Volume 2013, Article ID 909782,21pages
http://dx.doi.org/10.1155/2013/909782
Research Article
On the Spectral Asymptotics of Operators on Manifolds with
Ends
Sandro Coriasco and Lidia Maniccia
Dipartimento di Matematica, Universit`a degli Studi di Torino, V. C. Alberto, n. 10, I-10123 Torino, Italy
Correspondence should be addressed to Sandro Coriasco; sandro.coriasco@unito.it Received 28 September 2012; Accepted 16 December 2012
Academic Editor: Changxing Miao
Copyright Β© 2013 S. Coriasco and L. Maniccia. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
We deal with the asymptotic behaviour, forπ β +β, of the counting function ππ(π) of certain positive self-adjoint operators
P with double order(π, π), π, π > 0, π ΜΈ= π , defined on a manifold with ends M. The structure of this class of noncompact
manifolds allows to make use of calculi of pseudodifferential operators and Fourier integral operators associated with weighted
symbols globally defined onRπ. By means of these tools, we improve known results concerning the remainder terms of the Weyl
Formulae forππ(π) and show how their behaviour depends on the ratio π/π and the dimension of M.
1. Introduction
The aim of this paper is to study the asymptotic behaviour, for π β +β, of the counting function
ππ(π) = β
ππβ€π
1, (1)
where π1 β€ π2 β€ β β β is the sequence of the eigenvalues, repeated according to their multiplicities, of a positive order, self-adjoint, classical, elliptic SG-pseudodifferential operator π on a manifold with ends. Explicitly, SG-pseudodifferential operatorsπ = π(π₯, π·) = Op(π) on Rπcan be defined via the usual left-quantization
Pu(π₯) = 1
(2π)π β« πππ₯β ππ (π₯, π) Μπ’ (π) ππ, π’ β S (Rπ) , (2)
starting from symbols π(π₯, π) β πΆβ(Rπ Γ Rπ) with the property that, for arbitrary multiindices πΌ, π½, there exist constantsπΆπΌπ½β₯ 0 such that the estimates
σ΅¨σ΅¨σ΅¨σ΅¨
σ΅¨π·πΌππ·π₯π½π (π₯, π)σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨ β€ πΆπΌπ½β¨πβ©πβ|πΌ|β¨π₯β©πβ|π½| (3)
hold for fixedπ, π β R and all (π₯, π) β Rπ Γ Rπ, where β¨π¦β©= β1 + |π¦|2,π¦ β Rπ. Symbols of this type belong to the
class denoted byππ,π(Rπ), and the corresponding operators constitute the classπΏπ,π(Rπ) = Op(ππ,π(Rπ)). In the sequel we will sometimes writeππ,π and πΏπ,π, respectively, fixing once and for all the dimension of the (noncompact) base manifold toπ.
These classes of operators, introduced onRπ by Cordes [1] and Parenti [2], see also Melrose [3] and Shubin [4], form a graded algebra, that is,πΏπ,πβπΏπ,πβ πΏπ+π,π+π. The remainder elements are operators with symbols in πββ,ββ(Rπ) =
β(π,π)βR2ππ,π(Rπ) = S(R2π); that is, those having kernel
in S(R2π), continuously mapping SσΈ (Rπ) to S(Rπ). An operatorπ = Op(π) β πΏπ,πand its symbolπ β ππ,πare called SG-elliptic if there existsπ β₯ 0 such that π(π₯, π) is invertible for|π₯| + |π| β₯ π and
π(π₯, π)β1 = π (β¨πβ©βπβ¨π₯β©βπ) . (4) In such case we will usually writeπ β πΈπΏπ,π. Operators in πΏπ,π act continuously from S(Rπ) to itself and extend as continuous operators from SσΈ (Rπ) to itself and from
π»π ,π(Rπ) to π»π βπ,πβπ(Rπ), where π»π ,π(Rπ), π , π β R, denotes
the weighted Sobolev space
π»π ,π(Rπ) = {π’ β SσΈ (Rπ) : βπ’βπ ,π = σ΅©σ΅©σ΅©σ΅©Op (ππ ,π) π’σ΅©σ΅©σ΅©σ΅©πΏ2< β} ,
ππ ,π(π₯, π) = β¨πβ©π β¨π₯β©π.
(5) Continuous inclusionsπ»π ,π(Rπ) σ³¨ β π»π,π(Rπ) hold when π β₯ π and π β₯ π, compact when both inequalities are strict, and S (Rπ) = β (π ,π)βR2 π»π ,π(Rπ) , SσΈ (Rπ) = β (π ,π)βR2 π»π ,π(Rπ) . (6) An elliptic SG-operatorπ β πΏπ,πadmits a parametrixπΈ β
πΏβπ,βπsuch that
ππΈ = πΌ + πΎ1, πΈπ = πΌ + πΎ2, (7) for suitable πΎ1, πΎ2 β πΏββ,ββ = Op(πββ,ββ), and it turns out to be a Fredholm operator. In 1987, Schrohe [5] introduced a class of noncompact manifolds, the so-called SG-manifolds, on which it is possible to transfer from Rπ the whole SG-calculus. In short, these are manifolds which admit a finite atlas whose changes of coordinates behave like symbols of order (0, 1) (see [5] for details and additional technical hypotheses). The manifolds with cylindrical ends are a special case of SG-manifolds, on which also the concept of SG-classical operator makes sense; moreover, the principal symbol of an SG-classical operator π on a manifold with cylindrical endsπ, in this case a triple π(π) = (ππ(π), ππ(π), πππ(π)) = (ππ, ππ, πππ), has an invariant meaning onπ, see Egorov and Schulze [6], Maniccia and Panarese [7], Melrose [3], and Section 2. We indicate the subspaces of classical symbols and operators adding the subscriptclto the notation introduced above.
The literature concerning the study of the eigenvalue asymptotics of elliptic operators is vast and covers a number of different situations and operator classes, see, for example, the monograph by Ivrii [8]. Then, we only mention a few of the many existing papers and books on this deeply investigated subject, which are related to the case we consider here, either by the type of symbols and underlying spaces, or by the techniques which are used. We refer the reader to the corresponding reference lists for more complete infor-mations. On compact manifolds, well-known results were proved by HΒ¨ormander [9] and Guillemin [10], see also the book by Kumano-go [11]. On the other hand, for operators globally defined onRπ, see Boggiatto et al. [12], Helffer [13], HΒ¨ormander [14], Mohammed [15], Nicola [16], and Shubin [4]. Many other situations have been considered, see the cited book by Ivrii. On manifolds with ends, Christiansen and Zworski [17] studied the Laplace-Beltrami operator associ-ated with a scattering metric, while Maniccia and Panarese [7] applied the heat kernel method to study operators similar to those considered here.
Here we deal with the case of manifolds with ends for π β πΈπΏπ,πcl (π), positive and self-adjoint, such that
π, π > 0, π ΜΈ= π, focusing on the (invariant) meaning of the constants appearing in the corresponding Weyl formulae and on achieving a better estimate of the remainder term. Note that the situation we consider here is different from that of the Laplace-Beltrami operator investigated in [17], where continuous spectrum is present as well. In fact, in view ofTheorem 14, spec(π) consists only of a sequence of real
isolated eigenvalues{ππ} with finite multiplicity.
As recalled above, a first result concerning the asymptotic behaviour ofππ(π) for operators including those considered in this paper was proved by Maniccia and Panarese in [7], giving, forπ β +β, ππ(π) = { { { { { { { { { { { { { { { πΆ1ππ/π+ π (ππ/π) forπ < π, πΆ1
0ππ/πlogπ + π (ππ/πlogπ) for π = π,
πΆ2ππ/π+ π (ππ/π) forπ > π. (8)
Note that the constantsπΆ1, πΆ2, πΆ10 above depend only on the principal symbol ofπ, which implies that they have an invariant meaning on the manifoldπ, see Sections 2 and
3. On the other hand, in view of the technique used there, the remainder terms appeared in the form π(ππ/ min{π,π}) and π(ππ/πlogπ) for π ΜΈ= π and π = π, respectively. An improvement in this direction for operators onRπhad been achieved by Nicola [16], who, in the caseπ = π, proved that
ππ(π) = πΆ10ππ/πlogπ + π (ππ/π) , π σ³¨β +β, (9) while, forπ ΜΈ= π, showed that the remainder term has the form
π(π(π/ min{π,π})βπ) for a suitable π > 0. A further improvement
of these results in the caseπ = π has recently appeared in Battisti and Coriasco [18], where it has been shown that, for a suitableπ > 0,
ππ(π) = πΆ10ππ/πlogπ + πΆ20ππ/π+ π (π(π/π)βπ) ,
π σ³¨β +β. (10) Even the constant πΆ20 has an invariant meaning onπ, and bothπΆ10andπΆ20are explicitly computed in terms of trace operators defined onπΏπ,πcl (π).
