Korean J. Math. 22 (2014), No. 3, pp.

407417
http://dx.doi.org/10.11568/kjm.2014.22.3.407

SOME PROOFS OF THE CLASSICAL INTEGRAL

HARDY INEQUALITY
Iddrisu Mohammed Muniru , Okpoti Christopher Adjei,
and Gbolagade Kazeem Alagbe
Abstract. We present some proofs of the classical integral Hardy
inequality. Our approach makes use of continuous functions with
compact support in (0, ), homogeneity of the norm and Schurs
criterion for integral operators.

1. Introduction
The classical integral inequality announced by G. H. Hardy in 1920
is given by
p

p Z
Z  Z x
p
1
f (t)dt dx
f p (x)dx,
(1)
x 0
p1
0
0
where p > 1, x > 0, f is a nonnegative measurable function on (0, )
p
p
is the best possible [4]. This interesting reand the constant p1
sult (1) was later proved by Hardy himself in 1925 (see [1], [5], [7], [8],
[9] and the references therein.) Inequality (1) can also be written as

p Z
Z
p
p
(2)
F (x)dx
f p (x)dx,
p

1
0
0
Received May 5, 2014. Revised July 25, 2014. Accepted July 25, 2014.
2010 Mathematics Subject Classification: 26D10, 26D15.
Key words and phrases: Hardy inequality, Proofs, homogeneity, integral
operators.

Corresponding author.
c The Kangwon-Kyungki Mathematical Society, 2014.

This is an Open Access article distributed under the terms of the Creative commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by
-nc/3.0/) which permits unrestricted non-commercial use, distribution and reproduction in any medium, provided the original work is properly cited.

408

Iddrisu. M. M., Okpoti. C. A., and Gbolagade. K. A.

Rx
where 0 < F (x) = x1 0 f (t)dt < , f > 0.
The inequalities (1) and (2) are very popular in the research environment. See also [6].
Our task in this paper is mainly to deepen understanding of the Hardy
inequality (2) by providing elaborate proofs.
2. Preliminary Notes
We define continuous functions and present some auxilliary results.
Definition (Continuous functions) [11]. Let X be a subset of the set
of real numbers <, and let f : X < be a function. Let x0 X. We say
that f is continuous at x0 if and only if we have limxx0 f (x) = f (x0 )
for every x X. In other words, the limit of f (x) as x converges to x0
in X exists and is equal to f (x0 ).
Support of a function. Let I be a nonempty open set in <n , and let
f be a continuous function on I. The support of f , denoted by supp(f),
is defined to be the complement of the largest open set on which f is
zero. That is
supp(f) = {x I : f (x) 6= 0},
the closure of the set x I where f (x) 6= 0. (See [3], p. 134).
Fatous Lemma ([2], p. 52). Let X be a measure space with measure . Let {fn } be any sequence of measurable functions on X with
range in [0, ]. For each positive integer n,
Z
Z
(3)
(lim inf fn )d lim inf
fn d.
X

H
olders inequality ([2], p.182). Suppose 1 < p < and p1 + 1q = 1
(that is p + q = pq). Let X be a measure space with measure . If f
and g are measurable functions on X with range in [0, ], then
Z
 p1 Z
 1q
Z
p
q
(4)
f gd
f d
g d ,
X

Some proofs of the Classical Integral Hardy Inequality

409

In particular, if f Lp and g Lq , then f g L1 , and in this case

equality holds in (4) if and only if |f |p = |g|q almost everywhere for
some constants , with 6= 0. See [2] and also [10] for proofs of
inequalities (3) and (4).
Proposition 1. ([2], p.195). For f Lp and g Lq , let
Z +
Z x
1
x1 g(x)dx.
f (y)dy, Sg(y) =
T f (x) = x
y

Then for 1 < p and 1 q < ,

p
kT f kp
kf kp ,
p1

kSgkq qkgkq .

Proof. Let
K(x, y) =

1
x

0 otherwise .

if

0<y<x

Then
Z

|K(1, y)|y

p1

Z
dy =

where

1
p

1
q

y p dy =

p
= q,
p1

= 1, yielding the result.

