betaFeedback

WorksPhysics and Astronomy

Large particle number asymptotics for point-interacting bosons in the plane

Abstract

We consider NN bosons in two dimensions interacting via point interactions, modeled by the TMS Hamiltonian. We prove that their ground state energy has the asymptotic behavior −eCN-e^{C N } for large NN, for an explicit constant CC related to the sharp constant in the Ladyzhenskaya-Gagliardo-Nirenberg inequality. Such asymptotic behavior was predicted in the physics literature, based on a variant of the mean-field approximation.

Introduction and Main Results

Introduction

Point interactions play a prominent role in quantum physics, due to their universal nature. They have been predicted to be relevant effective models for cold atomic gases [6, 13]. In this paper, we shall consider point-interacting bosons in the two-dimensional plane. These are described by the TMS Hamiltonian [12, 3], which will be recalled in the next section. It is well known that this Hamiltonian is bounded from below for any particle number NN, but the best known lower bounds on its ground state energy behave as −eCN2-e^{CN^{2}} for a suitable constant C>0C>0 [3, 4]. On the other hand, a mean-field approximation predicts the asymptotic behavior −eCN-e^{CN} [10, 6] for an explicit CC involving the sharp constant a∗a_* in the Ladyzhenskaya–Gagliardo–Nirenberg inequality

∥∇u∥22∥u∥22≥a∗2∥u∥44,u∈H1(R2).(1)\lVert\nabla u\rVert_{2}^{2}\lVert u\rVert_{2}^{2}\ge\frac{a_*}{2}\lVert u\rVert_{4}^{4},\qquad u\in H^{1}(\mathbb{R}^{2}). \tag*{(1)}

For more on this inequality, including the numerical value a∗≈11.70a_*\approx11.70, see [14]. In this paper, we shall prove that the prediction from [10, 6] is indeed correct.

Asymptotics of ground state energy

Let eNe_N denote the ground state energy of the TMS Hamiltonian defined in (12) below. The latter depends on a parameter α∈R\alpha\in\mathbb{R} determining the two-body binding energy e2e_2, but the ratio eN/e2e_N/e_2 is independent of α\alpha, and only depends on NN. It has the following asymptotic behavior.

Theorem 1. Let eNe_N denote the ground state energy of the TMS Hamiltonian defined in (12), and a∗a_* the sharp constant in (1). Then

lim⁡N→∞1Nln⁡eNe2=8πa∗.(2)\lim_{N\to\infty}\frac{1}{N}\ln\frac{e_N}{e_2}=\frac{8\pi}{a_*}. \tag*{(2)}

The proof of this theorem will be given in § 3 (upper bound) and 4 (lower bound).

These bounds will be carried out in the charge formulation of the ground state energy problem, defined on functions ξ\xi on R2(N−1)\mathbb{R}^{2(N-1)} that are permutation symmetric in the last N−2N-2 variables. For the upper bound, following [10], we take ξ\xi to be a product state in these last N−2N-2 variables, multiplied by an appropriately chosen function of the remaining variable. This choice of product states is reminiscent of the Hartree approximation in the mean field limit, which was studied, in a rather general setting, in [8]. In the charge formulation, however, the effective Hamiltonian does not consist only of one-body and two-body terms, and consequently the approach from [8] is not applicable. For the lower bound we instead proceed by an ad-hoc method, where at a crucial point a weak form of a result of [7] enters. Interestingly, the proof in [7] relies on the De Finetti theorem for product observables with logarithmic dependence on the dimension of the one-particle space [2, 9], which is a variant of the De Finetti theorem lying at the basis of [8].

Universality for weakly interacting bosons

The point interaction model arises naturally for systems with genuine short range interactions in the weak coupling limit. Consider NN bosons in R2\mathbb{R}^{2}, with Hamiltonian

H=∑j=1N(−Δj)−λ∑1≤i<j≤NV(xi−xj)(3)H = \sum_{j=1}^{N}(-\Delta_j)-\lambda\sum_{1\le i<j\le N}V(x_i-x_j) \tag*{(3)}

in Lsym2(R2N)L_{\mathrm{sym}}^{2}(\mathbb{R}^{2N}), where V∈L1(R2)V\in L^{1}(\mathbb{R}^{2}) with ∫V>0\int V>0 is even and sufficiently regular (e.g. V∈L2(R2)V\in L^{2}(\mathbb{R}^{2}) and ∫R2∣V(x)∣(1+∣x∣)p dx<∞\int_{\mathbb{R}^{2}}|V(x)|(1+|x|)^{p}\,dx<\infty for p>0p>0 suffice). Let E0(N,λ)E_0(N,\lambda) denote the ground state energy of HH for λ>0\lambda>0. The following theorem shows universality in the weak coupling limit, and the appearance of point interactions as an effective model.

Theorem 2. With the definitions above,

lim⁡λ→0E0(N,λ)E0(2,λ)=ϵNϵ2,(4)\lim_{\lambda\to0}\frac{E_0(N,\lambda)}{E_0(2,\lambda)}=\frac{\epsilon_N}{\epsilon_2}, \tag*{(4)}

independently of VV.

As we shall explain momentarily, this theorem can be deduced in a straightforward way from the analysis in [5]. We remark that it follows from [11, 1] that

E0(2,λ)=−exp⁡(−8πλ∫V−2(∫V)2∫R4V(x)ln⁡∣x−y∣V(y) dx dy−2γE+ln⁡8+o(1)λ→0)(5)E_0(2,\lambda)=-\exp\left(-\frac{8\pi}{\lambda\int V}-\frac{2}{(\int V)^2}\int_{\mathbb{R}^{4}}V(x)\ln|x-y|V(y)\,dx\,dy-2\gamma_{\mathrm{E}}+\ln8+o(1)_{\lambda\to0}\right) \tag*{(5)}

where γE\gamma_{\mathrm{E}} denotes Euler’s constant.

Let us give the details of how to deduce Theorem 2 from the results of [5]. Their assertion is that

∑j=1N(−Δj)−ε−24π∫V1ln⁡(1/ε)∑1≤i<j≤NV((xi−xj)/ε)(6)\sum_{j=1}^{N}(-\Delta_j)-\varepsilon^{-2}\frac{4\pi}{\int V}\frac{1}{\ln(1/\varepsilon)}\sum_{1\leq i<j\leq N}V((x_i-x_j)/\varepsilon) \tag*{(6)}

converges, as ε→0+\varepsilon\to0_+, in norm resolvent sense to the TMS Hamiltonian with an explicit parameter α∈R\alpha\in\mathbb{R}; see [5 Eq. (4.34)]. While the value of α\alpha is irrelevant for us, what is crucial is that it is independent of NN. Thus, the bottom of the spectrum of the operator (1.6) is given by eN+o(1)e_N+o(1) as ε→0\varepsilon\to0. By scaling, it follows that the bottom of the spectrum of the operator

∑j=1N(−Δj)−4π∫V1ln⁡(1/ε)∑1≤i<j≤NV(xi−xj)(7)\sum_{j=1}^{N}(-\Delta_j)-\frac{4\pi}{\int V}\frac{1}{\ln(1/\varepsilon)}\sum_{1\leq i<j\leq N}V(x_i-x_j) \tag*{(7)}

is given by ε2eN+o(ε2)\varepsilon^2e_N+o(\varepsilon^2). Applying this fact with ε=exp⁡(−4π/(λ∫V))\varepsilon=\exp(-4\pi/(\lambda\int V)) proves Theorem 2.

