Abstract

A family π»πœ‡(𝑝), πœ‡>0, π‘βˆˆπ•‹2 of the Friedrichs models with the perturbation of rank one, associated to a system of two particles, moving on the two-dimensional lattice β„€2 is considered. The existence or absence of the unique eigenvalue of the operator π»πœ‡(𝑝) lying below threshold depending on the values of πœ‡>0 and π‘βˆˆπ‘ˆπ›Ώ(0)βŠ‚π•‹2 is proved. The analyticity of corresponding eigenfunction is shown.

1. Introduction

In celebrated work [1] of Simon and Klaus it has been considered a family of the SchrΓΆdinger operators 𝐻=βˆ’Ξ”+πœ‡π‘‰ and, a situation where as πœ‡ tends to πœ‡0 some eigenvalue 𝑒𝑖(πœ‡) tends to 0, that is, as πœ‡β†“πœ‡0 an eigenvalue is absorbed into continuous spectrum, and conversely, for any πœ‡βˆΆπœ‡>πœ‡0 continuous spectrum gives birth to a new eigenvalue. This phenomenon in [1] is called coupling constant threshold.

In [2] the Hamiltonian of a system of two identical quantum mechanical particles (bosons) moving on the 𝑑-dimensional lattice ℀𝑑,𝑑β‰₯3 and interacting via zero-range repulsive pair potentials has been considered. For the associated two-particle SchrΓΆdinger operator π»πœ‡(π‘˜),π‘˜βˆˆπ•‹π‘‘=(βˆ’πœ‹,πœ‹]𝑑 the existence of coupling constant threshold πœ‡0=πœ‡0(π‘˜)>0 has been proved: the operator has none eigenvalue for any 0<πœ‡β‰€πœ‡0, but for each πœ‡>πœ‡0 it has a unique eigenvalue 𝑧(πœ‡,π‘˜) above the upper threshold of the spectrum.

Note that in [1] the existence of a coupling constant threshold has been assumed, at the same time in [2] the coupling constant threshold has been definitely found by the given data of the Hamiltonian.

We remark that for the Hamiltonians of a system of two arbitrary particles moving on ℝ𝑑 or ℀𝑑,𝑑β‰₯1 the coupling constant threshold vanishes, if 𝑑=1,2 and the coupling constant threshold is positive, if 𝑑β‰₯3.

Notice also that for the Hamiltonians of a system of two identical particles moving on ℝ𝑑 or ℀𝑑,𝑑=1,2 the coupling constant threshold vanishes, if particles are bosons and the coupling constant threshold is positive, if particles are fermions.

In [3] for a wide class of the two-particle SchrΓΆdinger operators π»πœ‡(π‘˜) on the 𝑑-dimensional lattice ℀𝑑, 𝑑β‰₯3, π‘˜ being the two-particle quasimomentum, it has been proved that if the following two assumptions (i) and (ii) are satisfied, then for all π‘˜β‰ 0, the discrete spectrum of π»πœ‡(π‘˜) below its threshold is nonempty. The assumptions are (i) the two-particle SchrΓΆdinger operator π»πœ‡(0), corresponding to the zero value of the quasimomentum π‘˜, has a coupling constant threshold πœ‡0(0)>0 and (ii) the one-particle free Hamiltonians in the coordinate representation generate positivity preserving semigroups.

In [4] a family of the Friedrichs models π»πœ‡(𝑝),πœ‡>0,π‘βˆˆ(βˆ’πœ‹,πœ‹]3 with perturbation of rank one associated to a system of two particles on the three-dimensional lattice β„€3 has been considered. In some special case of the multiplication operator and under the assumption that the operator π»πœ‡(0),0βˆˆπ•‹3 has a coupling constant threshold πœ‡0(0)>0, the existence of a unique eigenvalue, below the threshold of the spectrum of π»πœ‡0(0)(𝑝),π‘βˆˆ(βˆ’πœ‹,πœ‹]3 for all nontrivial values of π‘βˆˆπ•‹3, has been proved.

In the present paper, a family of the Friedrichs models π»πœ‡(𝑝), πœ‡>0,π‘βˆˆπ‘ˆπ›Ώ(0)βŠ‚π•‹2, where π‘ˆπ›Ώ(0) is a 𝛿-neighborhood of the point 𝑝=0βˆˆπ•‹2 with perturbation of rank one associated to a system of two particles on the two-dimensional lattice β„€2 interacting via pair local potentials, is considered and the following results have been obtained.(i)If the parameters of the Friedrichs model satisfy some conditions (see Theorem 2.3), then there exists a coupling constant threshold πœ‡0=πœ‡0(𝑝)>0∢ for any 0<πœ‡β‰€πœ‡0(𝑝) the operator has none eigenvalue; at the same time for any πœ‡>πœ‡0(𝑝) it has a unique eigenvalue 𝑧(πœ‡,𝑝), lying below its threshold of the spectrum. Moreover an explicit expression for the corresponding eigenfunction is found and its analyticity is proven.(ii)If the parameters of the Friedrichs model do not satisfy conditions mentioned in (i), then the operator has none positive coupling constant threshold, that is, for any πœ‡>0 the operator π»πœ‡(𝑝) has a unique eigenvalue 𝑧(πœ‡,𝑝), lying below its threshold of the spectrum.(iii)A criterion for being the threshold π‘š(𝑝),π‘βˆˆπ‘ˆπ›Ώ(0) of the spectrum of π»πœ‡(𝑝) a virtual level of the operator π»πœ‡(𝑝) is proven.

Note that the generalized Friedrichs models appear in the problems of quantum mechanics [5], solid state physics [6], and quantum field theory [7, 8] and the existence of its eigenvalues and resonances have been studied in [9, 10].

In [11] a special family of generalized Friedrichs models has been considered and the existence of eigenvalues for some values of quasimomentum π‘βˆˆπ•‹π‘‘ of the system, lying in a neighborhood of some 𝑝0βˆˆπ•‹π‘‘, has been proved.

