Mean--field stability for the junction of quasi 1D systems with Coulomb interactions
MMean–field stability for the junction of quasi 1D systems withCoulomb interactions
Ling-Ling CAO
Universit´e Paris-Est, CERMICS (ENPC), F-77455 Marne-la-Vall´ee
April 25, 2019
Abstract
Junctions appear naturally when one studies surface states or transport properties of quasi one dimen-sional materials such as carbon nanotubes, polymers and quantum wires. These materials can be seen as1D systems embedded in the 3D space. In this article, we first establish a mean–field description of reducedHartree–Fock type for a 1D periodic system in the 3D space (a quasi 1D system), the unit cell of which is un-bounded. With mild summability condition, we next show that a quasi 1D system in its ground state can bedescribed by a mean–field Hamiltonian. We also prove that the Fermi level of this system is always negative.A junction system is described by two different infinitely extended quasi 1D systems occupying separatelyhalf spaces in 3D, where Coulombic electron-electron interactions are taken into account and without anyassumption on the commensurability of the periods. We prove the existence of the ground state for a junctionsystem, the ground state is a spectral projector of a mean–field Hamiltonian, and the ground state density isunique.
Contents a r X i v : . [ m a t h - ph ] A p r .8 Proof of Proposition 3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304.9 Proof of Lemma 4.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334.10 Proof of Proposition 3.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.11 Proof of Lemma 3.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384.12 Proof of Proposition 3.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384.13 Proof of Theorem 3.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Atomic junctions of quasi 1D systems appear for instance when studying the surface states of one-dimensional(1D) crystals [1, 67], quantum thermal transport in nanostructures [71], and p-n junctions [4, 60] which arethe foundation of the modern semiconductor electronic devices. Besides, electronic transport in carbon nan-otubes [50] and in molecular wires [59], which recently attracted a lot of interest, is often modeled by the junc-tion of two semi-infinite systems with different chemical potentials. In recent years, studies of various quantumHall effects and topological insulators focussed attention on 2D materials, see [37] and references therein. These2D materials often possess periodicity in one dimension and can therefore be reduced to quasi 1D materials bymomentum representation in the periodic direction [38]. Furthermore, when studying edge states properties(see [38, 3] and references therein) of 2D materials, they can be seen as a junction with the vacuum.Real world materials are often described by periodic [20, 15] or ergodically periodic [16] systems in mathe-matical modeling. In this article we consider a junction of two different quasi 1D periodic systems without anyassumption on the commensurability of the periods. Generally speaking, there are two regimes for the junctionof two different periodic systems: when the chemical potentials of the underlying periodic systems are separatedby some occupied bands (non-equilibrium regime, see Fig. 3), and when the chemical potential are in a commonspectral gap (equilibrium regime, see Fig. 4). The non-equilibrium regime models a persistent (non-perturbative)current in the junction system [8, 9, 10, 24], while the equilibrium regime can model either the ground state ofthe junction material or the presence of perturbative current in the linear response regime [23]. In this articlewe consider the equilibrium regime, and only briefly comment on the non-equilibrium regime in Section 3.2, asthe study of this situation requires different techniques.The most prominent feature of quasi 1D materials is the presence of strong electron-electron interactionsdue to low screening effect [11, 12] as electrons interact through the 3D space. For finite systems, one can usea N -body Schr¨odinger model to describe the electron-electron interactions. Nevertheless, this is impossible forinfinite systems. Mean–field theory is a good candidate for infinite systems: it consists in replacing the N -bodyinteractions by a -body interaction with an effective average field, leading to a quasi-particle description of thesystem. However, mean–field models are rarely available for quasi 1D periodic systems, as periodic systems areoften considered either in the 3D space (see [57, 21] for Thomas–Fermi type models and [20] for Hartree–Focktype models), or strictly in a 1D geometry (see [6] for Thomas–Fermi type models). To our knowledge, the workon Thomas–Fermi type model [5] for polymers is the only literature available for a 1D periodic system withinteractions through the 3D space. Furthermore, non-periodic infinite systems are difficult to handle mathemat-ically, as they do not possess any symmetry, hence the usual Bloch decomposition of periodic systems [62, 20]is not applicable, and the definition of the ground state energy needs to be examined [7].In this article, we establish a mean-field model to describe the junction of two different quasi 1D periodicsystems (see Fig. 2) in the 3D space with Coulomb interactions, under the framework of the reduced Hartree–Fock [68] (rHF) description. Remark that the rHF model is strictly convex in the density, and can be seen as agood approximation of Kohn-Sham LDA model [46, 2, 55], which is widely used in condensed matter physics.This non-linear model can be employed to describe the junction of two nanotubes, or a more realistic model ofjunction of two quasi 1D crystals for electronic structure calculations. It can be further explored to study thelinear response with respect to different Fermi levels between two semi-infinite chains: recall that the famousLandauer–B¨uttiker [51, 13] formalism for electronic (thermal) transport which is based on the lead-device-leaddescription, can be seen as the junction of two different quasi 1D systems (leads) with different chemical (ther-mal) potentials, and the device as a perturbation of this junction. Remark also that p-n junctions of carbonnanotubes without external battery [53, 52] correspond to the equilibrium regime, and can thus be described by2he model we consider. Futhermore, our model can also be easily adapted to describe 1D dislocation problems inthe 3D space, while the linear 1D dislocation problems have been studied in [47, 48] and some generalizationshave been provided for higher dimensional systems [26, 40, 41]. The organization of this article and the main results are as follows: in Section 2 we consider a quasi 1D peri-odic system, which is described by nuclei arranged periodically alongside the x -axis (see Fig. 1) with electronsoccupying the 3D space, as it is a building block for the junction system. We define a periodic rHF energy func- xy z Figure 1: An example of nuclei configuration of a quasi 1D periodic system.tional (2.15) by taking into account the real Coulomb interactions in the 3D space. In Theorem 2.6 we showthat this rHF functional admits minimizers, and that the ground state electronic density is unique. Remark thatthis is different from [20, 15] as the system is periodic only in the x -direction, the unit cell Γ being unboundedso that additional compactness proofs are needed when dealing with the ground state problem. With a mildsummability condition (2.18) on the unique density of minimizers, we are able to obtain a mean–field Hamil-tonian H per “ ´ ∆ ` V per to describe a quasi 1D periodic system, where V per is the mean–field potentialthat tends to in the r : “ p y, z q direction. In Theorem 2.7 we prove that the Fermi level (cid:15) F of the quasi 1Dsystem, which represents the highest energy attainable by electrons under this quasi-particle description, isalways negative. We also prove that the unique minimizer is a spectral projector of the mean–field Hamilto-nian γ per “ p´8 ,(cid:15) F s p H per q .In Section 3, under certain symmetry assumptions on the nuclear densities µ per ,L and µ per ,R of two different xy z a R a L Figure 2: Nuclei configuration of the junction system with period a L on p´8 , s ˆ R and a R on p , `8q ˆ R .quasi 1D periodic systems, the junction system is described by considering the following nuclear configuration(see Fig. 2): µ J : “ x ď ¨ µ per ,L ` x ą ¨ µ per ,R ` v, where v describes how the junction is initiated. We aim at establishing a quasi-particle description of thisinfinitely extended junction system with Coulomb interactions and show the existence of ground state. Aswe do not assume any commensurability of periods of the two quasi 1D systems, the junction system doesnot possess any translation-invariant symmetry. The main idea is to establish a well-suited reference systembased on the linear combination of periodic systems, and use perturbative techniques which have been widelyused for mean-field type models [36, 33, 35, 15, 30] to justify the construction. More precisely, we define areference Hamiltonian H χ “ χ H per ,L ` p ´ χ q H per ,R with χ a smooth cut-off function approximating x ď , where H per ,L and H per ,R are the mean–field Hamiltonians of the quasi 1D periodic systems. Denoteby σ p A q “ σ disc p A q Ť σ ess p A q the spectrum of A , where σ disc p A q (resp. σ ess p A q ) denotes the discrete (resp.essential) spectrum of an operator A . Denote also by σ ac p A q the purely absolutely continuous spectrum of A .We first show in Proposition 3.1 that σ ess p H χ q “ σ ess p H per ,L q ď σ ess p H per ,R q , σ ess p H χ q č p´8 , s Ď σ ac p H χ q . χ ,ii) the linear junction preserves the scattering channels of the underlying systems, since the purely absolutelycontinuous spectrum of Hamiltonian has not been modified, hence the linear junction can be used to study theelectronic conductance with the Landauer-B¨uttiker formalism (see for example [8, 9, 10] as well as the discussionfollowing Proposition 3.1).After introducing a reference state γ χ : “ p´8 ,(cid:15) F q p H χ q , we show in Proposition 3.2 that the electronic den-sity ρ χ is close to the linear combination of the underlying periodic electronic densities, and that the differencewith these reference densities decays exponentially fast. The quasi-particle description of the nonlinear junctionstate can be constructed by considering γ J “ γ χ ` Q χ , where Q χ is a trial density matrix which encodes the nonlinear effects of the junction system. Following theidea developed in [15], we associate Q χ with some minimization problem in Proposition 3.5, and denote by Q χ a minimizer. We prove in Theorem 3.6 that ρ γ J “ ρ χ ` ρ Q χ is independent of χ .This implies that the ground state of the junction system with Coulomb interactions exists and its density isindependent of the choice of the reference state, see Corollary 3.6.1. In this section, we give a mathematical description of a quasi 1D periodic system in the framework of the reducedHartree-Fock (rHF) approach. In Section 2.1, we introduce some mathematical preliminaries. In Section 2.2 weconstruct a periodic rHF energy functional for a quasi 1D system.Let us first introduce some notation. Unless otherwise specified, the functions on R d considered in thisarticle are complex-valued. Elements of R are denoted by x “ p x, r q , where x P R . For a given separableHilbert space H , we denote by L p H q the space of bounded linear operators acting on H , by S p H q the space ofbounded self-adjoint operators acting on H , and by S p p H q the Schatten class of operators acting on H . For ď p ă 8 , a compact operator A belongs to S p p H q if and only if (cid:107) A (cid:107) S p : “ p Tr p| A | p qq { p ă 8 . Operatorsin S p H q and S p H q are respectively called trace-class and Hilbert–Schmidt. If A P S p L p R d qq , there exists aunique function ρ A P L p R d q such that @ φ P L p R d q , Tr p Aφ q “ ż R d ρ A φ. The function ρ A is called the density of the operator A . If the integral kernel A p r , r q of A is continuous on R d ˆ R d , then ρ A p r q “ A p r , r q for all r P R d . This relation still stands in some weaker sense for a generictrace-class operator.An operator A P L p L p R d qq is called locally trace-class if the operator (cid:37)A(cid:37) is trace-class for any (cid:37) P C c p R d q . The density of a locally trace-class operator A P L p L p R d qq is the unique function ρ A P L p R d q such that @ φ P C c p R d q , Tr p Aφ q “ ż R d ρ A φ. Let S p R d q be the Schwartz space of rapidly decreasing functions on R d , and S p R d q the space of tempereddistributions on R d . We denote by p φ (resp. q φ ) the Fourier transform (resp. inverse Fourier transform) on S p R d q ,with the following normalization: @ φ P L p R d q , p φ p ζ q : “ p π q d { ż R d φ p x q e ´ i ζx dx, q φ p x q : “ p π q d { ż R d φ p ζ q e i ζx dζ. The normalization ensures that the Fourier transform defines a unitary operator on L p R d q . We first introduce a decomposition of the operator which is Z -translation invariant in the x -direction based onthe partial Bloch transform. In order to describe the 1D periodic system in the 3D space, we next introduce amixed Fourier transform. We also introduce a Green’s function which is periodic only in the x -direction. Finallywe introduce the kinetic energy space of density matrices and Coulomb interactions for quasi 1D systems.4 loch transform in the x -direction. For k P Z , we denote by τ xk the translation operator in the x -directionacting on L p R q : @ u P L p R q , p τ xk u qp¨ , r q “ u p¨ ´ k, r q for a.a. r P R . An operator A on L p R q is called Z -translation invariant in the x -direction if it commutes with τ xk for all k P Z . In order to decompose operators which are Z -translation invariant in the x -direction, let us without lossof generality choose a unit cell Γ : “ r´ { , { q ˆ R , and introduce the L p spaces and H spaces of functionswhich are -periodic in the x -direction: for ď p ď `8 , L p per ,x p Γ q : “ ! u P L p loc p R q ˇˇˇ (cid:107) u (cid:107) L p p Γ q ă `8 , τ xk u “ u, @ k P Z ) ,H ,x p Γ q : “ ! u P L ,x p Γ q ˇˇˇ ∇ u P ` L ,x p Γ q ˘ ) . Let us also introduce the following constant fiber direct integral of Hilbert spaces [62]: L p Γ ˚ ; L ,x p Γ qq : “ ż À Γ ˚ L ,x p Γ q dξ π , with the base Γ ˚ : “ r´ π, π qˆt u ” r´ π, π q . The partial Bloch transform B is a unitary operator from L p R q to L p Γ ˚ ; L ,x p Γ qq , defined on the dense subspace of C c p R q of L p R q : @p x, r q P Γ , @ ξ P Γ ˚ , p B φ q ξ p x, r q : “ ÿ k P Z e ´ i p x ` k q ξ φ p x ` k, r q . Its inverse is given, for f ‚ “ p f ξ q ξ P Γ ˚ by @ k P Z , for a.a. p x, r q P Γ , ` B ´ f ‚ ˘ p x ` k, r q : “ ż Γ ˚ e i p k ` x q ξ f ξ p x, r q dξ π . The partial Bloch transform has the property that any operator A on L p R q which commutes with τ xk for k P Z is decomposed by B : for any A P L p L p R qq such that τ xk A “ Aτ xk , there exists A ‚ P L p Γ ˚ ; L p L ,x p Γ qqq such that for all u P L p R q , p B p Au qq ξ “ A ξ p B u q ξ for a.a. ξ P Γ ˚ . We hence use the following notation for the decomposition of an operator A which is Z -translation invariantin the x -direction: A “ B ´ ˆż ‘ Γ ˚ A ξ dξ π ˙ B . In addition, (cid:107) A (cid:107) L p L p R qq “ (cid:13)(cid:13) (cid:107) A ‚ (cid:107) L p L ,x p Γ qq (cid:13)(cid:13) L p Γ ˚ q . In particular, if A is positive and locally trace-class, thenfor almost all ξ P Γ ˚ , A ξ is locally trace-class. The densities of these operators are related by the formula ρ A p x q “ π ż Γ ˚ ρ A ξ p x q dξ. (2.1)If A is a (not necessarily bounded) self-adjoint operator such that τ xk p A ` i q ´ “ p A ` i q ´ τ xk for all k P Z ,then A is decomposed by U (see [62, Theorems XIII.84 and XIII.85]). In particular, denoting by ∆ the Laplaceoperator acting on L p R q , the kinetic energy operator ´ ∆ on L p R q is decomposed by B as follows: ´
12 ∆ “ B ´ ˆż Γ ˚ ´
12 ∆ ξ dξ π ˙ B , ´ ∆ ξ “ p´ i ∇ ξ q “ p i B x ´ ξ q ´ ∆ r , (2.2)where ∆ r is the Laplace operator acting on L p R q . Mixed Fourier transform.
The mixed Fourier transform consists of a Fourier series transform in the x -direction and an integral Fourier transform in the r -direction. Denote by S per ,x p Γ q the space of functionsthat are C on R and Γ -periodic, decaying faster than any power of | r | when | r | tends to infinity, as well astheir derivatives. Denote by S per ,x p Γ q the dual space of S per ,x p Γ q . The mixed Fourier transform is the unitarytransform F : L ,x p Γ q Ñ (cid:96) ` Z , L p R q ˘ defined on the dense subspace S per ,x p Γ q of L ,x p Γ q by: @ φ P S per ,x p Γ q , @p n, k q P Z ˆ R , F φ p n, k q : “ π ż Γ φ p x, r q e ´ i p πnx ` k ¨ r q dx d r . (2.3)5ts inverse is given by, @ p ψ n p k qq n P Z , k P R P (cid:96) ` Z ; L p R q ˘ , F ´ ψ p x, r q : “ π ÿ n P Z ż R ψ n p k q e i p πnx ` k ¨ r q d k . Note that F can be extended from S per ,x p Γ q to S p R q . One can easily see that F is an isometry from L ,x p Γ q to (cid:96) ` Z , L p R q ˘ in the following sense: @ f, g P L ,x p Γ q , ż Γ f p x, r q g p x, r q dx d r “ ÿ n P Z ż R F f p n, k q F g p n, k q d k . (2.4)Moreover, it is easy to verify that for f, g P L ,x p Γ q , F p f ‹ Γ g q “ π p F f q p F g q , (2.5)where p f ‹ Γ g q p x q : “ ş Γ f p x ´ x q g p x q d x . As an application of the mixed Fourier transform, let us introducea Kato–Seiler–Simon type inequality [65] for the operator ´ i ∇ ξ “ p´ i B x ` ξ, ´ i B r q for all ξ P Γ ˚ , which willbe repeatedly used in the proofs. Lemma 2.1.
Fix ξ P Γ ˚ . Let ď p ď `8 and f, g P L p per ,x p Γ q . Then (cid:107) f p´ i ∇ ξ q g (cid:107) S p p L ,x p Γ q q ď p π q ´ { p (cid:107) g (cid:107) L p per ,x p Γ q ˜ ÿ n P Z (cid:107) f p πn ` ξ, ¨q (cid:107) pL p p R q ¸ { p , (2.6) for any ď p ă 8 and (cid:107) f p´ i ∇ ξ q g (cid:107) ď (cid:107) g (cid:107) L per ,x p Γ q sup n P Z (cid:107) f p πn ` ξ, ¨q (cid:107) L p R q , when p “ `8 . The proof of this lemma can be read in Section 4.1.
Periodic Green’s function.
We introduce a 3D Green’s function which is -periodic in the x -direction in thesame spirit as in [5, 57]. Definition 2.2 (Periodic Green’s function) . For p x, r q P R , the periodic Green’s function is defined as G p x, r q “ ´ p| r |q ` r G p x, r q , r G p x, r q : “ ÿ n ě K p πn | r |q cos p πnx q , (2.7)where K p α q : “ ş `8 e ´ α cosh p t q dt is the modified Bessel function of the second kind.The following lemma summarizes the properties of the periodic Green’s function defined in (2.7). Lemma 2.3.
1. The Green’s function G p x, r q defined in (2.7) satisfies the following Poisson’s equation: ´ ∆ G p x, r q “ π ÿ n P Z δ p x, r q“p n, q P S p R q , where δ a P S p R d q is the Dirac distribution at a P R d . Moreover G P S per ,x p Γ q and F p G qp n, k q “ π n ` | k | P S p R q . (2.8)
2. The function r G defined in (2.7) belongs to L p per ,x p Γ q for ď p ă and satisfies ş Γ r G ” . Moreover, thereexist positive constants d and d such that | r G p¨ , r q| ď d ´ π | r | ? | r | when | r | Ñ `8 , and | r G p¨ , r q| ď d | r | when | r | Ñ , uniformly with respect to x . Finally, the function r G p x, r q can also be written as r G p x, r q “ ÿ n P Z ˜ a p x ´ n q ` | r | ´ ż { ´ { a p x ´ y ´ n q ` | r | dy ¸ . (2.9)The proof of this lemma can be read in Section 4.2. 6 ne-body density matrices and kinetic energy space. In mean-field models, electronic states can be de-scribed by one-body density matrices (see e.g. [15, 30]). Recall that for a finite system with N electrons, adensity matrix is a trace-class self-adjoint operator γ P S p L p R qq X S p L p R qq satisfying the Pauli princi-ple ď γ ď and the normalization condition Tr p γ q “ ş R ρ γ “ N . The kinetic energy of γ is given by Tr p´ ∆ γ q : “ Tr p| ∇ | γ | ∇ |q (see [20, 15, 17]).Consider a 1D periodic system in the 3D space, where atoms are arranged periodically in the x -directionwith unit cell Γ and first Brillouin zone Γ ˚ . Since the rHF model is strictly convex in the density [68], we donot expect any spontaneous symmetry breaking. Therefore the electronic state of this quasi 1D system will bedescribed by a one-body density matrix which commutes with the translations t τ xk u k P Z , hence is decomposedby the partial Bloch transform B . In view of the decomposition (2.2), we define the following admissible set ofone-body density matrices, which guarantees that the number of electrons per unit cell and the kinetic energyper unit cell are finite: P per ,x : “ " γ P S p L p R qq ˇˇˇˇ ď γ ď , @ k P Z , τ xk γ “ γτ xk , ż Γ ˚ Tr L ,x `a ´ ∆ ξ γ ξ a ´ ∆ ξ ˘ dξ ă 8 * , (2.10)where γ “ B ´ ˆż Γ ˚ γ ξ dξ π ˙ B . (2.11)For any γ P P per ,x , it is easy to see that ρ γ P L ,x p Γ q . Moreover, a Hoffmann-Ostenhof type inequality [43]can also be deduced from [20, Equation (4.42)]: ż Γ ˇˇ ∇ ? ρ γ ˇˇ ď ż Γ ˚ Tr L ,x p´ ∆ ξ γ ξ q dξ π . (2.12)Therefore ? ρ γ is in H ,x p Γ q hence in L ,x p Γ q by Sobolev embeddings, so that ρ γ P L p per ,x p Γ q for ď p ď by an interpolation argument. Coulomb interactions.
