On the best possible remaining term in the Hardy inequality

Nassif Ghoussoub∗ and Amir Moradifam

Department of Mathematics, University of British Columbia,
Vancouver BC Canada V6T 1Z2
nassif@math.ubc.ca
a.moradi@math.ubc.ca

April 16, 2007

1 Abstract
We give a necessary and sufficient condition on a radially symmetric potential V on Ω that makes it an
admissible candidate for an improved Hardy inequality of the following form:
R |u|2
|∇u|2 dx − ( n−2 2
dx ≥ c Ω V (|x|)|u|2 dx for all u ∈ H01 (Ω).
R R
Ω 2 ) Ω |x|2
(1)

A characterization of the best possible constant c(V ) is also given. This result yields easily the improved
Hardy’s inequalities of Brezis-Vázquez [6], Adimurthi et al. [1], and Filippas-Tertikas [12], as well as the cor-
responding best constants. Our approach clarifies the issue behind the lack of an optimal improvement, while
yielding other interesting “dual” inequalities. Another consequence is the following substantial
Rr sharpening
of known integrability criteria: If a positive radial function V satisfies lim inf r→0 ln(r) 0 sV (s)ds > −∞,
then there exists ρ := ρ(Ω) > 0 such that the improved Hardy inequality (1) holds for the scaled potential
Rr
Vρ (x) = V ( |x|
ρ ). On the other hand, if limr→0 ln(r) 0 sV (s)ds = −∞, then there is no ρ > 0 for which (1)
holds for Vρ . This shows for example, that V (x) = |x|1α is an admissible potential for an improved Hardy
inequality when α < 2, while it is not so for α ≥ 2. All these results have immediate applications to the
corresponding Schrödinger equations. Analogous criteria for inequalities involving the bi-Laplacian will be
developed in a forthcoming paper [14].

2 Introduction
Let Ω be a bounded domain in Rn , n ≥ 3, with 0 ∈ Ω. The classical Hardy inequality asserts that
R |u|2
|∇u|2 dx ≥ ( n−2 2
dx for all u ∈ H01 (Ω).
R
Ω 2 ) Ω |x|2
(2)

This inequality and its various improvements are used in many contexts, such as in the study of the stability
of solutions of semi-linear elliptic and parabolic equations [6, 7, 22], in the analysis of the asymptotic behavior
of the heat equation with singular potentials [8, 23], as well as in the study of the stability of eigenvalues in
elliptic problems such as Schrödinger operators [11, 13].
Now it is well known that ( n−2 2
2 ) is the best constant for inequality (2), and that this constant is however not
1
attained in H0 (Ω). So, one could anticipate improving this inequality by adding a non-negative correction
term to the right hand side of (2) and indeed, several sharpened Hardy inequalities have been established in
recent years [4, 5, 12, 13, 23], mostly triggered by the following improvement of Brezis and Vázquez [6].
R |u|2
|∇u|2 dx ≥ ( n−2 2
dx + λΩ Ω |u|2 dx for every u ∈ H01 (Ω).
R R
Ω 2 ) Ω |x|2
(3)
∗ Partially supported by a grant from the Natural Sciences and Engineering Research Council of Canada.

1
The constant λΩ in (3) is given by
2
λΩ = z02 ωn2/n |Ω|− n , (4)
where ωn and |Ω| denote the volume of the unit ball and Ω respectively, and z0 = 2.4048... is the first zero of
the bessel function J0 (z). Moreover, λΩ is optimal when Ω is a ball, but is –again– not achieved in H01 (Ω).
This led to one of the open problems mentioned in [6] (Problem 2), which is whether the two terms on
the RHS of inequality (3) (i.e., the coefficients of |u|2 ) are just the first two terms of an infinite series of
correcting terms.
This question was addressed by several authors. In particular, Adimurthi et all [1] proved that for every
integer k, there exists a constant c depending on n, k and Ω such that
|u|2 Pk R |u|2 Qj ρ

−2
|∇u|2 dx ≥ ( n−2 2
log (i) |x| for u ∈ H01 (Ω),
R R
Ω 2 ) Ω |x|2
dx +c j=1 Ω |x|2 i=1 dx (5)
e(k−times)
..
ee
where ρ = (supx∈Ω |x|)(e ). Here we have used the notations log (1) (.) = log(.) and log (k) (.) =
(k−1)
log(log (.)) for k ≥ 2.
Also motivated by the question of Brezis and Vázquez, Filippas and Tertikas proved in [12] that the inequality
can be repeatedly improved by adding to the right hand side specific potentials which lead to an infinite
series expansion of Hardy’s inequality. More precisely, by defining iteratively the following functions,

X1 (t) = (1 − log(t))−1 , Xk (t) = X1 (Xk−1 (t)) k = 2, 3, ...,

they prove that for any D ≥ supx∈Ω |x|, the following inequality holds for any u ∈ H01 (Ω):
∞
n−2 2 |u|2 1X 1 |x| |x| |x|
Z Z Z
2
|∇u| dx ≥ ( ) dx + X 2 ( )X 2 ( )...Xi2 ( )|u|2 dx. (6)
Ω 2 Ω |x|2 4 i=1 Ω |x|2 1 D 2 D D

Moreover, they proved that the constant 14 is the best constant for the corresponding k−improved Hardy
inequality which is again not attained in H01 (Ω).
In this paper, we show that all the above results –and more– follow from a specific characterization of those
potentials V that yield an improved Hardy inequality. Here is our main result.

Theorem 2.1 Let V be a radial function on a smooth bounded radial domain Ω in Rn with radius R, in
such a way that V (x) = v(|x|) for some non-negative function v on (0, R). The following properties are then
equivalent:

1. The ordinary differential equation

y 0 (r)
(DV ) y 00 (r) + r + v(r)y(r) = 0

has a positive solution on the interval (0, R).

