Restriction Theorems on Métiver Groups Associated to Joint Functional Calculus∗
2018-11-19HepingLIUAnZHANG
Heping LIU An ZHANG
Abstract The authors get on Métivier groups the spectral resolution of a class of operators m(L,−∆z),the joint functional calculus of the sub-Laplacian and Laplacian on the centre,and then give some restriction theorems together with their asymptotic estimates,asserting the mix-norm boundedness of the spectral projection operatorsfor two classes of functions m(a,b)=(aα +bβ)γ or(1+aα +bβ)γ,with α,β >0, γ 6=0.
Keywords Restriction operator,Métivier group,Functional calculus
1 Introduction
In this paper,we extend the mix-norm boundedness obtained by V.Casarino and P.Ciatti[4]to general two classes of projection operators(we call projector for simplicity)on Métivier groups(a class of 2-step nilpotent Lie group, first defined and studied by G.Métivier in[10]).Métivier group class is strictly more general than the H-type group class introduced by A.Kaplan[5],with the Heisenberg group being the only special one(of H-type)with 1-dimensional centre.
The “restriction-type” operator(spectral projector)we study acts on the central variables by the Euclidean Fourier transform while acting on the “space-v” variables by the spectral projection of the twisted Laplacian.As the quotient of a Métiver group corresponding to the hyperplanes in the centre is isomorphic to the Heisenberg group,we can use the spectral projector on the Heisenberg group to estimate that on the Métivier group by a “partial” Radon transform.For the Fourier transform on the central variables,we use the famous Tomas-Stein theorem.Our result includes not only homogeneous(like L2−∆z)but also inhomogeneous operators(like full Laplacian∆G=L−∆z),and also cover the uniform-norm boundednessLp→Lp′with exponents in the corresponding range.
First we recall some histories.The restriction problem on Rn,denoted byRS(p→q),cares theLp(Rn)→Lq(S)boundedness of the Fourier transform for any hypersurfaceSwith boundary endowed with the Lebesgue surface measure dσ.It has many useful applications in both harmonic analysis and PDE.The restriction problem with respect to the unit sphere(more generally for any compact hypersurface with boundary and non-vanishing Gaussian curvature everywhere)is given in the following conjecture.
Conjecture 1.1(Stein’s Restriction Conjecture)

The conjecture has been proved forq=2 orn=2 by P.Tomas,E.Stein and C.Fefferman.Forn=3,J.Bourgain and L.Guth have recently gotten the best result so far in[2],where they proved the dual extension theoremESn−1(∞ →p′)forp′>.The authors used the method of multilinear theory from[1]together with the Kakeya maximal estimate due to T.Wolff[17]and improved a bit an old result of T.Tao?Tao’s bilinear approach givesp′>?.In a word,some exciting progresses have been made while the whole picture is still far from known.
Recalling the famous Tomas-Stein theorem,due to P.Tomas and E.Stein[16](Stein’s result for the endpoint is unpublished),they proved Conjecture 1.1 forq=2.It corresponds to theLp→Lp′boundedness of the convolution operatorf∗,which is just the spectral projector of the positive Laplacian −∆ whose symbol is|ξ|2.In[13],R.Strichartz study analogues of this in other settings from a viewpoint of harmonic analysis as spectral theory of sub-Laplacians.Motivated by this idea,D.Müller[11]proved a mix-norm1We will use the same mix-norm through:The(p,q)-type mix-norm of a function f on the measure space X×Y is defined by=Y ,which is reduced to the Lpnormwhen p=q.boundednessof the restriction operator associated to the sub-Laplacian on the Heisenberg group,using a bound of the spectral projector of the twisted Laplacian The exponent onthe centre is trivial because of the trivial 1-dimensional Tomas-Stein theorem.So when centre dimension is bigger than one,it is reasonable to get restriction theorems for exponents of wider range(see[8–9,14–15]for related results).In[4],Casarino and Ciatti used an improved sharp bound of kΛkkLp→Lqto get a greatly improved mix-norm bound for the restriction operators associated to the sub-Laplacian and full Laplacian on the Heisenberg group and simultaneously an analogue for sub-Laplacian on the Métivier group.More precisely,on a Métivir group with dimension 2n+d,wheredis the dimension of the centre,we have the following theorem of Casarino and Ciatti about the restriction operatorassociated to the sub-Laplacian L(details will be explained later).
Theorem 1.1Given1≤p≤2≤q≤2,1≤r≤,we have