Recall that the Coulomb interaction energy of charge densities f and g belonging to L { p R q can be written in real and reciprocal space as: D p f, g q : “ ż R ż R f p x q g p x q| x ´ x | d x d x “ π ż R p f p k q p g p k q| k | . In order to describe Coulomb interactions in the reciprocal space for a quasi 1D periodic system, we gatherthe results obtained in (2.4), (2.5) and (2.8), and define the Coulomb interaction energy per unit cell for chargedensities f, g belonging to S per ,x p Γ q as: D Γ p f, g q : “ π ÿ n P Z ż R F p f qp n, k q F p g qp n, k q| k | ` π n d k . (2.13)It is easy to see that D Γ p¨ , ¨q is a positive definite bilinear form on S per ,x p Γ q . Let us introduce the Coulombspace for the 1D periodic system in the 3D space as C Γ : “ (cid:32) f P S per ,x p Γ q ˇˇ @ n P Z , F p f qp n, ¨q P L p R q , D Γ p f, f q ă `8 ( , (2.14)which is a Hilbert space endowed with the inner product D Γ p¨ , ¨q . Remark . Remark that charge densities in C Γ are neutral in some weak sense. Indeed, for f P C Γ Ş L ,x p Γ q ,the condition ş R | F p f qp , k q| | k | d k ă `8 implies that F p f qp , q “ ş Γ f p x, r q dx d r “ . Based on the kinetic energy space and Coulomb interactions defined in the previous section, we construct herea rHF energy functional for a quasi 1D periodic system which is -periodic only in the x -direction. We showthat its ground state is given by the solution of some minimization problem. Denote by Z P N ˚ the total nuclear7harge in each unit cell. For the sake of technical reasons we model the nuclear density of a quasi 1D system bya smooth function (smeared nuclei) which is -periodic in the x -direction µ per p x, r q “ ÿ n P Z Z m p x ´ n, r q , where m p x, r q is a non-negative C c p Γ q function such that ş R m “ . In particular ş Γ µ per “ Z .For any trial density matrix γ which commutes with the translations τ xk in the x -direction, the periodic rHFenergy functional for a quasi 1D system associated with the nuclear density µ per is defined as: E per ,x p γ q : “ π ż Γ ˚ Tr L ,x p Γ q ˆ ´
12 ∆ ξ γ ξ ˙ dξ ` D Γ p ρ γ ´ µ per , ρ γ ´ µ per q . (2.15)Let us introduce the following set of admissible density matrices for this rHF energy functional, which guaranteesthat the kinetic energy and Coulomb interaction energy per-unit cell are finite: F Γ : “ t γ P P per ,x | ρ γ ´ µ per P C Γ u , where P per ,x is the kinetic energy space defined in (2.10) and C Γ is the Coulomb space defined in (2.14). Lemma 2.5.
The set F Γ is not empty. Moreover, for any γ P F Γ , ż Γ ρ γ “ ż Γ µ per . (2.16)The proof of Lemma 2.5 relies on an explicit construct of an element in F Γ , and can be read in Section 4.3.The periodic rHF ground state energy (per unit cell) of a quasi 1D system can then be written as the followingminimization problem: I per “ inf t E per ,x p γ q ; γ P F Γ u , (2.17)The minimization problem similar to (2.17) under the Thomas-Fermi type models has been studied in [5], wherethe authors proved the uniqueness of the minimizers, and justified the model by a thermodynamic limit argu-ment. For a 3D periodic crystal, the minimization problem (2.17) has been examined in [20], where the authorsshowed the existence of minimizers and the uniqueness of the density of the minimizers. The characterizationof the minimizers is given in [15, Theorem 1]: the minimizer is unique and is a spectral projector satisfying aself-consistent equation. The following theorem provides similar results for a quasi 1D system: we show that theminimizer of (2.17) exists, and that the density of the minimizers is unique. Let us emphasize that the unit cellof a quasi 1D system is an unbounded domain Γ , hence we need to deal with the possible escaping of electronsin the r -direction, a situation which needs not be considered for bounded unit cells as in [20, 15]. Theorem 2.6 (Existence of rHF ground state) . The minimization problem (2.17) admits a minimizer γ per withdensity ρ γ per belonging to L p per ,x p Γ q for ď p ď . Besides, all the minimizers share the same density. The proof of Theorem 2.6 relies on a classical variational argument, and can be read in Section 4.4.In order to treat the junction of quasi 1D systems in Section 3, it is useful to define and study the mean–field potential V per generated by the ground state electronic density ρ γ per and the nuclear density µ per . It isalso critical to obtain some decay estimates of V per in the r -direction. However, for V per satisfying Poisson’sequation ´ ∆ V per “ π p ρ γ per ´ µ per q , the L p integrability of ρ γ per obtained in Theorem 2.6 does not imply thedecay of the mean-field potential V per in the r -direction, given that the Green’s function defined in (2.7) has log -growth in the r -direction. Moreover, the uniform bound given by the energy functional (2.15) does not provideany L p bounds or decay property of V per . In this perspective, we introduce the following assumption on ρ γ per .Remark that this assumption, which we call “summability condition” is common when treating 2D Poisson’sequation [56, Theorem 6.21]. Assumption 1.
The unique ground state density ρ γ per of the problem (2.17) satisfies ż Γ | r | ρ γ per p x, r q dx d r ă `8 . (2.18)8ith this mild summability condition (2.18) on ρ γ per , we prove in Theorem 2.7 that the highest attainableenergy (Fermi level) of electrons for a quasi 1D system in its ground state is always negative. This coincides withthe physical reality: the additional summability condition on the density is sufficient to guarantee that the mean-field potential tends to in the r -direction. If the Fermi level is non-negative, electrons can escape to infinityin the r -direction, decreasing the energy of the system, hence the system is not at ground state. Furthermore,we are able to characterize the unique minimizer as a spectral projector of the mean-field Hamiltonian. Wecomment on Assumption 1 in Remark 2.9. Theorem 2.7 (Properties of the rHF ground state with summability condition on the density) . Assume thatAssumption 1 holds for the unique ground state density ρ γ per of the minimization problem (2.17) .1. (The integrability of mean-field potential.) The mean-field potential V per : “ p ρ γ per ´ µ per q ‹ Γ G belongs to L p per ,x p Γ q for ă p ď `8 . Moreover, V per is continuous and tends to zero in the r -direction.2. (Spectral properties of the mean-field Hamiltonian.) The mean-field Hamiltonian H per “ B ´ ˆż Γ ˚ H per ,ξ dξ π ˙ B “ ´
12 ∆ ` V per , H per ,ξ : “ ´
12 ∆ ξ ` V per , (2.19) is a self-adjoint operator acting on L p R q with domain H ` R ˘ and form domain H ` R ˘ . There exists N H P N ˚ which can be finite or infinite, and a sequence t λ n p ξ qu ξ P Γ ˚ , ď n ď N H such that σ ess p H per ,ξ q “ r , `8q , σ disc p H per ,ξ q “ ď ď n ď N H λ n p ξ q Ă r´ (cid:107) V per (cid:107) L , q . Moreover, the following spectral decomposition holds: σ p H per q “ σ ess p H per q “ ď ξ P Γ ˚ σ p H per ,ξ q , ď ξ P Γ ˚ σ disc p H per ,ξ q Ď σ ac p H per q . (2.20) In particular, r , `8q Ă σ ess p H per q .3. (The Fermi level is always negative.) The energy level counting function F p κ q : κ ÞÑ | Γ ˚ | ż Γ ˚ Tr L ,x p Γ q ` p´8 ,κ s p H per ,ξ q ˘ dξ “ | Γ ˚ | N H ÿ n “ ż Γ ˚ p λ n p ξ q ď κ q dξ is continuous and non-decreasing on p´8 , s . The following inequality always holds: N H “ F p q ě ż Γ µ per , which means that there are always enough negative energy levels for the electrons. Moreover, there exists (cid:15) F ă called Fermi level (chemical potential) such that F p (cid:15) F q “ ş Γ µ per “ Z , which represents the highestattainable energy level by electrons, and can be interpreted as the Lagrange multiplier associated with thecharge neutrality condition (2.16) .4. (The unique minimizer is a spectral projector.) The minimizer of the problem (2.17) is unique and satisfiesthe following self-consistent equation: γ per “ p´8 ,(cid:15) F s p H per q “ B ´ ˆż Γ ˚ γ per ,ξ dξ π ˙ B , γ per ,ξ : “ p´8 ,(cid:15) F s p H per ,ξ q . (2.21) Furthermore, there exist positive constants C (cid:15) F and α (cid:15) F which depend on the Fermi level (cid:15) F , such that ď ρ γ per p x, r q ď C (cid:15) F e ´ α (cid:15)F | r | . (2.22)The proof of Theorem 2.7 can be read in Section 4.6.9 emark . As the unit cell of the 1D system in the 3D space is an unbounded domain Γ , the decomposedmean-field Hamiltonian H per ,ξ does not have a compact resolvent, which is a significant difference compared tothe situation considered in [20, 15]. Remark . Let us comment on Assumption 1. Remark that the exponential decay of the density (2.22) impliesthe summability condition (2.18). However, we were not able to directly prove (2.18). This failure is mainlydue to the lack of a priori summability bounds for the density matrices in F Γ . One might argue that we canadd the condition (2.18) to the definition of F Γ . However, the set F Γ with the condition (2.18) is not closedfor the usual weak- ˚ topology when considering a minimizing sequence of (2.15). Another attempt is to usea Schauder fixed-point algorithm as in [58, 18] to prove that (2.21) admits a solution. The most crucial step isto guarantee that there are enough negative bound states to meet the charge neutrality constraint (2.16). Thenumber of bound states is controlled by the decay rate of potentials. With exponentially decaying densities wecan show that [5, Lemma 2.5] there exists C P R ` such that | V per p¨ , r q| ď C | r | ´ . Nevertheless this conditionis not sufficient to guarantee that the number of bound states is sufficient, as the critical decay rate for numbersof bound states to be finite or infinite is ´| r | ´ [62, Theorem XIII.6]. In other words, we do not have a uniformbound over the Fermi level (cid:15) F at each fixed-point iteration. On the other hand, the summability condition (2.18)is a sufficient but probably not a necessary condition for the negativity of the Fermi level and the characterizationof the minimizers. The main difficult is to control the decay of the mean-field potential V per in the r -direction byjust controlling the nuclear density µ per , given that the Green’s function defined in (2.7) has log -growth in the r -direction. Furthermore, different decay scenarios of V per in the r -direction lead to different characterizationsof the spectrum of the Hamiltonian H per : if V per is bounded from below, and positive with log -growth when | r | Ñ 8 , one can show that the spectrum of H per ,ξ is purely discrete and the spectrum of H per has a bandstructure. The Fermi level of the system could be positive in this case. We are not able to prove the abovestatements without Assumption 1.In order to describe the junction of quasi 1D systems, more specifically to guarantee that the Coulomb energygenerated by the perturbative state is finite, the integrability of the mean-field potential provided in Theorem 2.7is not sufficient in view of Lemma 3.3 below. In order to make use of this result, let us introduce a class of nucleardensities such that the x -averaged density is rotationally invariant in the r -direction: µ per , sym p x, r q “ ÿ n P Z Z m s p x ´ n, r q , where m s p x, r q is a non-negative C c p Γ q function such that ş R m s “ . Moreover, there exists m sym p| r |q P C c p R q such that @ r P R , ż { ´ { m s p x, r q dx ” m sym p| r |q . (2.23) Lemma 2.10.
Suppose that Assumption 1 holds. Under the symmetry condition (2.23) on the nuclear density µ per ,all the results of Theorem 2.7 hold for the minimization problem (2.17) . Besides, the mean–field potential V per belongsto L p per ,x p Γ q for ă p ď `8 . The proof of this lemma can be read in Section 4.7. The nuclei of many actual materials can be modeled witha smear nuclear density satisfying the condition (2.23): for instance nanotubes and polymers with rotationalsymmetry in the r -direction. In this section, we construct a rHF model for the junction of two different quasi 1D periodic systems. The junctionsystem is described by periodic nuclei satisfying the symmetry condition (2.23) with different periodicities andpossibly different charges per unit cell, occupying separately the left and right half spaces ( i.e. , p´8 , s ˆ R and p , `8q ˆ R ), see Fig. 2. We do not assume any commensurability of the different periodicities. Thejunction system is therefore a priori no longer periodic, making it impossible to define the periodic rHF energy.Inspired by perturbative approaches when treating infinitely extended systems [34, 33, 36, 35, 15], the idea is tofind an appropriate reference state which is close enough to the actual one. Section 3.1 gives a mathematicaldescription of the junction system. Section 3.2 is devoted to a rigorous construction of a reference Hamiltonian H χ and a reference one-particle density matrix defined as a spectral projector of H χ . In Section 3.3 we constructa perturbative state, which encodes the non-linear effects due to the electron-electron interaction in the rHF10pproximation, and associate the ground state energy of this perturbative state to some minimization problemin Section 3.4. Consider two quasi 1D periodic systems with periods a L ą and a R ą . The unit cells are respectively denotedby Γ L : “ r´ a L , a L q ˆ R and Γ R : “ r´ a R , a R q ˆ R with their duals Γ ˚ L : “ r´ πa L , πa L q and Γ ˚ R : “ r´ πa R , πa R q .We consider nuclear densities fulfilling the symmetry condition (2.23) and suppose that Assumption 1 holds forthe ground state densities of both quasi 1D periodic systems. More precisely, let m L p x, r q and m R p x, r q benon-negative C c functions with supports respectively in Γ L and Γ R such that ş R m L “ and ş R m R “ .Assume that there exist m sym ,L p| r |q , m sym ,R p| r |q P C c p R q such that @ r P R , ż a L { ´ a L { m L p x, r q dx ” m sym ,L p| r |q , ż a R { ´ a R { m R p x, r q dx ” m sym ,R p| r |q . Denoting by Z L , Z R P N zt u the total charges of the nuclei per unit cells, the smeared periodic nuclear densitiesare respectively defined as µ per ,L p x, r q : “ ÿ n P Z Z L m L p x ´ a L n, r q , µ per ,R p x, r q : “ ÿ n P Z Z R m R p x ´ a R n, r q . (3.1)The periodic Green’s functions with period Γ L and Γ R are separately defined as G a L p x, r q “ a ´ L G ˆ xa L , r ˙ , G a R p x, r q “ a ´ R G ˆ xa R , r ˙ , where G p¨q is the periodic Green’s function defined in (2.7). One can easily verify that ´ ∆ G a L p x, r q “ π ÿ n P Z δ p x, r q“p a L n, q P S p R q , ´ ∆ G a R p x, r q “ π ÿ n P Z δ p x, r q“p a R n, q P S p R q . According to the results of Theorem 2.7, the following self-consistent equations uniquely define the groundstates density matrices associated with the periodic nuclear densities µ per ,L and µ per ,R : γ per ,L : “ p´8 ,(cid:15) L s p H per ,L q , H per ,L : “ ´ ∆2 ` V per ,L , V per ,L : “ p ρ per ,L ´ µ per ,L q ‹ Γ L G a L ,γ per ,R : “ p´8 ,(cid:15) R s p H per ,R q , H per ,R : “ ´ ∆2 ` V per ,R , V per ,R : “ p ρ per ,R ´ µ per ,R q ‹ Γ R G a R , where the negative constants (cid:15) L and (cid:15) R are the Fermi levels of the quasi 1D systems. The junction of the quasi1D systems are described by considering the following nuclear density configuration (see Fig.2): µ J p x, r q : “ x ď ¨ µ per ,L p x, r q ` x ą ¨ µ per ,R p x, r q ` v p x, r q , (3.2)where v p x, r q P L { p R q describes how the junction switches between the underlying nuclear densities. Theassumption v P L { p R q ensures that D p v, v q ă `8 . Recall that D p f, g q “ ż R ż R f p x q g p y q| x ´ y | dx dy “ π ż R p f p k q p g p k q k dk describes the Coulomb interactions in the whole space. Once one sets the nuclear configuration (3.2), electronsare allowed to move in the 3D space. The infinite rHF energy functional for the junction system associated witha test density matrix γ J formally reads E p γ J q “ Tr ˆ ´
12 ∆ γ J ˙ ` D p ρ γ J ´ µ J , ρ γ J ´ µ J q . (3.3)Let us also introduce the Coulomb space C and its dual C (Beppo-Levi space [15]): C : “ (cid:32) ρ P S p R q ˇˇ p ρ P L p R q , D p ρ, ρ q ă 8 ( , C : “ (cid:32) V P L p R q | ∇ V P p L p R qq ( . (3.4)Remark that the ground state energy of the junction system, if it exists, is infinite and there is no periodicity inthis system, hence usual techniques which essentially consist in considering the energy per unit volume [20, 15]are not applicable. We next define a reference system such that the difference between the junction system andthe reference can be considered as a perturbation. This perturbative approach has been used in [36, 33, 35, 15]in various contexts. The next section is devoted to the rigorous mathematical construction of the reference stateand its rHF energy functional. 11 σ ( H per ,L ) — σ ( H per ,R ) 0 + ∞ ǫ R ǫ L Figure 3: Spectrum of H per ,L , H per ,R in the non-equilibrium regime. — σ ( H per ,L ) — σ ( H per ,R ) 0 + ∞ ǫ R ǫ L Σ a Σ b Figure 4: Spectrum of H per ,L , H per ,R in the equilibrium regime. In this section, we construct a reference Hamiltonian obtained by a linear combination of the periodic mean–field potentials V per ,L and V per ,R . We prove the validity of this approach by showing that the density generatedby this reference state is close to the linear combination of the periodic densities ρ per ,L and ρ per ,R . Hamiltonian of the reference state.
We introduce a class of smoothed cut-off functions. For x P R ,consider: X : “ ! χ P C p R q ˇˇˇ ď χ ď χ p x q “ if x P ´ ´8 , ´ a L ı ˆ R ; χ p x q “ if x P ” a R , `8 ¯ ˆ R ) . (3.5)Fix χ P X , let us introduce a reference potential V χ : “ χ V per ,L ` p ´ χ q V per ,R . We will show in Section 3.4 that the choice of χ P X is irrelevant. By Theorem 2.7 and Lemma 2.10 we knowthat V χ belongs to L p loc ` R , L p p R q ˘ for ă p ď 8 , is continuous in all directions and tends to zero in the r -direction. By the Kato–Rellich theorem (see for example [39, Theorem 9.10]), there exists a unique self-adjointoperator H χ : “ ´
12 ∆ ` V χ (3.6)on L p R q with domain H p R q and form domain H p R q . We next show that the essential spectrum of thereference Hamiltonian H χ is the union of the essential spectra of H per ,L and H per ,R , which implies that thereference system does not change essentially the unions of possible energy levels of quasi periodic systems,and that there are no surface states which propagate along the junction surface in the r -direction. Note thatthis is a priori not obvious as the cut-off function χ is r -translation invariant (hence not compact), thereforescattering states may occur at the junction surface and escape to infinity in the r -direction. Standard techniquesin scattering theory to prove this statement, such as Dirichlet decoupling [25, 41], are not applicable in oursituation since the junction surface is not compact. Proposition 3.1 (Spectral properties of the reference state H χ ) . For any χ P X , the essential spectrum of H χ defined in (3.6) satisfies σ ess p H χ q “ σ ess p H per ,L q ď σ ess p H per ,R q . In particular, r , `8q Ă σ ess p H χ q and σ ess p H χ q does not depend on the shape of the cut-off function χ P X definedin (3.5) . The proof relies on an explicit construction of Weyl sequences associated with H χ (see Section 4.8). Re-mark that Proposition 3.1 also implies that the reference system essentially preserves the scattering channels of12he underlying quasi 1D systems, since the scattering involves the purely absolutely continuous spectrum of aHamiltonian (see for example [28, 9, 10]). However, this does not exclude the existence of embedded eigenvaluesin the essential spectrum, which may cause additional scattering channels [63, 64]. We prove in Corollary 3.5.1that the results in Proposition 3.1 still hold for the nonlinear junction. Reference state as a spectral projector.
Before constructing the reference state, let us discuss differentregimes for junction system. From Theorem 2.7 we know that the chemical potentials (Fermi levels) (cid:15) L and (cid:15) R are negative. Introduce the energy interval I (cid:15) F : “ r min p (cid:15) L , (cid:15) R q , max p (cid:15) L , (cid:15) R qs . In view of Proposition 3.1,assume that the essential spectrum of H χ below is purely absolutely continuous, the non-equilibrium regime(Fig. 3) corresponds to σ ac p H χ q č I (cid:15) F ‰ H . In this regime, steady state currents occur and the Landauer-B¨uttiker conductance can be calculated [8, 9, 10].When µ per ,L and µ per ,R are identical, the junction system becomes periodic with different chemical potentials (cid:15) L and (cid:15) R on the left and right half lines. In this case the Thouless conductance [8] can be defined and it is givenby C T | σ ac p H χ q Ş I (cid:15) F || I (cid:15) F | ą , for some positive constant C T . However it is not the aim of this article to discuss steady state currents for non-equilibrium systems. We instead consider the equilibrium regime (see Fig. 4) with the following assumption. Assumption 2.
The chemical potential (cid:15) L and (cid:15) R are in a common spectral gap p Σ a , Σ b q (equilibrium regime, seeFig. 4), where Σ a is the maximum of the filled bands of H per ,L and H per ,R , and Σ b is the minimum of the unfilledbands of H per ,L and H per ,R . — σ ( H per ,ℓ ) — σ ( H per ,r ) 0 + ∞ — σ ( H χ ) × ××× ǫ F < b Σ a Figure 5: Spectrum of H per ,L , H per ,R and H χ below .Assumption 2 guarantees that the Fermi level of the junction system lies in a spectral gap of H χ in viewof Proposition 3.1, which is a common hypothesis [15, 33, 35] for 3D periodic insulating and semi-conductingsystems. We make this assumption for simplicity. Remark that with approaches proposed in [29, 30, 14] it ispossible to extend the results to metallic systems provided that the junction system is in its ground state and nosteady state current occurs.Let us without loss of generality choose the Fermi level (cid:15) F “ max p (cid:15) L , (cid:15) R q “ sup I (cid:15) F and define the referencestate γ χ as the spectral projector associated with the states of H χ below (cid:15) F : γ χ : “ p´8 ,(cid:15) F q p H χ q . (3.7)Remark that H χ can have discrete spectrum in the gap p Σ a , Σ b q , with eigenvalues possibly accumulating at Σ a and Σ b , and (cid:15) F can also be an eigenvalue of H χ . The definition of (3.7) however excludes the possible boundstates with energy (cid:15) F .The following proposition shows that the density ρ χ of γ χ is well defined in L p R q , and is close to thelinear combination of the periodic densities ρ per ,L and ρ per ,R , the difference decaying exponentially fast in the x -direction as | x | Ñ 8 . Proposition 3.2 (Exponential decay of density) . Under Assumption 2, the spectral projector γ χ is locally traceclass, so that its density ρ χ is well defined in L p R q . Moreover, χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ P L p p R q for ă p ď . urthermore, denote by w a the characteristic function of the unit cube centered at a P R . There exist positiveconstants C and t such that for all α “ p α x , , q P R , with either supp p w α q Ă p´8 , a L { s ˆ R or supp p w α q Ă r a R { , `8q ˆ R , it holds, ż R ˇˇ w α ` χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ ˘ w α ˇˇ ď C e ´ t | α | . The proof can be read in Section 4.10.