2. Notions and Assumptions: The Main Results

Let β„€ be the one-dimensional hypercubic lattice and 𝕋2=(ℝ/2πœ‹β„€)2=(βˆ’πœ‹,πœ‹]2 be the two-dimensional torus, the dual group of β„€2 (Brillion zone). Note that operations addition and multiplication by number of the elements of torus 𝕋2≑(βˆ’πœ‹,πœ‹]2βŠ‚β„2 is defined as operations in ℝ2 by the module (2πœ‹β„€)2.

Let 𝐿2(𝕋2) be the Hilbert space of square-integrable functions defined on the torus 𝕋2 and 𝐂1 be one-dimensional complex Hilbert space.

We consider the family of generalized Friedrichs model acting in 𝐿2(𝕋2) as follows: π»πœ‡(𝑝)=𝐻0(𝑝)βˆ’πœ‡Ξ¦βˆ—Ξ¦,πœ‡>0.(2.1) Here Φ∢𝐿2𝕋2ξ€ΈβŸΆπ‚1,Φ𝑓=(𝑓,πœ‘)𝐿2(𝕋2),Ξ¦βˆ—βˆΆπ‚1⟢𝐿2𝕋2ξ€Έ,Ξ¦βˆ—π‘“0=πœ‘(π‘ž)𝑓0,(2.2) where (β‹…,β‹…)𝐿2(𝕋2) is inner product in 𝐿2(𝕋2) and 𝐻0(𝑝),π‘βˆˆπ•‹2 is the multiplication operator by function 𝑀𝑝(β‹…)∢=𝑀(𝑝,β‹…), that is, 𝐻0ξ€Έ(𝑝)𝑓(π‘ž)=𝑀𝑝(π‘ž)𝑓(π‘ž),π‘“βˆˆπΏ2𝕋2ξ€Έ.(2.3) Note that for anyπ‘“βˆˆπΏ2(𝕋2) and 𝑔0βˆˆπ‚1 the equality Φ𝑓,𝑔0𝐂1=𝑓,Ξ¦βˆ—π‘”0𝐿2(𝕋2),(2.4) holds.

The following assumption will be needed throughout the paper.

Assumption 2.1. The following conditions are satisfied: (i)the function πœ‘(β‹…) is nontrivial real-analytic on 𝕋2;(ii)the function 𝑀(β‹…,β‹…) is real-analytic on (𝕋2)2=𝕋2×𝕋2 and has a unique nondegenerated minimum at (0,0)∈(𝕋2)2.

The perturbation 𝑣=Ξ¦βˆ—Ξ¦ is positive operator of rank one. Consequently, by well-known Weyl's theorem [12], the essential spectrum of π»πœ‡(𝑝) fills the following segment on the real axis: 𝜎essξ€·π»πœ‡ξ€Έ(𝑝)=𝜎ess𝐻0ξ€Έ=[π‘š],(𝑝)(𝑝),𝑀(𝑝)(2.5) where π‘š(𝑝)=minπ‘žβˆˆπ•‹2𝑀𝑝(π‘ž),𝑀(𝑝)=maxπ‘žβˆˆπ•‹2𝑀𝑝(π‘ž).(2.6)

By Assumption 2.1 there exist such 𝛿-neighborhood π‘ˆπ›Ώ(0)βŠ‚π•‹2 of the point 𝑝=0βˆˆπ•‹2 and analytic vector function π‘ž0βˆΆπ‘ˆπ›Ώ(0)→𝕋2 that for any π‘βˆˆπ‘ˆπ›Ώ(0) the point π‘ž0(𝑝)=(π‘ž0(1)(𝑝),π‘ž0(2)(𝑝))βˆˆπ•‹2 is a unique nondegenerated minimum of the function 𝑀𝑝(β‹…) (see Lemma 3.2).

Moreover, in the case πœ‘(π‘ž0(𝑝))=0,π‘βˆˆπ‘ˆπ›Ώ(0) the following integral ξ€œπ•‹2πœ‘2(𝑠)𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘š(𝑝)>0,(2.7) exists (see Lemma 3.4) and we introduce a parameter πœ‡(𝑝) as 1=ξ€œπœ‡(𝑝)𝕋2πœ‘2(𝑠)𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘š(𝑝)>0.(2.8) If πœ‘(π‘ž0(𝑝))β‰ 0,π‘βˆˆπ‘ˆπ›Ώ(0), then we define πœ‡(𝑝) as πœ‡(𝑝)=0.

Definition 2.2. The number 𝑧=π‘š(𝑝) is called a virtual level of the operator π»πœ‡(𝑝), if the equation π»πœ‡(𝑝)𝑓=π‘š(𝑝)𝑓 has a nonzero solution π‘“βˆˆπΏ1(𝕋2)⧡𝐿2(𝕋2), where 𝐿1(𝕋2) is the Banach space of integrable functions on 𝕋2. The corresponding solution 𝑓 is called a virtual state of the operator π»πœ‡(𝑝).

In the following theorem we assert that for any πœ‡>πœ‡(𝑝) there exists a unique eigenvalue 𝐸(πœ‡,𝑝), lying below the essential spectrum, of the operator π»πœ‡(𝑝),π‘βˆˆπ‘ˆπ›Ώ(0), but for 0<πœ‡β‰€πœ‡(𝑝),π‘βˆˆπ‘ˆπ›Ώ(0) the operator π»πœ‡(𝑝) has none eigenvalue outside the essential spectrum. It is proved that for fixed π‘βˆˆπ‘ˆπ›Ώ(0), the function 𝐸(β‹…,𝑝) is analytic in (πœ‡(𝑝),+∞).

Moreover, this theorem provides a criterion, for being the bottom π‘š(𝑝),π‘βˆˆπ‘ˆπ›Ώ(0) of the essential spectrum of π»πœ‡(𝑝), a virtual level of the operator π»πœ‡(𝑝).

