• No results found

Generalizations of the logarithmic Hardy inequality in critical Sobolev Lorentz spaces

N/A
N/A
Protected

Academic year: 2020

Share "Generalizations of the logarithmic Hardy inequality in critical Sobolev Lorentz spaces"

Copied!
14
0
0

Loading.... (view fulltext now)

Full text

(1)

R E S E A R C H

Open Access

Generalizations of the logarithmic Hardy

inequality in critical Sobolev-Lorentz spaces

Shuji Machihara

1

, Tohru Ozawa

2

and Hidemitsu Wadade

3*

*Correspondence:

[email protected]

3Faculty of Education, Gifu

University, 1-1 Yanagido, Gifu, 501-1193, Japan

Full list of author information is available at the end of the article

Abstract

In this paper, we establish the Hardy inequality of the logarithmic type in the critical Sobolev-Lorentz spaces. More precisely, we generalize the Hardy type inequality obtained in Edmunds and Triebel (Math. Nachr. 207:79-92, 1999). The generalized inequality allows us to take the exponents appearing in the inequality more flexibly, and its optimality is discussed in detail. O’Neil’s inequality and its reverse play an essential role for the proof.

MSC: Primary 46E35; secondary 26D10

Keywords: logarithmic Hardy inequality; critical Sobolev-Lorentz space; O’Neil’s inequality

1 Introduction and the main theorem

In this paper, we shall give a systematic treatment concerning the Hardy type inequal-ities on the critical Sobolev-Lorentz spacesHs

p,q(Rn) withn∈N,s∈R,  <p<∞and ≤q≤ ∞, where the spaceHs

p,q(Rn) can be characterized in terms of the Bessel potential such asHs

p,q(Rn) := ( –)– s

Lp,q(Rn) with the Lorentz spaceLp,q(Rn). We collect precise

definitions of those function spaces and related properties in Section .

We recall the Sobolev embedding theorem onH

n p

p,p(Rn), which states that the

continu-ous inclusionsH

n p

p,p(Rn)Lq,q(Rn) hold for allq∈[p,∞) andq∈[p,∞]. However,

the limiting caseq=∞in this embedding fails, provided that (p,q)= (,∞). This

im-plies that functions in the spaceH n p

p,p(R

n) can have a local singularity at some point inRn.

In fact, the critical Sobolev spaceH n p

p(Rn), which is identical with the critical

Sobolev-Lorentz spaceH

n p

p,p(Rn) withp=p=:p, admits a singularity of the logarithmic order,

see Adams and Fournier [] and Maz’ya []. As a characterization ofH

n p

p(Rn), Edmunds and Triebel [] proved the corresponding Hardy-type inequality with a logarithmic cor-rection as follows.

Theorem A(Edmunds-Triebel [, Theorem .]) Let n∈Nand <p<∞.Then there exists a positive constant C such that the inequality

{|x|<}

|

u(x)|

|log|x||

p dx

|x|n

p

Cu H

n p p

(.)

holds for all uH n p p(Rn).

(2)

The main purpose of this paper is to generalize (.) into two directions. First, we prove the corresponding logarithmic Hardy-type inequality in the critical Sobolev-Lorentz

spaceH

n p

p,p(R

n), which coincides with (.) whenp

=p=:p. Furthermore, we

investi-gate the possibility whether the exponents appearing in the inequalities can be taken more flexibly, including the consideration on its optimality. Indeed, our main result now reads as follows.

Theorem . Let n∈N,  <p<∞,  <q≤ ∞and <α,β<∞.Then the inequality

{|x|<}

|u(x)|α

|log|x||β dx

|x|n

α

Cu H

n p p,q

(.)

holds for all uH n p

p,q(Rn)if and only if one of the following conditions(i), (ii)and(iii)is fulfilled

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

(i)  +αβ< ;

(ii)  +αβ≥ and q< α

+αβ;

(iii)  +αβ> , q=+ααβ and αβ.

(.)

Remark . The condition (ii) in (.) allows us to take  +αβ= , which implies α

+αβ =∞. In the special case ofp=q=α=β, the inequality (.) is precisely inequal-ity (.) by Edmunds and Triebel []. Also note that the valueq=+ααβ is the critical ex-ponent in the sense that inequality (.) holds or not. Moreover, Theorem . states that

whenq= α

+αβ, inequality (.) holds ifαβand fails ifα<β. In particular, inequality (.) fails for the marginal caseq=∞and  +αβ= . Indeed, the functionudefined

byu(x) :=η(x)|log|x||belongs toH

n p

p,∞(Rn), whereηis a cut-off function supported near

the origin, while

{|x|<}

|u(x)|α |log|x||+α

dx

|x|n= +∞.