2. The following improved Hardy inequality holds
|u|2
|∇u|2 dx − ( n−2 2
V (|x|)|u|2 dx for u ∈ H01 (Ω).
R R R
(HV ) Ω 2 ) Ω |x|2
dx ≥ Ω


Moreover, the best constant c(V ) := sup c; (HcV ) holds can be characterized by the formula
y 0 (r)
c(V ) = sup c; y 00 (r) +


r + cv(r)y(r) = 0 has a positive solution on the interval 0, R . (7)

We note that the implication 1) implies 2) holds for any smooth bounded domain Ω in Rn containing 0,
provided v(r) + ( n−2 2 1
2 ) r 2 is non-increasing on (0, supx∈Ω |x|) and R is the radius of the ball which has the
same volume as Ω (i.e. R = ( |Ω|
1
ωn ) ).
n

It is therefore clear from the above discussion that in order to find what potentials are candidates for an
0
improved Hardy inequality, one needs to investigate the ordinary differential equation y 00 + yr + v(r)y(r) = 0.
We shall see that the results of Brezis-Vázquez, Adimurthi et al, and Filippas-Tertikas mentioned above can

2
be easily deduced by simply checking that the potentials V they consider, correspond to equations (DV )
where an explicit positive solution can be found.
Our approach turned out to be also useful for determining the best constants in the above mentioned
improvements. Indeed, the case when V ≡ 1 will follow immediately from Theorem 2.1. A slightly more
involved reasoning – but also based of the above characterization – will allow us to recover the best one
established by Filippas-Tertikas, and to show that the best constant in the improvement of Adimurthi et
al, is actually equal to 14 , which was conjectured by Chaudhuri in a recent paper [9] where he shows the
estimate 41 ≤ c ≤ 12 .
Since the existence of positive solutions for ODEs of the form (DV ) is closely related to the oscillatory
properties of second order equations of the form z 00 (s) + a(s)z(s) = 0, Theorem 2.1 also allows for the use of
the extensive literature on the oscillatory properties of such equations to deduce various interesting results
such as the following corollary.

Corollary 2.2 Let V be a positive radial function on a smooth bounded radial domain Ω in Rn .
Rr
1. If lim inf r→0 ln(r) 0 sV (s)ds > −∞, then there exists α := α(Ω) > 0 such that an improved Hardy
inequality (HVα ) holds for the scaled potential Vα (x) := α2 V (αx).
Rr
2. If limr→0 ln(r) 0 sV (s)ds = −∞, then there are no β, c > 0, for which (HVβ,c ) holds with Vβ,c =
cV (βx).

The following is a consequence of the two results above.

1
Corollary 2.3 For any α < 2, inequality (HcV ) holds on a bounded domain Ω for Vα (x) = |x|α and some
c > 0. Moreover, the best constant c( |x|1α ) is equal to the largest c such that the equation

1 1
y 00 (r) + y 0 (r) + c α y(r) = 0,
r r
has a positive solution on (0, R), where R is the radius of the ball wich has the same volume as Ω. Moreover,
if α ≥ 2 inequality (HV ) does not hold for Vα,c (x) = c |x|1α for any c > 0.

Note that the above corollary gives another proof of the fact that ( n−2 2
2 ) is the best constant for the classical
Hardy inequality.
Define now the class

AΩ = {v : R → R+ ;v is non-increasing on (0, supx∈Ω |x|), Dv has a positive solution on (0, ( |Ω|

1
ωn ) )}.
n

An immediate application of Theorem 2.1 coupled with Hölder’s inequality gives the following duality state-
ment, which should be compared to inequalities dual to those of Sobolev, recently obtained via the theory
of mass transport [2, 10].

Corollary 2.4 Suppose that Ω is a smooth bounded domain in Rn containing 0. Then for any 0 < p ≤ 2,
we have
 
2
Z 
n−2 2 |u| 1
Z  
inf |∇u|2 dx − ( ) 2
dx; u ∈ H 1
0 (Ω), ||u||p = 1 ≥ sup ; v ∈ A Ω . (8)
Ω 2 Ω |x|  ||v −1 || p . 
p−2

Finally, consider the following classes of radial potentials:

Z r
+
X = {V : Ω → R ; V ∈ L∞
loc (Ω \ {0}), lim inf ln(r) sV (s)ds > −∞}, (9)
r→0 0

and Z r
Y = {V : Ω → R+ ; V ∈ L∞
loc (Ω \ {0}), lim ln(r) sV (s)ds = −∞}. (10)
r→0 0

3
 (n−2) 2
For any 0 < µ < µn := 2 ) we consider the following weighted eigenvalue problem,
 µ
−∆u − |x| 2u = λV u in Ω,
(EV,µ ) (11)
u = 0 on Ω.

Our results above combine with standard arguments to yield the following.

Corollary 2.5 For any 0 < µ < µn , and V : Ω → R+ with V ∈ L∞ 2

loc (Ω \ {0}) and lim|x|→0 |x| V (x) = 0, the
problem (EV,µ ) admits a positive weak solution uµ ∈ H0 (Ω) corresponding to the first eigenvalue λ = λ1µ (V ).
1

Moreover, by letting λ1 (V ) = limµ↑µn λ1µ (V ), we have

• If V ∈ X, then there exists c > o such that λ1 (Vc ) > 0.

• If V ∈ Y , then λ1 (Vc ) = 0 for all c > 0,

where Vc (x) := V (cx).

3 Two dimensional inequalities

In this section, we start by establishing the following improvements of “two-dimensional” Poincaré and
Poincaré-Wirtinger inequalities.

