DOI: 10.24193/subbmath.2018.4.06
Sufficient conditions of boundedness of L-index and analog of Hayman’s Theorem for analytic functions in a ball
Andriy Bandura and Oleh Skaskiv
Abstract. We generalize some criteria of boundedness ofL-index in joint variables for analytic in an unit ball functions. Our propositions give an estimate maximum modulus of the analytic function on a skeleton in polydisc with the larger radii by maximum modulus on a skeleton in the polydisc with the lesser radii. An analog of Hayman’s Theorem for the functions is obtained. Also we established a connection between class of analytic in ball functions of boundedlj-index in every direction 1j, j ∈ {1, . . . , n} and class of analytic in ball of functions of boundedL-index in joint variables, whereL(z) = (l1(z), . . . , ln(z)), lj:Bn→R+ is continuous function,1j= (0, . . . ,0, 1
|{z}
j−th place
,0, . . . ,0)∈Rn+, z∈Cn.
Mathematics Subject Classification (2010):32A05, 32A10, 32A30, 32A40, 30H99.
Keywords:Analytic function, unit ball, boundedL-index in joint variables, max- imum modulus, partial derivative, boundedL-index in direction.
1. Introduction
Recently, there was introduced a concept of analytic function in a ball in Cn of bounded L-index in joint variables [8]. We also obtained criterion of boundedness of L-index in joint variables which describes a local behavior of partial derivatives on a skeleton in the polydisc and established other important properties of analytic functions in a ball of bounded L-index in joint variables. Those investigations used an idea of exhaustion of a ball inCn by polydiscs.
The presented paper is a continuation of our investigations from [8]. We set the goal to prove new analogues of criteria of boundedness ofL-index in joint variables for analytic in a ball functions. Particular, we prove an estimate of maximum modulus on a greater polydisc by maximum modulus on a lesser polydisc (Theorems 3.1, 3.2) and obtain an analog of Hayman’s Theorem for analytic functions in a ball of bounded
L-index in joint variables (Theorems 4.1 and 4.2). For entire functions similar propo- sitions were obtained by A. I. Bandura, M. T. Bordulyak, O. B. Skaskiv [4, 5] in a case L(z) = (l1(z), . . . , ln(z)), z∈Cn.Also A. I. Bandura, N.V. Petrechko, O. B. Skaskiv [6, 7] deduced same results for analytic in a polydisc functions. Hayman’s Theorem and its generalizations for different classes of analytic functions [1, 3, 5, 7, 12, 15, 20, 21]
are very important in theory of functions of bounded index. The criterion is helpful [1, 9] to investigate boundedness of index of entire solutions of ordinary or partial differential equations.
Note that the corresponding theorems for entire functions of bounded l-index and of boundedL-index in direction were also applied to investigate infinite products (see bibliography in [21, 1]). Thus, those generalizations for analytic in a ball functions are necessary to study L-index in joint variables of analytic solutions of PDE’s, its systems and multidimensional counterparts of Blaschke products. At the end of the paper, we present a scheme of application of Hayman’s Theorem to study properties of analytic solutions in the unit ball.
2. Main definitions and notations
We need some standard notations. Denote R+= (0,+∞),0= (0, . . . ,0), 1= (1, . . . ,1), 1j= (0, . . . ,0, 1
|{z}
j−th place
,0, . . . ,0)∈Rn+,
R= (r1, . . . , rn)∈Rn+, z= (z1, . . . , zn)∈Cn,|z|=q Pn
j=1|zj|2.
