I nternat. J. Math. & Math. Si. 503 Vol. 6 No. 3 (1983) 503-510 INFINITE MATRICES AND ABSOLUTE ALMOST CONVERGENCE MURSALEEN Department of Mathematics Aligarh Muslim University 202001, INDIA Aligarh (Received January 26, 1981 and in revised form August 16, 1982) ABSTRACT. In 1973, Stieglitz [5] introduced a notion of FB-Convergence which provided convergence due to Lorentz [i]. a wide generalization of the classical idea of almost The concept of strong almost convergence was introduced by Maddox [3] who later on generalized this concept analogous to Stieglitz’s extension of almost convergence [4]. In the present paper we define absolute the concept of FB-Convergence which naturally emerges from FB-Convergence. KEY WORDS AND PHRASES. In fine matces, almost convergence, stng almost convergence, FB-Convergence, absolute FB-Convergence. 1980 MATHEMATICS SUBJECT CLASSIFICATION CODES. 1 40C05, 40D05. INTRODUCT ION Let o’ c, and c o denote respectively the Banach spaces of bounded, convergent, (Xk) and null sequences x of complex numbers with norm lxll sp Xkl and let v be the space of sequences of bounded variation, that is, v Suppose that B Then x oo is {x: (B.) said to be lim n uniformly for i The space F B (Bi). llxll- k=O Ixk Xk_ll < 0}. + oo, x_ I is a sequence of infinite complex matrices with FB-Convergent [5], lira n p=O (Bix) n to the value Lim bnp(i)xP Bi=(bnp(i) Bx, if Lira Bx, 0,I,2 of B In case B FB-Convergent B 0 sequences depends on the fixed chosen sequence (I) (unit matrix), the space F B is same as c and, for 504 B MURSALEEN B B(1). i (I) (B i ), it is same as the space f of almost convergent sequences [i], where I (b(1) np (i)) with I b(l) (i) np I i n+l ! 0 -< p -< i + n elsewhere Maddox [4] generalized strong almost convergence by saying that x and only if _Dnp(i) IXp P 0 (n- oo, uniformly in i) s - P s[F B] if (i.1) assuming that the series in (i.i) converges for each n and i. In particular, if B B0, the [FB] c; if B strongly almost convergent sequences [3]. entry), 0 (i,I,i and e B i, then We shall write e [FB] [f], the space of (0,0,...,0,i (kth k ). Let s be the space of all complex sequences and d F B B {x s: {x (dB ): Lim Bx lira exists for each i} (BiX)n lim t (i x) exists uniformly in i n and the limit is independent of i}, where g bnp(i)x t n p p0 0p(i)Xp (i,x) (n >- I) (n 0) (n =-I) 0 and 0 otherwise. Let n (i x) x) t (i n tn-i (i,x) Therefore, we have [bnp(i) n bn_l, p (i)]Xp, (n -> l) (1.2) (i,x) Op(i) x P (n 0) INFINITE MATRICES AND ABSOLUTE ALMOST CONVERGENCE DEFINITION. A sequence x Let B (Bi) be a sequence of infinite matrices with B FB-Convergent is said to be absolutely 19 if (i x) 505 i (bnp(i)). converges n n=0 uniformly for i >- 0, and lim t (i,x) which must exist should take the same value for n->oo n all i. 2. We denote the space of absolute FB-Convergent sequences by v(B). THE MAIN RESULT. In this note, we denote by (v,v(B)) the set of matrices which give new classes of absolute B-conservative matrices and absolute almost B-conservative matrices. Let A be any infinite complex matrix for which the pth row-sum converges for a given x for all x in some class. We have A x P (Ax) apk P x k and p0 bnp(i) Xp (BiX)n Therefore, (BiAX)n bnp(i) p=o np A P k=O apk xk" and, assuming the interchange of order of summation can be justified (see lemma), we get that (BiAx) n k=O p--0 bnp (i) apk (2.1) x k Now, by (1.2) and (2.1), we have n(i’Ax) tn(i,Ax) tn_ l(i,Ax) p0 [bnp(i)- bn-l’ p (i) ]ApX, (n (n k=O where gnk (i) Xk, . I), 0), (2.2) 506 MURSALEEN p=O gnk(i) p=O HEOREM. SUPn Let A be an infinite matrix. sup p,k 30p (i) > (n apk p0= Then A: me =k < o, bnp (i) i), 0). for each i. v--> v(B) if and only if < there is an N such that for r,i (ii) (n bn_l, p (i)]apk (B i) be a sequence of infinite matrices with Let B (i) [bnp(i) 0,i,2,... r E gnk(i) n=N (apk)p>_0 (iii) (iv) k=0 Let A e (v,v(B)). < e v(B) for each k, For each k, let (Xk) and v(B) apk) p->0 apk be FB-Convergent with limit k" And let (In each case, limit is taken for p > 0). k=OE apk be FB-Convergent with limit c. If x (constant), K k=0 v, then lim t (i Ax) lira x k k-o n-o n + k=0 (xk k) k" lim x k-o We use the following lemma in the proof. LEMMA. If either the necessity part or the sufficiency part of the theorem holds, then, for x v, Xk p0 bnp(i) p=0 bnp(i) K=0 apkxk kOapk’= PROOF. If either A: v- v(B) or the conditions (i)-(iv) of the theorem hold, then by part_ i summation, for x k=O where dpk a %=k p%" v, apkXk k:O dpk (xk