Theorem 3.1 Let a < b, k is a differentiable function on (a, b), and ϕ be a strictly positive real valued
differentiable function on (a, b). Then, every h ∈ C 1 ([a, b]) with

ϕ0 (r) ϕ0 (r)
−∞ < lim k(r)|h(r)|2 = lim k(r)|h(r)|2 < ∞, (12)
r→a ϕ(r) r→b ϕ(r)
satisfies the following inequality:
Z b b
k0 (r)ϕ0 (r) + k(r)ϕ00 (r)
Z
0 2
|h (r)| k(r)dr ≥ −|h(r)|2 ( )dr. (13)
a a ϕ(r)

Moreover, assuming (12), the equality holds if and only if h(r) = ϕ(r) for all r ∈ (a, b).

Proof. Define ψ(r) = h(r)/ϕ(r), r ∈ [a, b]. Then

Z b Z b Z b Z b
|h0 (r)|2 k(r)dr = |ψ(r)|2 |ϕ0 (r)|2 k(r)dr + 2ϕ(r)ϕ0 (r)ψ(r)ψ 0 (r)k(r)dr + |ϕ(r)|2 |ψ 0 (r)|2 k(r)dr
a a a a
Z b Z b Z b
2 0 2 2 0 0
= |ψ(r)| |ϕ (r)| k(r)dr − |ψ(r)| (kϕϕ ) (r)dr + |ϕ(r)|2 |ψ 0 (r)|2 k(r)dr
a a a
Z b Z b
= |ψ(r)|2 (|ϕ0 (r)|2 k(r) − (kϕϕ0 )0 (r)dr + |ϕ(r)|2 |ψ 0 (r)|2 k(r)dr.
a a

Hence, we have
Z b b b
k0 (r)ϕ0 (r) + k(r)ϕ00 (r)
Z Z
|h0 (r)|2 k(r)dr = −|h(r)|2 ( )dr + |ϕ(r)|2 |ψ 0 (r)|2 k(r)dr (14)
a a ϕ a
b
k0 (r)ϕ0 (r) + k(r)ϕ00 (r)
Z
≥ −|h(r)|2 ( )dr. (15)
a ϕ(r)

Hence (13) holds. Note that the last inequality is obviously an idendity if and only if h(r) = ϕ(r) for all
r ∈ (a, b). The proof is complete. 

By applying Theorem 3.1 to the weight k(r) = r, we obtain the following generalization of the 2-dimensional
Poincaré inequality.

4
Corollary 3.2 (Generalized 2-dimensional Poincaré inequality) Let 0 ≤ a < b and ϕ be a strictly positive
real valued differentiable function on (a, b). Then every h ∈ C 1 ([a, b]) with
ϕ0 (r) ϕ0 (r)
−∞ < lim r|h(r)|2 = lim r|h(r)|2 < ∞, (16)
r→a ϕ(r) r→b ϕ(r)
satisfies the following inequality:
Z b b
ϕ0 (r) + rϕ00 (r)
Z
|h0 (r)|2 rdr ≥ −|h(r)|2 ( )dr. (17)
a a ϕ(r)
Moreover, under the assumption (16) the equality holds if and only if h(r) = ϕ(r) for all r ∈ (a, b).
By applying Theorem 3.1 to the weight k(r) = 1, we obtain the following generalization of the 2-dimensional
Poincaré-Wirtinger inequality.
Corollary 3.3 (Generalized Poincaré-Wirtinger inequality) Let a < b and ϕ be a strictly positive real valued
differentiable function on (a, b). Then, every h ∈ C 1 ([a, b]) with
ϕ0 (r) ϕ0 (r)
−∞ < lim |h(r)|2 = lim |h(r)|2 < ∞, (18)
r→a ϕ(r) r→b ϕ(r)
satisfies the following inequality:
b b
ϕ00 (r)
Z Z
|h0 (r)|2 dr ≥ −|h(r)|2 dr. (19)
a a ϕ(r)
Moreover, under assumption (18), the equality holds if and only if h(r) = ϕ(r) for all r ∈ (a, b).
Remark 3.4 Note that all of inequalities presented in the above theorems hold when we replace the condtions
(12), (16), and (18) with the following weaker conditions
ϕ0 (r) ϕ0 (r)
lim inf k(r)|h(r)|2 ≥ lim sup k(r)|h(r)|2 ,
r→b ϕ(r) r→a ϕ(r)
0
ϕ (r) ϕ0 (r)
lim inf r|h(r)|2 ≥ lim sup r|h(r)|2 ,
r→b ϕ(r) r→a ϕ(r)
0
ϕ (r) ϕ0 (r)
lim inf |h(r)|2 ≥ lim sup |h(r)|2 ,
r→b ϕ(r) r→a ϕ(r)
respectively, provided both sides in the above inequalities are not equal to −∞ or +∞.

4 Proof of the main theorem

We start with the sufficient condition of Theorem 2.1 by establishing the following.
Proposition 4.1 (Improved Hardy Inequality) Let Ω be a bounded smooth domain in Rn with 0 ∈ Ω, and
set R = (|Ω|/ωn )1/n . Suppose V is a radially symmetric function on Ω and ϕ is a C 2 -function on (0, R)
such that 0 00
0 ≤ V (|x|) ≤ − ϕ (|x|)+rϕ
|x|ϕ(|x|)
(|x|)
for all x ∈ Ω, 0 < |x| < R, (20)
0 0
lim inf r ϕϕ(r)
(r)
≥0 and lim sup ϕϕ(r)
(r)
< ∞, (21)
r→0 r→R

( n−2 2 1
2 ) |x|2 + V (|x|) is a decreasing function of |x|. (22)
Then for any u ∈ H01 (Ω), we have
n−2 2 |u|2
Z Z Z
|∇u|2 dx ≥ ( ) 2
dx + V (|x|)|u|2 dx. (23)
Ω 2 Ω |x| Ω

Moreover, if limr→0 rϕ(r)ϕ0 (r) = limr→R ϕ(r)ϕ0 (r) = 0, then equality holds if and only if u is a radial
function on Ω such that u(x) = ϕ(|x|) for all x ∈ Ω.