TMS Hamiltonian

The purpose of this section is two-fold. First, we introduce the TMS Hamiltonian, which describes N≥2N\geq2 bosons of mass 1/21/2 in two dimensions, interacting via zero range forces, as a self-adjoint operator in the Hilbert space Lsym2(R2N)L^2_{\mathrm{sym}}(\mathbb{R}^{2N}) of permutation symmetric functions. Its ground state energy eNe_N (that is, the bottom of the spectrum of this operator) is the main quantity of interest. Second, we introduce an ‘effective’ operator, acting in the Hilbert space

H:=L2(R2)⊗Lsym2(R2(N−2))(8)\mathcal{H}:=L^2(\mathbb{R}^2)\otimes L^2_{\mathrm{sym}}(\mathbb{R}^{2(N-2)}) \tag*{(8)}

of functions on R2(N−1)\mathbb{R}^{2(N-1)} that are symmetric in the last N−2N-2 variables. The elements of H\mathcal{H} are often called charges in the physics literature. The bottom of the spectrum of the effective operator in H\mathcal{H}, denoted by tNt_N, is directly related to the quotient eN/e2e_N/e_2. In the following sections we will study the asymptotics of tNt_N.

Let us recall the definition of the TMS Hamiltonian, following [12, 3]. Let μ>0\mu>0. In the Hilbert space H\mathcal{H} we introduce the quadratic form

Tμ(ξ)=π2∫∣ξ^(q,p^12)∣2ln⁡(12q2+∣p^12∣2+μ) dq dp3⋯dpN−2(N−2)∫ξ^(p1+p2,p^12)∗ξ^(p1+p3,p^13)∣p⃗∣2+μ dp1⋯dpN−(N−2)(N−3)2∫ξ^(p1+p2,p^12)∗ξ^(p3+p4,p^34)∣p⃗∣2+μ dp1⋯dpN.(9)\begin{aligned} T_\mu(\xi)={}&\frac{\pi}{2}\int\left|\widehat{\xi}(q,\widehat{p}_{12})\right|^2\ln\left(\frac{1}{2}q^2+|\widehat{p}_{12}|^2+\mu\right)\,dq\,dp_3\cdots dp_N\\ &-2(N-2)\int\frac{\widehat{\xi}(p_1+p_2,\widehat{p}_{12})^*\widehat{\xi}(p_1+p_3,\widehat{p}_{13})}{|\vec{p}|^2+\mu}\,dp_1\cdots dp_N\\ &-\frac{(N-2)(N-3)}{2}\int\frac{\widehat{\xi}(p_1+p_2,\widehat{p}_{12})^*\widehat{\xi}(p_3+p_4,\widehat{p}_{34})}{|\vec{p}|^2+\mu}\,dp_1\cdots dp_N. \tag*{(9)} \end{aligned}

Here, for p⃗=(p1,…,pN)∈R2N\vec{p}=(p_1,\ldots,p_N)\in\mathbb{R}^{2N} we denote p^ij=(p1,…,pi^,…,pj^,…,pN)∈R2(N−2)\widehat{p}_{ij}=(p_1,\ldots,\widehat{p_i},\ldots,\widehat{p_j},\ldots,p_N)\in\mathbb{R}^{2(N-2)}, and z∗z^* denotes the complex conjugate of a number z∈Cz\in\mathbb{C}. The domain of this quadratic form, dom⁡Tμ\operatorname{dom}T_\mu, is given by those ξ∈H\xi\in\mathcal{H} for which the first term on the right side is finite. The second and third terms are bounded quadratic forms in H\mathcal{H}, as we will see later on.

Now let α∈R\alpha\in\mathbb{R} and define a quadratic form Eα\mathcal{E}^{\alpha} in Lsym2(R2N)L^2_{\mathrm{sym}}(\mathbb{R}^{2N}) as follows. We fix a parameter μ>0\mu> 0 (later, we will see that the construction is independent of this parameter) and define the domain of Eα\mathcal{E}^{\alpha} to be

dom⁡Eα={ψ∈Lsym2(R2N): ψ=φ+Gμξ with φ∈Hsym1(R2N) and ξ∈dom⁡Tμ}.(10)\operatorname{dom}\mathcal{E}^{\alpha}=\left\{\psi\in L^2_{\mathrm{sym}}(\mathbb{R}^{2N}):\ \psi=\varphi+G_{\mu}\xi\ \text{with}\ \varphi\in H^1_{\mathrm{sym}}(\mathbb{R}^{2N})\ \text{and}\ \xi\in\operatorname{dom}T_{\mu}\right\}. \tag*{(10)}

Here GμξG_{\mu}\xi is defined through its Fourier transform

Gμξ^(p1,…,pN)=1∣p⃗∣2+μ∑i<jξ^(pi+pj,p^ij).(11)\widehat{G_{\mu}\xi}(p_1,\ldots,p_N)=\frac{1}{|\vec{p}|^2+\mu}\sum_{i<j}\widehat{\xi}(p_i+p_j,\widehat{p}_{ij}). \tag*{(11)}

Since Gμξ∉Hsym1(R2N)G_{\mu}\xi\notin H^1_{\mathrm{sym}}(\mathbb{R}^{2N}) for all ξ≠0\xi\ne0, the decomposition ψ=φ+Gμξ\psi=\varphi+G_{\mu}\xi of ψ∈dom⁡Eα\psi\in\operatorname{dom}\mathcal{E}^{\alpha} is unique and we can set

Eα(ψ)=−μ∥ψ∥2+∥∇φ∥2+μ∥φ∥2+N(N−1)2(Tμ(ξ)+α∥ξ∥2).(12)\mathcal{E}^{\alpha}(\psi)=-\mu\|\psi\|^2+\|\nabla\varphi\|^2+\mu\|\varphi\|^2+\frac{N(N-1)}{2}\left(T_{\mu}(\xi)+\alpha\|\xi\|^2\right). \tag*{(12)}

One easily checks that the definition of dom⁡Eα\operatorname{dom}\mathcal{E}^{\alpha} and the value of Eα(ψ)\mathcal{E}^{\alpha}(\psi) are independent of μ\mu. (To do so, one verifies that ξ\xi is uniquely determined by ψ\psi, independently of μ\mu.)

As shown in [3], the quadratic form Eα\mathcal{E}^{\alpha} is bounded from below and closed in the Hilbert space Lsym2(R2N)L^2_{\mathrm{sym}}(\mathbb{R}^{2N}) and, therefore, generates a self-adjoint operator. The bottom of its spectrum is

ϵN:=inf⁡0≠ψ∈dom⁡EαEα(ψ)∥ψ∥2.(13)\epsilon_N:=\inf_{0\ne\psi\in\operatorname{dom}\mathcal{E}^{\alpha}}\frac{\mathcal{E}^{\alpha}(\psi)}{\|\psi\|^2}. \tag*{(13)}

One easily sees that ϵN\epsilon_N is the largest −μ-\mu such that Tμ(ξ)+α∥ξ∥2T_{\mu}(\xi)+\alpha\|\xi\|^2 is non-negative. In particular, for N=2N=2, we see that ϵ2=−exp⁡(−2α/π)\epsilon_2=-\exp(-2\alpha/\pi). By scaling, ϵN/ϵ2\epsilon_N/\epsilon_2 is independent of α\alpha; in fact, we have the identity

ϵN=ϵ2e−2πtN.(14)\epsilon_N=\epsilon_2 e^{-\frac{2}{\pi}t_N}. \tag*{(14)}

where

tN=inf⁡0≠ξ∈dom⁡TμTμ(ξ)∥ξ∥2−π2ln⁡μ=inf⁡0≠ξ∈dom⁡T1T1(ξ)∥ξ∥2.(15)t_N=\inf_{0\ne\xi\in\operatorname{dom}T_{\mu}}\frac{T_{\mu}(\xi)}{\|\xi\|^2}-\frac{\pi}{2}\ln\mu =\inf_{0\ne\xi\in\operatorname{dom}T_1}\frac{T_1(\xi)}{\|\xi\|^2}. \tag*{(15)}

The second equality follows by scaling.

