1. Introduction
The concept of statistical convergence was introduced by Fast [1] and Steinhaus [2]. Besides, in this connection, Fridy [3] showed some relation to a Tauberian condition for the statistical convergence of xk. Subsequently, many researchers have worked in this area in several settings. For more recent works in this direction, one may refer to [4, 5]. Existing works in this field based on statistical convergence appears to have been restricted to real or complex sequences; however, Parida et al. [6] extended the idea for a locally convex Hausdorff topological linear space. Tauber [7] introduced the first Tauberian theorems for single sequences, that an Abel summable sequence is convergent with some suitable conditions. Later, a huge number of authors such as Landau [8], Hardy and Littlewood [9], and Schmidt [10] obtained some classical Tauberian theorems for Cessáro and Abel summability methods of single sequences. Recently, Braha [11] introduced some notions on statistical convergence by using the Nörlund-Cesáro summability method in a single sequence and proved some Tauberian theorems. In the last year, Canak and Totur [12], and Jena et al. [13] presented and studied several Tauberian theorems for single sequences. On the other hand, Knopp [14] obtained some classical type Tauberian theorems for Abel and C,1,1 summability methods of double sequences and proved that Abel and C,1,1 summability methods hold for the set of bounded sequences. Further, Moricz [15] proved some Tauberian theorems for Cesáro summable double sequences and deduced Tauberian theorems of Landau [16] and Hardy [17] type. Canak and Totur [18] have proved a Tauberian theorem for Cesáro summability of single integrals and also the alternative proofs of some classical type Tauberian theorems for the Cesáro summability of single integrals and later introduced by Parida et al. [6] for double integrals. Otherwise, the notion of C,1,1,1 summability of a triple sequence was originally introduced by Canak and Totur in 2016 [19]. Later, Canak et al. [20] studied some C,1,1,1 means of a statistical convergent triple sequence and gave some classical Tauberian theorems for a triple sequence that P-convergence follows from statistically C,1,1,1 summability under the two-sided boundedness conditions and slowly oscillating conditions in certain senses. Then, in 2020 Totur and Canak [21] proved Tauberian conditions under which convergence of triple integrals follows from C,1,1,1 summability. For more studies associated to the main topic of this paper, we refer the reader to [22, 23, 24].
Let pn,m,g and qn,m,g be any two non-negative real sequences with
Rn,m,g=∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k≠0nmg∈ℕ×ℕ×ℕ
and C,1,1,1-Cesáro summability method. Let xn,m,g be a sequence of real of complex numbers and set
Np,qn,m,gCn,m,g1,1,1=1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k1i+11j+11k+1∑u=0i∑v=0j∑y=0kxu,v,y
for nmg∈ℕ×ℕ×ℕ.
In this paper, we show necessary and sufficient conditions under which the existence of the limit limn,m,g→∞xn,m,g=L follows from that of limn,m,g→∞Np,qn,m,gCn,m,g1,1,1=L. These conditions are one-sided or two-sided if xn,m,g is a sequence of real or complex numbers, respectively.
Given two non-negative sequences pn,m,g and qn,m,g, the convolution p⋆q is defined by
Rn,m,g=p⋆qn,m,g=∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k=∑i=0n∑j=0m∑k=0gpn−i,m−j,g−kqi,j,k
with C,1,1,1 we will denote the triple Cesáro summability method. Now, let xn,m,g be a sequence, when p⋆qn,m,g≠0 for all nmg∈ℕ×ℕ×ℕ the generalized Nörlund-Cesáro transform of the sequence xn,m,g is the sequence Np,qn,m,gCn,m,g1,1,1 obtained by putting
Np,qn,m,gCn,m,g1,1,1=1p⋆qn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k1i+11j+11k+1∑u=0i∑v=0j∑y=0kxu,v,y.E1
We say that the sequence xn,m,g is generalized Nörlund-Cesáro summable to L determined by the sequences pn,m,g and qn,m,g (or simply summable Np,qn,m,gCn,m,g1,1,1) to L if
limn,m,g→∞Np,qn,m,gCn,m,g1,1,1=L.E2
Throughout this paper, we will assume that the sequences pn,m,g and qn,m,g are satisfying the following conditions
qn,m,g≥1,∑i=0n∑j=0m∑k=0gpi,j∼nmg,nmg∈ℕ×ℕ×ℕ,E3
qλn−i,m−j,g−k≤2qn−i,m−j,g−k,i=1,2,…,n;j=1,2,…,m;λ>1k=1,2,…;λ>1,E4
qn−i,m−j,g−k≤2qλn−i,m−j,g−ki=1,2,…,λn;j=1,2,…,λm;k=1,2,…,;0<λ<1,E5
where λn=λn, λm=λm and λg=λg. On the other hand, an,m,g∼bn,m,g means that there are constants C,C1 such that an,m,g≤Cbn,m,g≤C1an,m,g. If
limn,m,g→∞xn,m,g=LE6
implies (2), then the method Np,qn,m,gCn,m,g1,1,1 is said to be regular. Nevertheless, the converse is not always true as can be seen in the following example:
Let us consider that pn,m,g=qn,m,g=1 for all nmg∈ℕ×ℕ×ℕ. Besides, we define the following sequence x=xi,j,k=−1i+j+k, then we get
1n+1m+1g+1∣∑i=0n∑j=0m∑k=0g1i+1j+1k+1∑u=0i∑v=0j∑y=0k−1u+v+y∣≤1n+1m+1g+1∑i=0n∑j=0m∑k=0g1i+1j+1k+1∑u=0i∑v=0j∑y=0k1→1asn,m,g→∞.