5
Proof: We first prove the inequality for smooth radial positive functions on the ball Ω = BR . For such
u ∈ C02 (BR ), we define
v(r) = u(r)r (n−2)/2 , r = |x|.
In view of Corollary 3.2, we can write
Z 2 R
n−2 2 u (x) n − 2 −n/2
Z Z
|∇u(x)|2 dx − ( ) 2
dx = ωn | r v(r) − r 1−n/2 v 0 (r)|2 r n−1 dr
Ω 2 Ω |x| 0 2
Z R 2
n−2 2 v (r)
− ( ) ωn dr
2 0 r
n−2 2 R 2 2v 0 (r)r 2 dr
Z
= ωn ( ) v ((1 − ) − 1)
2 0 (n − 2)v(r) r
Z R Z R
n−2
= ωn (v 0 (r))2 r − ωn ( ) v(r)v 0 (r)dr
0 2 0
Z R
= ωn (v 0 (r))2 r
0
R
ϕ0 (r) + rϕ00 (r)
Z
≥ ωn −v 2 (r)( )dr
0 ϕ(r)
R
ϕ0 (r) + rϕ00 (r) n−2
Z
= ωn −u2 (r)( )r dr
ϕ(r)
Z0
ϕ0 (|x|) + |x|ϕ00 (|x|)
= − u2 (x)( )dx.
Ω |x|ϕ(|x|)
Hence, the inequality (23) holds for radial smooth positive functions. By density arguments, inequality (23)
is valid for any u ∈ H01 (Ω). For general domain Ω, we use symmetrization arguments. Let BR be a ball
having the same volume as Ω with R = (|Ω|/ωn )1/n and let u∗ be the symmetric decreasing rearrangement of
the function u. Now note that for any u ∈ H01 (Ω), u∗ ∈ H01 (BR ). It is well known that the symmetrization
not change the Lp -norm, and that it decreases the Dirichlet energy, while increasing the integrals
Rdoes n−2
Ω
(( 2 )2 |x|1 2 + V (|x|)|u|2 dx, since the weight ( n−2 2 1
2 ) |x|2 + V (|x|) is a decreasing function of |x|. Hence,
(23) holds for any u ∈ H01 (Ω).
We shall need the following lemmas.

Lemma 4.1 Let x(r) be a function in C 1 (0, R] that is a solution of

rx0 (r) + x2 (r) ≤ −F (r), 0 < r ≤ R, (24)

where F is a nonnegative continuous function. Then

lim x(r) = 0. (25)

r↓0

Proof: Divide equation (24) by r and integrate once. Then we have

Z R Z R
|x(s)|2 F (s)
x(r) ≥ ds + x(1) + ds. (26)
r s r s
It follows that limr↓0 x(r) exists. In order to prove that this limit is zero, we claim that
Z R 2
x (s)
ds < ∞. (27)
r s
RR 2
Indeed, otherwise we have G(r) := r x s(s) ds → ∞ as r → 0. From (24) we have
R
F (s)
Z
1
(−rG0 (r)) 2 ≥ G(r) + x(1) + ds.
r s

6
Note that F ≥ 0, and G goes to infinity as r goes to zero. Thus, for r sufficiently small we have −rG0 (r) ≥
1 2 1 0 1 0
2 G (r) hence, ( G(r) ) ≥ 2 (ln(r)) , which contradicts the fact that G(r) goes to infinity as r tends to zero.
Thus, our claim is true and the limit in (25) is indeed zero. 
0
Lemma 4.2 If the equation ϕ00 + ϕr + v(r)ϕ = 0 has a positive solution on some interval (0, R), then we
have necessarily,
0 0
lim inf r ϕϕ(r)
(r)
≥ 0 and lim sup ϕϕ(r)
(r)
< ∞. (28)
r→0 r→R

Proof: Since ϕ(δ) ≥ 0 and ϕ(r) > 0 for 0 < r < δ, it is obvious that ϕ satisfies the second condition. To
0
obtain the first condition, set x(r) = r ϕϕ(r)
(r)
. one may verify that x(r) satisfies the ODE:

rx0 (r) + x2 (r) = −F (r), for 0 < r ≤ δ,

0
where F (r) = r 2 v(r) ≥ 0. By Lemma 4.1 we conclude that limr↓0 r ϕϕ(r)
(r)
= limr↓0 x(t) = 0. 

Lemma 4.3 Let V be positive radial potential on the ball Ω of radius R in Rn (n ≥ 3). Assume that
R  2

2 |u|
Ω
|∇u|2 − ( n−2
2 ) |x|2 − V (|x|)|u|
2
dx ≥ 0 for all u ∈ H01 (Ω).

