betaFeedback

WorksMathematics

Analyticity of Sine⁡β\operatorname{Sine}_β pair correlations

Abstract

We prove that the pair correlation function of Sine⁡β\operatorname{Sine}_β extends holomorphically in ββ to ℜβ>0\Reβ>0 at every fixed nonzero separation. Our proof combines a random walk representation of the stochastic sine phase moments, stretched-exponential Fourier decay obtained by coupling, and weighted energy estimates for complex ββ.

Introduction

Let Ξβ\Xi_\beta be the Sineβ_\beta point process, normalized to have intensity 1/(2π)1/(2\pi). Its pair correlation function ρβ\rho_\beta is defined by

E∑x,y∈Ξβx≠yF(x,y)=∫R2F(x,y)ρβ(x,y) dx dy\mathbb{E}\sum_{\substack{x,y\in\Xi_\beta\\x\ne y}}F(x,y)=\int_{\mathbb{R}^2}F(x,y)\rho_\beta(x,y)\,\mathrm{d}x\,\mathrm{d}y

for nonnegative compactly supported Borel functions FF, where the sum runs over ordered pairs of distinct points.

Theorem 1 (Analyticity of the pair correlation). For every β>0\beta>0, the pair correlation function ρβ\rho_\beta has a continuous version off the diagonal. For this version and each λ>0\lambda>0, the function

β⟼ρβ(0,λ)\beta\longmapsto\rho_\beta(0,\lambda)

extends holomorphically from (0,∞)(0,\infty) to {Re⁡β>0}\{\operatorname{Re}\beta>0\}. The extension is jointly continuous in (β,λ)(\beta,\lambda) on {Re⁡β>0}×(0,∞)\{\operatorname{Re}\beta>0\}\times(0,\infty).

Figure 1 shows the continuation numerically.

Contour plots and line plots of the holomorphic continuation

Figure 1. The holomorphic continuation of 4π2ρβ(0,λ)4\pi^2\rho_\beta(0,\lambda), computed numerically from (5). The black curves on the contour plots are where 4π2∣ρβ(0,λ)∣=14\pi^2\lvert\rho_\beta(0,\lambda)\rvert=1.

Dependence on inverse temperature has long been studied through exact formulas and perturbation theory. For β=1,2,4\beta=1,2,4, the correlation functions of Sineβ_\beta have classical explicit formulas: determinantal for β=2\beta=2 and Pfaffian for β=1,4\beta=1,4 [10 Chapters 5–7]. Forrester [8, 9] obtained correlation formulas at even integer β\beta, and Ha [12] extended these to rational coupling. Okounkov [19] subsequently obtained

Selberg-type integral formulas for nn-point correlations at rational β\beta. Forrester, Jancovici and McAnally (2001) computed first-order perturbations around β=2\beta=2 and 44, following earlier work on the two-dimensional plasma [14]. More recently, Qu and Valkó (2025) obtained a stochastic representation for all β>0\beta>0 and proved continuity in β\beta. Theorem 1 strengthens continuity to holomorphic dependence throughout the right half-plane.

Translation and reflection invariance reduce the pair correlation to ρβ(0,λ)\rho_\beta(0,\lambda), λ>0\lambda>0. Our proof uses the stochastic phase representation of Sineβ_\beta [16]. We express the pair correlation as the phase density at zero divided by 2π2\pi, and use a random walk representation of the phase moments to prove analyticity.

Let Z=B1+iB2Z=B_1+iB_2, where B1B_1, B2B_2 are independent two-sided standard real Brownian motions. For real β>0\beta>0 and λ∈R\lambda\in\mathbb{R}, let αλ,β\alpha_{\lambda,\beta} solve the phase SDE

dαλ,β(t)=βλ4eβt/4 dt+Re⁡[(e−iαλ,β(t)−1)dZt],t≤0,\mathrm{d}\alpha_{\lambda,\beta}(t)=\frac{\beta\lambda}{4}e^{\beta t/4}\,\mathrm{d}t+\operatorname{Re}\left[\left(e^{-i\alpha_{\lambda,\beta}(t)}-1\right)\mathrm{d}Z_t\right],\qquad t\leq0,
lim⁡t→−∞αλ,β(t)=0.(1)\lim_{t\to-\infty}\alpha_{\lambda,\beta}(t)=0. \tag*{(1)}

Equation (1) is the Killip–Stoiciu equation in logarithmic time, as considered in Valkó and Virág [17]. All spectral parameters are driven by the same ZZ. We often write αλ\alpha_\lambda when β\beta is fixed or clear from context. If UU is uniform on [0,2π)[0,2\pi) and independent of the phase field, then

Ξβ={λ∈R:αλ,β(0)=Umod⁡2π}(2)\Xi_\beta=\{\lambda\in\mathbb{R}:\alpha_{\lambda,\beta}(0)=U\mathbin{\operatorname{mod}}2\pi\} \tag*{(2)}

has the Sineβ_\beta distribution (Killip and Stoiciu [16]). Write T=R/(2πZ)\mathbb{T}=\mathbb{R}/(2\pi\mathbb{Z}) and let pβ,λp_{\beta,\lambda} denote the density of αλ,β(0)mod⁡2π\alpha_{\lambda,\beta}(0)\mathbin{\operatorname{mod}}2\pi with respect to dθd\theta on T\mathbb{T}, for λ>0\lambda>0. For k∈Zk\in\mathbb{Z}, define its Fourier coefficients

qk(β,λ)=Eeikαλ,β(0).q_k(\beta,\lambda)=\mathbb{E}e^{ik\alpha_{\lambda,\beta}(0)}.

Theorem 2 (Pair correlation and phase density). For real β,λ>0\beta,\lambda>0, the pair correlation off the diagonal and the phase density have continuous versions, and

ρβ(0,λ)=12πpβ,λ(0)=14π2∑k∈Zqk(β,λ),(3)\rho_\beta(0,\lambda)=\frac{1}{2\pi}p_{\beta,\lambda}(0)=\frac{1}{4\pi^2}\sum_{k\in\mathbb{Z}}q_k(\beta,\lambda), \tag*{(3)}

with absolute convergence.

The existence of correlation functions of all orders for Sineβ_\beta, β>0\beta>0, was proved by Qu and Valkó (2025). More recently, Assiotis and Najnudel [1] represented these functions in terms of the Hua–Pickrell stochastic zeta function introduced by Li and Valkó (2022), and established their continuity and boundedness in the spatial variables. For higher correlations, we obtain the following representation at the level of correlation measures.