Theorem 2.3. Let Assumption 2.1 holds and π‘βˆˆπ‘ˆπ›Ώ(0). Then the following assertions are true. (i)If πœ‡>πœ‡(𝑝), then the operator π»πœ‡(𝑝) has a unique eigenvalue 𝐸(πœ‡,𝑝), lying below the essential spectrum of π»πœ‡(𝑝). The function 𝐸(β‹…,𝑝) is monotonously decreasing real-analytic function in the interval (πœ‡(𝑝),+∞) and the function 𝐸(πœ‡,β‹…) is real-analytic in π‘ˆπ›Ώ(0). The corresponding eigenfunction Ξ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝))=πΆπœ‡πœ‘(β‹…)𝑀𝑝,(β‹…)βˆ’πΈ(πœ‡,𝑝)(2.9) is analytic on 𝕋2, where 𝐢≠0 is a normalizing constant. Moreover, the mappings Ξ¨πœ‡βˆΆπ‘ˆπ›Ώ(0)⟢𝐿2𝕋2ξ€Έ,π‘βŸΌΞ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝))∈𝐿2𝕋2ξ€Έ,Ξ¨π‘βˆΆ(πœ‡(𝑝),+∞)⟢𝐿2𝕋2ξ€Έ,πœ‡βŸΌΞ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝))∈𝐿2𝕋2ξ€Έ,(2.10) are vector-valued analytic functions in π‘ˆπ›Ώ(0) and (πœ‡(𝑝),+∞), respectively.(ii)If πœ‘(π‘ž0(𝑝))=0 and 0<πœ‡<πœ‡(𝑝), then the operator π»πœ‡(𝑝) has none eigenvalue in (βˆ’βˆž,π‘š(𝑝)].(iii)If πœ‘(π‘ž0(𝑝))=0,β€‰βˆ‡πœ‘(π‘ž0(𝑝))=((πœ•πœ‘/πœ•π‘ž1)(π‘ž0(𝑝)),(πœ•πœ‘/πœ•π‘ž2)(π‘ž0(𝑝)))β‰ 0 and πœ‡=πœ‡(𝑝), then the number 𝑧=π‘š(𝑝) is a virtual level of the operator π»πœ‡(𝑝) and the corresponding virtual state is of the form: 𝑓𝑝(β‹…)=πΆπœ‡(𝑝)πœ‘(β‹…)𝑀𝑝(β‹…)βˆ’π‘š(𝑝)∈𝐿1𝕋2⧡𝐿2𝕋2ξ€Έ,(2.11) where 𝐢≠0 is a normalizing constant.(iv)If πœ‘(π‘ž0(𝑝))=0,β€‰βˆ‡πœ‘(π‘ž0(𝑝))=((πœ•πœ‘/πœ•π‘ž1)(π‘ž0(𝑝)),(πœ•πœ‘/πœ•π‘ž2)(π‘ž0(𝑝)))=0 and πœ‡=πœ‡(𝑝), then the number π‘š(𝑝)=𝑀𝑝(π‘ž0(𝑝)) is an eigenvalue of the operator π»πœ‡(𝑝) and the corresponding eigenfunction is of the form 𝑓𝑝(π‘ž)=πΆπœ‡(𝑝)πœ‘(β‹…)𝑀𝑝(β‹…)βˆ’π‘š(𝑝)∈𝐿2𝕋2ξ€Έ,(2.12) where 𝐢≠0 is a normalizing constant.

Remark 2.4. Notice that if πœ‘(π‘ž0(𝑝))β‰ 0, then πœ‡(𝑝)=0. So, in this case the number 𝑧=π‘š(𝑝) is neither a virtual level nor an eigenvalue for the operator π»πœ‡(𝑝).

Remark 2.5. From the positivity of Ξ¦βˆ—Ξ¦ it follows that the operator π»πœ‡(𝑝) has none eigenvalue lying above 𝑀(𝑝).

3. Proof of the Results

We postpone the proof of the theorem after several lemmas and remarks.

For any πœ‡>0 and π‘βˆˆπ•‹2 we define in β„‚β§΅[π‘š(𝑝);𝑀(𝑝)] an analytic function Ξ”(πœ‡,𝑝;β‹…) (the Fredholm determinant Ξ”(πœ‡,𝑝;β‹…), associated to the operator π»πœ‡(𝑝)) as Ξ”(πœ‡,𝑝;β‹…)=1βˆ’πœ‡Ξ©(𝑝;β‹…),(3.1) where ξ€œΞ©(𝑝;𝑧)=𝕋2πœ‘2(𝑠)𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘§,π‘βˆˆπ•‹2[].,π‘§βˆˆβ„‚β§΅π‘š(𝑝);𝑀(𝑝)(3.2)

Lemma 3.1. For any πœ‡βˆˆ(πœ‡(𝑝),+∞) and π‘βˆˆπ‘ˆπ›Ώ(0) the number π‘§βˆˆβ„‚β§΅πœŽess(π»πœ‡(𝑝)),π‘βˆˆπ•‹2 is an eigenvalue of the operator π»πœ‡(𝑝) if and only if, when Ξ”(πœ‡,𝑝;𝑧)=0.(3.3) The corresponding eigenfunction π‘“πœ‡,𝑝(β‹…)=πΆπœ‡πœ‘(β‹…)𝑀𝑝,(β‹…)βˆ’π‘§(3.4) is analytic on 𝕋2, where 𝐢≠0 is a normalizing constant [4].

Lemma 3.2. Let Assumption 2.1 holds. Then there exist such a 𝛿-neighborhood π‘ˆπ›Ώ(0)βŠ‚π•‹2 of the point 𝑝=0 and analytic function π‘ž0βˆΆπ‘ˆπ›Ώ(0)→𝕋2 that for any π‘βˆˆπ‘ˆπ›Ώ(0) the point π‘ž0(𝑝) is a unique non degenerated minimum of 𝑀𝑝(β‹…).

Proof. By Assumption 2.1, the square matrix ξƒ©πœ•π΄(0)=2𝑀0πœ•π‘žπ‘–πœ•π‘žπ‘—ξƒͺ(0)2𝑖,𝑗=1>0,(3.5) is positively defined and βˆ‡π‘€0(0)=0. Then by the implicit function theorem in analytic case there exist a 𝛿-neighborhood π‘ˆπ›Ώ(0)βŠ‚π•‹2 of 𝑝=0βˆˆπ•‹2 and a unique analytic vector function π‘ž0(β‹…)βˆΆπ‘ˆπ›Ώ(0)→𝕋2 such that βˆ‡π‘€π‘(π‘ž0(𝑝))=0 and ξƒ©πœ•π΄(𝑝)=2π‘€π‘πœ•π‘žπ‘–πœ•π‘žπ‘—ξ€·π‘ž0ξ€Έξƒͺ(𝑝)2𝑖,𝑗=1>0,π‘βˆˆπ‘ˆπ›Ώ(0).(3.6) Hence for any π‘βˆˆπ‘ˆπ›Ώ(0) the point π‘ž0(𝑝) is a unique non degenerated minimum of the function 𝑀𝑝(β‹…).