with

In[8],the first author advised to compute some joint functional calculus of the Laplacians including the full Laplacian, first done by M.Song,on groups of H-type.Combining the two works,we obtain the mix-norm bound in[4]for the joint functional calculus in[8].V.Casarino and P.Ciatti[3]also got a similar bound for the full Laplacian without asymptotic characterization.Our main result is the following theorem about two classes of operatorsm(L,−∆z)from the joint functional calculus of the sub-Laplacian L and Laplacian on the centre−∆z.
Theorem 1.2Given α,β>0,γ6=0,1≤p≤ 2≤q≤ ∞,(d,p,q)6=(1,2,2),1≤r≤m2(a,b)=(1+aα+bβ)γ,we have

with

and

The range of exponent r is already sharp from the Tomas-Stein theorem and the restriction operatorand constants A,B,C,D dependent on n,d,p,q,r,α,β,γ will be given later by(2.5)and(2.7)in Section2.
We arrange the remaining part of our paper in the following order.In Section 2,we will give the main result without proof after introducing basic notations and the restriction operators by functional calculus on the Métivier group.Following the theorems,some remarks will also be given.In Section 3,we will put our attention to the detailed proofs of the main result given in Section 2.
Concerning the boundedness we care in this note,we introduce two notations:We will always useto indicate that the left side is less than a constant multiple2We do not care the detailed expression of the constant in the present formula.of the right side,whileto mean that the constant is dependent onλ,···;We also usefor “almost equal”,accurately,the quotient is bounded both from above and below.
2 Main Results
In Subsection 2.1,we introduce the Laplacians on the Métivier Group and give the definition and explicit formula of the restriction operators associated to the joint functional calculus of the sub-Laplacian and Laplacian on the central variables.By the partial Randon transform,we can connect our Métivier group with the Heisenberg group and then use the bound of spectral projector of the twisted Laplacian on the Heisenberg group(scaled special Hermite projector)to estimate the mix-norm bound of the restriction operators associated to the Laplacians on the Métivier group.In Subsection 2.2,we list two detailed main theorems,but leave their proofs in the next section.
2.1 Restriction operators on the Métivier group
First,we will give some definitions.We take many notations and terminologies from[4,15],to which the reader can refer if is not very familiar with them.LetGbe a connected,simply connected,two-step nilpotent Lie group,associated with Lie algebra g,endowed with an inner product h·,·i.The Lie algebra g can be decomposed into the direct sum g=z+v,with the centre z and its complement v.Take

both which we assume are always positive integers,and z∗denotes the dual of z with dual norm|·|induced by the inner productin z.The unit ball in the dual space z∗is denoted byS={ω∈ z∗,|ω|=1}.For eachω∈S,we can find a unique unitZω∈ z such thatω(Zω)=1.Then the centre z can be decomposed as z=RZω+kerω,with the quotient isomorphism z/kerωRZω.Denote gω=RZω+v,then we have isomorphism g/kerωgω.As kerωis an ideal of z,gωis a Lie subalgebra.We denote byGωthe connected simply connected Lie subgroup ofG,associated with Lie algebra gω.We define the Métivier property by the following non-degeneracy description.
Definition 2.1Bilinear function Bω(U,V)ω([U,V])with U,V∈v,ω∈S,is called non-degenerate,if

