Ostrowski Type Inequalities Related to the Generalized Baouendi-Grushin Vector Fields
JINGBODOU(*) - YAZHOUHAN(**)
ABSTRACT- In this paper, we employ a new method to prove a representation for- mula related to the generalized Baouendi-Grushin vector fields, and then the Ostrowski type inequalities is established in the ball and bounded domain, re- spectively, via the representation formula andL1norm of the horizontal gra- dient. In addition, in the same spirit, we show the Hardy inequalities with boundary term related to the generalized Baouendi-Grushin vector fields.
MATHEMATICSSUBJECTCLASSIFICATION(2010). 26D10, 35J70.
KEYWORDS. Generalized Baouendi-Grushin vector fields; representation formula;
Ostrowski type inequality; Hardy inequality.
1. Introduction
In 1938, A. Ostrowski [21] established the following sharp integral in- equality
1 b a
Zb
a
f(y)dy f(x)
1
4(x ab2 )2 (b a)2 (b a)
kf0k1 (1:1)
for f 2C1[a;b] and x2[a;b], which is known as the Ostrowski type in- equality. Because the Ostrowski type inequality is useful in some fields, it
(*) Indirizzo dell'A.: School of Statistics, XI'AN University of Finance and Economics, XI'AN, Shaanxi, 710100, P. R. China.
E-mail: djbmn@126.com
(**) Indirizzo dell'A.: Department of Mathematics, College of Science, China Jiliang University, Hangzhou, 310018, P. R. China.
E-mail: yazhou.han@gmail.com
was extended from intervals to rectangles and general domains inRN(see [3] and [4]). Moreover, many authors improved inequality (1.1) to the higher-order derivatives case (see [5, 17, 22] and the references therein).
Recently, Lian and Yang [16] obtained the Ostrowski type inequality on the H-type group by the representation formula.
Our main interesting in this paper is to prove the Ostrowski type in- equality related to the generalized Baouendi-Grushin (B-G) vector fields by an improved method. To do this, we first introduce the generalized B-G vector fields.
Let x(x1;x2; ;xn)2Rn;y(y1;y2; ;ym)2Rm;a>0: The generalized B-G vector fields is defined as
Zi @
@xi; Znj jxja @
@yj (i1;2;. . .;n;j1;2;. . .;m):
(1:2)
Denote the horizontal gradient
rL(Z1;. . .;Zn;Zn1. . .;Znm);
and divL(u1;u2;. . .;un;un1 . . .;unm)nmP
i1 Ziui denotes the general- ized divergence. Then the second order degenerate elliptic operator andp- degenerate sub-elliptic operator are defined as
LaDx jxj2aDynmX
i1
Zi rL rL; and Lp;audivL jrLujp 2rLu
rL jrLujp 2rLu
; p>1;
respectively, whereDx andDy are Laplace operators onRn andRm, re- spectively.
La is usually called the B-G operator. This second order operator be- longs to the wide class of sub-elliptic operators introduced and studied by Franchi and Lanconelli in [13, 15]. Whenx60,Lais elliptic and becomes degenerate on the manifoldf0g Rm. Moreover, Fora1, the B-G op- erator can be viewed as Tricomi operator for transonic flow restricted to subsonic regions.
In particular, Caffarelli and Silvestre [6] established an equivalent defi- nition of the fractional LaplacianDs;s2(0;1), which can be written to a nondivergence form asLawithm1 by the suitable change of variables.
Thus it seems conveniently to study fractional Laplacian equation in some sense. Furthermore, a series of results obtained for B-G operator can be carried over to fractional Laplacian. For example, the fundamental solution of equationLau0 can apply to the fractional Laplacian problem, see [6, 19].
In recent years, the integral inequalities and partial differential equa- tions related to the generalized B-G vector fields have been paid much attention by many scholars (see, e.g., [8, 9, 11, 10, 19, 20]).
