=1 k
a
k
cos kx ,
and let f(x) = lim
n
S
n
(x), if it exists. The convergence of the series (1.1) in the metric space L, has been studied by
various authors under different conditions on the coefficients a
k
.
Theorem A [8]. If {a
k
} is convex (A
2
a
k
> 0 , for every k) null sequence, then the cosine series (1.1) is the Fourier series of
its sum f, and
S
n
(x) f(x) = o(1) , if and only if a
n
log n = o(1), n .
Kolmogorov extended the definition of convex null sequence. A sequence {a
k
} is said to be quasiconvex if a
k
= o(1), k
, and the series
=1 k
k  A
2
a
k
 < .
Theorem B[4]. If {a
k
} is a quasiconvex null sequence, then the cosine series (1.1) is the Fourier series of its sum f, and
S
n
(x) f(x) = o(1) , if and only if a
n
log n = o(1), n .
Sidon generalized the concept of convex (quasiconvex) null sequence. We say a sequence {a
k
} belongs to the class S[6], if
there exists a monotonically decreasing sequence {A
k
} such that E A
k
< , and A a
k
 s A
k
, for every k.
A quasiconvex null sequence satisfies the class S if we take A
n
=
=n m
 A
2
a
m
 .
Theorem C[7]. Let {a
k
} belong to the class S. Then the cosine series (1.1) is the Fourier series of its sum f and
S
n
(x) f(x) = o(1), if and only if a
n
log n = o(1), n .
Fomin[2] generalized Teljakovskis Theorem by introducing a new class F
p
. A null sequence {a
k
} belongs to the class F
p
,
if for some 1 < p s 2, we have
=
=

.

\

A
n
k
p
n k
p
k
n a
1
1
< .
and proved that F
p
c BV , where BV denotes the class of bounded variation and
if {a
k
} c F
p
, then f c L
1
(o, t) and
A
1
1
k
p
k
p
a k < .
Theorem D [2]. Let {a
k
} satisfy the conditions of F
p ,
then the cosine series (1.1) is the Fourier series of its sum f and
S
n
(x) f(x) = o(1) if and only if a
n
log n = o(1), n .
Kumari and Ram [5]introduced a new modified trigonometric cosine sum
2000 AMS Mathematics Subject Classification : 42A20, 42A32
On L
1
Convergence of Cosine Sums
59
(1.2) f
n
(x) = (a
0
/2) +
= =
n
k j
n
k 1
A(a
j
/j) k cos kx ,
and proved
Theorem E[5]. Let the sequence {a
k
} satisfy the conditions of the class S.
If a
n
log n = o(1), n , then f(x) f
n
(x) = o(1), n .
In this paper, we generalize Theorem E by obtaining a necessary and sufficient condition for the convergence of
the limit of (1.2) in the metric space L by taking {a
k
} in the class F
p
. We also deduce Fomins Theorem D as a corollary
from our result.
II. LEMMA
We require the following Lemma for the proof of our result:
Lemma 1[3]. If D
n
(x),D
n
(x) and F
n
(x) denote Dirichlet, conjugate Dirichlet and Fejrs kernel respectively, then F
n
(x) =
D
n
(x)
(1/n+1)D'
n
(x) .
III. MAIN RESULT
Theorem 1. Let {a
k
} belong to the class F
p
. Then
f(x) f
n
(x) = o(1) if and only if a
n
log n = o(1), n.
Proof. Application of Lemma 1 and summation by parts give
(3.1) f
n
(x)
=
2
a
0
+
=
n
1 k
=
n
k j
A


.

\

j
a
j
k cos kx
= S
n
(x) 
.

\

+
+
1
1
n
a
n
D'
n
(x)
= S
n
(x) a
n+1
D
n
(x) + a
n+1
F
n
(x)
=
=
n
k 0
Aa
k
D
k
(x) + a
n+1
F
n
(x)
Since D
n
(x) = O(1/x
2
) and a
n
0 as n, lim
n
f
n
(x) exists and
f(x) = lim
n
f
n
(x). Now,
(3.2)
}
t
0
f
n
(x) f(x) dx =
}
n
1
0
f
n
(x) f(x)  dx +
}
t
n
1
f
n
(x) f(x) dx.
For the first integral of right hand side of (3.2),we have
}
n
1
0
f
n
(x) f(x)  dx s
}
n
1
0
o
n
(x) f(x)dx +
}
n
1
0
f
n
(x) o
n
(x)dx,
where o
n
(x) is the Fejr sum of S
n
(x), and
}
n
1
0
o
n
(x) f(x)dx = O(o
n
(x) f(x) ) , n .
From (3.1), we get
f
n
(x) o
n
(x) = a
n+1
F
n
(x) +

.