There is a number of both mathematical and physical applications of Hardy-type in-equalities. Among others, we refer the reader to Adimurthiet al.[], Beckner [], Bradley [], Brézis and Marcus [], Edmunds and Triebel [], García and Peral [], Gurka and Opic [], Herbst [], Kalf and Walter [], Kerman and Pick [–], Ladyzhenskaya [],

Machiharaet al.[], Matsumura and Yamagata [], Nagayasu and Wadade [], Ozawa

and Sasaki [], Pick [], Reed and Simon [], Triebel [] and Zhang []. Especially, in Bradley [] and Edmunds and Triebel [], the similar type inequalities to (.) were considered in terms of Besov-type spaces.

This paper is organized as follows. Section  is devoted to the definition of the Sobolev-Lorentz space, as well as several lemmas needed for the proof of Theorem .. We shall prove Theorem . in Section .

2 Preliminaries

In this section, we first recall the definition of the Lorentz spaces. To this end, we define the

(3)

f∗: [,∞)→[,∞] denotes the distribution function off given by

f∗(λ) := x∈Rn; f(x) >λ forλ≥,

and then the rearrangementf∗: [,∞)→[,∞] off is defined by

f∗(t) :=infλ> ;f∗(λ)≤t fort≥.

Moreover,f∗∗: (,∞)→[,∞] denotes the average function off∗defined by

f∗∗(t) := t

t

f∗(τ)dτ fort> .

In what follows, we assume thatf∗(t) < +∞for allt> . Thenf∗is right-continuous and non-increasing on (,∞), and hence,f∗∗is continuous and non-increasing on (,∞) with f∗(t)≤f∗∗(t) for allt> . We now introduce the Lorentz space by using the rearrangement. Let ≤p<∞and ≤q≤ ∞. Then the Lorentz spaceLp,q(Rn) is defined as a function space, equipped with the following norm,

fLp,q:=

⎧ ⎨ ⎩

(∞(tpf(t))q dt t)

q if q<;

supt>(tpf(t)) ifq=. (.)

We can takef∗replaced byf∗∗in definition (.) as another equivalent norm onLp,q(Rn) ifp= . Indeed, the following Hardy inequality guarantees its equivalence

tp t

t

f(τ)

q dt

t

q

p

tpf(t)qdt t

q

(.)

for non-negative measurable functionsf, for which the integral on the right-hand side in (.) is finite. Remark that inequality (.) is still valid for the caseq=∞by replacing the integral by the supremum. For the proof of (.), see O’Neil [, Lemma .] and

references therein. Furthermore, sincef∗andf∗∗are both monotonically non-increasing

functions in (,∞), we easily get the following decay estimates. For anyt> , we have

f∗(t)≤

q p

q tpf

Lp,q (.)

and ifp> , together with inequality (.), we also have for anyt> ,

f∗∗(t)≤p

q p

q t–pf

Lp,q. (.)

Note that inequalities (.) and (.) are also valid for the marginal caseq=∞, and we will utilize them frequently for the proof of the main theorem in Section .

We also make use of the celebrated Hardy-Littlewood inequality

Rn

f(x)g(x) dx

(4)

for all measurable functionsfandg. The proof of (.) can be found in Bennett and Sharp-ley [, Theorem .].

Next, we recall the pointwise rearrangement inequality for the convolution of functions

proved by O’Neil [, Theorem .]. In fact, for measurable functionsf andgonRn, we

have

(f ∗g)∗∗(t)≤tf∗∗(t)g∗∗(t) +

t

f∗(τ)g∗(τ)dτ fort> . (.)

Moreover, we make use of the reverse O’Neil inequality, established in Kozonoet al.[,

Lemma .]. Indeed, there exists a positive constantCsuch that the inequality

(f ∗g)∗∗(t)≥C

tf∗∗(t)g∗∗(t) +

t

f∗(τ)g∗(τ)

(.)

holds for allt>  and for all measurable functionsf andgonRn, which are both non-negative, radially symmetric and non-increasing in the radial direction.