In view of these facts, Theorem 1 is equivalent to the statement

lim⁡N→∞1NtN=−(2π)2a∗.(16)\lim_{N\to\infty}\frac{1}{N}t_N=-\frac{(2\pi)^2}{a_*}. \tag*{(16)}

which we shall prove in the following. It will be convenient to introduce the operators S0S_0, S1S_1 and S2S_2 in H\mathcal{H}, corresponding to the three terms in (9) at μ=1\mu=1, in such a way that

T1(ξ)=⟨ξ∣S0∣ξ⟩−2(N−2)⟨ξ∣S1∣ξ⟩−(N−2)(N−3)2⟨ξ∣S2∣ξ⟩.(17)T_1(\xi)=\langle\xi|S_0|\xi\rangle-2(N-2)\langle\xi|S_1|\xi\rangle-\frac{(N-2)(N-3)}{2}\langle\xi|S_2|\xi\rangle. \tag*{(17)}

For computing its infimum, we shall use that it clearly suffices to consider functions ξ\xi with non-negative Fourier transform, ξ^≥0\widehat{\xi}\ge0.

Upper Bound

Our upper bound on tNt_N in (15) is based on a product ansatz, following [10]. We shall pick a normalized ξ∈H\xi\in\mathcal{H} such that ξ^≥0\hat{\xi} \ge0. Then the second term in (17) is non-positive,

−⟨ξ∣S1∣ξ⟩=−∫ξ^(p1+p2,p^12)ξ^(p1+p3,p^13)∣p⃗∣2+1 dp1⋯dpN≤0.(18)-\langle\xi|S_1|\xi\rangle=-\int\frac{\hat{\xi}(p_1+p_2,\hat{p}_{12})\hat{\xi}(p_1+p_3,\hat{p}_{13})}{|\vec{p}|^2+1}\,dp_1\cdots dp_N\le0. \tag*{(18)}

Moreover, by Jensen’s inequality

−⟨ξ∣S2∣ξ⟩=−∫ξ^(p1+p2,p^12)ξ^(p3+p4,p^34)∣p⃗∣2+1 dp1⋯dpN≤−(∫ξ^(p1+p2,p^12)ξ^(p3+p4,p^34) dp1⋯dpN)2∫ξ^(p1+p2,p^12)ξ^(p3+p4,p^34)(∣p⃗∣2+1) dp1⋯dpN.(19)\begin{aligned} -\langle\xi|S_2|\xi\rangle=-\int\frac{\hat{\xi}(p_1+p_2,\hat{p}_{12})\hat{\xi}(p_3+p_4,\hat{p}_{34})}{|\vec{p}|^2+1}\,dp_1\cdots dp_N \\ &\le-\frac{\left(\int\hat{\xi}(p_1+p_2,\hat{p}_{12})\hat{\xi}(p_3+p_4,\hat{p}_{34})\,dp_1\cdots dp_N\right)^2}{\int\hat{\xi}(p_1+p_2,\hat{p}_{12})\hat{\xi}(p_3+p_4,\hat{p}_{34})(|\vec{p}|^2+1)\,dp_1\cdots dp_N}. \tag*{(19)} \end{aligned}

Likewise,

2π⟨ξ∣S0∣ξ⟩=∫∣ξ^(q,p^12)∣2ln⁡(12q2+∣p^12∣2+1) dqdp3⋯dpN≤ln⁡(∫∣ξ^(q,p^12)∣2(12q2+∣p^12∣2+1) dqdp3⋯dpN).(20)\begin{aligned} \frac{2}{\pi}\langle\xi|S_0|\xi\rangle=\int|\hat{\xi}(q,\hat{p}_{12})|^2\ln\left(\frac{1}{2}q^2+|\hat{p}_{12}|^2+1\right)\,dqdp_3\cdots dp_N \\ &\le\ln\left(\int|\hat{\xi}(q,\hat{p}_{12})|^2\left(\frac{1}{2}q^2+|\hat{p}_{12}|^2+1\right)\,dqdp_3\cdots dp_N\right). \tag*{(20)} \end{aligned}

We shall make the ansatz

ξ^(q,p1,…,pN−2)=v^(q)∏j=1N−2u^(pj)(21)\hat{\xi}(q,p_1,\ldots,p_{N-2})=\hat{v}(q)\prod_{j=1}^{N-2}\hat{u}(p_j) \tag*{(21)}

for L2L^2-normalized u,v∈H1(R2)u,v\in H^1(\mathbb{R}^2) such that u^,v^≥0\hat{u},\hat{v}\ge0. Then

∫∣ξ^(q,p^12)∣2(12q2+∣p^12∣2+1) dqdp3⋯dpN=1+12∥∇v∥2+(N−2)∥∇u∥2.(22)\int|\hat{\xi}(q,\hat{p}_{12})|^2\left(\frac{1}{2}q^2+|\hat{p}_{12}|^2+1\right)\,dqdp_3\cdots dp_N=1+\frac{1}{2}\|\nabla v\|^2+(N-2)\|\nabla u\|^2. \tag*{(22)}

Moreover,

∫ξ^(p1+p2,p^12)ξ^(p3+p4,p^34) dp1⋯dpN=(∫v^(p1+p2)u^(p1)u^(p2) dp1dp2)2(23)\int\hat{\xi}(p_1+p_2,\hat{p}_{12})\hat{\xi}(p_3+p_4,\hat{p}_{34})\,dp_1\cdots dp_N=\left(\int\hat{v}(p_1+p_2)\hat{u}(p_1)\hat{u}(p_2)\,dp_1dp_2\right)^2 \tag*{(23)}

and

∫ξ^(p1+p2,p^12)ξ^(p3+p4,p^34)∣p⃗∣2 dp1⋯dpN=(N−4)∥∇u∥2(∫v^(p1+p2)u^(p1)u^(p2) dp1dp2)2+4(∫v^(p1+p2)u^(p1)u^(p2) dp1dp2)(∫v^(p1+p2)u^(p1)u^(p2)∣p1∣2 dp1dp2).(24)\begin{aligned} \int\hat{\xi}(p_1+p_2,\hat{p}_{12})\hat{\xi}(p_3+p_4,\hat{p}_{34})|\vec{p}|^2\,dp_1\cdots dp_N ={}&(N-4)\|\nabla u\|^2\left(\int\hat{v}(p_1+p_2)\hat{u}(p_1)\hat{u}(p_2)\,dp_1dp_2\right)^2\\ &+4\left(\int\hat{v}(p_1+p_2)\hat{u}(p_1)\hat{u}(p_2)\,dp_1dp_2\right) \left(\int\hat{v}(p_1+p_2)\hat{u}(p_1)\hat{u}(p_2)|p_1|^2\,dp_1dp_2\right). \tag*{(24)} \end{aligned}

We thus conclude that

lim sup⁡N→∞1NT1(ξ)≤−12(∫v^(p1+p2)u^(p1)u^(p2) dp1dp2)2∥∇u∥2.(25)\limsup_{N\to\infty}\frac{1}{N}T_1(\xi)\le-\frac{1}{2}\frac{\left(\int\hat{v}(p_1+p_2)\hat{u}(p_1)\hat{u}(p_2)\,dp_1dp_2\right)^2}{\|\nabla u\|^2}. \tag*{(25)}

Since

∫v^(p1+p2)u^(p1)u^(p2) dp1dp2=2π∫v∗u2,(26)\int\widehat{v}(p_1+p_2)\widehat{u}(p_1)\widehat{u}(p_2)\,dp_1dp_2=2\pi\int v^*u^2, \tag*{(26)}

the optimal choice of vv, given uu, is v=u2/∥u∥42v=u^2/\|u\|_4^2, and, in view of (15), this gives the desired upper bound

lim sup⁡N→∞1NtN≤−(2π)2a∗.(27)\limsup_{N\to\infty}\frac{1}{N}t_N\leq-\frac{(2\pi)^2}{a_*}. \tag*{(27)}

Here we used the fact that the sharp constant a∗a_* in (1) remains the same when restricted to functions with nonnegative Fourier transform: indeed, replacing any function uu by u~=F−1∣u^∣\widetilde{u}=\mathcal{F}^{-1}|\widehat{u}| preserves its L2L^2 norm and Dirichlet integral, while ∣u~∣4≥∣u∣4|\widetilde{u}|_4\geq|u|_4 by Plancherel’s theorem and the pointwise inequality ∣u^∗u^∣≤∣u^∣∗∣u^∣|\widehat{u}*\widehat{u}|\leq|\widehat{u}|*|\widehat{u}|. Moreover, for an optimizer uu, u2∈H1(R2)u^2\in H^1(\mathbb{R}^2).

