13.10) 1 - - I Z o - ~ - r e '~ 1 - - z o - - - - 2 r z o cosO

-- 1 - - (Zo-}-r) 2 - ~ - 2 z o r ( l - - c o s 0 ) ;

1 (1 - r). Then Suppose first Zo < 9~

�9 1 1 1 - - 1 Z o ~ - r e ' ~ : (1 - - r), 0 < ~ <: 2 rq 2

I (1 --- r), (13.10) gives so that ( 1 3 . 9 ) i s trivial. Next if z o > -~-

�9 1 (l_[Zoq_re,O[,) l - - [ Z o + r e ' ~ > ~ -

1 > Zor ( l - - c o s 0 ) > ~- r(1 - - - r ) ( l - - cos0),

since zo-1-r < 1, so that we have in t h i s case

fo ;~ 1 2 f ] ~ 1 log . ~- log d 0 1 [Zo-~-re'~ dO<zrcl~ - - r(1 - - r ) l - - c o s 0

< A log ~-1 < A l o g r (1 - - r ) r (1 - - r)"

Thus again (13.9) follows. This proves (13.9) and lemma 11 follows from this and (13.8).

14) The proof of Theorem VII now follows from standard arguments.

We have from Theorem B as n ~-co

q ( 1 (q--2)T(z~,r~,f)< Z N z~,r~,r_av

Page 31: On Nevanlinna’s second theorem and extensions

376 w. x . .AY M^r~

w h e r e

q

u /" ) + o(z).

f - - - 12v

It thus fo l lows from lemma 11 that

l 1_~ S (z,,r., f ) - - O l og 1 -- r . t - - - o { T(z , , r , , f ) l, a sn -~oo ,

since the circles I z - z. I < r.. exhaus t [ z i < 1 p roper ly . This p r o v e s (i).

Nex t if a~ (z), a~ (z), a.~ (z) are unequal func t ions of b o u n d e d character is t ic

in l z l < 1 put

(z)-- f ( z ) - a t(z) a 2 ( z ) - a 3 ( z ) f (~) - a= (z)' a, (~) -- a, (z)"

Then �9 (z) z 0, 1, ~ on ly if f (z) == a, (z), a2 (z), as (z) or a I (z) = az (z), a~ - - a ,

a.~ :-= at. T h u s

(14,1)

( ) ( , ) 1 -i- N z . , r . , - ~ - N r . ,

' ( '~) ( ' ) < v~ N z,, r,, f _ a , (z Jr- ~ N z., r., a~ (z) - av (z ) ----- l ~p .<v<z I

' ( , ) < ~ ~., r., + o(l). ,=, f - at (z)

For s ince the a~ (z) - a , (3) have b o u n d e d character is t ic in I z l < 1, w e have

a s ~ -),- ~:)

( '2 ( ' ) z . , r . , - - ~ 0, 1, - < co. a~ a F a v

For a s im i l a r reason we have as n ~ cx~

(14.2) T (z,, i",, [ ) : T (z,, r,,, �9 (z)) q-- 0 (1).

Page 32: On Nevanlinna’s second theorem and extensions

ON NEV/INLINNA~S SECOND THEOREM AND EXTENSIONS 377

We can n o w apply (9.8) to (I) (z) with q --:- 3, a I ~--- 0, as • 1, a8 = oo and the

results for f(z)follow f rom (14.1) and (14.2). T h u s (if) is proved.

It remains to prove (i i i). We note fo l lowing MiUoux(s) that

r. ~,r. , 7 _<,,, ~,,,r., 7 + . , z.,r,,,

( , (14.3) ~ lV (z~, r,,, +) -[-- N z,,, &, ~ - - _ - {

where

S(z . , r . ) - -m z.,&, -~-2m z~,r~, ~-m z~,&,~__ 1 q- 0 ( 1 ) .