In this paper, we frequently use the Bessel potentialGs∗:= ( –)– s

and the Riesz

poten-tialIs∗:= (–)– s

for  <s<n. More precisely, the kernel functionsIsandGsare defined

respectively by

⎧ ⎪ ⎨ ⎪ ⎩

Is(x) :=

(ns) sπn (s)|x|

–(ns);

Gs(x) := 

(π)s (s)

e

π|x|t–πt tn–sdt

t

forx∈Rn\ {}, where denotes the gamma function. Based on the Lorentz space, we

de-fine the Sobolev-Lorentz spaceHs

p,q(Rn) byHps,q(Rn) := (I–)– s

Lp,q(Rn) =GsLp,q(Rn),

equipped with the norm uHps,q :=(I–) s

uLp

,q. The spaceHps,q(Rn) is a generaliza-tion of the usual Sobolev spaceHs

p(Rn), since we haveLp,p(Rn) =Lp(Rn) due to the norm-invariance ofuLp,p=uLp. We now collect the elementary properties ofIsandGsin the following lemma.

Lemma . Let n∈Nand <s<n.

(i) IsandGsare non-negative,radially symmetric and non-increasing in the radial

direction,so thatIs∗(t) =Is(x)andGs∗(t) =Gs(x)if|x|= (ωnt )

n> ,whereωn:= π

π

n (n)

denotes the volume of the unit ball inRn.

(ii) Gs(x)≤Is(x)for allx∈Rn\ {},which implies theGs(t)≤Is∗(t),G∗∗s (t)≤I∗∗s (t)for

allt> ,andlim|x|↓Gs(x)

Is(x) =limt↓

Gs(t) Is∗(t) = .

(iii) GsL= and there exists a positive constantCsuch that the following inequalities

hold

Gs(x)≤

⎧ ⎨ ⎩

C|x|–(ns) forxRn\ {}; Ce–|x| forxRnwith|x| ≥.

(5)

In the end of this section, we shall show the following one-dimensional Hardy inequality of logarithmic type.

Lemma . Let <α,β<∞.Then there exists a positive constant C such that the

inequal-ity

 

 

t

φ(s) ds

α

|logt|–βdt t

α

C

 

φ(t) α|logt|–βdt t

α

(.)

holds for all measurable functionsφ such that the integral on the right-hand side of (.) is finite.

Furthermore, we can show the following dual variant of inequality (.).

Lemma . Let <βα<∞and q:=+ααβ.Then there exists a positive constant C such that the inequality

 

 

t

φ(s) ds

α

|logt|–βdt t

α

C

 

t φ(t) qdt t

q

(.)

holds for all measurable functionsφsuch that the integral on the right-hand side of (.) is finite.

We shall apply Lemma . for the proof of the sufficiency part of Theorem . in Sec-tion , and Lemma . will be used for the proof of the necessity part of Theorem . in Section . Lemma . and Lemma . can be obtained as corollaries of the following weighted inequalities obtained in Bradley [] and Muckenhoupt [].

Theorem B(Bradley [], Muckenhoupt []) Let <ρσ<∞and let U and V be

mea-surable weights.

(i) There exists a positive constantCsuch that the inequality

U(t)

t

ψ(s) ds σ

dt

σ

C

V(t)ψ(t) ρdt

ρ

(.)

holds for all measurable functionsψsuch that the integral on the right-hand side of

(.)is finite if and only if

sup

r>

r

U(t) σdt

σ r

V(t) –ρdt

ρ

< +∞.

(ii) There exists a positive constantCsuch that the inequality

U(t)

t

ψ(s) ds σ

dt

σ

C

V(t)ψ(t) ρdt

ρ

(6)

holds for all measurable functionsψsuch that the integral on the right-hand side of (.)is finite if and only if

sup

r>

r

U(t) σdt

σ

r

V(t) –ρdt

ρ

< +∞.

Now we shall show Lemma . and Lemma . by applying Theorem B(i) and Theo-rem B(ii), respectively.

Proof of Lemma. Define the weightsUandVby

U(t) :=

⎧ ⎨ ⎩

|logt|–βαt–+αα for  <t< ;

 fort

and

V(t) :=

⎧ ⎨ ⎩

|logt|–βαt–α for  <t< ;

 fort.

Then the direct calculation shows

sup

r>

r

U(t)

α dt

α r

V(t) –

α dt

α

< +∞.