Remark 3.3. We note that by the parametrical Morse lemma for any π‘βˆˆπ‘ˆπ›Ώ(0) there exists a map 𝑠=πœ“(𝑦,𝑝) of the sphere π‘Šπ›Ύ(0)βŠ‚β„2 with radius 𝛾>0 and center at 𝑦=0 to a neighborhood π‘ˆ(π‘ž0(𝑝)) of the point π‘ž0(𝑝) that in π‘ˆ(π‘ž0(𝑝)) the function 𝑀𝑝(πœ“(𝑦,𝑝)) can be represented as 𝑀𝑝(πœ“(𝑦,𝑝))=π‘š(𝑝)+𝑦2.(3.7) Here the function πœ“(𝑦,β‹…) (resp., πœ“(β‹…,𝑝)) is analytic in π‘ˆπ›Ώ(0) (resp., π‘Šπ›Ύ(0)) and πœ“(0,𝑝)=π‘ž0(𝑝). Moreover, the Jacobian 𝐽(πœ“(𝑦,𝑝)) of the mapping 𝑠=πœ“(𝑦,𝑝) is analytic in π‘Šπ›Ύ(0) and positive, that is 𝐽(πœ“(𝑦,𝑝))>0 for all π‘βˆˆπ‘ˆπ›Ώ(0) and for all π‘¦βˆˆπ‘Šπ›Ύ(0).

Lemma 3.4. Let Assumption 2.1 holds. Then the integral πœ‰ξ€œ(𝑝)=𝕋2πœ‘2(𝑠)βˆ’πœ‘2ξ€·π‘ž0ξ€Έ(𝑝)𝑀𝑝(𝑠)βˆ’π‘š(𝑝)𝑑𝑠,(3.8) exists and defines an analytic function in π‘ˆπ›Ώ(0).

Proof. We represent the function πœ‰ξ€œ(𝑝,𝑧)=𝕋2πœ‘2(𝑠)βˆ’πœ‘2ξ€·π‘ž0ξ€Έ(𝑝)𝑀𝑝(𝑠)βˆ’π‘§π‘‘π‘ ,(3.9) as πœ‰(𝑝,𝑧)=πœ‰1(𝑝,𝑧)+πœ‰2(𝑝,𝑧),(3.10) where πœ‰1ξ€œ(𝑝,𝑧)=π‘ˆ(π‘ž0(𝑝))πœ‘2(𝑠)βˆ’πœ‘2ξ€·π‘ž0ξ€Έ(𝑝)π‘€π‘πœ‰(𝑠)βˆ’π‘§π‘‘π‘ ,2ξ€œ(𝑝,𝑧)=𝕋2β§΅π‘ˆ(π‘ž0(𝑝))πœ‘2(𝑠)βˆ’πœ‘2ξ€·π‘ž0ξ€Έ(𝑝)𝑀𝑝(𝑠)βˆ’π‘§π‘‘π‘ ,(3.11) and π‘ˆ(π‘ž0(𝑝)) is a neighborhood of π‘ž0(𝑝).
Observe that by Assumption 2.1 for any π‘βˆˆπ‘ˆπ›Ώ(0) the function πœ‰2(𝑝,𝑧) is analytic at the point 𝑧=π‘š(𝑝).
According to Remark 3.3 in the integral for πœ‰1(𝑝,𝑧) a change of variables 𝑠=πœ“(𝑦,𝑝) implies πœ‰1ξ€œ(𝑝,𝑧)=π‘Šπ›Ύ(0)πœ‘2(πœ“(𝑦,𝑝))βˆ’πœ‘2ξ€·π‘ž0ξ€Έ(𝑝)𝑦2𝐽+π‘š(𝑝)βˆ’π‘§(πœ“(𝑦,𝑝))𝑑𝑦,(3.12) where 𝐽(πœ“(𝑦,𝑝)) is the Jacobian of the mapping πœ“(𝑦,𝑝).
Passing to spherical coordinates as 𝑦=π‘Ÿπœˆ, we obtain πœ‰1ξ€œ(𝑝,𝑧)=𝛾0π‘Ÿπ‘Ÿ2ξ‚»ξ€œ+π‘š(𝑝)βˆ’π‘§Ξ©2ξ€Ίπœ‘2(πœ“(π‘Ÿπœˆ,𝑝))βˆ’πœ‘2ξ€·π‘ž0ξ‚Ό(𝑝)𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆπ‘‘π‘Ÿ,(3.13) where Ξ©2 is a unit sphere in ℝ2 and π‘‘πœˆ its element. Inner integral can be represented as ξ€œΞ©2ξ€Ίπœ‘2(πœ“(π‘Ÿπœˆ,𝑝))βˆ’πœ‘2ξ€·π‘ž0(𝑝)𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆ=βˆžξ“π‘›=1πœπ‘›(𝑝)π‘Ÿ2𝑛,(3.14) where the Pizetti coefficients πœπ‘›(𝑝),𝑛=1,2,… are analytic in π‘ˆπ›Ώ(0) [13].
Thus we have that πœ‰1(𝑝,𝑧)=βˆžξ“π‘›=1πœπ‘›(ξ€œπ‘)𝛾0π‘Ÿ2𝑛+1π‘‘π‘Ÿπ‘Ÿ2.+π‘š(𝑝)βˆ’π‘§(3.15) From (3.15) it follows that the following limit exists πœ‰1(𝑝)=limπ‘§β†’π‘š(𝑝)βˆ’0πœ‰1(𝑝,𝑧)=limβˆžπ‘§β†’π‘š(𝑝)βˆ’0𝑛=1πœπ‘›(ξ€œπ‘)𝛾0π‘Ÿ2𝑛+1π‘‘π‘Ÿπ‘Ÿ2=+π‘š(𝑝)βˆ’π‘§βˆžξ“π‘›=1𝛾2π‘›πœ2𝑛𝑛(𝑝),(3.16) and, consequently, πœ‰(𝑝)=limπ‘§β†’π‘š(𝑝)βˆ’0πœ‰(𝑝,𝑧)=πœ‰1(𝑝)+πœ‰2(𝑝),(3.17) where πœ‰2(𝑝)=πœ‰2(𝑝,π‘š(𝑝)). Observe that the functions in the right hand side of (3.17) are analytic in π‘βˆˆπ‘ˆπ›Ώ(0). So, the function πœ‰(𝑝) is analytic in π‘βˆˆπ‘ˆπ›Ώ(0).