This follows on app ly ing Theorem B to + (z). N o w

l

* (z) ----- ~ a , (z) f(') (:),

and since the a v (z) have by hypo thes i s b o u n d e d characteristic and so finite

order in I z I < 1, and f(z) and all its derivatives have finite order in I z I < 1, qJ (z) also has finite order in l z I < 1. Also

m z,,r,, : m z , , r , , ~ a,(z) (z) ,=o f (z)

( ' ) - ~ a s / / . - ~ c ~ := O log 1 ..... r,,

as iemma 11 shows , On apply ing lemma 11 deduce therefore that

14.4 ( ' / S(z , , , 1 , , )= O log I - - r ,0 '

also to ~ ( Z ) - - 1

a s / / . ~ o o .

and ~ (z) we

Page 33: On Nevanlinna’s second theorem and extensions

378 w. K. HAYMAN

Again the poles of + (z) occur when f (z) or one of the av (z) has a pole so that

1

(z., r., ~ (z)) ~ N (z., r., f) -" ~ N (z., r., a, (z))

<_ N ( z . , , . . , f )+O( l ) .

Using (14.3), (14.4) we obtain

and adding N (: . ,

( , ) ( N ( z , , , r . , f ) + N z, ,r , , ,~--~-~ 4 - 0 log

') r., [ : to both sides we obtain

~_N(z,,,r.,f)~-N(z~,r.,, -~-N z,.,r,,,~-(z)_

+o(,og, _1,)

, ) )

1 &

,)

) / [ (1 - - !".) 2 - - [ z . !~] A (z . , r . , f ) -~ oo .

Then the Riemann surface of f (z) is regularly exhaustible in Ahlfors' sense, in

particular this is always the case i f there exists a sequence of circles properly

exhausting[ z [ < 1 in the sense of (9.4).

Corollary: I f (9.5) or (9.6) holds the Riemann surface of f (z) is regularly exhaustible.

Our proof is based on

(15.1)

Theorem VIIL Suppose that f (z) is meromorphic in [ z I < 1 and that there

exist a sequence of circles ] z - - z, I < r, exhausting I z I < 1 so that

Now Theorem VII (i i i) follows on making n ~ c~. This completes the proof of Theorem VII.

THE AHLFORS EXHAUSTION THEORY.

15) We recall the definitions and notation of the introduction, paragraph 3.

We proceed to prove the following

Page 34: On Nevanlinna’s second theorem and extensions

ON N E V A N L I N N A ' S S E C O N D T H E O R E M AND E X T E N S I O N S 370

1 L e m m a m2. S u p p o s e t h a t 0 ~ [ z o [ < 1 - - r, - - ~ r < 1 a n d t h a t

2

(15.2) A (Zo, r, f) > 4 ~ h ~

(1 - r ) ' - I ~o I ~

T h e n t h e r e e x i s t s a c i r c l e [ z - - z~ I < ri c o n t a i n e d in I z I < 1 a n d c o n t a i n i n g

[ z - - Zo ] < r s u c h t h a t

(12.3) A (zt, r~, f ) > h L (z~, r~, D .

This result is a slight generalisation of a corresponding one of Ahlfors, (~a)

w h o showed that (12.3) holds for some rl, r < r i < 1 and zl ---- o if

(15.4) A (0, r, f ) > - - 2 z c h ~

- - r o

W e proceed to derive our result from that of Ahlfors. W e may suppose wi thout

loss in generality that Zo is real. Then Zo -~- r, Zo - - r are ends of a diameter of the cir-

cle I z - - z [ , < r . W e map I z [ < 1 onto [ w [ < 1 byab i l inea r t r ans fo rmat ion

which keeps the real axis invariant and maps Zo ~ r on to two points ~ to.

Then the circle [ z ~ Zo [ < r b e c o m e s ] w I < ro and

(15.2) A (Zo, r, f (z)) = A (0, ro, f [z (w)])

L (zo, r, f (z)) = L (0, ro, f [z (w)]).

Also the crossratio of four points remains invariant under the transformation so that

l - - 1, Z o - r, zo-~-r , 1 1 - - - / - - l , - - r o , ro, l f

(1 - - r -+- Zo) (1 - - r - - Z o ) _ (1 - - ro) ~

2 �9 2 r 2 �9 2ro

i~ e~

r _ _ r o

(15.6) (1 - - r) ~ - Zo ~ - (1 - - r o ) ~

1 It is clear from (15.6) that ro ~ r. Suppose n o w that r ~ - - and (15.2)

2 holds.