In this paper the remainder estimates in the caseπ ΜΈ= π are further improved. More precisely, we first consider the power π = π1/ max{π,π}ofπ (see Maniccia et al. [19] for the properties
of powers of SG-classical operators). Then, by studying the asymptotic behaviour inπ of the trace of the operator Μππ(βπ), ππ(π‘) = π(π‘)πβππ‘π,π β πΆβ
0 (R), defined via a Spectral Theorem
and approximated in terms of Fourier Integral Operators, we prove the following.
Theorem 1. Let π be a manifold with ends of dimension π and
π, π > 0, π ΜΈ= π, with domain π»π,π(π) σ³¨ β πΏ2(π). Then, the
following Weyl formulae hold forπ β +β:
ππ(π) = { { { { { { { { { { { { { { { { { { { πΆ1ππ/π+ π (ππ/π) + π (π(π/π)β(1/π)) = πΆ1ππ/π+ π (π(π/π)βπ1) πππ π < π, πΆ2ππ/π+ π (ππ/π) + π (π(π/π)β(1/π)) = πΆ2ππ/π+ π (π(π/π)βπ2) πππ π > π, (11)
whereπ1 = min{1/π, π((1/π) β (1/π))} and π2 = min{1/
π, π((1/π) β (1/π))}.
The order of the remainder is then determined by the ratio ofπ and π and the dimension of π, since
π πβ 1 π β€ π π, for π < π ββ 1 < π πβ€ 1 + 1 π, π πβ 1 π β€ π π, for π > π ββ 1 < π π β€ 1 + 1 π. (12)
In particular, when max{π, π}/ min{π, π} β₯ 2, the remainder is alwaysπ(ππ/ max{π,π}).
Examples include operators of SchrΒ¨odinger type onπ, that is,π = βΞπ+ π, Ξπthe Laplace-Beltrami operator in π associated with a suitable metric π, π a smooth potential that, in the local coordinatesπ₯ β ππβ Rπon the cylindrical end growths asβ¨π₯β©π, with an appropriate π > 0 related to π. Such examples will be discussed in detail, together with the sharpness of the results inTheorem 1, in the forthcoming paper [20], see also [21].
The key point in the proof ofTheorem 1is the study of the asymptotic behaviour forπ β +β of integrals of the form
πΌ (π) = β« ππ(βπ‘π+π(π‘;π₯,π)βπ₯π)π (π‘) π (π‘; π₯, π) ππ‘Lπππ₯, (13) whereπ and π satisfy certain growth conditions in π₯ and π (seeSection 3for more details). The integralsπΌ(π) represent in fact the local expressions of the trace of Μππ(βπ), obtained through the so-called βgeometric optic method,β specialised to the SG situation, see, for example, Coriasco [22, 23], Coriasco and Rodino [24]. To treat the integrals πΌ(π) we proceed similarly to Grigis and SjΒ¨ostrand [25], Helffer and Robert [26], see also Tamura [27].
The paper is organised as follows.Section 2is devoted to recall the definition of SG-classical operators on a manifold with ends π. In Section 3 we show that the asymptotic behaviour ofππ(π), π β +β, for a positive self-adjoint operatorπ β πΏπ,πcl (π), π, π > 0, is related to the asymp-totic behaviour of oscillatory integrals of the formπΌ(π). In
Section 4we conclude the proof ofTheorem 1, investigating the behaviour ofπΌ(π) for π β +β. Finally, some technical details are collected in the Appendix.
2. SG-Classical Operators on
Manifolds with Ends
From now on, we will be concerned with the subclass of SG-operators given by those elementsπ β πΏπ,π(Rπ), (π, π) β R2, which are SG-classical, that is, π = Op(π) with π β
ππ,πcl (Rπ) β ππ,π(Rπ). We begin recalling the basic definitions
and results (see, e.g., [6,19] for additional details and proofs).
Definition 2. (i) A symbol π(π₯, π) belongs to the class
ππ,πcl(π)(Rπ) if there exist π
πβπ,β (π₯, π) β ΜHπβππ (Rπ), π = 0, 1, . . .,
positively homogeneous functions of orderπ β π with respect to the variableπ, smooth with respect to the variable π₯, such that, for a0-excision function π,
π (π₯, π) βπβ1β
π=0π (π) ππβπ,β (π₯, π) β π
πβπ,π(Rπ) ,
π = 1, 2, . . . . (14)
(ii) A symbolπ(π₯, π) belongs to the class ππ,πcl(π₯)(Rπ) if there exist πβ ,πβπ(π₯, π) β ΜHπβππ₯ (Rπ), π = 0, . . ., positively homogeneous functions of orderπ β π with respect to the variableπ₯, smooth with respect to the variable π, such that, for a0-excision function π,
π (π₯, π) βπβ1β
π=0π (π₯) πβ ,πβπ
(π₯, π) β ππ,πβπ(Rπ) , π = 1, 2, . . . .
(15)
Definition 3. A symbolπ(π₯, π) is SG-classical, and we write
π β ππ,πcl(π₯,π)(Rπ) = ππ,πcl (Rπ) = ππ,πcl , if
(i) there existππβπ,β (π₯, π) β ΜHπβππ (Rπ) such that for a 0-excision functionπ, π(π)ππβπ,β (π₯, π) β πcl(π₯)πβπ,π(Rπ) and
π (π₯, π) βπβ1β
π=0π (π) ππβπ,β
(π₯, π) β ππβπ,π(Rπ) , π = 1, 2, . . . ;
(16)
(ii) there existπβ ,πβπ(π₯, π) β ΜHπβππ₯ (Rπ) such that for a 0-excision functionπ, π(π₯)πβ ,πβπ(π₯, π) β πcl(π)π,πβπ(Rπ) and
π (π₯, π) βπβ1β
π=0
π (π₯) πβ ,πβπβ ππ,πβπ(Rπ) , π = 1, 2, . . . .
(17) We setπΏπ,πcl(π₯,π)(Rπ) = πΏπ,πcl = Op(ππ,πcl ).
Remark 4. The definition could be extended in a natural way
from operators acting between scalars to operators acting between (distributional sections of) vector bundles. One should then use matrix-valued symbols whose entries satisfy the estimates (3).
Note that the definition of SG-classical symbol implies a condition of compatibility for the terms of the expansions
with respect toπ₯ and π. In fact, defining ππβππ andππβππ onππ,πcl(π)
andππ,πcl(π₯), respectively, as
πππβπ(π) (π₯, π) = ππβπ,β (π₯, π) , π = 0, 1, . . . ,
πππβπ(π) (π₯, π) = πβ ,πβπ(π₯, π) , π = 0, 1, . . . .
(18) It is possibile to prove that
ππβπ,πβπ= ππππβπ,πβπ(π) = πππβπ(πππβπ(π))
= πππβπ(ππβππ (π)) , π = 0, 1, . . . , π = 0, 1, . . . . (19) Moreover, the composition of two SG-classical operators is still classical. Forπ = Op(π) β πΏπ,πcl the triple π(π) = (ππ(π), ππ(π), πππ(π)) = (ππ,β , πβ ,π, ππ,π) = (ππ, ππ, πππ) is called the principal symbol of π. The three components are also called theπ-, π- and ππ-principal symbol, respectively. This definition keeps the usual multiplicative behaviour, that is, for anyπ β πΏπ,πcl ,π β πΏπ ,πcl,(π, π), (π , π) β R2, π(π π) = π(π)π(π), with component-wise product in the right-hand side. We also set
Symπ(π) (π₯, π) = Symπ(π) (π₯, π) = πm(π₯, π) = π (π) ππ(π₯, π)
+π (π₯) (ππ(π₯, π) β π (π) πππ(π₯, π)) ,
(20)
for a fixed0-excision function π.Theorem 5allows to express the ellipticity of SG-classical operators in terms of their principal symbol.
Theorem 5. An operator π β πΏπ,πππ is elliptic if and only if
each element of the tripleπ(π) is nonvanishing on its domain
of definition.
As a consequence, denoting by {ππ} the sequence of eigenvalues ofπ, ordered such that π β€ π β ππ β€ ππ, with each eigenvalue repeated accordingly to its multiplicity, the counting functionππ(π) = βππβ€π1 is well defined for a SG-classical elliptic self-adjoint operatorπ see, for example, [16, 18,20,21]. We now introduce the class of noncompact manifolds with which we will deal.
