Professional Documents
Culture Documents
Aletheia University
G. S. Srivastava†
Department of Mathematics, Jaypee Institute of Information Technology,
A-10, Sector-62, NOIDA,201307, India
and
Ramesh Ganti
Department of Mathematics, National Institute of Technology Silchar,
Assam-788010, India
Abstract
In the present paper, we study the polynomial approximation of
entire functions of two complex variables over Jordan domains by using
Faber polynomials. The coefficient characterizations of order and type
of entire functions have been obtained in terms of the approximation
errors.
1. Introduction
Let Γ1 and Γ2 be given Jordan curves in the complex plane C and Dj , Ej , j =
1,2., be the interior and exterior respectively, of Γj . Let ϕj map Ej conformally
∗
2000 Mathematics Subject Classification. Primary 30B10, 30D15.
†
E-mail: girssfma@iitr.ernet.in
350 G. S. Srivastava and Ramesh Ganti
0
onto {wj : |wj | > 1} such that ϕj (∞) = ∞ and ϕj (∞) > 0. Then by [4], for
sufficiently large |zj |, ϕj (zj ) can be expressed as
z1 c1 c2
w1 = ϕ1 (z1 ) = + c0 + + 2 + ... (1.1)
d1 z1 z1
0 0
z2 0 c c
w2 = ϕ2 (z2 ) = + c0 + 1 + 22 + .... (1.2)
d2 z2 z2
where d1 andd2 > 0. Let us put D = D1 ×D2 and E = E1 ×E2 in C 2 and let the
function ϕ map E conformally onto the unit bidisc U = {|w1 | > 1, |w2 | > 1}
such that ϕ(z1 , z2 ) = ϕ1 (z1 ) ϕ2 (z2 ) satisfies the conditions ϕ(∞, ∞) = ∞ and
0
ϕ (∞, ∞) > 0. Then for sufficiently large values of |z1 |and |z2 |, ϕ(z1 , z2 ) can
be expressed as
∞
z1 z2 X cm,n
w1 w2 = ϕ(z1 , z2 ) = + . (1.3)
d1 d2 m,n=0 z1m z2n
An arbitrary Jordan curve can be approximated from the inside as well
as from the outside by analytic Jordan curves. Since Γ is analytic, ϕ is holo-
morphic on Γ as well. The (m,n) th Faber polynomial Fm,n (z1 , z2 ) of Γ is the
principal part of (ϕ(z1 , z2 ))m+n at (∞, ∞), so that
m n
z1 z2
Fm,n (z1 , z2 ) = + ....
d1 d2
Following Faber [3] for the one variable case, we can easily see that as
m, n → ∞,
where
Approximation of Entire Functions of Two Complex Variables 351
Z Z
1 −(m+1) −(n+1)
am,n = f (ϕ−1 −1
1 (w1 ), ϕ2 (w2 ))w1 w2 dw1 dw2
(2πi)2 |w1 |=r1 |w2 |=r2
and 0 < r1 , r2 < 1 are sufficiently close to 1 so that for j = 1, 2., ϕ−1
j are
holomorphic and univalent in |wj | ≥ rj respectively, the series converging
uniformly on compact subsets of D. Let M (r1 , r2 ) = max |f (z1 , z2 )|, j =
|zj |=rj
1, 2 be the maximum modulus of f (z1 , z2 ). The growth of an entire function
f (z1 , z2 ) is measured in terms of its order ρ and type τ (see [1]) defined as under
ln ln M (r1 , r2 )
lim sup = ρ, (1.7)
r1 ,r2 →∞ ln(r1 r2 )
ln M (r1 , r2 )
lim sup = τ, (1.8)
r1 ,r2 →∞ r1ρ + r2ρ
for 0 < ρ < ∞.
Let Lp (D) denote the set of functions f holomorphic in D and such that
Z Z 1/p
1 p
kf kLp (D) = |f (z1 , z2 )| dx1 dy1 dx2 dy2 <∞
A D
p p
Em,n = Em,n (f ; D) = min kf − πm,n kLp (D)
πm,n
ln mm nn
µ = lim sup . (2.1)
m,n→∞ − ln |am,n |
is finite and then the order ρ of f is equal to µ.