Proposition 3.5. For 𝜁<0 the following equalities hold: πΌπ‘›ξ€œ(𝜁)=𝛿0π‘Ÿ2𝑛+1π‘‘π‘Ÿπ‘Ÿ21βˆ’πœ=βˆ’2πœπ‘›ξπΌln(βˆ’πœ)+𝑛(𝜁),𝑛=0,1,2,…,(3.18) where ln(βˆ’πœ) is real for 𝜁<0 and 𝐼𝑛(𝜁) is an analytic function in a neighborhood of the origin [14].

In the following lemma we establish an expansion of Ξ”(πœ‡,𝑝;𝑧) in a half neighborhood (π‘š(𝑝)βˆ’πœ€,π‘š(𝑝)) of the point 𝑧=π‘š(𝑝).

Lemma 3.6. Assume Assumption 2.1. Then for any πœ‡>0,β€‰π‘βˆˆπ‘ˆπ›Ώ(0) and sufficiently small π‘š(𝑝)βˆ’π‘§>0 the function Ξ”(πœ‡,𝑝;β‹…) can be represented as the following convergent series: Ξ”(πœ‡,𝑝;𝑧)=1βˆ’πœ‡π›Ό0(πœ‡π‘)ln(π‘š(𝑝)βˆ’π‘§)+2ln(π‘š(𝑝)βˆ’π‘§)βˆžξ“π‘›=1𝛼𝑛(𝑝)(π‘š(𝑝)βˆ’π‘§)π‘›π›Όβˆ’πœ‡πΉ(𝑝,𝑧),01(𝑝)=βˆ’2πœ‘2ξ€·π‘ž0ξ€Έπ½ξ€·π‘ž(𝑝)0ξ€Έ,𝐹(𝑝)(𝑝,𝑧)=βˆžξ“π‘›=0𝑐𝑛(𝑝)(π‘š(𝑝)βˆ’π‘§)𝑛,(3.19) where 𝛼𝑛(𝑝),𝑐𝑛(𝑝),𝑛=0,1,2,… are real numbers.

Proof. The function Ξ©(𝑝;β‹…) can be written as a sum of the following functions: Ξ©(𝑝;β‹…)=𝐼1(𝑝,β‹…)+𝐼2(𝑝,β‹…)+𝐼3(𝑝,β‹…),(3.20) where 𝐼1(𝑝,𝑧)=πœ‘2ξ€·π‘ž0(ξ€Έξ€œπ‘)π‘ˆ(π‘ž0(𝑝))𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘§,𝐼2(𝑝,𝑧)=πœ‘2ξ€·π‘ž0(ξ€Έξ€œπ‘)𝕋2β§΅π‘ˆ(π‘ž0(𝑝))𝑑𝑠𝑀𝑝,𝐼(𝑠)βˆ’π‘§3ξ€œ(𝑝,𝑧)=𝕋2ξ€·πœ‘2(𝑠)βˆ’πœ‘2ξ€·π‘ž0(𝑝)𝑑𝑠𝑀𝑝,(𝑠)βˆ’π‘§(3.21) and π‘ˆ(π‘ž0(𝑝)) is a neighborhood of the point π‘ž0(𝑝),π‘βˆˆπ‘ˆπ›Ώ(0).
Since minπ‘žβˆˆπ•‹2𝑀𝑝(π‘ž)=𝑀𝑝(π‘ž0(𝑝)) for any π‘βˆˆπ‘ˆπ›Ώ(0), the function 𝐼2(𝑝,𝑧) is analytic at 𝑧=π‘š(𝑝). According to Lemma 3.4 the function 𝐼3(𝑝,π‘š(𝑝)) is analytic in π‘ˆπ›Ώ(0).
A change of variables 𝑠=πœ“(𝑦,𝑝) in the integral 𝐼1(𝑝,𝑧) yields 𝐼1(𝑝,𝑧)=πœ‘2ξ€·π‘ž0(ξ€Έξ€œπ‘)π‘Šπ›Ύ(0)𝐽(πœ“(𝑦,𝑝))π‘‘π‘¦π‘š(𝑝)+𝑦2.βˆ’π‘§(3.22) Passing to spherical coordinates by 𝑦=π‘Ÿπœˆ we obtain 𝐼1(𝑝,𝑧)=πœ‘2ξ€·π‘ž0ξ€Έξ€œ(𝑝)𝛾0ξ€œΞ©2𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘Ÿπ‘‘πœˆπ‘‘π‘Ÿπ‘š(𝑝)+π‘Ÿ2,βˆ’π‘§(3.23) and hence 𝐼1(𝑝,𝑧)=πœ‘2ξ€·π‘ž0ξ€Έξ€œ(𝑝)𝛾0ξ‚΅ξ€œΞ©2𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆπ‘Ÿπ‘‘π‘Ÿπ‘š(𝑝)+π‘Ÿ2,βˆ’π‘§(3.24) where Ξ©2 is unit sphere in ℝ2. Since ξ€œΞ©2𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆ=βˆžξ“π‘›=0𝛼𝑛(𝑝)π‘Ÿ2𝑛,(3.25) where 𝛼𝑛(𝑝), 𝑛=0,1,… are the Pizetti coefficients, we get 𝐼1(𝑝,𝑧)=πœ‘2ξ€·π‘ž0(𝑝)βˆžξ“π‘›=0𝛼𝑛(ξ€œπ‘)𝛾0π‘Ÿ2𝑛+1π‘‘π‘Ÿπ‘š(𝑝)+π‘Ÿ2,βˆ’π‘§(3.26) where 𝛼0(𝑝)=𝐽(π‘ž0(𝑝)). Using Proposition 3.5 we have βˆžξ“π‘›=0𝛼𝑛(ξ€œπ‘)𝛾0π‘Ÿ2𝑛+1π‘‘π‘Ÿπ‘š(𝑝)+π‘Ÿ21βˆ’π‘§=βˆ’2ln(π‘š(𝑝)βˆ’π‘§)βˆžξ“π‘›=0𝛼𝑛(𝑝)(π‘š(𝑝)βˆ’π‘§)𝑛+Ξ¦(𝑝,𝑧),(3.27) where βˆ‘Ξ¦(𝑝,𝑧)=βˆžπ‘›=0𝛽𝑛(𝑝)(π‘š(𝑝)βˆ’π‘§)𝑛 and 𝛼𝑛(𝑝)=(βˆ’1)𝑛𝛼𝑛(𝑝). Using relations (3.27) and (3.21) and putting (3.26) in (3.20) we get required relation (3.19).

