1 Introduction

There are many applications with well-known Hardy’s inequality in analytics, which refers to the following: let a k >0 (k=1,2,), p>1, then

( p p 1 ) p n = 1 a n p > n = 1 ( 1 n k = 1 n a k ) p .
(1.1)

In recent decades, there have also been many results due to the extension and refinement of this inequality (cf. [15]), especially the monograph [6], which summarized part of the research done before 2005. In research on the coefficient of (1.1), the following conclusion in the case p=2 was drawn in [6]:

4 n = 1 ( 1 1 3 n + 5 ) a n 2 > n = 1 ( 1 n k = 1 n a k ) 2 .
(1.2)

In [7], by using the method of weight-coefficient, the following inequality is proved with p[ 7 6 ,2]:

( p p 1 ) p n = 1 ( 1 15 196 1 n 1 1 / p + 3 , 436 ) a n p > n = 1 ( 1 n k = 1 n a k ) p .
(1.3)

In the following section, let p>1 and

Z p ={ p 1 ( p 1 ) 2 p 2 1 p , 1 < p 2 , 1 ( p 1 p ) p 1 2 p 1 p , p > 2 .

We shall strengthen Hardy’s inequality to

( p 1 p ) p n = 1 ( 1 Z p 2 ( n 1 2 ) 1 1 p ) a n p > n = 1 ( 1 n k = 1 n a k ) p .

2 Relevant lemmas

Some characters of a convex function will be cited in this section.

Definition 2.1 LR is an interval, and f:IR is continuous. If

f ( x + y 2 ) () f ( x ) + f ( y ) 2

holds for all x,yI, then f is called a convex (concave) function.

The sufficient and necessary condition for a second-order differential function f to be convex (concave) function is that f (x)()0 always holds for any xI. The famous Hadamard inequality is as follows. Let f be a convex (concave) function on [a,b], then the equality

f ( b + a 2 ) () 1 b a a b f(x)dx() f ( b ) + f ( a ) 2
(2.1)

holds if and only if f is a linear function.

Lemma 2.1 Let p>1, and Z p is defined in the first section, then 0< Z p < 1 2 .

Proof The proof includes two parts.

Part 1: When 1<p2, if we can obtain f(p):= 2 1 p 1+ 1 p >0, then we can get Z p >0 obviously. Since

f (p)= 1 p 2 2 1 p ln2 1 p 2 = 1 p 2 2 1 p ( ln 2 2 1 p ) <0,

f(p) is monotone decreasing on (1,2], then f(p)f(2)= 2 2 1 2 >0.

Meanwhile, Z p < 1 2 is equivalent to

p 3 2 < ( p 1 ) 2 p 2 1 p , 3 2 < ( p 2 + 1 p ) 2 1 p p.

The above two inequalities obviously hold with 1<p< 3 2 . If 3 2 <p<2, then p2+ 1 p is increasing about p, 2 1 p and −p are decreasing in relation to p. Then

Part 2: When p>2, then it should be proved that ( p 1 p ) p 1 2 p 1 p <1 and ( p 1 p ) p 1 2 p 1 p > 1 2 are respectively equivalent to 2< ( 1 + 1 p 1 ) p and ( 1 + 1 p 1 ) p < 2 2 + 1 p 1 . Since ( 1 + 1 p 1 ) p is strictly decreasing for p(2,+), then it is proved that

( 1 + 1 p 1 ) p > lim p ( 1 + 1 p 1 ) p =e>2

and

( 1 + 1 p 1 ) p < ( 1 + 1 2 1 ) 2 =4< 2 2 + 1 p 1 .

The proof of Lemma 2.1 is completed. □

Lemma 2.2

  1. (i)

    Let x1, p>2, then

    ( 1 Z p x 1 p 1 ) 2 p p 1 ( 1 p 2 p 1 Z p x 1 p 1 ) >1.
    (2.2)
  2. (ii)

    Let p>1, then f:x[1,+) x 1 p 2 Z p x 2 p 3 is a convex function.

Proof (i) The proposition is equivalent to

1 p 2 p 1 Z p x 1 p 1 > ( 1 Z p x 1 p 1 ) p 2 p 1 .