Proposition 3 (Phase representation of correlation measures). Let β>0\beta>0, k≥2k\ge2, and let I0,…,Ik−1I_0,\ldots,I_{k-1} be pairwise disjoint bounded intervals. Write αλ=αλ,β(0)\alpha_\lambda=\alpha_{\lambda,\beta}(0). Then

E∏j=0k−1Ξβ(Ij)=12πlim⁡ε↓0∫I0×⋯×Ik−1P(max⁡1≤j≤k−1dist⁡(αλj−λ0,2πZ)<ε)(2ε)k−1 dλ0⋯dλk−1.(4)\mathbb{E}\prod_{j=0}^{k-1}\Xi_\beta(I_j) =\frac{1}{2\pi}\lim_{\varepsilon\downarrow0}\int_{I_0\times\cdots\times I_{k-1}} \frac{\mathbb{P}\left(\max_{1\le j\le k-1}\operatorname{dist}(\alpha_{\lambda_j-\lambda_0},2\pi\mathbb{Z})<\varepsilon\right)} {(2\varepsilon)^{k-1}}\,d\lambda_0\cdots d\lambda_{k-1}. \tag*{(4)}

The proof is in § 5. When k=2k=2, continuity of the phase density turns this limit into (3).

Itô’s formula and the scaling invariance of the phase SDE (1) yield

λ∂λqk=ikλqk+2k2β(qk+1+qk−1−2qk),q0=1,(5)\lambda\partial_\lambda q_k=ik\lambda q_k+\frac{2k^2}{\beta}(q_{k+1}+q_{k-1}-2q_k),\qquad q_0=1, \tag*{(5)}

where all coefficients are evaluated at (β,λ)(\beta,\lambda). The derivation is given in § 3. To extend (3) to complex β\beta, we first continue each coefficient and then establish locally uniform

summability. The key tool is a representation by a random walk whose law does not depend on β\beta.

Proposition 4 (Random walk representation). Let (St)t≥0(S_t)_{t\geq0} be the continuous-time nearest-neighbor random walk on Z≥0\mathbb{Z}_{\geq0}, absorbed at zero. From k≥1k\geq1, its jump rate to each of k−1k-1 and k+1k+1 is k2/2k^2/2. Let Ek\mathbb{E}_k denote expectation for the walk started at kk. For real β>0\beta>0, λ>0\lambda>0, and k≥1k\geq1,

qk(β,λ)=EkeiλAβ,Aβ=β4∫0∞e−βt/4St dt.(6)q_k(\beta,\lambda)=\mathbb{E}_k e^{i\lambda A_\beta},\qquad A_\beta=\frac{\beta}{4}\int_0^\infty e^{-\beta t/4}S_t\,\mathrm{d}t. \tag*{(6)}

Section 3 proves this formula by applying the countable-state Feynman–Kac formula on [−T,0][-T,0] and then sending T→∞T\to\infty. The embedded simple random walk is absorbed after finitely many jumps almost surely, so SS is nonexplosive and the integral in (6) is almost surely finite. Write τj=inf⁡{t≥0:St=j}\tau_j=\inf\{t\geq0:S_t=j\}. For 1≤k≤n1\leq k\leq n, the finite-space cutoff is

qk,n(β,λ)=Ek[eiλAβ1{τ0<τn+1}].(7)q_{k,n}(\beta,\lambda)=\mathbb{E}_k\left[e^{i\lambda A_\beta}\mathbf{1}_{\{\tau_0<\tau_{n+1}\}}\right]. \tag*{(7)}

Gambler’s ruin gives the uniform real-parameter estimate

∣qk(β,λ)−qk,n(β,λ)∣≤kn+1,β>0,λ>0.(8)\left|q_k(\beta,\lambda)-q_{k,n}(\beta,\lambda)\right|\leq\frac{k}{n+1},\qquad\beta>0,\quad\lambda>0. \tag*{(8)}

The cutoffs are holomorphic on Re⁡β>0\operatorname{Re}\beta>0. A weighted energy estimate provides a bound independent of nn, allowing Vitali’s theorem to remove the cutoff. The details are in Section 6.

Proposition 5 (Holomorphic Fourier coefficients). For every fixed λ>0\lambda>0 and k∈Zk\in\mathbb{Z}, qk(β,λ)q_k(\beta,\lambda) has a unique holomorphic extension to Re⁡β>0\operatorname{Re}\beta>0. These extensions satisfy

∑k=1∞∣qk(β,λ)∣2k2≤4∣β∣2(Re⁡β)2,λ>0,Re⁡β>0.(9)\sum_{k=1}^{\infty}\frac{|q_k(\beta,\lambda)|^2}{k^2}\leq4\frac{|\beta|^2}{(\operatorname{Re}\beta)^2},\qquad\lambda>0,\quad\operatorname{Re}\beta>0. \tag*{(9)}

Here q0(β,λ)=1q_0(\beta,\lambda)=1 and q−k(β,λ)=qk(β‾,λ)‾q_{-k}(\beta,\lambda)=\overline{q_k(\overline{\beta},\lambda)} for $k\geq 1.

To sum these holomorphic extensions in (3), we need locally uniform bounds that are summable in kk. For real β>0\beta>0, a coupling argument for the random walk in Proposition 4 gives stretched-exponential decay (Section 4). Combining this decay with the energy bound (9), we obtain the same type of estimate for complex β\beta with Re⁡β>0\operatorname{Re}\beta>0 (Section 7):

Proposition 6 (Stretched-exponential Fourier decay for complex β\beta). For every compact K⊂{Re⁡β>0}K\subset\{\operatorname{Re}\beta>0\} and compact interval L⊂(0,∞)L\subset(0,\infty), there are constants a,b>0a,b>0 such that

sup⁡β∈Kλ∈L∣qk(β,λ)∣≤be−a∣k∣1/3,k∈Z.\sup_{\substack{\beta\in K\\ \lambda\in L}}|q_k(\beta,\lambda)|\leq b e^{-a|k|^{1/3}},\qquad k\in\mathbb{Z}.

Thus the series in (3) converges locally uniformly for Re⁡β>0\operatorname{Re}\beta> 0 and gives the extension in Theorem 1.

§ 2 records the phase representation and the probabilistic and analytic preliminaries. § 3 and § 4 establish the random walk representation and real Fourier decay. § 5 identifies the pair correlation. § 6 and § 7 complete the analytic continuation argument. Section 8 gives explicit formulas for the Fourier moments and phase density at β=2\beta= 2, recovering the sine-kernel pair correlation. The resulting moment asymptotics show that the decay exponent 1/31/3 in Proposition 6 cannot be increased beyond 1/21/2 in general.