Definition 6. A manifold with a cylindrical end is a triple
(π, π, [π]), where π = MβπΆC is a π-dimensional smooth manifold and
(i)M is a smooth manifold, given by M = (π0\ π·) βͺ πΆ with a π-dimensional smooth compact manifold without boundary π0, π· a closed disc of π0, and πΆ β π· a collar neighbourhood of ππ· in π0;
(ii)C is a smooth manifold with boundary πC = π, with π diffeomorphic to ππ·;
(iii)π : [πΏπ, β)ΓSπβ1 β C, πΏπ> 0, is a diffeomorphism, π({πΏπ} Γ Sπβ1) = π and π({[πΏ
π, πΏπ+ ππ)} Γ Sπβ1),
ππ> 0, is diffeomorphic to πΆ;
(iv) the symbolβπΆ means that we are gluingM and C, through the identification ofπΆ and π({[πΏπ, πΏπ+ππ)}Γ Sπβ1);
(v) the symbol[π] represents an equivalence class in the set of functions
{π : [πΏπ, β) Γ Sπβ1σ³¨β C : π is a diffeomorphism, π ({πΏπ} Γ Sπβ1) = π and π ([πΏπ, πΏπ+ ππ) Γ Sπβ1) ,
ππ> 0, is diffeomorphic to πΆ} ,
(21)
whereπ βΌ π if and only if there exists a diffeomorphism Ξ β Diff(Sπβ1) such that
(πβ1β π) (π, πΎ) = (π, Ξ (πΎ)) , (22) for allπ β₯ max{πΏπ, πΏπ} and πΎ β Sπβ1.
We use the following notation: (i)ππΏπ= {π₯ β Rπ: |π₯| > πΏπ};
(ii)Cπ= π([π, β)ΓSπβ1), where π β₯ πΏπ. The equivalence condition (22) implies thatCπis well defined; (iii)π : Rπ \ {0} β (0, β) Γ Sπβ1 : π₯ σ³¨β π(π₯) =
(|π₯|, π₯/|π₯|);
(iv)ππ = πβπ : ππΏπ β C is a parametrisation of the end. Let us notice that, settingπΉ = πβ1π βππ, the equivalence condition (22) implies
πΉ (π₯) = |π₯| Ξ (|π₯|π₯ ) . (23) We also denote the restriction of ππ mappingππΏπ onto ΜC = C \ π by Μππ.
The couple ( ΜC, Μππβ1) is called the exit chart. If A = {(Ξ©π, ππ)}ππ=1is such that the subset{(Ξ©π, ππ)}πβ1π=1 is a finite atlas forM and (Ξ©π, ππ) = ( ΜC, Μππβ1), then π, with the atlas A, is a SG-manifold (see [4]). An atlasA of such kind is called
admissible. From now on, we restrict the choice of atlases on
π to the class of admissible ones. We introduce the following spaces, endowed with their natural topologies,
S (ππΏ) = {π’ β πΆβ(π πΏ) : βπΌ, π½ β Nπ βπΏσΈ > πΏ sup π₯βππΏσΈ σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨π₯ πΌππ½π’ (π₯)σ΅¨σ΅¨σ΅¨σ΅¨ σ΅¨ < β} , S0(ππΏ) = β πΏσΈ βπΏ {π’ β S (Rπ) : supp π’ β ππΏσΈ } , S (π) = {π’ β πΆβ(π) : π’ β Μππ β S (ππΏπ) for any exit mapππ} , SσΈ (π) denotes the dual space of S (π) .
Definition 7. The set ππ,π(ππΏπ) consists of all the symbols
π β πΆβ(π
πΏπ) which fulfill (3) for(π₯, π) β ππΏπ Γ R
π only.
Moreover, the symbolπ belongs to the subset SGπ,πcl (ππΏπ) if it admits expansions in asymptotic sums of homogeneous symbols with respect toπ₯ and π as in Definitions2and 3, where the remainders are now given by SG-symbols of the required order onππΏπ.
Note that, sinceππΏπis conical, the definition of homoge-neous and classical symbol onππΏπmakes sense. Moreover, the elements of the asymptotic expansions of the classical sym-bols can be extended by homogeneity to smooth functions onRπ\ {0}, which will be denoted by the same symbols. It is a fact that, given an admissible atlas{(Ξ©π, ππ)}ππ=1onπ, there exists a partition of unity{ππ} and a set of smooth functions {ππ} which are compatible with the SG-structure of π, that is,
(i) suppππβ Ξ©π, suppππβ Ξ©π,ππππ= ππ,π = 1, . . . , π; (ii)|ππΌ(ππβ Μππ)(π₯)| β€ πΆπΌβ¨π₯β©β|πΌ|and|ππΌ(ππβ Μππ)(π₯)| β€
πΆπΌβ¨π₯β©β|πΌ|for allπ₯ β π
πΏπ.
Moreover,ππandππcan be chosen so thatππβ Μππand ππβ Μππare homogeneous of degree0 on ππΏ. We denote byπ’β the composition ofπ’ : ππ(Ξ©π) β Rπ β C with the coordinate patchesππ, and byVβthe composition ofV : Ξ©π β π β C withπβ1π ,π = 1, . . . , π. It is now possible to give the definition of SG-pseudodifferential operator onπ.
Definition 8. Letπ be a manifold with a cylindrical end.
A linear operator π : S(π) β SσΈ (π) is an SG-pseudodifferential operator of order(π, π) on π, and we writeπ β πΏπ,π(π), if, for any admissible atlas {(Ξ©π, ππ)}ππ=1 onπ with exit chart (Ξ©π, ππ):
(1) for allπ = 1, . . . , π β 1 and any ππ, ππβ πΆβπ (Ξ©π), there exist symbolsππ(π₯, π) β ππ(ππ(Ξ©π)) such that (ππππππ’β)β(π₯) = β¬ ππ(π₯βπ¦)β πππ(π₯, π) π’ (π¦) ππ¦ππ₯, π’ β πΆβ(ππ(Ξ©π)) ;
(25)
(2) for anyππ, ππof the type described above, there exists a symbolππ(π₯, π) β SGπ,π(ππΏπ) such that
(ππππππ’β)β(π₯) = β¬ ππ(π₯βπ¦)β πππ(π₯, π) π’ (π¦) ππ¦ππ₯, π’ β S0(ππΏπ) ;
(26)
(3)πΎπ, the Schwartz kernel ofπ, is such that
πΎπβ πΆβ((π Γ π) \ Ξ) β S (( ΜC Γ ΜC) \ π) , (27) whereΞ is the diagonal of π Γ π and π = ( ΜππΓ
Μ
ππ)(π) with any conical neighbourhood π of the
diagonal ofππΏπΓ ππΏπ.
The most important local symbol of π is ππ. Our definition of SG-classical operator onπ differs slightly from the one in [7].
Definition 9. Letπ β πΏπ,π(π). π is an SG-classical operator
onπ, and we write π β πΏπ,πcl (π), if ππ(π₯, π) β ππ,πcl (ππΏπ) and the operatorπ, restricted to the manifold M, is classical in the usual sense.
The usual homogeneous principal symbolππof an SG-classical operatorπ β πΏπ,πcl (π) is of course well defined as a smooth function onπβπ. In order to give an invariant definition of the principal symbols homogeneous inπ₯ of an operator π β πΏπ,πcl (π), the subbundle ππβπ = {(π₯, π) β πβπ : π₯ β π, π β ππ₯βπ} was introduced. The notions of ellipticity can be extended to operators onπ as well.
Definition 10. Letπ β πΏπ,πcl (π) and let us fix an exit map ππ.
We can define local objectsππβπ,πβπ, πβ ,πβπas
ππβπ,πβπ(π, π) = πππβπ,πβπ(π, π) , π β Sπβ1, π β Rπ\ {0} ,
πβ ,πβπ(π, π) = ππβ ,πβπ(π, π) , π β Sπβ1, π β Rπ.
(28)
Definition 11. An operatorπ β πΏπ,πcl (π) is elliptic, and we
writeπ β πΈπΏπ,πcl (π), if the principal part of ππ β ππ,π(ππΏπ) satisfies the SG-ellipticity conditions on ππΏπ Γ Rπ and the operatorπ, restricted to the manifold M, is elliptic in the usual sense.
Proposition 12. The properties π β πΏπ,π(π) and π β
πΏπ,πππ (π), as well as the notion of SG-ellipticity, do not depend
on the (admissible) atlas onπ. Moreover, the local functions ππ
andπππgive rise to invariantly defined elements ofπΆβ(ππβπ)
andπΆβ(ππβπ \ 0), respectively.
Then, with any π β πΏπ,πcl (π), it is associated an invariantly defined principal symbol in three components π(π) = (ππ, ππ, πππ). Finally, through local symbols given by
ππ ,ππ (π₯, π) = β¨πβ©π ,π = 1, . . . , π β 1, and πππ ,π(π₯, π) = β¨πβ©π β¨π₯β©π, π , π β R, we get a SG-elliptic operator Ξ π ,π β πΏπ ,π
cl(π) and
introduce the (invariantly defined) weighted Sobolev spaces π»π ,π(π) as
π»π ,π(π) = {π’ β SσΈ (π) : β
π ,ππ’ β πΏ
2(π)} . (29)
The properties of the spacesπ»π ,π(Rπ) extend to π»π ,π(π) without any change, as well as the continuity of the linear mappingsπ : π»π ,π(π) β π»π βπ,πβπ(π) induced by π β πΏπ,π(π), mentioned inSection 1.