Then there exists a C 2 -supersolution to the equation

 2
n−2 u
−∆u − − V (|x|)u = 0, in Ω, (29)
2 |x|2
u > 0 in Ω \ {0}, (30)
u = 0 in ∂Ω. (31)

Proof: Define
2
ψ|
|∇ψ|2 − ( n−2 )2 | |x| 2
R
2 − V |ψ|
λ1 (V ) := inf{
Ω
R2 ; ψ ∈ C0∞ (Ω \ {0})}.
|ψ|2
Ω

By our assumption λ(V ) ≥ 0. Let (ϕn , λn1 ) be the first eigenpair for the problem

(L − λ1 (V ) − λn1 )ϕn = 0 on Ω \ B R
n

ϕ(r) = 0 on ∂(Ω \ B R ),
n

where L = −∆ − ( n−2 2 1
2 ) |x|2 − V , and B R is a ball of radius R
n , n ≥ 2 . The eigenfunctions can be chosen
n
in such a way that ϕn > 0 on Ω \ B R and ϕn (b) = 1, for some b ∈ Ω with R 2 < |b| < R.
n
Note that λ1 ↓ 0 as n → ∞. Harnak’s inequality yields that for any compact subset K, max
n K ϕn
minK ϕn ≤ C(K)
with the later constant being independant of ϕn . Also standard elliptic estimates also yields that the family
(ϕn ) have also uniformly bounded derivatives on compact sets Ω − B R .
n
Therefore, there exists a subsequence (ϕnl2 )l2 of (ϕn )n such that (ϕnl2 )l2 converges to some ϕ2 ∈ C 2 (Ω \
B( R R
2 )). Now consider (ϕnl2 )l2 on Ω \ B( 3 ). Again there exists a subsequence (ϕnl3 )l3 of (ϕnl2 )l2 which
converges to ϕ3 ∈ C (Ω \ B( 3 )), and ϕ3 (x) = ϕ2 (x) for all x ∈ Ω \ B( R
2 R
2 ). By repeating this argument we
get a supersolution ϕ ∈ C 2 (Ω \ {0}) i.e. Lϕ ≥ 0, such that ϕ > 0 on Ω \ {0}. 

Lemma 4.4 Let a be a locally integrable function on R, then the following statements are equivalent.
1. z 00 (s) + a(s)z(s) = 0, has a strictly positive solution on (b, ∞).
2. There exists a function ψ ∈ C 1 (b, ∞) such that ψ 0 (r) + ψ 2 (r) + a(t) ≤ 0, f or r > b.
Consequently, the equation y 00 + r1 y 0 + v(r)y = 0 has a positive supersolution on (0, δ) if and only if it has a
positive solution on (0, δ).

7
Proof: That 1) and 2) are equivalent follows from the work of Wintner [24, 25], a proof of which may be
found in [15]).
To prove the rest, we note that the change of variable z(s) = y(e−s ) maps the equation y 00 + r1 y 0 + v(r)y = 0
−e−t y 0 (e−t )
into z 00 + e−2s v(e−s )z(s) = 0. On the other hand, the change of variables ψ(t) = y(e−t ) maps
y 00 + r1 y 0 + v(r)y into ψ 0 (t) + ψ 2 (t) + e−2t v(e−t ). This proves the lemma. 

Proof of Theorem 2.1: The implication 1) implies 2) follows immediately from Proposition 4.1 and Lemma
4.2. It is valid for any smooth bounded domain provided v is assumed to be non-decreasing on (0, R). this
condition is not needed if the domain is a ball of radius R.
To show that 2) implies 1), we assume that inequality (HV ) holds on a ball Ω of radius R, and then apply
Lemma (4.3) to obtain a C 2 -supersolution for the equation (29). Now take the surface average of u, that is
1 1
Z Z
w(r) = u(x)dS = u(rω)dω > 0, (32)
nωw r n−1 ∂Br nωn |ω|=1
where ωn denotes the volume of the unit ball in Rn . We may assume that the unit ball is contained in Ω
(otherwise we just use a smaller ball). By a standard calculation we get
n−1 0 1
Z
w00 (r) + w (r) ≤ ∆u(x)dS. (33)
r nωn r n−1 ∂Br
Since u(x) is a supersolution of (29), w satisfies the inequality:
n−1 0 n − 2 2 w(r)
w00 (r) + w (r) + ( ) ≤ −v(r)w(r), f or 0 < r < R. (34)
r 2 r2
Now define n−2
ϕ(r) = r 2 w(r), in 0 < r < R. (35)
Using (34), a straightforward calculation shows that ϕ satisfies the following inequality
ϕ0 (r)
ϕ00 (r) + ≤ −ϕ(r)v(r), f or 0 < r < R. (36)
r
By Lemma 4.4 we may conclude that the equation y 00 (r) + 1r y 0 + v(r)y = 0 has actually a positive solution
ϕ on (0, R).
To establish formula (7), it is clear that by the sufficient condition c(V ) ≥ c whenever y 00 (r)+ r1 y 0 +cv(r)y = 0
has a positive solution on (0, R). On the other hand, the necessary condition yields that y 00 (r) + r1 y 0 +
c(V )v(r)y = 0 has a positive solution on (0, R). The proof is now complete. 

5 Applications
In this section we start by applying Theorem 2.1 to recover in a relatively simple and unified way, all
previously known improvements of Hardy’s inequality. For that we need to investigate whether the ordinary
differential equation
y0
y 00 + + v(r)y(r) = 0, (37)
r
corresponding to a potential v has a positive solution ϕ on (0, δ) for some δ > 0. In this case, ψ(r) = ϕ( δr
R ) is
δ2 δ2
00 1 0
a solution for y (r)+ r y + R2 v( R r)y = 0 on (0, R), which means that the scaled potential Vδ (x) = R2 V ( Rδ x)
δ

yields an improved Hardy formula (HVδ ) on a ball of radius R. Here is an immediate application of this
criterium.