Definition 2.2Group G discussed above is called a Métivier Group,if Bωis non-degenerate for all ω∈S.
D.Müller and A.Seeger[12]gave an example that is a Métivier group but not of H-type.For completion,we explain it here:Given Lie algebra g=v+z=R8+R2,with the Lie bracket

whereu,vare the coordinates ofU,Vin an orthonormal basis,and matrix

As|Jz|=0 unlessz=0,we see that g is Métivier with non-degeneracy property(2.1).Now we try to say that g is not of H-type.Actually,assume that there is another H-type Lie algebra g′=v′+z′(in abuse of notation,a map3above,is orthogonal whenever|z′|=1,and we sometimes identify z and z∗)and a Lie algebra isomorphismα:g → g′,then under orthonormal basis,α=(notice thatαpreserves the centers and induces an isomorphism z → z′).By the conservation of Lie brackets,for anyu,v,z,we have



which implies a contradiction

For Métivier groupG,the dimension of v is even from the non-degeneracy and skew-symmetry,which we denote by dimv=2n,then Lie subgroupGωHn,the Heisenberg group with Lie algebra hn=Cn+R.We will use the spectral decomposition of the Laplacians on the Heisenberg group to get its counterpart on the Métivier group,and finally obtain the corresponding restriction operators.
By the nilpotency ofG,we can parameterizeGby its Lie algebra z+v,through the surjective exponential map.Fix a basis of Lie algebra g,

then we can endow every point of groupGwith an exponential coordinate(V,Z)or(v,z)∈R2n×Rd.By Baker-Campbell-Hausdorff formula,we get the multiplication law

withV,V′∈v,Z,Z′∈ z.Simple computation gives the left-invariant vector fields

associated respectively to one-parameter subgroups{(sVj,0)|s∈R}and{(0,tZi)|t∈R}.These 2n+dvector fields form a basis of the tangent bundle ofG.Now we can define onGthe sub-Laplacian,the Laplacian on the centre,and the full Laplacian respectively to be

Hömander’s theorem tells that the sub-Laplacian and full Laplacian are hypoelliptic,positive,and essentially self-adjoint,and the Laplacians play an important role in harmonic analysis on the group.
The partial Radon transform on the central variables is a powerful tool for us to get the spectral decomposition on Métivier group from that on the Heisenberg group.
Definition 2.3For ω∈S,f∈S(G),the Schwartz space on G,we define the partial Radon transform of f to be

Lemma 2.1For Schwartz functions f∈S(G),g∈S(Gω),we have the following formulas:
(1)Sub-Laplacians on G and Gω:

whereare respectively the left-invariant vectorfield and sub-Laplacian on Gω.
(2)Sub-Laplacian and twisted Laplacian on Gω:

wheremeans the inverse Fourier transform on the one dimensional centre,and=are respectively the λ-twisted left-invariant vectorfield and λ-twisted Laplacian on Gω.
(3)Sub-Laplacian and λω-twisted Laplacian on G:

wheremeans the inverse Fourier transform on the centre,andare respectively the λω-twisted left-invariant vector field and λω-twistedLaplacian on G.
Remark 2.1Lemma 2.1 tells us that theλω-twisted Laplacian onGis nothing but theλ-twisted Laplacian onGω.This gives the idea how to get the spectral projection of the twisted Laplacian onG.
Observing the non-degeneracyproperty(2.1)of skew-symmetric bilinear functionBω,we can use an invertible linear transformAωto change the bilinear function to the standard symplectic formIn this new coordinates,denoted e.g.by,theλ-twisted Laplacian onGωis then

which is just the usualλ-twisted LaplacianLλon the Heisenberg group Hn,and actually we have educed an isomorphism(non-isometric)betweenGωand Hn.Then we get the spectral decomposition of theλω-twisted Laplacian ofGin the following proposition.
Proposition 2.1For g∈ S(v),ω∈S,take gω=g◦(Aω)−1and denote bythe spectralprojector of the λω-twisted Laplacian on G,then

whereis the spectral projection of the λ-twisted Laplacian Lλon the Heisenberg groupHn,given by the twisted convolution

for z∈Cn,and the special Hermite function

−1and degree k.
Now,we use the spectral projectorto give the restriction operators associated to the joint functional calculus of L and−∆z.From the inverse Fourier transform formula on the central variables,polar coordinates transformation,and spectral expansion(2.2),we can get the following expansion forf∈S(G),