Bernoulli’s inequality refers to the following: ( 1 + t ) α 1+αt (t>1, 0<α<1) holds if t=0. If t= Z p x 1 + 1 p , α= p 2 p 1 , then formula (2.2) holds.

  1. (ii)
    f ( x ) = ( 1 p 2 ) ( 1 p 3 ) x 1 p 4 Z p ( 2 p 3 ) ( 2 p 4 ) x 2 p 5 = ( 2 p 1 ) x 2 p 5 p 2 [ ( 3 p 1 ) x 1 1 p 2 Z p ( 3 p 2 ) ] ( 2 p 1 ) x 2 p 5 p 2 [ ( 3 p 1 ) 2 Z p ( 3 p 2 ) ] .

According to Lemma 2.1,

f (x) ( 2 p 1 ) x 2 p 5 p 2 [ ( 3 p 1 ) ( 3 p 2 ) ] >0,

so, f is a convex function. The proof of Lemma 2.2 is completed. □

Lemma 2.3 Let n be a positive natural number, p>1.

  1. (i)

    If 1<p2, then

    k = 1 n 1 ( k 1 2 ) 1 p p p 1 ( n 1 1 p Z p p 1 ) .
    (2.3)
  2. (ii)

    If p>2, then

    k = 1 n 1 ( k 1 2 ) 1 p p p 1 n 1 2 p ( n 1 1 p Z p ) 1 p 1 .
    (2.4)

Proof (i) If n=1, inequality (2.3) is proved easily. Assume that when n=m1, the following equality holds for n=m+1:

k = 1 m + 1 1 ( k 1 2 ) 1 p p p 1 ( m 1 1 p Z p p 1 ) + 1 ( m + 1 2 ) 1 p = p p 1 [ ( m + 1 ) 1 1 p Z p p 1 ] m m + 1 x 1 p d x + 1 ( m + 1 2 ) 1 p .

Because x 1 p is a convex function on [m,m+1], then according to Hadamard’s inequality of a convex function, formula (2.3) also holds if n=m+1.

  1. (ii)

    If n=1, inequality (2.4) is proved easily. Assume that when n=m1, the inequality holds. For n=m+1,

    k = 1 m + 1 1 ( k 1 2 ) 1 p p p 1 m 1 2 p ( m 1 1 p Z p ) 1 p 1 + 1 ( m + 1 2 ) 1 p = p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 m m + 1 [ p p 1 x 1 2 p ( x 1 1 p Z p ) 1 p 1 ] d x + 1 ( m + 1 2 ) 1 p = p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 + 1 ( m + 1 2 ) 1 p p p 1 m m + 1 [ ( 1 2 p ) x 2 p ( x 1 1 p Z p ) 1 p 1 + 1 p 1 ( 1 1 p ) x 1 2 p ( x 1 1 p Z p ) 1 p 1 1 x 1 p ] d x = p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 + 1 ( m + 1 2 ) 1 p p p 1 m m + 1 x 2 p ( x 1 1 p Z p ) 2 p p 1 × [ ( 1 1 p ) x 1 1 p ( 1 2 p ) Z p ] d x = p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 + 1 ( m + 1 2 ) 1 p m m + 1 x 1 p ( 1 Z p x 1 + 1 p ) 2 p p 1 [ 1 p 2 p 1 Z p x 1 + 1 p ] d x .

By inequality (2.2) and the fact that x 1 p is a convex function on [m,m+1], we have

k = 1 m + 1 1 ( k 1 2 ) 1 p < p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 + 1 ( m + 1 2 ) 1 p m m + 1 x 1 p d x p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 + 1 ( m + 1 2 ) 1 p ( m + 1 2 ) 1 p = p p 1 ( m + 1 ) 1 2 p ( ( m + 1 ) 1 1 p Z p ) 1 p 1 .

Thus, the inequality (2.4) also holds if n=m+1. □

Lemma 2.4 Let i be any positive natural number and p>1, then

n = i ( n 2 + 1 p Z p n 3 + 2 p ) < p p 1 [ ( i 1 2 ) 1 p 1 Z p 2 ( i 1 2 ) 2 p 2 ] .

Proof Let i=1,2, and

f(i)= p p 1 [ ( i 1 2 ) 1 p 1 Z p 2 ( i 1 2 ) 2 p 2 ] n = i ( n 1 p 2 Z p n 2 p 3 ) .