We now present our main results which are basically the different
approaches to the proof of inequality (2). We denote by Cc (0, +), the
set of all continuous functions with compact support in (0, +).

3. First Proof
Integration by parts and Holders inequality are essentially applied
here.
3.1. Case 1: Let p > 1, f is positive and continuous with compact
support in (0, +) and F is positive and differentiable on [0, +).
Setting u = F p and dv = dx implies du = pF p1 F 0 dx and v = x.

410

Iddrisu. M. M., Okpoti. C. A., and Gbolagade. K. A.

Consider (a, A0 ) with 0 < a < A0 < so that the suppf [a, A0 ].
RA
Integration by parts of a 0 F p dx gives
Z A0
Z A0
p
p A0
xF p1 F 0 dx.
F dx = [xF ]a p
(5)
a

RA

R A0

Ra

RA

But 0 F p dx = 0 F p dx + a F p dx + A0 F p dx for 0 < A0 < A < .

Since f Cc (0, +) with suppf [a, A0 ] then F (x) = 0 for 0 x < a
and A0 < x A. Thus
Z A
Z A0
p
F p (x)dx
F (x)dx =
0

and (5) becomes

Z A
Z
p
p
(6)
F (x)dx = AF (A) p
0

xF p1 (x)F 0 (x)dx.

R
1 x

Rx
But F (x) = x 0 f (t)dt implies xF (x) = 0 f (t)dt. Differentiating gives
xF 0 (x) + F (x) = f (x)
Thus (6) becomes
Z A
Z A
p
p
F p1 (x){f (x) F (x)}dx
F (x)dx = AF (A) p
0
0
Z A
Z A
p1
p
F p (x)dx,
F (x)f (x)dx + p
= AF (A) p
0

0
A

AF p (A)
p
F (x)dx =
+
1p
p1
p

(7)
0

F p1 (x)f (x)dx.

By Holders inequality,
Z
(8)

F p1 (x)f (x)dx

0
1
p

Z
0

1
q

p
.
p1

F (p1)q (x)dx

 1q Z

f p (x)dx

 p1

where + = 1 or q =
Putting (8) into (7), then
(9)
Z A
 1q Z A
 p1
Z A
p
AF p (A)
p
(p1)q
p
F (x)dx
+
F
(x)dx
f (x)dx .
1p
p1
0
0
0

Some proofs of the Classical Integral Hardy Inequality

RA
Let I = 0 f (t)dt so that F (A) =
as A . Thus (9) simplies to
Z

(10)
0

Since

R
0

p
F (x)dx
p1
p

F p (x)dx

(1 p1 )

Z

Z

I
.
A

Then AF p (A) =
(1 p1 ) Z

F (x)dx

Ip
Ap1

411

turns to 0

f (x)dx

 p1
.

> 0, then

 p1
F (x)dx

p
p1

Z

 p1
f (x)dx ,
p

Thus
Z

F (x)dx
0

p
p1

p Z

f p (x)dx.

3.2. Case 2: We consider f not necessarily positive. Let f Lp (0, +)

and set f+ = max(f, 0) and f = min(0, f ). Consider sequence
(fn )nN of functions in Cc (0, +)
such that fn f in Lp (0, +).
R
x
For a fixed x > 0, F (x) = x1 0 f (t)dt is well defined for f 0. By
Holders inequality,
 Z x

Z
1
 1 x


f (t)dt
|f (t)|dt
x
x 0
0
 p1 Z x
Z x
 1q
1
p
q
|f (t)| dt

1 dt
x
0
0
Z x
 p1
p1
p
x
|f (t)| dt .
0

Hence, by Fubini, we get


p Z
Z
p
p
|F (x)| dx
|(f (x)|p dx,
p

1
0
0
 p
p
where the constant p1
is the best possible.
Also set f = f+ f , |f | = f+ + f and F = F+ F , |F | = F+ + F

412

Iddrisu. M. M., Okpoti. C. A., and Gbolagade. K. A.