Proof. Let f (z1 , z2 ) = ∞
P
m,n=0 am,n Fm,n (z1 , z2 ) be an entire function of two
complex variables in z1 and z2 , where
Z Z
1 −(m+1) −(n+1)
am,n = f (ϕ−1 −1
1 (w1 ), ϕ2 (w2 ))w1 w2 dw1 dw2
(2πi)2 |w1 |=r1 |w2 |=r2
1 Z Z
−1 −1 −(m+1) −(n+1)
|am,n | = f (ϕ (w ), ϕ (w ))w w dw dw .
2 1 1 2 2 1 2 1 2
(2πi) |w1 |=r1 |w2 |=r2
ϕ1 (z1 ) = w1 ⇒ ϕ−1
1 (w1 ) = z1
ϕ2 (z2 ) = w2 ⇒ ϕ−1
2 (w2 ) = z2 .
Hence
|am,n | ≤ M (r1 , r2 ) r1−m r2−n (2.2)
where M (r1 , r2 ) = max|zt |≤rt |f (z1 , z2 )|, t = 1, 2.
Now first we show that ρ ≥ µ and we assume that µ > 0.Let ε > 0be
chosen such that ε < µ < ∞. Then from (2.1), we have
1
⇒ ln |am,n | ≥ − (m ln m + n ln n)
(µ − ε)
for an infinite sequence of values of m and n. From (2.2), we have
1
≥− (m ln m + n ln n) + m ln r1 + n ln r2
(µ − ε)
1 1
= m ln r1 − ln m + n ln r2 − ln n .
(µ − ε) (µ − ε)
Choosing
1 1
r1 = (em) (µ−ε) , r2 = (en) (µ−ε) .
in the above inequality, we have
m n rµ−ε + r2µ−ε
ln M (r1 , r2 ) ≥ + = 1 .
(µ − ε) (µ − ε) e(µ − ε)
Since µ − ε is independent of r1 and r2 , therefore
ln ln M (r1 , r2 )
ρ = lim sup ≥ µ − ε.
r1 ,r2 →∞ ln(r1 r2 )
and since ε is arbitrary, therefore we have
ρ ≥ µ. (2.3)
Conversely, let
ln(mm nn )
lim sup = σ.
m,n→∞ − ln |am,n |
Suppose σ < ∞. Then for every ε > 0, ∃ X(ε), Y (ε) such that for all
m ≥ X and n ≥ Y , we have
m n
|am,n | ≤ K m− σ+ε n− σ+ε .
354 G. S. Srivastava and Ramesh Ganti
P∞
Since f (z1 , z2 ) = m,n=0 am,n Fm,n (z1 , z2 ), therefore
∞
X m n
|f (z1 , z2 )| ≤ K m− σ+ε n− σ+ε |Fm,n (z1 , z2 )|.
m,n=0
|z1 | |z2 |
|ϕ1 (z1 )| ≤ , and |ϕ2 (z2 )| ≤ .
d1 − ε d2 − ε
By applying these inequalities, for all sufficiently large |z1 | and |z2 |, we
have
∞ m n
X m
− σ+ε n
− σ+ε |z1 | |z2 |
|f (z1 , z2 )| ≤ K m n .
m,n=0
d1 − ε d2 − ε
Hence
M X
N m n
X m
− σ+ε n
− σ+ε r1 r2
M (r1 , r2 ) ≤ m n
m=0 n=0
d1 − ε d2 − ε
∞ ∞ m n
X X m
− σ+ε n
− σ+ε r1 r2
+ m n
m=M +1 n=N +1
d1 − ε d2 − ε
M ∞ m n
X X m
− σ+ε n
− σ+ε r1 r2
+ m n
m=0 n=N +1
d1 − ε d2 − ε
∞ N m n
X X m
− σ+ε n
− σ+ε r1 r2
+ m n .
m=M +1 n=0
d1 − ε d2 − ε
Next we prove
n 1
ρ o m+n
Theorem 2. Let α = lim supm,n→∞ mm nn |admm,n n
|
. If 0 < α < ∞,
1 d2
the function f is the restriction to domain D of an entire function of finite
order ρ and type τ if and only if
α = eτ ρ. (2.5)
ρ ρ
≤ r1−m r2−n e(τ +ε)((d1 +ε)r1 ) e(τ +ε)((d2 +ε)r2 )
for all r1 , r2 sufficiently large. To minimize the right hand side of this inequal-
ity, we select
h i1/ρ h i1/ρ
1 m 1 n
r1 = d1 +ε ρ(τ +ε) , and r2 = d2 +ε ρ(τ +ε) .