Lemma 3.7. Let Assumption 2.1 hold. Then for any π‘βˆˆπ‘ˆπ›Ώ(0) consider (i)if πœ‘(π‘ž0(𝑝))=βˆ‡πœ‘(π‘ž0(𝑝))=0, then 𝑓𝑝(π‘ž)=πœ‘(β‹…)𝑀𝑝(β‹…)βˆ’π‘š(𝑝)∈𝐿2𝕋2ξ€Έ;(3.28)(ii)if πœ‘(π‘ž0(𝑝))=0,βˆ‡πœ‘(π‘ž0(𝑝))β‰ 0, then π‘“π‘βˆˆπΏ1(𝕋2)⧡𝐿2(𝕋2).

Proof. We consider the following integral: ξ€œπΌ(𝑝)=𝕋2𝐹(𝑠)𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘š(𝑝)π‘˜,(3.29) where 𝐹(β‹…) is a continuous function on 𝕋2 and π‘˜βˆˆπ‘. By Lemma 3.2 for any π‘βˆˆπ‘ˆπ›Ώ(0) the function 𝑀𝑝(β‹…) has a unique non degenerated minimum at π‘ž=π‘ž0(𝑝). Then there exist a neighborhood π‘ˆ(π‘ž0(𝑝))βŠ‚π•‹2 of the point π‘ž=π‘ž0(𝑝) and positive number 𝑐𝑝>0 that 𝑐𝑝≀𝑀𝑝(π‘ž)βˆ’π‘š(𝑝),π‘žβˆˆπ•‹2ξ€·π‘žβ§΅π‘ˆ0ξ€Έ.(𝑝)(3.30)
We represent the function 𝐼(β‹…) as a sum of two functions: 𝐼(β‹…)=𝐼1(β‹…)+𝐼2(β‹…),(3.31) where 𝐼1(ξ€œπ‘)=π‘ˆ(π‘ž0(𝑝))𝐹(𝑠)𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘š(𝑝)π‘˜,𝐼2(ξ€œπ‘)=𝕋2β§΅π‘ˆ(π‘ž0(𝑝))𝐹(𝑠)𝑑𝑠𝑀𝑝(𝑠)βˆ’π‘š(𝑝)π‘˜.(3.32) From (3.30) we get that 𝐼2(𝑝)<∞. In the integral for 𝐼1(𝑝) making a change of variables π‘ βˆΆ=πœ“(𝑦,𝑝) one obtains 𝐼1(ξ€œπ‘)=π‘Šπ›Ύ(0)𝐹(πœ“(𝑦,𝑝))𝐽(πœ“(𝑦,𝑝))𝑑𝑦𝑦2π‘˜,(3.33) where 𝐽(πœ“(𝑦,𝑝)) is the Jacobian of the mapping 𝑠=πœ“(𝑦,𝑝).
(i) Let 𝐹(𝑠)=πœ‘2(𝑠),β€‰π‘˜=2. Then from (3.33) we get 𝐼1ξ€œ(𝑝)=π‘Šπ›Ύ(0)πœ‘2(πœ“(𝑦,𝑝))𝐽(πœ“(𝑦,𝑝))𝑑𝑦𝑦4.(3.34)
Passing to spherical coordinates by 𝑦=π‘Ÿπœˆ we get 𝐼1ξ€œ(𝑝)=𝛾0ξ‚΅ξ€œΞ©2πœ‘2ξ‚Άπ‘Ÿ(πœ“(π‘Ÿπœˆ,𝑝))𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆβˆ’3π‘‘π‘Ÿ.(3.35) Expanding the function πœ‘(πœ“(π‘Ÿπœˆ,𝑝)) to the Taylor series at π‘Ÿ=0 we obtain ξ€·π‘žπœ‘(πœ“(π‘Ÿπœˆ,𝑝))=πœ‘0ξ€Έ+(𝑝)2𝑖=1πœ•πœ‘πœ•πœ“(𝑖)ξ€·π‘ž0(𝑝)2𝑗=1πœ•πœ“(𝑖)πœ•π‘¦π‘—(0,𝑝)πœˆπ‘—ξƒͺπ‘Ÿ+𝑔(π‘Ÿ,𝜈)π‘Ÿ2,𝑦𝑗=π‘Ÿπœˆπ‘—,(3.36) where 𝑔(β‹…,𝜈) is continuous in π‘Šπ›Ύ(0) and 𝜈21+𝜈22=1. By condition of part (i) of this lemma and from equality (3.36) it follows that (3.35) has the following form: 𝐼1ξ€œ(𝑝)=𝛾0ξ€œπΊ(𝑝,π‘Ÿ)π‘‘π‘Ÿ,𝐺(𝑝,π‘Ÿ)=π‘ŸΞ©2𝑔2(π‘Ÿ,𝜈)𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆ.(3.37)
Since the function 𝐺(𝑝,β‹…) is continuous in [0,𝛾], we have 𝐼1(𝑝)<∞. Taking into account ‖𝑓‖2𝐿2(𝕋2)=𝐼(𝑝), from (3.31) we get that π‘“βˆˆπΏ2(𝕋2).
Now we show that if the conditions of part (i) of Lemma 3.7 are not satisfied, that is, πœ‘(π‘ž0(𝑝))β‰ 0 or βˆ‡πœ‘(π‘ž0(𝑝))β‰ 0, then the function defined by (3.28) does not belong to 𝐿2(𝕋2).