Lower bound

The proof of the lower bound is substantially more involved. We shall split it into several steps.

Step 1. Reduction to S2S_2

Step 1. Reduction to S2S_2. We shall first show that it suffices to consider the third term S2S_2 in (17). For the first term, we simply have S0≥0S_0\geq0. For the second, we use the following lemma.

Lemma 1. For any 0<β<10<\beta<1, λ∈R\lambda\in\mathbb{R} and ψ∈L2(R2)\psi\in L^2(\mathbb{R}^2),

0≤∫R4ψ(x)∗ψ(y)∣x∣2+∣y∣2+λ2 dxdy≤π2sin⁡πβ∫R2∣ψ(x)∣2(∣x∣2∣x∣2+λ2)β dx.(28)0\leq\int_{\mathbb{R}^4}\frac{\psi(x)^*\psi(y)}{|x|^2+|y|^2+\lambda^2}\,dxdy\leq\frac{\pi^2}{\sin\pi\beta}\int_{\mathbb{R}^2}|\psi(x)|^2\left(\frac{|x|^2}{|x|^2+\lambda^2}\right)^\beta\,dx. \tag*{(28)}

Proof. By writing (∣x∣2+∣y∣2+λ2)−1=∫0∞e−t(∣x∣2+∣y∣2+λ2) dt(|x|^2+|y|^2+\lambda^2)^{-1}=\int_0^\infty e^{-t(|x|^2+|y|^2+\lambda^2)}\,dt, we obtain the left inequality in (28). By Cauchy–Schwarz, we have for any positive function hh,

∫R4ψ(x)∗ψ(y)∣x∣2+∣y∣2+λ2 dxdy≤∫R2∣ψ(x)∣21h(x)∫R2h(y)∣x∣2+∣y∣2+λ2 dydx.(29)\int_{\mathbb{R}^4}\frac{\psi(x)^*\psi(y)}{|x|^2+|y|^2+\lambda^2}\,dxdy\leq\int_{\mathbb{R}^2}|\psi(x)|^2\frac{1}{h(x)}\int_{\mathbb{R}^2}\frac{h(y)}{|x|^2+|y|^2+\lambda^2}\,dydx. \tag*{(29)}

We choose h(y)=∣y∣−2βh(y)=|y|^{-2\beta}. This implies the claim by scaling and the fact that

∫R2∣z∣−2β1+∣z∣2 dz=2π∫0∞r−2β+11+r2 dr=π∫0∞s−β1+s ds=π2sin⁡πβ.(30)\int_{\mathbb{R}^2}\frac{|z|^{-2\beta}}{1+|z|^2}\,dz=2\pi\int_0^\infty\frac{r^{-2\beta+1}}{1+r^2}\,dr=\pi\int_0^\infty\frac{s^{-\beta}}{1+s}\,ds=\frac{\pi^2}{\sin\pi\beta}. \tag*{(30)}

For the last integral we have substituted t=s1+st=\frac{s}{1+s} to arrive at a beta function, which we have simplified using Euler’s reflection formula for the gamma function.

Shifting the integration variable p1p_1 to p1−p2−p3p_1-p_2-p_3, we can write

⟨ξ∣S1∣ξ⟩=∫ξ^(p1−p3,p^12)∗ξ^(p1−p2,p^13)∣p1−p2−p3∣2+∑j=2N∣pj∣2+1 dp1⋯dpN(31)\langle\xi|S_1|\xi\rangle=\int\frac{\widehat{\xi}(p_1-p_3,\widehat{p}_{12})^*\widehat{\xi}(p_1-p_2,\widehat{p}_{13})}{|p_1-p_2-p_3|^2+\sum_{j=2}^{N}|p_j|^2+1}\,dp_1\cdots dp_N \tag*{(31)}

and hence

∣⟨ξ∣S1∣ξ⟩∣≤∫∣ξ^(p1−p3,p^12)∣∣ξ^(p1−p2,p^13)∣∣p2∣2+∣p3∣2+∑j=4N∣pj∣2 dp1⋯dpN.(32)|\langle\xi|S_1|\xi\rangle|\leq\int\frac{|\widehat{\xi}(p_1-p_3,\widehat{p}_{12})||\widehat{\xi}(p_1-p_2,\widehat{p}_{13})|}{|p_2|^2+|p_3|^2+\sum_{j=4}^{N}|p_j|^2}\,dp_1\cdots dp_N. \tag*{(32)}

Applying Lemma 1 for the integration over the variables p2p_2 and p3p_3, we obtain, for normalized ξ∈H\xi\in\mathcal{H},

∣⟨ξ∣S1∣ξ⟩∣≤π2sin⁡βπ∫∣ξ^(p1−p3,p^12)∣2(∣p3∣2∣p3∣2+∑j=4N∣pj∣2)β dp1dp3⋯dpN=π2sin⁡βπ∫∣ξ^(p1,p^12)∣2(∣p3∣2∑j=3N∣pj∣2)β dp1dp3⋯dpN≤π2sin⁡βπ(∫∣ξ^(p1,p^12)∣2∣p3∣2∑j=3N∣pj∣2 dp1dp3⋯dpN)β=π2sin⁡βπ(N−2)−β.(33)\begin{aligned} |\langle\xi|S_1|\xi\rangle| &\le\frac{\pi^2}{\sin\beta\pi}\int|\widehat{\xi}(p_1-p_3,\widehat{p}_{12})|^2\left(\frac{|p_3|^2}{|p_3|^2+\sum_{j=4}^{N}|p_j|^2}\right)^\beta\,dp_1dp_3\cdots dp_N\\ &= \frac{\pi^2}{\sin\beta\pi}\int|\widehat{\xi}(p_1,\widehat{p}_{12})|^2\left(\frac{|p_3|^2}{\sum_{j=3}^{N}|p_j|^2}\right)^\beta\,dp_1dp_3\cdots dp_N\\ &\le\frac{\pi^2}{\sin\beta\pi}\left(\int|\widehat{\xi}(p_1,\widehat{p}_{12})|^2\frac{|p_3|^2}{\sum_{j=3}^{N}|p_j|^2}\,dp_1dp_3\cdots dp_N\right)^\beta\\ &= \frac{\pi^2}{\sin\beta\pi}(N-2)^{-\beta}. \tag*{(33)} \end{aligned}

In the next-to-last step, we used Jensen’s inequality and, in the last step, the bosonic symmetry of ξ\xi in the last N−2N-2 variables. Optimizing over the choice of β\beta thus yields ∥S1∥≤O(N−1ln⁡N)\|S_1\|\le O(N^{-1}\ln N).

Remark 3. Using a product trial state similarly as in the last section, one can show that the NN-dependence in the obtained bound ∥S1∥≤O(N−1ln⁡N)\|S_1\|\le O(N^{-1}\ln N) is actually optimal.