Page 35: On Nevanlinna’s second theorem and extensions

380 W . K . ItAYMAN

Then we deduce from (15.5), (15.6) that

A (0, ~o, f [z (w)]) > 4 ~ h~ t 1/(1 r - - r ) ~ - I Zo i ~

Y i r~ > 2 ~ h ~ 2 - - 4 ~ h ~ 1 - - r o ) 2 1 - - ro'

and this is (15.4). Thus for some p such that ro < p < 1 we have by Ahlfors ' result

A [0, p, f [z (w)]] > h L (0, p, f [z (w)]),

and so (15.3) follows, where I z -- : i I < ri is the inverse image of the circle

[ w I < P in the w plane. This proves lemma 12.

16) We can now complete the proof of Theorem VIII. Suppose that there

exists a sequence of circles [ z - - z, I < r, satisfying (15.1) and exhaust ing

I z l < 1. It then follows from lemma 12, taking

h,2 :: I/[-~ - - r-) 2 - - I z,~ 1 ~] A (z,, r.,, f (z)) 4 ~

that for a sequence of circles I z - - z," I < r,,', which contain ] z - - z,, ! < r ,

and so also exhaus t J z ] < 1 we have

L (4, ,:,, i (z)~ A (z~,, r'~, f (z))

I �9 < - - - ~ 0 .

//.

Thus the Riemann surface of f (z) is regularly exhaustible in this case.

It remains to s h o w that if a Riemann surface is properly exhaustible in the

sense (9.4), then there exists a sequence of circles exhaust ing I z J < 1 so that

(15.1) holds. Suppose contrary to this that there is no such sequence. Then we

can find constants ro, O < ro < 1 and K > 0 such that if r o < r < l a n d

I z, [ < 1 - - r we have

(10.1) A (Zo, r, D < K

r - r ) ~ - I Zo I '

Page 36: On Nevanlinna’s second theorem and extensions

O N N E V A N L I N N A ' S S E C O N D THEORE1VI AND E X T E N S I O N S 381

N o w we have by the Ahlfors-Shimizu identity(24)

T (Zo, r, f ) = A (zo, t, f) T -~- O (1) = -[- 0 (1). o o

1 Hence if I zo I < 1 - r, ~ (1 + r,,) < r < 1, we have

~ dt (16.2) T(zo, r , f )= A (Zo, t,f) T + 0(1).

r

1 In fact if [ zo I < 1 - - r, r > ~- (l -+- ro), t ~ r o then

1 1 (1 + to), I z o l + t < r o + l - r ~ r o + - ~ ( 1 - - r o ) : : : ~-

. 1 ( l + r o ) < l and so so that the c i r c l e l z - - z o[ < t is conta ined i n , z l < _ 2

A (Zo, t, f ) is uni formly bounded . We n o w apply (16.1) with t > ro instead of r t

in (16.2) and deduce

f r Kd t T(zo, r , D ~ "~ 1/[( 1 -- 0 2 - 1 z o l 2 ]

+ o 0 )

1 = K log + 0 (1)

1 - - r + l/I(1 - - r) ~ - - I z o I~A

1 < K log - - -~- 0 (1).

1 - - r

1 This result ho lds for ~ (1 --{-ro)< r < 1 and [ Zo I < 1 - - r with a u n i f o r m O(1)

2 and this contradicts the existence of a sequence of circles which e : 'haust I z ] <: 1

properly. This comple tes the proof of Theo rem VIII. The corollary n o w fol lows

from Theorem VI. 25 - R e n d . C i r c . M a f c m . P a l e r m o , - - s e r i e I f - t o m o I I - a n n o ~953

Page 37: On Nevanlinna’s second theorem and extensions

3 8 2 w . K. HAyblAN

P r o o f of T h e o r e m IX. We now prove Theorem IX. We shall denote by K a constant independent of zo and r, not necessarily the same each time it occurs. We need two more lemmas.

L e m m a 13, With the hypotheses of Theorem IX, given ~ > 0 we can

find ro < l, such that if ro < r < 1,1zo l < 1 - - r we have

q

N Zo, r, < (b-l- log ~=, [ a,, ') - - 1 - - I Z o l - - r "

.It follows from our assumption (3.8) that if we write

nlzo, r):=Y~,n zo, r,[__ov,

then for r > re (~), I Zo I < 1 - - r, we have

n(Zo, r ) ~ n ( O , I zo I + r ) < b + ~

l - - l z o l - - r

Hence i f l ( 1 + q ) ~ r < l then 2

q( " f+I/ E N zo, r, < n(zo, t)--~ = V ~=1 �9 ~" I"

