P N Natarajan
Another Theorem on Weighted Means
Abstract. In this short paper, which is a continuation of [2], we prove another interesting result concerning weighted means.
2000 Mathematics Subject Classification: 40,46.
Key words and phrases: Regular matrix, weighted means.
1. Introduction. For the sake of completeness, we recall the following. For a given infinite matrix A = (ank), n, k = 0, 1, 2, . . . and a given sequence x = {xk}, k = 0, 1, 2, . . . , by the A-transform of x = {xk}, we mean the sequence Ax = {(Ax)n},
(Ax)n= X∞ k=0
ankxk, n = 0, 1, 2, . . . , where it is assumed that the series on the right converges. If lim
n→∞(Ax)n = `, we say that x = {xk} is A-summable or summable A to `. If limn→∞(Ax)n= `, whenever
klim→∞xk = `, we say that A is regular. The following result is well-known (see [3], Theorem II.1, pp. 11-12).
Theorem 1.1 A = (ank)is regular if and only if (i) sup
n
X∞ k=0
|ank| < ∞;
(ii) lim
n→∞ank= 0, k = 0, 1, 2, . . . ; and
(iii) lim
n→∞
X∞ k=0
ank= 1.
An infinite series X∞ k=0
xk is said to be A-summable to ` if {sn} is A-summable
to `, where sn= Xn k=0
xk, n = 0, 1, 2, . . . .
2. Weighted Means.
Definition 2.1 ([3], p.16) The weighted mean method or ( ¯N , pn) method is defi- ned by the infinite matrix A = (ank), where
ank= (p
k
Pn, k¬ n;
0, k > n,
Pn= Xn k=0
pk, n = 0, 1, 2, . . . , Pn 6= 0, n = 0, 1, 2, . . . .
Theorem 2.2 ([3], p.16) The weighted mean method ( ¯N , pn)is regular if and only if
(i) Xn k=0
|pk| = O(Pn), n → ∞;
and
(ii) Pn → ∞, n → ∞.
Remark 2.3
|Pn| ¬ Xn k=0
|pk|
¬
n+mX
k=0
|pk|
¬ L|Pn+m|, for some L > 0, m = 0, 1, 2, . . . ; n = 0, 1, 2, . . . .
3. Main Result. Following Móricz and Rhoades [1], we prove the main result of the paper, which supplements [2].
Theorem 3.1 Let ( ¯N , pn), ( ¯N , qn)be two regular weighted mean methods and
(1) Pn = O(pnQn), n → ∞,
i.e., pnPQnn
¬ M for some M > 0, n = 0, 1, 2, . . . . Let
X∞ n=0
xn be ( ¯N , pn) summable to `. Then X∞ n=0
bn converges to ` if and only if
sup
n
"
|Qn| X∞ k=n
Pk
Qk+1
qk+1
pkQk − qk+2
pk+1Qk+2
#
<∞,
where bn= qn
X∞ k=n
xk
Qk, n = 0, 1, 2, . . .. Proof Let
sn = Xn k=0
xk and
tn =p0s0+ p1s1+ · · · + pnsn
Pn
, n = 0, 1, 2, . . . . Then
s0= t0 and sn = 1
pn
(Pntn− Pn−1tn−1), n = 1, 2, . . . .
Let X∞ n=0
xn be ( ¯N , pn) summable to ` so that lim
n→∞tn= `. Now, sn
Qn
= 1
pnQn
(Pntn− Pn−1tn−1)
= 1
pnQn
[Pn(tn− `) − Pn−1(tn−1− `) + `(Pn− Pn−1)]
= 1
pnQn
[Pn(tn− `) − Pn−1(tn−1− `) + `pn]
= Pn
pnQn
(tn− `) − Pn−1
pnQn
(tn−1− `) + ` Qn
so that sn
Qn
¬ M [|tn− `| + L|tn−1− `|] + |`|
|Qn|,
since |Pn−1| ¬ L|Pn|, using Remark 2.3
→ 0, n → ∞, since limn
→∞tn= ` and lim
n→∞Qn= ∞,
( ¯N , qn) being regular, using Theorem 2.2.
As already worked out in [2],
bn= −qnsn−1
Qn
+ qn
X∞ k=n
cksk,
where
ck = 1 Qk − 1
Qk+1, k = 0, 1, 2, . . . Now,
Bn= sn−1+ Qn −sn−1
Qn
+ X∞ k=n
cksk
!