Step 2. Energy localization

The operator K=e2πS0K=e^{\frac{2}{\pi}S_0} in H\mathcal{H} is given by the Fourier multiplier K(q,p3,…,pN)=12q2+∣p^12∣2+1K(q,p_3,\ldots,p_N)=\frac{1}{2}q^2+|\widehat{p}_{12}|^2+1. Let us write

⟨ξ∣S2∣ξ⟩=∫ξ^(p1+p2,p^12)∗ξ^(p3+p4,p^34)∣p⃗∣2+1(1R+1R⊥) dp1⋯dpN(34)\langle\xi|S_2|\xi\rangle=\int\frac{\widehat{\xi}(p_1+p_2,\widehat{p}_{12})^*\widehat{\xi}(p_3+p_4,\widehat{p}_{34})}{|\vec p|^2+1}\left(\mathbb{1}_R+\mathbb{1}_{R}^{\perp}\right)\,dp_1\cdots dp_N \tag*{(34)}

with 1R\mathbb{1}_R the characteristic function of ∣ln⁡(K(p1+p2,p^12)/K(p3+p4,p^34))∣≤R|\ln(K(p_1+p_2,\widehat{p}_{12})/K(p_3+p_4,\widehat{p}_{34}))|\le R for some R>0R>0. The term involving 1R⊥=1−1R\mathbb{1}_{R}^{\perp}=1-\mathbb{1}_R can be bounded with the aid of the following lemma.

Lemma 2. For any R>0R>0, λ≥0\lambda\ge0 and ψ∈L2(R2)\psi\in L^2(\mathbb{R}^2),

∣∫R4ψ(x)∗ψ(y)∣x∣2+∣y∣2+λ21(∣ln⁡(∣x∣2+λ2∣y∣2+λ2)∣>R) dxdy∣≤Ce−R/2∥ψ∥2.(35)\left|\int_{\mathbb{R}^4}\frac{\psi(x)^*\psi(y)}{|x|^2+|y|^2+\lambda^2}\mathbb{1}\left(\left|\ln\left(\frac{|x|^2+\lambda^2}{|y|^2+\lambda^2}\right)\right|>R\right)\,dxdy\right|\le Ce^{-R/2}\|\psi\|^2. \tag*{(35)}

Proof. Without loss of generality, we can restrict our attention to non-negative radial functions ψ\psi. Introducing logarithmic coordinates ∣x∣2+λ2=eu|x|^2+\lambda^2=e^u, ∣y∣2+λ2=ev|y|^2+\lambda^2=e^v, and bounding the denominator by ∣x∣2+∣y∣2+λ2≥12(∣x∣2+∣y∣2+2λ2)|x|^2+|y|^2+\lambda^2\ge\frac{1}{2}(|x|^2+|y|^2+2\lambda^2), the desired bound follows from

∣∫R2f(u)∗f(v)cosh⁡(12(u−v))1(∣u−v∣>R) dudv∣≤Ce−R/2∥f∥2.(36)\left|\int_{\mathbb{R}^2}\frac{f(u)^*f(v)}{\cosh\left(\frac{1}{2}(u-v)\right)}\mathbb{1}(|u-v|>R)\,dudv\right|\le Ce^{-R/2}\|f\|^2. \tag*{(36)}

for functions f∈L2(R)f\in L^2(\mathbb{R}), which is a simple consequence of the Cauchy–Schwarz inequality. ▪

We shall apply this lemma to the integration over p1−p2p_1-p_2 and p3−p4p_3-p_4, with the result that

∫ξ^(p1+p2,p^12)∗ξ^(p3+p4,p^34)∣p⃗∣2+11R dp1⋯dpN≤Ce−R/2∥ξ∥2.(37)\int\frac{\widehat{\xi}(p_1+p_2,\widehat{p}_{12})^*\widehat{\xi}(p_3+p_4,\widehat{p}_{34})}{|\vec{p}|^2+1}\mathbb{1}_R\,dp_1\cdots dp_N \leq Ce^{-R/2}\|\xi\|^2. \tag*{(37)}

To bound the term involving 1R\mathbb{1}_R in (34), we pick a nonnegative χ∈C0∞((−1,1))\chi\in C_0^\infty((-1,1)) with ∫χ2=1\int\chi^2=1 (independent of RR). For L>0L>0 and s∈Rs\in\mathbb{R}, define the Fourier multipliers

χs=L−1/2χ((ln⁡K−s)/L)(38)\chi_s=L^{-1/2}\chi((\ln K-s)/L) \tag*{(38)}

satisfying ∫Rχs2 ds=1\int_{\mathbb{R}}\chi_s^2\,ds=1. For p⃗\vec{p} in the support of 1R\mathbb{1}_R, we shall use that

1−∫Rχs(p1+p2,p^12)χs(p3+p4,p^34) ds=12∥χ−χ(⋅+L−1ln⁡(K(p1+p2,p^12)/K(p3+p4,p^34)))∥22≤CL−2∣ln⁡(K(p1+p2,p^12)/K(p3+p4,p^34))∣2≤C(R/L)2.(39)\begin{aligned} 1-\int_{\mathbb{R}}\chi_s(p_1+p_2,\widehat{p}_{12})\chi_s(p_3+p_4,\widehat{p}_{34})\,ds \\ &=\frac{1}{2}\left\|\chi-\chi\left(\mathord{\cdot}+L^{-1}\ln\left(K(p_1+p_2,\widehat{p}_{12})/K(p_3+p_4,\widehat{p}_{34})\right)\right)\right\|_2^2 \\ &\leq CL^{-2}\left|\ln\left(K(p_1+p_2,\widehat{p}_{12})/K(p_3+p_4,\widehat{p}_{34})\right)\right|^2\leq C(R/L)^2. \tag*{(39)} \end{aligned}

For ξ\xi with ξ^≥0\widehat{\xi}\geq0 we thus conclude that, if R/LR/L is sufficiently small,

⟨ξ∣S2∣ξ⟩≤(1−C(R/L)2)−1⟨ξ∣∫RχsS2χs ds∣ξ⟩+Ce−R/2∥ξ∥2.(40)\langle\xi\mid S_2\mid\xi\rangle\leq\left(1-C(R/L)^2\right)^{-1}\left\langle\xi\left|\int_{\mathbb{R}}\chi_sS_2\chi_s\,ds\right|\xi\right\rangle+Ce^{-R/2}\|\xi\|^2. \tag*{(40)}

In order to bound S2S_2, it thus suffices to consider functions ξ\xi whose Fourier transform ξ^(q,p^12)\widehat{\xi}(q,\widehat{p}_{12}) is supported on a set

es−L≤K(q,p^12)≤es+L(41)e^{s-L}\leq K(q,\widehat{p}_{12})\leq e^{s+L} \tag*{(41)}

for some s∈Rs\in\mathbb{R}.