3. Spectral Asymptotics for
SG-Classical Elliptic Self-Adjoint
Operators on Manifolds with Ends
In this section we illustrate the procedure to proveTheorem 1, similar to [13,25,27]. The result will follow from the Trace formula (39), (41), the asymptotic behaviour (42), and the Tauberian Theorem 19. The remaining technical points, in particular the proof of the asymptotic behaviour of the integrals appearing in (41), are described inSection 4and in the Appendix.
Let the operator π β πΈπΏπ,πcl (π) be considered as an unbounded operatorπ : S(π) β π»0,0(π) = πΏ2(π) β πΏ2(π). The following proposition can be proved by reducing to the local situation and using continuity and ellipticity of π, its parametrix, and the density of S(π) in the π»π ,π(π)
spaces.
Proposition 13. Every π β πΈπΏπ,πππ (π), considered as an
unbounded operatorπ : S(π) β πΏ2(π) β πΏ2(π), admits
a unique closed extension, still denoted byπ, whose domain is
D(π) = π»π,π(π).
From now on, when we writeπ β πΈπΏπ,πcl (π) we always mean its unique closed extension, defined inProposition 13. As standard, we denote byσ°(π) the resolvent set of π, that is, the set of allπ β C such that ππΌβπ maps π»π,π(π) bijectively ontoπΏ2(π). The spectrum of π is then spec(π) = C \ σ°(π). The next theorem was proved in [7].
Theorem 14 (Spectral theorem). Let π β πΈπΏπ,πππ (π) be
regarded as a closed unbounded operator onπΏ2(π) with dense
domainπ»π,π(π). Assume also that π, π > 0 and πβ = π.
Then
(i)(ππΌ β π)β1 is a compact operator onπΏ2(π) for every
π β σ°(π). More precisely, (ππΌ β π)β1is an extension by
continuity fromS(π) or a restriction from SσΈ (π) of
an operator inπΈπΏβπ,βπππ (π).
(ii) spec(π) consists of a sequence of real isolated
eigenval-ues{ππ} with finite multiplicity, clustering at infinity;
the orthonormal system of eigenfunctions {ππ}πβ₯1 is
complete inπΏ2(π) = π»0,0(π). Moreover, ππ β S(π)
for allπ.
Given a positive self-adjoint operator π β πΈπΏπ,πcl (π), π, π > 0, π ΜΈ= π, we can assume, without loss of generality (considering, if necessary,π + π in place of π, with π β R a suitably large constant),1 β€ π1 β€ π2. . .. Define the counting functionππ(π), π β R, as
ππ(π) = β
ππβ€π
1 = # (spec (π) β© (ββ, π]) . (30) Clearly,ππis nondecreasing, continuous from the right and supported in[0, +β). If we set π = π1/π,π = max{π, π} (see [19] for the definition of the powers of π), π turns out to be a SG-classical elliptic self-adjoint operator with
π(π) = (π1/π
π , π1/ππ , π1/πππ). We denote by {ππ} the sequence of
eigenvalues ofπ, which satisfy ππ = π1/ππ . We can then, as above, considerππ(π). It is a fact that ππ(π) = π(ππ/π), see [7].
From now on we focus on the case π > π > 0. The caseπ > π > 0 can be treated in a completely similar way, exchanging the role ofπ₯ and π. So we can start from a closed positive self-adjoint operatorπ β πΈπΏπ,1cl (π) with domain D(π) = π»π,1(π), π β (0, 1). For π’ β π»π,1(π), π‘ β R, we set π (π‘) π’ =ββ π=1π ππ‘ππ (π’, π π)πΏ2(π)ππ, (31)
and the series converges in theπΏ2(π) norm (cf., e.g., [25]). Clearly, for allπ‘ β R, π(π‘) is a unitary operator such that
π (0) = πΌ, π (π‘ + π ) = π (π‘) π (π ) , π‘, π β R. (32) Moreover, if π’ β π»ππ,π(π) for some π β N, π(π‘)π’ β πΆπ(R, π»0,0(π)) β© β β β β© πΆ0(R, π»ππ,π(π)) and, for π’ β
π»π,1(π), we have π·
π‘π(π‘)π’ β ππ(π‘)π’ = 0, π(0)π’ = π’, which
implies thatV(π‘, π₯) = π(π‘)π’(π₯) is a solution of the Cauchy problem
(π·π‘β π) V = 0, V|π‘=0= π’. (33)
Let us fixπ β S(R). We can then define the operator Μπ(βπ) either by using the formula
Μπ (βπ) π’ =ββ
π=1Μπ (βππ
) (π’, ππ)πΏ2(π)ππ, (34)
or by means of the vector-valued integral(β« π(π‘)π(π‘)ππ‘)π’ = β« π(π‘)π(π‘)π’ ππ‘, π’ β π»0,0(π). Indeed, there exists π0 β N such that ββπ=1πβπ0
π < β, so the definition makes sense
and gives an operator inL(πΏ2(π)) with norm bounded by βπβπΏ1(R). The following lemma, whose proof can be found in
the Appendix, is an analog onπ of Proposition 1.10.11 in [13].
Lemma 15. Μπ(βπ) is an operator with kernel πΎπ(π₯, π¦) = βπ Μπ(βππ)ππ(π₯)ππ(π¦) β S(π Γ π).
Clearly, we then have β«
ππΎπ(π₯, π₯) ππ₯ = βπ Μπ (βππ) . (35)
By the analysis in [22β24,28] (see also [29]), the above Cauchy problem (33) can solve moduloS(π) by means of a smooth family of operatorsπ(π‘), defined for π‘ β (βπ, π), π > 0 suitably small, in the sense that (π·π‘βπ)βπ is a family of smoothing operators andπ(0) is the identity on SσΈ (π). More explicitly, the following theorem holds (see the Appendix for some details concerning the extension to the manifoldπ of the results onRπproved in [22β24,28]).
Theorem 16. Define π(π‘)π’ = βππ=1πππ΄π(π‘)(πππ’), where ππ
andππare as inDefinition 8, withππππ = ππ,π = 1, . . . , π,
while theπ΄π(π‘) are SG FIOs which, in the local coordinate open
setππ= ππ(Ξ©π) and with V β S(Rπ), are given by
(π΄π(π‘) V) (π₯) = β« ππππ(π‘;π₯,π)π
π(π‘; π₯, π) ΜV (π) Lπ. (36) Eachπ΄π(π‘) solves a local Cauchy problem (π·π‘β ππ) β π΄πβ
πΆβ((βπ, π), πΏββ,ββ(Rπ)), π΄
π(0) = πΌ, with ππ= ππ(ππ) and
{ππ} β ππΊπ,1ππ (Rπ) local (complete) symbol of π associated with {ππ}, {ππ}, with phase and amplitude functions such that
ππ‘ππ(π‘; π₯, π) β ππ(π₯, ππ₯ππ(π‘; π₯, π)) = 0, ππ(0; π₯, π) = π₯π, ππβ πΆβ((βπ, π) , ππΊ0,0ππ (Rπ)) , ππ(0; π₯, π) = 1. (37) Then,π(π‘) satisfies (π·π‘β π) β π β πΆβ((βπ, π) , πΏββ,ββ(π)) , π (0) = πΌ, (38) andπ β π β πΆβ((βπ, π), πΏββ,ββ(π)).
Remark 17. Trivially, forπ = 1, . . . , π β 1, ππ andππ can be
considered SG-classical, since, in those cases, they actually have order ββ with respect to π₯, by the fact that ππ(π₯, π) vanishes forπ₯ outside a compact set.
Remark 18. Notation likeπ β πΆβ((βπ, π), ππ,π(Rπ)), π΅ β
πΆβ((βπ, π), πΏπ,π(π)), and similar, in Theorem 16 and in
the sequel, also mean that the seminorms of the involved elements in the corresponding spaces (induced, in the men-tioned cases, by (3)), are uniformly bounded with respect to π‘ β (βπ, π).