Conversely, let
ρ 1/(m+n)
1 m n |am,n |
lim sup m n = σ.
m,n→∞ eρ dm
1 d2
n
Suppose σ < ∞. Then for given ε > 0, ∃ M (ε), N (ε) such that for all
m ≥ M and n ≥ N , we have
∞
X
|f (z1 , z2 )| ≤ K m−m/ρ n−n/ρ dm n
1 d2 [eρ(σ + ε)]
(m+n)/ρ
|Fm,n (z1 , z2 )|.
m,n=0
From (1.5), by using the estimate of Fm,n (z1 , z2 ) in the above inequality,
we have
|f (z1 , z2 )|
∞ m n
X
−m/ρ −n/ρ m n (m+n)/ρ |z1 | |z2 |
≤ K m n d1 d2 [eρ(σ + ε)]
m,n=0
d1 − ε d2 − ε
∞ m n
X
−m/ρ −n/ρ (m+n)/ρ d1 |z1 | d2 |z2 |
≤ K m n [eρ(σ + ε)]
m,n=0
d1 − ε d2 − ε
∞
X
≤ K m−m/ρ n−n/ρ [eρ(σ + ε)](m+n)/ρ r1m r2n .
m,n=0
To estimate the right hand side of the above inequality, we proceeded on the
similar lines of proof of Theorem V of Bose and Sharma [1, p 224], and we
obtain
ρ ρ
|f (z1 , z2 )| ≤ 0{e(σ+ε)(r1 +r2 ) }.
Approximation of Entire Functions of Two Complex Variables 357
Hence
ρ ρ
M (r1 , r2 ) ≤ 0{e(σ+ε)(r1 +r2 ) }.
ln M (r1 , r2 )
⇒ ≤ σ + ε.
r1ρ + r2ρ
On proceeding to limits, we obtain
ln M (r1 , r2 )
lim sup ≤ σ. (2.7)
m,n→∞ r1ρ + r2ρ
From (2.6) and (2.7), we get the required result. This completes the proof
of Theorem 2.
3. Lp - Approximation
In this section we consider the approximations of an entire function over the
domain D. Consider the polynomials
1/2
(m + 1)(n + 1)A1 A2 0 0
pm,n (z1 , z2 ) ∼ ϕ1 (z1 )(ϕ1 (z1 ))m ϕ2 (z2 )(ϕ2 (z2 ))n
π2
(3.1)
as m, n → ∞, uniformly for z1 ∈ E1 and z2 ∈ E2 . Any function f ∈ L2 (D)
can be expanded in terms of these polynomials as
∞ X
X ∞
f (z1 , z2 ) = bm,n pm,n (z1 , z2 ) (3.2)
m=0 n=0
where
358 G. S. Srivastava and Ramesh Ganti
Z Z
1
bm,n = f (z1 , z2 ) pm,n (z1 , z2 ) dx1 dy1 dx2 dy2
A D
∞ ∞
!1/2
X X
2
Em,n = |bk,l |2 . (3.3)
k=m+1 l=n+1
Before going to main results here we state and prove two lemmas which
are more useful in the proof of main theorems. We now prove
Lemma 1.
ln(mm nn ) ln(mm nn )
lim sup = lim sup 2
. (3.4)
m,n→∞ − ln |bm,n | m,n→∞ − ln(Em,n )
2
|bm+1,n+1 | ≤ Em,n ,
2
⇒ − ln |bm+1,n+1 | ≥ − ln(Em,n ).
Proceeding to limits, we have
ln(mm nn ) ln(mm nn )
lim sup 2 )
≤ lim sup . (3.5)
m,n→∞ − ln(Em,n m,n→∞ − ln |bm,n |
Conversely, let
ln(mm nn )
lim sup = σ.
m,n→∞ − ln |bm,n |
Suppose σ < ∞. Then for each ε > 0, ∃ M, N such that for all m ≥ M,
and n ≥ N , we have
m n
|bm,n | ≤ K m− σ+ε n− σ+ε
so that
Approximation of Entire Functions of Two Complex Variables 359
2
(Em,n )2
∞ ∞ ∞ ∞
X X 2k 2l X X 2k 2l
− σ+ε − σ+ε
≤ K k l ≤K (m + 1)− σ+ε (n + 1)− σ+ε
k=m+1 l=n+1 k=m+1 l=n+1
−1 −1
−
2(m+1)
−
2(n+1) 1 1
≤ K(m + 1) σ+ε (n + 1) σ+ε 1− 1−
(m + 1)2/(σ+ε) (n + 1)2/(σ+ε)
2(m+1) 2(n+1)
≤ O(1)K(m + 1)− σ+ε (n + 1)− σ+ε .