Step 3. Simple upper bound on S2S_2

Using the change of variables

P=p1+p2,Q=p3+p4,p=(p1−p2)/2,q=(p3−p4)/2,(42)P=p_1+p_2,\quad Q=p_3+p_4,\quad p=(p_1-p_2)/2,\quad q=(p_3-p_4)/2, \tag*{(42)}

we write

⟨ξ∣S2∣ξ⟩=∫ξ^(P,Q/2+q,Q/2−q,p5,…,pN)∗ξ^(Q,P/2+p,P/2−p,p5,…,pN)12(P2+Q2)+2(p2+q2)+∑j=5Npj2+1×dP dp dQ dq dp5⋯dpN.(43)\begin{aligned} \langle\xi\mid S_2\mid\xi\rangle &=\int\frac{\widehat{\xi}(P,Q/2+q,Q/2-q,p_5,\ldots,p_N)^*\widehat{\xi}(Q,P/2+p,P/2-p,p_5,\ldots,p_N)} {\frac{1}{2}(P^2+Q^2)+2(p^2+q^2)+\sum_{j=5}^{N}p_j^2+1}\\ &\qquad\times dP\,dp\,dQ\,dq\,dp_5\cdots dp_N. \tag*{(43)} \end{aligned}

For fixed PP, QQ, p5,…,pNp_5,\ldots,p_N, a simple Cauchy–Schwarz inequality (using the positivity of the operator on L2(R2)L^2(\mathbb{R}^2) with integral kernel 1/(∣x∣2+∣y∣2+λ2)1/(|x|^2+|y|^2+\lambda^2); see Lemma 1) implies that

∫ξ^(P,Q/2+q,Q/2−q,p5,…,pN)∗ξ^(Q,P/2+p,P/2−p,p5,…,pN)12(P2+Q2)+2(p2+q2)+∑j=5Npj2+1 dp dq≤∫ξ^(P,Q/2+q,Q/2−q,p5,…,pN)∗ξ^(P,Q/2+p,Q/2−p,p5,…,pN)12(P2+Q2)+2(p2+q2)+∑j=5Npj2+1 dp dq∫ξ^(Q,P/2+q,P/2−q,p5,…,pN)∗ξ^(Q,P/2+p,P/2−p,p5,…,pN)12(P2+Q2)+2(p2+q2)+∑j=5Npj2+1 dp dq.(44)\int\frac{\widehat{\xi}(P,Q/2+q,Q/2-q,p_5,\ldots,p_N)^*\widehat{\xi}(Q,P/2+p,P/2-p,p_5,\ldots,p_N)}{\frac{1}{2}(P^2+Q^2)+2(p^2+q^2)+\sum_{j=5}^{N}p_j^2+1}\,dp\,dq \leq \sqrt{\int\frac{\widehat{\xi}(P,Q/2+q,Q/2-q,p_5,\ldots,p_N)^*\widehat{\xi}(P,Q/2+p,Q/2-p,p_5,\ldots,p_N)}{\frac{1}{2}(P^2+Q^2)+2(p^2+q^2)+\sum_{j=5}^{N}p_j^2+1}\,dp\,dq} \sqrt{\int\frac{\widehat{\xi}(Q,P/2+q,P/2-q,p_5,\ldots,p_N)^*\widehat{\xi}(Q,P/2+p,P/2-p,p_5,\ldots,p_N)}{\frac{1}{2}(P^2+Q^2)+2(p^2+q^2)+\sum_{j=5}^{N}p_j^2+1}\,dp\,dq}. \tag*{(44)}

We integrate this inequality with respect to PP and QQ and apply another Cauchy–Schwarz inequality with the two square roots on the right side. After integrating with respect to p5,…,pNp_5,\ldots,p_N, we arrive at the bound

⟨ξ∣S2∣ξ⟩≤∫ξ^(P,Q/2+q,Q/2−q,p5,…,pN)∗ξ^(P,Q/2+p,Q/2−p,p5,…,pN)12(P2+Q2)+2(p2+q2)+∑j=5Npj2+1 dP dp dQ dq dp5⋯dpN.(45)\langle\xi\mid S_2\mid\xi\rangle \leq\int\frac{\widehat{\xi}(P,Q/2+q,Q/2-q,p_5,\ldots,p_N)^*\widehat{\xi}(P,Q/2+p,Q/2-p,p_5,\ldots,p_N)}{\frac{1}{2}(P^2+Q^2)+2(p^2+q^2)+\sum_{j=5}^{N}p_j^2+1}\,dP\,dp\,dQ\,dq\,dp_5\cdots dp_N. \tag*{(45)}

Moreover, for ξ^≥0\widehat{\xi}\geq0 we can further bound the denominator from below by 12(P2+Q2)+p2+q2+∑j=5Npj2+1\frac{1}{2}(P^2+Q^2)+p^2+q^2+\sum_{j=5}^{N}p_j^2+1, which can be written as

112(P2+Q2)+p2+q2+∑j=5Npj2+1=∫0∞e−t2(K(p1+p2,p^12)+K(p3+p4,p^34)) dt,(46)\frac{1}{\frac{1}{2}(P^2+Q^2)+p^2+q^2+\sum_{j=5}^{N}p_j^2+1} =\int_{0}^{\infty}e^{-\frac{t}{2}\left(K(p_1+p_2,\widehat{p}_{12})+K(p_3+p_4,\widehat{p}_{34})\right)}\,dt, \tag*{(46)}

where on the right side we undid the change of variables (4.15). The resulting expression is thus given by

(2π)2∫0∞⟨ξ∣e−tK/2δ34e−tK/2∣ξ⟩ dt,(47)(2\pi)^2\int_{0}^{\infty}\langle\xi\mid e^{-tK/2}\delta_{34}e^{-tK/2}\mid\xi\rangle\,dt, \tag*{(47)}

where δ34\delta_{34} denotes multiplication by δ(x3−x4)\delta(x_3-x_4). That is, we have shown that

⟨ξ∣S2∣ξ⟩≤(2π)2∫0∞⟨ξ∣e−tK/2δ34e−tK/2∣ξ⟩ dt(48)\langle\xi\mid S_2\mid\xi\rangle\leq(2\pi)^2\int_{0}^{\infty}\langle\xi\mid e^{-tK/2}\delta_{34}e^{-tK/2}\mid\xi\rangle\,dt \tag*{(48)}

if ξ^≥0\widehat{\xi}\geq0.

Step 4. Mollifying the interaction.

We shall now replace the interaction δ(x3−x4)\delta(x_3-x_4) in (48) by a mollified interaction w(x3−x4)w(x_3-x_4). Assume that ww is a smooth, nonnegative, compactly supported, even function on R2\mathbb{R}^2, with ∫R2w=1\int_{\mathbb{R}^2}w=1 and

∫R2w(x)eipx dx≥1−Cw∣p∣2\int_{\mathbb{R}^2}w(x)e^{ipx}\,dx\geq1-C_w|p|^2

for all p∈R2p\in\mathbb{R}^2. For ξ^\widehat{\xi} supported on p⃗\vec{p} satisfying (41), all momenta p3,…,pNp_3,\ldots,p_N in the support of ξ^\widehat{\xi} satisfy ∣pj∣≤e(s+L)/2|p_j|\leq e^{(s+L)/2}. In particular, ∣p−q∣≤2e(s+L)/2|p-q| \le2e^{(s+L)/2} in the integrand on the right hand side of (45). Consequently, with wκ(x)=κ2w(κx)w^{\kappa}(x)=\kappa^{2}w(\kappa x) for some κ>0\kappa>0

∫R2wκ(x)ei(p−q)x dx≥1−Cwκ−2∣p−q∣2≥1−4Cwκ−2es+L.(49)\int_{\mathbb{R}^{2}} w^{\kappa}(x)e^{i(p-q)x}\,dx \ge1-C_{w}\kappa^{-2}|p-q|^{2} \ge1-4C_{w}\kappa^{-2}e^{s+L}. \tag*{(49)}

We thus conclude that

⟨ξ∣e−tK/2δ34e−tK/2∣ξ⟩≤(1−4Cwκ−2es+L)−1⟨ξ∣e−tK/2w34κe−tK/2∣ξ⟩(50)\langle\xi\lvert e^{-tK/2}\delta_{34}e^{-tK/2}\rvert\xi\rangle\le\left(1-4C_{w}\kappa^{-2}e^{s+L}\right)^{-1}\langle\xi\lvert e^{-tK/2}w_{34}^{\kappa}e^{-tK/2}\rvert\xi\rangle \tag*{(50)}

for any ξ\xi with ξ^≥0\widehat{\xi}\ge0 supported in (41). Here w34κw_{34}^{\kappa} is short for multiplication by wκ(x3−x4)w^{\kappa}(x_{3}-x_{4}).