ForA= (a1, . . . , an)∈Rn, B= (b1, . . . , bn)∈Rn we will use formal notations without violation of the existence of these expressions
AB= (a1b1,· · · , anbn), A/B= (a1/b1, . . . , an/bn), AB =ab11ab22·. . .·abnn, kAk=a1+· · ·+an,
and the notation A < B means that aj < bj, j ∈ {1, . . . , n}; the relation A ≤ B is defined similarly. For K = (k1, . . . , kn) ∈ Zn+ denote K! = k1!·. . . ·kn!. The polydisc{z∈Cn : |zj−zj0|< rj, j= 1, . . . , n}is denoted by Dn(z0, R),its skeleton {z ∈ Cn : |zj −z0j| = rj, j = 1, . . . , n} is denoted by Tn(z0, R), and the closed polydisc {z ∈ Cn : |zj −z0j| ≤ rj, j = 1, . . . , n} is denoted by Dn[z0, R]. The open ball{z ∈Cn : |z−z0|< r} is denoted by Bn(z0, r),its boundary is a sphere Sn(z0, r) ={z∈Cn : |z−z0|=r},the closed ball{z∈Cn : |z−z0| ≤r}is denoted byBn[z0, r],Bn =Bn(0,1),D=B1={z∈C: |z|<1}.
For K = (k1, . . . , kn) ∈ Zn+ and the partial derivatives of an analytic in Bn functionF(z) =F(z1, . . . , zn) we use the notation
F(K)(z) =∂kKkF
∂zK = ∂k1+···+knF
∂zk11. . . ∂znkn
.
Let L(z) = (l1(z), . . . , ln(z)), where lj(z) : Bn → R+ is a continuous function such that
(∀z∈Bn) : lj(z)> β/(1− |z|), j∈ {1, . . . , n}, (2.1)
whereβ >√
nis a some constant. For a polydisc A.I. Bandura, N.V. Petrechko and O.B. Skaskiv [6, 7] imposed the restriction (∀z ∈ Dn(0,1)) : lj(z) > β/(1− |zj|), j∈ {1, . . . , n}.A similar condition is used in one-dimensional case by S.N. Strochyk, M.M. Sheremeta, V.O. Kushnir [22, 14, 21].
Note that if R ∈Rn+, |R| ≤ β, z0 ∈Bn and z ∈ Dn[z0, R/L(z0)] then z ∈ Bn (see Remark 1 in [8]).
An analytic function F: Bn → C is said to be of bounded L-index (in joint variables),if there exists n0∈Z+such that for allz∈Bn and for allJ ∈Zn+
|F(J)(z)|
J!LJ(z) ≤max
|F(K)(z)|
K!LK(z) : K∈Zn+, kKk ≤n0
. (2.2)
The least such integer n0 is called the L-index in joint variables of the function F and is denoted byN(F,L,Bn) (see [8]). Entire and analytic in polydisc functions of boundedL-index in joint variables are considered in [4, 5, 6, 7, 10, 13, 19, 18, 16, 17].
ByQ(Bn) we denote the class of functionsL, which satisfy (2.1) and the following condition
(∀R∈Rn+,|R| ≤β, j∈ {1, . . . , n}) : 0< λ1,j(R)≤λ2,j(R)<∞, (2.3) where λ1,j(R) = inf
z0∈Bninf
lj(z)/lj(z0) :z∈Dn
z0, R/L(z0) , λ2,j(R) = sup
z0∈Bn
sup
lj(z)/lj(z0) :z∈Dn
z0, R/L(z0) . Λ1(R) = (λ1,1(R), . . . , λ1,n(R)), Λ2(R) = (λ2,1(R), . . . , λ2,n(R)).
We need the following results.
Theorem 2.1 ([8]). LetL∈Q(Bn). An analytic inBn functionF has boundedL-index in joint variables if and only if for eachR∈Rn+,|R| ≤β,there existn0∈Z+,p0>0 such that for everyz0∈Bn there existsK0∈Zn+,kK0k ≤n0, and
max
|F(K)(z)|
K!LK(z):kKk ≤n0, z∈Dn
z0, R/L(z0)
≤p0
|F(K0)(z0)|
K0!LK0(z0). (2.4) DenoteL(z) = (ee l1(z), . . . ,eln(z)). The notationLLe means that there exist
Θ1= (θ1,j, . . . , θ1,n)∈Rn+, Θ2= (θ2,j, . . . , θ2,n)∈Rn+
such that∀z∈Bn θ1,jelj(z)≤lj(z)≤θ2,jelj(z) for eachj∈ {1, . . . , n}.
Theorem 2.2 ([8]). LetL∈Q(Bn),LL, β|Θe 1|>√
n.An analytic inBn functionF has bounded L-index in joint variables if and only ife F has boundedL-index in joint variables.