Fictitious nuclear density of the reference state.
The density ρ χ associated with γ χ is fixed once the Fermilevel (cid:15) F is chosen. We can therefore define a fictitious nuclear density µ χ by imposing that the total electronicdensity ρ χ ´ µ χ generates the potential V χ . The fictitious nuclear density µ χ is given by ´ ∆ V χ “ π p ρ χ ´ µ χ q , µ χ : “ ρ χ ´ ` χ p ρ per ,L ´ µ per ,L q ` p ´ χ qp ρ per ,R ´ µ per ,R q ` η χ ˘ , (3.8)where η χ has compact support in the x -direction: η χ : “ ´ π ` B x p χ q p V per ,L ´ V per ,R q ` B x p χ qB x p V per ,L ´ V per ,R q ˘ . (3.9)Let us emphasize that the Poisson’s equation (3.8) is defined on the whole space R . The nuclear density of the junction is a fictitious nuclear density plus a perturbation.
Once we havedefined fictitious nuclear density, we can treat the difference between the real nuclear density of the junctionsystem µ J and the fictitious nuclear density µ χ as a perturbative nuclear density. By doing so we can definea finite renormalized energy with respect to the perturbative nuclear density. Note that this idea is similarto the definition of the defect state in [15] for defects in crystals, and the polarization of the vacuum in theBogoliubov–Dirac–Fock model [36, 33, 35]. Introduce ν χ : “ µ J ´ µ χ “ ` x ď ´ χ ˘ p µ per ,L ´ µ per ,R q ` ` χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ ˘ ` η χ ` v. (3.10)In order to guarantee that the perturbative state has a finite Coulomb energy, we need D p ν χ , ν χ q ă `8 . Asufficient condition is that ν χ belongs to L { p R q . This motivates the following L p -estimate on η χ . Lemma 3.3.
The function η χ defined in (3.9) belongs to L p ` R ˘ for ă p ă . The proof can be read in Section 4.11. In view of Lemma 3.3 and Proposition 3.2, together with the factthat p x ď ´ χ q p µ per ,L ´ µ per ,R q has compact support and v belongs to L { ` R ˘ , it is easy to see that ν χ belongs to L { ` R ˘ , and hence to the Coulomb space C defined in (3.4). This means that the perturbative stategenerated by the nuclear density ν χ has finite Coulomb energy. Remark . Remark that the integrability of V per provided by Lemma 2.10 is crucial to deduce Lemma 3.3. In this section we define a perturbative state associated with the perturbative density ν χ following the ideasdeveloped in [15]. We formally derive the rHF energy difference between the junction state γ J and the referencestate γ χ by writing γ J “ γ χ ` Q χ with Q χ a trial density state. In view of (3.3), we formally have E p γ J q ´ E p γ χ q formally hnlj “ Tr ˆ ´
12 ∆ p γ χ ` Q χ q ˙ ` D p ρ J ´ µ J , ρ J ´ µ J q´ Tr ˆ ´
12 ∆ γ χ ˙ ´ D p ρ χ ´ µ χ , ρ χ ´ µ χ q“ Tr ˆ ´
12 ∆ Q χ ˙ ` D p ρ χ ´ µ χ , ρ Q χ q ´ D p ρ Q χ , ν χ q ` D p ρ Q χ , ρ Q χ q´ D p ρ χ ´ µ χ , ν χ q ` D p ν χ , ν χ q“ Tr p H χ Q χ q ´ D p ρ Q χ , ν χ q ` D p ρ Q χ , ρ Q χ q ´ D p ρ χ ´ µ χ , ν χ q ` D p ν χ , ν χ q . (3.11)14e next give a mathematical definition of the terms in the last equality of (3.11). We expect Q χ to be a per-turbation of the reference state γ χ . More precisely, we expect Q χ to be Hilbert-Schmidt. This is usually calledthe “Shale-Stinespring” condition [66], see [54, 69] for a detailed discussion. Moreover, we also expect the ki-netic energy of Q χ to be finite. Let Π be an orthogonal projector on the Hilbert space H such that both Π and Π K : “ ´ Π have infinite ranks. A self-adjoint compact operator A on H is said to be Π -trace class if A P S p H q and both Π A Π and Π K A Π K are in S p H q . For an operator A we define its Π -trace as Tr Π p A q : “ Tr p Π A Π q ` Tr ` Π K A Π K ˘ , and denote by S Π1 p H q the associated set of Π -trace class operators. Since the reference state γ χ defined in (3.7)is an orthogonal projector on L p R q , we can define associated γ χ -trace class operators. For any trial densitymatrix Q χ , let us denote by Q `` χ : “ γ K χ Q χ γ K χ and Q ´´ χ : “ γ χ Q χ γ χ , and introduce a Banach space of operatorswith finite γ χ -trace and finite kinetic energy as follows: Q χ : “ (cid:32) Q χ P S γ χ p L p R qq ˇˇ Q ˚ χ “ Q χ , | ∇ | Q χ P S p L p R qq , | ∇ | Q `` χ | ∇ | P S p L p R qq , | ∇ | Q ´´ χ | ∇ | P S p L p R qq ( , equipped with its natural norm (cid:107) Q χ (cid:107) Q χ : “ (cid:107) Q χ (cid:107) S ` (cid:13)(cid:13) Q `` χ (cid:13)(cid:13) S ` (cid:13)(cid:13) Q ´´ χ (cid:13)(cid:13) S ` (cid:107) | ∇ | Q χ (cid:107) S ` (cid:13)(cid:13) | ∇ | Q `` χ | ∇ | (cid:13)(cid:13) S ` (cid:13)(cid:13) | ∇ | Q ´´ χ | ∇ | (cid:13)(cid:13) S . By construction Tr γ χ p Q χ q “ Tr ` Q `` χ ˘ ` Tr ` Q ´´ χ ˘ . For Q to be an admissible perturbation of the referencestate γ χ , Pauli’s principle requires that ď γ χ ` Q χ ď . Let us introduce the following convex set of admissibleperturbative states: K χ : “ t Q χ P Q χ | ´ γ χ ď Q χ ď ´ γ χ u . Remark that K χ is not empty since it contains at least . Note also that K χ is the convex hull of states in Q χ of the special form γ ´ γ χ , where γ is an orthogonal projector [15]. Furthermore, for any Q χ P K χ a simplealgebraic calculation shows that Q `` χ ě , Q ´´ χ ď , ď Q χ ď Q `` χ ´ Q ´´ χ . As mentioned in the previous section, the Fermi level (cid:15) F can be an eigenvalue of H χ . Consider N P N ˚ such that (cid:15) F P p Σ N,χ , Σ N ` ,χ s , where Σ N,χ ă Σ N ` ,χ are two eigenvalues of H χ in the gap p Σ a , Σ b q , and let Σ N,χ “ Σ a and Σ N ` ,χ “ Σ b whenever there is no such element. For any κ P p Σ N,χ , (cid:15) F q , let us introduce the followingrHF kinetic energy of a state Q χ P Q χ : Tr γ χ p H χ Q χ q : “ Tr ´ | H χ ´ κ | { ` Q `` χ ´ Q ´´ χ ˘ | H χ ´ κ | { ¯ ` κ Tr γ χ p Q χ q . By [15, Corollary 1], the above expression is independent of κ P p Σ N,χ , (cid:15) F q . In view of the last line of (3.11) weintroduce the following minimization problem E κ,χ “ inf Q χ P K χ (cid:32) E χ p Q χ q ´ κ Tr γ χ p Q χ q ( , (3.12)where E χ p Q χ q : “ Tr γ χ p H χ Q χ q ´ D ` ρ Q χ , ν χ ˘ ` D ` ρ Q χ , ρ Q χ ˘ . (3.13) The following result shows that the minimization problem (3.12) is well posed and admits minimizers.
Proposition 3.5. ( Existence of the perturbative ground state ) Assume that Assumption 2 holds. Then thereexist minimizers for the problem (3.12) . There may be several minimizers, but they all share the same density.Moreover, any minimizer Q χ of (3.12) satisfies the following self-consistent equation: $&% Q χ “ p´8 ,(cid:15) F q p H Q χ q ´ γ χ ` δ χ ,H Q χ “ H χ ` p ρ Q χ ´ ν χ q ‹ | ¨ | ´ , (3.14) where δ χ is a finite-rank self-adjoint operator satisfying ď δ χ ď and Ran p δ χ q Ď Ker p H Q χ ´ (cid:15) F q .
15e proof is a direct adaptation of several results obtained in [15], see a short summary in Section 4.12 forcompleteness. Remark that p ρ Q χ ´ ν χ q ‹ | ¨ | ´ P L p R q by [17, Lemma 16], therefore p ´ ∆ q ´ p ρ Q χ ´ ν χ q ‹ | ¨ | ´ belongs to S by the Kato–Seiler–Simon inequality (4.1), hence p ρ Q χ ´ ν χ q ‹ | ¨ | ´ is ´ ∆ -compacthence H χ -compact by the boundedness of V χ , leaving the essential spectrum unchanged. Therefore in view ofProposition 3.1, the following corollary holds. Corollary 3.5.1.
For any χ P X , and H Q χ solution of (3.14) , it holds σ ess ´ H Q χ ¯ “ σ ess p H per ,L q ď σ ess p H per ,R q , σ ess ´ H Q χ ¯ č p´8 , s Ď σ ac ´ H Q χ ¯ . In particular, r , `8q Ă σ ess ´ H Q χ ¯ and σ ess ´ H Q χ ¯ does not depend on the shape of the cut-off function χ P X defined in (3.5) . The result of Proposition 3.5 can be interpreted as follows: given a cut-off function χ belonging to the class X defined in (3.5), we can construct a reference state γ χ and a perturbative ground state Q χ , the sum of whichforms the ground state of the junction system. However it is artificial to introduce cut-off functions χ since thereare infinitely many possible choices. In view of (3.2), the ground state of the junction system should not dependon the choice of cut-off functions. The following theorem shows that the electronic density of the junctionsystem is indeed independent of the choice of the cut-off function χ . Theorem 3.6 ( Independence of the reference state and uniqueness of ground state density ) . The groundstate density of the junction system with nuclear density defined in (3.2) under the rHF description is independent ofthe choice of the cut-off function χ P X , i.e., the total electronic density ρ J “ ρ χ ` ρ Q χ is independent of χ , where ρ χ is the density associated with the spectral projector γ χ defined in (3.7) , and ρ Q χ is the unique density associatedwith the solution Q χ of the minimization problem (3.14) . The proof can be read in Section 4.13. Theorem 3.6 and Proposition 3.5 together imply that
Corollary 3.6.1.
The ground state of the junction system (3.2) is of the form p´8 ,(cid:15) F q ´ H χ ` p ρ Q χ ´ ν χ q ‹ | ¨ | ´ ¯ ` δ χ , ď δ χ ď , Ran p δ χ q Ď Ker ´ H χ ` p ρ Q χ ´ ν χ q ‹ | ¨ | ´ ´ (cid:15) F ¯ , and its density is independent of the choice of χ . Remark that an extension to junctions of 2D materials may be done by similar constructions as above, see [19]for more details.
Acknowledgements
I would like to express my deep gratitude to ´Eric Canc`es and Gabriel Stoltz for many useful discussions andadvice for the article, as well as their critical readings of the manuscript.16
Proofs of the results
In order to simplify the notation, in Section 4.1 to 4.11 when treating the quasi 1D periodic system we denoteby S p the Schattern class S p p L ,x p Γ qq for ď p ď `8 . Unless otherwise specified, starting from Section 4.8we use S p instead of S p p L p R qq for the proofs of the junction system.First of all, let us recall the following Kato–Seiler–Simon (KSS) inequality: Lemma 4.1. ([65, Lemma 2.1]) Let ď p ď 8 . For g, f belonging to L p p R q , the following inequality holds: (cid:107) f p´ i ∇ q g p x q (cid:107) S p p L p R qq ď p π q ´ { p (cid:107) g (cid:107) L p p R q (cid:107) f (cid:107) L p p R q . (4.1) The proof is an easy adaptation of the proof of the classical Kato–Seiler–Simon inequality (4.1) by replacing theFourier transform with the mixed Fourier transform F . Let us prove separately (2.6) for p “ and p “ `8 .The conclusion then follows by an interpolation argument. We use the following kernel representation duringthe proofs. For x “ p x, r q and y “ p y, r q belonging to Γ , symbolic calculus shows that the Schwartz kernel K f,ξ pp x, r q , p y, r qq of the operator f p´ i ∇ ξ q acting on L ,x p Γ q formally reads K f,ξ ` p x, r q , p y, r q ˘ “ π ` F ´ ˝ τ x ´ ξ f ˘ ` p x ´ y q , p r ´ r q ˘ “ π ÿ n P Z ż R e i p πn p x ´ y q` k ¨p r ´ r qq f p πn ` ξ, k q d k . (4.2)Let p “ . In view of the isometry identity (2.4), the convolution equality (2.5) and the kernel representation (4.2),the following estimate holds (cid:107) f p´ i ∇ ξ q g (cid:107) S “ π ż Γ ˆ Γ ˇˇ` F ´ ˝ τ x ´ ξ f ˘ p x ´ y q g p y q ˇˇ d x d y ď π ż Γ | g p y q| ˆż Γ ˇˇ` F ´ ˝ τ x ´ ξ f ˘ p x ´ y q ˇˇ d x ˙ d y “ π ż Γ | g p y q| ÿ n P Z ˆż R | f p πn ` ξ, k q| d k ˙ d y “ π (cid:107) g (cid:107) L ,x p Γ q ÿ n P Z (cid:107) f p πn ` ξ, ¨q (cid:107) L p R q , which proves (2.6) for p “ . For p “ `8 , it suffices to prove that for any φ P L ,x p Γ q , (cid:107) f p´ i ∇ ξ q gφ (cid:107) L ,x p Γ q ď (cid:107) g (cid:107) L per ,x p Γ q sup n P Z (cid:107) f pp πn ` ξ, ¨qq (cid:107) L p R q (cid:107) φ (cid:107) L ,x p Γ q . By arguments similar to those used when p “ , we obtain by the isometry (2.4) that (cid:107) f p´ i ∇ ξ q gφ (cid:107) L ,x p Γ q “ ÿ n P Z ż R | F p gφ q p n, k q| ˇˇˇ f ´ p πn ` ξ q ` k ¯ˇˇˇ d k ď sup n P Z (cid:107) f pp πn ` ξ, ¨qq (cid:107) L p R q (cid:107) gφ (cid:107) L ,x p Γ q ď (cid:107) g (cid:107) L per ,x p Γ q sup n P Z (cid:107) f pp πn ` ξ, ¨qq (cid:107) L p R q (cid:107) φ (cid:107) L ,x p Γ q , which is (2.6) for p “ `8 . Therefore, following the same interpolation arguments as in [65, Lemma 2.1] weobtain (2.6) for ď p ď `8 . For n P Z , let us consider the 2D equation: ´ ∆ r G n ` π n G n “ πδ r “ in S p R q .
17t is well known (see for example [56, 49]) that the solution of the above equation is G n p| r |q “ ´ log p| r |q , n ” ,K p π | n || r |q , | n | ě , where K p α q : “ ş `8 e ´ α cosh p t q dt is the modified Bessel function of the second kind. Therefore the Green’sfunction G p x, r q defined in (2.7) can be rewritten as G p x, r q “ ÿ n P Z e πnx G n p r q P S per ,x p Γ q . (4.3)Applying the Laplacian operator to both sides, ´ ∆ G p x, r q “ π ÿ n P Z e πnx δ r “ P S per ,x p Γ q . On the other hand, by the Poisson summation formula ř n P Z δ x “ n “ ř n P Z e πnx P S p R q , we conclude thatthe Green’s function G p x, r q defined in (2.7) satisfies ´ ∆ G p x, r q “ π ÿ n P Z δ p x, r q“p n, q . Taking the Fourier transform F on both sides of (4.3) we obtain that F G p n, k q “ π n ` | k | P S p R q . Let us now give some estimates on r G defined in (2.7). Recall that there exist two positive constants C and C such that [27] ď K p α q ď C | log p α q| , when α ď π,C e ´ α p π { α q { , when α ą π. For | r | ą , it holds that ˇˇˇ r G p x, r q ˇˇˇ ď C `8 ÿ n “ e ´ πn | r | a n | r | ď C a | r | e ´ π | r | ´ e ´ π | r | ď C ´ e ´ π e ´ π | r | a | r | . (4.4)For | r | ď fixed, there exists N ě such that N ď | r | ă N ` . In particular, for n ą N ` we have πn | r | ą π . There exists therefore a positive constant C such that ˇˇˇ r G p x, r q ˇˇˇ ď C ˇˇˇˇˇ N ÿ n “ log p πn | r |q ˇˇˇˇˇ ` C ÿ n “ N ` e ´ πn | r | a n | r | ď C ˇˇˇˇˇż | r | log p πt | r |q dt ˇˇˇˇˇ ` C ż π e ´ t dt “ C | r | | log p π q ´ ´ | r | ` | r | log p π | r |q| ` C e ´ π ď C | r | . (4.5)Together with (4.4) we deduce that r G p x, r q P L p per ,x p Γ q for ď p ă . Note that for all r P R zt u , it holds ş { ´ { G p x, r q dx ” . Consider, for r ‰ , G p x, r q “ ÿ n P Z ˜ a p x ´ n q ` | r | ´ ż { ´ { a p x ´ y ´ n q ` | r | dy ¸ . From [5, Equation (1.8)], ´ ∆ ` G p x, r q ´ p| r |q ˘ “ π ÿ k P Z δ p x, r q“p k, q P S p R q , with G p x, r q “ O p | r | q when | r | Ñ 8 by [5, Lemma 2.2]. Denoting by u p x, r q “ r G p x, r q´ G p x, r q we thereforeobtain that ´ ∆ u p x, r q ” . As u p x, r q belongs to L p R q , by Weyl’s lemma for the Laplace equation we obtainthat u p x, r q is C p R q . On the other hand, by the decay properties of r G and G , we deduce that | u p¨ , r q| Ñ when | r | Ñ 8 uniformly in x , hence by the maximum modulus principle for harmonic functions we canconclude that u ” , hence r G p x, r q “ G p x, r q . 18 .3 Proof of Lemma 2.5 We prove this lemma by an explicit construction of a density matrix belonging to F Γ . Consider a cut-off function (cid:37) P C c p Γ q such that ď (cid:37) ď and ş Γ (cid:37) “ , and define (cid:37) per “ ř n P N ρ p¨ ´ n q . Let ω ě be a parameter tobe made precise later. Define γ ω “ B ´ ˆż Γ ˚ γ ω,ξ dξ π ˙ B , γ ω,ξ “ A ω,ξ A ˚ ω,ξ , A ω,ξ : “ r ,ω s p´ ∆ ξ q (cid:37) per . (4.6)It is easy to see that ď γ ω ď , and that τ xk γ ω “ γ ω τ xk for all k P Z by construction. Let us prove that thekinetic energy per unit cell of γ ω is finite. Denote by F ξ p n, k q “ b p πn ` ξ q ` k . By the Kato–Seiler–Simontype inequality (2.6), ż Γ ˚ Tr L ,x `a ´ ∆ ξ γ ω,ξ a ´ ∆ ξ ˘ dξ “ ż Γ ˚ (cid:107) a ´ ∆ ξ A ω,ξ (cid:107) S dξ ď π ż Γ ˚ (cid:107) (cid:37) per (cid:107) L ,x p Γ q ˜ ÿ n P Z (cid:13)(cid:13)(cid:13)(cid:13) b ` F ξ p n, ¨q r ,ω s ` F ξ p n, ¨q ˘ (cid:13)(cid:13)(cid:13)(cid:13) L p R q ¸ dξ ă `8 . The last estimate follows by the condition p πn ` ξ q ` k ď ω implies that the sum on n is finite and theintegration on k occurs in a compact domain. Hence γ ω belongs to P per ,x . Let us now show that there exists ω ˚ ě such that ρ γ ω ˚ ´ µ per P C Γ . It is easy to see that the density ρ γ ω is smooth and compactly supported in Γ by definition. Moreover, in view of the kernel representation (4.2), the kernel K A ω,ξ of the operator A ω,ξ is K A ω,ξ ` p x, r q , p y, r q ˘ “ π ÿ n P Z ż R e i p πn p x ´ y q` k ¨p r ´ r qq r ,ω s ` F ξ p n, k q ˘ (cid:37) per p y, r q d k . Remark that the non-negative function ω ÞÑ ş Γ ˚ ´ř n P Z ş R r ,ω s ´ F ξ p n, k q ¯ d k ¯ dξ is monotonic non-decreasingin ω , equals when ω “ and tends to `8 when ω Ñ `8 . Hence there exists ω ˚ ą such that ż Γ ρ γ ω ˚ “ π ż Γ ˚ (cid:107) A ω ˚ ,ξ (cid:107) S dξ “ π ż Γ ˚ (cid:107) K A ω,ξ (cid:107) L ,x p Γ ˆ Γ q dξ “ p π q ż Γ ˚ ż Γ ż R ÿ n P Z r ,ω ˚ s ` F ξ p n, k q ˘ (cid:37) p y, r q dy d r d k dξ “ p π q ż Γ ˚ ˜ ÿ n P Z ż R r ,ω ˚ s ` F ξ p n, k q ˘ d k ¸ dξ “ ż Γ µ per ą . (4.7)This condition is equivalent to F p ρ γ ω ˚ ´ µ per qp , q “ . As k ÞÑ F p ρ γ ω ˚ ´ µ per qp , k q is C p R q and bounded,the function k ÞÑ | k | ´ F p ρ γ ω ˚ ´ µ per qp , k q is in L p R q . In view of this, there exists a positive constant C such that ÿ n P Z ż R ˇˇ F p ρ γ ω ˚ ´ µ per qp n, k q ˇˇ | k | ` π n d k ď ż | k |ď π ˇˇ F p ρ γ ω ˚ ´ µ per qp , k q ˇˇ | k | d k ` π ¨˝ż | k |ą π ˇˇ F p ρ γ ω ˚ ´ µ per qp , k q ˇˇ d k ` ÿ n P Z zt u ż R ˇˇ F p ρ γ ω ˚ ´ µ per qp n, k q ˇˇ d k ˛‚ ď C ` π ż Γ ˇˇ ρ γ ω ˚ ´ µ per ˇˇ ă `8 . (4.8)In view of the definition of the Coulomb energy (2.13), we can therefore conclude that D Γ p ρ γ ω ˚ ´ µ per , ρ γ ω ˚ ´ µ per q ă `8 . This concludes the proof that the state γ ω ˚ P F Γ . Hence F Γ is not empty. As any density ρ γ associated with γ P P per ,x is integrable, we can conclude that (2.16) holds in view of Remark 2.4.19 .4 Proof of Theorems 2.6 Let us start by giving a convenient equivalent formulation of the minimization problems (2.17). The opera-tor ´ ∆ ξ is not invertible, but the operator ´ ∆ ξ ´ κ is positive definite and ˇˇ ´ ∆ ξ ´ κ ˇˇ ´ is bounded forany κ ă . Therefore in view of the charge neutrality constraint (2.16), we rewrite the periodic rHF energyfunctional (2.15) as follows: @ γ P F Γ , E per ,x p γ q “ π ż Γ ˚ Tr L ,x p Γ q ˆ ´
12 ∆ ξ γ ξ ˙ dξ ` D Γ p ρ γ ´ µ per , ρ γ ´ µ per q“ E per ,x,κ p γ q ` κ π ż Γ ˚ Tr L ,x p Γ q p γ ξ q dξ “ E per ,x,κ p γ q ` κ ż Γ µ per , with E per ,x,κ p γ q : “ π ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ¸ dξ ` D Γ p ρ γ ´ µ per , ρ γ ´ µ per q . The parameter κ can be interpreted as the Lagrangian multiplier associated with the charger neutrality con-straint. Therefore by fixing κ ă , the minimization problem (2.17) is equivalent to the problem inf (cid:32) E per ,x,κ p γ q ; γ P F Γ ( . (4.9)We prove the existence of minimizers and the uniqueness of the density of minimizers for the problem (4.9)(hence of (2.17)) by considering a minimizing sequence, and show that there is no loss of compactness. Thisapproach is rather classical for rHF type models [20, 15, 16, 14]. But in our case we need to be careful as electronsmight escape to infinity in the r -direction. We show that this is impossible thanks to the Coulomb interactions(Lemma 4.3). Weak convergence of the minimizing sequence.