To state our main results, we need also some notations. There exists a natural family of anisotropic dilations attached toLa;i.e.,
dl(x;y)(lx;la1y); l>0; (x;y)2Rnm: (1:3)
The homogeneous dimension with respect to the dilations (1.3) is Qn(a1)m:
The distance function is written as
d:d(x;y) j(x;y)j (jxj2(a1)(a1)2jyj2)2(a1)1 : Takej(x;y)2Rnm;jjj d(j);x x
jjj;y y
jjja1;j(x;y);and write rjdr(j)dr(x;y). The open ball of radius R and centered at (0;0)2 Rnm is denoted by BL(R)BL((0;0);R) f(x;y)2Rnmjd(x;y)5Rg. Let
S: f(x;y)2Rnmjd(x;y)1gbe the B-G unit sphere andjSjbe the vo-
lume of theS. Obviously,j2S. Forf 2C(BL(R));let
~f(r) 1 jSj
Z
S
f(rj)dm(j)
be the avenge of f over the B-G sphere, here 05r5R;dm is the surface measureonS.DenoteN(f) kf ~fk1 kf ~fkL1(BL(R)) sup
j2BL(R)jf ~f(r)j.
One of the main results in this paper is
THEOREM 1.1. Assume that f 2C1(BL(R)) and 05a5n. Then for jrj, there holds
f(j) 1 jBL(R)j
Z
BL(R)
f(h)dh
N(f)G 2(a1)n a
G 2(a1)Q G 2(a1)n
G 2(a1)Q a
Q
Q1R r 2rQ1 (Q1)RQ
krLfk1: (1:4)
It seems that the estimate (1.4) is not sharp since jrLdj 61 with (x;y)6(x;0) (see next section).
Suppose thatf(j)f(d(j)) is the radial function. Then, we can improve the result of Theorem 1.1 as follows.
COROLLARY 1.2. Let f 2C1(BL(R)) be a radial function. Then for jrj, it holds
(1:5)
f(j) 1 jBL(R)j
Z
BL(R)
f(h)dh
N(f)
Q
Q1R r 2rQ1 (Q1)RQ
kf0k1:
LetVRnmbe a bounded domain, and let Lip(V) fu2C(V): ju(x) u(y)j Kjx yj
for someK>0 and for anyx;y2Vg:
We define
C10(V) fu2C1(V);rLu0 on@Vg;
and
W1;1(V): fu2C1(V);rLu2Lip(V)g:
Iff 2W1;1(V)\C10(V), we can extend the functionf by zero toFonBL(R), the smallest ball centered at the origin and containing V. We state the following theorem.
THEOREM 1.3. Assume that 05a5n and f 2W1;1(V)\C10(V).
Let RinffR0>0jVBL(R0)g. Then for all jrj2V and F2W1;1(BL(R)),
f(j) 1 jVj
Z
V
f(h)dh
N(F)
1 jVj
jBL(R)j 1
jVj Z
V
f(h)dh
G 2(a1)n a
G 2(a1)Q G 2(a1)n
G 2(a1)Q a Q
Q1R r 2rQ1 (Q1)RQ
krLfkL1(V): (1:6)
Note that Liu and Luan [18] established the Ostrowski type inequality associated with Carnot-CaratheÂodory distance dcc on the Grushin plane (a2;nm1). Since jrLdccj 1, the estimate (1.4) is sharp for Grushin plane.
A key process in the proof of our results is to establish the re- presentation formula related to the generalized B-G vector fields. We employ a new technique to prove the representation formula which is different from Lianet al.[16] and Donget al.[12], bases on the appropriate use of a suitable vector field and elementary calculations.
We also can obtain the following Hardy inequalities with boundary term by the above mentioned technique. It has the advantage that it
allows us to compute explicit constants for the remainder term. We point out that our method is different from M. Zhu and Z. Wang [23] and Adimurthiet al. [1, 2].
Let D1;p0 (V) be the completion of C10 (V) with respect to the norm k ku D1;p R
VjrLujpdjp1 .
THEOREM 1.4. Let R>0;15p51 and p6Q. Then for all u2D1;p(Rnm),
Z
BL(R)
jrLujpdj Q p
p p Z
BL(R)
cpajujp dp dj
Q p
p
Q p p
p 2 1 Rp 1
Z
@BL(R)
cpa jujp
jrdjdHQ 1; (1:7)
and
Z
BcL(R)
jrLujpdj Q p
p p Z
BcL(R)
cpajujp dp dj
Q p p
Q p p
p 2 1 Rp 1
Z
@BL(R)
cpa jujp
jrdjdHQ 1; (1:8)
whereBcL(R):RnmnBL(R);dHQ 1is the (Q 1)-dimensional Hausdorff measure and r is usual gradient onRnm. For further improvement of these inequalities see Sec. 4.