and as we know, x=xi,j,k is not convergent. Notice that (6) can imply (2) under a certain condition, which is called a Tauberian conditions. Any theorem which states that convergence of a sequence follows from its Np,qn,m,gCn,m,g1,1,1 summability and some Tauberian conditions are said to be a Tauberian theorems for the Np,qn,m,gCn,m,g1,1,1 summability method.
Next, we will find some conditions under which the converse implication holds, for defined convergence. Exactly, we will prove under which conditions statistical convergence of sequences xn,m,g, follows from statistically Nörlund-Cesáro summability method.
A sequence xn,m,g is weighted Np,qn,m,gCn,m,g1,1,1-statistically convergent to L if for every ε>0,
limn,m,g→∞1p⋆qn,m,g∣{i,j,k≤p⋆qn,m,g:1p⋆qn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k1i+11j+11k+1∑u=0i∑v=0j∑y=0kxu,v,y−L∣≥ε}∣=0.
And we say that the sequence xn,m,g is statistically summable to L by the weighted summability method Np,qn,m,gCn,m,g1,1,1 if st−limn,m,gNp,qn,m,gCn,m,g1,1,1=L. We will denote by Np,qn,m,gCn,m,g1,1,1st the set of all sequences which are statistically summable, by the weighted summability method Np,qn,m,gCn,m,g1,1,1.
Theorem 1.1 Let x=xn,m,g be a sequence Np,qn,m,gCn,m,g1,1,1 summable to L, then the sequence x=xn,m,g is Np,qn,m,gCn,m,g1,1,1-statistically convergent to L, but not conversely.
Proof: The first part of the proof is obvious. To prove the second part, we will show the following example:
Let us define
xi,j,k=xyz,fori=n2j=m2andk=g20,otherwise
and pn,m,g=1=qn,m,g. Under this conditions we obtain,
1n+1m+1g+1∣{i,j,k≤n+1,m+1,g+1:∣1n+1m+1g+1∑i=0n∑j=0m∑k=0g1Pi,j,k∑u=0i∑v=0j∑y=0kpu,v,yxu,v,y−0∣≥ε}∣≤n+1m+1g+1n+1m+1g+1→0.
On the other hand, for i=n2, j=m2 and k=g2, we have
1n+1m+1g+1∑i=0n∑j=0m∑k=0g1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y→∞,asn,m,g→∞.
From last relation follows that x=xn,m,g is not Np,qn,m,gCn,m,g1,1,1 summable to 0.
Theorem 1.2 Let x=xn,m,g be a sequence statistically convergent to L and ∣xn,m,g−L∣≤M for every nmg∈ℕ×ℕ×ℕ. Then, it converges Np,qn,m,gCn,m,g1,1,1-statistically to L.
Proof: From the fact that xn,m,g converges statistically to L, we have
limn,m,g→∞∣i,j,k≤n,m,g:∣xi,j,k−L∣≥ε}∣nmg=0.