Since thek-term in the sum is the joint eigenfunction4See for L from Lemma 2.1(3)and Proposition 2.1 and it is obvious for−∆z.of L and−∆z,associated to the spectrum rayRk=((2k+n)λ,λ2),we can naturally define the functional calculus operator associated to a functionmby given a “proper”msuch thatm?(2k+n)λ,λ2?is differentiable,positive,and strictly monotonic on R+with regards toλ.Generally,given a spectral decomposition of operatorD,


the associated restriction operator(spectral projector)can be defined as

and similarly,the restriction operator associated tom(D)can be defined as

Now,we will care the operatorm(L,−∆z)onG.∀µ∈ R+,µkdenotes the solutionλof the equationm((2k+n)λ,λ2)=µanddenotes the derivative relative toµ,and we will write simplyto mean the restriction operator,whose formula is given in the following theorem.
Theorem 2.1∀f∈S(G),f have the following expansion:
under the spectral decomposition

whereis the restriction operator defined by

ProofBy changing variables from(2.3)or(2.4),

From Lemma 2.1 and Proposition 2.1,we have,after taking

that

and

which finally gives

2.2 Mix-norm boundedness of the restriction operator
In[4],V.Casarino and P.Ciatti have given the mix-norm boundedness of the restriction operatorassociated to the sub-Laplacian L.We now use the similar method to get the mixnorm boundedness of the restriction operators associated to a class of operatorsm(L,−∆z).In this subsection,we will state our main theorems of this paper in detail.
First give some notations.We define a functionφon??by

wheres∗=,from which we see that bothφands∗are related to dimensionn.Givenα,β∈R+,γ∈R∗,1≤p≤ 2≤q≤ ∞,1≤r≤,we define four numbers relative to(α,β,γ,p,q,r,n,d):

Now,we can state the boundedness theorems of restriction operators Pmµ.
Theorem 2.2Given α,β>0,γ6=0,1≤p≤ 2≤q≤ ∞,1≤r≤,(d,p,q)6=(1,2,2)and m(a,b)=(aα+bβ)γ,we have for all f∈ S(G),

with

Remark 2.2(1)The theorem can be described in several cases relative to parameters,γ,and alsoµ,which can be seen from the following table:

(2)The theorem contains the inhomogeneous operator— full Laplacian∆G,whenα=β=γ=1(also homogeneous ones like L2−∆z).
(3)The mix-norm bound covers the uniform-norm bound whenp=q′=r,especially that on H-type groups,when the exponent function is degenerated to an easy form.Actually,from the Clifford algebra of H-type groups,we have a dimension relationd<2n(see[6]),which tellsp=q′=ra critical point forAandB,more precisely,see(2.6),and

which coincides with the result of[8],just like the case for the sub-Laplacian,when the result of[4]coincides with that of[9]on H-type groups although slightly different arguments are applied.
Theorem 2.3Given α,β>0,γ6=0,1≤p≤ 2≤q≤ ∞,1≤r≤,(d,p,q)6=(1,2,2)and m(a,b)=(1+aα+bβ)γ,we have for all f∈ S(G),

with

Remark 2.3(1)We have a similar table as last theorem(we give the caseγ<0,which we care more about):

(2)The new approximating situationµ→ 1± is similar,as we get a similar control ofµkand.Our result includes many useful operators like the resolvent(I+L)−1or(I+ ∆G)−1.
3 Proof of the Main Results
The following important sharp estimate due to H.Koch and F.Ricci[7],about theLp→L2bound of the spectral projector of twisted Laplacian for 1≤p≤2,is critical in our proof,

3.1 Series bound for general m
Lemma 3.1Let,given in Proposition2.1,be the spectral projection operator on the Heisenberg groupHn.Then for1≤p≤2≤q≤∞,we have

ProofFrom duality and the projection property of Λk=,we see Λk==Λk,and the sharpLp→L2estimate(3.1)for 1≤p≤2 gives the followingLp→Lqestimate for general exponents 1≤p≤2≤q≤∞,


where we use dilationδλg(·)=g(λ·).So we have

and the hidden ignored constant is dependent onp,q,nfrom(3.1).Therefore,we get the expected bound of.
Theorem 3.1Given1≤p≤2≤q≤ ∞,1≤r≤and“proper”m(·,·),then for allf∈S(G),we have

with

If m(·,·)is good enough,the sharp constant is finite.
ProofUsing the relation between the spectral projection of theλω-twisted Laplacian onGand that on Heisenberg group Hnin Proposition 2.1,we have

with regards toωon the unit sphere,we can assume|Aω|∼ 1,so from(3.2)in last Lemma 3.1,we have