Therefore
(m+1) (n+1)
2
Em,n ≤ (m + 1)− σ+ε (n + 1)− σ+ε
2 1
⇒ − ln(Em,n ) ≥ ln((m + 1)m+1 ) ((n + 1)n+1 ).
σ+ε
Proceeding to limits and since ε is arbitrary, therefore we have
ln(mm nn ) ln(mm nn )
σ = lim sup ≥ lim sup 2
. (3.6)
m,n→∞ − ln |bm,n | m,n→∞ − ln(Em,n )
From (3.5) and (3.6), we obtain the required result. This completes the
proof of Lemma 1.
Next we prove
ρ 1/(m+n) ( !ρ )1/(m+n)
2
Em,n
1 m n |bm,n | 1 m n
lim sup m n = lim sup m n .
m,n→∞ eρ dm n
1 d2 m,n→∞ eρ dm n
1 d2
(3.7)
ρ 2 ρ
Em,n
|bm+1,n+1 |
≤
dm
1 d2
n
dm n
1 d2
ρ 1/(m+n) (
2 ρ )1/(m+n)
Em,n
|bm,n |
⇒ mm nn ≤ mm nn
dm n
1 d2 dm n
1 d2
n ρ o1/(m+n) n 2 ρ o1/(m+n)
E
or 1
eρ
mm nn |bdmm,n
d n
|
≤ 1
eρ
mm nn dmm,n
dn
.
1 2 1 2
ρ 1/(m+n) ( !ρ )1/(m+n)
2 )
Em,n
1 m n |bm,n | 1 m n
lim sup m n ≤ lim sup m n .
m,n→∞ eρ dm n
1 d2 m,n→∞ eρ dm n
1 d2
(3.8)
Conversely, let
ρ 1/(m+n)
1 m n |bm,n |
lim sup m n = σ.
m,n→∞ eρ dm n
1 d2
Suppose σ < ∞. Then for each ε > 0, ∃ M (ε), N (ε) such that for all
m ≥ M and n ≥ N , we have
1/ρ m n
|bm,n | ≤ (eρ(σ + ε))m+n m−m n−n
d1 d2
so that
∞
X ∞ n
X o2/ρ
2
(Em,n )2 ≤K (eρ(σ + ε))k+l k −k l−l d2k 2l
1 d2
k=m+1 l=n+1
n o2/ρ
(m+1)+(n+1) −(s+1) 2(m+1) 2(n+1)
≤ K (eρ(σ + ε)) (s + 1) d1 d2
n o2/ρ
2(m+1) 2(n+1)
≤ O(1)K (eρ(σ + ε))(m+1)+(n+1) (s + 1)−(s+1) d1 d2
2/ρ −1
(eρ(σ+ε))2 dρ1
−(s+1) −(m+1) −(n+1)
where (s+1) = (m+1) (n+1) , X1 = 1 − (m+1)(m+1)
,
2/ρ −1
(eρ(σ+ε))2 dρ2
and X2 = 1 − (n+1)(n+1) . Therefore
n o1/ρ
(m+1) (n+1)
2
Em,n ≤ O(1)K (eρ(σ + ε))(m+1)+(n+1) (s + 1)−(s+1) d1 d2 .
(3.9)
From (3.8) and (3.9), we get the required result. This completes the proof
of Lemma 2.
4. Main Results
Now we prove
ln(mm nn )
lim sup p = ρ. (4.1)
m,n→∞ − ln(Em,n )
Proof. We prove the result in two steps. First we consider the case p = 2. Let
us assume that f is an entire function having finite order ρ. Then by Theorem
1, we have
m n
|am,n | ≤ K m− ρ+ε n− ρ+ε .
Now, by considering the property of orthonormality of polynomials pm,n (z1 , z2 ),
we have
∞ ∞ Z Z
X X 1
bm,n = ak,l Fk,l (z1 , z2 )pm,n (z1 , z2 ) dx1 dy1 dx2 dy2 .
k=m+1 l=n+1
A D
362 G. S. Srivastava and Ramesh Ganti
Hence
∞
X ∞
X
|bm,n | ≤ |ak,l | max |Fk,l (z1 , z2 )|.
z1 ,z2 ∈Γ
k=m+1 l=n+1
m n
≤ Km− ρ+ε n− ρ+ε (1 + ε)(m+n)
for all sufficiently large m and n. Therefore, we have
1
− ln |bm,n | ≥ ln(mm nn ).