1) The Brezis-Vázquez improvement [6]: Here we need to show that we can have an improved inequality
with a constant potential. In this case, the best constant for which the equation
y0
y 00 + + cy(r) = 0, (38)
r

8
 1 2/n
has a positive solution on (0, R), with R = (|Ω|/ωn ) n is z02 ωn |Ω|−2/n , where z0 = 2.4048... is the first zero
of bessel function J0 (z). We need to show that if α is the first root of an arbitrary solution of the Bessel
0
equation y 00 + yr + y(r) = 0, then we necessarily have α ≤ z0 . To see this, we let x(t) = aJ0 (t) + bY0 (t), where
J0 and Y0 are the two standard linearly independent solutions of Bessel equation, and a and b are constants.
Assume the first zero of x(t) is larger than z0 . Since the first zero of Y0 is smaller than z0 , we have a ≥ 0.
Also b ≤ 0, because Y0 (t) → −∞ as t → 0. Finally note that Y0 (z0 ) > 0, so that if b < 0, then x(z0 + ) < 0
for  sufficiently small, which is impossible. This means that b = 0 which is also a contradiction, and we are
done proving the result of Brezis-Vázquez mentioned in the introduction.

2) The best constant in the improvement of Adimurthi et al. [1]: In this case, one easily sees that
the functions ϕj (r) = ( ji=1 log (i) ρr ) 2 is a solution of the equation
Q 1

n
ϕ0j (r) + rϕ00j (r) 1 Y ρ
− = 2 ( log(i) )−2 ,
rϕj (r) 4r i=1 r

1
Qn (i) ρ −2
on (0, R), which means that the inequality (HV ) holds for the potential V (x) = 4|x|2( i=1 log |x| ) which
yields the result of Adimurthi et al. In the following, we use our characterization to show that the constant
appearing in the above improvement is indeed the best constant in the following sense:
|u|2 Pm−1 R 2 Qj
−2
|∇u|2 dx − ( n−2 2
dx − 41 j=1 Ω |u| ρ
log (i) |x|
R R
1 Ω 2 ) Ω |x|2 |x|2 i=1
= inf , (39)
(i) R −2
R |u|2 Qm
4 u∈H01 (Ω)\{0}


Ω |x|2 i=1 log |x|

fo all 1 ≤ m ≤ k. We proceed by contradiction, and assume that

R |u|2 Pm−1 R 2 Qj (i) ρ −2
|∇u|2 dx − ( n−2 2
dx − 14 j=1 Ω |u|
R

1 Ω 2 ) Ω |x|2 |x|2 i=1 log |x|
+λ= inf R |u|2 Qm
−2 ,
4 u∈H01 (Ω)\{0}
2 log (i) ρ
Ω |x| i=1 |x|

and λ > 0. From Theorem 2.1 we deduce that there exists a positive function ϕ such that
m−1 j m
ϕ0 (r) + rϕ00 (r) 1 X 1 Y (i) ρ
−2 1 1 Y (i) ρ
−2
− = log + ( + λ) log .
ϕ(r) 4 j=1 r i=1 r 4 r i=1 r

ϕ(r)
Now define f (r) = ϕm (r) > 0, and calculate,
m
ϕ0 (r) + rϕ00 (r) ϕ0 (r) + rϕ00m (r) f 0 (r) + rf 00 (r) f 0 (r) X 1
= m + − .
ϕ(r) ϕm (r) f (r) f (r) i=1 ij=1 logj ( ρ )
Q
r

Thus,
m m
f 0 (r) + rf 00 (r) f 0 (r) X 1 1 Y (i) ρ
−2
− = −λ log . (40)
f (r) f (r) i=1 ij=1 logj ( ρ ) r i=1 r
Q
r

If now f 0 (αn ) = 0 for some sequence {αn }∞ ∞

n=1 that converges to zero, then there exists a sequence {βn }n=1
00 0
that also converges to zero, such that f (βn ) = 0, and f (βn ) > 0. But this contradicts (40), which means
that f is eventually monotone for r small enough. We consider the two cases according to whether f is
increasing or decreasing:

Case I: Assume f 0 (r) > 0 for r > 0 sufficiently small. Then we will have
m
(rf 0 (r))0 X 1
0
≤ Qi j ρ
.
rf (r) i=1 r j=1 log ( r )

Integrating once we get

c
f 0 (r) ≥ Qm j ρ ,
r j=1 log ( r )

9
for some c > 0. Hence, limr→0 f (r) = −∞ which is a contradiction.

Case II: Assume f 0 (r) < 0 for r > 0 sufficiently small. Then
m
(rf 0 (r))0 X 1
0
≥ Qi j ρ
.
rf (r) i=1 r j=1 log ( r )

Thus,
c
f 0 (r) ≥ − Qm , (41)
r j=1 logj ( ρr )
for some c > 0 and r > 0 sufficiently small. On the other hand
m j
f 0 (r) + rf 00 (r) X 1 Y (i) R
−2 1
≤ −λ log ≤ −λ( Qm j ρ
)0 .
f (r) j=1
r i=1
r j=1 log ( r )

Since f 0 (r) < 0, there exists l such that f (r) > l > 0 for r > 0 sufficiently small. From the above inequality
we then have
1 1
bf 0 (b) − af 0 (a) < −λl( Qm j ρ
− Qm j ρ
).
j=1 log ( b ) j=1 log ( a )

From (41) we have lima→0 af 0 (a) = 0. Hence,

λl
bf 0 (b) < − Qm j ρ
,
j=1 log ( b )

for every b > 0, and

λl
f 0 (r) < − Qm ,
r j=1 logj ( ρr )
for r > 0 sufficiently small. Therefore,
lim f (r) = +∞,
r→0

and by choosing l large enouph (e.g., l > λc ) we get to contradict (41) and the proof is now complete.