If we writeππ(π‘) = π(π‘)πβππ‘πin place ofπ(π‘), for a chosen π β πΆβ
0 ((βπ, π)), the trace formula (35) becomes
β«
ππΎππ(π₯, π₯) ππ₯ = β Μπ (π β ππ) . (39)
Let us denote the kernel of π β π by π(π‘; π₯, π¦) β πΆβ((βπ, π), S(π Γ π)). Then, the distribution kernel of
β« πβππ‘ππ(π‘) π(π‘)ππ‘ = Μπ π(βπ) is πΎππ(π₯, π¦) =βπ π=1 ππ(π₯) β¬ π (π‘) ππ(βπ‘π+ππ(π‘;π₯,π)βπ¦π) Γ ππ(π‘; π₯, π) ππ‘Lπππ(π¦) + β« πβππ‘ππ (π‘) π (π‘; π₯, π¦) ππ‘, (40)
where the local coordinates in the right-hand side depend on π and, to simplify the notation, we have omitted the
corresponding coordinate maps. By the choices ofπ, ππand ππwe obtain β π Μπ (π β ππ ) =βπ π=1β π (π‘) π π(βπ‘π+ππ(π‘;π₯,π)βπ₯π) Γ ππ(π‘; π₯, π) ππ(π₯) ππ‘Lπππ₯ + β¬ πβππ‘ππ (π‘) π (π‘; π₯, π₯) ππ‘ππ₯ =βπ π=1 β π (π‘) ππ(βπ‘π+ππ(π‘;π₯,π)βπ₯π) Γ ππ(π‘; π₯, π) ππ(π₯) ππ‘Lπππ₯ + π (|π|ββ) . (41)
Letπ β πΆβ0 ((βπ, π)), π > 0, be such that π(0) = 1 and Μπ β₯ 0, Μπ(0) > 0 (e.g., set π = π β Μπ with a suitable π β πΆβ
0 ((βπ, π))). By the analysis of the asymptotic behaviour
of the integrals appearing in (41), described inSection 4, we finally obtain β π Μπ (π β ππ ) = { { { { { { { { { { { { { π ππ0π(π/π)β1+ π (ππ ββ1 ) forπ σ³¨β +β, π (|π|ββ) for π σ³¨β ββ, (42)
with πβ = min{π, (π/π) β 1}. The following Tauberian theorem is a slight modification of Theorem 4.2.5 of [13] (see the Appendix).
Theorem 19. Assume that
(i)π β πΆβ0 (R) is an even function satisfying π(0) = 1, Μπ β₯ 0, Μπ(0) > 0;
(ii)ππ(π) is a nondecreasing function, supported in [0, +β), continuous from the right, with polynomial
growth at infinity and isolated discontinuity points of
first kind{ππ}, π β N, such that ππ β +β;
(iii) there existsπ0β₯ 0 such that β π Μπ (π β ππ ) = β« Μπ (π β π) πππ(π) = { { { { { { { { { { { { { { { π ππ0 π(π/π)β1+ π (ππ ββ1 ) πππ π σ³¨β +β, π (|π|ββ) πππ π σ³¨β ββ, (43) withπ β (0, 1), πβ= min{π, (π/π) β 1}. Then ππ(π) = π0 2πππ/π+ π (ππ β ) , πππ π σ³¨β +β. (44)
Remark 20. The previous statement can be modified as
follows: withπ, ππ, andπ as inTheorem 19, when β« Μπ (π β π) πππ(π) = { { { { { { { { { { { { { π ππ0 π(π/π)β1+ π (π(π/π)β2) + π (ππβ1) for π σ³¨β +β, π (|π|ββ) for π σ³¨β ββ, (45) withπ β (0, 1), then ππ(π) = (π0/2π)ππ/π+ π(π(π/π)β1) + π(ππ), for π β +β.
4. Proof of
Theorem 1
In view ofTheorem 19andRemark 20, to complete the proof ofTheorem 1we need to show that (42) holds. To this aim, as explained previously, this section will be devoted to studying the asymptotic behaviour for|π| β +β of
πΌ (π) = β« ππΞ¦(π‘;π₯,π;π)π (π‘) π (π‘; π₯, π) ππ‘Lπππ₯, (46) where π β πΆ0β((βπ, π)), π(0) = 1, π β πΆβ((βπ, π), π0,0(Rπ)), π(0; π₯, π) = 1, and Ξ¦ (π‘; π₯, π; π) = π (π‘; π₯, π) β π₯π β π‘π, π β πΆβ((βπ, π) , π1,1cl (Rπ)) , (47) such that (i)ππ‘π(π‘; π₯, π) = π(π₯, ππ₯π(π‘; π₯, π)), π(0; π₯, π) = π₯π; (ii)πΆβ1β¨πβ©β€ β¨ππ₯π(π‘; π₯, π)β©β€ πΆβ¨πβ©, for a suitable
constantπΆ > 1;
(iii)π β ππ,1cl (Rπ), 0 < π < 1, SG-elliptic.
Sinceπβ1(π₯, π) β π(β¨π₯β©β1β¨πβ©βπ) for |π₯| + |π| β₯ π > 0, it is not restrictive to assume that this estimate holds on the whole phase space, so that, for a certain constantπ΄ > 1,
π΄β1β¨π₯β© β¨πβ©πβ€ π (π₯, π) β€ π΄ β¨π₯β© β¨πβ©π. (48)
Remark 21. The assumption onπβ1above amounts, at most,
to modifying π by adding and subtracting a compactly supported symbol, that is, an element ofπββ,ββ(Rπ). The corresponding solutionsπ and π of the eikonal and transport equations, respectively, would then change, at most, by an element ofπΆβ((βπ, π), πββ,ββ(Rπ)), see [23, 24, 28]. It is immediate, by integration by parts with respect to t, that an integral as (46) isπ(|π|ββ) for π β πΆβ((βπ, π), πββ,ββ(Rπ)). Then, the modifiedπ obviously keeps the same sign every-where.
For two functionsπ, π, defined on a common subset π ofRπ1and depending on parametersπ¦ β π β Rπ2, we will
writeπ βΊ π or π(π₯, π¦) βΊ π(π₯, π¦) to mean that there exists a suitable constantπ > 0 such that |π(π₯, π¦)| β€ π|π(π₯, π¦)| for all(π₯, π¦) β π Γ π. The notation π βΌ π or π(π₯, π¦) βΌ π(π₯, π¦) means that bothπ βΊ π and π βΊ π hold.
Remark 22. The ellipticity ofπ yields, for π < 0,
ππ‘Ξ¦ (π‘; π₯, π; π) = π (π₯, ππ₯π (π‘; π₯, π)) β π β» β¨π₯β© β¨πβ©π+ |π|
(49) which, by integration by parts, impliesπΌ(π) = π(|π|ββ) when π β ββ.
From now on any asymptotic estimate is to be meant for π β +β.
We will make use of a partition of unity on the phase space. The supports of its elements will depend on suitably large positive constantsπ1, π2 > 1. We also assume, as it is possible,π β₯ π0, again with an appropriateπ0 β« 1. As we will see below, the values ofπ1,π2, andπ0depend only onπ and its associated seminorms.
Proposition 23. Let π»1be any function inπΆβ0 (R) such that
suppπ»1β [(2π1)β1, 2π1], 0 β€ π»1β€ 1 and π»1β‘ 1 on [πβ11 , π1],
whereπ1> 1 is a suitably chosen, large positive constant. Then
πΌ (π) = π (πββ) + β« ππΞ¦(π‘;π₯,π;π)π (π‘) π»1(β¨π₯β© β¨πβ© π π ) Γ π (π‘; π₯, π) ππ‘Lπππ₯. (50) Proof. Write πΌ (π) = β« ππΞ¦(π‘;π₯,π;π)π (π‘) [1 β π»1(β¨π₯β© β¨πβ© π π )] Γ π (π‘; π₯, π) ππ‘Lπππ₯ + β« ππΞ¦(π‘;π₯,π;π)π (π‘) π»1(β¨π₯β© β¨πβ© π π ) Γ π (π‘; π₯, π) ππ‘Lπππ₯, (51)
and observe that, by π΄β1β¨π₯β©β¨πβ©π β€ π(π₯, π) β€ π΄β¨π₯β©β¨πβ©π, π₯, π β Rπ, we find σ΅¨σ΅¨σ΅¨σ΅¨ππ‘Ξ¦ (π‘; π₯, π; π)σ΅¨σ΅¨σ΅¨σ΅¨ β₯ π2 + (π21 β π΄πΆ) β¨π₯β© β¨πβ©π when β¨π₯β© β¨πβ© π π β€ πβ11 , (52) σ΅¨σ΅¨σ΅¨σ΅¨ππ‘Ξ¦ (π‘; π₯, π; π)σ΅¨σ΅¨σ΅¨σ΅¨ β₯ (π΄πΆ) β1 2 β¨π₯β©β¨πβ© π+ [(π΄πΆ)β1 2 π1β 1] π when β¨π₯β© β¨πβ© π π β₯ π1. (53) Thus, ifπ1> 2π΄πΆ we have |ππ‘Ξ¦(π‘; π₯, π; π)| βΌ π + β¨π₯β©β¨πβ©π on the support of1βπ»1(β¨π₯β©β¨πβ©π/π), and the assertion follows integrating by parts with respect toπ‘ in the first integral of (51).