Let πœ‘(π‘ž0(𝑝))=0 and βˆ‡πœ‘(π‘ž0(𝑝))β‰ 0. We will show that 𝐢(𝜈)=2𝑖=1πœ•πœ‘πœ•πœ“(𝑖)ξ€·π‘ž0(𝑝)2𝑗=1πœ•πœ“(𝑖)πœ•π‘¦π‘—(0,𝑝)πœˆπ‘—ξƒͺβ‰ 0,𝜈∈Ω2.(3.38) Assume the converse, let 2𝑖=1𝑐𝑖2𝑗=1π‘’π‘–π‘—πœˆπ‘—=22𝑗=1𝑖=1π‘π‘–π‘’π‘–π‘—β‹…πœˆπ‘—=0,(3.39) where 𝑐𝑖=(πœ•πœ‘/πœ•πœ“(𝑖))(π‘ž0(𝑝)) and 𝑒𝑖𝑗=(πœ•πœ“(𝑖)/πœ•π‘¦π‘—)(0,𝑝),𝑖,𝑗=1,2. Since the function πœˆπ‘—,𝑗=1,2 are linearly independent, the last equality holds if and only if, when 2𝑖=1𝑐𝑖𝑒𝑖𝑗=0,𝑗=1,2.(3.40) Observe that det(𝑒𝑖𝑗)2𝑖,𝑗=1=𝐽(π‘ž0(𝑝))β‰ 0. Consequently, the equalities (3.40) hold if and only if, when 𝑐1=𝑐2=0. This contradicts the fact that βˆ‡πœ‘(π‘ž0(𝑝))β‰ 0. Thus, 𝐢(𝜈)β‰ 0. Hence the equality (3.35) has the form 𝐼1ξ€œ(𝑝)=𝛾0𝐺(𝑝,π‘Ÿ)π‘‘π‘Ÿ,𝐺(𝑝,π‘Ÿ)=π‘Ÿβˆ’1ξ€œΞ©2̃𝑔2(π‘Ÿ,𝜈)𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆ,̃𝑔(π‘Ÿ,𝜈)=𝐢(𝜈)+𝑔(π‘Ÿ,𝜈)π‘Ÿ.(3.41) Since ξ€œπ›Ύ0π‘Ÿβˆ’1π‘‘π‘Ÿ=∞,limπ‘Ÿβ†’0𝐺(𝑝,π‘Ÿ)π‘Ÿβˆ’1ξ€·π‘ž=𝐽0ξ€Έξ€œ(𝑝)Ξ©2𝐢2(𝜈)π‘‘πœˆ>0,(3.42) by the theorem on comparison of improper integrals, we get that 𝐼1(𝑝)=∞ and thereforeπ‘“βˆ‰πΏ2(𝕋2).
In case of πœ‘(π‘ž0(𝑝))β‰ 0 the relation π‘“βˆ‰πΏ2(𝕋2) can be proven analogously.
(ii) Let 𝐹(𝑠)=|πœ‘(𝑠)|,π‘˜=1. Then from (3.33) we get 𝐼1ξ€œ(𝑝)=π‘Šπ›Ύ(0)||||πœ‘(πœ“(𝑦,𝑝))𝐽(πœ“(𝑦,𝑝))𝑑𝑦𝑦2.(3.43) Passing to spherical coordinates by 𝑦=π‘Ÿπœˆ we obtain 𝐼1ξ€œ(𝑝)=𝛾0ξ‚Έξ€œΞ©2||||ξ‚Ήπ‘Ÿπœ‘(πœ“(π‘Ÿπœˆ,𝑝))𝐽(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆβˆ’1π‘‘π‘Ÿ.(3.44) By the condition of part (ii) of Lemma 3.7 and from (3.36) we get 𝐼1ξ€œ(𝑝)=𝛾0ξ‚Έξ€œΞ©2||||𝐢(𝜈)+𝑔(π‘Ÿ,𝜈)π‘Ÿπ½(πœ“(π‘Ÿπœˆ,𝑝))π‘‘πœˆπ‘‘π‘Ÿ.(3.45) Since the function under the integral sign is continuous in [0,𝛾], it follows that 𝐼1(𝑝)<∞. Thus 𝐼(𝑝)<∞. Taking into account ‖𝑓‖𝐿1(𝕋2)=𝐼(𝑝) we obtain π‘“βˆˆπΏ1(𝕋2). Consequently, from part (i) of Lemma 3.7 it follows that π‘“βˆˆπΏ1(𝕋2)⧡𝐿2(𝕋2).

Lemma 3.8. Let the point 𝑠=π‘ž0(𝑝),π‘βˆˆπ‘ˆπ›Ώ(0) a unique non degenerated minimum of the function 𝑀𝑝(𝑠), and πœ‘(π‘ž0(𝑝))=0. Then for any πœ‡>0 the equation π»πœ‡(𝑝)𝑓=π‘š(𝑝)𝑓.(3.46) has a nonzero solution if and only if ξ€œΞ”(πœ‡,𝑝;π‘š(𝑝))=1βˆ’πœ‡π•‹2πœ‘2(π‘ž)π‘‘π‘žπ‘€π‘(π‘ž)βˆ’π‘š(𝑝)=0.(3.47) In this case the nonzero function π‘“πœ‡,𝑝(β‹…)=πΆπœ‡πœ‘(β‹…)𝑀𝑝(β‹…)βˆ’π‘š(𝑝)∈𝐿1𝕋2ξ€Έ,(3.48) is a solution of (3.46), where 𝐢≠0 a normalizing constant.

This lemma can be proved as Lemma 3.1 taking into account part (ii) of Lemma 3.7.

Now we prove the main results.