3) The Filippas and Tertikas improvement [12]: Let D ≥ supx∈Ω |x|, and define
r r r r 1
ϕk (r) = (X1 ( )X2 ( ) . . . Xi−1 ( )Xi ( ))− 2 , i = 1, 2, . . . .
D D D D
Using the fact that Xk0 (r) = 1r X1 (r)X2 (r) . . . Xk−1 (r)Xk2 (r) for k = 1, 2, . . ., we get

ϕ0k (r) + rϕ00k (r) 1 r r r r

− = X12 ( )X22 ( ) . . . Xk−1
2
( )Xk2 ( ).
ϕk (r) 4r D D D D
1 2 |x| 2 |x| 2 |x| 2 |x|
This means that the inequality (HV ) holds for the potential V (x) = 4|x|2 X1 ( D )X2 ( D ) . . . Xk−1 ( D )Xk ( D ),

which yields the result of Filippas and Tertikas [12]. We now identify the best constant by showing that:
R |u|2 Pm−1 R 2
|∇u|2 dx − ( n−2 2
dx − 41 j=1 Ω |u| 2 |x| 2 |x| 2 |x| 2 |x|
R
1 Ω 2 ) Ω |x|2 |x|2 X1 ( D )X2 ( D ) . . . Xj−1 ( D )Xj ( D )
= inf R |u|2 2 |x| 2 |x| ,
4 u∈H01 (Ω)\{0} X ( )X2 ( D ) . . . Xm−1
Ω |x|2 1 D
2 ( |x| 2 |x|
D )Xm ( D )
(42)
fo all 1 ≤ m ≤ k. We proceed again by contradiction and in a way very similar to the above case. Indeed,
assuming that
R |u|2 Pm−1 R |u|2 2 |x| 2 |x|
|∇u|2 dx − ( n−2 2
dx − 41 j=1 2
( |x| 2 |x|
R
1 Ω 2 ) Ω |x|2
X ( )X2 ( D ) . . . Xj−1
Ω |x|2 1 D D )Xj ( D )
+λ= inf |u|2
,
4 2 |x| )X 2 ( |x| ) . . . X 2 |x| 2 ( |x| )
R
u∈H01 (Ω)\{0} 2 X1 (
Ω |x| D2 D m−1 (D )Xm D

10
and λ > 0, we use again Theorem 2.1 to find a positive function ϕ such that
m−1
ϕ0 (r) + rϕ00 (r) 1 X 1 2 r r r r
− = X ( )X 2 ( ) . . . Xj−1
2
( )Xj2 ( )
ϕ(r) 4 j=1 r 1 D 2 D D D
1 1 r r r 2 r
+ ( + λ) X12 ( )X22 ( ) . . . Xm−1
2
( )Xm ( ).
4 r D D D D
ϕ(r)
Setting f (r) = ϕm (r) > 0, we have

m i
ϕ0 (r) + rϕ00 (r) ϕ0 (r) + rϕ00m (r) f 0 (r) + rf 00 (r) f 0 (r) X Y r
= m + − Xj ( ).
ϕ(r) ϕm (r) f (r) f (r) i=1 j=1 D

Thus,
m i m
f 0 (r) + rf 00 (r) f 0 (r) X Y r 1Y 2 r
− Xj ( ) = −λ X ( ). (43)
f (r) f (r) i=1 j=1 D r j=1 j D

Arguing as before, we deduce that f is eventually monotone for r small enough, and we consider two cases:

Case I: If f 0 (r) > 0 for r > 0 sufficiently small, then we will have
m i
(rf 0 (r))0 X 1 Y r
0
≤ Xj ( ).
rf (r) i=1
r j=1 D

Integrating once we get

m
cY r
f 0 (r) ≥ Xj ( ),
r j=1 D

for some c > 0, and therefore limr→0 f (r) = −∞ which is a contradiction.

Case II: Assume f 0 (r) < 0 for r > 0 sufficiently small. Then
m i
(rf 0 (r))0 X 1 Y r
0
≥ Xj ( )
rf (r) i=1
r j=1 D

Thus,
m
cY r
f 0 (r) ≥ − Xj ( ), (44)
r j=1 D

for some c > 0 and r > 0 sufficiently small. On the other hand
m j m
f 0 (r) + rf 00 (r) X 1Y 2 Y r
≤ −λ Xj ≤ −λ( Xj ( ))0 .
f (r) j=1
r i=1 j=1
D

Since f 0 (r) < 0, we may assume f (r) > l > 0 for r > 0 sufficiently small, and from the above inequality we
have m m
Y b Y a
bf 0 (b) − af 0 (a) < −λl( Xj ( ) − Xj ( )).
j=1
D j=1
D

From (44) we have lima→0 af 0 (a) = 0. Hence,

m
λl Y r
f 0 (r)) < − Xj ( ),
r j=1 D

11
for r > 0 sufficiently small. Therefore,
lim f (r) = +∞,
r→0

and by choosing l large enouph (i.e. l > λc ) we contradict (44) and the proof is complete. 

We shall now make the connection between improved Hardy inequalities and the existence of non-oscillatory
solutions (i.e., those z(s) such that z(s) > 0 for s > 0 sufficiently large) for the second order linear differential
equations
z 00 (s) + a(s)z(s) = 0. (45)
Interesting results in this direction were established by many authors (see [15, 16, 24, 25, 26]). Here is a
typical criterium about the oscillatory properties of equation (45):
R∞
1. If lim supt→∞ t t a(s)ds < 14 , then Eq. (45) is non-oscillatory.
R∞
2. If lim inf t→∞ t t a(s)ds > 41 , then Eq. (45) is oscillatory.

This result combined with Theorem 2.1 and Lemma 4.4 clearly yields Corollary 2.2.