3. Local behaviour of maximum modulus of analytic in ball function
For an analytic inBn functionF we put
M(r, z0, F) = max{|F(z)|:z∈Tn(z0, r)},
where z0 ∈ Bn, r ∈Rn+. Then M(R, z0, F) = max{|F(z)|: z ∈ Dn[z0, R]}, because the maximum modulus for an analytic function in a closed polydisc is attained on its skeleton.
The following proposition uses an idea about the possibility of replacing universal quantifier by existential quantifier in sufficient conditions of index boundedness [2]. To prove an analog of Hayman’s Theorem we need this theorem which has an independent interest.
Theorem 3.1. Let L ∈ Qn, F : Bn → C be analytic function. If there exist R0, R00∈Rn+, R0< R00,|R00|< β andp1=p1(R0, R00)≥1such that for every z0∈Cn
M R00
L(z0), z0, F
≤p1M R0
L(z0), z0, F
(3.1) thenF is of boundedL-index in joint variables.
Proof. At first, we assume that0< R0 <1< R00.
Letz0∈Bn be an arbitrary point. We expand a functionF in power series F(z) = X
K≥0
bK(z−z0)K = X
k1,...,kn≥0
bk1,...,kn(z1−z01)k1. . .(zn−zn0)kn, (3.2)
wherebK=bk1,...,kn= F(K)K!(z0).
Let µ(R, z0, F) = max{|bK|RK: K ≥ 0} be a maximal term of power series (3.2) and
ν(R) =ν(R, z0, F) = (ν10(R), . . . , νn0(R)) be a set of indices such that
µ(R, z0, F) =|bν(R)|Rν(R),
kν(R)k=
n
X
j=1
νj(R) = max{kKk:K≥0, |bK|RK =µ(R, z0, F)}.
In view of inequality (3.8) we obtain for any|R|<1− |z0|, µ(R, z0, F)≤M(R, z0, F).
Then for givenR0 andR00 with 0<|R0|<1<|R00|< β we conclude M(R0R, z0, F)≤X
k≥0
|bk|(R0R)k ≤X
k≥0
µ(R, z0, F)(R0)k
=µ(R, z0, F)X
k≥0
(R0)k =
n
Y
j=1
1
1−r0jµ(R, z0, F).
Besides,
lnµ(R, z0, F) = ln{|bν(R)|Rν(R)}= ln
|bν(R)|(RR00)ν(R) 1 (R00)ν(R)
= ln{|bν(R)|(RR00)ν(R)}+ ln 1
(R00)ν(R)
≤lnµ(R00R, z0, F)− kν(R)kln min
1≤j≤nrj00. This implies that
kν(R)k ≤ 1
ln min1≤j≤nrj00(lnµ(R00R, z0, F)−lnµ(R, z0, F))
≤ 1
ln min1≤j≤nr00j
lnM(R00R, z0, F)−ln(
n
Y
j=1
(1−r0j)M(R0R, z0, F))
≤ 1
ln min1≤j≤nr00j lnM(R00R, z0, F)−lnM(R0R, z0, F)
− Pn
j=1ln(1−rj0) min1≤j≤nrj00
= 1
min1≤j≤nr00j lnM(R00R, z0, F) M(R0R, z0, F) −
Pn
j=1ln(1−rj)
min1≤j≤nr00j . (3.3) PutR= L(z10).Now letN(F, z0,L) be theL-index of the functionFin joint variables at point z0 i. e. it is the least integer for which inequality (2.2) holds at point z0. Clearly that
N(F, z0,L)≤ν 1
L(z0), z0, F
=ν(R, z0, F). (3.4)
But
M R00/L(z0), z0, F
≤p1(R0, R00)M R0/L(z0), z0, F
. (3.5)
Therefore, from (3.3), (3.4), (3.5) we obtain that∀z0∈Bn N(F, z0,L)≤−P2
j=1ln(1−r0j)
ln min{r001, r200} + lnp1(R0, R00) ln min{r001, r200}.
This means that F has bounded L-index in joint variables, if 0 < R0 < 1 < R00,
|R00|< β.