|F | < . Let kfn f k < n1 . Then

 Z x

Z x
1

1
|Fn (x) F (x)| = 
fn (t)dt
f (t)dt
x 0
x 0
Z x
1

|fn (t) f (t)| dt

x 0
Z x
 p1 Z x
 1q
1
|fn (t) f (t)|p dt

1q dt
x
0
0

and

by (4)

x p kfn f kp
1

x p
<
0 as n .
n
This shows pointwise convergence: |Fn (x)| |F (x)| as . Therefore

p Z
Z
p
p
(11)
|Fn (x)| dx
|fn (x)|p dx.
p

1
0
0
Suppose that f is positive and so is F . Suppose also that fn is positive
which implies Fn is positive. Then
Z +
Z +
p
lim |Fn (x)|p dx
|F (x)| dx =
n+
0
0
Z +
|Fn (x)|p dx (Fatous lemma)
lim
n+ 0


p Z +
p
p
lim
|fn (x)| dx
n+
p1
0
p Z +

p

|f (x)|p dx.
p1
0
4. Second Proof
The approach here makes use of homogeneity of a norm and the use
of a kernel.
4.1. Case 1: Homogeneity of a norm.
Let
Z
Z 1
1 x
f (t)dt =
f (tx)dt.
F (x) =
x 0
0

Some proofs of the Classical Integral Hardy Inequality

413

Set ft (x) = f (tx). By Minkowski inequality for integrals,

Z 1
kft (x)kp dt
kF (x)kp
0
1

Z

|ft (x)|p dx

 p1
dt

By change of variables with s = tx, ds = tdx, and also by Fubinis

theorem we have
 p1
Z 1 Z 1
p ds
kF (x)kp
|f (s)|
dt
t
0
0
 p1
Z 1
Z 1
p1
p
t dt

|f (s)| ds
0

kF (x)kp

p
kf (s)kp
p1

Hence
kF (x)kpp

p
p1

p

kf kpp .

4.2. Case 2: Kernels approach.

Let K > 0 be a Lebesgue measurable function on (0, ) x (0, ) and
define
1
if
0<y<x
x
K(x, y) =

0 otherwise .
Let f Lp and consider the integral operator
Z
T f (x) =
|K(x, y)f (y)|dy.
0

By change of variables with y = xz, dy = xdz, we have

Z
T f (x) =
|K(x, xz)|f (xz)xdz.
0

414

Iddrisu. M. M., Okpoti. C. A., and Gbolagade. K. A.

Set fz (x) = f (xz) and by Minkowski inequality for integrals, we have

Z
kT f (x)kp
|K(1, z)|kfz (x)kp dz
0

Z

|fz (x)|p dx

 p1
dz

Again by change of variables with xz = y, zdx = dy and by Fubinis

theorem, we have
 p1
Z 1 Z
p dy
|f (y)|
kT f (x)kp
dz
z
0
0
 p1
Z
Z 1
p1
p

z dz
|f (y)| dy
0
0


p

kf (y)kp
p1
Hence
kT f kpp

p
p1

p

kf kpp .

Remark 1. Let us remark that the proofs for the cases 1 and 2 here
can be described as almost the same, except that the mention of a kernel
and its application is demonstrated in case 2 .

5. Third Proof
Let K be a Lebesgue measurable function on (0, ) x (0, ). For
1 < p < , and p1 + 1q = 1, define a nonnegative kernel
1
if
0<y<x
x
K(x, y) =

0 otherwise .
and consider the integral operator
Z +
Z
1 x
(12)
T f (x) =
K(x, y)f (y)dy =
f (y)dy = F (x)
x 0
0

Some proofs of the Classical Integral Hardy Inequality

415

Let h(t) = t for 1 < < 0. By Proposition 1, we compute

Z

(13)

K(x, y)h(y)dy =
0

1
1 x(+1)
x
y dy =
=
x
x +1
+1

and
+

K(x, y)h(x)dx =

(14)

1
x dx =
x

Z
y

1
x(1) dx = y .