Proof of Corollary 2.4: It follows from Hölder’s inequality that

R 2
|u(x)| s dx
Z
1
2
( V (|x|)|u(x)| dx) s ≥ R Ω − r 1 ,
Ω ( Ω V s (|x|)dx) r
1 1
where s ≥ 1 and s + r = 1. Letting p = 2s , we get

1
Z Z
2
V (|x|)|u(x)|2 dx ≥ ( |u(x)|p dx) p −1 (|x|)||
.
Ω Ω ||V p
L 2−p (Ω)

Inequality (2.4) now follows from Theorem 2.1. 

Proof of Corollary 2.5: Define the functional

Z 2
u (x)
Z
Fµ (u) = |∇u(x)|2 dx − µ 2
dx, (46)
Ω Ω |x|

which is continuous, Gateaux differentiable and coercive on H01 (Ω). Let uµ > 0 be a minimizer of Fµ over
the manifold M = {u ∈ H0 (Ω)| Ω u (x)V (x) = 1} and assume λ1µ is the infimum. It is clear that λ1µ > 0.
1
R 2

By standard arguments we can conclude that uµ is a weak solution of (EV,µ ). The rest of the proof follows
from Corollary 2.2 and the fact that
2
(|∇u|2 − µn u|x|(x)
R
Ω 2 )dx
λ1 (V ) = lim λ1 = inf .
µ→µn µ
R
u∈H01 (Ω)\{0} 2
|u(x)| V (x)dx
Ω

References
[1] Adimurthi, N. Chaudhuri, and N. Ramaswamy, An improved Hardy Sobolev inequality and its appli-
cations, Proc. Amer. Math. Soc. 130 (2002), 489-505.
[2] M. Agueh, N. Ghoussoub, X. S. Kang: Geometric inequalities via a general comparison principle for
interacting gases, Geom. And Funct. Anal., Vol 14, 1 (2004) p. 215-244

[3] H. Brezis, E. H. Lieb, Sobolev inequalities with remainder terms, J. Funct. Anal. 62 (1985), 73-86.
[4] H. Brezis, M. Marcus, Hardy’s inequality revisited, Ann. Scuola. Norm. Sup. Pisa 25 (1997), 217-237.

12
 [5] H. Brezis, M. Marcus, and I. Shafrir, Extremal functions for Hardy’s inequality with weight, J. Funct.
Anal. 171 (2000), 177-191.
[6] H. Brezis and J. L. Vázquez, Blowup solutions of some nonlinear elliptic problems, Revista Mat. Univ.
Complutense Madrid 10 (1997), 443-469.
[7] X. Cabré, Y. Martel, Weak eigenfunctions for the linearization of extremal elliptic problems, J. Funct.
Anal. 156 (1998), 30-56.
[8] X. Cabré, Y. Martel, Existence versus explosion instantanée pour des équations de la chaleur linéaires
avec potential singulier, C. R. Acad. Sci. Paris Sér. I 329 (1999), 973-978.
[9] N. Chaudhuri, Bounds for the best constant in an improved Hardy-Sobolev inequality, Zeitschrift für
Analysis und ihre Anwendungen, Vol. 22, No. 4 (2003), 757-765
[10] D. Cordero-Erausquin, B. Nazaret, and C. Villani. A mass-transportation approach to sharp Sobolev
and Gagliardo-Nirenberg inequalities, Adv. Math. 182, 2 (2004), 307-332.
[11] E. B. Davies, A review of Hardy inequalities, Oper. Theory Adv. Appl. 110 (1999), 55-67.
[12] S. Filippas, A. Tertikas, Optimizing improved Hardy inequalities. J. Funct. Anal. 192 (2002), no. 1,
186-233.

[13] J. Fleckinger, E. M. Harrell II, and F. Thelin, Boundary behaviour and estimates for solutions of
equations containing the p-Laplacian, Electron. J. Differential Equations 38 (1999), 1-19.
[14] N. Ghoussoub, A. Moradifam, Improved Hardy inequalities for the p-Laplacian and the bi-Laplacian,
In preparation.
[15] P. Hartman, Ordinary differential equations, Wiley, New York, 1964.
[16] C. Huang, Oscillation and Nonoscillation for second order linear differential equations, J. Math. Anal.
Appl. 210 (1997) 712-723.
[17] B. Kawohl, Rearrangements and Convexity of Level Sets in PDE, Lecture Notes in Mathematics, Vol.
1150, Springer, Berlin, 1985.
[18] B. Opic, and A. Kufner, Hardy type Inequalities, Pitman Research Notes in Mathematics, Vol. 219,
Longman, New York, 1990.
[19] I. Peral and J. L. Vázquez, On the stability and instability of the semilinear heat equation with expo-
nential reaction term, Arch. Rat. Mech. Anal. 129 (1995), 201-224.
[20] B. Simon, Schrödinger semigroups, Bull. Amer. Math. Soc. 7 (1982), 447-526.
[21] A. Tertikas, Critical phenomena in linear elliptic problems, J. Funct. Anal. 154 (1998), 42-66.
[22] J. L. Vázquez, Domain of existence and blowup for the exponential reaction diffusion equation, Indiana
Univ. Math. J. 48 (1999), 677-709.
[23] J. L. Vázquez and E. Zuazua, The Hardy inequality and the asymptotic behaviour of the heat equation
with an inverse-square potential, J. Funct. Anal. 173 (2000), 103-153.

[24] A. Wintner, On the nonexistence of conjugate points, Amer. J. Math. 73 (1951) 368-380.
[25] A. Wintner, On the comparision theorem of Knese-Hille, Math. Scand. 5 (1957) 255-260.
[26] James S. W. Wong, Oscillation and nonoscillation of solutions of second order linear differential equations
with integrable coefficients. Trans. Amer. Math. Soc. 144 (1969) 197-215.

13