(see [2])
= sn−1− sn−1+ Qn
X∞ k=n
cksk
= Qn
X∞ k=n
cksk
= Qn lim
m→∞
Xm k=n
cksk
= Qn lim
m→∞
Xm k=n
ck· 1
pk {Pktk− Pk−1tk−1}
= Qn lim
m→∞
"
cmPmtm
pm −cnPn−1tn−1 pn
+
mX−1 k=n
Pktk
ck
pk − ck+1
pk+1
#
(2) .
Let
A1= (
{xk} : X∞ k=0
xk is ( ¯N , pn) summable )
;
A2= (
{xk} : X∞ k=0
bk converges )
.
Note that A1, A2are BK spaces with respect to the norms defined by
||x||A1= sup
n0|tn|, x = {xk} ∈ A1; and
||x||A2= sup
n0|Bn|, x = {xk} ∈ A2, respectively. In view of Banach-Steinhaus theorem,
(3) ||x||A2 ¬ U||x||A1, for some U > 0.
For every fixed k = 0, 1, 2, . . . , define the sequence x = {xn}, where
xn=
1, if n = k;
−1, if n = k + 1;
0, otherwise.
For this sequence x,
||x||A1= pk
Pk
and ||x||A2 = |Qkck|.
Using (3), we have, for k = 0, 1, 2, . . . ,
|Qkck| ¬ U|pk|
|Pk|, so that
ckPk
pk
¬ U
|Qk|
→ 0, k → ∞, since limk
→∞Qk= ∞, in view of Theorem 2.2.
Consequently
(4) lim
k→∞
ckPk
pk
= 0.
Using (4) in (2), we have,
Bn= −cnPn−1tn−1
pn
Qn+ Qn
X∞ k=n
Pktk
ck
pk −ck+1
pk+1
= X∞ k=0
anktk,
where (ank) is defined by
ank=
0, if 0 ¬ k < n − 1;
−QncnpnPn−1, k = n− 1;
QnPk
ck
pk −cpk+1k+1
, k n.
We first note that lim
n→∞ank= 0, k = 0, 1, 2, . . . and X∞ k=0
ank= 1, n = 0, 1, 2, . . . so
that lim
n→∞
X∞ k=0
ank= 1. Thus, appealing to Theorem 1.1, X∞ n=0
bn converges to ` if and only if
(5) sup
n
"
|Qn|
cnPn−1
pn
+
X∞ k=n
Pk
ck
pk −ck+1
pk+1
!#
<∞.
However,
QncnPn−1 pn
¬ L
QncnPn
pn
since |Pn−1| ¬ L|Pn|, using Remark 2.3
= L|Qn| Pn
X∞ k=n
ck
pk −ck+1
pk+1
, using (4)
¬ L2|Qn|
X∞ k=n
Pk
ck
pk − ck+1
pk+1
since|Pn| ¬ L|Pk|, k n, using Remark 2.3
¬ L2|Qn|X
kn
Pk
ck
pk −ck+1
pk+1
. (6)
Using (6), it is now clear that (5) is equivalent to
sup
n0|Qn|
X
kn
Pk
ck
pk −ck+1
pk+1
< ∞.
Now,
ck
pk −ck+1
pk+1 = 1 pk
1 Qk − 1
Qk+1
− 1
pk+1
1
Qk+1 − 1 Qk+2
= qk+1
pkQkQk+1 − qk+2
pk+1Qk+1Qk+2.
Thus X∞ n=0
bn converges to ` if and only if
sup
n0|Qn|
X
kn
Pk
qk+1
pkQkQk+1 − qk+2
pk+1Qk+1Qk+2
< ∞,
i.e., sup
n0|Qn|
X
kn
Pk
Qk+1
qk+1
pkQk − qk+2
pk+1Qk+2
< ∞,
which completes the proof of the theorem. ■
References
[1] F. Móricz and B.E. Rhoades, An equivalent reformulation of summability by weighted mean methods, revisited, Linear Algebra Appl.349 (2002), 187–192.
[2] P.N. Natarajan, A generalization of a theorem of Móricz and Rhoades on weighted means (Communicated for publication).
[3] A. Peyerimhoff, Lectures on summability, Lecture Notes in Mathematics,107, Springer, 1969.
P N Natarajan
Old No. 2/3, New No. 3/3, Second Main Road, R.A. Puram, Chennai 600 028, India E-mail: pinnangudinatarajan@gmail.com
(Received: 4.10.2010)