Remark 24. We actually chooseπ1> 4π΄πΆ > 2π΄πΆ, since this
will be needed in the proof ofProposition 28; see also Section C in the Appendix.
Let us now pickπ»2 β πΆβ0 (R) such that 0 β€ π»2(π) β€ 1, π»2(π) = 1 for |π| β€ π2andπ»2(π) = 0 for |π| β₯ 2π2, where π2 > 1 is a constant which we will choose big enough (see below). We can then write
(π) = π (πββ) + β« ππΞ¦(π‘;π₯,π;π)π (π‘) π» 1(β¨π₯β© β¨πβ© π π ) Γ π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨)π(π‘;π₯,π)ππ‘Lπππ₯ + β« ππΞ¦(π‘;π₯,π;π)π (π‘) π»1(β¨π₯β© β¨πβ© π π ) Γ [1 β π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨)]π(π‘;π₯,π)ππ‘Lπππ₯ = π (πββ) + πΌ1(π) + πΌ2(π) . (54)
In what follows, we will systematically use the notation ππ,π = ππ,π(π¦, π), π¦ β Rπ, π β Rπ, to generally denote functions depending smoothly on π¦ and π and satisfying SG-type estimates of orderπ, π in π¦, π. In a similar fashion, ππ,ππ = πΆβ((βπ, π), ππ,π(π¦, π)) will stand for some function of the same kind which, additionally, depends smoothly on π‘ β (βπ, π), and, for all π β Z+,π·π‘π πΆβ((βπ, π), ππ,π(π¦, π)) satisfies SG-type estimates of orderπ, π in π¦, π, uniformly with respect toπ‘ β (βπ, π).
To estimate πΌ1(π), we will apply the stationary phase theorem. We begin by rewriting the integralπΌ1(π), using the fact thatπ is solution of the eikonal equation associated with π and that π is a classical SG-symbol. Note that then ππ‘2π β πΆβ((βπ, π), π2πβ1,1cl (Rπ)) β πΆβ((βπ, π), ππ,1cl (Rπ)), since ππ‘2π (π‘; π₯, π) =βπ π=1 (ππππ) (π₯, ππ₯π (π‘; π₯, π)) Γ ππ₯π(π (π₯, ππ₯π (π‘; π₯, π))) . (55)
In view of the Taylor expansion ofπ at π‘ = 0, recalling the propertyπ(π₯, π) = π(π₯)ππ(π₯, π) + ππ,0(π₯, π), π a fixed 0-excision function, we have, for some0 < πΏ1< 1,
Ξ¦ (π‘; π₯, π; π) = βππ‘ β π₯π + π (0; π₯, π) + π‘ππ‘π (0; π₯, π) +π‘2 2 ππ‘2π (π‘πΏ1; π₯, π) = βππ‘ + π‘π (π₯, π) + π‘2π2πβ1,1π (π₯, π) = βππ‘ + π‘π (π₯) ππ(π₯, π) + π‘ππ,0(π₯, π) + π‘2π2πβ1,1π (π₯, π) = βππ‘ + π‘π (π₯) ππ(π₯, π) + π‘ππ,0(π₯, π) + π‘2π (π₯) π2πβ1,1π,π (π₯, π) + π‘2π2πβ1,0π (π₯, π) , (56) where the subscript π denotes the π₯-homogeneous (exit) principal parts of the involved symbols, which are all SG-classical and real-valued, see [28].
Observe that|π₯| βΌ π on the support of the integrand in πΌ1(π), so that we can, in fact, assume π(π₯) β‘ 1 there. Indeed, recalling that, by definition,π β πΆβ(Rπ), π(π) β‘ 0 for |π| β€ π΅, π(π) β‘ 1 for |π| β₯ 2π΅, with a fixed constant π΅ > 0, it is enough to observe that
σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨ βΊ 1, β¨π₯β© β¨πβ©πβΌ π σ³¨β β¨π₯β© βΌ π, (57) which of course impliesβ¨π₯β©βΌ |π₯|, provided that π0 β€ π is large enough. Moreover, by the ellipticity ofπ, writing π₯ = |π₯|π, π β Sπβ1, with the constantπ΄ > 1 of (48),
π΄β1β¨π₯β© β¨πβ©πβ€ π (π₯, π) = π (π₯) ππ(π₯, π) + ππ,0(π₯, π) β€ π΄ β¨π₯β© β¨πβ©π σ³¨β π΄β1β¨π₯β© |π₯|β¨πβ© πβ€ π (π₯) π π(π, π) +ππ,0(π₯, π) |π₯| β€ π΄β¨ π₯β© |π₯|β¨πβ© π σ³¨β π΄β1β¨πβ©πβ€ ππ(π, π) β€ π΄β¨πβ©π, π β Sπβ1, π β Rπ (58)
taking the limit for|π₯| β +β. Then, setting π₯ = πππ, π β [0, +β), π β Sπβ1,π β₯ π
0β« 1, in πΌ1(π), by homogeneity and
the previous remarks, we can write
Ξ¦ (π‘; πππ, π; π) = βππ‘ + π‘π (πππ) ππ(πππ, π) + π‘ππ,0(πππ, π) + π‘2π (πππ) π2πβ1,1π,π (πππ, π) + π‘2π2πβ1,0π (πππ, π) = βππ‘ + πππ‘ππ(π, π) + πππ‘2π2πβ1,1π,π (π, π) + π‘ππ,0(πππ, π) + π‘2π2πβ1,0 π (πππ, π) = π [βπ‘ + ππ‘ππ(π, π) + ππ‘2ππ,π2πβ1,1(π, π)] + πΊ1(π; π‘, π; π, π) = ππΉ1(π‘, π; π, π) + πΊ1(π; π‘, π; π, π) , (59) and find, in view of the compactness of the support of the integrand (see the proof ofProposition 25 below) and the hypotheses πΌ1(π) = ππβ« ππππΉ1(π‘,π;π,π)πππΊ1(π;π‘,π;π,π)π (π‘) Γ π (π‘; πππ, π) π»1(β¨ππβ© β¨πβ© π π ) Γ π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨)ππβ1ππ‘ππLπππ = ππ (2π)πβ« ππππΉ1(π,π)π1(π, π; π) ππππ, (60)
Proposition 25. Choosing the constants π1, π0 > 1 large
enough andπ > 0 suitably small, one has, for any π2 > 1 and
for a certain sequenceππ,π = 0, 1, . . .,
πΌ1(π) βΌ+ββ
π=0
ππππβ1βπ, (61)
that is,πΌ1(π) = π0ππβ1+ π(ππβ2), with
π0= 1
(2π)πβ1β«Rπβ«Sπβ1
π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨)
ππ(π, π)πππππ. (62)
Proof. It is easy to see that, on the support ofπ1, the phase
functionπΉ1(π, π) admits a unique, nondegenerate, stationary pointπ0 = π0(π) = (0, ππ(π, π)β1), that is, πΉ1,πσΈ (π0(π), π) = 0 for all π such that (π, π) β supp π1, provided thatπ >
0 is chosen suitably small (see, e.g., [25, page 136]), and the Hessian det(πΉσΈ σΈ
1,π(π0(π), π)) equals βππ(π, π)2< 0. Moreover,
the amplitude function
π1(π, π; π) = π (π‘) π»1(β¨ππβ© β¨πβ©
π
π ) π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨) Γ π (π‘; πππ, π) ππβ1πππΊ(π;π‘,π;π,π)
(63)
is compactly supported with respect to the variablesπ and π and satisfies, for allπΎ β Z2+,
π·πΎππ1(π, π; π) βΊ 1 (64)
for allπ, π, π β₯ π0. In fact,
(1)π β πΆ0β((βπ, π)), π β Sπβ1, supp[π»2(|π|)] β {π : |π| β€ 2π2}, and (2π1)β1β€ β¨πβ©πβ 1 π2 + π2β€ 2π1 σ³¨β 0 < β 1 4π2 1β¨2π2β©2π β 1 π2 0 β€ π β€ 2π1, (65) whereπ0> 2π1β¨2π2β©π;
(2) all the factors appearing in the expression of π1 are uniformly bounded, together with all their π-derivatives, forπ β ππ = supp π Γ [π0, π1], π β ππ= Sπβ1Γ {π : |π| β€ 2π
2}, and π β₯ π0.
Of course, (2) trivially holds for the cutoff functions π(π‘) and π»2(|π|), and for the factor ππβ1. Since π(π‘; π₯, π) β
π0,0π (π₯, π), on ππΓ ππwe have, for allπΎ β Z2+andπ β₯ π0> 1,
π·πΎππ (π‘; πππ, π) βΊ β¨ππβ©βπΎ2ππΎ2β¨πβ©π βΊ 1
((1/π2) + π2)πΎ2/2
< 1 ππΎ2 βΊ 1.