(ρ + ε)
Proceeding to limits and since ε is arbitrary, we obtain
ln(mm nn )
lim sup ≤ ρ. (4.2)
m,n→∞ − ln |bm,n |
Conversely, let
ln(mm nn )
lim sup = σ.
m,n→∞ − ln |bm,n |
Suppose σ < ∞. Then for each ε > 0, ∃ L(ε), Z(ε) such that for all m > L
and n > Z, we have
m n
|bm,n | ≤ K m− σ+ε n− σ+ε .
From (3.1), we have
0 0
|pm,n (z1 , z2 )| ≤ K (m + 1)1/2 (n + 1)1/2 |ϕ1 (z1 )||ϕ1 (z1 )|m |ϕ2 (z2 )||ϕ2 (z2 )|n
0 0 0
|ϕ1 (z1 )| ≤ K , |ϕ2 (z2 )| ≤ K ”
0
where K , K ” are fixed positive constants, and
|z1 | |z2 |
|ϕ1 (z1 )| ≤ , |ϕ2 (z2 )| ≤
d1 − ε d2 − ε
for all z1 , z2 with sufficiently large modulus. Hence
∞ X
∞ m n
X m
− σ+ε n
− σ+ε 1/2 1/2 |z1 | |z2 |
|f (z1 , z2 )| ≤ K m n (m+1) (n+1)
m=0 n=0
d1 − ε d2 − ε
∞ X
∞ m n
X m
− σ+2ε n
− σ+2ε |z1 | |z2 |
≤K m n .
m=0 n=0
d1 − ε d2 − ε
To estimate the right hand side of above inequality, following the method
used in Theorem 1, we have
ln ln M (r1 , r2 )
ρ = lim sup ≤ σ. (4.3)
r1 ,r2 →∞ ln(r1 r2 )
From (4.2) and (4.3), we have
ln(mm nn )
lim sup = ρ.
m,n→∞ − ln |bm,n |
ln(mm nn )
lim sup 2 )
= ρ. (4.4)
m,n→∞ − ln(Em,n
Now we consider the case for p > 2. Since
364 G. S. Srivastava and Ramesh Ganti
2 p ∞
Em,n ≤ Em,n ≤ Em,n f or2 ≤ p ≤ ∞, (4.5)
it is sufficient to consider the case p = ∞. Suppose f is an entire function of
order ρ. Then
m X
n
X
∞
Em,n ≤ max f (z1 , z2 ) − ak,l Fk,l (z1 , z2 )
z1 ,z2 ∈Γ
k=0 l=0
m
X ∞
X ∞
X n
X
≤ |ak,l | max |Fk,l (z1 , z2 )| + |ak,l | max |Fk,l (z1 , z2 )|
z1 ,z2 ∈Γ z1 ,z2 ∈Γ
k=0 l=n+1 k=m+1 l=0
∞
X ∞
X
+ |ak,l | max |Fk,l (z1 , z2 )| . (4.6)
z1 ,z2 ∈Γ
k=m+1 l=n+1
The first two summations in the above inequality (4.6) are bounded. It is
sufficient to estimate the last summation. Since f is an entire function of finite
order ρ, therefore bym Theorem
n
1, we have
− ρ+ε − ρ+ε
|am,n | ≤ K m n and max |Fk,l (z1 , z2 )| ≤ (1 + ε)k+l .
z1 ,z2 ∈Γ
Therefore the above inequality (4.6) becomes,
∞ ∞
X X k l
∞
Em,n ≤ k − ρ+ε l− ρ+ε (1 + ε)k+l
k=m+1 l=n+1
∞ ∞ k/(ρ+ε) l/(ρ+ε)
X X (1 + ε)ρ+ε (1 + ε)ρ+ε
≤K
k=m+1 l=n+1
m+1 n+1
m/(ρ+ε) n/(ρ+ε)
(1 + ε)ρ+ε (1 + ε)ρ+ε
≤K
m n
ln(mm nn ) ln(mm nn )
⇒ ∞ )
≤ .