REMARK 1.5. The Hardy inequalities related to the generalized B-G vector fields have been established in space C01(V) or C10 (Rnm), see [8, 9, 11]. In this work, we extend these inequalities to D1;p(V) or D1;p(Rnm).
This paper is organized as follows. In section 2, we give some elemen- tary facts and provide a representation formula related to the generalized B-G vector fields for function without compact support. In section 3, we prove Theorem 1.1 and 1.3. In section 4, we derive a family of Hardy in- equalities with boundary term.
2. Arepresentation formula related to the generalized B-G vector fields Let us note that some essential differences appearing unavoidably. We recall some known facts about the generalized B-G vector fields (see, e.g., [15, 19, 8]).
We need the following concepts. A functionu:V Rnm!Ris said cylindrical, ifu(x;y)u(r;s) (i.e.,udepends only onr jxjands jyj), and in particular, if u(x;y)u(d(x;y)), we call u is radial. Let VBL(R2)nBL(R1) with 0R1R2 1;u2L1(V). Then, the change of polar coordinates defined in [19, 10]
(x1;. . .;xn;y1;. . .;ym)!(r;u;u1;. . .;un 1;g1;. . .;gm 1) allows the following formula to hold
Z
V
u(x;y)dxdy ZR2
R1
Z
S
u(dr(x;y))rQ 1dmdr
sn;m
ZR2
R1
rQ 1u rjsinuja11; ra1 a1jcos uj
dr;
(2:1)
where sn;m 1 a1 m
vnvmRa2
a1
jsinuja1n 1jcos ujm 1du, vn;vm are the Lebesgue measure of the unitary Euclidean sphere in Rn and Rm; re- spectively,a1;a2depend onnandm, see [19].
Let us recall some elementary facts about distance functiondd(x;y).
rLd jxja
d2a1 jxjax1;jxjax2;. . .;jxjaxn;(a1)y1;. . .;(a1)ym;
jrLdjpcpa jxjpa
dpa 1; Ladc2aQ 1
d ; Lp;adQ 1
d cpa; p1:
It is easy to seedandrLdare homogeneous of degree one and zero with respect to the dilations (1.3), respectively. From the formula ofrLd, it holds
rL(d2(a1))2(a1)d2a1rLd
2(a1)jxja jxjax1;jxjax2; ;jxjaxn;(a1)y1; ;(a1)ym (2:2) :
NoticingrL(d2(a1))2(a1)d2a1rLd, it is not difficult to obtain jrL(d2(a1))j2[2(a1)d2a1]2jrLdj2[2(a1)]2jxj2ad2(a1): (2:3)
Introduce the following two matrix Aa In O
O jxj2aIm
; sa In O
O jxjaIm
;
whereInandImare thenandmorder unit matrix, respectively. Thus, we can denote
LaudivL(ru)div(Aaru)div(sTasaru); and rLsar:
Letj(x;y)2Rnm;x x
jjj;y y
jjja1;j(x;y) andrjdr(j).
We havej2S f(x;y)2Rnmjd(x;y)1g. Arguing as in [14], we can obtain the following result from the polar coordinates and dilations (1.3).