Thus, Theorem B(i) implies that

U(t)

t

ψ(s) ds α

dt

α

C

V(t)ψ(t)

α dt

α

,

namely,

 

t

t

ψ(s) ds

α

|logt|–βdt t

α

C

 

ψ(t) α|logt|–βdt t +

 

ψ(t) αdt

α

for all measurable functionsψ. Takingφ=χ(,

)ψyields the desired inequality (.).

Proof of Lemma. Define the weightsUandVby

U(t) :=

⎧ ⎨ ⎩

|logt|–βαt–α for  <t< ;

 fort

and

V(t) :=

⎧ ⎨ ⎩

tq–q for  <t<

;

tqq– fort≥ 

(7)

Then the direct calculation shows

sup

r>

r

U(t)

α dt

α

r

V(t) –q

dt

q < +∞.

Sinceαβimpliesαq, by applying Theorem B(ii), we obtain

 

t

ψ(s) ds

α

|logt|–βdt t

α

C

 

t ψ(t) qdt t +

 

tq– ψ(t) qdt

q

for all measurable functionsψ. Takingφ=χ(,

)ψyields the desired inequality (.).

3 Proof of the sufficiency part of Theorem 1.1

In this section, we consider the sufficiency part of Theorem .. To this end, it suffices to show the following key lemmas.

Lemma . Let n∈N,  <p<∞,  <q≤ ∞,and let <α,β <∞.Assume one of the

conditions(i), (ii)and(iii)in(.).Then there exists a positive constant C such that the inequality

 

u∗(t)α

|logt|β dt

t

α

Cu H

n p p,q

(.)

holds for all uH n p p,q(Rn).

Lemma . Let n∈N,  <p<∞,  <q≤ ∞, and let <α,β <∞.Assume one of the conditions(i), (ii)and(iii)in(.).Then there exists a positive constant C such that the inequality

Rn

w(x)u(x) α dx

α

C

sup <t<

|logt|βαw(t)

u H

n p p,q

(.)

holds for all uH n p

p,q(Rn)and for all measurable function w satisfying

|suppw|< 

and <supt< 

tα|logt| β

αw(t) <.

Remark . By takingw(x) :=|log|x||–βα|x|–αnχ{|x|<ε}(x) with smallε>  in Lemma ., we

can prove the sufficiency part of Theorem ., whereχ{|x|<ε}is a characteristic function on

{|x|<ε}.

(8)

Proof of Lemma. By using inequality (.) with|suppw|< and applying Lemma ., we see

Rn

w(x)u(x) αdx=

 

(wu)∗(t)αdt

 

w∗(t)αu∗(t)αdt

=

 

|logt| β

αw(t)α u

(t)α

|logt|β dt

t

≤sup <t<

t|logt|βw∗(t)α

 

u∗(t)α

|logt|β dt

t

C

sup <t<

t|logt|βw∗(t)α

H n p p,q

,

which is exactly the inequality (.).

We are now in a position to prove Lemma ..

Proof of Lemma. First, by letting ( –)npu=fL

p,q(Rn), Lemma . can be rewritten as the following equivalent form

 

(Gn pf)

(t)α

|logt|β dt

t

α

CfLp,q (.)

forfLp,q(Rn). Hence, we concentrate our attention on the proof of (.) below. By the O’Neil inequality (.) and decay estimates (.) and (.), we have for  <t<,

(Gn pf)

(t)

≤(Gn pf)

∗∗(t)

tG∗∗n p(t)f

∗∗(t) +t

Gn p(s)f

(s)ds

=tG∗∗n p(t)f

∗∗(t) +

 

Gn p(s)f

(s)ds+

 

t Gn

p(s)f ∗(s)ds

C

Gn

pLp,∞fLp,q+GnpLfLp,q

 

s–(+p)ds

+

 

t Gn

p(s)f ∗(s)ds

=CfLp,q+

 

t Gn

p(s)f

(s)ds. (.)

Thus, from (.), we obtain

 

(Gn pf)

(t)α

|logt|β dt tαC   

|logt|–βdt t

α

fLp,q

+      t Gn

p(s)f

(s)dsα|logt|βdt t

α

, (.)

(9)

Note that the conditions (i), (ii) and (iii) in (.) can be rewritten equivalently as follows

(i) β>  +α

q or (ii)  +

α

q =β and αβ. (.)