2 2

< n(O, I Zo I + r , ) log 2 + - - - - - b--]-~ f ~ d t

1 i z o l - - t " r i .~

1 (1 + rt) < 1, so Also I zo l < 1 - - r ~ l _ _ ( l _ r e ) a n d h e n c e l Z o l + r e ~ _ 2

that the first term on the right hand side is uniformly bounded�9 Further we

may suppose that 1/r e < 1 -t- ~. We then obtain for re < r < 1, I Zo I < 1 - - r

q

~=~ av < (1 --[- e)(b -l- e) log + K. - - 1 - - I Z o l - r

Page 38: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THEOREM AND EXTENSIONS 383

Next we can find rz such that r, < rz '< 1 and

1 K <( ~ log

1 - - r e

Then (17.1) gives if r e < r . < l , [ z o[ < l - - r

q ( l ) N Zo, r, F=--a~ < [(1 + ~)(b --I- ~) -t- ~] log

1

1 - - I z o l - - r

Since ~ is arbitrary lemma 13 follows.

Our next result is

L e m m a 14. I f ro is defined as in l emma 12 and ), as in Theorem IX, then we have /'or ro < r < I, I Zo l < I - - r

1 1 (17.2) m (zo, r , / ' ) < (~. -4- r log . . . . . . . . . . ; K log - -;

1 - - [ Z o [ - - r 1 - - r

(17.3) N (0, r, ]') .< K log - - 1 - - r

We have by Theorem B and lemma 11

(17.4) q ( ':/ ' (q - z ) r (zo, r, f) < ~:~ X Zo, r, I' :--rJ + / r l o g r . . . . . . . . (1 - - r)"

Taking Zo - - 0 and us ing lemma 13 for r > ro this gives

N(0, r , f ) < ( b + r 1 1

- - --l- K log . . . . . . . , 1 - r 1 - - r

which proves (17.3). Also (17.4) gives, us ing lemma 13,

(q - 2) m (zo, r, f) < (b -t- ~) log 1 1

- - r - - II'Zo' - + - x l ~ . . . . . l - - r " 1

This proves (17.2) if one of the a, is 0% so that ~ , - - - - b . If a~ to aq a r e all

q - - 2

Page 39: On Nevanlinna’s second theorem and extensions

384 w. K. HAYM&N

finite we put aq+, : oo and apply (17.4) with ai to aq+l instead

a, to a~, This gives

' ( ,) (q--1) T(zo, r, f) < ~../=l N ZO IF, /" =-a: + N(z*I r, f) -+- K log . . . . . 1 - - r

Subtracting ( q - 1) N (Zo, r, f ) from both sides we obtain

( q - - l ) m ( z . , r , f ) < N Zor, f f - - - - ( q - - 2 ) N(zo, r , f ) + K l o g ........ 1 - - r

V=I

of

1 1 < (b ~- e) log - - -~- K log - - ,

I - - [ z o l 1 - - r

using iemma 13. Now again (17.2) follows, since X - - - -

proves lemma 14. q - - 1

in this case. This

18) We can now complete the proof of (3.12). We take a fixed k ~ 4 and

given p, 0 < p < 1, we define p, by

1 - - p (18.1) I - - p, - -

k

Then for any fixed k and = > 0 we can find p as near 1 as we please such that

( 18 .2 ) n (o , p~ f ) < K k K

1 - - P k 1 - - p

For i f (18,2) is false for all p, > p' then for p' < p < I we have

~1_ d t N (0, p, f ) --- n (0, t, f ) --

t 2

n ( O , t , f ) d t > K d t t l - - t

- - K l o g . . . . . .

1 - - p t

I - - p '

Page 40: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THEOREM AND EXTENSIONS 385

and if K is large enough this contradicts (17.3).

Suppose then that p is so chosen that (18.2) holds. We next choose z : z (p)

so that I z (p) I - - P and

I f , (z) I - - M (p, f , (z)),

W e also take a fixed r so that ro < r < 1, where ro is defined as in lemma 13.

W e suppose p > r and choose Zo so that

(18.3) arg zo --- arg z (p),

(18.4) 1 - - I z , , I - - r = l I p ~ .