Now we will prove the theorem for any 0 < R0 < R00, |R00| < β. From (3.1) with 0< R1< R2 it follows that
max
|F(z)|:z∈Tn
z0, 2R00 R0+R00
R0+R00 2L(z0)
≤P1max
|F(z)|:z∈Tn
z0, 2R0 R0+R00
R0+R00 2L(z0)
.
DenotingL(z) =e R2L(z)0+R00,we obtain max
(
|F(z)|:z∈Tn z0, 2R00 (R0+R00)eL(z0)
!)
≤P1max (
|F(z)|:z∈Tn z0, 2R00 (R0+R00)eL(z0)
!) ,
where 0< R02R+R000 <1< R2R0+R0000.Taking into account the first part of the proof, we conclude that the function F has bounded L-index in joint variables. By Theoreme 2.2, the functionF is of boundedL-index in joint variables.
Also the corresponding necessary conditions are valid.
Theorem 3.2. LetL∈Q(Bn).If an analytic in Bn functionF has boundedL-index in joint variables then for anyR0, R00∈Rn+, R0< R00,|R00| ≤β,there exists a number p1=p1(R0, R00)≥1 such that for everyz0∈Bn inequality (3.1)holds.
Proof. LetN(F,L) =N <+∞.Suppose that inequality (3.1) does not hold i.e. there existR0, R00,0<|R0|<|R00|< β,such that for eachp∗≥1 and for somez0=z0(p∗)
M R00
L(z0), z0, F
> p∗M R0
L(z0), z0, F
. (3.6)
By Theorem 2.1, there exists a number p0=p0(R00)≥1 such that for everyz0∈Bn and some K0 ∈Zn+,kK0k ≤N, (i.e.n0 =N, see proof of necessity of Theorem 2.1 in [8]) one has
M R00
L(z0), z0, F(K0)
≤p0|F(K0)(z0)|. (3.7) We put
b1=p0
n
Y
j=2
λN2,j(R00)
(N!)n−1
N
X
j=1
(N−j)!
(r100)j
r100r200. . . rn00 r01r20. . . r0n
N ,
b2=p0
n Y
j=3
λN2,j(R00)
(N!)n−2 N
X
j=1
(N−j)!
(r002)j
r002. . . rn00 r02. . . r0n
N 1, 1
(r10)N
, . . .
bn−1=p0λN2,n(R0)N!
N
X
j=1
(N−j)!
(r00n−1)j
rn−100 r00n rn−10 r0n
N
max
1, 1
(r10. . . rn−20 )N
,
bn=p0
N
X
j=1
(N−j)!
(rn00)j
rn00
rn0 N
max
1, 1
(r01. . . r0n−1)N
and
p∗= (N!)np0
r100r002. . . rn00 r01r02. . . r0n
N
+
n
X
k=1
bk+ 1.
Letz0=z0(p∗) be a point for which inequality (3.6) holds andK0be such that (3.7) holds and
M R0
L(z0), z0, F
=|F(z∗)|, M r00
L(z0), z0, F(J)
=|F(J)(z∗J)|
for everyJ ∈Zn+,kJk ≤N.We apply Cauchy’s inequality
|F(J)(z0)| ≤J!
L(z0) R0
J
|F(z∗)| (3.8)
for estimate the difference
|F(J)(zJ,1∗ , zJ,2∗ , . . . , zJ,n∗ )−F(J)(z10, zJ,2∗ , . . . , zJ,n∗ )|
=
Z zJ,1∗ z01
∂kJk+1F
∂z1j1+1∂z2j2. . . ∂znjn
(ξ, zJ,2∗ , . . . , zJ,n∗ )dξ
≤
∂kJk+1F
∂zj11+1∂zj22. . . ∂znjn
(z(j∗
1+1,j2,...,jn))
r100
l1(z0). (3.9) Taking into account (z10, zJ,2∗ , . . . , z∗J,n)∈Dn[z0,L(zR000)],for allk∈ {1, . . . , n},
|z∗J,k−zk0|= rk00
lk(z0), lk(z10, zJ,2∗ , . . . , z∗J,n)≤λ2,k(R00)lk(z0) and (3.8) withJ =K0,by Theorem 2.1 we have
|F(J)(z01, z∗J,2, . . . , zJ,n∗ )|
≤J!lj11(z10, zJ,2∗ , . . . , zJ,n∗ )Qn
k=2ljkk(z10, zJ,2∗ , . . . , z∗J,n)
K0!LK0(z0) p0|F(K0)(z0)|
≤ J!LJ(z0)Qn
k=2λj2,kk (R00) K0!LK0(z0) p0K0!