Preliminaries

The phase and its boundary angle

Proposition 7 (The stochastic sine phase). For real β>0\beta> 0, the phase SDE (1) has a pathwise unique solution for the prescribed driving Brownian motion. Almost surely λ↦αλ(0)\lambda\mapsto\alpha_\lambda(0) is a real analytic increasing bijection of R\mathbb{R}, with α0(0)=0\alpha_0(0) = 0 and ∂λαλ(0)>0\partial_\lambda\alpha_\lambda(0) > 0. Moreover, for λ∈R\lambda\in\mathbb{R} and t≤0t \le0,

αλ(t)=dαλeβt/4(0),Eαλ(t)=λeβt/4.(10)\alpha_\lambda(t) \mathrel{\mathop{=}^{d}} \alpha_{\lambda e^{\beta t/4}}(0), \qquad\mathbb{E}\alpha_\lambda(t) = \lambda e^{\beta t/4}. \tag*{(10)}

Let UU be uniform on [0,2π)[0,2\pi) and independent of ZZ, and define Ξβ\Xi_\beta by (2). Then Ξβ\Xi_\beta has the Sineβ_\beta distribution. The corresponding normalized spectral weights

wλ=1∂λαλ(0),λ∈Ξβ,w_\lambda= \frac{1}{\partial_\lambda\alpha_\lambda(0)}, \qquad\lambda\in\Xi_\beta,

are mutually independent gamma variables of shape β/2\beta/2 and mean one, and are independent of Ξβ\Xi_\beta. These weights are half the right spectral weights in Valkó and Virág (2023, Proposition 3 and Lemma 17). The phase and shooting representations follow from Valkó and Virág (2022, Corollary 51); see also Valkó and Virág (2023, Definition 16 and Lemma 17). Time translation in (1) gives the scaling in (10). The expectation identity follows from Killip and Stoiciu (2009, Proposition 4.5) after the time change s=eβt/4s = e^{\beta t/4}.

The phase representation was introduced by Killip and Stoiciu [16]; its equivalence with the Gaussian bulk limit of Valkó and Virág (2009) was proved by Nakano [18] and Valkó and Virág (2017, Theorem 28).

Lemma 8 (Phase and counting moments). For each 0<L<∞0 < L < \infty there is ηL>0\eta_L > 0 such that

sup⁡β>0, 0≤λ≤L, t≤0EeηLαλ(t)<∞.(11)\sup_{\beta>0,\,0\leq\lambda\leq L,\,t\leq0} \mathbb{E}e^{\eta_L\alpha_\lambda(t)} < \infty. \tag*{(11)}

Thus phase moments of every positive order are locally bounded in λ\lambda, uniformly in β>0\beta> 0 and t≤0t \le0. Counts in bounded intervals have moments of every positive order as well.

Proof. Put Nλ=Eβ((0,λ])N_\lambda= \mathbb{E}_\beta((0,\lambda]). The counting bound in Valkó and Virág (2009, Proposition 9(viii)), applied with total drift λ\lambda, gives

P(Nλ≥ak)≤2(λ2πa)k,a,k∈N.\mathbb{P}(N_\lambda\ge ak) \le2\left(\frac{\lambda}{2\pi a}\right)^k,\qquad a,k\in\mathbb{N}.

Choose an integer a>L/(2π)a > L/(2\pi). This gives a positive exponential moment for NλN_\lambda uniformly for λ≤L\lambda\le L and β>0\beta> 0. By (2), almost surely,

Nλ=⌊αλ(0)+2π−U2π⌋.(12)N_\lambda=\left\lfloor\frac{\alpha_\lambda(0)+2\pi-U}{2\pi}\right\rfloor. \tag*{(12)}

Consequently, ∣2πNλ−αλ(0)∣≤2π|2\pi N_\lambda-\alpha_\lambda(0)|\le2\pi, so the phase has the same moment property. The scaling identity in (10) gives the bound for t≤0t\le0. Stationarity gives the count assertion for any bounded interval.

Probabilistic and analytic tools

Lemma 9 (Feynman–Kac). Let SS be a nonexplosive continuous-time Markov chain on a countable state space EE, with finite total jump rates and generator GG. Fix T>0T>0, and let v:[0,T]×E→Cv:[0,T]\times E\to\mathbb{C} be continuous in time, with Re⁡v(t,k)≤C\operatorname{Re}v(t,k)\le C for some C≥0C\ge0. Every bounded solution u:[0,T]×E→Cu:[0,T]\times E\to\mathbb{C} of

∂tu(t,⋅)=Gu(t,⋅)+v(t,⋅)u(t,⋅),u(0,⋅)=g(13)\partial_tu(t,\cdot)=Gu(t,\cdot)+v(t,\cdot)u(t,\cdot),\qquad u(0,\cdot)=g \tag*{(13)}

with u(⋅,k)∈C1([0,T])u(\cdot,k)\in C^1([0,T]) for each k∈Ek\in E, satisfies

u(t,k)=Ek[exp⁡(∫0tv(t−s,Ss) ds)g(St)],0≤t≤T.(14)u(t,k)=\mathbb{E}_k\left[\exp\left(\int_0^t v(t-s,S_s)\,\mathrm{d}s\right)g(S_t)\right],\qquad0\le t\le T. \tag*{(14)}

In particular, there is at most one bounded solution.

Proof. Fix t≤Tt\le T, and set

Mr=exp⁡(∫0rv(t−s,Ss) ds)u(t−r,Sr),0≤r≤t,M_r=\exp\left(\int_0^r v(t-s,S_s)\,\mathrm{d}s\right)u(t-r,S_r),\qquad0\le r\le t,

and let σn\sigma_n be the exit times from increasing finite subsets exhausting EE. The jump-process product rule shows that Mr∧σnM_{r\wedge\sigma_n} is a martingale, since its drift before σn\sigma_n is proportional to −∂tu+Gu+vu=0-\partial_tu+Gu+vu=0. The finite jump rates and boundedness of uu justify the generator sums. Moreover,

∣Mr∧σn∣≤eCT∥u∥∞.|M_{r\wedge\sigma_n}|\le e^{CT}\lVert u\rVert_\infty.

Nonexplosion gives σn→∞\sigma_n \to\infty almost surely, so bounded convergence yields EkMt=M0\mathbb{E}_k M_t = M_0, proving (14). See also Fitzsimmons and Pitman (1999) for general Markov-process formulations.

Lemma 10 (Hardy’s inequality in difference form). For z1,…,zn∈Cz_1,\ldots,z_n \in\mathbb{C}, with z0=zn+1=0z_0=z_{n+1}=0,

∑k=1n∣zk∣2k2≤4∑k=0n∣zk+1−zk∣2.(15)\sum_{k=1}^{n}\frac{|z_k|^2}{k^2}\leq4\sum_{k=0}^{n}|z_{k+1}-z_k|^2. \tag*{(15)}

The constant is independent of nn.

This is the discrete Hardy inequality in difference form; see Hardy, Littlewood and Pólya (1952, Chapter IX, Theorem 326).