It fo l lows that

(18 .5) r-- l z ( p ) - - z o l - - r - - p ~ l z o l - - p ~ - - p - - (l - - 1 ) ( l - - P ) �9

W e next apply lemma 9 with z - - z (p). W e deduce, using (17.2)

log M (p, f , (z)) < r + I z (p) - - z o I m (Zo, r, [ ) -~- A,, n (Zo, t, f ) t r 1 I z (~) - zo I

2k [ (18.0) < (k . . . . l ) (1 p) (~" -~- r log --] - - iZo I - - r

+ A. -~- - t o to - - - n (0, I zo I -t- r, [ (Z))].

/ 1 N o w to - - max ~-2

so that by (18.5)

A lso

r, I z (p) - - Zo I ) - - I z (p) - - Zo I, if p is suff ic ient ly near 1,

. . . . ( 'i Ao r - - t o Ao 1 - - - - (1 - p). to to k

k K n (o, I Zo I + r,/') - n (o, p~, f ) < 1-U~p"

Page 41: On Nevanlinna’s second theorem and extensions

386 w . 1:. ,~A*MA.~

T h u s

Ao . . . . r - - t o

lo

Ao ( k - - 1) K n ( O , l z , , l + r , [ ) < . . . . . . . . . . < K k ,

to

1 since r is fixed and to ~-_- .... r. Again by (18.4), (18.1)

2

1 1 (~, + s) log = (~. q-- ~) log . . . .

1 - - I Zo [ - - r 1 - - , o k

1 . . . . . . . . . ~ (;~ + ~) log k - 4 . , ) t o g :

1 P

Thus (18.6) yields finally

(18.7) log/14 (~, f , (z)) < 2 k [ 1 q _ O ( 1 ) ] (k - - 1) (1 - - `O) ()' -~- 6 ) l o g 1 ~

provided `o is sufficiently near 1 and (18.2) holds. If f ( z ) has only a finite num-

ber of poles, and we put I z [ - - - ` o , then we have (18.2) and f (z) = f , (z) for

all p sufficiently near 1. Hence also (18.7) holds for all O sufficiently near to 1.

Since ~ can be made as small as we please and k as large as we please we deduce

(3.12) with lira instead lira, as required in Theorem IX.

Suppose next that f (z) has infinitely many poles in I z I < 1. We choose .~

1 so that (18.2) holds with 2 k instead of k. Then if O -~_ r __~ - - (1 --[- `O), we

2 have rk ~ ~k, so that

2 k K k K n (0, r~, f ) <_ - - - - - - .

1 - - rk 1 - - r ~

Thus it follows that we can find `o as near 1 as we please such thai

2 k [ 1 + 0 ( 1 ) 1 (18.8) log M (r , / ' , (z)) < ik .... 1 ~-(i---" r) O, + ~) log 1 ~

1 for all r such that p ~ r ~ - - ( l -Jr P). It now follows from lemma 10 that

2

Page 42: On Nevanlinna’s second theorem and extensions

ON NEVANLINNA'S SECOND THEOEEM AND EXTENSIONS 387

there is an r in this range for which

3 ~ ~,f). log 34 (r, f) < log M (r, f, (z)) + A n O, 4

We have assumed k ~ 4 so that ~ ~ - - 3 + ~ and so 4

/'/ 0, 3--[-~ ) K k K k 4 ' f ~ - - < = - - " 1 ~ 1 ~ r

Since also (18.8) holds for some r arbitrarily near 1 we deduce

iim (1 - - r) log M (r, f)

r..~l log 1 - - r

2 k O. + ~) k - - I

and since k is large and ~ small as we please, (3.12) follows.

19) It remains to show that equality is possible in (3.12) under the hypothesis

(3.11) for any finite set of values a t to aq and b < c~. We may suppose without

loss in generality that ai :--O, since otherwise we can consider f ( z ) - an instead of ]'(z). Consider then

(19.1)

We put

and quote the following

1-{--z z

f (z) -" M e x ~ log

J + z Z = X - ~ i Y = l - - z

L e m m a x$. (~5) The /unction

- - ~ + i ~ = :p (z ) = z log (1 + z )

is schlicht in X > 0 and further if ~ ~ 0 in this region, then

X ~ + Y~ 2 X

Page 43: On Nevanlinna’s second theorem and extensions

388 w. K. HAYMAN

Using this lemma it follows that if I w l < M, the equation

f (z) - - M e xz'~ - - w

has exactly one root in z for each root of the equation

W X Z l o g (1 -]- Z) -~- log~.l~ + 2 n ~ i

for some positive integer n, i. e. one root for each integer n at most. Again if

the principal value of the logarithm is taken the integers n satisfy the inequality