L(z0) R0
K0
|F(z∗)|
= p0J!LJ(z0)Qn
k=2λj2,kk (R00)
(R0)K0 |F(z∗)|. (3.10)
From inequalities (3.9) and (3.10) it follows that
∂kJk+1F
∂z1j1+1∂z2j2. . . ∂zjnn
(z(j∗
1+1,j2,...,jn))
≥ l1(z0) r100
n|F(J)(zj∗)| − |F(J)(z10, zJ,2∗ , . . . , zJ,n∗ )|o
≥ l1(z10)
r100 |F(J)(z∗j)| −p0J!L(j1+1,j2,...,jn)(z0)Qn
k=2λj2,kk (R00)
r001(R0)K0 |F(z∗)|.
Then
|F(K0)(zK∗0)| ≥l1(z0) r100
∂kK0k−1f
∂z1k10−1∂z2k02. . . ∂znk0n
(z∗(k0
1−1,k02,...,k0n))
−p0(k10−1)!k20!. . . k0n!LK0(z0)Qn i=2λk
0 i
2,i(R00) r100(R0)K0 |F(z∗)|
≥ l12(z0) (r001)2
∂kK0k−2f
∂z1k01−2∂z2k02. . . ∂znk0n
(z∗(k0
1−2,k02,...,k0n))
−p0(k10−2)!k20!. . . k0n!LK0(z0)Qn
i=2λk2,i0i(R00) (r001)2(R0)K0 |F(z∗)|
−p0(k01−1)!k20!. . . k0n!LK0(z0)Qn i=2λk
0 i
2,i(r00i) r100(R0)K0 |F(z∗)|
. . .
≥ l1k01(z0) (r001)k01
∂kK0k−k01f
∂zk202. . . ∂zkn0n
(z(0,k∗ 0 2,...,k0n))
− p0
(R0)K0LK0(z0)
n
Y
i=2
λk2,i0i(R00)
!
k20!. . . k0n!
k01
X
j1=1
(k10−j1)!
(r100)j1 |F(z∗)|. . .
≥lk
0 1
1 (z0) (r100)k10
lk
0 2
2 (z0) (r002)k02
∂kK0k−k01−k02f
∂zk303. . . ∂zkn0n
(z(0,0,k∗ 0 3,...,kn0))
−lk101(z0)p0L(0,k20,...,k0n)(z0) (r001)k01(R0)K0
n
Y
i=3
λk
0 i
2,i(R00)
!
k30!. . . k0n!
k02
X
i2=1
(k02−j2)!
(r200)j2 |F(z∗)|
− p0
(R0)K0LK0(z0)
n
Y
i=2
λk2,i0i(R00)
!
k20!. . . k0n!
k01
X
j1=1
(k10−j1)!
(r100)j1 |F(z∗)|
. . .
≥
L(z0) R00
|F(z∗0)| − |F(z∗)|
b
X
i=1
˜bi, (3.11)
where in view of the inequalitiesλ2,i(R00)≥1 andR00≥R0 we have
˜b1= p0
(R0)K0LK0(z0)
n
Y
i=2
λk
0 i
2,i(R00)
!
k02!. . . kn0!
k10
X
j1=1
(k01−j1)!
(r001)j1
=
L(z0) R00
K0R00 R0
K0 p0
n
Y
i=2
λk2,i0i(R00)
!
k02!. . . kn0!
k01
X
j1=1
(k01−j1)!
(r001)j1 ≤
L(z0) R00
K0 b1,
˜b2= p0
(R0)K0LK0(z0) n
Y
i=3
λk2,ii0(R00)
k03!. . . kn0! (r001)k01
k02
X
j2=1
(k20−j2)!