We will denote Bε=ijk≤nmg:xi,j,k−L≥ε and B¯ε=ijk≤nmg:xi,j,k−L≤ε. Then,
∣1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L∣=∣1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L∣
≤1Rn,m,g∑i=0n∑j=0m∑k=0g¯ijk∈Bεpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0k∣xu,v,y−L∣+1Rn,m,g∑i=0n∑j=0m∑k=0g¯ijk∈B¯εpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0k∣xu,v,y−L∣≤M1Rn,m,g∑i=0n∑j=0m∑k=0g¯ijk∈Bε1+ε≤MC2nmg∑i=0n∑j=0m∑k=0g¯ijk∈Bε1+ε→0+ε,asn,m,g→∞,
for some constant C2.
Converse of Theorem 1.2 is not true as can be seen in the following example.
Consider that pn,m,g=n+1m+1g+1, qn,m,g=1 for some nmg∈ℕ∪0×ℕ∪0×ℕ∪0 and define the sequence x=xn,m,g as follows:
xi,j,k={1,fori=p2−p,…,p2−j=t2−t,…,t2−1andk=o2−o,…,o2−1;−1pto,fori=p2,p=2,..j=t2,t=2,..andk=o2,o=2,…0,otherwise
Under this conditions, after some basic calculations we get that x=xn,m,g is Np,qn,m,gCn,m,g1,1,1-summable to 1. Therefore, by Theorem 1.2, x=xn,m,g is Np,qn,m,gCn,m,g1,1,1-statistically convergent. On the other hand, the sequences p2;p=2,3,…, t2;t=2,3,… and o2;o=2,3,… have natural density zero and it is clear that st-liminfn,m,gxn,m,g=0 and st-limsupn,m,gxn,m,g=1. Hence, xi,j,k is not statistically convergent.
Advertisement
2. Tauberian theorems under Np,qn,m,gCn,m,g1,1,1-statistically convergence
In this section, we show the results that we obtained. Throughout this paper, Rλn,m,g and Rλn,λm,λg will have the same meaning.
Consider that st-limi,j,kxi,j,k=L; xn,m,g is Np,qn,m,gCn,m,g1,1,1-statistically convergent and (13) satisfies, then for every t>1, is valid the following relation
st−limi,j,k1Rλi,j,k−Ri,j,k∑w=i+1λi∑e=j+1λj∑r=k+1λkpw,e,rqλi−w,λj−e,λk−r1w+1e+1r+1∑u=0i∑v=0j∑y=0kxw,e,r−xi,j,k=0E7
and in case where 0<t<1,
st−limi,j,k1Ri,j,k−Rλi,j,k∑w=λi+1i∑e=λj+1j∑r=λk+1kpw,e,rqi−w,j−e,k−r1w+1e+1r+1∑u=0i∑v=0j∑y=0kxi,j,k−xw,e,r=0.E8
The condition given by relation (13) is equivalent to the condition
st−limn,m,g→∞Rn,m,gRλn,m,g>1,0<λ<1.E9
Proof: Suppose that relation (13) is valid, 0<λ<1, w=λn=λn, e=λm=λm and r=λg=λg, nmg∈ℕ×ℕ×ℕ. Then, it follows that
1λ>1⇒wλ=λnt≤n,1λ>1⇒eλ=λmt≤mand1λ>1⇒rλ=λgt≤g.
From above relation and definition of sequences pn,m,g and qn,m,g, we have
Rn,m,gRλn,m,g≥Rnλ,mλ,gλRλn,m,g⇒st−liminfn,m,g→∞Rn,m,gRλn,m,g≥st−liminfn,m,g→∞Rnλ,mλ,gλRλn,m,g>1.
Conversely, suppose that (9) is valid. Now, let λ>1 be given and let λ1,λ2,λ3 be chosen such that 1<λ1,λ2,λ3<λ. Set w=λn=λn, e=λm=λm and r=λg=λg. From 0<1λ<1λ1,1λ2,1λ3<1, it follows that
n≤λn−1λ1<λnλ1=wλ1,m≤λm−1λ2<λmλ2=eλ2andg≤λg−1λ3<λgλ3=rλ3
provided λ1,λ2,λ3≤λ−1n,λ−1m,λ−1g, which is a case where if n, m and g are large enough. Under this condition, we obtain
Rλn,m,gRn,m,g≥Rλn,m,gRwλ1,eλ2,rλ3⇒st−liminfn,m,g→∞Rλn,m,gRn,m,g≥st−liminfn,m,g→∞Rλn,m,gRwλ1,eλ2,rλ3>1.