LEMMA2.1. Iff 2L1(Rnm), then Z
Rnm
f(j)dj Z1
0
Z
S
f(rj)rQ 1dm(j)dr:
(2:4)
On the other hand, by Federer's coarea formula Z
RN
f(x)dx Z1
1
ds Z
gs
f(x)
jrg(x)jdHN 1;
where g2Lip(RN), dHN 1 denotes the (N 1)-dimensional Hausdorff measure andris usual gradient. It follows
(2:5) Z
BL(R)
f(j)dj ZR
0
dr Z
@BL(r)
f(j)
jrdjdHQ 1 ZR
0
dr Z
S
f(j)
jrdjrQ 1dHQ 1dr(j):
Combining (2.4) with (2.5), it yields Z
S
f(rj)dm Z
S
f(j)
jrdjdHQ 1dr(j) Z
@BL(r)
f(j)
dQ 1jrdjdHQ 1: (2:6)
Using the polar coordinates and (2.1), as in [7], we have LEMMA2.2. Suppose thatg> n, and
Cg Z
S
jxjgdm:
Then,
Cg 1 a1
m4pnm2 G2(a1)ng G(n2)G2(a1)Qg: In particular,
jSj C0 Z
S
dm 1 a1
m4pnm2 G2(a1)n G(n2)G2(a1)Q : PROOF. Forg> Q, it follows from (2.4) that
Cg Z
S
jxjgdm(Qg) Z1
0
rQg 1dr Z
S
jxjgdm
(Qg) Z1
0
Z
S
rQ 1jrxjgdmdr
(Qg) Z
BL(1)
jxjgdxdy:
Now, invoking (2.1), as in [19], it gets forg> n Cg (Qg)
Z
BL(1)
jxjgdxdy
(Qg) 1 a1 m
vnvm Z1
0
rQg 1dr Za2
a1
jsinujnga1 1jcos ujm 1du
1
a1 m
vnvm Za2
a1
jsinujnga1 1jcos ujm 1du
1
a1
m 4pnm2 G2(a1)ng G(n2)G2(a1)Qg: The proof of Lemma 2.2 is finished.
LEMMA2.3. Assume05R15R251and f 2C1(BL(R2)nBL(R1)). Then (2:7)
Z
S
f(R2j)dm Z
S
f(R1j)dm
Z
BL(R2)nBL(R1)
1 dQ(j)jxj2a
rLf;rL
d2(a1)(j) 2(a1)
dj:
PROOF. Let T be a C1 vector field on U:BL(R2)nBL(R1) and be specified later. For anyf 2C1(U), integrating by parts, we have
(2:8) Z
U
divLTfdj Z
U
divsTaTfdj Z
U
hT;rLfidj Z
@U
fsTahT;nidHQ 1;
wherenis the outer unit normal vector to the boundary.
Choose nowT 1 dQ(j)jxj2arL
d2(a1)(j) 2(a1)
. From (2.2), it is easy to see T rLd
dQ 2a 1jxj2a sard
dQ 2a 1jxj2a. We compute
divsTaTdivsTa
rLd dQ 2a 1jxj2a
Lad
dQ 2a 1jxj2a hrLd;rL(d2a1 Qjxj 2a)i
jxj2a d2a
Q 1
ddQ 2a 1jxj2a(2a1 Q)jxj 2ad2a QhrLd;rLdi
d2a1 QhrLd;rL(jxj 2a)i
Q 1
dQ 2a1 Q
dQ 2ad2a1 Qjxja
d2a1 jxj 2a 2jxja2
2a dQ
2a dQ0:
(2:9)
Note that n rd
jrdj and jrLdj2Aahrd;rdi. Combining (2.8) and (2.9), we obtain
(2:10) Z
U
1 dQ(j)jxj2a
rLf;rL
d2(a1)(j) 2(a1)
dj
Z
@U
fsTa
rLd dQ 2a 1jxj2a;n
dHQ 1
Z
@U
fsTa
rLd
dQ 2a 1jxj2a; rd jrdj
dHQ 1
Z
@U
f jrdj
jrLdj2
dQ 2a 1jxj2adHQ 1
Z
@BL(R2)
f
dQ 1jrdjdHQ 1
Z
@BL(R1)
f
dQ 1jrdjdHQ 1:
Invoking (2.6), we deduce the formula (2.7). Hence, this completes the proof
of Lemma 2.3. p
3. Ostrowski type inequalities
In this section, we prove Theorem 1.1 and 1.3. To begin with, we need the following estimates.
LEMMA3.1. Let05R15R251;05a5n and f 2C1(BL(R2)nBL(R1)).