\

+1
1
n
=
A
n
k
k k
x D a k
1
) ( 

.

\

+1
1
n
=
n
k
k k
x D a
1
) (
On L
1
Convergence of Cosine Sums
60
Therefore,
}
n
1
0
f
n
(x) o
n
(x) dx s (t/2) a
n+1
+

.

\

+1
1
n
}
=
A
n
k
n
k k
x D a k
1
1
0
) ( dx
+

.

\

+1
1
n
}
=
n
k
n
k k
x D a
1
1
0
) ( dx
or
}
n
1
0
f
n
(x) o
n
(x) dx =O
(
A

.

\

=
n
k
k
a k
1
n
1
, n .
Hence, for the first integral in (3.2) we have
}
n
1
0
f
n
(x) f(x) dx = O ( ) ) ( ) ( x f x
n
o + O
(
A

.

\

=
n
k
k
a k
1
n
1
, n .
For the second integral of right hand side of (3.2), we have
(3.3)
}
t
n
1
f
n
(x) f(x) dx s
}
t
n
1
a
n+1
F
n
(x) dx +
}
t
n
1

+ = 1 n k
Aa
k
D
k
(x) dx
s
}
t
0
a
n+1
F
n
(x)dx +
}
t
n
1

+ = 1 n k
Aa
k
D
k
(x) dx
= (t/2)a
n+1
+
}
t
n
1

+ = 1 n k
Aa
k
D
k
(x) dx .
Combining all these estimates as in [2], we obtain
}
t
0
f
n
(x) f(x)dx = O 
.

\

A + A +
+ =
= 1
1
1
) ( ) (
n k
p
k
p
n
k
k n
a k a k x f x o
= o(1) , n .
This completes the proof of our result.
Corollary. If {a
k
} belongs to F
p
, then
S
n
(x) f(x) = o(1) if and only if a
n
log n = o(1), n .
Proof . We have
}
t
0
 f(x) S
n
(x)  dx =
}
t
0
 f(x) f
n
(x) + f
n
(x) S
n
(x)  dx
s
}
t
0
 f(x) f
n
(x) dx +
}
t
0
 f
n
(x) S
n
(x) dx
s
}
t
0
 f(x) f
n
(x) dx +
}
t
0
a
n+1
D
n
(x)dx +
}
t
t
a
n+1
K
n
(x) dx
and
On L
1
Convergence of Cosine Sums
61
}
t
0
 a
n+1
D
n
(x) dx s
}
t
0
 f
n
(x)S
n
(x) dx +
}
t
t
 a
n+1
K
n
(x)  dx
}
t
0
 f(x) S
n
(x) dx +
}
t
t
 a
n+1
K
n
(x) dx .
Since
}
t
t
 D
n
(x) dx behave like log n for large values of nand
}
t
0
) ( ) ( x f x f
n
dx=0 by our theorem for n , the
corollary follows.
REFERENCES
[1]. N.K.Bari A Treatise on trigonometric series (London : Pregamon Press) Vol.1,Vol. II (1964).
[2]. G.A .Fomin,On linear methods for summing Fourier series, Mat. Sb., 66 (107) (1964) 114152.
[3]. N.Hooda and B.Ram, Convergence of certain modified cosine sum, Indian J. Math,44(1)(2002),4146.
[4]. A.N.Kolmogorov. Sur I' ordre de grandeur des coefficients de la serie de Fourierlebesque, Bull. Acad. Polan. Sci.
Ser. Sci. Math., Astronom Phys., (1923) 8386.
[5]. S.Kumari and B.Ram L
1
convergence of a modified cosine sum, Indiaj J. pure appl. Math., 19 (1988) 11011104.
[6]. S.Sidon,Hinreichende be dingeingen fur den Fouriercharacter einer trigonometris
[7]. S. A. Teljakovski, A sufficient condition for Sidon for the integrability of trigonometric series, Mat. Zametki, 14
(1973), 317328.
[8]. W.H.Young. On the Fourier series of bounded functions, Proc. London Math. Soc., 12 (1913), 4170.