First of all it is easy to see that the functional E per ,x,κ p¨q is well defined on the non-empty set F Γ . Consider a minimizing sequence of E per ,x,κ p¨q " γ n : “ B ´ ˆż Γ ˚ γ n,ξ dξ π ˙ B * n ě on F Γ . There exists C ą such that for all n ě : ď ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ¸ dξ ď C, ď D Γ p ρ γ n ´ µ per , ρ γ n ´ µ per q ď C. (4.10)The kinetic energy bound (4.10) together with the inequality (2.12) implies that the sequence (cid:32) ? ρ γ n ( n ě isuniformly bounded in H ,x p Γ q hence in L ,x p Γ q by Sobolev embeddings. Therefore for all n P N ˚ , thedensity ρ γ n belongs to L p per ,x p Γ q for ď p ď . On the other hand, for almost all ξ P Γ ˚ , the operator γ n,ξ is atrace-class operator on L ,x p Γ q . As ď γ n,ξ ď γ n,ξ ď , we obtain that ď ż Γ ˚ (cid:107) γ n,ξ (cid:107) S dξ “ ż Γ ˚ Tr L ,x p Γ q ` γ n,ξ ˘ dξ ď ż Γ ˚ Tr L ,x p Γ q p γ n,ξ q dξ “ πZ. This implies that the operator-valued function ξ ÞÑ γ n,ξ is uniformly bounded in L p Γ ˚ ; S q . Furthermore,the uniform boundedness ď γ n,ξ ď also implies that the operator-valued function ξ ÞÑ γ n,ξ belongs to L p Γ ˚ ; S p L ,x p Γ qqq . Combining these remarks with the uniform energy bound (4.10), we deduce that thereexist (up to extraction): γ “ B ´ ˆż Γ ˚ γ ξ dξ π ˙ B , ρ γ P L p per ,x p Γ q , Ă ρ γ ´ µ per P C Γ , (4.11)such that γ n ˚ Ýá γ in the following sense: for any operator-valued function ξ ÞÑ U ξ P L p Γ ˚ ; S q ` L p Γ ˚ ; S q , ż Γ ˚ Tr L ,x p Γ q p U ξ γ n q dξ ÝÝÝÑ n Ñ8 ż Γ ˚ Tr L ,x p Γ q p U ξ γ q dξ. (4.12)20e density ρ γ n á ρ γ weakly in L p per ,x p Γ q for ă p ď . The total density ρ γ n ´ µ per á Ă ρ γ ´ µ per weakly in C Γ . The convergence (4.12) is due to the fact that the predual of L p Γ ˚ ; S p L ,x p Γ qqq is L p Γ ˚ ; S q , and that L p Γ ˚ ; S q is a Hilbert space. Remark . The convergence of γ n ˚ Ýá γ in the sense of (4.12) can also be reformulated as the convergence inthe sense of the following weak- ˚ topology: B γ n B ´ ÝÝÝÑ n Ñ8 B γ B ´ for the weak- ˚ topology of L ` Γ ˚ ; S ` L ,x p Γ q ˘˘ č L p Γ ˚ ; S q . (4.13)Denote by D per ,x p Γ q the functions which are C on R , -periodic in the x -direction, and have compact sup-port in the r -direction. Denote by D per ,x p Γ q the dual space of D per ,x p Γ q . The following lemma guarantees thatthe densities obtained by different weak limit processes coincide. In particular there is no loss of compactnessin the r -direction when | r | Ñ 8 . Lemma 4.3 (Consistency of densities) . Denote by ρ γ the density associated with the density matrix γ obtainedin the weak limit (4.11) . Then ρ γ “ ρ γ “ Ă ρ γ in D per ,x p Γ q . In particular, ρ γ “ ρ γ as elements in L p per ,x p Γ q for ă p ď and ρ γ ´ µ per “ Ă ρ γ ´ µ per as elements in C Γ . We postpone the proof to Section 4.5.
The state γ is a minimizer. Let us first show that the kinetic energy of γ obtained by the weak limit (4.12) isfinite. To achieve this, consider an orthonormal basis t e i u i P N Ă H ,x p Γ q of L ,x p Γ q , and define the followingfamily of operators for N P N ˚ : M Nξ : “ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ˜ N ÿ i “ | e i y x e i | ¸ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { . An easy computation shows that for all ξ P Γ ˚ , the operator M Nξ belongs to S . Moreover, the function ξ ÞÑ M Nξ can be seen as an operator-valued function belonging to L p Γ ˚ ; S q as Γ ˚ “ r´ π, π q is a finite interval.In view of the convergence (4.12) and by choosing U ξ “ M Nξ we obtain that (recalling that Tr p AB q “ Tr p BA q when A, B are Hilbert-Schmidt operators, and ˇˇ ´ ∆ ξ ´ κ ˇˇ { γ ξ is Hilbert-Schmidt for almost all ξ P Γ ˚ ) ď ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ˜ N ÿ i “ | e i y x e i | ¸¸ dξ “ ż Γ ˚ Tr L ,x p Γ q ` M Nξ γ ξ ˘ dξ “ lim n Ñ8 ż Γ ˚ Tr L ,x p Γ q ` M Nξ γ n,ξ ˘ dξ “ lim n Ñ8 ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ˜ N ÿ i “ | e i y x e i | ¸¸ dξ “ lim n Ñ8 ż Γ ˚ N ÿ i “ C e i ˇˇˇˇˇ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ˇˇˇˇˇ e i G dξ ď lim inf n Ñ8 ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ¸ dξ ď C , where the last step we have used the uniform energy bound (4.10). Therefore, by passing to the limit N Ñ `8 ,by Fatou’s lemma we have ď ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ¸ dξ ď lim inf n Ñ8 ż Γ ˚ Tr L ,x p Γ q ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ¸ dξ ď C . (4.14)21emark that the bound (4.14) also implies that ? ρ γ belongs to H ,x p Γ q by the Hoffmann–Ostenhof inequal-ity (2.12). Hence ρ γ P L p per ,x p Γ q for ď p ď . Since ρ γ ´ µ per is an element in C Γ by Lemma 4.3, this impliesthat the charge is neutral by Remark 2.4. That is, ż Γ ρ γ “ ż Γ µ per . As D Γ p¨ , ¨q defines an inner product on C Γ , by the weak convergence (4.11) of ρ γ n ´ µ per to Ă ρ γ ´ µ per in C Γ andthe consistency of densities given by Lemma 4.3, we obtain that D Γ p ρ γ ´ µ per , ρ γ ´ µ per q ď lim inf n Ñ8 D Γ p ρ γ n ´ µ per , ρ γ n ´ µ per q . (4.15)In view of (4.14) and (4.15), we conclude that E per ,x p γ q ď lim inf n Ñ8 E per ,x p γ n q , which shows that the state γ obtained in (4.11) is a minimizer of the problem (2.17). Let us prove that allminimizers share the same density: consider two minimizers γ and γ . By the convexity of F Γ it holds that p γ ` γ q P F Γ . Moreover E per ,x ˆ γ ` γ ˙ “ E per ,x p γ q ` E per ,x p γ q ´ D Γ ` ρ γ ´ ρ γ , ρ γ ´ ρ γ ˘ , which shows that D Γ ` ρ γ ´ ρ γ , ρ γ ´ ρ γ ˘ ” , hence all the minimizers of the problem (2.17) share the samedensity. Equality of Ă ρ γ and ρ γ . The proof follows ideas similar to the ones used for the proof of [14, Lemma 3.5]by considering a test function in D per ,x p Γ q and replacing the Fourier transform by the mixed Fourier transformdefined in (2.3). Consider a test function w P D per ,x p Γ q . The weak convergence of ρ γ n á ρ γ in L p per ,x p Γ q with ă p ď implies that x ρ γ n ´ µ per , w y D per ,x , D per ,x ÝÝÝÑ n Ñ8 x ρ γ ´ µ per , w y D per ,x , D per ,x . On the other hand, x ρ γ n ´ µ per , w y D per ,x , D per ,x “ ż Γ p ρ γ n ´ µ per q w “ ÿ n P Z ż R F p ρ γ n ´ µ per q p n, k q F w p n, k q d k “ π ÿ n P Z ż R F p ρ γ n ´ µ per q p n, k q F f p n, k q π n ` | k | d k . (4.16)where f “ ´ π ∆ w . Note that f belongs to the Coulomb space C Γ defined in (2.14) since for all n P Z , F f p n, ¨q P L p R q , and D Γ p f, f q “ π ÿ n P Z ż R | F f p n, k q| | k | ` π n d k “ π ÿ n P Z ż R ` | k | ` π n ˘ | F w p n, k q| d k “ π ÿ n P Z ż R | F p ∇ w q p n, k q| d k “ π ż Γ | ∇ w | ă `8 . Therefore in view of (4.16), the convergence D Γ p ρ γ n ´ µ per , f q ÝÝÝÑ n Ñ8 D Γ p Ă ρ γ ´ µ per , f q implies that x ρ γ n ´ µ per , w y ÝÝÝÑ n Ñ8 x Ă ρ γ ´ µ per , w y . The uniqueness of limit in the sense of distribution allows us to conclude.22 quality of ρ γ and ρ γ . Let us prove that ρ γ “ ρ γ in D per ,x p Γ q . The fact that ρ γ n á ρ γ weakly in L p per ,x p Γ q implies that x ρ γ n , w y ÝÝÝÑ n Ñ8 x ρ γ , w y . Therefore it suffices to prove that the operator-valued function ξ ÞÑ wγ n,ξ P L p Γ ˚ ; S q converges in thefollowing sense: π ż Γ ˚ Tr L ,x p Γ q p wγ n,ξ q dξ “ x ρ γ n , w y ÝÝÝÑ n Ñ8 x ρ γ , w y “ π ż Γ ˚ Tr L ,x p Γ q p wγ ξ q dξ. (4.17)The weak convergence (4.12) does not guarantee the above convergence since the function w does not belong toany Schatten class. We prove (4.17) by using the kinetic energy bound (4.10), which implies that the operator-valued function ξ ÞÑ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ “ ˜ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { ¸ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ ´ { is uniformly bounded in L p Γ ˚ ; S q as ˇˇ ´ ∆ ξ ´ κ ˇˇ ´ { is uniformly bounded with respect to ξ P Γ ˚ . Moreover,the energy bounded (4.10) also implies that ξ ÞÑ ˇˇ ´ ∆ ξ ´ κ ˇˇ { ? γ n,ξ is uniformly bounded in L p Γ ˚ ; S q .Hence the operator-valued function ξ ÞÑ ˇˇ ´ ∆ ξ ´ κ ˇˇ { γ n,ξ “ ˇˇ ´ ∆ ξ ´ κ ˇˇ { ? γ n,ξ ? γ n,ξ is uniformly boundedin L p Γ ˚ ; S q as ď γ n,ξ ď . Therefore ξ ÞÑ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ is uniformly bounded in L p Γ ˚ ; S q č L p Γ ˚ ; S q , hence in L q p Γ ˚ ; S q q for ď q ď by interpolation. Therefore, up to extraction, the following weak conver-gence holds: for any operator-valued function ξ ÞÑ W ξ P L q ` Γ ˚ ; S q ˘ where q “ qq ´ for ă q ď , ż Γ ˚ Tr L ,x p Γ q ˜ W ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ¸ dξ ÝÝÝÑ n Ñ8 ż Γ ˚ Tr L ,x p Γ q ˜ W ξ ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ ξ ¸ dξ. (4.18)On the other hand by the inequality (2.6) we obtain that for any ξ P Γ ˚ , (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) w ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ ´ { (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S q ď p π q { q (cid:107) w (cid:107) L q per ,x p Γ q ˜ ÿ n P Z ż R q { pp πn ` ξ q ` | r | ` | κ |q q { d r ¸ { q . Upon choosing for example q “ , the right hand side of the above quantity is finite. Therefore upon taking W ξ “ w ˇˇ ´ ∆ ξ ´ κ ˇˇ ´ { in (4.18) we obtain that π ż Γ ˚ Tr L ,x p Γ q p wγ n,ξ q dξ “ π ż Γ ˚ Tr L ,x p Γ q ˜ w ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ ´ { ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ n,ξ ¸ dξ ÝÝÝÑ n Ñ8 π ż Γ ˚ Tr L ,x p Γ q ˜ w ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ ´ { ˇˇˇˇ ´
12 ∆ ξ ´ κ ˇˇˇˇ { γ ξ ¸ dξ “ π ż Γ ˚ Tr L ,x p Γ q p wγ ξ q dξ. Hence (4.17) holds. Therefore ρ γ “ ρ γ in D per ,x p Γ q , which concludes the proof of the lemma. We first define a mean-field Hamiltonian associated with the problem (2.17), and then show that the Fermi level isalways negative. Moreover, the minimizer of (2.17) is uniquely given by the spectral projector of the mean-fieldHamiltonian. In the end we show that the density of the minimizer decays exponentially fast in the r -direction.23 roperties of the mean–field potential and Hamiltonian. We begin with the definition of a mean-fieldpotential and a mean-field Hamiltonian, and next study the spectrum of the mean-field Hamiltonian. Considera minimizer γ per of (2.17) with the unique density ρ γ per P L p per ,x p Γ q where ď p ď . Define the mean-fieldpotential V per : “ q per ‹ Γ G, q per : “ ρ γ per ´ µ per . which is the solution of Poisson’s equation ´ ∆ V per “ πq per . Let us prove that V per belongs to L p per ,x p Γ q for ă p ď `8 . As µ per is smooth and has compact support in the r -direction and ρ γ per belongs to L p per ,x p Γ q for ď p ď , hence F q per p , ¨q belongs to L p R q X L p R q X C p R q by classical Fourier theory (see forexample [61] ). Moreover, as ş Γ | r | ρ γ per p x, r q ă `8 by (2.18), (cid:13)(cid:13)(cid:13)(cid:13) B k F q per p , k q (cid:13)(cid:13)(cid:13)(cid:13) L p R q “ ˇˇˇˇż Γ e ´ i r ¨ k r ¨ q per p x, r q dx d r ˇˇˇˇ ď ˇˇˇˇż Γ | r | ˇˇ ρ γ per ` µ per ˇˇ p x, r q dx d r ˇˇˇˇ ă `8 . (4.19)This implies that F q per p , k q is C p R q and bounded. Remark also that F q per p , q “ by the charge neutralityand that F q per p , ¨q belongs to L p R q X L p R q X C p R q , hence for ď α ă , ż R | F V per p , k q| α d k “ ż R | F q per p , k q| α | k | α d k ď ż | k |ă | F q per p , k q| α | k | α d k ` ˜ż | k |ě | F q per p , k q| α d k ¸ { ˜ż | k |ě | k | α d k ¸ { ă `8 . (4.20)Remark that the mixed Fourier transform F is an isometry from L ,x p Γ q to (cid:96) ` Z , L ` R ˘˘ by (2.4). On theother hand, @ φ P (cid:96) ` Z , L ` R ˘˘ , (cid:107) F ´ φ (cid:107) L p Γ q “ sup x P Γ ˇˇˇˇˇ π ÿ n P Z ż R φ n p k q e i p πnx ` k ¨ r q d k ˇˇˇˇˇ ď π (cid:107) φ (cid:107) (cid:96) p Z ,L p R qq , By the Riesz–Thorin interpolation theorem (see for example [61] and [56, Theorem 5.7]), we can deduce aHausdorff-Young inequality for F ´ : for ď α ď there exists a constant C α depending on α such that @ φ P (cid:96) α ` Z , L α ` R ˘˘ , (cid:107) F ´ φ (cid:107) L α per ,x p Γ q ď C α (cid:107) φ (cid:107) (cid:96) α p Z ,L α p R qq . (4.21)where α : “ α {p α ´ q . Hence in view of (4.20) and (4.21), for ď α ă and ă α “ α {p α ´ q ď `8 thereexist positive constants C α, , C α, and C α, such that (cid:107) V per (cid:107) αL α per ,x p Γ q ď C αα (cid:107) F V per (cid:107) α(cid:96) α p Z ,L α p R qq “ C αα π ÿ n P Z ż R | F V per p n, k q| α d k “ C αα π ż R | F V per p , k q| α d k ` C αα π ÿ n ‰ ż R | F q per p n, k q| α p πn ` | k | q α d k ď C α, ` C αα π ˜ ÿ n ‰ ż R | F q per p n, k q| α d k ¸ { ˜ ÿ n ‰ ż R p πn ` | k | q α d k ¸ { ď C α, ` C α, (cid:107) F q per (cid:107) α(cid:96) α p Z ,L α p R qq ď C α, ` C α, (cid:107) q per (cid:107) αL q per ,x p Γ q ă `8 , (4.22)where the last step we have used estimates similar to (4.21) for F and the fact that q per belongs to L q per ,x p Γ q for { ă q : “ α {p α ´ q ď . Therefore V per belongs to L p per ,x p Γ q for ă p ď `8 . By the elliptic regularitywe know that V per belongs to the Sobolev space W ,p per ,x p Γ q for ă p ď , where W ,p per ,x p Γ q is the space offunctions, together with their gradients and hessians, belong to L p per ,x p Γ q . Approximating V per by functions in D per ,x p Γ q , we also deduce that V per tends to when | r | tends to infinity. The mean-field potential V per defines a ´ ∆ -bounded operator on L p R q with relative bound zero, hence by the Kato–Rellich theorem (see for example[39, Theorem 9.10]) we know that H per “ ´ ∆ ` V per uniquely defines a self-adjoint operator on L p R q withdomain H ` R ˘ and form domain H ` R ˘ . As H per is Z -translation invariant in the x -direction, H per “ B ´ ˆż Γ ˚ H per ,ξ dξ π ˙ B , H per ,ξ : “ ´
12 ∆ ξ ` V per . H per ,ξ does not have a compact resolvent as Γ is not a bounded domain.It is easy to see that σ p´ ∆ ξ q “ σ ess p´ ∆ ξ q “ r , `8q . On the other hand, by the inequality (2.6) we have (cid:13)(cid:13)(cid:13)(cid:13) V per p ´ ∆ ξ q ´ (cid:13)(cid:13)(cid:13)(cid:13) S ď π (cid:107) V per (cid:107) L ,x p Γ q ˜ ÿ n P Z ż R pp πn ` ξ q ` | k | ` q d k ¸ { ă `8 . In particular V per is a compact perturbation of ´ ∆ ξ , and therefore introduces at most countably many eigen-values below which are bounded from below by ´ (cid:107) V per (cid:107) L . Denote by t λ n p ξ qu ď n ď N H these (negative)eigenvalues for N H P N ˚ ( N H can be finite or infinite). Then for all ξ P Γ ˚ , σ ess p H per ,ξ q “ σ ess p´ ∆ ξ q “ r , `8q , σ disc p H per ,ξ q “ ď ď n ď N H λ n p ξ q . In view of the decomposition (2.19), a result of [62, Theorem XIII.85] gives the following spectral decomposition: σ ess p H per q Ě ď ξ P Γ ˚ σ ess p H per ,ξ q “ r , `8q , σ disc p H per q Ď ď ξ P Γ ˚ σ disc p H per ,ξ q “ ď ξ P Γ ˚ ď ď n ď N H λ n p ξ q . We also obtain from [62, Item (e) of Theorem XIII.85] that λ P σ disc p H per q ô t ξ P Γ ˚ | λ P σ disc p H per ,ξ q u has non-trivial Lebesgue measure.By the regular perturbation theory of the point spectra [44] (see also [62, Section XII.2]) and the approach ofThomas [70, Lemma 1], we know that the eigenvalues λ n p ξ q below are analytical functions of ξ and cannot beconstant, so that t ξ P Γ ˚ | λ P σ disc p H per ,ξ q u has trivial Lebesgue measure, and the essential spectrum of H per below is purely absolutely continuous. As a conclusion, σ p H per q “ σ ess p H per q “ ď ξ P Γ ˚ σ p H per ,ξ q . The Fermi level is always negative.