(r200)j2 ≤ L(z0)
R00 K0
b2, . . .
˜bn−1= p0
(R0)K0LK0(z0)λk2,nn0 (R00) k0n!
(r001)k01. . .(r00n−2)kn−20 ×
×
k0n−1
X
jn−1=1
(kn−10 −jn−1)!
(r00n−1)jn−1 ≤
L(z0) R00
K0 bn−1,
˜bn= p0
(R0)K0LK0(z0) 1
(r100)k10. . .(rn−100 )k0n−1
k0n
X
jn=1
(k0n−jn)!
(r00n)jn ≤
L(z0) R00
K0 bn. Thus, (3.11) implies that
|F(K0)(zK∗0)| ≥
L(z0) R00
K0
|F(z∗)|
|F(z∗0)|
|F(z∗)|−
n
X
j=1
bj
. But in view of (3.6) and a choice ofp∗ we have
|F(z0∗)|
|F(z∗)| ≥p∗>
n
X
j=1
bj. Thus, (3.7) and (3.8) imply
|F(K0)(z∗K0)| ≥
L(z0) R00
K0
|F(z∗)|
p∗−
n
X
j=1
bj
≥
L(z0) R00
K0
p∗−
n
X
j=1
bj
|F(K0)(z0)|(R0)K0 K0!LK0(z0)
≥
r01. . . r0n r100. . . rn00
N
p∗−
n
X
j=1
bj
|F(K0)(z∗K0)|
p0(n!)n . Hence, we havep∗≤p0
r01...r0n r001...r00n
N
(N!)n+Pn
j=1bj,but this contradicts the choice of
p∗.
4. Analogue of Theorem of Hayman for analytic in a ball function of bounded L-index in joint variables
Theorem 4.1. Let L∈Q(Bn). An analytic functionF in Bn has bounded L-index in joint variables if and only if there existp∈Z+ andc∈R+ such that for eachz∈Bn
max
|F(J)(z)|
LJ(z) : kJk=p+ 1
≤c·max
|F(K)(z)|
LK(z) : kKk ≤p
. (4.1)
Proof. LetN=N(F,L,Bn)<+∞.The definition of the boundedness ofL-index in joint variables yields the necessity withp=N andc= ((N+ 1)!)n.
We prove the sufficiency. ForF ≡0 theorem is obvious. Thus, we suppose that F 6≡0.Denote β= (√βn, . . . ,√βn).
Assume that (4.1) holds,z0∈Bn, z∈Dn[z0,L(zβ0)].For allJ ∈Zn+,kJk ≤p+ 1, one has
|F(J)(z)|
LJ(z0) ≤ΛJ2(β)|F(J)(z)|
LJ(z) ≤c·ΛJ2(β) max
|F(K)(z)|
LK(z) : kKk ≤p
≤c·ΛJ2(β) max
Λ−K1 (2)|F(K)(z)|
LK(z0) :kKk ≤p
≤BG(z), (4.2)
whereB=c·max{ΛK2 (β) : kKk=p+ 1}max{Λ−K1 (β) : kKk ≤p},and G(z) = max
|F(K)(z)|
LK(z0) : kKk ≤p
. We choose
z(1)= (z1(1), . . . , zn(1))∈Tn(z0, 1 2β√
nL(z0)) and
z(2)= (z1(2), . . . , zn(2))∈Tn(z0, β L(z0)) such thatF(z(1))6= 0 and
|F(z(2))|=M β
L(z0), z0, F
6= 0. (4.3)
These points exist, otherwise ifF(z)≡0 on skeleton Tn
z0, 1 2β√
nL(z0)
or Tn
z0, β L(z0)
then by the uniqueness theorem F ≡0 in Bn. We connect the points z(1) and z(2) with plane
α=
z2=k2z1+c2, z3=k3z1+c3, . . . zn=knz1+cn, where
ki= zi(2)−zi(1) z1(2)−z1(1)
, ci =zi(1)z(2)1 −z(2)i z1(1) z(2)1 −z(1)1
, i= 2, . . . , n.
It is easy to check thatz(1) ∈αandz(2)∈α.LetG(ze 1) =G(z)|α be a restriction of the functionGontoα.