Since condition (i) holds, dpk Xk_ I) is bounded for all p,k. Thus . p=O INFINITE MATRICES AND ABSOLUTE ALMOST CONVERGENCE bnp (i) apkxk d bnp (i) p=O k=0 (xk = k=O Xk-l) pk (x 507 Xk-i k bnp(i) ape’ (where the inversion is justified by absolute convergence) bnp(i) apk xk k=0 since l im x k_o PROOF OF THEOREM. Since k p0 b np (i) d 0. pk Condition (i) follows from the fact that A: v Necessity. v, necessity of (iii) and (iv) is obvious. ek,e It is clear that, for fixed p and j, J x / k=0 apk x k We are given that, for all x e v, it tends to is a continuous linear functional on v. (for fixed p) and hence, by the Banach-Steinhaus Theorem [2], this a limit as j limit A x is also a continuous linear functional on v. P We observe that, although Y. n=0 ln(i,Ax) for r i and >- o, {qN,i i >- o, qm, i n=l is pointwise bounded on v. i_> o Such an N exists. For suppose not. qr i(Xr) By the principle of condensation of singularities [6], v: sup i_>o qr o but not uniformly v, and there is an integer N such v with sup e o on the behaviour of i_>O {x -> (i,Ax) [n(i’Ax)[" is a continuous seminorm on 0,1,2,... ther exists x r n it needs Thus, by definition, there is an m such that qm,i(x) that (i,Ax) is uniformly convergent in i (i,Ax) for sufficiently large n. For m t Now, we can say that uniforn convergence bears only bounded. n l n-o is uniformly convergent in i For example, if not be uniformly bounded in i. (n -> i and i), then In (i,Ax) i (x) for r 0,1,2,.. Then 508 MURSALEEN is of second category in v and hence nonempty, i.e., there is x sup q i>o r, i (x) for v with 0,1,2 r But this contradicts the fact that to each x e v there exists an integer N i->oSUp qNx,i(x x with < o. Now, by another application of the Banach-Steinhaus Theorem, there exists a con- stant M such that -< M qN,i(x) Apply (2.3) with x (xk) defined by x Ixll. i for k k -< r and o for k (2.3) > r. Hence (ii) must hoid. Sufficiency. Suppose that the conditions (i)-(iv) hold and that x defined v(B) as a subspace of % Thus, in order to show that Ax necessary to prove that Ax is bounded. v(B), v. We have it is By virtue of condition (i), this follows im- mediately. Now, it follows from (iv) and the lemma that gnk (i) k=O converges for all i,n. Hence, if we write hnk(i) then hnk(i) %--k gnu(i), is defined, also for fixed i,n, hnk (i) --> 0 as k--. (2.4) Now condition (iv) gives us that (2.5) n=O converges aniformly in i, and, for suitable chosen N, n=N is bounded. lhno (i) (2.6) By virtue of condition (iii), for fixed k, we get that n=O converges uniformly in i. Since k-i hnk(i) hn0(i) %=0 gnu(i)’ (2.7) INFINITE MATRICES AND ABSOLUTE ALMOST CONVERGENCE 509 it follows that, for fixed k, n=O lhnk (i) (2.8) converges uniformly in i. Now (i,Ax) gnk(i) k=O x k [hnk(i) k=O hnk(i) k=O h (xk n k+l(i)] x k (2.9) Xk_l) by (2.4) and the boundedness of x k. Condition (ii) and the boundedness of (2.6) show that n=N is bounded for all k,i. lhnk(i) (2.10) We can make k=k +1 o arbitrarily small by choosing k c> o, we can choose k o o sufficiently large. It therefore follows that, given so that, for all i, k=k +1 n=N hnk(i)(xk < Xk_l) o - (2.11) By the uniform convergence of (2.8), it follows that, once k has been chosen, we can o choose n o so that, for all i, k n=n +i o o k=o hnk(i)(xk < g. Xk_l) It follows from (2.11) that the same inequality holds when 7. is replaced by n=N 7. n=no+l hence, for all i, n=n +1 o hnk(i)(xk k=O Xk_l) Hence, n=n +i o ln(i,Ax) Thus n=O I@ n(i,Ax) < 2g. < 2g. (2.12) 510 _ MURSALEEN converges uniformly. Now, by virtue of (2.9), we have lim n-oo tn(i,Ax) - tN_l(i,Ax) __-- hnk(i (x k=0 N= hnk(i) Xk-l) (xk Xk_l) k n=N k=O (2.13) the assertion being justified by absolute convergence because of the boundedness of (2.10). By (2.7), we have n=N Thus, k-i hnk(i)= lim t (i n-o n n=N hno(i)=0 Ax) % lira x k k-o %=0 %=k p=O + >- k=0 gn%(i) n=N (xk bN_ 1 p lim k-o (i) ap " Xk) k This completes the proof. REFERENCES i. LORENTZ, G.G. A contribution to the theory of divergent series, Acta Math. 80 (1948), 167-190. Press, 1970. 2. MADDOX, l.J. Elements of Functional Analysis, Cambridge University 3. MADDOX, l.J. 61-64. A new type of convergence, Math. Proc. Camb. Phil. Soc. 83 (1978), 4. IviADDOX, l.J. On strong almost convergence, Math. Proc. Camb. Phil. Soc. 85 (1979), 345-350. 5. STIEGLITZ, M. Eine Verallgemeinerung des Begriffs der Fastkonvergenz, Math. Japon. 18 (1973), 53-70. 6. YOSIDA, K. Functional Analys__i__s, Springer-Verlag, Berlin-Heidelberg-New York, 196