Proof of Theorem 2.3. (i) Observe that Ξ”(πœ‡,𝑝;β‹…) is continuous and monotonously decreasing in (βˆ’βˆž,π‘š(𝑝)). Moreover, limπ‘§β†’βˆ’βˆžΞ”(πœ‡,𝑝;𝑧)=1.(3.49)
By definition, if πœ‘(π‘ž0(𝑝))β‰ 0, then πœ‡(𝑝)=0. So, for πœ‡>πœ‡(𝑝)=0, Lemma 3.6 gives that limπ‘§β†’π‘š(𝑝)βˆ’0Ξ”(πœ‡,𝑝;𝑧)=βˆ’βˆž.(3.50)
Analogously, if πœ‘(π‘ž0(𝑝))=0, then for πœ‡>πœ‡(𝑝)>0 the inequality limπ‘§β†’π‘š(𝑝)βˆ’0πœ‡Ξ”(πœ‡,𝑝;𝑧)=1βˆ’πœ‡(𝑝)<0,(3.51) holds.
The continuity of function Ξ”(πœ‡,𝑝;β‹…) in (βˆ’βˆž,π‘š(𝑝)) yields that the equation Ξ”(πœ‡,𝑝;𝑧)=0 has a unique solution 𝑧=𝐸(πœ‡,𝑝)<π‘š(𝑝) and hence Lemma 3.1 yields that the operator π»πœ‡(𝑝),β€‰π‘βˆˆπ‘ˆπ›Ώ(0) has a unique eigenvalue 𝐸(πœ‡,𝑝).
Since for any π‘βˆˆπ‘ˆπ›Ώ(0) and πœ‡βˆˆ(πœ‡(𝑝),+∞) the number 𝑧=𝐸(πœ‡,𝑝) is a solution of the equation Ξ”(πœ‡,𝑝;𝑧)=0 and the function Ξ”(πœ‡,β‹…;𝑧) (resp., Ξ”(β‹…,𝑝;𝑧)) is real analytic in π‘ˆπ›Ώ(0) (resp., (πœ‡(𝑝),+∞)), the implicit function theorem implies that 𝐸(πœ‡,β‹…) (resp., 𝐸(β‹…,𝑝)) is real analytic in π‘ˆπ›Ώ(0) (resp., (πœ‡(𝑝),+∞)).
Note that the function Ξ”(β‹…,𝑝;𝑧) monotonously decreases in (πœ‡(𝑝),∞) and hence for any πœ‡1>πœ‡2>πœ‡(𝑝) the eigenvalues 𝐸(πœ‡1,𝑝) and 𝐸(πœ‡2,𝑝) satisfy the relations: ξ€·πœ‡0=Ξ”1ξ€·πœ‡,𝑝;𝐸1ξ€·πœ‡,𝑝=Ξ”2ξ€·πœ‡,𝑝;𝐸2ξ€·πœ‡,𝑝>Ξ”1ξ€·πœ‡,𝑝;𝐸2,𝑝.(3.52)
Using the monotonicity of the function Ξ”(πœ‡1,𝑝;β‹…) in (βˆ’βˆž,π‘š(𝑝)) we obtain that 𝐸(πœ‡1,𝑝)<𝐸(πœ‡2,𝑝), that is, 𝐸(β‹…,𝑝) is monotonously decreases in (πœ‡(𝑝),∞).
Lemma 3.1 implies that if for any πœ‡βˆˆ(πœ‡(𝑝),+∞) and π‘βˆˆπ‘ˆπ›Ώ(0) the number 𝐸(πœ‡,𝑝) is an eigenvalue of the operator π»πœ‡(𝑝),β€‰π‘βˆˆπ‘ˆπ›Ώ(0), then the function Ξ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝))=πΆπœ‡πœ‘(β‹…)𝑀𝑝(β‹…)βˆ’πΈ(πœ‡,𝑝)∈𝐿2𝕋2ξ€Έ,(3.53) is a solution of the equation π»πœ‡(𝑝)Ξ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝))=𝐸(πœ‡,𝑝)Ξ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝)),(3.54) where 𝐢≠0 is a a normalizing constant.
Analyticity of the function Ξ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝)) follows from the analyticity of the functions πœ‘(β‹…) and [𝑀𝑝(β‹…)βˆ’πΈ(πœ‡,𝑝)]βˆ’1 in 𝕋2 and the representation (3.53). The functions 𝐸(πœ‡,β‹…) and 𝑀(β‹…,π‘ž) are analytic in π‘ˆπ›Ώ(0) and for any π‘žβˆˆπ•‹π‘‘ the inequality 𝑀𝑝(π‘ž)βˆ’πΈ(πœ‡,𝑝)>0 holds, therefore representation (3.53) yields that the mapping 𝑝↦Ψ(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝)) is analytic in π‘ˆπ›Ώ(0). Analogously the analyticity of the function 𝐸(β‹…,𝑝)) implies that the mapping πœ‡β†¦Ξ¨(πœ‡;𝑝,β‹…,𝐸(πœ‡,𝑝)) is analytic in (πœ‡(𝑝),+∞).
(ii) Let πœ‘(π‘ž0(𝑝))=0 and 0<πœ‡<πœ‡(𝑝). Since limπ‘§β†’π‘š(𝑝)βˆ’0πœ‡Ξ”(πœ‡,𝑝;𝑧)=Ξ”(πœ‡,𝑝;π‘š(𝑝))=1βˆ’πœ‡(𝑝)>0,(3.55) we have Ξ”(πœ‡,𝑝;𝑧)>0,π‘§βˆˆ(βˆ’βˆž,π‘š(𝑝)] and Lemma 3.1 yields that the operator π»πœ‡(𝑝),β€‰π‘βˆˆπ‘ˆπ›Ώ(0) does not have any eigenvalue in (βˆ’βˆž,π‘š(𝑝)].
The statements (iii) and (iv) of Theorem 2.3 follows from Lemmas 3.7 and 3.8.

Acknowledgments

The first author wishes to thank the Mathematics Department of Faculty of Computer Science and Mathematics, MARA University of Technology (Malaysia), where the paper has been finished, for the invitation and hospitality. Authors are grateful to the referee(s) for useful remarks.