For every K ∈ Zn+ the function F(K)(z)
α is analytic function of variable z1
and ˜G(z(1)1 ) = G(z(1))
α 6= 0 becauseF(z(1))6= 0. Hence, all zeros of the function F(K)(z)
αare isolated as zeros of a function of one variable. Thus, zeros of the function G(z˜ 1) are isolated too. Therefore, we can choose piecewise analytic curveγontoαas following
z=z(t) = (z1(t), k2z1(t) +c2, . . . , knz1(t) +cn), t∈[0,1],
which connect the pointsz(1), z(2) and such thatG(z(t))6= 0 and Z 1
0
|z10(t)|dt≤ 2β
√nl1(z01).
For a construction of the curve we connectz1(1) andz1(2) by a line z∗1(t) = (z1(2)−z1(1))t+z1(1), t∈[0,1].
The curveγcan cross pointsz1at which the functionG(ze 1) = 0.The number of such pointsm=m(z(1), z(2)) is finite. Let (z1,k∗ ) be a sequence of these points in ascending order of the value |z(1)1 −z∗1,k|, k∈ {1,2, ..., m}. We choose
r < min
1≤k≤m−1{|z1,k∗ −z1,k+1∗ |,|z1,1∗ −z(1)1 |,|z1,m∗ −z1(2)|, 2β2−1 2π√
nβl1(z0)}.
Now we construct circles with centers at the points z∗1,k and corresponding radii r0k< 2rk such thatG(ze 1)6= 0 for allz1 on the circles. It is possible, becauseF 6≡0.
Every such circle is divided onto two semicircles by the linez∗1(t). The required piecewise-analytic curve consists with arcs of the constructed semicircles and segments of linez1∗(t), which connect the arcs in series between themselves or with the points z1(1), z1(2). The length ofz1(t) in C(but notz(t) inCn!) is lesser than
β/√ n
l1(z0) + 1 2√
nβl1(z0)+πr≤ 2β
√nl1(z0). Then
Z 1 0
|zs0(t)|dt=|ks| Z 1
0
|z10(t)|dt≤ |zs(2)−zs(1)|
|z1(2)−z1(1)|
√ 2β nl1(z0)
≤ 2β2+ 1 2√
nβls(z0) 2√
nβl1(z0) 2β2−1
√ 2β
nl1(z0) ≤ 2β(2β2+ 1) (2β2−1)√
nls(z0), s∈ {2, . . . , n}.
Hence,
Z 1 0
n
X
s=1
ls(z0)|zs0(t)|dt≤ 2β(2β2+ 1)√ n
2β2−1 =S. (4.4)
Since the functionz=z(t) is piece-wise analytic on [0,1], then for arbitraryK∈Zn+, J ∈Zn+,kKk ≤p,either
|F(K)(z(t))|
LK(z0) ≡ |F(J)(z(t))|
LJ(z0) , (4.5)
or the equality
|F(K)(z(t))|
LK(z0) = |F(J)(z(t))|
LJ(z0) (4.6)
holds only for a finite set of pointstk ∈[0; 1].
Then for function G(z(t)) as maximum of such expressions |F(J)LJ(z(z(t))|0) by all kJk ≤ptwo cases are possible:
1. In some interval of analyticity of the curve γ the function G(z(t)) identically equals simultaneously to some derivatives, that is (4.5) holds. It means that G(z(t)) ≡ |F(J)LJ(z(z(t))|0) for some J, kJk ≤ p. Clearly, the function F(J)(z(t)) is analytic. Then |F(J)(z(t))| is continuously differentiable function on the in- terval of analyticity except points where this partial derivative equals zero
|F(j1,j2)(z1(t), z2(t))| = 0. However, there are not the points, because in the opposite caseG(z(t)) = 0.But it contradicts the construction of the curveγ.
2. In some interval of analyticity of the curve γ the function G(z(t)) equals si- multaneously to some derivatives at a finite number of points tk, that is (4.6) holds. Then the pointstk divide interval of analyticity onto a finite number of segments, in which of them G(z(t)) equals to one from the partial derivatives, i. e. G(z(t)) ≡ |FL(J)J(z(t))|(z0) for some J, kJk ≤ p. As above, in each from these segments the functions|F(J)(z(t))|, andG(z(t)) are continuously differentiable except the pointstk.