Lemma 11 (Expectations of random analytic functions). Let D⊂CD \subset\mathbb{C} be a domain, and let F(z)F(z), z∈Dz \in D, be a jointly measurable random function whose sample functions are almost surely holomorphic on DD. Suppose that, for every compact K⊂DK \subset D,

sup⁡z∈KE∣F(z)∣<∞.\sup_{z\in K}\mathbb{E}|F(z)|<\infty.

Then f(z)=E[F(z)]f(z)=\mathbb{E}[F(z)] is holomorphic on DD.

Proof. Choose a closed disk {∣z−b∣≤r}⊂D\{|z-b|\leq r\}\subset D. For ∣z−b∣<r|z-b|<r, Cauchy’s integral formula and Fubini’s theorem give

f(z)=E[12πi∮∣w−b∣=rF(w)w−z dw]=12πi∮∣w−b∣=rf(w)w−z dw.f(z)=\mathbb{E}\left[\frac{1}{2\pi i}\oint_{|w-b|=r}\frac{F(w)}{w-z}\,dw\right]=\frac{1}{2\pi i}\oint_{|w-b|=r}\frac{f(w)}{w-z}\,dw.

The bound on E∣F(w)∣\mathbb{E}|F(w)| along the circle justifies Fubini’s theorem. On every smaller disk the denominator stays bounded away from zero, so the last integral can be differentiated with respect to zz. Thus ff is holomorphic inside the disk. Such disks cover DD.

Lemma 12 (Vitali’s theorem). Let fnf_n be holomorphic on a connected open set D⊂CD \subset\mathbb{C} and locally bounded there, uniformly in nn. If fnf_n converges on a set with an accumulation point in DD, then it converges locally uniformly on DD to a holomorphic function.

See Conway (1978, Chapter VII).

Lemma 13 (Two-constants theorem on a slit domain). Let DD be a bounded smooth domain and let JJ be a nondegenerate closed line segment in DD such that D∖JD \setminus J is connected. Let ω(z)\omega(z) be the harmonic measure of JJ in D∖JD \setminus J. If ff is holomorphic on a neighborhood of D‾\overline{D}, and ∣f∣≤b0|f| \leq b_0 on JJ, ∣f∣≤b1|f| \leq b_1 on ∂D\partial D, where b0,b1>0b_0,b_1>0, then

∣f(z)∣≤b0ω(z)b11−ω(z),z∈D∖J.(16)|f(z)|\leq b_0^{\omega(z)}b_1^{1-\omega(z)},\qquad z\in D\setminus J. \tag*{(16)}

Moreover ω\omega is positive in D∖JD \setminus J and extends continuously as 11 on JJ. With this extension, inf⁡Kω>0\inf_K \omega> 0 for every compact K⊂DK \subset D.

This is the two-constants theorem in its harmonic-measure form, applied to log⁡∣f∣\log|f|; see [21 Sections 4.2–4.3]. The smooth outer boundary and the line-segment slit are regular for the Dirichlet problem, so the stated boundary continuity and positivity of ω\omega apply.

Fourier moments and the random walk

For real β,λ>0\beta, \lambda> 0 and t≤0t \le0, write qk(β,λ,t)=Eeikαλ,β(t)q_k(\beta,\lambda,t) = \mathbb{E}e^{ik\alpha_{\lambda,\beta}(t)}. Time translation of the phase SDE (1) gives

qk(β,λ,t)=qk(β,λeβt/4).(17)q_k(\beta,\lambda,t)=q_k(\beta,\lambda e^{\beta t/4}). \tag*{(17)}

Itô’s formula applied to the angular exponential gives

∂tqn(β,λ,t)=n22(qn+1+qn−1−2qn)(β,λ,t)+inλβ4eβt/4qn(β,λ,t).(18)\partial_t q_n(\beta,\lambda,t)=\frac{n^2}{2}\left(q_{n+1}+q_{n-1}-2q_n\right)(\beta,\lambda,t)+in\lambda\frac{\beta}{4}e^{\beta t/4}q_n(\beta,\lambda,t). \tag*{(18)}

The noise integrand is bounded, so its stochastic integral has mean zero on each finite time interval. The right-hand side is continuous by bounded convergence. By (17), equation (18) gives differentiability in λ>0\lambda> 0 and (5). The same scaling identity and the entrance condition imply qk(β,λ)→1q_k(\beta,\lambda) \to1 as λ↓0\lambda\downarrow0, by bounded convergence.

Proof of Proposition 4. The embedded simple random walk hits zero after finitely many jumps almost surely ([4 Chapter 4]). Thus SS is nonexplosive and AβA_\beta is almost surely finite. For T>0T > 0, applying Lemma 9 on [−T,0][-T,0] and using (17) gives

qn(β,λ)=En[exp⁡(iλβ4∫0Te−βt/4St dt)qST(β,λe−βT/4)].(19)q_n(\beta,\lambda)=\mathbb{E}_n\left[\exp\left(i\lambda\frac{\beta}{4}\int_0^T e^{-\beta t/4}S_t\,dt\right)q_{S_T}(\beta,\lambda e^{-\beta T/4})\right]. \tag*{(19)}

For T≥τ0T \ge\tau_0, the terminal factor is q0=1q_0=1 and the exponential is eiλAβe^{i\lambda A_\beta}. Since τ0<∞\tau_0 < \infty almost surely and the integrand has modulus at most one, bounded convergence as T→∞T \to\infty gives (6).

The same Feynman–Kac argument gives uniqueness for the moment system among uniformly bounded solutions.

Corollary 14. (Uniqueness of bounded solutions). For every real β>0\beta> 0, the system (5), with q0≡1q_0 \equiv1, has a unique solution (qk)k∈Z(q_k)_{k\in\mathbb{Z}} of continuously differentiable functions on (0,∞)(0,\infty) satisfying

sup⁡k∈Z, λ>0∣qk(λ)∣<∞.\sup_{k\in\mathbb{Z},\,\lambda>0}|q_k(\lambda)|<\infty.

Proof. The phase Fourier coefficients provide such a solution. Let rk=qk−q~kr_k=q_k-\widetilde{q}_k be the difference of two solutions, so that r0=0r_0=0 and ∣rk(λ)∣≤c\lvert r_k(\lambda)\rvert\le c for some finite cc. For fixed λ,T>0\lambda,T>0, the functions t↦rk(λeβt/4)t\mapsto r_k(\lambda e^{\beta t/4}) satisfy (18) on [−T,0][-T,0] by (5). Lemma 9 therefore gives

rk(λ)=Ek[exp⁡(iλβ4∫0Te−βt/4St dt)rST(λe−βT/4)],k≥1.r_k(\lambda)=\mathbb{E}_k\left[\exp\left(i\lambda\frac{\beta}{4}\int_0^T e^{-\beta t/4}S_t\,dt\right)r_{S_T}\left(\lambda e^{-\beta T/4}\right)\right],\qquad k\ge1.