Then Z
S
f(R2j)dm Z
S
f(R1j)dm
Z
BL(R2)nBL(R1)
1
dQ (a1)jxjajrLfjdj C ajR2 R1j krLfk1:
(3:1)
PROOF. Using the pointwise Schwartz inequality, it follows from Lemma 2.3 and (2.3) that
Z
S
f(R2j)dm Z
S
f(R1j)dm
Z
BL(R2)nBL(R1)
1
dQjxj2ajrLfj rL
d2(a1) 2(a1)
dj
Z
BL(R2)nBL(R1)
1
dQ (a1)jxjajrLfjdj
krLfk1 Z
BL(R2)nBL(R1)
1
dQ (a1)jxjadj :
Therefore, invoking Lemma 2.2, we get for 05a5n Z
S
f(R2j)dm Z
S
f(R1j)dm
krLfk1 Z
BL(R2)nBL(R1)
1 dQ 1jxjadj
krLfk1 ZR2
R1
dr Z
S
jxj adm
C ajR2 R1jkrLfk1:
The lemma is proved. p
PROOF OFTHEOREM1.1. Since
jBL(R)j jSjRQ Q ; it follows from this and Lemma 2.1 that
f(j) 1 jBL(R)j
Z
BL(R)
f(h)dh
jf ~f(r)j
1
jSj Z
S
f(rj)dm Q RQjSj
ZR
0
Z
S
f(rj)rQ 1dmdr
:N(f)I:
(3:2)
By Lemma 3.1 and 2.2, we have I
1 jSj
Z
S
f(rj)dm Q RQjSj
ZR
0
Z
S
f(rj)rQ 1dmdr
Q
RQjSj ZR
0
Z
S
f(rj)rQ 1dmdr ZR
0
Z
S
f(rj)rQ 1dmdr QC a
RQjSj ZR
0
jr rjrQ 1drkrLfk1
QC a
RQjSj RQ1
Q1 rRQ
Q 2rQ1 (Q1)Q
krLfk1
C a jSj
Q
Q1R r 2rQ1 (Q1)RQ
krLfk1
G 2(a1)n a
G 2(a1)Q G 2(a1)n
G 2(a1)Q a Q
Q1R r 2rQ1 (Q1)RQ
krLfk1: (3:3)
Combining (3.3) and (3.2), we obtain the conclusion. p
PROOF OF COROLLARY 1.2. Let f 2C1(BL(R)) be a radial function.
From Lemma 2.3 and 2.2, the following estimate holds Z
S
f(R2j)dm Z
S
f(R1j)dm
Z
BL(R2)nBL(R1)
1
dQjxj2ajrLfkrL( d2(a1) 2(a1))jdj
Z
BL(R2)nBL(R1)
1
dQ (a1)jxjajf0krLdjdj kf0k1
Z
BL(R2)nBL(R1)
1 dQ 1dj
C0jR2 R1jkf0k1: Similar to (3.3), we can get
I 1
jSj Z
S
f(rj)dm Q RQjSj
ZR
0
Z
S
f(rj)rQ 1dmdr C0
jSj Q
Q1R r 2rQ1 (Q1)RQ
kf0k1
Q
Q1R r 2rQ1 (Q1)RQ
kf0k1:
Hence, we obtain (1.5) from the proof of Theorem 1.1. p PROOF OFTHEOREM 1.3. Assume thatRinffR0>0jVBL(R0)g.
Then
F(j): f(j); if j2V;
0; if j2BL(R)nV;
(
and satisfiesF2W1;1(BL(R))\C01(BL(R)):We can splitf(j) 1 jVj
R
Vf(h)dh as
(3:4)
f(j) 1 jVj
Z
V
f(h)dh
F(j) 1 jBL(R)j
Z
BL(R)
F(h)dh
1 jBL(R)j
Z
BL(R)
F(h)dh 1 jVj
Z
V
f(h)dh
:I1I2:
By the definition ofR, we can obtain I2
1 jBL(R)j
Z
BL(R)
F(h)dh 1 jVj
Z
V
f(h)dh
1
jBL(R)j Z
V
f(h)dh 1 jVj
Z
V
f(h)dh
1
jBL(R)j 1 jVj
Z
V
f(h)dh
1 jVj
jBL(R)j 1
jVj Z
V
f(h)dh : (3:5)
Further, replacingBL(R) byVin Theorem 1.1, it follows
(3:6) I1
F(j) 1 jBL(R)j
Z
BL(R)
F(h)dh kF Fk~ 1G 2(a1)n a
G 2(a1)Q G 2(a1)n
G 2(a1)Q a Q
Q1R r 2rQ1 (Q1)RQ
krLfkL1(V):
Substituting (3.5) and (3.6) into (3.4), we conclude (1.6). Theorem 1.3 is
established. p
4. Hardy inequalities with boundary term
In this section, we present some Hardy inequalities with boundary term which improves the Hardy inequality established in [8, 9, 11]. Here we point out that our interests is to establish inequalities, the best constants of inequalities are not discussed (due to the optimality of the constants in the inequalities is of its independent interest).