X ~ + Y~ (19.2) i 2 n ,~ I <-- ~ + x

2 X

by lemma 15. Now if z, Z are related as in (19.1), we have

4 X 4 X 1 - - ] z [ ~ _ _ - - <

(X-~- 1) ~ + Y~ X ~ -[- Y~

Thus if z,~ is the root of the equation f ( z ) : w corresponding to the integer

n, then (19.2) gives

2~;~

and hence if this root lies in ] z ] < ~ we must have

(10.3) In I < z ~ - + 1 - ~"

The number of integers positive, negative or zero satisfying (10.3) is at most

2 - ] - 2 ~ . / (1 - ~e). We now suppose M so large that those a v which are finite

satisfy Iav I < M. Also the equations f (z) - - 0, oo have no roots in [ z ] .< 1.

Hence if p (~) denotes the total number of roots of the equations [ (z) __-- 0, a~,..., a~,

we have

p (p) < (q - - 2) 2 + ~ _ or p (p) < (q - - l) 2 + ~ _ 1

Page 44: On Nevanlinna’s second theorem and extensions

ON NEVANI'.tNNAJS SECOND THEOIIE.'~I AND EXTENSIONS ~

according as one of the a~ is infinite or not. Thus

iim (1 - - ~ ) p ( p ) ~ ) , ( q - - 2) o r ) , ( q - - 1),

according as one of the c v is infinite or not. This is equivalent to (3.11) with

the relation be tween k, b, given in Theorem IX. Also we have clearly from (19.1)

log M (~, I) ~:: ~, I -1 -~ log __2 __ _~_ log M, 1 - - O 1 --,~

so that equality holds in (3.12) in this case. The inequality (3.12) is therefore

sharp. This completes the proof of Theorem IX.

P r o o f of T h e o r e m X. 20) In proving Theorem X, we shall denote b y K 1

constants independent of Zo for [ zo I < - - . W e also put 2

(20.1) g (z) = f (z)/,~ (z).

Then g ( z ) : l : 0 , 1 in I z l < 1 a n d g ( z ) has poles only at the zeros of ~ ( z ) .

Clearly g (z) has finite order in [ z [ < 1, since �9 (z) does b y hypothesis . W e

therefore have b y (1.6), (1.7), taking q - - 3, at = 0, a 2 - - 1, a 3 - - oo

C, Zo, 2' 7' m - - g ( z ) < m zo, + r n z,,, -FK,

1 where ~ (z) : g (z) (g (z) - - 1). Using lemma 11 with r - - - - this gives

2

( ' ) (20.2) m Zo,~-,g(z) < K, z., I <-2- .

1 W e now suppose - - :~ r < p < 1 and choose 0 = 0 (r) so that

2

(20.3) . f (r e i~ = NI (r,/').

1 Clearly Theorem X is trivial if r < - - .

2

Page 45: On Nevanlinna’s second theorem and extensions

300 w. ~ . nAYr,,A~

Further let Z o - Zo (r) be such that

(20.4) 1 �9 1

I zo 1 - + - ~ = p, l r r 1 7 6 - Z o I - r - I z o I = r -k- ~ - - ~.

This is a l w a y s poss ib l e on c h o o s i n g a r g z o - - O . Let m,, v - - 1 to N be the

zeros of r (z) in l z - - Zo [ < 1__ wi th correct multiplici ty. Let 2

Then as

on ly at z = av while

Wri te

2 (z - - av) L (z, av) = I - - 4 ( Z - - Z o ) ( a v - - Z o ) "

a funct ion of z, L ( z , a ~ ) is regular in I z - - Zo I ~ 1 and van i shes 2

1 I L(z, av)] - - l, [ z - - z o l - - - ~ .

N

(:zo.5) go (z) = p (z) g (z).