Since r0=0r_0=0 and τ0<∞\tau_0<\infty almost surely,

∣rk(λ)∣≤cPk(τ0>T)⟶0as T→∞.\lvert r_k(\lambda)\rvert\le c\mathbb{P}_k(\tau_0>T)\longrightarrow0\qquad\text{as }T\to\infty.

Applying the same argument to r‾−k\overline{r}_{-k} proves uniqueness for negative indices.

Real Fourier decay from the walk

We use the representation (6) to prove decay of the phase moments on the real parameter intervals needed for analytic continuation.

Proposition 15 (Real Fourier decay). Let J,K⊂(0,∞)J,K\subset(0,\infty) be compact sets for λ\lambda and β\beta, respectively. There are a,b>0a,b>0, depending only on J,KJ,K, such that

sup⁡λ∈J,β∈K∣qk(β,λ)∣≤be−a∣k∣1/3,k∈Z.(20)\sup_{\lambda\in J,\beta\in K}\lvert q_k(\beta,\lambda)\rvert\le be^{-a\lvert k\rvert^{1/3}},\qquad k\in\mathbb{Z}. \tag*{(20)}

Consider the walk SS from Proposition 4. At height kk, the walk fluctuates by order khk\sqrt{h} over a short time hh, producing area fluctuations of order kh3/2kh^{3/2}. Taking h=k−2/3h=k^{-2/3} makes these fluctuations of order one. This motivates the study of the process on this time scale.

Lemma 16 (Short-time path limit). Put h=k−2/3h=k^{-2/3}. If k/2≤j≤2kk/2\le j\le2k and j/k→xj/k\to x as k→∞k\to\infty, then under Pj\mathbb{P}_j,

(k−2/3(Shu−j):0≤u≤1)⟹(xBu:0≤u≤1),\left(k^{-2/3}(S_{hu}-j):0\le u\le1\right)\Longrightarrow\left(xB_u:0\le u\le1\right),

where BB is standard Brownian motion and convergence is in D([0,1])D([0,1]) with its Skorokhod topology.

Proof. Let YY be a simple symmetric continuous-time walk started at zero, with rate 1/21/2 to each neighbor. Changing its holding times realizes the walk started at jj as

St=j+Y∫0tSs2 dsS_t=j+Y_{\int_0^t S_s^2\,ds}

see Ethier and Kurtz [6] for the general random time-change construction. Set

Wk(v)=k−2/3Yk4/3v,ck(u)=∫0u(Shvk)2 dv.W_k(v)=k^{-2/3}Y_{k^{4/3}v},\qquad c_k(u)=\int_0^u\left(\frac{S_{hv}}{k}\right)^2\,dv.

Donsker’s theorem [4 Theorem 8.1.4], together with the law of large numbers for the Poisson jump count, gives Wk⇒BW_k \Rightarrow B on every bounded time interval. Let ζk\zeta_k be the exit time of SS from [k/4,3k][k/4,3k]. Before exit, Ss/k≤3S_s/k \le3, so the rescaled clock runs at rate at most 9 and

sup⁡u≤1∣Shu∧ζk−j∣k≤k−1/3sup⁡v≤9∣Wk(v)∣→P0.\sup_{u\le1}\frac{\left|S_{hu\wedge\zeta_k}-j\right|}{k}\le k^{-1/3}\sup_{v\le9}\left|W_k(v)\right|\xrightarrow{\mathbb{P}}0.

Since exit requires displacement at least k/4k/4, Pj(ζk≤h)→0\mathbb{P}_j(\zeta_k\le h)\to0. Thus Shu/k→xS_{hu}/k\to x uniformly in probability on [0,1][0,1], and consequently ck(u)→x2uc_k(u)\to x^2u uniformly in probability. The clock limit is deterministic, so joint convergence and continuity of composition at continuous paths give

k−2/3(Sh−j)=Wk∘ck⇒Bx2=dxB.k^{-2/3}(S_h-j)=W_k\circ c_k\Rightarrow B_{x^2}\mathrel{\overset{d}{=}}xB.

Lemma 17 (Area before hitting zero). Let Dn⇒σBD_n\Rightarrow\sigma B in D([0,1])D([0,1]) for some σ>0\sigma>0. Suppose that almost surely DnD_n cannot change sign without visiting zero. Let ξn\xi_n be its first zero in [1/2,1][1/2,1], or 1 if there is none. For every nonempty open interval I⊂RI\subset\mathbb{R},

lim inf⁡n→∞P(ξn<1, ∫0ξnDn(u) du∈I)>0.\liminf_{n\to\infty}\mathbb{P}\left(\xi_n<1,\ \int_0^{\xi_n}D_n(u)\,du\in I\right)>0.

Proof. Choose aa and r>0r>0 with [a−r,a+r]⊂I[a-r,a+r]\subset I. On [0,1/2][0,1/2] let dd be the triangular path from zero to 4a4a at time 1/41/4 and back to zero; its integral is aa. On [1/2,1][1/2,1] let dd run linearly from zero to rr at time 3/43/4 and then to −r-r at time 1.

On the tube ∥Dn−d∥∞<r/2\lVert D_n-d\rVert_\infty<r/2, the values at 3/43/4 and 1 have opposite signs, so ξn<1\xi_n<1. Since ∣d∣≤r\lvert d\rvert\le r on the second half,

∣∫0ξnDn(u) du−a∣≤∥Dn−d∥∞+∫1/21∣d(u)∣ du<r.\left|\int_0^{\xi_n}D_n(u)\,du-a\right|\le\lVert D_n-d\rVert_\infty+\int_{1/2}^1\lvert d(u)\rvert\,du<r.

The Brownian limit has positive probability of lying in this tube: dd is polygonal and starts at zero, so this follows from Cameron–Martin and the positive probability of a ball about zero. The map f↦∥f−d∥∞f\mapsto\lVert f-d\rVert_\infty is continuous at continuous paths in the Skorokhod topology. The continuous mapping theorem and the open-set bound thus give the same positive lower bound for the liminf of the tube probabilities.

If the walk SS starts near a large kk, then over time h=k−2/3h=k^{-2/3} its area fluctuations are of order one. The next lemma turns these fluctuations into a contraction of the weighted transition operator in the supremum norm, by coupling paths with the same endpoint but different accumulated phases.

Lemma 18 (Uniform short-time contraction). Fix $0<b_0\le b_1<\infty$ and ℓ0≥0\ell_0\ge0. Put h=k−2/3h=k^{-2/3}, Ik=[k/2,2k]∩ZI_k=[k/2,2k]\cap\mathbb{Z}, and σk=inf⁡{t≥0:St∉Ik}\sigma_k=\inf\{t\ge0:S_t\notin I_k\}. For a real function bb on $[0,h]

satisfying b0≤b(t)≤b1b_0 \le b(t) \le b_1 and ∣b(t)−b(r)∣≤ℓ0∣t−r∣|b(t)-b(r)| \le\ell_0|t-r|, define

(Tkbf)(j)=Ej[ei∫0hb(t)St dtf(Sh)1{σk>h}],j∈Ik.(T_k^b f)(j)=\mathbb{E}_j\left[e^{\mathrm{i}\int_0^h b(t)S_t\,\mathrm{d}t}f(S_h)1_{\{\sigma_k>h\}}\right],\qquad j\in I_k.