Then

f ( i ) f ( i + 1 ) = p p 1 [ ( i 1 2 ) 1 p 1 Z p 2 ( i 1 2 ) 2 p 2 ] ( i 1 p 2 Z p i 2 p 3 ) p p 1 [ ( i + 1 2 ) 1 p 1 Z p 2 ( i + 1 2 ) 2 p 2 ] = p p 1 i 1 2 i + 1 2 ( x 1 p 1 Z p 2 x 2 p 2 ) d x ( i 1 p 2 Z p i 2 p 3 ) = i 1 2 i + 1 2 ( x 1 p 2 Z p x 2 p 3 ) d x ( i 1 p 2 Z p i 2 p 3 ) .

According to the conclusion of Lemma 2.2 and Hadamard’s inequality of a convex function, the sequence { f ( i ) } i = 1 + is a strictly decreasing sequence. It is also known that lim i + f(i)=0, then f(i)>0 always holds. The proof of Lemma 2.4 is completed. □

3 A new strengthened version of Hardy’s inequality

Theorem 3.1 Let a k >0, n1, nN, p>1, and

Z p ={ p 1 ( p 1 ) 2 p 2 1 p , 1 < p 2 , 1 ( p 1 p ) p 1 2 p 1 p , p > 2 ,

then

n = 1 ( 1 n k = 1 n a k ) p < ( p p 1 ) p n = 1 [ 1 Z p 2 ( n 1 2 ) 1 1 p ] a n p .
(3.1)

Proof Let r= p p 1 , then 1 r + 1 p =1. According to Holder’s inequality,

( 1 n k = 1 n a k ) p = 1 n p ( k = 1 n ( k 1 2 ) p 1 p 2 ( k 1 2 ) p 1 p 2 a k ) p 1 n p ( k = 1 n ( k 1 2 ) p 1 p 2 r ) p r ( k = 1 n ( k 1 2 ) p 1 p 2 p a k p ) p p = 1 n p ( k = 1 n ( k 1 2 ) 1 p ) p 1 k = 1 n ( k 1 2 ) p 1 p a k p .

If λ n = 1 n p ( k = 1 n ( k 1 2 ) 1 p ) p 1 and μ k = ( k 1 2 ) p 1 p , then

n = 1 ( 1 n k = 1 n a k ) p n = 1 λ n k = 1 n μ k a k p = n = 1 ( k = n λ k ) μ n a n p .
(3.2)

Thus,

k = n λ k = k = n 1 k p ( j = 1 k ( j 1 2 ) 1 p ) p 1 .

If 1<p2, according to formula (2.3) and Bernoulli’s inequality,

k = n λ k = ( p p 1 ) p 1 k = n 1 k p ( k 1 1 p Z p p 1 ) p 1 = ( p p 1 ) p 1 k = n k 2 + 1 p ( 1 Z p p 1 k 1 + 1 p ) p 1 < ( p p 1 ) p 1 k = n k 2 + 1 p ( 1 Z p k 1 + 1 p ) = ( p p 1 ) p 1 k = n ( k 2 + 1 p Z p k 3 + 2 p ) .

If p>2, by using formula (2.4), we obtain

k = n λ k ( p p 1 ) p 1 k = n 1 k p ( k 1 2 p ( k 1 1 p Z p ) 1 p 1 ) p 1 = ( p p 1 ) p 1 k = n ( k 2 + 1 p Z p k 3 + 2 p ) .

Therefore, for any p>1, the following inequality holds:

k = n λ k ( p p 1 ) p 1 k = n ( k 2 + 1 p Z p k 3 + 2 p ) .

By Lemma 2.4, we get

n = k λ k ( p p 1 ) p ( ( n 1 2 ) 1 p 1 Z p 2 ( n 1 2 ) 2 p 2 ) .
(3.3)

So, from inequalities (3.2) and (3.3), the following result can be obtained:

n = 1 ( 1 n k = 1 n a k ) p ( p p 1 ) p n = 1 ( ( n 1 2 ) 1 p 1 Z p 2 ( n 1 2 ) 2 p 2 ) ( n 1 2 ) p 1 p a n p = ( p p 1 ) p n = 1 ( 1 Z p 2 ( n 1 2 ) 1 1 p ) a n p .

The proof of Theorem 3.1 is completed. □