Consider that (13) is satisfied and let x=xi,j,k be a sequence of complex numbers which is Np,qn,m,gCn,m,g1,1,1-statistically convergent to L. Then,
st−limn,m,g1Rλn,m,g−Rn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y=Lforλ>1E10
and
st−limn,m,g1Rn,m,g−Rλn,m,g∑i=λn+1n∑j=λm+1m∑k=λg+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y=Lfor0<λ<1.E11
Proof: We begin proving the case (10), i.e. when λ>1. Then, we have
1Rλn,m,g−Rn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L=Rλn,m,gRλn,m,g−Rn,m,g1Rλn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L−Rn,m,gRλn,m,g−Rn,m,g1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L=Rλn,m,gRλn,m,g−Rn,m,g1Rλn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−j1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L−Rn,m,gRλn,m,g−Rn,m,g1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqλn−i,λm−j,λg−k+qn−i,m−j,g−k−qn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L=Rλn,m,gRλn,m,g−Rn,m,g1Rλn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L−Rn,m,gRλn,m,g−Rn,m,g1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−Rn,m,gRλn,m,g−Rn,m,g1Rn,m,g∑i=0n∑j=0m∑k=0gpi,j,kqλn−i,λm−j,λg−k−qn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−L.E12
From (12), definition of the sequence qn,m,g and relation limsupn,m,gRλn,m,gRλn,m,g−Rn,m,g<∞, we get (10).
Prove of (11) is made similarly to the prove of (10).
In the following theorem, we characterize the converse implication when the statistically convergence follows from its Np,qn,m,gCn,m,g1,1,1− statistically convergence.
Theorem 1.3 Let pn,m,g and qn,m,g be two non-negative real sequences and
st−liminfn,m,g→∞Rλn,m,gRn,m,g>1foreveryλ>1,E13
where λn,m,g=λnλmλg=λnλmλg denotes the integral part of λnλmλg for every nmg∈ℕ×ℕ×ℕ, and let xn,m,g be a sequence of real numbers which is Np,qn,m,gCn,m,g1,1,1-statistically convergent to a finite number L. Then, xn,m,g is st-convergent to the same number L if and only if the following two conditions hold
infλ>1limsupn,m,g1Rn,m,g∣{i,j,k≤Rn,m,g:1Rλi,j,k−Ri,j,k∑w=i+1λi∑e=j+1λj∑r=k+1λkpw,e,rqλi−w,λj−e,λk−r1w+1e+1r+1∑u=0i∑v=0j∑y=0kxw,e,r−xi,j,k≤−ε}∣=0,E14
and
inf0<λ<1limsupn,m,g1Rn,m,g∣{i,j,k≤Rn,m,g:1Ri,j,k−Rλi,j,k∑w=λi+1i∑e=λj+1j∑r=λk+1kpw,e,rqi−w,j−e,k−r1w+1e+1r+1∑u=0i∑v=0j∑y=0kxi,j,k−xw,e,r≤−ε}∣=0.E15
Proof: Necessity: Suppose that limn,m,g→∞xn,m,g=L and (13) holds. By Proposition 2, we have
limn,m,g→∞1Rλn,m,g−Rn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−xn,m,g=limn,m,g→∞{(1Rλn,m,g−Rn,m,g∑i=n+1λn∑j=m+1λm∑k=g+1λgpi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y)−xn,m,g}=0,
for every λ>1. In case where 0<λ<1, we have that
limn,m,g→∞1Rn,m,g−Rλn,m,g∑i=λn+1n∑j=λm+1m∑k=λg+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxn,m,g−xu,v,y=limn,m,g→∞{xn,m,g−(1Rn,m,g−Rλn,m,g∑i=λn+1n∑j=λm+1m∑k=λg+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y)}=0.
Sufficiency: Consider that (14) and ((15) are satisfied. In what follows, we will prove that limn,m,g→∞xn,m,g=L. Given any ε>0, by (14) we can choose λ1>0 such that
liminfn,m,g→∞1Rλn1,λm1,λg1−Rn,m,g∑i=n+1λn1∑j=m+1λm1∑k=g+1λg1pi,j,kqλn−i,λm−j,λg−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−xn,m,g≥−ε,E16
where λn1=λn1, λm1=λm1 and λg1=λg1. By the assumed summability Np,qn,m,gCn,m,g1,1,1 of xn,m,g, Proposition 2 and (16), we have
limsupn,m,g→∞xn,m,g≤L+ε,E17
for any λ>1.