There are δ>0\delta>0 and k0k_0, depending only on b0,b1,ℓ0b_0,b_1,\ell_0, such that ∥Tkb∥∞→∞≤1−δ\|T_k^b\|_{\infty\to\infty}\le1-\delta for k≥k0k\ge k_0.

Proof. Start two independent walks S1,t,S2,tS_{1,t},S_{2,t} at j∈Ikj\in I_k. Form coupled walks S^1,t,S^2,t\widehat S_{1,t},\widehat S_{2,t} by letting the second follow the first after their first meeting at or after h/2h/2. On the independent paths put

Dk(u)=b(hu)k−2/3(S1,hu−S2,hu),0≤u≤1,D_k(u)=b(hu)k^{-2/3}(S_{1,hu}-S_{2,hu}),\qquad0\le u\le1,

and let ζ~k\widetilde{\zeta}_k be its first zero in [1/2,1][1/2,1], or 11 if none exists. The coupled phase difference is ∫0ζ~kDk(u) du\int_0^{\widetilde{\zeta}_k}D_k(u)\,\mathrm{d}u. Uniformly over jj and bb, there is p>0p>0 such that

P(ζ~k<1,2π/3<∫0ζ~kDk(u) du<4π/3)≥p\mathbb{P}\left(\widetilde{\zeta}_k<1,\quad2\pi/3<\int_0^{\widetilde{\zeta}_k}D_k(u)\,\mathrm{d}u<4\pi/3\right)\ge p

for all sufficiently large kk. Indeed, otherwise take a sequence with k→∞k\to\infty and success probabilities tending to zero. Pass to a subsequence with j/k→x∈[1/2,2]j/k\to x\in[1/2,2] and b(0)→a∈[b0,b1]b(0)\to a\in[b_0,b_1]. Since sup⁡u∣b(hu)−b(0)∣≤ℓ0h\sup_u|b(hu)-b(0)|\le\ell_0h, Lemma 16 and independence give Dk⇒a2xBD_k\Rightarrow a\sqrt{2}xB. The difference cannot change sign without visiting zero, since the independent walks almost surely never jump simultaneously and bb is positive. Lemma 17 contradicts the choice of sequence.

Let GG denote the event in the preceding display, so that P(G)≥p\mathbb{P}(G)\ge p, and let σ^ℓ\widehat{\sigma}_\ell be the exit time of S^ℓ\widehat S_\ell from IkI_k. For ∥f∥∞≤1\|f\|_\infty\le1, extending ff by zero outside IkI_k, set

Vℓ=exp⁡(i∫0hb(t)S^ℓ,t dt)f(S^ℓ,h)1{σ^ℓ>h},ℓ=1,2.V_\ell=\exp\left(\mathrm{i}\int_0^h b(t)\widehat S_{\ell,t}\,\mathrm{d}t\right)f(\widehat S_{\ell,h})1_{\{\widehat{\sigma}_\ell>h\}},\qquad\ell=1,2.

The marginal laws give EV1=EV2=(Tkbf)(j)\mathbb{E}V_1=\mathbb{E}V_2=(T_k^b f)(j), and ∣Vℓ∣≤1|V_\ell|\le1. On GG, the walks agree from time hζ~kh\widetilde{\zeta}_k onward. If both survive until hh, they have the same endpoint and phase difference

Θ=∫0ζ~kDk(u) du∈(2π/3,4π/3).\Theta=\int_0^{\widetilde{\zeta}_k}D_k(u)\,\mathrm{d}u\in(2\pi/3,4\pi/3).

Hence

∣V1+V2∣=∣f(S^1,h)∣∣1+eiΘ∣≤2∣cos⁡(Θ/2)∣≤1.|V_1+V_2|=|f(\widehat S_{1,h})||1+e^{\mathrm{i}\Theta}|\le2|\cos(\Theta/2)|\le1.

If either walk is killed, at most one contribution is nonzero, so again ∣V1+V2∣≤1|V_1+V_2|\le1. Thus this bound holds throughout GG, while ∣V1+V2∣≤2|V_1+V_2|\le2 always. Consequently,

∣(Tkbf)(j)∣≤12E∣V1+V2∣≤12P(G)+P(Gc)≤1−p2.|(T_k^b f)(j)|\le\frac12\mathbb{E}|V_1+V_2|\le\frac12\mathbb{P}(G)+\mathbb{P}(G^c)\le1-\frac p2.

Taking the supremum over j∈Ikj\in I_k and ∥f∥∞≤1\|f\|_\infty\le1 proves the assertion with δ=p/2\delta=p/2.

Lemma 19 (Exponential exit bound). For the walk started at kk, every T>0T>0, and 0<θ≤10<\theta\le1,

Pk(σk≤T)≤2exp⁡(−θk2+4k2Tθ2).(21)\mathbb{P}_k(\sigma_k\le T)\le2\exp\left(-\frac{\theta k}{2}+4k^2T\theta^2\right). \tag*{(21)}

Proof. Put f(i)=cosh⁡(θ(i−k))f(i)=\cosh(\theta(i-k)). For i∈Iki\in I_k, the generator satisfies

(Gf)(i)=i2(cosh⁡θ−1)f(i)≤4k2θ2f(i),(Gf)(i)=i^2(\cosh\theta-1)f(i)\le4k^2\theta^2f(i),

since cosh⁡θ−1≤θ2\cosh\theta-1\le\theta^2 for 0≤θ≤10\le\theta\le1. Dynkin’s formula stopped at σk\sigma_k and Gronwall’s inequality give

Ekf(ST∧σk)≤e4k2θ2T.\mathbb{E}_k f(S_{T\wedge\sigma_k})\le e^{4k^2\theta^2T}.

On {σk≤T}\{\sigma_k\le T\}, the displacement at exit is at least k/2k/2, so f(ST∧σk)≥cosh⁡(θk/2)f(S_{T\wedge\sigma_k})\ge\cosh(\theta k/2). Therefore

Pk(σk≤T)≤e4k2θ2Tcosh⁡(θk/2)≤2exp⁡(−θk2+4k2Tθ2).\mathbb{P}_k(\sigma_k\le T)\le\frac{e^{4k^2\theta^2T}}{\cosh(\theta k/2)}\le2\exp\left(-\frac{\theta k}{2}+4k^2T\theta^2\right).