− ln(Em,n [1/(ρ + ε)] ln(mm nn ) − ln K − (m + n) ln(1 + ε)
ln(mm nn )
lim sup ∞ )
≤ ρ.
m,n→∞ − ln(Em,n
Approximation of Entire Functions of Two Complex Variables 365
In view of inequalities(4.5) and the fact that (4.1) holds for p = 2, this
last inequality actually is an equality. Finally assuming (4.1) with p = ∞, we
deduce from (4.5), that (4.1) will hold for p = 2 and hence that f is of order
ρ. This completes the proof of Theorem 3.
Proof. We prove the theorem in two steps. First we consider the case p = 2.
Let us assume that f is an entire function having finite order ρ and finite type
τ . Then by Theorem 2, we have
∞
X ∞
X
|bm,n | ≤ K k −(k/ρ) l−(l/ρ) dk1 dl2 (eρ(τ + ε))(k+l)/ρ (1 + ε)(k+l)
k=m+1 l=n+1
≤ Km−(m/ρ) n−(n/ρ) dm n
1 d2 (eρ(τ + ε))
(m+n)/ρ
(1 + ε)(m+n)
for all sufficiently large m and n. Therefore, we have
mm nn |bm,n |ρ ≤ K(dm n ρ
1 d2 ) (eρ(τ + ε))
(m+n)
.
By applying limits, we have
1
ρ m+n
m n |bm,n |
lim sup m n ≤ eρτ. (4.8)
m,n→∞ dm n
1 d2
Conversely ,let
1
ρ m+n
1 m n |bm,n |
lim sup m n = σ.
m,n→∞ eρ dm n
1 d2
366 G. S. Srivastava and Ramesh Ganti
Suppose σ < ∞. Then for each ε > 0, ∃ H(ε), G(ε) such that for all
m > H and n > G, we have
|f (z1 , z2 )|
∞ X ∞ m n
X
−(g/ρ) m n (m+n)/ρ 1/2 |z1 | |z2 |
≤ L g d1 d2 (eρ(σ + ε)) (s + 1)
m=0 n=0
d1 − ε d2 − ε
∞ X ∞ m n
X d1 |z1 | d2 |z2 |
≤ L g −(g/ρ) (eρ(σ + 2ε))(m+n)/ρ
m=0 n=0
d1 − ε d2 − ε
∞ X
X ∞
≤ L g −(g/ρ) (eρ(σ + 2ε))(m+n)/ρ r1m r2n
m=0 n=0
ln M (r1 , r2 )
τ = lim sup ≤ σ. (4.9)
r1 ,r2 →∞ r1ρ + r2ρ
From (4.8) and (4.9), we have
1
ρ m+n
m n |bm,n |
lim sup m n = eρτ.
m,n→∞ dm n
1 d2
Now we consider the case for p > 2. From (4.5), it is sufficient to consider
the case p = ∞. Suppose f is an entire function having finite order ρ and of
type τ . Then from (4.6), the first two summations of the above inequality are
bounded. It is sufficient to estimate the last summation. Since f is an entire
function of finite type τ , therefore by Theorem 2, we have
∞
X ∞
X
∞
Em,n ≤K k −(k/ρ) l−(l/ρ) dk1 dl2 (eρ(τ + ε))(k+l)/ρ (1 + ε)k+l
k=m+1 l=n+1
∞ ∞ k/ρ l/ρ
X X (1 + ε)ρ (1 + ε)ρ
≤K dk1 dl2 (eρ(τ + ε))k+l
k=m+1 l=n+1
m + 1 n + 1
m/ρ n/ρ
(1 + ε)ρ (1 + ε)ρ
≤K dm n
1 d2 (eρ(τ + ε))
m+n
m+1 n+1
∞ ρ 1/(m+n)
Em,n
m n
⇒ m n ≤ (1 + ε) (eρ(τ + ε))
dm n
1 d2
n 1
∞ ρ o m+n
m n Em,n
Hence lim sup m n dm dn ≤ eρτ.
m,n→∞ 1 2
In view of inequalities(4.5) and the fact that (4.7) holds for p = 2, this
last inequality actually is an equality. Finally assuming (4.7) with p = ∞, we
deduce from (4.5), that (4.7) will hold for p = 2 and hence that f is of type
τ . This completes the proof of Theorem 4.
368 G. S. Srivastava and Ramesh Ganti
References
[1] S. K. Bose and D. Sharma, Integral functions of two complex variables,
Compositio Math. 15 (1963), 210-216.