Denote by5In abuse of notation.the dual action of two functions respectively in two dualLpandLp′spaces or mix-norm spaces onG,i.e.,

By the formula(2.5)of the restriction operatorin Theorem 2.1,changing integral orders,and then using orderly the Höder inequality,Lp→Lqbound(3.4)of,Cauchy-Schwartz inequality,Minkowski inequality(glancing at the exponentp,q′≤ 2),and finally the
Tomas-Stein theorem,we get for anyf,g∈S(G),

By duality,we have proved the bound(3.3)in the theorem.
3.2µ-dependent bound for two special classes of m
Now,with the series bound control(3.3)associated to general proper functional calculus in Theorem 3.1,we are going to get more sophisticatedµ-dependent control for two special cases of functionals(Lα+(−∆z)β)γand(1+Lα+(−∆z)β)γwithα,β>0,γ6=0.In short,we come to prove our main results:Theorems 2.2–2.3.
Proof of Theorem 2.2For Theorem 2.2,we consider operators(Lα+(−∆z)β)γ,associated to

Then we have the following easy estimates forµk,the solutionλof equation((2k+n)λ)α+λ2β=µ1γ,

So from(3.3)in Theorem 3.1,we have

We consider in two cases6The two cases can also be divided into several more detailed cases,see table in first term of the remark following Theorem 2.2..
Case 1When≤ 1.
In this case,the first term in(3.6)can be discarded,so after inserting the estimate(3.5),we have

Actually,in order to derive the last inequality,it suffices to check the exponent of the power series.First,we denote7p∗=and the exponent

Using 1≤r≤we check it in four cases corresponding to the piecewise functionφin(2.6):

this case is equivalent to item(b).

ν=−−1,andν<−1 unlessd=1,r=1,p=q=2,which is just the bad endpoint case on the Heisenberg group.
Case 2When>1.

Here,we need to check the exponent of the series for I1.We denote the exponent by

thenν1>−1,as the worst case is thatp→p∗andq→,then

Then from the formulas of the exponents in(2.7)–(2.8)and hence Theorem 2.2 is proved.
Proof of Theorem 2.3For Theorem 2.3,we consider operators(1+Lα+(−∆z)β)γ,associated to

We may assumeγ<0,µ∈ (0,1)andµkis the solution of equation 1+?(2k+n)λ?α+λ2β=,which is easily seen to be strictly decreasing relative toµ.We naturally consider two boundary cases

First we have similar boundµk≤ minor given in specific cases,

By the decreasing ofµkcorresponding to not onlyµbut alsok,we also have

From the last two estimates of{µk},which is similar to that in the proof of Theorem 2.2,we can repeat the proof there,and so similar is the form of the conclusions in two main theorems.For caseµ→ 0+,the estimate of{µk}is absolutely the same as that in Theorem 2.2,so is the proof process.For caseµ→ 1−,the process is similar.By(3.3)in Theorem 3.1 and(3.7)–(3.8),we have

Again we discussed it in two cases.
Case 1When≤1.
Still the first term in(3.9)can be omitted,then
The exponent of the series is the same as that in the proof of Theorem 2.2,i.e.,it equalsνand<−1,wiping off the bad endpoint in the Heisenberg case.
Case 2When>1.
There are also two terms to estimate.

Then from(2.7)and(2.9),therefore,Theorem 2.3 is proved.
AcknowledgementThe authors thank sincerely the anonymous referees for careful reading of the manuscript.
杂志排行
Chinese Annals of Mathematics,Series B的其它文章
- A Schwarz Lemma for Harmonic Mappings Between the Unit Balls in Real Euclidean Spaces∗
- Ribbon Hopf Superalgebras and Drinfel’d Double∗
- On the Waring-Goldbach Problem for Six Cubes and Two Biquadrates∗
- Some Weighted Norm Inequalities on Manifolds∗
- Strongly Gauduchon Spaces
- Variation Inequalities Related to Schrödinger Operators on Morrey Spaces∗