Case . Assume (i) in (.). For  <t< , by Lemma .(ii) and Hölder’s inequality, we see

 

t Gn

p(s)f

(s)dsC

 

t

spf(s)ds s

C

 

t ds

s

q 

t

spf(s)qds s

q

C|logt|

qf Lp,q.

Note that the calculation above is also valid for the caseq=∞. Thus, we have

 

 

t Gn

p(s)f

(s)dsα|logt|βdt t

α

C

 

 |logt|

α

qβdt t

α

fLp,qCfLp,q, (.)

where we have used the conditionβ>  +qα, which ensures that the integral on the middle-hand side of (.) is finite. Thus, combining (.) with (.), we obtain the desired estimate.

Case . Assume (ii) in (.). By Lemma .(ii) and Lemma ., we have

 

 

t Gn

p(s)f

(s)dsα|logt|βdt t

α

C

 

tGn p(t)f

(t)qdt t

q

C

 

tpf(t)qdt t

q

CfLp,q. (.)

Thus, combining (.) with (.), we obtain the desired estimate.

4 Proof of the necessity part of Theorem 1.1

In this final section, we shall prove the necessity part of Theorem .. To this end, we shall

construct a concrete function in the critical Sobolev-Lorentz spaceH

n p p,q(Rn).

Proof of the necessity part of Theorem. First, by putting ( –)npu=f, inequality (.) can be rewritten as

{|x|<} |Gn

pf(x)| α

|log|x||β dx

|x|n

α

(10)

Therefore, it is enough to show the breakdown of the inequality (.) under the following conditions, which are the negations of (.) or (.),

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

(i) β<  +α

q and q<∞;

(ii) β≤ +qα (=  +α) and q=∞;

(iii) β=  + α

q, q<∞ and α<β.

(.)

Case . Assume (i) in (.). In this case, we define the functionby

(x) := log|x|

–+εq |x|–npχ{|

x|<ε}(x) (.)

for smallε> . Then we see that for sufficiently smallε> ,becomes non-negative and non-increasing with respect to the radial direction|x|. Thus, we have for smallt> 

∗(t) =f˜ε

t

ωn

n

|logt|–+εq tp=:g

ε(t), (.)

where f˜ε(|x|) :=(x). More precisely, (.) implies that there exist positive constantsδ small enough,CandC˜ such that the inequalities

Cgε(t)≤∗(t)≤ ˜Cgε(t) (.)

hold for all  <t<δ. By using (.), it is easy to seeLp,q(Rn). Indeed, from (.), we obtain

δ

tpfε(t)

qdt t ≤ ˜C

δ

tpg ε(t)

qdt t =C˜

δ

 |

logt|–(+ε)dt t <∞.

On the other hand, sinceis non-negative and non-increasing with respect to the radial direction, so isGn

p. Thus, notingGnp(x) = (Gnp) ∗(ω

nrn) if|x|=r> , we see by changing a variableωnrn=t,

{|x|<} |Gn

p(x)| α

|log|x||β dx

|x|n

=nωn

 

(Gn p)

(ω nrn)α

|logr|β dr

r

C

δ

(Gn p)

(t)α

|logt|β dt

t (.)

for smallδ> . Furthermore, by using Lemma . and the reverse O’Neil inequality (.), we have

δ

(Gn p)

(t)α

|logt|β dt

t

C

δ

(Gn p)

∗∗(t)α

|logt|β dt

(11)

C

δ

(tG∗∗n p(t)f

∗∗ ε (t) +

t Gnp(τ)fε∗(τ))α

|logt|β

dt t

C

δ

(tδGn p(τ)f

ε(τ))

α

|logt|β

dt

t . (.)

Thus, by Lemma . (ii) and (.), we have for smallδ> ,

δ

(tδGn p(τ)f

ε(τ))α

|logt|β

dt

tC

δ

(tδIn p(τ)f

ε(τ))α

|logt|β

dt t

C

δ

(tδτ– 

pg

ε(τ))α

|logt|β dt

t . (.)

Takeε>  small enough, so that  –+qε> , which is possible sinceq> . Thus, we have for any  <t<δ

with smallδ> ,

δ t

τ– 

pg

ε(τ) = q q– ( +ε)

|logt|–+εq |logδ|–+εqC|logt|–+εq . (.)