PROOF OFTHEOREM1.4. LetTbe aC1 vector field onBL(R) and let it be specified later. For any u2D1;p(Rnm), integrating by parts, we have
(4:1) Z
BL(R)
divLTjujpdj p Z
BL(R)
hT;jujp 2urLuidj Z
@BL(R)
jujpsTahT;nidHQ 1:
Using HoÈlder's inequality and Young's inequality we obtain
(4:2) p
Z
V
hT;rLuijujp 2udjp Z
BL(R)
jrLujpdj 1p Z
BL(R)
jTjpp1jujpdj pp1
Z
BL(R)
jrLujpdj(p 1) Z
BL(R)
jTjpp1jujpdj:
Thereby, substituting (4.2) into (4.1), it follows (4:3)
Z
BL(R)
divLT (p 1)jTjpp1
jujpdj Z
BL(R)
jrLujpdj Z
@BL(R)
jujpsTahT;nidHQ 1:
Choose nowTAjAjp 2jrLdjp 2rLd
dp 1 , where AQ p
p forp6Q:A di- rect computation gets
divLT (p 1)jTjpp1jujp jAjp 1jrLdjp dp ; and note thatn rd
jrdj:Combining these with (4.3) yields jAjp
Z
BL(R)
cpajujp dp dj
Z
BL(R)
jrLujpdjAjAjp 2 1 Rp 1
Z
@BL(R)
cpa jujp
jrdjdHQ 1: Thus, inequality (1.7) is proved.
A similar argument can prove inequality (1.8), so we omit the details
here. p
In the same spirit with Theorem 1.4, we can obtain the following Hardy inequality with boundary term on bound domain.
THEOREM4.1. Let15p5Q. IfVRnmis a smooth domain with@V being bounded and062@V, then there exists R0inffR>0jVBL(R)g such that for all u2D1;p(Rnm)
Z
V
jrLujpdj Q p
p pZ
V
cpajujp dp dj
Q p p
p 1 1 Rp0 1
Z
@V
cpa jujp
jrdjdHQ 1; (4:4)
In addition, ifVRnm is a bounded and star-sharped with respect to the origin, there exists R1inffR>0jV BL(R)g such that for all u2D1;p(Rnm)
Z
Vc
jrLujpdj Q p
p pZ
Vc
cpajujp dp dj
Q p p
p 1 1 Rp1 1
Z
@V
cpa jujp
jrdjdHQ 1; (4:5)
whereVc:RnmnV:
In addition, J. Dou, Q. Guo and P. Niu [11] had established the following Hardy inequalities with remainder term.
PROPOSITION 4.2. Let 02V be a bounded domain in Rnm and 15p51.
(1): If p6Q, then there exists a positive constant R0sup
j2Vd(j)such that for any R>R0 and all u2D1;p0 (Vnf0g),
(4:6) Z
V
jrLujpdj Q p
p pZ
V
cpajujp dp dj
p 1 2p
Q p p
p 2Z
V
cpajujp dp
lnR
d 2
dj:
In particular, if 2p5Q, one can takesup
j2Vd(j)R0. (2): If pQ, then there exists R>sup
j2Vd(j) such that for all u2D1;p0 (Vnf0g)
Z
V
jrLujpdj
p 1 p
pZ
V
cpa jujp dln Rdpdj:
(4:7)
We note that a simple density argument can show thatD1;p0 (Vnf0g) D1;p0 (V). Hence, inequalities (4.6) and (4.7) can be extended to the space D1;p(V).
THEOREM4.3. Let02V be a bounded domain inRnmwith smooth boundary and15p51.