Let us prove that the inequality N H “ F p q ě ş Γ µ per always holds.The physical meaning of this statement is that the Fermi level of the quasi 1D system at ground state is alwaysnegative when the mean-field potential tends to in the r -direction. We prove this by contradiction: assume that F p q ă ş Γ µ per , then we can always construct (infinitely many) states belonging to F Γ with positive energiesarbitrarily close to and they decrease the ground state energy of the problem (2.17).Let us first define a spectral projector representing all the states of H per below : for any ξ P Γ ˚ and H per ,ξ defined in (2.19), define γ ´ per : “ p´8 , s p H per q , γ ´ per ,ξ : “ p´8 , s p H per ,ξ q . Therefore, N H “ F p q “ | Γ ˚ | ż Γ ˚ Tr L ,x p Γ q ´ γ ´ per ,ξ ¯ dξ “ ż Γ ρ γ ´ per . (4.23)The inequality that F p q ă ş Γ µ per implies that Z diff : “ ż Γ µ per ´ N H “ ż Γ µ per ´ ż Γ ρ γ ´ per P N ˚ . (4.24)The condition (4.24) implies in particular that N H ă `8 , i.e. , that there are at most finitely many states below .Let us construct states of H per with positive energies. These states belonging to L p R q approximate the planewaves of H per traveling in the r -direction. For R ą , recall that B R is the ball in R centered at . Consider asmooth function t p x, r q supported in B , equal to one in B { and such that (cid:107) t (cid:107) L p R q “ . For n P N ˚ , let usdefine ψ n p x, r q : “ n ´ { t ˜ p x, r q ´ ` n , ` n , n ˘˘ n ¸ . It is easy to see that ψ n belongs to L p R q , converges weakly to when n tends to infinity and (cid:107) ψ n (cid:107) L p R q “ .Moreover, as V per tends to in the r -direction, hence for any ε ą there exists an integer N ε such that25 ˇ V per p¨ , p n , n qq ˇˇ ď ε when n ě N ε . Denote by t ψ n,ξ u n P N ˚ ,ξ P Γ ˚ the Bloch decomposition B in the x -direction(see Section 2.1 for the definition) of t ψ n u n P N ˚ which belong to L ,x p Γ q . For n ě N ε , it holds (cid:13)(cid:13)(cid:13)(cid:13) H per ψ n (cid:13)(cid:13)(cid:13)(cid:13) L p R q “ π ż Γ ˚ (cid:13)(cid:13)(cid:13)(cid:13) H per ,ξ ψ n,ξ (cid:13)(cid:13)(cid:13)(cid:13) L ,x p Γ q dξ “ (cid:13)(cid:13)(cid:13)(cid:13) ´ n ´ { ∆ t ˜ ¨ ´ ` n , ` n , n ˘˘ n ¸ ` V per ψ n (cid:13)(cid:13)(cid:13)(cid:13) L p R q ď ˆ n ` ε ˙ . (4.25)Remark that γ ´ per ,ξ H per ,ξ “ ř N H n “ P t λ n p ξ qu p H per ,ξ q is a compact operator, where P t¨u p H per ,ξ q the spectral pro-jector of H per ,ξ . There exists an orthonormal basis t e n,ξ u n ě of L ,x p Γ q with elements in H ,x p Γ q such that γ ´ per ,ξ H per ,ξ e n,ξ “ λ n p ξ q e n,ξ for ď n ď N H , and γ ´ per ,ξ H per ,ξ e n,ξ ” for n ą N H . Let us construct test den-sity matrices composed by all the states of H per with negative energies and some states with positive energies.More precisely, for N P N ˚ to be made precise later, consider a test density matrix γ N “ B ´ ˆż Γ ˚ γ N ,ξ dξ π ˙ B , where γ N ,ξ : “ γ ´ per ,ξ ` N ` Z diff ÿ n “ N ` p ´ γ ´ per ,ξ q | ψ n,ξ yx ψ n,ξ |“ `8 ÿ n “ γ ´ per ,ξ | e n,ξ yx e n,ξ | ` N ` Z diff ÿ n “ N ` p ´ γ ´ per ,ξ q | ψ n,ξ yx ψ n,ξ | . Lemma 4.4.
For any N P N ˚ , the state γ N belongs to the admissible set F Γ .Proof. It is easy to see that ď γ N ď . Remark also that Ran ´ γ ´ per ,ξ ¯ “ Span t e n,ξ u ď n ď N H for all ξ P Γ ˚ .The density of γ N can be written as ρ γ N “ π ż Γ ˚ N H ÿ n “ | e n,ξ | dξ ` π N ` Z diff ÿ n “ N ` ż Γ ˚ | ψ n,ξ | dξ. The density ρ γ N belongs to L p per p Γ q for ď p ď as t e n,ξ u n ě and t ψ n,ξ u n ě belong to H ,x p Γ q . Besides, inview of (4.23), ż Γ ρ γ N “ π ż Γ ˚ Tr L ,x p Γ q ´ γ ´ per ,ξ ¯ dξ ` π N ` Z diff ÿ n “ N ` ż Γ ż Γ ˚ | ψ n,ξ | dξ “ N H ` N ` Z diff ÿ n “ N ` ż R | ψ n | “ N H ` Z diff “ ż Γ µ per . (4.26)A simple calculation shows that | ∇ | γ N ,ξ | ∇ | is trace-class on L ,x p Γ q . Hence γ N belongs to P per ,x . Let usshow that ρ γ N ´ µ per belongs to C Γ . Following calculations similar to the ones leading to (4.8), we only needto prove that k ÞÑ | k | ´ F p ρ γ N ´ µ per qp , k q is square-integrable near k “ since ρ γ N ´ µ per belongs to L ,x p Γ q . Remark that F p ρ γ N ´ µ per qp , q “ ş Γ ρ γ N ´ µ per “ and ˇˇˇ B k F p ρ γ N ´ µ per qp , q ˇˇˇ “ ˇˇˇˇż Γ r ´ ρ γ N ´ µ per ¯ p x, r q dx d r ˇˇˇˇ ď π N H ÿ n “ ż Γ ˚ ż Γ | r | | e n,ξ | p x, r q dx d r dξ ` N ` Z diff ÿ n “ N ` ż R | r | | ψ n | p x, r q dx d r ` ż Γ | r | µ per p x, r q dx d r ă `8 , where we have used the fact that the eigenfunctions of H per ,ξ associated with negative eigenvalues decay expo-nentially (see [42, Theorem 3.4] and [22, Theorem 1]) so that ş Γ | r | | e n,ξ | p x, r q dx d r ă `8 for ď n ď N H ξ P Γ ˚ , and the fact that t ψ n u N ` ď n ď N ` Z diff have compact support in the r -direction by definition.Therefore F p ρ γ N ´ µ per qp , k q is C near k “ . The conclusion then follows by arguments similar to thoseleading to (4.8) in Section 4.3.Lemma 4.4 implies that we can construct many admissible states in F Γ by varying N . Let us show that wecan always find N such that γ N decreases the ground state energy of (2.17) if N H ă ş Γ µ per .Given a minimizer γ of (2.17), simple expansion of the energy functional around minimal shows that γ alsominimizes the functional (see [15]) γ ÞÑ ż Γ ˚ Tr L ,x p Γ q p H per ,ξ γ ξ q dξ on F Γ . Therefore, given N P N ˚ we have ď ż Γ ˚ Tr L ,x p Γ q p H per ,ξ p γ N ,ξ ´ γ ξ qq dξ “ ż Γ ˚ Tr L ,x p Γ q ´ γ ´ per ,ξ H per ,ξ p γ N ,ξ ´ γ ξ q ¯ dξ ` ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ p γ N ,ξ ´ γ ξ q ¯ dξ “ M ` ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ γ N ,ξ ¯ dξ ´ ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ γ ξ ¯ dξ, (4.27)where, since ď γ ξ ď and t λ n p¨qu ď n ď N H ă , M : “ ż Γ ˚ Tr L ,x p Γ q ´ γ ´ per ,ξ H per ,ξ p γ N ,ξ ´ γ ξ q ¯ dξ “ ż Γ ˚ N H ÿ n “ λ n p ξ q A e n,ξ ˇˇˇ ´ γ ´ per ,ξ γ ξ ˇˇˇ e n,ξ E dξ “ ż Γ ˚ N H ÿ n “ λ n p ξ q x e n,ξ | ´ γ ξ | e n,ξ y dξ ď . (4.28)In view of (4.25) and by a Cauchy-Schwarz inequality, we deduce that, for N ě N ε : ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ γ N ,ξ ¯ dξ “ ż Γ ˚ N ` Z diff ÿ n “ N ` `8 ÿ m “ x e m,ξ | H per ,ξ | ψ n,ξ yx ψ n,ξ | e m,ξ y dξ ď ż Γ ˚ N ` Z diff ÿ n “ N ` ˜ `8 ÿ m “ |x ψ n,ξ | e m,ξ y| ¸ { ˜ `8 ÿ m “ |x e m,ξ | H per ,ξ | ψ n,ξ y | ¸ { dξ “ ż Γ ˚ N ` Z diff ÿ n “ N ` (cid:107) ψ n,ξ (cid:107) L ,x p Γ q (cid:13)(cid:13)(cid:13)(cid:13) H per ,ξ ψ n,ξ (cid:13)(cid:13)(cid:13)(cid:13) L ,x p Γ q dξ ď π N ` Z diff ÿ n “ N ` ˆ π ż Γ ˚ (cid:107) ψ n,ξ (cid:107) L ,x p Γ q dξ ˙ { ˜ π ż Γ ˚ (cid:13)(cid:13)(cid:13)(cid:13) H per ,ξ ψ n,ξ (cid:13)(cid:13)(cid:13)(cid:13) L ,x p Γ q dξ ¸ { ď ? π N ` Z diff ÿ n “ N ` ˆ n ` ε ˙ { ď ? πZ diff ˆ N ` ε ˙ { . (4.29)Moreover, by definition of γ ´ per ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ γ ξ ¯ “ ż Γ ˚ Tr L ,x p Γ q ´ | H per ,ξ | { p ´ γ ´ per ,ξ q γ ξ p ´ γ ´ per ,ξ q | H per ,ξ | { ¯ ě . (4.30)We distinguish in the inequality (4.28) the cases M ” or M ă . When M ” , the inequality (4.28) impliesthat γ ξ γ ´ per ,ξ “ γ ´ per ,ξ for almost all ξ P Γ ˚ . In view of the the inequalities (4.29) and (4.30), the inequality (4.27)implies that, @ N ě N ε , ď ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ γ ξ ¯ dξ ď ? πZ diff ˆ N ` ε ˙ { . (4.31)27y letting N tend to infinity, it is easy to deduce that p ´ γ ´ per ,ξ q γ ξ “ for almost all ξ P Γ ˚ . Together withthe fact that γ ξ γ ´ per ,ξ “ γ ´ per ,ξ we deduce that γ ξ ” γ ´ per ,ξ for almost all ξ P Γ ˚ . In view of (4.23) and (4.24), bythe charge neutrality we obtain that Z diff “ ż Γ µ per ´ N H “ ż Γ ρ γ ´ N H “ | Γ ˚ | ż Γ ˚ Tr L ,x p Γ q ´ γ ´ per ,ξ ¯ dξ ´ N H ” . Hence ş Γ µ per “ N H “ F p q . This also implies that the minimizer of the problem (2.17) equals to γ ´ per when N H “ F p q “ ş Γ µ per and Z diff ” . When M ă and Z diff ‰ , we can always find ε ą and N ě N ε suchthat ? πZ diff ´ N ` ε ¯ { ď ´ M { . In view of the the inequalities (4.29) and (4.30), the inequality (4.27)implies that @ N ě N ε , ď ż Γ ˚ Tr L ,x p Γ q ´ p ´ γ ´ per ,ξ q H per ,ξ γ ξ ¯ dξ ď M { ă , which leads to contradiction. We can finally conclude that F p q ě ş Γ µ per , so that the Fermi level of the quasi1D system is always non-positive. In the following paragraph we show that the Fermi level can be chosen to bestrictly negative. Form of the minimizer and decay of the density of minimizers.
We have already shown that if N H “ F p q ” ş Γ µ per then the 1D system has a unique minimizer which is equal to γ ´ per . This also implies that foralmost all ξ P Γ ˚ , the operator H per ,ξ has N H strictly negative eigenvalues below , therefore we can alwayschoose the Fermi level (cid:15) F P ` max ξ P Γ ˚ λ N H p ξ q , ˘ . If F p q ą ş Γ µ per , it is clear that there exists (cid:15) F ă such that F p (cid:15) F q “ ş Γ µ per as F p κ q is a non-decreasing function on p´8 , s with range in r , F p qs . The formof the minimizer and the uniqueness is a direct adaptation of [15, Theorem 1] by using a spectral projectordecomposition similar to (A.2) of [15, Theorem 1], that is, the unique minimizer can be written as γ per “ p´8 ,(cid:15) F s p H per q “ B ´ ˆż Γ ˚ γ per ,ξ dξ π ˙ B “ B ´ ˜ż Γ ˚ N H ÿ n “ p λ n p ξ q ď (cid:15) F q | e n,ξ yx e n,ξ | ¸ B , (4.32)where γ per ,ξ : “ p´8 ,(cid:15) F s p H per ,ξ q . The Fermi level (cid:15) F ă can be considered as the Lagrange multiplier associ-ated with the charge neutrality condition F p (cid:15) F q “ ż Γ ρ γ per “ ż Γ µ per . Once the unique minimizer is shown to be a spectral projector, we can use the exponential decay property ofthe eigenfunctions of H per ,ξ in the r -direction via the Combe–Thomas estimate [22, Theorem 1]: for almost all ξ P Γ ˚ , there exist positive constant C p ξ q and α p ξ q such that @p x, r q P Γ , @ ď n ď N H , | e n,ξ p x, r q| ď C p ξ q e ´ α p ξ q| r | . On the other hand, the fact that ş Γ µ per “ F p (cid:15) F q ă `8 implies that there exist only finitely many states of H per ,ξ below (cid:15) F for all ξ P Γ ˚ . Therefore there exist positive constants C (cid:15) F and α (cid:15) F such that ď ρ γ per p x, r q ď π ż Γ ˚ N H ÿ n “ p λ n p ξ q ď (cid:15) F q C p ξ q e ´ α p ξ q| r | dξ ď C (cid:15) F e ´ α (cid:15)F | r | . Remark that this exponential decay property coincides with Assumption 1 that ş Γ | r | ρ γ per p x, r q dx d r ă `8 . Assume that (2.23) holds, that is µ per p x, r q ” µ per p x, | r |q has radial symmetry in the r -direction. It is clearthat the results of Theorem 2.7 hold. We employ the same notations as in Theorem 2.7 in the sequel. By theuniqueness of density, ρ γ per enjoys the same radial symmetry in the r -direction. Recall that q per “ ρ γ per ´ µ per .Together with the facts that ş Γ | r | ρ γ per p x, r q dx d r ă `8 and that µ per has compact support in the r -direction.The radial symmetry in the r -direction implies that ż Γ r ¨ q per p x, r q dx d r ” . (4.33)28emark also that the exponential decay of density implies that ş Γ | r | | q per p x, r q| dx d r ă `8 . Following cal-culations similar to the those in (4.19), (4.20) and (4.22), it is easy to deduce that B k F q per p , k q ” and B k F q per p , k q is continuous and bounded, so that F V per p , ¨q belongs to L p R q , and V per also belongs to L ,x p Γ q . Let us prove that V per P L p per ,x p Γ q for ă p ă (for which we can conclude that V per belongs to L p per ,x p Γ q for ă p ď `8 ). Let us rewrite V per as V per p x, r q “ p q per ‹ Γ G q p x, r q “ ´ q per ‹ Γ r G ¯ p x, r q ` T p r q , (4.34)where T p r q “ ´ ż R q per p r q log ` | r ´ r | ˘ d r , q per p r q : “ ż { ´ { q per p x, r q dx. Recall that µ per has compact support in the r -direction, hence there exist positive constants C q , α q such that @p x, r q P R , | q per p¨ , r q| ď C q e ´ α q | r | , | q per p r q| ď C q e ´ α q | r | . As r G belongs to L p per ,x p Γ q for ď p ă , by Young’s convolution inequality we deduce that q per ‹ Γ r G belongsto L t per ,x p Γ q for ď t ď `8 .It remains to prove that T p r q belongs to L p p R q for ă p ă . Let us use the partition R “ t| r | ď R u Yt| r | ą R u for the integration domain of T p r q . Note first that log p| r |q is L t loc p R q for ď t ă `8 . Therefore,by a Cauchy-Schwarz inequality, there exists a positive constant C R, such that for p “ p {p p ´ q P p , `8q : ˜ż | r |ď R | T p r q| p d r ¸ { p “ ˜ż | r |ď R ˇˇˇˇż R q per p r q log `ˇˇ r ´ r ˇˇ˘ d r ˇˇˇˇ p d r ¸ { p ď ˜ż | r |ď R ˇˇˇˇˇż | r |ď R q per p r q log `ˇˇ r ´ r ˇˇ˘ d r ˇˇˇˇˇ p d r ¸ { p ` C q ˜ż | r |ď R ˇˇˇˇˇż | r |ą R e ´ α q | r | ˇˇ log `ˇˇ r ´ r ˇˇ˘ˇˇ d r ˇˇˇˇˇ p d r ¸ { p ď ˜ż | r |ď R | q per | p ¸ { p ¨˝ż | r |ď R ˜ż | r |ď R ˇˇ log `ˇˇ r ´ r ˇˇ˘ˇˇ p d r ¸ p { p d r ˛‚ { p ` C q ˜ż | r |ď R ˇˇˇˇˇż | r |ą R e ´ α q | r | ˇˇ log p|p R, R q ´ r |q ˇˇ d r ˇˇˇˇˇ p d r ¸ { p ď C R, . (4.35)Let us look at the integration domain t| r | ą R u . Remark that by the charge neutrality condition and the radialsymmetry condition (4.33), it holds, for any r ‰ , ż R q per p r q log p| r |q d r ” , ż R q per p r q r r | r | d r ” . Denote by Q p r , r q : “ log `ˇˇ r ´ r ˇˇ˘ ´ log p| r |q ´ r r | r | “
12 log ˆ ´ rr | r | ` | r | | r | ˙ ´ r r | r | . Then T p r q “ ´ ş R q per p r q Q p r , r q d r . Remark that when | r | ą R and | r || r | ď ε R for ε R ą fixed. ATaylor expansion shows that there exists a positive constant C such that | Q p r , r q| ď C | r | | r | . This motivates thefollowing partition of R given | r | ą R : R “ B ε R Y B A ε R , B ε R : “ " r P R ˇˇˇˇ | r || r | ď ε R * . Hence T p r q “ T int p r q ` T ext p r q , T int p r q : “ ż B εR q per p r q Q p r , r q d r , T ext p r q : “ ż B A εR q per p r q Q p r , r q d r . ă p ă , ż | r |ą R | T int p r q| p d r ď C p ż | r |ą R ˇˇˇˇˇż B εR q per p r q | r | | r | d r ˇˇˇˇˇ p d r ď C ż | r |ą R ˇˇˇˇˇż | r |ď ε R | r | e ´ α q | r | | r | d r ˇˇˇˇˇ p | r | ´ p d r ă `8 . (4.36)Similarly, ż | r |ą R | T ext p r q| p d r ď C ż | r |ą R ˇˇˇˇˇż | r |ą ε R | r | e ´ α q | r | ˇˇ Q p r , r q ˇˇ d r ˇˇˇˇˇ p d r ď C ż | r |ą R ˇˇˇˇˇż | r |ą ε R | r | e ´ α q ε R | r |{ e ´ α q | r |{ ˇˇ Q p r , r q ˇˇ d r ˇˇˇˇˇ p d r ă `8 . (4.37)In view of (4.35), (4.36) and (4.37) we conclude that T p r q belongs to L p p R q for ă p ă . This leads to theconclusion that V per belongs to L p per ,x p Γ q for ă p ď `8 . Let us emphasize that the function χ being translation-invariant in the r -direction makes it difficult to control thecompactness in the r -direction across the junction surface. Our geometry is very different from the cylindricalgeometry considered in [41] for instance which automatically provides compactness in the r -direction.The proof of σ ess p H per ,L q Y σ ess p H per ,R q Ď σ ess p H χ q is relatively easier than the converse inclusion. Intu-itively, the a L -periodicity (resp. a R -periodicity) implies that the Weyl sequences of H per ,L (resp. H per ,R ) aftera translation by na L (resp. na R -translation) are still Weyl sequences of H per ,L (resp. H per ,R ) for any n P Z .This suggests that one can construct Weyl sequences of H χ by properly translating Weyl sequences of H per ,L or H per ,R , as H χ is a linear interpolation of H per ,L and H per ,R hence behaves like H per ,L or H per ,R away fromthe junction surface. The construction of Weyl sequences of either H per ,L or H per ,R from Weyl sequences of H χ is much more difficult, as the support of Weyl sequences is essentially away from any compact set, makingit difficult to control their behaviors. One naive approach should be to cut-off Weyl sequences of H χ by thefunction χ in order to construct Weyl sequences of H per ,L . However, one quickly remarks that the commutator r´ ∆ , χ s is not ´ ∆ -compact, hence it is difficult to ensure that the cut-off sequence is a Weyl sequence of H per ,L .Another naive approach is to use a cut-off function χ c which has compact support in the r -direction. Howeveras mentioned before, it is also difficult to ensure that the Weyl sequences leave any mass in the support of χ c as Weyl sequences essentially have supports away from any compact. We construct a special Weyl sequence of H χ from H per ,L (or H per ,R ) by a suitable cut-off function which introduces ´ ∆ -compact perturbations to solvethis difficulty.The proof is organized as follows: we first prove that r , `8q Ă σ ess p H χ q by an explicit construction ofWeyl sequences with positive energies, as r , `8q belongs to the essential spectrum of H per ,L and H per ,R . Wenext prove for any λ ă such that λ P σ ess p H per ,L q Y σ ess p H per ,R q , λ also belongs to σ ess p H χ q . Finally weprove that σ ess p H χ q in included in σ ess p H per ,L q Y σ ess p H per ,R q by a rather technical construction of a Weylsequence. The essential spectrum σ ess p H χ q contains r , `8q . Recall that a sequence t ψ n u n P N ˚ is a Weyl sequenceof an operator O on L p R q associated with λ P R if the following properties are satisfied:• for all n , ψ n is in the domain of the operator O and (cid:107) ψ n (cid:107) L p R q “ ;• the sequence ψ n á weakly in L p R q ;• lim n Ñ`8 (cid:107) p O ´ λ q ψ n (cid:107) L p R q “ .Consider a C c p R q function f p z q supported on r , s with ş R | f | “ , and g P C c p R q supported on the unitdisk centered at and such that ş R | g | “ . For any λ ą , consider a sequence of functions t ψ n u n P N ˚ definedas follows: ψ n p x, y, z q : “ n ´ { e i ? λz f pp z ´ n q{ n q g p x { n, y { n q .