Introduce y into (12) and applying Holders inequality, we have

Z

f (y)
dy
y
0
 1q Z
Z +
q
K(x, y)y dy

K(x, y)y

T f (x) =

p  p1
|f (y)|
K(x, y)
dy
y
0
0
 q  1q Z +

p  p1
x
|f (y)|

K(x, y)
{by (13)}
dy
q + 1
y
0
 1q


p  p1
Z +
1
|f (y)|

x
K(x, y)
dy
q + 1
y
0

(15)

Let =

(16)

1
q+1

 1q

. Thus (15) becomes

p p

Z

|T f (x)| x


K(x, y)

|f (y)|
y

p


dy

By Fubinis theorem, we get

Z

(17)

p 
|f (y)|
x
K(x, y)
dy
dx
|T f (x)| dx
y
0
0

p
Z + Z +
|f (y)|
p
p

K(x, y)x dx
dy.
y
0
0
p

Z

416

Iddrisu. M. M., Okpoti. C. A., and Gbolagade. K. A.

From (14), T h(y) =

Z +
0

R +

K(x, y)xp dx = yp . Thus (17) becomes

Z +
y p |f (y)|p
p
p
|T f (x)| dx

dy
p y p
0
Z
p +
|f (y)|p dy

p 0
Z +
|f (y)|p dy
C
0

where C =

p
p

for 1 < < 0. Equivalently

kT f (x)kpP Ckf (y)kpp .

Remark 2. The Schur criterion discussed above shows that the operator T is bounded on Lp (0, +) with kT k C. See ([12], p. 45) for
discussions on Schur test.
6. Conclusion
Precise proofs for the classical integral Hardy inequality were presented. A number of useful applications of some important theorems
such as Fatous lemma and Fubinis theorem as well as Holder and
Minkowski inequalities were provided.

Acknowledgement
The first author wish to thank Emeritus Prof. Aline BONAMI and
Prof. Frederic SYMESAK for their care and attention during his research period at the University of Angers, France. Also, his deepest and
sincere appreciation goes to the French Government for the financial
support during the research period.
References
[1] S. Abramovich, K. Krulic, J. Pecaric and L.-E. Persson, Some new refined Hardy
type inequalities with general kernels and measures, Aequat. Math. 79 (2010),
157172.
[2] G. B. Folland, Real Analysis: Modern Techniques and Their Applications, 2nd
ed., John Willey and Sons, Inc., New York, 1999.

Some proofs of the Classical Integral Hardy Inequality

417

[3] C. Gasquet and P. Witomski, Fourier Analysis and Applications: Filtering,

Numerical Computation, Wavelets, Springer-Verlag New York, Inc., 1999.
[4] G. H. Hardy, Note on a theorem of Hilbert, Math. Z. 6 (1920), 314317.
[5] G. H. Hardy, J. E. Littlewood and G.Polya, Inequalities, 2nd ed., Cambridge
University Press, Cambridge, 1952.
[6] S. Kaijser, L. Nikolova, L.- E. Persson and A. Wedestig, Hardy-Type Inequalities
Via Convexity, J. Math. Ineq. Appl., 8 (3) (2005), 403417.
[7] A. Kufner, L. Maligranda and L.- E. Persson, The Hardy inequality: About its
History and Some Related Results, Vydavatelsky Sevis Publishing House, Pilsen,
2007.
[8] A. Kufner, L. Maligranda and L.-E. Persson, The Prehistory of the Hardy Inequality, Amer. Math. Monthly, 113 (8) (2006), 715732.
[9] J. A. Oguntuase and L.- E. Persson, Refinement of Hardy inequalities via superquadratic and subquadratic functions, J. Math. Anal. Appl., 339 (2) (2008),
13051312.
[10] W. Rudin, Real and Complex Analysis, 2nd edition, McGraw-Hill Inc, 1974.
[11] T. Tao, Analysis I, Hindustan Book Agency, Volume I, 2006.
[12] K. Zhu, Spaces of Holomorphic Functions in the Unit Ball, Springer, NewYork,
2005.

Iddrisu Mohammed Muniru

Department of Mathematics
University for Development Studies
Navrongo Campus
P. O. Box 24, Navrongo, UE/R, Ghana
E-mail : immuniru@gmail.com
Okpoti Christopher Adjei
Department of Mathematics
University of Education
Winneba, Ghana
E-mail : caokpoti@yahoo.com
Gbolagade Kazeem Alagbe
Department of Computer Science
University for Development Studies
Navrongo Campus
P. O. Box 24, Navrongo, UE/R, Ghana
E-mail : gkazy1@yahoo.com