Proof of Proposition 15. Since q0=1q_0=1 and q−k=qk‾q_{-k}=\overline{q_k} for real β\beta, it suffices to take k≥1k\ge1. Fix λ∈J\lambda\in J and β∈K\beta\in K, and put

h=k−2/3,m=⌊k1/3⌋,T=mh≤k−1/3.h=k^{-2/3},\qquad m=\lfloor k^{1/3}\rfloor,\qquad T=mh\le k^{-1/3}.

Set b(t)=λ(β/4)e−βt/4b(t)=\lambda(\beta/4)e^{-\beta t/4} and f(j)=qj(β,λe−βT/4)f(j)=q_j(\beta,\lambda e^{-\beta T/4}). On [0,1][0,1], these bb have common positive lower and upper bounds and a common Lipschitz constant, uniformly over λ∈J\lambda\in J and β∈K\beta\in K. Also ∥f∥∞≤1\|f\|_\infty\le1. By (19),

qk(β,λ)=Ek[ei∫0Tb(t)St dtf(ST)].q_k(\beta,\lambda)=\mathbb{E}_k\left[e^{i\int_0^T b(t)S_t\,\mathrm{d}t}f(S_T)\right].

For r=0,…,m−1r=0,\ldots,m-1, define b[r](t)=b(rh+t)b^{[r]}(t)=b(rh+t) on [0,h][0,h]. Each b[r]b^{[r]} satisfies the same bounds in Lemma 18. Restricting to paths staying in IkI_k until TT, the Markov property gives

Ek[ei∫0Tb(t)St dtf(ST)1{σk>T}]=(Tkb[0]Tkb[1]⋯Tkb[m−1]f)(k).\mathbb{E}_k\left[e^{i\int_0^T b(t)S_t\,\mathrm{d}t}f(S_T)\mathbf{1}_{\{\sigma_k>T\}}\right] =\left(T_k^{b^{[0]}}T_k^{b^{[1]}}\cdots T_k^{b^{[m-1]}}f\right)(k).

For k≥k0k\ge k_0, where k0k_0 depends only on J,KJ,K, Lemma 18 therefore yields

∣qk(β,λ)∣≤(1−δ)m+Pk(σk≤T).|q_k(\beta,\lambda)|\le(1-\delta)^m+\mathbb{P}_k(\sigma_k\le T).

Apply Lemma 19 with θ=εk−2/3\theta=\varepsilon k^{-2/3}, where 0<ε<1/80<\varepsilon<1/8 is fixed. Since T≤k−1/3T\le k^{-1/3},

Pk(σk≤T)≤2exp⁡[−(ε/2−4ε2)k1/3].\mathbb{P}_k(\sigma_k\le T)\le2\exp\left[-\left(\varepsilon/2-4\varepsilon^2\right)k^{1/3}\right].

Also (1−δ)m≤c1e−c2k1/3(1-\delta)^m \le c_1 e^{-c_2 k^{1/3}}. Enlarging the prefactor to cover the finitely many remaining kk proves (20).

Corollary 20 (Smoothness of the phase density). For β,λ>0\beta,\lambda>0, the law of αλ,β(0)\alpha_{\lambda,\beta}(0) mod 2π2\pi has a smooth density pβ,λ(θ)p_{\beta,\lambda}(\theta). Its angular derivatives of every order are jointly continuous in (β,λ,θ)(\beta,\lambda,\theta) and locally uniformly bounded.

Proof. By Proposition 15, the Fourier series

pβ,λ(θ)=12π∑k∈Zqk(β,λ)e−ikθp_{\beta,\lambda}(\theta)=\frac{1}{2\pi}\sum_{k\in\mathbb{Z}}q_k(\beta,\lambda)e^{-ik\theta}

and all its termwise angular derivatives converge absolutely and locally uniformly. By Fourier inversion [15 Chapter I, §7.1, Corollary, and §7.4], this defines a smooth density of the phase law.

The walk in (6) has a law independent of β\beta and λ\lambda. Each path is absorbed after finitely many jumps, so AβA_\beta is almost surely continuous in β>0\beta>0. Thus, if (βn,λn)→(β,λ)(\beta_n,\lambda_n)\to(\beta,\lambda) in (0,∞)2(0,\infty)^2, then

eiλnAβn⟶eiλAβalmost surely,∣eiλnAβn∣=1.e^{i\lambda_n A_{\beta_n}}\longrightarrow e^{i\lambda A_\beta}\qquad\text{almost surely},\qquad\left|e^{i\lambda_n A_{\beta_n}}\right|=1.

Bounded convergence in (6) gives joint continuity of each qkq_k in (β,λ)(\beta,\lambda). Locally uniform convergence of every termwise angular derivative series gives the claimed joint continuity and local boundedness.

From boundary angles to correlation measures

Proof of Proposition 3. Write αλ=αλ,β(0)\alpha_\lambda=\alpha_{\lambda,\beta}(0). For a boundary angle uu, put

μu=∑λ: αλ=u mod 2πδλ∂λαλ.\mu_u=\sum_{\lambda:\,\alpha_\lambda=u\ \mathrm{mod}\ 2\pi}\frac{\delta_\lambda}{\partial_\lambda\alpha_\lambda}.

For any continuous compactly supported gg on R\mathbb{R} and bounded measurable hh on T\mathbb{T}, changing variables u=αλ mod 2πu=\alpha_\lambda\ \mathrm{mod}\ 2\pi on each inverse branch gives

∫Th(u)∫Rg(λ) μu(dλ) du=∫Rg(λ)h(αλ) dλ,\int_{\mathbb{T}}h(u)\int_{\mathbb{R}}g(\lambda)\,\mu_u(d\lambda)\,du=\int_{\mathbb{R}}g(\lambda)h(\alpha_\lambda)\,d\lambda,

since du=∂λαλ dλdu=\partial_\lambda\alpha_\lambda\,d\lambda cancels the spectral weight. Thus, as measures on R\mathbb{R},

∫Th(u) μu(dλ) du=h(αλ) dλ.\int_{\mathbb{T}}h(u)\,\mu_u(d\lambda)\,du=h(\alpha_\lambda)\,d\lambda.

Let U0=UU_0=U be the uniform angle in (2), and put Uj=U+εVj mod 2πU_j=U+\varepsilon V_j\ \mathrm{mod}\ 2\pi, where the VjV_j are independent uniform variables on [−1,1][-1,1], independent of UU and the phases. For 0<ε<π0<\varepsilon<\pi the angles have joint density

12π(2ε)k−1on{max⁡1≤j≤k−1dist⁡(uj−u0,2πZ)<ε}\frac{1}{2\pi(2\varepsilon)^{k-1}}\quad\text{on}\quad\left\{\max_{1\le j\le k-1}\operatorname{dist}(u_j-u_0,2\pi\mathbb{Z})<\varepsilon\right\}

and zero elsewhere. By the substitution above, E∏j=0k−1μUj(Ij)\mathbb{E}\prod_{j=0}^{k-1}\mu_{U_j}(I_j) is therefore the integral on the right of (4), including its prefactor, with αλj−αλ0\alpha_{\lambda_j}-\alpha_{\lambda_0} in place of αλj−λ0\alpha_{\lambda_j-\lambda_0}.