The inequality
d
dt|f(t)| ≤
df(t) dt
holds for complex-valued functions of real argument outside a countable set of points.
In view of this fact and (4.2) we have d
dtG(z(t))≤maxn 1 LJ(z0)
d
dtF(J)(z(t))
: kJk ≤po
≤maxnXn
s=1
∂kJk+1F
∂z1j1. . . ∂zsjs+1. . . ∂zjnn
(z(t))
|zs0(t)|
Lj(z0) : kJk ≤po
≤maxnXn
s=1
∂kJk+1F
∂zj11. . . ∂zsjs+1. . . ∂znjn
(z(t))
ls(z0)|zs0(t)|
l1j1(z0). . . ljs1+1(z0). . . ljnn(z0) : kJk ≤po
≤Xn
s=1
ls(z0)|zs0(t)|
maxn|F(j)(z(t))|
LJ(z0) : kJk ≤p+ 1o
≤Xn
s=1
ls(z0)|z0s(t)|
BG(z(t)).
Therefore, (4.4) yields
lnG(z(2)) G(z(1)) =
Z 1 0
1 G(z(t))
d
dtG(z(t))dt ≤B
Z 1 0
n
X
s=1
ls(z0)|zs0(t)|dt≤S·B.
Using (4.3), we deduce M
β
L(z0), z0, F
≤G(z(2))≤G(z(1))eSB.
Sincez(1)∈Tn(z0,2β√nL(z1 0)),the Cauchy inequality holds
|F(J)(z(1))|
LJ(z0) ≤J!(2β√ n)kJkM
1 2β√
nL(z0), z0, F
. for allJ ∈Zn+.Therefore, forkJk ≤pwe obtain
G(z(1))≤(p!)n(2β√ n)pM
1 2β√
nL(z0), z0, F
,
M β
L(z0), z0, F
≤eSB(p!)n(2β√ n)pM
1 2β√
nL(z0), z0, F
.
Hence, by Theorem 3.1 the functionF has boundedL-index in joint variables.
The following result was also obtained for other classes of holomorphic functions in [21, 11, 7].
Theorem 4.2. Let L∈Q(Bn). An analytic functionF in Bn has bounded L-index in joint variables if and only if there exist c ∈ (0; +∞) and N ∈ N such that for each z∈Bn the inequality
N
X
kKk=0
|F(K)(z)|
K!LK(z) ≥c
∞
X
kKk=N+1
|F(K)(z)|
K!LK(z). (4.7)
Proof. Let 1β < θj < 1, j ∈ {1, . . . , n}, Θ = (θ1, . . . , θn). If the function F has bounded L-index in joint variables then by Theorem 2.2 F has bounded L-indexe in joint variables, where Le = (el1(z), . . . ,eln(z)),elj(z) = θjlj(z), j ∈ {1, . . . , n}. Let Ne =N(F,L,e Bn). Therefore,
max
|F(K)(z)|
K!LK(z): kKk ≤Ne
= max
(ΘK|F(K)(z)|
K!eLK(z) :kKk ≤Ne )
≥
n
Y
s=1
θNsemax
|F(K)(z)|
K!eLK(z):kKk ≤Ne
≥
n
Y
s=1
θNse|F(J)(z)|
J!eLJ(z) =
n
Y
s=1
θNse−js|F(J)(z)|
J!LJ(z) for allJ ≥0and
∞
X
kJk=N+1e
|F(J)(z)|
J!Lj(z) ≤max
|F(K)(z)|
K!LK(z):kKk ≤Ne ∞
X
kJk=Ne+1
θsjs−Ne
=
n
Y
i=1
θs
1−θs
max
|F(K)(z)|
K!LK(z):kKk ≤Ne
≤
n
Y
i=1
θs
1−θs Ne
X
kKk=0
|F(K)(z)|
K!LK(z). Hence, we obtain (4.7) withN =Ne and
c=
n
Y
i=1
θs
1−θs.