On the other hand, if 0<λ<1, for every ε>0, we can choose 0<λ2<1 such that
mliminfn,m,g→∞1Rn,m,g−Rλn2,λm2,λg2∑i=λn2+1n∑j=λm2+1m∑k=λg2+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxn,m,g−xu,v,y≥−ε,E18
where λn2=λn2, λm2=λm2 and λg2=λg2. By the assumed summability Np,qn,m,gCn,m,g1,1,1 of xn,m,g, Proposition 2 and (18), we have
liminfn,m,g→∞xn,m,g≥L−ε,E19
for any 0<λ<1.
Since ε>0 is arbitrary, combining (17) and (19), we obtain
limn,m,g→∞xn,m,g=L.
In the following theorem, we will consider the case where x=xn,m,g is a sequence of complex numbers.
Theorem 1.4 Let (13) be satisfied and let xn,m,g be a sequence of complex numbers which is Np,qn,m,gCn,m,g1,1,1-statistically convergent to a finite number L. Then, xn,m,g is convergent to the same number L if and only if the following two conditions hold
infλ>1limsupn,m,g1Rn,m,g∣{i,j,k≤Rn,m,g:1Rλi,j,k−Ri,j,k∑w=i+1λi∑e=j+1λj∑r=k+1λkpw,e,rqλi−w,λj−e,λk−r1w+1e+1k+1∑u=0i∑v=0j∑y=0kxw,e,r−xi,j,k≥ε}∣=0,E20
and
inf0<λ<1limsupn,m,g1Rn,m,g∣{i,j,k≤Rn,m,g:1Ri,j,k−Rλi,j,k∑w=λi+1i∑e=λj+1j∑r=λk+1kpw,e,rqi−w,j−e,k−r1w+1e+1r+1∑u=0i∑v=0j∑y=0kxi,j,k−xw,e,r≥ε}∣=0.E21
Proof: Necessity: If both (2) and (6) hold, then Proposition 2 yields (20) for every λ>1 and (21) for every 0<λ<1.
Sufficiency: Suppose that (2), (13) and one of the conditions (20) and (21) are satisfied. For any given ε>0, there exists λ1>0 such that
limsupn,m,g→∞∣1Rλn1,λm1,λg1−Rn,m,g∑i=n+1λn1∑j=m+1λm1∑k=g+1λg1pi,j,kqλn1−i,λm1−j,λg1−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−xn,m,g∣≤ε,
where λn1=λn1, λm1=λm1 and λg1=λg1. Taking into account fact that xn,m,g is Np,qn,m,gCn,m,g1,1,1 summbale to L and Proposition 2, we have the following estimation
limsupn,m,g→∞∣L−xn,m,g∣≤limsupn,m,g→∞∣L−1Rλn1,λm1,λg1−Rn,m∑i=n+1λn1∑j=m+1λm1∑k=g+1λg1pi,j,kqλn1−i,λm1−j,λg1−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y∣+limsupn,m,g→∞∣L−1Rλn1,λm1,λg1−Rn,m∑i=n+1λn1∑j=m+1λm1∑k=g+1λg1pi,j,kqλn1−i,λm1−j,λg1−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y−xn,m,g∣≤ε.
For a given ε>0, there exists λ2>0 such that
limsupn,m,g→∞∣1Rn,m,g−Rλn2,λm2,λg2∑i=λn2+1n∑j=λm2+1m∑k=λg2+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxn,m,g−xu,v,y∣≤ε,
where λn2=λn2, λm2=λm2 and λg2=λg2. Taking into account fact that xn,m,g is Np,qn,m,gCn,m,g1,1,1 summbale to L and Proposition 2, we obtain the following
limsupn,m,g→∞∣L−xn,m,g∣limsupn,m→∞∣L−1Rn,m,g−Rλn2,λm2,λg2∑i=λn2+1n∑j=λm2+1m∑k=λg2+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxu,v,y∣+limsupn,m,g→∞∣1Rn,m,g−Rλn2,λm2,λg2∑i=λn2+1n∑j=λm2+1m∑k=λg2+1gpi,j,kqn−i,m−j,g−k1i+1j+1k+1∑u=0i∑v=0j∑y=0kxn,m,g−xu,v,y∣≤ε.
Since ε>0 in either case, we get
limn,m,g→∞xn,m,g=L.