30t is easy to see that ş R | ψ n | “ for all n P N ˚ , and that ψ n tends weakly to in L p R q . On the other hand p H χ ´ λ q ψ n p x, y, z q “ ´ ´ n ´ f pp z ´ n q{ n q ´ ? λn ´ f pp z ´ n q{ n q ¯ n ´ { e i ? λz g p x { n, y { n q´ n ´ { e i ? λz f pp z ´ n q{ n q p ∆ g q p x { n, y { n q ` V χ ψ n p x, y, z q . Recall also that by the results of Theorem 2.7, there exists for any (cid:15) ą an integer N (cid:15) such that | V χ p x, y, n q| ď (cid:15) when n ě N (cid:15) . Therefore, for n ě N (cid:15) , there exists a positive constant C such that (cid:107) p H χ ´ λ q ψ n (cid:107) L p R q ď n (cid:107) f (cid:107) L p R q (cid:107) g (cid:107) L p R q ` ? λn (cid:107) f (cid:107) L p R q (cid:107) g (cid:107) L p R q ` n (cid:107) f (cid:107) L p R q (cid:107) ∆ g (cid:107) L p R q ` ˆ n ´ ż R ˆż n n | V χ p x, y, z q f pp z ´ n q{ n q| dz ˙ | g p x { n, y { n q| dx dy ˙ { ď Cn ` ˆż R ż | V χ p nx, ny, n ` nz q f p z q| | g p x, y q| dz dx dy ˙ { ď Cn ` (cid:15) (cid:107) f (cid:107) L p R q (cid:107) g (cid:107) L p R q “ Cn ` (cid:15). This shows that t ψ n u n P N ˚ is a Weyl sequence of H χ associated with λ ą . Since that is an accumulationpoint of σ ess p H χ q , it holds r , `8q Ă σ ess p H χ q . The union of σ ess p H per ,L q Y σ ess p H per ,R q is included in σ ess p H χ q . Without loss of generality we provethat a negative λ L belonging to σ ess p H per ,L q also belongs to σ ess p H χ q . Consider a Weyl sequence t w n u n P N ˚ for H per ,L associated with λ L . Let us construct a Weyl sequence for H χ from t w n u n P N ˚ . Fix n P N ˚ , there existsa sequence t v k,n u k P N ˚ belonging to C c p R q such that for all ε ą , there exists a K n P N ˚ such that for any k ě K n , (cid:107) v k,n ´ w n (cid:107) H p R q ď ε. (4.38)It is easy to see that v K n ,n tends weakly to in L p R q as n Ñ 8 since w n converges weakly to . As v K n ,n has compact support, for any fixed n P N ˚ and for m P N ˚ large enough, supp ` τ xa L m v K n ,n ˘ č ´` r´ a L { , `8q ˆ R ˘ ď B n ¯ “ H , (4.39)where B n denotes the ball of radius n centered at in R . Remark that the above equality also ensures that τ xa L m v K n ,n tends weakly to in L p R q when m Ñ `8 for n fixed. In view of (4.38) and (4.39), we introduce Ă w n : “ τ xa L m n v K n ,n for n P N ˚ so that (4.39) is satisfied. This implies that Ă w n tends weakly to in L p R q when n Ñ `8 . Moreover, in view of (4.38) and by the definition of the Weyl sequence (cid:13)(cid:13)(cid:13)(cid:13) p H χ ´ λ L q Ă w n (cid:13)(cid:13)(cid:13)(cid:13) L “ (cid:13)(cid:13)(cid:13)(cid:13) p H χ ´ λ L q τ xa L m n v K n ,n (cid:13)(cid:13)(cid:13)(cid:13) L “ (cid:13)(cid:13)(cid:13)(cid:13) p H per ,L ´ λ L q τ xa L m n v K n ,n (cid:13)(cid:13)(cid:13)(cid:13) L ď (cid:13)(cid:13)(cid:13)(cid:13) τ xa L m n p H per ,L ´ λ L q p v K n ,n ´ w n q (cid:13)(cid:13)(cid:13)(cid:13) L ` (cid:13)(cid:13)(cid:13)(cid:13) τ xa L m n p H per ,L ´ λ L q w n (cid:13)(cid:13)(cid:13)(cid:13) L ď p ` (cid:107) V per ,L (cid:107) L ` | λ L |q (cid:13)(cid:13)(cid:13)(cid:13) v K n ,n ´ w n (cid:13)(cid:13)(cid:13)(cid:13) H ` (cid:13)(cid:13)(cid:13)(cid:13) p H per ,L ´ λ L q w n (cid:13)(cid:13)(cid:13)(cid:13) L ÝÝÝÝÑ n Ñ`8 . Therefore the sequence Ă w n { (cid:107) Ă w n (cid:107) L is a Weyl sequence of H χ associated with λ L . This leads to the conclusionthat σ ess p H per ,L q Y σ ess p H per ,R q Ď σ ess p H χ q . (4.40) The essential spectrum σ ess p H χ q in included in σ ess p H per ,L q Y σ ess p H per ,R q . We prove that for strictlynegative λ P σ ess p H χ q , it holds that λ P σ ess p H per ,L q Y σ ess p H per ,R q . The main technique is to use spreadingsequences (Zhislin sequences) [32, Definition 5.12], which are special Weyl sequences for which the supports ofthe functions move off to infinity. More precisely, a sequence t ψ n u n P N ˚ is a spreading sequence of an operator O on L p R q associated with λ if the following properties are satisfied:• for all n , ψ n is in the domain of the operator O and (cid:107) ψ n (cid:107) L p R q “ ;31 for any bounded set G Ă R , supp p ψ n q X G “ H for n sufficiently large. As a consequence, ψ n á weakly in L p R q ;• lim n Ñ`8 (cid:107) p O ´ λ q ψ n (cid:107) L p R q “ .As V per ,L and V per ,R are continuous and bounded by Theorem 2.7, it holds by [32, Theorem 5.14] that σ ess p H θ q “ t λ P C | there is a spreading sequence for H θ and λ u with H θ being H χ , H per ,L or H per ,R . The results of Theorem 2.7 also imply that, for any ε ą , there exists aconstant R ε such that max ´ (cid:107) V per ,L | r |ą R ε (cid:107) L p R q , (cid:107) V per ,R | r |ą R ε (cid:107) L p R q ¯ ă ε. (4.41)Consider a spreading sequence t φ n u n P N ˚ for H χ associated with λ ă . For all n P N ˚ , either (cid:107) φ n (cid:107) L pp , `8qˆ R q ě { or (cid:107) φ n (cid:107) L pp´8 , sˆ R q ě { . Without loss of generality, we assume in the sequel that there exists a sub-sequence t φ n u n P N ˚ such that, for n sufficiently large, (cid:107) φ n (cid:107) L pp , `8qˆ R q ě { . We next construct a Weylsequence of H per ,R from t φ n u n P N ˚ by constructing a special cut-off function ρ which has a non-trivial mass on p , `8q ˆ R , and whose derivatives decay rapidly: ρ p x q : “ ş x ´8 η p y q dy (cid:107) η (cid:107) L p R q , where η satisfies the following one-dimensional Yukawa equation ´ η ´ λη “ e ´ ?´ λ | x | . (4.42)The following lemma summarizes some properties of the cut-off function ρ . Lemma 4.5.
It holds that ď ρ ď and lim n Ñ8 (cid:107) ρφ n (cid:107) L ‰ . Moreover, (cid:107) ρ φ n ` ρ B x φ n (cid:107) L ÝÝÝÑ n Ñ8 , ρχ p V per ,L ´ V per ,R q P L p R q . (4.43)We postpone the proof of this lemma to Section 4.9. Define w n : “ ρφ n { (cid:107) ρφ n (cid:107) L . Let us show that t w n u n P N ˚ is a spreading sequence of H per ,R associated with λ . First of all, the sequence t w n u n P N ˚ is well defined at leastfor large n as lim n Ñ8 (cid:107) ρφ n (cid:107) L ‰ . It is also easy to see that (cid:107) w n (cid:107) L “ for all n P N ˚ . For any bounded set G P R , it holds supp p w n q X G “ H for n sufficiently large as t φ n u n P N ˚ is a spreading sequence. Note that p H per ,R ´ λ q w n “ (cid:107) ρφ n (cid:107) L p ρ p H χ ´ λ q φ n ` A χ φ n q , (4.44)where A χ : “ ´ ρ ´ ρ B x ` ρχ p V per ,L ´ V per ,R q . As lim n Ñ8 (cid:107) p H χ ´ λ q φ n (cid:107) L “ by definition of thespreading sequence, it follows that lim n Ñ8 (cid:107) ρ p H χ ´ λ q φ n (cid:107) L ď lim n Ñ8 (cid:107) p H χ ´ λ q φ n (cid:107) L “ . It thereforesuffices to prove that (possibly up to extraction) lim n Ñ8 (cid:107) A χ φ n (cid:107) L “ . By the Kato–Seiler–Simon inequality(4.1) we obtain that (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) p ´ ∆ q ´ ρχ p V per ,L ´ V per ,R q (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ď ? π (cid:13)(cid:13)(cid:13)(cid:13) ρχ p V per ,L ´ V per ,R q (cid:13)(cid:13)(cid:13)(cid:13) L p R q . In particular ρχ p V per ,L ´ V per ,R q is ´ ∆ –compact, hence H per ,R –compact by the boundedness of V per ,R . Asthe sequence H per ,R w n is bounded in L p R q , the H per ,R –compactness of ρχ p V per ,L ´ V per ,R q implies that ρχ p V per ,L ´ V per ,R q w n converges strongly to some function v P L p R q . On the other hand, for any f P L p R q , p v, f q L “ lim n Ñ`8 ` ρχ p V per ,L ´ V per ,R q w n , f ˘ L “ lim n Ñ`8 ` w n , ρχ p V per ,L ´ V per ,R q f ˘ L “ , where we have used the fact that w n á and ρχ p V per ,L ´ V per ,R q P L p R q so ρχ p V per ,L ´ V per ,R q f P L p R q . Therefore by the uniqueness of weak limit, ρχ p V per ,L ´ V per ,R q w n converges strongly to v ” .Together with (4.43) we conclude that lim n Ñ8 (cid:107) Aφ n (cid:107) L “ . Therefore in view of (4.44), it holds lim n Ñ8 (cid:107) p H per ,R ´ λ q w n (cid:107) L “ . t w n u n P N is a spreading sequence of H per ,R associated with λ ă . This implies that for any spreadingsequence of H χ associated with λ ă , we can construct a spreading sequence for either H per ,L or H per ,R associated with λ ă , depending on whether (up to extraction) the spreading sequence of H χ has non-trivialmass on ` p´8 , s ˆ R ˘ or ` p , `8q ˆ R ˘ . This allows us to conclude that σ ess p H χ q Ď σ ess p H per ,L q Y σ ess p H per ,R q . (4.45)By gathering (4.40) and (4.45) we conclude that σ ess p H χ q ” σ ess p H per ,L q Y σ ess p H per ,R q . In particular, σ ess p H χ q is independent of the function χ P X . The solution of the one-dimensional Yukawa equation (4.42) is ă η p x q “ ż R e ´?´ λ | x ´ y | ?´ λ e ´ ?´ λ | y | dy ď ż R e ´?´ λ p| x |`| y |q ?´ λ dy ď ´ λ e ´?´ λ | x | . (4.46)This implies that η belongs to L p p R q for ď p ď `8 . As η ą is integrable, the cut-off function ρ p x q “ ş x ´8 η p y q dy (cid:107) η (cid:107) L p R q is well defined. It is easy to see that ď ρ ď . Since (cid:107) φ n (cid:107) L pp , `8qˆ R q ě { for n large enoughand ρ p x q ě for x ě , hence lim n Ñ8 (cid:107) ρφ n (cid:107) L p R q ‰ .Let us next prove that (cid:107) ρ φ n ` ρ B x φ n (cid:107) L ÝÝÝÑ n Ñ8 . In view of (4.46), ˇˇ η p x q ˇˇ “ ˇˇˇˇż e ´?´ λ | x ´ y |´ ?´ λ | y | p x ď y ´ x ą y q dy ˇˇˇˇ ď ż `8´8 e ´?´ λ | x ´ y |´ ?´ λ | y | dy “ ?´ λη p x q ď ?´ λ e ´?´ λ | x | . Combining the above inequality with (4.46) we obtain that (cid:107) ρ φ n ` ρ B x φ n (cid:107) L p R q “ (cid:107) η φ n ` η B x φ n (cid:107) L p R q (cid:107) η (cid:107) L p R q ď (cid:107) η (cid:107) L p R q ż R ´` η ˘ | φ n | ` η |B x φ n | ¯ ď | λ | (cid:107) η (cid:107) L p R q ż R e ´ ?´ λ |¨| | φ n | ` (cid:107) η (cid:107) L (cid:107) η (cid:107) L p R q ż R η |B x φ n | . (4.47)It suffices to prove that each of the previous integrals tends to when n Ñ 8 . Remark that the integrandsof these terms appear when one does an explicit calculation of p η p´ ∆ ´ λ q φ n , p´ ∆ ´ λ q φ n q L p R q . Let ustherefore first prove that p η p´ ∆ ´ λ q φ n , p´ ∆ ´ λ q φ n q L p R q ÝÝÝÑ n Ñ8 . For this purpose, we prove that (cid:107) η p´ ∆ ´ λ q φ n (cid:107) L p R q ÝÝÝÑ n Ñ8 strongly and p´ ∆ ´ λ q φ n ÝÝÝá n Ñ8 weakly in L p R q . In view of (4.46), for any ε ą there exists x ε ą such that when | x | ą x ε , ď η p x q ď ε . Togetherwith (4.41) and the fact that t φ n u n P N ˚ is a spreading sequence, it holds (cid:107) ηV χ φ n (cid:107) L p R q “ ż R ˆt| r |ą R ε u | ηV χ φ n | ` ż r´ x ,x sˆt| r |ď R ε u | ηV χ φ n | ` ż r´ x ,x s A ˆt| r |ď R ε u | ηV χ φ n | ď ε (cid:107) η (cid:107) L p R q ` (cid:107) ηV χ (cid:107) L p R q (cid:13)(cid:13)(cid:13)(cid:13) φ n r´ x ,x sˆt| r |ď R ε u (cid:13)(cid:13)(cid:13)(cid:13) L p R q ` ε (cid:107) V χ (cid:107) L p R q ÝÝÝÑ n Ñ8 . The above convergence allows us to conclude that (cid:107) η p´ ∆ ´ λ q φ n (cid:107) L p R q ď (cid:107) η p H χ ´ λ q φ n (cid:107) L p R q ` (cid:107) ηV χ φ n (cid:107) L p R q ÝÝÝÑ n Ñ8 . (4.48)Moreover, φ n P H p R q ã Ñ L p R q with continuous embedding for all n P N ˚ . Approximating φ n by C c p R q functions we deduce that lim | x |Ñ8 | φ n p x q| “ . Furthermore, remark that for any ψ P C c p R q it holds pp´ ∆ ´ λ q φ n , ψ q L p R q “ p φ n , p´ ∆ ´ λ q ψ q L p R q ÝÝÝÑ n Ñ8 , F c Σ a Σ b ´ σ ess p H χ q “ σ ess p H per , L q Y σ ess p H per , R q C Figure 6: The essential spectrum of H χ , H per ,L and H per ,R , and the contour C .which implies the weak convergence p´ ∆ ´ λ q φ n ÝÝÝá n Ñ8 by the density of C c p R q in L p R q . Together withthe strong convergence (4.48) and an integration by parts, this leads to, p η p´ ∆ ´ λ q φ n , p´ ∆ ´ λ q φ n q L p R q “ ż R η | ∆ φ n | ` ηλ ∆ φ n φ n ` λη ∆ φ n φ n ` ηλ | φ n | “ ż R η | ∆ φ n | ´ λη B x ´ | φ n | ¯ ´ ηλ | ∇ φ n | ` λ η | φ n | “ ż R η | ∆ φ n | ` ` λη ` λ η ˘ | φ n | ` η | λ | | ∇ φ n | ÝÝÝÑ n Ñ8 . In view of (4.42) we have ` λη ` λ η ˘ “ ´ λ e ´ ?´ λ |¨| ą , hence the integrand of the above integral is positive.This implies that ż R e ´ ?´ λ |¨| | φ n | ÝÝÝÑ n Ñ8 , ż R η | ∇ φ n | ÝÝÝÑ n Ñ8 . (4.49)In view of (4.47) and (4.49), we deduce that (cid:107) ρ φ n ` ρ B x φ n (cid:107) L p R q ÝÝÝÑ n Ñ8 . Let us finally prove that ρχ p V per ,L ´ V per ,R q P L p R q . By definition of χ , the function ρχ has support in p´8 , a R { s and is equal to ρ p x q when x P p´8 , ´ a L { q .Remark also that (4.46) implies that @ x ă , ρ p x q ď (cid:107) η (cid:107) L ż x ´8 | λ | e ´?´ λ | y | dy ď | λ | { (cid:107) η (cid:107) L e ?´ λx . Hence, as V per ,L P L s per ,x p Γ L q for ă s ď 8 by Theorem 2.7, ż R p ρχ V per ,L q “ ż p´8 , ´ a L { qˆ R p ρV per ,L q ` ż r´ a L { ,a R { sˆ R p ρχ V per ,L q ď (cid:107) V per ,L (cid:107) L p Γ aL q | λ | (cid:107) η (cid:107) L p R q `8 ÿ n “ e ´ ?´ λa L n ` ż r´ a L { ,a R { sˆ R V ,L ă `8 . This implies that ρχ V per ,L P L p R q . Similar arguments show that ρχ V per ,R P L p R q , which concludes theproof of the lemma. In view of Proposition 3.1, consider a contour C in the complex plan enclosing the spectrum of H χ below theFermi level (cid:15) F without intersecting it, crossing the real axis at c ă inf t´ (cid:107) V per ,L (cid:107) L , ´ (cid:107) V per ,R (cid:107) L u (See Fig.6). This is possible even if (cid:15) F is an eigenvalue: one can always slightly move the curve C below (cid:15) F in orderbypass (cid:15) F but still enclose all the spectrum of H χ below (cid:15) F . Let us introduce the following estimates, which areuseful to characterize the decay property of densities. Lemma 4.6 (Combes-Thomas estimate [45, 22, 31]) . Consider H : “ ´ ∆ ` V with V P L p R q . Let p, q bepositive integers such that pq ą { . Then there exists ε ą and a positive constant C p p, q q such that for any ζ R σ p H q , and any p α, β q P Z ˆ Z , (cid:13)(cid:13)(cid:13)(cid:13) w α p ζ ´ H q ´ p w β (cid:13)(cid:13)(cid:13)(cid:13) S q ď C p p, q q ˆ ` θ p ζ, V q ˙ p e ´ εθ p ζ,V q| α ´ β | , (4.50) where θ p ζ, V q “ dist p ζ, σ p H qq| ζ | ` (cid:107) V (cid:107) L ` . V χ belongs to L p R q , the following lemma is a direct adaption of [15, Lemma 1]: Lemma 4.7.