Step 5. Ground state energy in the mean-field limit

With ww as above, pick 0<a<a∗0<a<a_{*} (defined in (1)), and let

En(a)=inf⁡spec⁡[∑i=1n(−Δi)−an−1∑1≤i<j≤nw(xi−xj)](51)E_{n}(a)=\inf\operatorname{spec}\left[\sum_{i=1}^{n}(-\Delta_{i})-\frac{a}{n-1}\sum_{1\le i<j\le n}w(x_{i}-x_{j})\right] \tag*{(51)}

for n≥2n\ge2, where the operator acts in Lsym2(R2n)L_{\mathrm{sym}}^{2}(\mathbb{R}^{2n}). Then, by a simple rescaling, we have the bound

∑1≤i<j≤nwκ(xi−xj)≤n−1a(−κ2En(a)+∑i=1n(−Δi))(52)\sum_{1\le i<j\le n}w^{\kappa}(x_{i}-x_{j})\le\frac{n-1}{a}\left(-\kappa^{2}E_{n}(a)+\sum_{i=1}^{n}(-\Delta_{i})\right) \tag*{(52)}

for any κ>0\kappa>0. We shall apply this in (50), for n=N−2n=N-2. By the bosonic symmetry of ξ\xi in the last N−2N-2 variables,

∫0∞⟨ξ∣e−tK/2w34κe−tK/2∣ξ⟩ dt≤2a(N−2)(−κ2EN−2(a)⟨ξ∣1K∣ξ⟩+∑j=3N⟨ξ∣−ΔjK∣ξ⟩)≤2a(N−2)(1−eL−sκ2EN−2(a)),(53)\begin{aligned} \int_{0}^{\infty}\langle\xi\lvert e^{-tK/2}w_{34}^{\kappa}e^{-tK/2}\rvert\xi\rangle\,dt \\ &\le\frac{2}{a(N-2)}\left(-\kappa^{2}E_{N-2}(a)\langle\xi\lvert\frac{1}{K}\rvert\xi\rangle+\sum_{j=3}^{N}\langle\xi\lvert\frac{-\Delta_{j}}{K}\rvert\xi\rangle\right) \\ &\le\frac{2}{a(N-2)}\left(1-e^{L-s}\kappa^{2}E_{N-2}(a)\right), \tag*{(53)} \end{aligned}

where we used the lower bound on KK in (41) in the first term on the right hand side.

Step 6. Combining the bounds

A combination of (48), (50) and (53) shows that, for any normalized ξ\xi with ξ^≥0\widehat{\xi}\ge0 supported in (41) for some s∈Rs\in\mathbb{R},

⟨ξ∣S2∣ξ⟩≤2(2π)2a(N−2)(1−4Cwκ−2es+L)−1(1−eL−sκ2EN−2(a))(54)\langle\xi\lvert S_{2}\rvert\xi\rangle\le\frac{2(2\pi)^{2}}{a(N-2)}\left(1-4C_{w}\kappa^{-2}e^{s+L}\right)^{-1}\left(1-e^{L-s}\kappa^{2}E_{N-2}(a)\right) \tag*{(54)}

for any κ>0\kappa>0. Given ε>0\varepsilon>0, we shall choose κ\kappa such that κ−2es+L=ε\kappa^{-2}e^{s+L}=\varepsilon, hence this becomes

⟨ξ∣S2∣ξ⟩≤2(2π)2a(N−2)(1−4Cwε)−1(1−e2Lε−1EN−2(a)),(55)\langle\xi\lvert S_{2}\rvert\xi\rangle\le\frac{2(2\pi)^{2}}{a(N-2)}(1-4C_{w}\varepsilon)^{-1}\left(1-e^{2L}\varepsilon^{-1}E_{N-2}(a)\right), \tag*{(55)}

which is now independent of ss. Further combining this with (40), we finally conclude that

∥S2∥≤2(2π)2a(N−2)(1−C(R/L)2)−1(1−4Cwε)−1(1−e2Lε−1EN−2(a))+Ce−R/2.(56)\lVert S_2\rVert\le\frac{2(2\pi)^2}{a(N-2)}\left(1-C(R/L)^2\right)^{-1}\left(1-4C_w\varepsilon\right)^{-1}\left(1-e^{2L\varepsilon^{-1}E_{N-2}(a)}\right)+Ce^{-R/2}. \tag*{(56)}

This bound holds for all R,L>0R,L>0 such that R/LR/L is sufficiently small and for all ε>0\varepsilon>0. For the first error term to be small, we choose R/L=ε1/2R/L=\varepsilon^{1/2}. For the last one we need e−R/2=ε/Ne^{-R/2}=\varepsilon/N. This leaves us with showing that

ε−1(N/ε)4ε−1/2EN−2(a)(57)\varepsilon^{-1}(N/\varepsilon)^{4\varepsilon^{-1/2}}E_{N-2}(a) \tag*{(57)}

is small for large NN, i.e., that EN−2(a)E_{N-2}(a) goes to zero faster than any power of NN. This will be deduced in the final step from the work in [7]. Accepting this for the moment, we deduce that

∥S2∥≤2(2π)2a(N−2)(1+o(1))(58)\lVert S_2\rVert\le\frac{2(2\pi)^2}{a(N-2)}(1+o(1)) \tag*{(58)}

as N→∞N\to\infty.

Combining this with the operator inequality S0≥0S_0\ge0 and the norm bound on S1S_1 from Step 1, we deduce that

N−1tN≥−2(1−2/N)∥S1∥−12(N−2)(1−3/N)∥S2∥≥−(2π)2a−1(1+o(1)).(59)N^{-1}t_N\ge-2(1-2/N)\lVert S_1\rVert-\frac{1}{2}(N-2)(1-3/N)\lVert S_2\rVert\ge-(2\pi)^2a^{-1}(1+o(1)). \tag*{(59)}

Since a<a∗a<a_* is arbitrary, this implies

lim inf⁡N→∞tNN≥−(2π)2a∗,(60)\liminf_{N\to\infty}\frac{t_N}{N}\ge-\frac{(2\pi)^2}{a_*}, \tag*{(60)}

which is the claimed lower bound.

Step 7. Rough bound on ground state energy

In [7 Thm. 2], it was shown that for any 0<a<a∗0<a<a_* and κn=enβ\kappa_n=e^{n^\beta} for some β<1\beta<1,

∑i=1n(−Δi+V(xi))−an−1∑1≤i<j≤nwκn(xi−xj)≥−nK(61)\sum_{i=1}^{n}(-\Delta_i+V(x_i))-\frac{a}{n-1}\sum_{1\le i<j\le n}w^{\kappa_n}(x_i-x_j)\ge-nK \tag*{(61)}

for some constant KK depending only on aa, ww, VV and β\beta. Here VV is a confining potential growing like a power of ∣x∣|x|. After rescaling by κn\kappa_n, this becomes

∑i=1n(−Δi+κn−2V(xi/κn))−an−1∑1≤i<j≤nw(xi−xj)≥−nκn−2K.(62)\sum_{i=1}^{n}\left(-\Delta_i+\kappa_n^{-2}V(x_i/\kappa_n)\right)-\frac{a}{n-1}\sum_{1\le i<j\le n}w(x_i-x_j)\ge-n\kappa_n^{-2}K. \tag*{(62)}