(66)
Moreover, sinceπΊ1β ππ,0π (π₯, π) is actually in πββ,0π (π₯, π) β π0,0π (π₯, π) on ππ Γ ππ, the same holds for exp(ππΊ1), by an
application of the Fa`a di Bruno formula for the derivatives of compositions of functions, so also this factor fulfills the desired estimates. Finally, another straightforward computa-tion shows that, for allπΎ2β Z+andπ β₯ π0> 1,
π·πΎ2
ππ»1(β¨ππβ© β¨πβ© π
π ) βΊ 1 (67)
on ππ Γ ππ. The proposition is then a consequence of the stationary phase theorem (see [30, Proposition 1.2.4], [31, Theorem 7.7.6]), applied to the integral with respect to π = (π‘, π). In particular, the leading term is given by ππ/(2π)πβ1 times the integral with respect to π of πβ1| det
(πΉ1,πσΈ σΈ (π0(π), π))|β1/2π1(π0(π), π; π), that is πΌ1(π) = π πβ1 (2π)πβ1β«Rπβ«Sπβ1 1 ππ(π, π) π (0) Γ π»1(β¨π/ππ(π, π)β© β¨πβ© π π ) Γ π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨) ππ(π, π)πβ1π (0; ππ ππ(π, π), π) ππππ + π (ππβ2) = ππβ1 (2π)πβ1β«Rπβ«Sπβ1π»1( β¨π/ππ(π, π)β© β¨πβ©π π ) Γ π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨) ππ(π, π)πππππ + π (ππβ2) = ππβ1 (2π)πβ1β«Rπβ«Sπβ1 π»2(σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨) ππ(π, π)πππππ + π (ππβ2) , (68) recalling thatπ(0) = 1, π(0; π₯, π) = 1 for all π₯, π β Rπ.
Indeed, having chosenπ1 > 2π΄, π0 > 2π1β¨2π2β©π, (58) implies πβ11 < π΄β1 < β¨π/ππ(π, π)β© β¨πβ© π π = β(β¨πβ©π π ) 2 + ( β¨πβ©π ππ(π, π)) 2 < β4π12 1 + π΄2< π 1, (69)
uniformly forπ β Sπβ1, π β supp[π»2(|π|)], π β₯ π0. This concludes the proof.
Let us now considerπΌ2(π). We follow a procedure close to that used in the proof of Theorem7.7.6 of [31]. However, since here we lack the compactness of the support of the amplitude with respect to π₯, we need explicit estimates to show that all the involved integrals are convergent, so we give the argument in full detail in what follows.
We initially proceed as in the analysis ofπΌ1(π) mentioned previously. In view of the presence of the factor1 β π»2(|π|) in the integrand, we can now assume|π| β₯ π2 > max{π΅, 1}, π΅ > 0 the radius of the smallest ball in Rπincluding supp(1 β
π), so that π(π₯, π) = π(π)ππ(π₯, π) + ππβ1,1(π₯, π) = π
π(π₯, π) +
ππβ1,1(π₯, π). Then, with some 0 < πΏ
2< 1, Ξ¦ (π‘; π₯, π; π) = βππ‘ β π₯π + π (0; π₯, π) + π‘ππ‘π (0; π₯, π) +π‘22 ππ‘2π (π‘πΏ2; π₯, π) = βππ‘ + π‘π (π₯, π) + π‘2π2πβ1,1π (π₯, π) = βππ‘ + π‘ππ(π₯, π) + π‘ππβ1,1(π₯, π) + π‘2π2πβ1,1π (π₯, π) . (70) Settingπ = (ππ)1/ππ, π β [0, +β), π β Sπβ1,π β₯ π0, we can rewriteπΌ2(π) as πΌ2(π) = ππ(2π)ππ/ππ Γ β« πππ(βπ‘+ππ‘ππ(π₯,π)+π‘πβ1ππβ1,1(π₯,(ππ)1/ππ)+π‘2πβ1π2πβ1,1π (π₯,(ππ)1/ππ)) Γ π (π‘) π (π‘; π₯, (ππ)1/ππ) Γ π»1(β¨π₯β© β¨(ππ) 1/ππβ©π π ) Γ [1 β π»2((ππ)1/π)] π(π/π)β1ππ‘ππππππ₯ = ππ(2π)ππ/ππβ« ππππΉ2(π,π;π)π 2(π, π; π) ππππ, (71) π = (π‘, π), π = (π, π₯), where we have set
πΉ2(π, π; π) = β π‘ + ππ‘ππ(π₯, π) + π‘πβ1ππβ1,1(π₯, (ππ)1/ππ) + π‘2πβ1π2πβ1,1π (π₯, (ππ)1/ππ) π2(π, π; π) = π (π‘) π»1(β¨π₯β© β¨(ππ) 1/ππβ©π π ) Γ [1 β π»2((ππ)1/π)] Γ π (π‘; π₯, (ππ)1/ππ) π(π/π)β1. (72) On the support ofπ2, we have
β¨π₯β© β¨(ππ)1/ππβ©π π βΌ 1, (ππ)1/πβ» 1 σ³¨β β¨(ππ)1/ππβ©π= β¨(ππ)1/πβ©πβΌ ππ, (73) so that β¨π₯β© ππ π βΌ 1 ββ π βΌ β¨π₯β©β1, |π₯| < β¨π₯β© β€ 2π1(π2)βππ = Μππ. (74)
For any fixedπ β Sπβ1Γ Rπ, we then haveπ belonging to a compact set, uniformly with respect toπ β₯ π0, say suppπ Γ [πβ1β¨π₯β©β1, πβ¨π₯β©β1], for a suitable π > 1.
Remark 26. Incidentally, we observe that a rough estimate of
ππ/ππΌ 2(π) is β« ππππΉ2(π,π;π)π 2(π, π; π) ππ βΊ β¨π₯β©β(π/π)+1β«πβ¨π₯β© β1 πβ1β¨π₯β©β1ππ βΊ β¨π₯β© βπ/π σ³¨β ππ/πβ« ππππΉ2(π,π;π)π 2(π, π; π) ππππ βΊ ππ, π σ³¨β +β. (75)
An even less precise result would be the boundππ/π, using the convergence of the integral with respect toπ₯ in the whole Rπ, given byβ(π/π) + π < 0.
The next lemma is immediate, and we omit the proof.
Lemma 27. ππ ,π
π (π₯, (ππ)1/ππ) = πππ ,π(π₯, (ππ)1/π) for any π β
[0, +β), π₯ β Rπ,π β Sπβ1,π β₯ π0,π β (0, 1), and, for all
πΎ β Z2
+,
π·πΎπππ ,ππ (π₯, (ππ)1/π) = πβπΎ2ππ ,π
π (π₯, (ππ)1/π) . (76)
The main result of this section is as follows.
Proposition 28. If π1, π2, π0> 1 are chosen large enough, one
has πΌ2(π) = π π π0π(π/π)β1+ π (ππβ1) + π (π(π/π)β2) . (77) Explicitly, π0=(2π)1πβ1β«Rπβ«Sπβ1 1 ππ(π₯, π)π/π ππππ₯. (78) We will prove Proposition 28through various interme-diate steps. First of all, arguing as in the proof of (58), exchanging the role ofπ₯ and π, we note that, for all π₯ β Rπ, π β Sπβ1,
π΄β1β¨π₯β© β€ ππ(π₯, π) β€ π΄ β¨π₯β© , (79)
πΉ2,πσΈ (π, π; π) = (ππ‘πΉ2(π, π; π) πππΉ2(π, π; π) ) = ( β1 + π π0 + πβ1ππβ1,1(π₯, (ππ)1/π) + π‘πβ1π2πβ1,1π (π₯, (ππ)1/π) π‘ (ππ(π₯, π) + πβ1πβ1ππβ1,1(π₯, (ππ)1/π) +π‘πβ1πβ1π2πβ1,1 π (π₯, (ππ)1/π)) ) , (80) π = (π‘, π) β ππ = supp π Γ [πβ1β¨π₯β©β1, πβ¨π₯β©β1], π = (π, π₯) β ππ = Sπβ1 Γ Rπ, π β₯ π
0, where we have used
Lemma 27. By the symbolic calculus, remembering thatππ β₯ ππ
2 > 1 on supp π2, we can rewrite the expressions mentioned
previously as ππ‘πΉ2(π, π; π) = β 1 +ππ 0 + π(ππ) β1ππβ1,1(π₯, (ππ)1/π) + π‘π(ππ)β1π2πβ1,1π (π₯, (ππ)1/π) = β 1 +ππ 0 + π[(ππ) 1/π]βπππβ1,1(π₯, (ππ)1/π) + π‘π[(ππ)1/π]βππ2πβ1,1π (π₯, (ππ)1/π) = β 1 + π π0 + ππβ1,1(π₯, (ππ)1/π) + π‘πππβ1,1π (π₯, (ππ)1/π) , πππΉ2(π, π; π) = π‘ (ππ(π₯, π) + πβ1,1(π₯, (ππ)1/π) +π‘ππβ1,1π (π₯, (ππ)1/π)) . (81)
It is clear thatπ βΌ β¨π₯β©β1 impliesππβ1,1(π₯, (ππ)1/π) = πβ1,0 (π₯, (ππ)1/π) and πππβ1,1π (π₯, (ππ)1/π) = ππβ1,0π (π₯, (ππ)1/π), so that we finally have
ππ‘πΉ2(π, π; π) = β 1 + π π0 + πβ1,0(π₯, (ππ)1/π) + π‘ππβ1,0π (π₯, (ππ)1/π) , πππΉ2(π, π; π) = π‘ (ππ(π₯, π) + πβ1,1(π₯, (ππ)1/π) + π‘ππβ1,1π (π₯, (ππ)1/π)) . (82)
We now prove that, modulo anπ(|π|ββ) term, we can consider an amplitude such that, on its support, the ration π/π0is very close to1. To this aim, take π»3β πΆβ0 (R) such that
0 β€ π»3(π) β€ 1, π»3(π) = 1 for |π| β€ (3/2)π and π»3(π) = 0 for
|π| β₯ 2π, with an arbitrarily fixed, small enough π β (0, 1/2), and set π1(π, π; π) = π2(π, π; π) β [1 β π»3(ππ 0 β 1)] , π2(π, π; π) = π2(π, π; π) β π»3(ππ 0 β 1) , π½1(π) = β« ππππΉ2(π,π;π)π 1(π, π; π) ππππ, π½2(π) = β« ππππΉ2(π,π;π)π 2(π, π; π) ππππ. (83)
Proposition 29. With the choices of π, π1, π0, for any π β
(0, 1/2), one can find π2 > 1 large enough such that π½1(π) = π(πββ).