Under Assumption 2, there exist two positive constants c , c such that @ ζ P C , c p ´ ∆ q ď | H χ ´ ζ | ď c p ´ ∆ q as operators on L p R q . In particular (cid:13)(cid:13)(cid:13)(cid:13) | H χ ´ ζ | { p ´ ∆ q ´ { (cid:13)(cid:13)(cid:13)(cid:13) ď ? c , (cid:13)(cid:13)(cid:13)(cid:13) | H χ ´ ζ | ´ { p ´ ∆ q { (cid:13)(cid:13)(cid:13)(cid:13) ď ? c . Moreover, p H χ ´ ζ qp ´ ∆ q ´ and its inverse are bounded operators. Let us turn to the proof of Proposition 3.2. First of all let us show that γ χ is locally trace class. Consider (cid:37) P C c p R q . Remark that γ χ is a spectral projector. In view of Lemma 4.7, by Cauchy’s resolvent formula andthe Kato–Seiler–Simon inequality (4.1), there exists a positive constant C χ such that (cid:107) (cid:37)γ χ (cid:37) (cid:107) S “ (cid:107) (cid:37)γ χ γ χ (cid:37) (cid:107) S “ (cid:107) (cid:37)γ χ (cid:107) S “ (cid:13)(cid:13)(cid:13)(cid:13) (cid:37) ¿ C π ζ ´ H χ dζ (cid:13)(cid:13)(cid:13)(cid:13) S ď C χ (cid:13)(cid:13)(cid:13)(cid:13) (cid:37) ´ ∆ (cid:13)(cid:13)(cid:13)(cid:13) S ď C χ π (cid:107) (cid:37) (cid:107) L p R q . This implies that γ χ is locally trace class so that its density ρ χ is well defined in L p R q . Let us provethat χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ belongs to L p p R q for ă p ď . It is difficult to directly comparethe difference of χ ρ per ,L ` p ´ χ q ρ per ,R and ρ χ . We construct to this end a density operator γ d whose density ρ d is equal to χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ : γ d : “ γ d, ` γ d, , γ d, : “ χ p γ per ,L ´ γ χ q χ, γ d, : “ a ´ χ p γ per ,R ´ γ χ q a ´ χ . (4.51)Remark that if γ d P S , then Tr L p R q p γ d q “ χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ . The density ρ d is in L p p R q for ă p ď . The proof that ρ d P L p p R q relies on duality arguments:denoting by q “ pp ´ P r , `8q , we prove that for any W P L q p R q there exists some K q ą such that | Tr L p R q p γ d W q| ď K q (cid:107) W (cid:107) L q . By Cauchy’s formula we have γ d, “ π i ¿ C χ ˆ z ´ H per ,L ´ ζ ´ H χ ˙ χ dζ,γ d, “ π i ¿ C a ´ χ ˆ ζ ´ H per ,R ´ ζ ´ H χ ˙ a ´ χ dζ. (4.52)Let us prove that there exists K q ą such that ˇˇ Tr L p R q p γ d, W q ˇˇ ď K q (cid:107) W (cid:107) L q . It is easily shown that a similar inequality holds for γ d, . Denote by V d : “ p ´ χ qp V per ,L ´ V per ,R q P L p R q .Remark that the function V d χ has compact support in the x -direction and that V d χ belongs to L r p R q for ă r ď `8 by Theorem 2.7. For any ζ P C , the integrand of γ d, writes D p ζ q : “ χ ˆ ζ ´ H per ,L ´ ζ ´ H χ ˙ χ “ χ ζ ´ H per ,L V d ζ ´ H χ χ. Remark that χ being translation-invariant in the r -direction, it is not in any L p space in R , which prevents usfrom using the standard techniques such as calculating the commutator r´ ∆ , χ s to give Schatten class estimateson γ d, . By writing “ γ per ,L ` γ K per ,L and “ γ χ ` γ K χ , the following decomposition holds D p ζ q “ χ γ per ,L ζ ´ H per ,L V d ζ ´ H χ χ ` χ γ K per ,L ζ ´ H per ,L V d γ χ ζ ´ H χ χ ` χ γ K per ,L ζ ´ H per ,L V d γ K χ ζ ´ H χ χ. (4.53)By the residue theorem, ż C χ γ K per ,L ζ ´ H per ,L V d γ K χ ζ ´ H χ χ dζ ” . (4.54)To estimate other terms in (4.53) we rely on the following Lemmas 4.8 and 4.9.35 emma 4.8. Consider a self-adjoint operator H “ ´ ∆ ` V defined on L p R q with domain H p R q and V P L p R q . For E P R z σ p H q denote by γ “ p´8 ,E s p H q . Then for any a, b P R , the operator p ´ ∆ q a γ p ´ ∆ q b isbounded. Moreover, if γ P S k for some k ě , then p ´ ∆ q a γ p ´ ∆ q b P S k .Proof. Similarly as in Lemma 4.7 it can be shown that for any ζ P R z σ p H q the operator p ζ ´ H q ´ a p ´ ∆ q a andits inverse are bounded. Fix δ ą and define λ : “ ´ (cid:107) V (cid:107) L ´ δ . Then λ R σ p H q . By writing γ “ γ , thereexists a positive constant C such that (cid:13)(cid:13)(cid:13)(cid:13) p ´ ∆ q a γ p ´ ∆ q b (cid:13)(cid:13)(cid:13)(cid:13) S k ď C (cid:13)(cid:13)(cid:13)(cid:13) p λ ´ H q a γ p λ ´ H q b (cid:13)(cid:13)(cid:13)(cid:13) S k “ C (cid:13)(cid:13)(cid:13)(cid:13) γ p λ ´ H q a ` b (cid:13)(cid:13)(cid:13)(cid:13) S k ă `8 , as γ P S k and γ p λ ´ H q a ` b is a bounded operator. The proof of the boundedness in operator norm follows thesame lines. Lemma 4.9.
For any ă p ď , there exist positive constants d p, and d p, , such that @ ζ P C , (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) χ γ per ,L ζ ´ H per ,L V d (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S p ď d p, (cid:107) V d χ (cid:107) L p p R q , (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) V d γ χ ζ ´ H χ χ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S p ď d p, (cid:107) V d χ (cid:107) L p p R q . (4.55) Proof.
Let us prove the statement for χγ per ,L p ζ ´ H per ,L q ´ V d , the proof of the bound of V d γ χ p ζ ´ H χ q ´ χ follows similar arguments. Fix R ą . Recall that B R is the ball in R centered at with radius R . Denoteby ϕ R the characteristic function of B R . For any R ą , by the Kato–Seiler–Simon inequality (4.1) and theboundedness of p ´ ∆ qp ζ ´ H per ,L q ´ it is easy to see that ϕ R χ p ζ ´ H per ,L q ´ and p ζ ´ H per ,L q ´ V d ϕ R belongto S . The operator γ per ,L p ζ ´ H per ,L q m is a bounded for any m P R in view of Lemma 4.9. Therefore ϕ R ˆ χ γ per ,L ζ ´ H per ,L V d ˙ ϕ R “ ˆ ϕ R χ ζ ´ H per ,L ˙ γ per ,L p ζ ´ H per ,L q ˆ ζ ´ H per ,L V d ϕ R ˙ P S . Let us first prove that for any ď p ď , there exists a positive constant d p, only depending on p such that forany R ą , (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ϕ R χ γ per ,L ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S p ď d p, (cid:107) V d ϕ R χ (cid:107) L p p R q . (4.56)We first prove (4.56) for p “ and p “ , and conclude by an interpolation argument for ď p ď . Consider p “ . By the cyclicity of the trace and the Kato–Seiler–Simon inequality (4.1), (cid:13)(cid:13)(cid:13)(cid:13) ϕ R χ γ per ,L ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) ϕ R χ ζ ´ H per ,L γ per ,L p ζ ´ H per ,L q ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) γ per ,L p ζ ´ H per ,L q ζ ´ H per ,L V d ϕ R χ ζ ´ H per ,L (cid:13)(cid:13)(cid:13)(cid:13) S ď c (cid:13)(cid:13)(cid:13)(cid:13) | ´ ∆ | ˇˇ V d ϕ R χ ˇˇ | ´ ∆ | (cid:13)(cid:13)(cid:13)(cid:13) S “ c (cid:13)(cid:13)(cid:13)(cid:13) | ´ ∆ | ˇˇ V d ϕ R χ ˇˇ { (cid:13)(cid:13)(cid:13)(cid:13) S ď d , (cid:107) V d ϕ R χ (cid:107) L . Let us next prove (4.56) for p “ . Use again the cyclicity of the trace and the Kato–Seiler–Simon inequality (4.1), (cid:13)(cid:13)(cid:13)(cid:13) ϕ R χ γ per ,L ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) V d ϕ R γ per ,L ζ ´ H per ,L ϕ R χ γ per ,L ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) γ per ,L ζ ´ H per ,L ϕ R χ γ per ,L ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13) S ď c (cid:13)(cid:13)(cid:13)(cid:13) ϕ R χ ζ ´ H per ,L γ per ,L p ζ ´ H per ,L q ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13) S “ c (cid:13)(cid:13)(cid:13)(cid:13) γ per ,L p ζ ´ H per ,L q ζ ´ H per ,L V d ϕ R χ ζ ´ H per ,L (cid:13)(cid:13)(cid:13)(cid:13) S ď c (cid:13)(cid:13)(cid:13)(cid:13) ζ ´ H per ,L V d ϕ R χ ζ ´ H per ,L (cid:13)(cid:13)(cid:13)(cid:13) S “ c (cid:13)(cid:13)(cid:13)(cid:13) ˇˇ V d ϕ R χ ˇˇ ζ ´ H per ,L (cid:13)(cid:13)(cid:13)(cid:13) S ď d , (cid:107) V d ϕ R χ (cid:107) L .
36y the interpolation arguments we can conclude (4.56) for ď p ď . Remark that for ă p ď the followinguniform bound holds: (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ϕ R χ γ per ,L ζ ´ H per ,L V d ϕ R (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S p ď d p, (cid:13)(cid:13) V d ϕ R χ (cid:13)(cid:13) L p p R q ď d p, (cid:107) V d χ (cid:107) L p p R q . By passing the limit R Ñ `8 we can conclude the proof.Consider W P L q p R q for q “ pp ´ P r , `8q . In view of (4.53), (4.54) and (4.55), by manipulations similarto the ones used in the proof of Lemma 4.9, and the H¨older’s inequality for Schatten class operators (see forexample [61, Proposition 5]), we obtain that (cid:107) γ d, W (cid:107) S “ π (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ¿ C D p ζ q dζW (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ¿ C χ γ per ,L ζ ´ H per ,L V d ζ ´ H χ χW ` χ γ K per ,L ζ ´ H per ,L V d γ χ ζ ´ H χ χW dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ¿ C ˆ χ γ per ,L ζ ´ H per ,L V d ˙ ζ ´ H χ p ´ ∆ q ˆ ´ ∆ χW ˙ dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ¿ C ˆ W χ ´ ∆ ˙ p ´ ∆ q γ K per ,L ζ ´ H per ,L ˆ V d γ χ ζ ´ H χ χ ˙ dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ď C (cid:107) V d χ (cid:107) L p p R q (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) ´ ∆ χW (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S q ď K q (cid:107) W (cid:107) L q p R q , (4.57)where we have used the Kato–Seiler–Simon inequality (4.1) as well as the fact that (cid:107) χ (cid:107) L “ . Similar estimateshold for γ d, . We therefore can conclude that ρ d “ χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ belongs to L p p R q for ă p ď . Decay rate in the x -direction. Let us show that the density difference χ ρ per ,L ` p ´ χ q ρ per ,R ´ ρ χ decaysexponentially fast in the x -direction. Note that there exists N L P N such that N L ´ ă a L { ď N L . Denoteby D a L : “ r´ a L { , `8q ˆ R , we prove the exponential decay when supp p w α q Ă R z D a L . Denote by α : “ p α x , , q P p R , , q , β “ p β x , β y , β z q P Z , β x ě ´ N L . We have D aL ¨˝ `8 ÿ β x ě´ N L ÿ β y ,β z P Z w β ˛‚ “ D aL , D aL V d “ V d , α x ă ´ a L ă ´ N L ` ď β x ` . The above relations imply, together with (4.55), the Combes–Thomas estimate (4.50) and arguments similar toones used in (4.57), that there exist a positive constants C and t such that, for ă p ď and q “ pp ´ ě , (cid:107) w α γ d w α (cid:107) S “ (cid:107) w α γ d, w α (cid:107) S “ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) π i ¿ C ˜ w α χ γ per ,L ζ ´ H per ,L V d ζ ´ H χ χ w α ` w α χ γ K per ,L ζ ´ H per ,L V d γ χ ζ ´ H χ χ w α ¸ dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) π i ¿ C ˆ w α χ γ per ,L ζ ´ H per ,L V d ˙ ˆ D aL ζ ´ H χ χ w α ˙ dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) π i ¿ C ˜ w α χ γ K per ,L ζ ´ H per ,L D aL ¸ ˆ V d γ χ ζ ´ H χ χ w α ˙ dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ď K `8 ÿ β x ě´ N L `8 ÿ β y ,β z P Z e ´ t p β x ´ α x q e ´ t | β y | e ´ t | β z | ď C e ´ t | α x | . The last step relies on the uniform distance of ζ P C to σ p H χ q and σ p H per ,L q . Similar estimates hold when thesupport of w α is in r a R { , `8q ˆ R . There exist therefore positive constants C and t such that (cid:107) w α γ d w α (cid:107) S “ ş R | w α ρ d w α | ď C e ´ t | α | , which concludes the proof. 37 .11 Proof of Lemma 3.3 From the last item of the Theorem 2.7 we know that V per ,L P L p per ,x p Γ L q (resp. V per ,R P L p per ,x p Γ R q ) for ă p ď `8 . Remark also that B x p χ q , B x p χ q are uniformly bounded and have support in r´ a L { , a R { s ˆ R . Ittherefore suffices to obtain the L p -estimates on B x V per ,L and B x V per ,R . We treat B x V per ,L , the L p -estimates of B x V per ,R following similar arguments. First of all in view of the form of the minimizer (4.32), by the Cauchy–Schwarz inequality B x ρ per ,L “ B x ˜ π ż Γ ˚ L ÿ n ě p λ n p ξ q ď (cid:15) L q | e n p ξ, ¨q| dξ ¸ ď π ż Γ ˚ L ˜ ÿ n ě p λ n p ξ q ď (cid:15) L q |B x | e n |p ξ, ¨q| ¸ { ˜ ÿ n ě p λ n p ξ q ď (cid:15) L q | e n p ξ, ¨q| ¸ { dξ ď π a K ξ,L ? ρ per ,L , where K ξ,L p x q : “ ş Γ ˚ L ř n ě p λ n p ξ q ď (cid:15) L q |B x e n p ξ, x q dξ | . We also have used the fact that | ∇ | f || ď | ∇ f | forany complex-valued function f . In view of the potential decomposition (4.34), the term T p r q does not contributeto the x -directional derivative, hence |B x V per ,L | “ ˇˇˇ pB x p ρ per ,L ´ µ per ,L qq ‹ Ą G a L ˇˇˇ ď ˆ π a K ξ,L ? ρ per ,L ` |B x µ per ,L | ˙ ‹ ˇˇˇ Ą G a L ˇˇˇ . On the other hand, finite kinetic energy condition (2.10) implies that K ξ,L P L ,x p Γ L q . Moreover, ? ρ per ,L belongs to H ,x p Γ L q hence to L s per ,x p Γ L q for ď s ď . Therefore, by H¨older’s inequality, for p, m ě : ż Γ L p K ξ,L ρ per ,L q p { ď ˆż Γ L K pm { ξ,L ˙ { m ˆż Γ L ρ pm {p p m ´ qq per ,L ˙ p m ´ q{ m , with the conditions pm “ and ď pm {p m ´ q ď . This is the case for { ď m ď and ď p ď { sothat p K ξ,L ρ per ,L q { belongs to L p per ,x p Γ L q for ď p ď { . As B x µ per ,L is in L p per ,x p Γ L q for any ď p ď `8 and Ą G a L P L q per ,x p Γ L q for ď q ă by Lemma 2.3, we obtain by Young’s convolution inequality that B x V per ,L P L s per ,x p Γ L q for ď s ă . This allows us to conclude the lemma. The following statements assure that the problem (3.12) is well-defined: a duality argument shows that densitiesof operators in Q χ are well defined. The energy functional (3.13) is well defined and is bounded from below. Theproofs are direct adaptations of [15, Proposition 1, Lemma 2, Corollary 1 and Corollary 2].1. For any Q χ P Q χ , it holds that Q χ W P S γ χ for W “ W ` W P C ` L p R q . Moreover, there exists apositive constant C χ such that: | Tr γ χ p Q χ W q | ď C χ (cid:107) Q χ (cid:107) Q χ p (cid:107) W (cid:107) C ` (cid:107) W (cid:107) L p R q q . Moreover, there exists a uniquely defined function ρ Q χ P C Ş L p R q such that @ W “ W ` W P C ` L p R q , Tr γ χ p Q χ W q “ x ρ Q χ , W y C , C ` ż R ρ Q χ W . The linear map Q χ P Q χ ÞÑ ρ Q χ P C Ş L p R q is continuous: (cid:107) ρ Q χ (cid:107) C ` (cid:107) ρ Q χ (cid:107) L p R q ď C χ (cid:107) Q χ (cid:107) Q χ . Moreover, if Q χ P S Ă S γ χ , then ρ Q χ p x q “ Q χ p x , x q where Q χ p x , x q the integral kernel of Q χ .2. For any κ P p Σ N,χ , (cid:15) F q and any state Q χ P K χ , the following inequality holds ď c Tr ´ p ´ ∆ q { ` Q `` χ ´ Q ´´ χ ˘ p ´ ∆ q { ¯ ď Tr γ χ p H χ Q χ q ´ κ Tr γ χ p Q χ qď c Tr ´ p ´ ∆ q { p Q `` χ ´ Q ´´ χ qp ´ ∆ q { ¯ , where c and c are the same constants as in Lemma 4.7.38. Assume that Assumption 2 holds. There are positive constants r d , r d , such that E χ p Q χ q ´ κ Tr γ χ p Q χ q ě r d ` (cid:107) Q `` χ (cid:107) S ` (cid:107) Q ´´ χ (cid:107) S ` (cid:107) | ∇ | Q `` χ | ∇ | (cid:107) S ` (cid:107) | ∇ | Q ´´ χ | ∇ | (cid:107) S ˘ ` r d ` (cid:107) | ∇ | Q χ (cid:107) S ` (cid:107) Q χ (cid:107) S ˘ ´ D p ν χ , ν χ q . Hence E χ p¨q ´ κ Tr γ χ p¨q is bounded from below and coercive on K χ .The existence and the form of the minimizers are direct adaptations of [15, Theorem 2]. We prove this theorem by taking two arbitrary cut-off functions χ , χ belonging to X , and prove that ρ χ ` ρ Q χ “ ρ χ ` ρ Q χ . For i “ , , consider the reference states associated with the Hamiltonian H χ i . Denote by γ χ i the spectral projector of H χ i below (cid:15) F and by Q χ i the solutions of (3.14) associated with χ i . Consider a teststate r Q : “ γ χ ` Q χ ´ γ χ . (4.58)We show that r Q is a minimizer of the problem (3.12) associated with the cut-off function χ , so that ρ r Q ” ρ Q χ by the uniqueness of the density of the minimizer provided by Proposition 3.5. Note that Assumption 2 andProposition 3.1 guarantee that there is a common spectral gap for H χ i and σ ess p H χ q “ σ ess p H χ q . We firstshow that the test state r Q belongs to the convex set K χ : “ t Q P Q χ | ´ γ χ ď Q ď ´ γ χ u , hence is an admissible state for the minimization problem (3.12) associated with χ . We next show that r Q is aminimizer. The test state r Q belongs to K χ . We begin by proving that r Q is in Q χ . Let us prove that r Q is γ χ -traceclass. The following lemma will be useful. Lemma 4.10.
The difference of the spectral projectors γ χ ´ γ χ belongs to S γ χ . Moreover, | ∇ | p γ χ ´ γ χ q P S , p γ χ ´ γ χ q | ∇ | P S . (4.59) Proof.