To get a lower bound on En(a)E_n(a), we can use IMS localization. Let Hn(a)H_n(a) denote the operator on the right-hand side of (51). It is convenient to view it as an operator on all of L2(R2n)L^2(\mathbb{R}^{2n}), it is well-known that the infimum of its spectrum is unchanged upon restriction to the bosonic subspace. IMS localization implies that for any ℓ>0\ell>0 there exist zi∈R2z_i\in\mathbb{R}^2 and a ψ∈L2(R2n)\psi\in L^2(\mathbb{R}^{2n}) supported in ∏i=1nBℓ(zi)\prod_{i=1}^{n}B_\ell(z_i) such that ⟨ψ∣Hn(a)∣ψ⟩≤En(a)+nCℓ−2\langle\psi|H_n(a)|\psi\rangle\le E_n(a)+nC\ell^{-2}, with Bℓ(z)⊂R2B_\ell(z)\subset\mathbb{R}^2 denoting the ball of radius ℓ\ell centered at z∈R2z\in\mathbb{R}^2. Since ww is compactly supported by assumption, an interaction term w(xi−xj)w(x_i-x_j) will be zero if ∣zi−zj∣>2ℓ+d|z_i-z_j|>2\ell+d, with dd the diameter of the support of ww. We thus slightly enlarge the balls BℓB_\ell to Bℓ+d/2B_{\ell+d/2}, and write their union as a disjoint union of mm clusters, with 1≤m≤n1\le m\le n, each consisting of finitely many balls: ⋃i=1nBℓ+d/2(zi)=⋃j=1mSj\bigcup_{i=1}^{n}B_{\ell+d/2}(z_i)=\bigcup_{j=1}^{m}S_j, the latter sets being mutually disjoint. Since there are no interactions between particles in different clusters, we have

⟨ψ∣Hn(a)∣ψ⟩=∑j=1m⟨ψ∣HSj∣ψ⟩(63)\langle\psi|H_n(a)|\psi\rangle=\sum_{j=1}^{m}\langle\psi|H^{S_j}|\psi\rangle \tag*{(63)}

where HSjH^{S_j} is the Hamiltonian for the particles in cluster SjS_j.

Each cluster SjS_j is a connected union of 1≤mj≤n1\le m_j\le n balls of radius ℓ+d/2\ell+d/2, and is therefore contained in a ball BD(wj)B_D(w_j) for some wj∈R2w_j\in\mathbb{R}^2, where D:=n(ℓ+d/2)D:=n(\ell+d/2). We now translate all the coordinates in SjS_j by wjw_j, i.e., center all the balls at the origin. This leaves the interactions within clusters invariant, but may introduce additional attractive interactions between particles in different clusters. The translated wave function ψ~\widetilde{\psi} hence satisfies ⟨ψ~∣Hn(a)∣ψ~⟩≤⟨ψ∣Hn(a)∣ψ⟩\langle\widetilde{\psi}|H_n(a)|\widetilde{\psi}\rangle\le\langle\psi|H_n(a)|\psi\rangle. Since by construction ψ~\widetilde{\psi} is supported in ⨂i=1nBD(0)\bigotimes_{i=1}^{n}B_D(0), we can add one-body potential terms D−2∑k=1nV(xk/D)D^{-2}\sum_{k=1}^{n}V(x_k/D) to the Hamiltonian Hn(a)H_n(a) at the energy cost of at most nD−2sup⁡B1(0)VnD^{-2}\sup_{B_1(0)}V. Using (4.35) with κn=D\kappa_n=D, i.e., choosing ℓ=κn/n−d/2\ell=\kappa_n/n-d/2, we conclude that

En(a)≥−Cnℓ2−nD2(K+sup⁡B1(0)V)≥−Cn3κn−2(64)E_n(a)\ge-\frac{Cn}{\ell^2}-\frac{n}{D^2}\left(K+\sup_{B_1(0)}V\right)\ge-Cn^3\kappa_n^{-2} \tag*{(64)}

in the case κn≥dn\kappa_n\ge dn. This completes the proof.

Acknowledgments. R.S. is grateful for inspiring discussions with Artem Volosniev. R.L.F. acknowledges partial support by the German Research Foundation grants EXC-2111-390814868 and TRR 352–Project-ID 470903074, as well as through the Chat-GPT for Academic Researchers program of OpenAI. R.S. acknowledges partial support by the Austrian Science Fund (FWF) grant DOI: 10.55776/I6427 (as part of the SFB/TRR 352).

AI declaration. The entire text of the manuscript was written by the authors without assistance of AI tools. Crucial ideas for the proof of the lower bound in § 4 were obtained with the help of GPT-6 Astra. Claude Opus 5.5 was used for proofreading.

Rupert L. Frank, Mathematisches Institut, Ludwig-Maximilians Universität München, Theresienstr. 39, 80333 München, Germany, and Munich Center for Quantum Science and Technology, Schellingstr. 4, 80799 München, Germany

Email address: [email protected]

Robert Seiringer, Institute of Science and Technology Austria, Am Campus 1, 3400 Klosterneuburg, Austria

Email address: [email protected]

References

  1. [1]F. Bentosela, R.M. Cavalcanti, P. Exner, V.A. Zagrebnov, Anomalous electron trapping by localized magnetic fields, J. Phys. A 32, 3029–3039 (1999).DOI
  2. [2]F.G. Brandão, A.W. Harrow, Quantum de Finetti theorems under local measurements with applications, Commun. Math. Phys. 353, 469–506 (2017).
  3. [3]G. Dell-Antonio, R. Figari, A. Teta, Hamiltonians for systems of N particles interacting through point interactions, Ann. Inst. Henri Poincaré 60, 253–290 (1994).
  4. [4]J. Dimock, S.G. Rajeev, Multi-particle Schrödinger operators with point interactions in the plane, J. Phys. A: Math. Gen. 37, 9157 (2004).DOI
  5. [5]M. Griesemer, M. Hofacker, From Short-Range to Contact Interactions in Two-dimensional Many-Body Quantum Systems, Ann. Henri Poincaré 23, 2769–2818 (2022).
  6. [6]H.-W. Hammer, D.T. Son, Universal Properties of Two-Dimensional Boson Droplets, Phys. Rev. Lett. 92, 250408 (2004).pubmed.ncbi.nlm.nih.gov/15697882
  7. [7]L. Junge, F.L.A. Visconti, Derivation of Hartree theory for two-dimensional attractive Bose gases in almost Gross–Pitaevskii regime, J. Math. Phys. Appl. 205, 103800 (2026).arxiv.org/abs/2410.23150
  8. [8]M. Lewin, P.T. Nam, N. Rougerie, Derivation of Hartree’s theory for generic mean-field Bose systems, Adv. Math. 254, 570–621 (2014).arxiv.org/abs/1303.0981
  9. [9]K. Li, G. Smith, Quantum de Finetti theorem under fully-one-way adaptive measurements, Phys. Rev. Lett. 114, 160503 (2015).pubmed.ncbi.nlm.nih.gov/25955040
  10. [10]S.G. Rajeev, A Condensation of Interacting Bosons in Two Dimensional Space, unpublished, preprint arXiv:9905120
  11. [11]B. Simon, The Bound State of Weakly Coupled Schrödinger Operators in One and Two Dimensions, Ann. Phys. 97, 279–288 (1976).
  12. [12]G.V. Skorniakov, K.A. Ter-Martirosian, Three body problem for short range forces. I. Scattering of low energy neutrons by deuterons, Sov. Phys. JETP 4, 648–661 (1957).
  13. [13]M. Suchorowski, F. Brauneis, H.-W. Hammer, M. Tomza, A.G. Volosniev, Generalized Gross–Pitaevskii Equation for 2D Bosons with Attractive Interactions, Phys. Rev. Lett. 136, 243402 (2026).DOI
  14. [14]M.I. Weinstein, Nonlinear Schrödinger equations and sharp interpolation estimates, Comm. Math. Phys. 87, 567–576 (1983).

Paper details

vlnc:W_ke6nsyu6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 6

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

GPT-6 Astra, Claude Opus 5.5

View statement
Looking for related papers…
Contents