(1): If p6Q, then there exists a small positive constant M0such that for any R>R0:eM10sup
j2Vd(j)and all u2D1;p(V), Z
V
jrLujpdj Q p
p pZ
V
cpajujp dp dj p 1
2p
Q p p
p 2Z
V
cpajujp dp
lnR
d 2
dj
Cp;Q;R0 Z
@V
cpa jujp
jrdjdHQ 1; (4:8)
where Cp;Q;R0
p Q
p
Q p p
p 2 1Q pp 1M0aM20
Rp0 1 for 15p52;05a5(2 p)(p 1) 6p2A2 ; p Q
p
Q p p
p 2 1Q pp 1M0
Rp0 1 for 2p5Q;
Q p p
Q p p
p 2 Q pp 1M0aM02
Rp0 1 for p>Q;0>a>(2 p)(p 1) 6p2A2 : 8>
>>
>>
>>
>>
>>
><
>>
>>
>>
>>
>>
>>
:
(2): If pQ, then there exists R>sup
j2Vd(j)and the definition of M0
as above such that for all u2D1;p(V) (4:9)
Z
V
jrLujpdj
p 1 p
pZ
V
cpa jujp dln Rdpdj
p 1 p
p 1M0p 1 Rp0 1
Z
@V
cpa jujp
jrdjdHQ 1:
PROOF. LetB(s) 1
ln (s);s2(0;1) andAQ p
p . ForR>sup
j2Vd(j), there exists a constantM>0 such that
0 B d(j)
R
M; j2V:
Furthermore, rLBg
d R
gBg1 Rd rLd
d ; dBg rR
dr gBg1 Rr
r ; for allg2R;
and Zb
a
Bg1(s) s ds1
g
Bg(b) Bg(a) :
(1) Forp6Q:Let Tbe aC1 vector field onV and let it be specified later. For anyu2D1;p(V), similar to (4.3), it holds
(4:10) Z
V
[divLT (p 1)jTjpp1]jujpdj Z
V
jrLujpdj
Z
@V
jujpsTahT;nidHQ 1:
We choose
T(d)AjAjp 2jrLdjp 2rLd
dp 1 1p 1 pA B
d R
aB2 d
R
!
; whereais a parameter to be chosen later.
As in [11], take (i) a>(2 p)(p 1)
6p2A2 >0 for 15p525Q, (ii) a0 for 2p5Q,
(iii) a5(2 p)(p 1)
6p2A2 50 forp>Q.
Thus, it is not difficult to chooseM0(small enough) in all cases such that for 05B M0, and then
divLT (p 1)jTjpp1 jAjpjrLdjp dp
1p 1 2pA2B2
d R
: (4:11)
Noting that the conditionB M0 is equivalent toRR0:eM10sup
j2Vd(j) andn rd
jrdj;ifp5Q, then
Z
@V
jujpsTahT;nidHQ 1
Z
@V
AjAjp 2jrLdjp dp 1
1p 1 pA B
d R
aB2 d
R jujp
jrdjdHQ 1
AjAjp 2 1ppA1M0aM20 Rp0 1
Z
@V
cpa jujp
jrdjdHQ 1; (4:12)
and substituting (4.11) and (4.12) into (4.10), we conclude inequality (4.8). If p>Q, we have
(4:13) Z
@V
jujpsTahT;nidHQ 1AjAjp 2
p 1
pA M0aM20 Rp0 1
Z
@V
cpa jujp
jrdjdHQ 1;
where the factsA50;a50 is used.
(2) ForpQ:ChooseT(d) p 1
p p 1
jrLdjp 2rLd dp 1 Bp 1
d R
, as in [11], we can obtain
divLT (p 1)jTjpp1
p 1 p
p Bp
d R
jrLdjp dp : (4:14)
Notice
(4:15) Z
@V
jujpsTahT;nidHQ 1 Z
@V
p 1 p
p 1 jrLdjp
dp 1 Bp 1 d
R jujp
jrdjdHQ 1
p 1
p p 1
Mp0 1 Rp0 1
Z
@V
cpa jujp
jrdjdHQ 1:
Combining (4.14) and (4.15) yields inequality (4.9). p Acknowledgements. This work was partially supported by the National Natural Science Foundation of China (Grant No. 11101319), the Foundation of Shaanxi Province Education Department (Grant No. 2010JK549) and Zhejiang Provincial Natural Science Foundation of China (Grant No.
Y6110118).
REFERENCES
[1] ADIMURTHI,Hardy-Sobolev inequality in H1(V)and its applications, Com- mun. Contemp. Math.,4(2002), pp. 409-434.