1 Then go (z) is regular in I z - - Zo [ <~ and l go (z) ] --- [ g (z) [ for [ z - - Zo [ - - -~ - ,

so that

( ' ) m Zo,~,go(z) = m T h u s l emma 0 gives

1

(20.0) l o g l go (r egO)] < ~ - @ I r e i~ - - Zo ! - - m

• jr e,0- ol 2

1 (z)). Zo, ~ - , g

1 us ing (20.2), (20.4). Fur ther on [ z - - Zo [ = - - w e have

2

I r (z)) - - log A4 (~, r (z;), logl ~ ( z ) l ~ l o g M Izol-Jr-~- ,

1 and also I P (z) I - - l. Thus w e have for I z - - Zo I = - -

2

Page 46: On Nevanlinna’s second theorem and extensions

ON N E V A N L I N N A ,~ S E C O N D T H E O I ~ E M A N D E X T E N S I O N S 391

(20.7) log cI~ (z--) i - / l o g M (p, ,I) (z), p (z)

! and since (P (z) /p (z) is regular in [ z - - zo I ~ "~-,(20.7) ho ld s a lso in

I z - - Zo I < 1 and in part icular for z : r e 10. Also us ing (20.1), (20.5) we have 2

log I f (z) I - - log I g (z) a) (z) I : log I go (z) �9 (z)/p (z) I

: log leo (z) I + log -(I)(z) " p (z)

N o w Theorem X fol lows, pu t t ing z : r e i0 and us ing (20.3), (20.6) and (20.7).

Exeter, May 1952.

1 ) . - -

2).--

3 ) . - -

4) . - -

5 ) . - -

0) . - -

7 ) - - $) . - -

B I B L I O O R A P H V

L. V. AaLFORS, Zur Theorie der ~)berlagerungsfldchen. Aeta Malh. 65 (1935) 157-I94.

L. V. AHLPOR8, Uber die Anwendung DifferentialKeometrischer Methoden zur Untersuchuns

yon Crberlagerungsfldchen. Acta Soc. Sci. Fenn. Nova Series A, Tom. !!, nr. 6.

A. DmOHAS, Fine Bemerkung zur Ahlforschen Theorie der ~)berlagerungsfldchen. Math. Zeitschr. 44 (1938), 568-572.

W. K. FIAYMAN, Maximum Modulus and Valency of Functions Meromorphic in the Unit Circle. Acta Mat. 86 (1951), 89-257.

J. E. LITTL~WOOD, Lectures on the Theory of Functions, Oxford (1944).

H. MmLoux, Les Fonctions M#omorphes el leur Dgrivds, Paris (1940).

R. NmVAt~r.m~A, Le Th6or~me de Picard.Borel et les Fonctions Mdromorphes, Paris (1929).

R. N~vAsr~tI~sA, Eindeutige Analytische Funkctionen, Berlin (1936).

N O T E S

(') Nevanlinna 8), p. 157. We shall denote this book by E. A. F. in the sequel. (2) E. A. F., p. 158. (a) E . A . F . , p. 232. (4) E. A. F., p. 268. ~5) See Nevanlinna 7) p. 138. (6) Littlewood 5) p. 90. F) See Nevanlinna 7).

Page 47: On Nevanlinna’s second theorem and extensions

302 w. x . ~AYMAN

(s) See Milloux 6). (9) Ahlfors 1). See also E. A. F. Chapter 13, from which the following results are quoted. (i0) E. A. F., p. 332. We write A instead of Ahlfors' S. (li) E. A. F., p. 334. (~2) E. A. F., p. 335. (13) E. A. F., p. 336. (i4) E. A. F., p. 343. (i5) Ahlfors 2). See also Dinghas (3). (i6) If the a~ reduce to 0,I oo, (3.12) with A ~. instead of X follows from Theorem V

p. 174 of Hayman 4). This paper will be denoted by M. M. in the sequel. (i~) See e. g. Littlewood 5), Theorem !!0, p. 114.

(i8) If f (Re i0) ---- 0, I or oo we put ~R (0) : 0. Thus ~R (0) > 0 except possibly for a

finite number of values of 0. (~9) A proof is given in M. M. p. 108. The result follows from the arithmetic-geometric

mean Theorem.

(s0) M.M. lemma 7 p. 109. For n (0), d (0), d o (0) in M. M. read n R (0), d R (0), ~R (0) here.

(:~) See for instance Nevanlinna 7) p. 138. (23) M. M. Theorem VI, p. 131. The difference in the two definitions o f f . (z) does not

affect the proof. (23) E. A. F. p. 343, formula (30). (~4) See e. g. E. A. F. p. 167. 1~5) M. M. p. 195, lemma 1.