Summing up all estimates (.), (.), (.), and (.), we obtain

{|x|<} |Gn

p(x)| α

|log|x||β dx

|x|nC

δ

|logt|(–+εq )αβdt

t . (.)

However, the integral on the right-hand side of (.) diverges, provided thatε>  is taken small enough, so that ( – +qε)αβ+ ≥, which is possible since qαβ+  >  by the assumption. Thus, inequality (.) fails under the condition (i) in (.).

Case . Assume (ii) in (.). In this case, we utilizef(x) :=|x|–

n

p instead off

ε(x) used

in Case . Then it is easily seenf∈Lp,∞(Rn). On the other hand, in a quite similar way carried out in Case , we see

{|x|<}

|Gn pf(x)|

α

|log|x||β dx

|x|nC

δ

(Gn pf)

(t)α

|logt|β dt tC δ  (Gn pf)

∗∗(t)α

|logt|β dt

tC

δ

(tδGn p(τ)f

(τ))α |logt|β

dt t

C

δ

(tδIn p(τ)f

(τ))α |logt|β

dt

tC

δ

|logt|αβdt t

for smallδ, where the last integral diverges ifαβ+ ≥, that is,β≤ +α. Thus, in-equality (.) fails under the condition (ii) in (.).

Case . Assume (iii) in (.), which implies that α

q=  +αβ< , namely,q>α. In this case, we make use of the functionwith smallε>  defined by

(x) := log|x|

q

log log|x| –

+ε

(12)

Sinceis non-negative and non-increasing in the radial direction|x|with smallε> , we see

∗(t) |logt|

–q log|log

t| –+εq t–p=:g ε(t)

for smallt> , namely, there exist positive constantsδsmall enough,CandC˜ such that the inequalities

Cgε(t)≤∗(t)≤ ˜Cgε(t) (.)

hold for all  <t<δ. By using (.), it is easy to seeLp,q(Rn). Indeed,

δ

tpfε(t)

qdt t ≤ ˜C

δ

tpg ε(t)

qdt t

=C˜

δ

 |

logt|– log|logt| –(+ε)dt t <∞.

On the other hand, in the same estimates from below as in (.), (.) and (.) in Case , we obtain

{|x|<}

|Gn p(x)|

α

|log|x||β dx

|x|nC

δ

(tδτ– 

pg

ε(τ))α

|logt|β dt

t . (.)

Furthermore, we can easily see

δ t

τ– 

pg

ε(τ) =

δ t

|logτ|–q log|logτ| –

+ε

q

τ |logt|

–qlog|logt| –

+ε

q

for smallt> . In particular, for any  <t<δ with smallδ> , we have

δ t

τ– 

pg

ε(τ)C|logt|–

q log|logt| –

+ε

q . (.)

Thus, combining (.) with (.), we see

{|x|<} |Gn

p(x)| α

|log|x||β dx

|x|nC

δ

(tδτ– 

pg

ε(τ))α

|logt|β dt

t

C

δ

 |logt|

α

qβ log|logt| –

+ε

q αdt t

=C

δ

|logt|– log|logt| –

+ε

q αdt

t . (.)

However, the last integral in (.) diverges, provided thatε>  is taken small, so that –+qεα+ ≥, which is possible sinceq>α. Thus, the inequality (.) fails under the

(13)

Remark . (.) in Lemma . is equivalent to (.) in Theorem .. Indeed, we have already seen in Section  that Lemma . implies Theorem .. On the other hand, (.) is equivalent to (.), and since the weighted norm in the left-hand side of (.) is non-decreasing under the rearrangement, (.) can be reduced to (.), which is equivalent to (.).

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

HW drafted the manuscript. All authors computed to complete the proof of main theorems, and they read and approved the final manuscript.

Author details

1Faculty of Education, Saitama University, 255 Shimookubo, Saitama, Sakuraku 338-8570, Japan.2Department of Applied

Physics, Waseda University, Shinjuku, Tokyo 169-8555, Japan.3Faculty of Education, Gifu University, 1-1 Yanagido, Gifu,

501-1193, Japan.

Acknowledgements

The authors are grateful to the referees for their valuable comments.