Proof. Since0 < π < 1, in view of (3), (74), and (79), we can
chooseπ2> 1 so large that, for an arbitrarily fixed π β (0, 1/2), for anyπ β₯ π0,π β (0, +β) satisfying |π| = (ππ)1/πβ₯ π2,
inππ‘πΉ2(π, π; π) , σ΅¨σ΅¨σ΅¨σ΅¨ σ΅¨πβ1,0(π₯, (ππ)1/π)σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨ β€ 2π, σ΅¨σ΅¨σ΅¨σ΅¨ σ΅¨π‘πππβ1,0(π₯, (ππ)1/π)σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨ β€ π2, σ΅¨σ΅¨σ΅¨σ΅¨ σ΅¨σ΅¨σ΅¨σ΅¨π0πππ πβ1,0(π₯, (ππ)1/π)σ΅¨σ΅¨σ΅¨σ΅¨ =σ΅¨σ΅¨σ΅¨σ΅¨ σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨π0πβ1πβ1,0(π₯, (ππ)1/π))σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨ β€ π0< 1, (84a) inπππΉ2(π, π; π) , σ΅¨σ΅¨σ΅¨σ΅¨ σ΅¨πβ1,1(π₯, (ππ)1/π) + π‘ππβ1,1π (π₯, (ππ)1/π)σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨ β€ π΄β1 2 β¨π₯β© , (84b)
uniformly with respect to(π, π) β ππΓ ππ β supp π2(β ; π). Then,πΉ2 is nonstationary on suppπ1, since there we have |(π/π0) β 1| β₯ (3/2)π, while
σ΅¨σ΅¨σ΅¨σ΅¨
σ΅¨πβ1,0(π₯, (ππ)1/π) + π‘ππβ1,0π (π₯, (ππ)1/π)σ΅¨σ΅¨σ΅¨σ΅¨σ΅¨ β€ π, (85)
which impliesππ‘πΉ2(π, π; π) β» 1. Observing that, on supp π1, ππ‘πΉ2(π, π; π) = π0,0
π (π₯, (ππ)1/π), as well as π1(π, π; π) =
π0,0π (π₯, (ππ)1/π), the assertion follows by repeated integrations by parts with respect toπ‘, using the operator
πΏπ‘=ππ 1
π‘πΉ2(π, π; π)π·π‘σ³¨β πΏ1π
πππΉ2(π,π;π)
= ππππΉ2(π,π;π),
(86)
and recallingRemark 26.
Proposition 30. With the choices of π, π > 0, π1, π2, π0 > 1,
one can assume, modulo anπ(ππβ1) term, that the integral with
respect toπ₯ in π½2(π) is extended to the set {π₯ β Rπ : β¨π₯β© β€ ππ},
with
π = (1 β π2) [π΄(2π2)π]β1. (87)
Proof. Indeed ifπ < Μπ = 2π1β¨π2β©βπ, we can splitπ½2(π) into
the sum β« ππβ€β¨π₯β©β€Μππβ«Sπβ1β« π πππΉ2π 2ππππππ₯ + β« β¨π₯β©β€ππβ«Sπβ1β« π πππΉ2π 2ππππππ₯, (88)
since the inequalityπ < Μπ is true when π2is sufficiently large. Observing that, on suppπ2,
β¨π₯β© βΌ π σ³¨β β¨πβ©π= β¨π₯β© β¨πβ©
π
π π
β¨π₯β©βΌ 1 σ³¨β σ΅¨σ΅¨σ΅¨σ΅¨πσ΅¨σ΅¨σ΅¨σ΅¨ β€ π3, (89) switching back to the original variables, the first integral in (88) can be treated as πΌ1(π), and gives, in view of
Proposition 25, anπ(ππβ1) term, as stated.
Now we can show that πΉ2(π, π; π) admits a unique, nondegenerate stationary pointπβ0 = πβ0(π, π) belonging to suppπ2forβ¨π₯β©β€ ππ. Under the same hypotheses, πβ0 lies in a circular neighbourhood ofπ0 = (0, π0) = (0, ππ(π₯, π)β1) of arbitrarily small radius.
Proposition 31. With π β (0, 1/2), π > 0, π1, π2, π0 > 1
fixed previously,πΉ2,πσΈ (π, π; π) vanishes on supp π2only forπ =
πβ
0(π; π) = (0, π0β(π; π)), that is, πΉ2,πσΈ (πβ0(π; π), π; π) = 0 for allπ such that (π, π; π) β supp π2. Moreover,
det(πΉ2,πσΈ σΈ (πβ0(π; π) , π)) βΌ β¨π₯β©2, σ΅¨σ΅¨σ΅¨σ΅¨πβ 0(π; π) β π0(π)σ΅¨σ΅¨σ΅¨σ΅¨ =σ΅¨σ΅¨σ΅¨σ΅¨π0β(π; π) β π0(π)σ΅¨σ΅¨σ΅¨σ΅¨ β€ π΄π2 β¨π₯β©β1 (90) holds on suppπ2.
Proof. We have to solve
0 = β 1 +ππ 0 + π β1,0(π₯, (ππ)1/π) + π‘ππβ1,0π (π₯, (ππ)1/π) 0 = π‘ (ππ(π₯, π) + πβ1,1(π₯, (ππ)1/π) +π‘ππβ1,1π (π₯, (ππ)1/π)) , (91) (π, π; π) β supp π2. By (79) and (84a) and (84b), with the choices ofπ, π > 0, π1, π2, π0, the coefficient ofπ‘ in the second equation does not vanish at any point of suppπ2. Thenπ‘ = 0, andπ must satisfy
β 1 + π
π0 + πβ1,0(π₯, (ππ)1/π) = 0
ββ π = π0(1 + πβ1,0(π₯, (ππ)1/π))
= πΊ (π; π; π) .
(92)
Since, by the choice ofπ2,|πππΊ(π; π; π)| β€ π0< 1, uniformly with respect toπ β Sπβ1Γ {π₯ β Rπ : β¨π₯β© β€ ππ}, π β₯ π0,πΊ has a unique fixed pointπβ0 = π0β(π; π), smoothly depending on the parameters; see the Appendix for more details. Since
π2π‘πΉ2(π, π; π) = ππβ1,0π (π₯, (ππ)1/π) , ππ‘πππΉ2(π, π; π) = ππ(π₯, π) (1 + π0πβ1(πβ1,0(π₯, (ππ)1/π) +π‘ππβ1,0 π (π₯, (ππ)1/π))) , π2ππΉ2(π, π; π) = π‘πβ1(πβ1,1(π₯, (ππ)1/π) +π‘πππβ1,1(π₯, (ππ)1/π)) , (93)
we can assume thatππ β₯ ππ2 and the choices of the other parameters imply, on suppπ2,
ππ‘2πΉ2(π, π; π) βΊ π2, ππ‘πππΉ2(π, π; π) βΌ β¨π₯β© , ππ2πΉ2(π, π; π) βΊ π
2β¨π₯β©2.
(94)