By Cauchy’s resolvent formula and the Kato–Seiler–Simon inequality (4.1), (cid:107) γ χ ´ γ χ (cid:107) S “ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) π ¿ C p ζ ´ H χ q ´ p χ ´ χ qp V per ,L ´ V per ,R qp ζ ´ H χ q ´ dζ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) S ď C (cid:13)(cid:13)(cid:13)(cid:13) p ´ ∆ q ´ p χ ´ χ qp V per ,L ´ V per ,R q (cid:13)(cid:13)(cid:13)(cid:13) S ď C ? π (cid:13)(cid:13)(cid:13)(cid:13) p χ ´ χ qp V per ,L ´ V per ,R q (cid:13)(cid:13)(cid:13)(cid:13) L ă `8 . (4.60)The results of Lemma 4.7 imply that | ∇ |p ζ ´ H χ i q ´ is uniformly bounded with respect to ζ P C . By calculationssimilar to (4.60), (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | p γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ď c (cid:13)(cid:13)(cid:13)(cid:13) p χ ´ χ qp V per ,L ´ V per ,R qp ´ ∆ q ´ (cid:13)(cid:13)(cid:13)(cid:13) S ă `8 . (4.61)Hence p γ χ ´ γ χ q | ∇ | also belongs to S since it is the adjoint of | ∇ | p γ χ ´ γ χ q . On the other hand, as γ χ isa bounded operator, in view of (4.60) and by writing γ χ ´ γ χ “ γ K χ ´ γ K χ and using the fact that γ χ i ` γ K χ i “ , γ K χ p γ χ ´ γ χ q γ K χ “ γ K χ γ χ γ K χ “ p γ χ ´ γ χ q γ χ p γ χ ´ γ χ q P S ,γ χ p γ χ ´ γ χ q γ χ “ ´ γ χ γ K χ γ χ “ ´ p γ χ ´ γ χ q γ K χ p γ χ ´ γ χ q P S . (4.62)Together with (4.60) we conclude that γ χ ´ γ χ belongs to S γ χ .The following lemma is a consequence of [33, Lemma 1] and the fact that γ χ ´ γ χ P S .39 emma 4.11. Any self-adjoint operator A is in S γ χ if and only if A is in S γ χ . Moreover Tr γ χ p A q “ Tr γ χ p A q . The fact that Q χ P S γ χ implies that | ∇ | Q χ P S , and Q χ P S γ χ by Lemma 4.11. In view of thisand Lemma 4.10 we know that r Q “ γ χ ´ γ χ ` Q χ belongs to S γ χ . The inequality (4.61) implies that | ∇ | r Q “ | ∇ | Q χ `| ∇ |p γ χ ´ γ χ q P S . It remains to prove that | ∇ | γ K χ r Qγ K χ | ∇ | P S and | ∇ | γ χ r Qγ χ | ∇ | P S .In view of (4.58) we have | ∇ | γ K χ r Qγ K χ | ∇ | “ | ∇ | γ K χ Q χ γ K χ | ∇ | ` | ∇ | γ K χ p γ χ ´ γ χ q γ K χ | ∇ | , | ∇ | γ χ r Qγ χ | ∇ | “ | ∇ | γ χ Q χ γ χ | ∇ | ` | ∇ | γ χ p γ χ ´ γ χ q γ χ | ∇ | . (4.63)We estimate (4.63) term by term. By Lemma 4.8 we know that (cid:107) | ∇ | γ χ i (cid:107) ď (cid:107) | ∇ |p ´ ∆ q ´ (cid:107)(cid:107) p ´ ∆ q γ χ i (cid:107) ă 8 .Moreover, by writing γ K χ “ ´ γ χ ` γ χ ´ γ χ and γ χ “ γ χ ` γ χ ´ γ χ , in view of (4.61), (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ K χ Q χ γ K χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ K χ Q χ γ K χ | ∇ | ` | ∇ | γ K χ Q χ p γ χ ´ γ χ q| ∇ | ` | ∇ |p γ χ ´ γ χ q Q χ γ K χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ K χ Q χ γ K χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | Q χ p γ χ ´ γ χ q| ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ Q χ p γ χ ´ γ χ q| ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q Q χ γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q Q χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ K χ Q χ γ K χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | Q χ (cid:13)(cid:13)(cid:13)(cid:13) S (cid:13)(cid:13)(cid:13)(cid:13) p γ χ ´ γ χ q| ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) Q χ (cid:13)(cid:13)(cid:13)(cid:13) S (cid:13)(cid:13)(cid:13)(cid:13) p γ χ ´ γ χ q| ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) Q χ (cid:13)(cid:13)(cid:13)(cid:13) S (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) Q χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ă 8 , and similarly (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ Q χ γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ Q χ γ χ | ∇ | ` | ∇ | γ χ Q χ p γ χ ´ γ χ q| ∇ | ` | ∇ |p γ χ ´ γ χ q Q χ γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ Q χ γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ Q χ p γ χ ´ γ χ q| ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q Q χ γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ Q χ γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) Q χ (cid:13)(cid:13)(cid:13)(cid:13) S (cid:13)(cid:13)(cid:13)(cid:13) p γ χ ´ γ χ q| ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) Q χ (cid:13)(cid:13)(cid:13)(cid:13) S (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ă 8 , From (4.62) we know that (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ K χ p γ χ ´ γ χ q γ K χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | p γ χ ´ γ χ q γ χ p γ χ ´ γ χ q | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ă 8 , (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | γ χ p γ χ ´ γ χ q γ χ | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) | ∇ | p γ χ ´ γ χ q γ K χ p γ χ ´ γ χ q | ∇ | (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) | ∇ |p γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ă 8 . This shows that | ∇ | γ K χ r Qγ K χ | ∇ | P S and | ∇ | γ χ r Qγ χ | ∇ | P S . In view of (4.63), this allows us to concludethat r Q P Q χ . On the other hand, it is easy to see that ´ γ χ ď r Q “ γ χ ` Q χ ´ γ χ ď ´ γ χ , which showsthat r Q belongs to the convex set K χ . The state r Q is a minimizer. We now prove that r Q is a minimizer of the problem (3.12) associated with thecut-off function χ . As r Q P K χ , the fact that Q χ is a minimizer implies that E χ ´ r Q ¯ ´ κ Tr γ χ ´ r Q ¯ ě E χ p Q χ q ´ κ Tr γ χ p Q χ q . (4.64)Define Θ : “ r Q ´ Q χ “ Q χ ´ Q χ ` γ χ ´ γ χ . The inequality (4.64) can therefore also be written as E χ p Θ q ´ κ Tr γ χ p Θ q ` D ` ρ Θ , ρ Q χ ˘ ě . (4.65)It is easy to see that ´ ď Θ ď and Θ belongs to Q χ (but not necessarily to the convex set K χ ), whichalso implies that the density ρ Θ of Θ is well defined and belongs to the Coulomb space C (see Section 4.12).40erefore (4.65) is well defined. Introduce another state by exchanging the indices and in the definition of r Q : rr Q : “ γ χ ` Q χ ´ γ χ . Proceeding as before, it can be shown that rr Q P K χ . By definition Q χ “ Θ ` rr Q . Since Q χ minimizes theproblem (3.12) associated with χ and rr Q P K χ , E χ ˆ rr Q ˙ ´ κ Tr γ χ ˆ rr Q ˙ ě E χ ˆ Θ ` rr Q ˙ ´ κ Tr γ χ ˆ Θ ` rr Q ˙ . The above equation can be simplified as E χ p Θ q ´ κ Tr γ χ p Θ q ` D ˆ ρ Θ , ρ ĂĂ Q ˙ ď . (4.66)Let us show that the left hand sides of (4.65) and (4.66) are equal. First of all as Θ belongs to Q χ , we know that Tr γ χ p Θ q “ Tr γ χ p Θ q by Lemma 4.11. Remark also that ρ ĂĂ Q “ ρ χ ´ ρ χ ` ρ Q χ . By Lemma 4.11 E χ p Θ q ´ κ Tr γ χ p Θ q ` D ` ρ Θ , ρ Q χ ˘ ´ ˆ E χ p Θ q ´ κ Tr γ χ p Θ q ` D ˆ ρ Θ , ρ ĂĂ Q ˙˙ “ Tr γ χ pp´ ∆ ` V χ q Θ q ´ D p ρ Θ , ν χ q ` D p ρ Θ , ρ Θ q ´ Tr γ χ pp´ ∆ ` V χ q Θ q ` D p ρ Θ , ν χ q´ D p ρ Θ , ρ Θ q ´ κ ` Tr γ χ p Θ q ´ Tr γ χ p Θ q ˘ ` D ` ρ Θ , ρ Q χ ˘ ´ D ˆ ρ Θ , ρ ĂĂ Q ˙ “ Tr γ χ pp´ ∆ ` V χ q Θ q ´ Tr γ χ pp´ ∆ ` V χ q Θ q ` D p ρ Θ , ρ χ ` ν χ ´ ρ χ ´ ν χ q“ Tr γ χ pp V χ ´ V χ q Θ q ` D p ρ Θ , ρ χ ` ν χ ´ ρ χ ´ ν χ q . (4.67)We show that (4.67) is equal to zero by first showing that p V χ ´ V χ q Θ P S Ă S γ χ . We start by showingthat p ´ ∆ q Q χ i P S . By Cauchy’s resolvent formula, Q χ i “ π ¿ C ˜ z ´ H Q χi ´ z ´ H χ i ¸ dζ “ Q ,i ` Q ,i ` Q ,i , where Q ,i “ π ¿ C z ´ H χ i ˆ´ ρ Q χi ´ ν χ i ¯ ‹ | ¨ | ˙ z ´ H χ i dζ,Q ,i “ π ¿ C z ´ H χ i ˆ´ ρ Q χi ´ ν χ i ¯ ‹ | ¨ | ˙ z ´ H χ i ˆ´ ρ Q χi ´ ν χ i ¯ ‹ | ¨ | ˙ z ´ H χ i dζQ ,i “ π ¿ C z ´ H Q χi ˆ´ ρ Q χi ´ ν χ i ¯ ‹ | ¨ | ˙ z ´ H χ i ˆ´ ρ Q χi ´ ν χ i ¯ ‹ | ¨ | ˙ z ´ H χ i ¨ ˆ´ ρ Q χi ´ ν χ i ¯ ‹ | ¨ | ˙ z ´ H χ i dζ. Following arguments similar to the ones used in the proof of [15, Proposition 2] we obtain that ´ ρ Q χi ´ ν χ i ¯ ‹ |¨| belongs to L p R q` C , and we can conclude that p ´ ∆ q Q χ i P S . Remark that V χ ´ V χ “ p χ ´ χ qp V per ,L ´ V per ,R q P L p R q X L p R q . By definition of Θ , by the Kato–Seiler–Simon inequality and use calculationssimilar to (4.60) (cid:13)(cid:13)(cid:13)(cid:13) p V χ ´ V χ q Θ (cid:13)(cid:13)(cid:13)(cid:13) S “ (cid:13)(cid:13)(cid:13)(cid:13) p V χ ´ V χ q p Q χ ´ Q χ ` γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ď (cid:13)(cid:13)(cid:13)(cid:13) p V χ ´ V χ q p ´ ∆ q ´ (cid:13)(cid:13)(cid:13)(cid:13) S ˜ (cid:13)(cid:13)(cid:13)(cid:13) p ´ ∆ q p Q χ ´ Q χ q (cid:13)(cid:13)(cid:13)(cid:13) S ` (cid:13)(cid:13)(cid:13)(cid:13) p ´ ∆ qp γ χ ´ γ χ q (cid:13)(cid:13)(cid:13)(cid:13) S ¸ ď ? π (cid:13)(cid:13)(cid:13)(cid:13) V χ ´ V χ (cid:13)(cid:13)(cid:13)(cid:13) L ˜ (cid:13)(cid:13)(cid:13)(cid:13) p ´ ∆ q p Q χ ´ Q χ q (cid:13)(cid:13)(cid:13)(cid:13) S ` C (cid:13)(cid:13)(cid:13)(cid:13) p χ ´ χ qp V per ,L ´ V per ,R qp ´ ∆ q ´ (cid:13)(cid:13)(cid:13)(cid:13) S ¸ ă 8 , p V χ ´ V χ q Θ belongs to S , hence Tr γ χ pp V χ ´ V χ q Θ q “ Tr pp V χ ´ V χ q Θ q . On the other hand, by the definition of V χ i in (3.8) and ν i in (3.10) for i “ , , we deduce that Tr pp V χ ´ V χ q Θ q “ D p ρ Θ , p ρ χ ´ µ χ q ´ p ρ χ ´ µ χ qq “ D p ρ Θ , ρ χ ´ ρ χ ` ν χ ´ ν χ q . The above equation implies that the quantity (4.67) equals to . Hence, in view of (4.65) and (4.66), E χ p Θ q ´ κ Tr γ χ p Θ q ` D ` ρ Θ , ρ Q χ ˘ “ E χ p Θ q ´ κ Tr γ χ p Θ q ` D ˆ ρ Θ , ρ ĂĂ Q ˙ ” . We conclude with (4.64) that E χ p r Q q ´ κ Tr γ χ p r Q q ” E χ p Q χ q ´ κ Tr γ χ p Q χ q . Therefore r Q is a minimizer of the problem (3.12) associated with the cut-off function χ . From Theorem 3.5 weknow that ρ r Q ” ρ Q χ , which is equivalent to that ρ Q χ ` ρ χ “ ρ χ ` ρ Q χ . By the arbitrariness of the choiceof χ , χ we deduce that ρ χ ` ρ Q χ is independent of the cut-off function χ P X . References [1] E. Aerts. Surface states of one-dimensional crystals (iii).
Physica , 26(12):1063 – 1072, 1960.[2] A. Anantharaman and ´E. Canc`es. Existence of minimizers for Kohn–Sham models in quantum chemistry.
Ann. I H. Poincar´e-AN , 26(6):2425 – 2455, 2009.[3] J. C. Avila, H. Schulz-Baldes, and C. Villegas-Blas. Topological invariants of edge states for periodic two-dimensional models.
Math. Phys. Anal. Geom. , 16(2):137–170, 2013.[4] B. W. H. Baugher, H. O. H. Churchill, Y. F. Yang, and P. Jarillo-Herrero. Optoelectronic devices based onelectrically tunable p–n diodes in a monolayer dichalcogenide.
Nature Nano. , 9:262–267, 2014.[5] X. Blanc and C. Le Bris. Thomas-Fermi type theories for polymers and thin films.
Adv. Differential Equations ,5(7-9):977–1032, 2000.[6] X. Blanc and C. Le Bris. Periodicity of the infinite-volume ground state of a one-dimensional quantummodel.
Nonlinear Anal. , 48(6):791–803, 2002.[7] X. Blanc, C. Le Bris, and P.-L. Lions. A definition of the ground state energy for systems composed ofinfinitely many particles.
Comm. Partial Differential Equations , 28(1-2):439–475, 2003.[8] L. Bruneau, V. Jakˇsi´c, Y. Last, and C.-A. Pillet. Landauer–B¨uttiker and Thouless conductance.
Commun.Math. Phys. , 338(1), 2015.[9] L. Bruneau, V. Jakˇsi´c, Y. Last, and C.-A. Pillet. Conductance and absolutely continuous spectrum of 1dsamples.
Commun. Math. Phys. , 344(3):959–981, 2016.[10] L. Bruneau, V. Jakˇsi´c, Y. Last, and C.-A. Pillet. Crystalline conductance and absolutely continuous spectrumof 1d samples.
Lett. Math. Phys. , 106(6):787–797, 2016.[11] L. Brus. Commentary: Carbon nanotubes, CdSe nanocrystals, and electron-electron interaction.
Nano Lett. ,10(2):363–365, 2010.[12] L. Brus. Size, dimensionality, and strong electron correlation in nanoscience.
Acc. Chemi. Res. , 47(10):2951–2959, 2014.[13] M B¨uttiker, Y. Imry, R. Landauer, and S. Pinhas. Generalized many-channel conductance formula withapplication to small rings.
Phys. Rev. B , 31:6207–6215, 1985.4214] ´E. Canc`es, L.-L. Cao, and G. Stoltz. Removing a slab from the Fermi sea: the reduced Hartree-Fock model. arXiv:1807.06960 , 2018.[15] ´E. Canc`es, A. Deleurence, and M. Lewin. A new approach to the modeling of local defects in crystals: Thereduced Hartree-Fock case.
Commun. Math. Phys. , 281(1):129–177, 2008.[16] ´E. Canc`es, S. Lahbabi, and M. Lewin. Mean-field models for disordered crystals.
J. Math. Pures. Appl. ,100(2):241–274, 2013.[17] ´E. Canc`es and G. Stoltz. A mathematical formulation of the random phase approximation for crystals.
Ann.I. H. Poincare-An. , 29(6):887 – 925, 2012.[18] ´E Canc`es, G. Stoltz, G. E. Scuseria, V. N. Staroverov, and E. R. Davidson. Local Exchange Potentials forElectronic Structure Calculations.
Maths. In Action , 2(1):1–42, 2009.[19] L. Cao.
Analyse math´ematique du transport thermo- ´electronique dans les solides d´esordonn´es . Universit´eParis Est, 2019.[20] I. Catto, C. Le Bris, and P-L. Lions. On the thermodynamic limit for Hartree–Fock type models.
Ann. I. H.Poincar´e-AN , 18(6):687 – 760, 2001.[21] I. Catto, C. Le Bris, and P.-L. Lions.
The Mathematical Theory of Thermodynamic Limits: Thomas-Fermi TypeModels . Oxford mathematical monographs. Clarendon Press, 1998.[22] J. M. Combes and L. Thomas. Asymptotic behaviour of eigenfunctions for multiparticle Schr¨odinger oper-ators.
Comm. Math. Phys. , 34(4):251–270, 1973.[23] H. D. Cornean, P. Duclos, G. Nenciu, and R. Purice. Adiabatically switched-on electrical bias and theLandauer–B¨uttiker formula.
J. Math. Phys. , 49(10):102106, 2008.[24] H. D. Cornean, P. Duclos, and R. Purice. Adiabatic non-equilibrium steady states in the partition freeapproach.
Ann. Henri Poincar´e , 13(4):827–856, 2012.[25] P. Deift and B. Simon. On the decoupling of finite singularities from the question of asymptotic complete-ness in two body quantum systems.
J. Funct. Anal. , 23(3):218 – 238, 1976.[26] T. Dohnal, M. Plum, and W. Reichel. Surface gap soliton ground states for the nonlinear Schr¨odingerequation.
Commu. Math. Phys. , 308(2):511, 2011.[27] R. J. Duffin. Yukawa potential theory.
J. Math. Anal. Appl. , 35:105–130, 1971.[28] P. Exner and R. L. Frank. Absolute continuity of the spectrum for periodically modulated leaky wires in R . Ann. Henri. Poincar´e , 8(2):241–263, 2007.[29] R.L. Frank, M. Lewin, E.H. Lieb, and R. Seiringer. Energy cost to make a hole in the Fermi sea.
Phys. Rev.Lett. , 106:150402, 2011.[30] R.L. Frank, M. Lewin, E.H. Lieb, and R. Seiringer. A positive density analogue of the Lieb–Thirring inequal-ity.
Duke Math. J. , 162(3):435–495, 2013.[31] F. Germinet and A. Klein. Operator kernel estimates for functions of generalized Schr¨odinger operators.
Proceedings of the American Mathematical Society , 131:911–920, 2003.[32] S. J. Gustafson and I. M. Sigal.
Mathematical Concepts of Quantum Mechanics . Universitext (Berlin. Print).Springer, 2003.[33] Ch. Hainzl, M. Lewin, and ´E. S´er´e. Existence of a stable polarized vacuum in the Bogoliubov-Dirac-Fockapproximation.
Commun. Math. Phys. , 257(3):515–562, 2005.[34] Ch. Hainzl, M. Lewin, and ´E. S´er´e. Self-consistent solution for the polarized vacuum in a no-photon QEDmodel.
J. Phys. A-Math. Gen. , 38(20):4483, 2005.[35] Ch. Hainzl, M. Lewin, and ´E. S´er´e. Existence of atoms and molecules in the mean-field approximation ofno-photon quantum electrodynamics.
Arch. Ration. Mech. An. , pages 453–499, 2009.4336] Ch. Hainzl, M. Lewin, and J.P. Solovej. The mean-field approximation in quantum electrodynamics. Theno-photon case.
Commun. Pur. Appl. Math. , LX:546–596, 2007.[37] M. Z. Hasan and C. L. Kane. Colloquium: Topological insulators.
Rev. Mod. Phys. , 82:3045–3067, 2010.[38] Y. Hatsugai. Edge states in the integer quantum Hall effect and the Riemann surface of the Bloch function.
Phys. Rev. B , 48:11851–11862, 1993.[39] B. Helffer.
Spectral Theory and its Applications . Cambridge University Press, 2013.[40] R. Hempel and M. Kohlmann. A variational approach to dislocation problems for periodic Schr¨odingeroperators.
J. Math. Anal. Appl. , 381(1):166–178, 2011.[41] R. Hempel, M. Kohlmann, M. Stautz, and J. Voigt. Bound states for nano-tubes with a dislocation.
J. Math.Anal. Appl. , 431:202 – 227, 2014.[42] P.D. Hislop and I.M. Sigal.
Introduction to Spectral Theory: With Applications to Schr¨odinger Operators .Applied Mathematical Sciences. Springer New York, 1995.[43] M. Hoffmann-Ostenhof and T. Hoffmann-Ostenhof. “Schr¨odinger inequalities” and asymptotic behavior ofthe electron density of atoms and molecules.
Phys. Rev. A , 16:1782–1785, 1977.[44] T. Kato.
Perturbation Theory for Linear Operators . Springer, Berlin, Heidelberg, 1995.[45] F. Klopp. An asymptotic expansion for the density of states of a random Schr¨odinger operator withBernoulli disorder.
Random Oper. Stochastic Equations , 3:315–332, 1995.[46] W. Kohn and L. J. Sham. Self-consistent equations including exchange and correlation effects.
Phys. Rev. ,140:A1133–A1138, 1965.[47] E. Korotyaev. Lattice dislocations in a 1-dimensional model.
Commun. Math. Phys. , 213(2):471–489, 2000.[48] E. Korotyaev. Schr¨odinger operator with a junction of two 1-dimensional periodic potentials.
Asymptot.Anal. , 45, 2003.[49] S. Lahbabi. The reduced Hartree–Fock model for short-range quantum crystals with nonlocal defects.
Ann.Henri. Poincar´e , 15(7):1403–1452, 2014.[50] E. A. Laird, F. Kuemmeth, G. A. Steele, K. Grove-Rasmussen, J. Nyg˚ard, K. Flensberg, and L. P. Kouwen-hoven. Quantum transport in carbon nanotubes.
Rev. Mod. Phys. , 87:62, 2015.[51] R. Landauer. Electrical resistance of disordered one-dimensional lattices.
Philos. Mag. , 21:863–867, 1970.[52] J. U. Lee, P. P. Gipp, and C. M. Heller. Carbon nanotube p-n junction diodes.
Appl. Phys. Lett. , 85:145–147,2004.[53] F. L´eonard and J. Tersoff. Novel length scales in nanotube devices.
Phys. Rev. Lett. , 83:5174–5177, 1999.[54] M. Lewin. Large Quantum Systems: a Mathematical and Numerical Perspective, 2009. Habilitation `aDiriger des Recherches, University of Cergy-Pontoise.[55] M. Lewin, E.H. Lieb, and R. Seringer. The local density approximation in density functional theory. arXivpreprint,1903.04046 , 2019.[56] E.H. Lieb and M. Loss.
Analysis . CRM Proceedings & Lecture Notes. Amer. Math. Soc., 2001.[57] E.H. Lieb and B. Simon. The Thomas-Fermi theory of atoms, molecules and solids.
Adv. Math. , 23(1):22 –116, 1977.[58] P.-L. Lions. Solutions of Hartree-Fock equations for Coulomb systems.
Comm. Math. Phys. , 109(1):33–97,1987.[59] A. Nitzan and M. A. Ratner. Electron transport in molecular wire junctions.
Science , 300:1384–1389, 2003.4460] A. Pospischil, M. M. Furchi, and T. Mueller. Solar-energy conversion and light emission in an atomicmonolayer p–n diode.
Nature. Nano. , 9:257–261, 2014.[61] M. Reed and B. Simon.
Method of Modern Mathematical Physics. Vol.2: Fourier Analysis, Self-Adjointness .Academic Press, San Diego, 1975.[62] M. Reed and B. Simon.
Method of Modern Mathematical Physics. Vol.4: Analysis of Operators . AcademicPress, San Diego, 1978.[63] L. F. Rupert and G. S. Roman. On the scattering theory of the laplacian with a periodic boundary condition.i. existence of wave operators.
Doc. Math. , 8:547–565, 2003.[64] L. F. Rupert and G. S. Roman. On the scattering theory of the laplacian with a periodic boundary condition.ii. additional channels of scattering.
Doc. Math. , 9:57–77, 2004.[65] E. Seiler and B. Simon. Bounds in the Yukawa2 quantum field theory: Upper bound on the pressure,Hamiltonian bound and linear lower bound.
Commun. Math. Phys. , 45(2):99–114, 1975.[66] D. Shale and W. F. Stinespring. Spinor representations of infinite orthogonal groups.
J. Math. Mech. ,14(2):315–322, 1965.[67] W. Shockley. On the surface states associated with a periodic potential.
Phys. Rev. , 56:317–323, 1939.[68] J. P. Solovej. Proof of the ionization conjecture in a reduced Hartree–Fock model.
Invent. Math. , 104(1):291–311, 1991.[69] J. P. Solovej. Lecture notes on Many Body Quantum Mechanics, 2007.[70] L. E. Thomas. Time dependent approach to scattering from impurities in a crystal.
Commu. Math. Phys. ,33(4):335–343, 1973.[71] J.-S. Wang, J. Wang, and J. T. L¨u. Quantum thermal transport in nanostructures.