Received: 6 February 2013 Accepted: 31 July 2013 Published: 14 August 2013 References

1. Adams, RA, Fournier, JJF: Sobolev Spaces, 2nd edn. Academic Press, Amsterdam (2003) 2. Maz’ya, VG: Sobolev Spaces. Springer, Berlin (1985)

3. Edmunds, DE, Triebel, H: Sharp Sobolev embeddings and related Hardy inequalities: the critical case. Math. Nachr.

207, 79-92 (1999)

4. Adimurthi, Chaudhuri, N, Ramaswamy, M: An improved Hardy-Sobolev inequality and its application. Proc. Am. Math. Soc.130, 489-505 (2002)

5. Beckner, W: Pitt’s inequality with sharp convolution estimates. Proc. Am. Math. Soc.136, 1871-1885 (2008) 6. Bradley, JS: Hardy inequalities with mixed norms. Can. Math. Bull.21, 405-408 (1978)

7. Brézis, H, Marcus, M: Hardy’s inequalities revisited. Ann. Sc. Norm. Super. Pisa, Cl. Sci.25, 217-237 (1997)

8. García, JPA, Peral, IA: Hardy inequalities and some critical elliptic and parabolic problems. J. Differ. Equ.144, 441-476 (1998)

9. Gurka, P, Opic, B: Sharp embeddings of Besov-type spaces. J. Comput. Appl. Math.208, 235-269 (2007) 10. Herbst, WI: Spectral theory of the operator (p2+m2)1/2Ze2/r. Commun. Math. Phys.53, 285-294 (1977)

11. Kalf, H, Walter, J: Strongly singular potentials and essential self-adjointness of singular elliptic operators in C0(Rn\ {0}). J. Funct. Anal.10, 114-130 (1972)

12. Kerman, R, Pick, L: Optimal Sobolev embeddings. Forum Math.18, 535-570 (2006) 13. Kerman, R, Pick, L: Optimal Sobolev embedding spaces. Stud. Math.192, 195-217 (2009)

14. Kerman, R, Pick, L: Explicit formulas for optimal rearrangement-invariant norms in Sobolev imbedding inequalities. Stud. Math.206, 97-119 (2011)

15. Ladyzhenskaya, OA: The Mathematical Theory of Viscous Incompressible Flow. Second English edition, revised and enlarged. Translated from the Russian by Richard A. Silverman and John Chu. Mathematics and its Applications, vol. 2. Gordon & Breach, New York (1969)

16. Machihara, S, Ozawa, T, Wadade, H: Hardy type inequalities on balls. Tohoku Math. J. (to appear)

17. Matsumura, A, Yamagata, N: Global weak solutions of the Navier-Stokes equations for multidimensional compressible flow subject to large external potential forces. Osaka J. Math.38, 399-418 (2001)

18. Nagayasu, S, Wadade, H: Characterization of the critical Sobolev space on the optimal singularity at the origin. J. Funct. Anal.258, 3725-3757 (2010)

19. Ozawa, T, Sasaki, H: Inequalities associated with dilations. Commun. Contemp. Math.11, 265-277 (2009)

20. Pick, L: Optimal Sobolev embeddings. In: Nonlinear Analysis, Function Spaces and Applications (Prague, 1998), vol. 6, pp. 156-199. Acad. Sci. Czech Repub., Prague (1999)

21. Reed, M, Simon, B: Methods of Modern Mathematical Physics. II. Fourier Analysis, Self-Adjointness. Academic Press, New York (1975)

22. Triebel, H: Sharp Sobolev embeddings and related Hardy inequalities: the sub-critical case. Math. Nachr.208, 167-178 (1999)

23. Zhang, J: Extensions of Hardy inequality. J. Inequal. Appl.2006, Article ID 69379 (2006) 24. O’Neil, R: Convolution operators andL(p,q) spaces. Duke Math. J.30, 129-142 (1963) 25. Bennett, C, Sharpley, R: Interpolation of Operators. Academic Press, New York (1988)

26. Kozono, H, Sato, T, Wadade, H: Upper bound of the best constant of a Trudinger-Moser inequality and its application to a Gagliardo-Nirenberg inequality. Indiana Univ. Math. J.55, 1951-1974 (2006)

27. Stein, EM: Singular Integrals and Differentiability Properties of Functions. Princeton University Press, Princeton (1970) 28. Almgren, FJ, Lieb, EH: Symmetric decreasing rearrangement is sometimes continuous. J. Am. Math. Soc.2, 683-773

(1989)

(14)

doi:10.1186/1029-242X-2013-381

References

Related documents