[2] ADIMURTHI- M. ESTEBAN,An improved Hardy-Sobolev inequality in W1;p(V) and its application to Schrodinger operator, NoDEA Nonl. Diff. Equa. Appl., 12(2005), pp. 243-263.
[3] G. A. ANASTASSIOU,Quantitative Approximations, Chapman and Hall/CRC, Boca Raton/New York, 2001.
[4] G. A. ANASTASSIOU - J. A. GOLDSTEIN, Ostrowski type inequalities over Euclidean domains. Rend. Linc. Matem. Appl.,18(2007), pp. 305-310.
[5] G. A. ANASTASSIOU - J. A. GOLDSTEIN, Higher order Ostrowski type inequalities over Euclidean domains, J. Math. Anal. Appl., 337 (2008), pp. 962-968.
[6] L. CAFFARELLI- L. SILVESTRE,An extension problem related to the fractional Laplacian, Comm. Part. Diff. Eqs.,32(2007), pp. 1245-1260.
[7] W. COHN- G. LU, Best constants for Moser-Trudinger inequalities on the Heisenberg group, India. Univ. Math. J.,50(2001), pp. 1567-1591.
[8] L. D'AMBROSIO,Hardy inequalities related to Grushin type operators, Proc.
Amer. Math. Soc.,132(2004), pp. 725-734.
[9] L. D'AMBROSIO,Hardy type inequalities related to second order degenerate differential operators, Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 4 (2005), pp. 451-486.
[10] L. D'AMBROSIO- S. LUCENTE,Nonlinear Liouville theorems for Grushin and Tricomi operators, J. Diff. Eqs.,193(2003), pp. 511-541.
[11] J. DOU- G. GUO- P. NIU,Hardy inequalities with remainder terms for the generalized Baouendi-Grushin vector fields, Math. Ineq. Appl., 13 (2010), pp. 555-570.
[12] Y. DONG- G. LU- L. SUN, Global Poincare Inequalities on the Heisenberg Group and Applications, Acta Mathematica Sinica, English Series.,23(2007), pp. 735-744.
[13] B. FRANCHI- E. LANCONELLI,HoÈlder regularity theorem for a class of linear nonuniformly elliptic operators with measurable coefficients, Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4)10(1983), pp. 523-541.
[14] G. B. FOLLAND - E. M. STEIN, Hardy spaces on homogeneous groups, Princeton University Press, Princeton, NJ, 1982.
[15] B. FRANCHI - E. LANCONELLI, An embedding theorem for Sobolev spaces related to non-smooth vector fields and Harnack inequality, Comm. Part.
Diff. Eqs.,9(1984), pp. 1237-1264.
[16] B. LIAN- Q. YANG,Ostrowski type inequalities on H-type groups, J. Math.
Anal. Appl.,365(2010), pp. 158-166.
[17] Z. LIU, Some Ostrowski type inequalities, Math. Comp. Model., 48 (2008), pp. 949-960.
[18] H. LIU- J. LUAN,Ostrowski type inequalities in the Grushin plane, J. Ineq.
Appl. 2010, Article ID 987484, pp. 1-9.
[19] P. NIU- J. DOU- H. ZHANG,Nonexistence of weak solutions for the p-degenerate subelliptic inequalities constructed by the generalized Baouendi-Grushin vector fields, Georgian Math. J.,12(2005), pp. 723-738.
[20] P. NIU- Y. OU- J. HAN,Several Hardy type inequalities with weights related to generalized greiner operator, Canad. Math. Bull.,53(2010), pp. 153-162.
[21] A. OSTROWSKI,UÈ ber die Absolutabweichung einer differentiebaren Funktion von ihrem Integralmittelwert. Comment. Math. Helv.,10(1938), pp. 226-227.
[22] M. WANG- X. ZHAO,Ostrowski type inequalities for higher-order derivatives, J. Ineq. Appl. 2009, Article ID 162689, pp. 1-8.
[23] M. ZHU- Z. WANG,Hardy inequalities with boundary terms, Electron. J. Diff.
Eqs., No. 43 (2003), 8 pp.
Manoscritto pervenuto in redazione il 3 Gennaio 2012.