The two vectors have the same joint law: subtract the equation for λ0\lambda_0 in (1) from the equations for λj\lambda_j. The differences solve the phase SDE (1) at λj−λ0\lambda_j-\lambda_0, all driven by the same rotated Brownian motion dZ~t=e−iαλ0,β(t)dZtd\widetilde{Z}_t=e^{-i\alpha_{\lambda_0,\beta}(t)}dZ_t.

Almost surely no endpoint of the IjI_j has angle UU, since UU is independent and uniform. Only finitely many inverse branches meet each interval near UU, and their locations and weights vary continuously. Thus μUj(Ij)→μU(Ij)\mu_{U_j}(I_j)\to\mu_U(I_j) almost surely. Each UjU_j is itself uniform and independent of the phases. Given nn points in IjI_j, the rrth moment of μU(Ij)\mu_U(I_j) is at most nrEwrn^r\mathbb{E}w^r for r≥1r\ge1, by convexity, where ww is one normalized weight. Proposition 7 and Lemma 8 thus give moments of all orders; Hölder’s inequality bounds the products uniformly in L2L^2. We may therefore pass to expectations. At the common angle UU every tuple in the product consists of distinct points because the intervals are disjoint. Given the points, their weights are independent with mean one, so the limiting expectation is E∏j=0k−1Ξβ(Ij)\mathbb{E}\prod_{j=0}^{k-1}\Xi_\beta(I_j).

Proof of Theorem 2. Take k=2k=2 in Proposition 3, with sup⁡I0<inf⁡I1\sup I_0<\inf I_1. Writing λ=λ1−λ0\lambda=\lambda_1-\lambda_0, the quotient in (4) is

12ε∫−εεpβ,λ(θ) dθ⟶pβ,λ(0)\frac{1}{2\varepsilon}\int_{-\varepsilon}^{\varepsilon}p_{\beta,\lambda}(\theta)\,d\theta\longrightarrow p_{\beta,\lambda}(0)

locally uniformly in λ\lambda, by Corollary 20. Thus ρβ(0,λ)=pβ,λ(0)/(2π)\rho_\beta(0,\lambda)=p_{\beta,\lambda}(0)/(2\pi) gives a continuous version for λ>0\lambda>0. Translation invariance and symmetry give the density off the diagonal, and the Fourier series in Corollary 20 gives (3).

Remark 21. For Nλ=Ξβ((0,λ])N_\lambda=\Xi_\beta((0,\lambda]), we note in passing that ρβ(0,λ)=12∂λ2Var⁡(Nλ)+1/(4π2)\rho_\beta(0,\lambda)=\frac{1}{2}\partial_\lambda^2\operatorname{Var}(N_\lambda)+1/(4\pi^2), where

Var⁡(Nλ)=16Eν[1−cos⁡(λAβ)]+2βπ2E1[Cin⁡(λAβ)].\operatorname{Var}(N_\lambda)=\frac{1}{6}\mathbb{E}_\nu[1-\cos(\lambda A_\beta)]+\frac{2}{\beta\pi^2}\mathbb{E}_1[\operatorname{Cin}(\lambda A_\beta)].

Here Cin⁡(x)=∫0x(1−cos⁡t) dt/t\operatorname{Cin}(x)=\int_0^x(1-\cos t)\,dt/t, and ν(k)=6/(π2k2)\nu(k)=6/(\pi^2k^2), k≥1k\ge1, is the initial law of the walk SS in the first term. The formula follows by conditioning on the phase in (12), using Fourier expansion, and applying Itô’s formula to αλ2\alpha_\lambda^2.

Holomorphic Fourier coefficients

We now extend the Fourier moments to complex β\beta with Re⁡β>0\operatorname{Re}\beta>0 using the representations (6) and (7). We first construct a holomorphic function for each cutoff, and then remove the cutoff.

Holomorphy before removing the cutoff

Lemma 22 (Holomorphy and ODE for the cutoff moments). For 1≤k≤n1\le k\le n, λ>0\lambda>0, and Re⁡β>0\operatorname{Re}\beta>0, define qk,n(β,λ)q_{k,n}(\beta,\lambda) by (7), with AβA_\beta given by the integral in (6). This function is holomorphic in β\beta and continuously differentiable in λ\lambda. It satisfies

λ∂λqk,n=ikλqk,n+2k2β(qk+1,n+qk−1,n−2qk,n),q0,n=1,qn+1,n=0.(22)\lambda\partial_\lambda q_{k,n}=ik\lambda q_{k,n}+\frac{2k^2}{\beta}(q_{k+1,n}+q_{k-1,n}-2q_{k,n}),\qquad q_{0,n}=1,\qquad q_{n+1,n}=0. \tag*{(22)}

As λ↓0\lambda\downarrow0,

qk,n(β,λ)⟶1−kn+1.(23)q_{k,n}(\beta,\lambda)\longrightarrow1-\frac{k}{n+1}. \tag*{(23)}

Proof. Write En={τ0<τn+1}E_n=\{\tau_0<\tau_{n+1}\}. The walk and this event are independent of β\beta. On a retained path, τ0<∞\tau_0<\infty and Sv≤nS_v\le n before absorption. The integral

Aβ=β4∫0∞e−βv/4Sv dvA_\beta=\frac{\beta}{4}\int_0^\infty e^{-\beta v/4}S_v\,\mathrm{d}v

is supported on the finite interval [0,τ0][0,\tau_0] and is entire in β\beta. This follows from Lemma 11 applied to normalized Lebesgue measure on [0,τ0][0,\tau_0]. On the right half-plane it satisfies the path-independent bound

∣Aβ∣≤n∣β∣4∫0∞e−(Re⁡β)v/4 dv=n∣β∣Re⁡β.(24)|A_\beta|\le\frac{n|\beta|}{4}\int_0^\infty e^{-(\operatorname{Re}\beta)v/4}\,\mathrm{d}v=\frac{n|\beta|}{\operatorname{Re}\beta}. \tag*{(24)}

Thus

∣eiλAβ1En∣≤exp⁡(nλ∣β∣Re⁡β).\left|e^{i\lambda A_\beta}\mathbf{1}_{E_n}\right|\le\exp\left(\frac{n\lambda|\beta|}{\operatorname{Re}\beta}\right).

Lemma 11 gives holomorphy of qk,nq_{k,n} in β\beta, and dominated convergence gives continuity in λ\lambda. The same bound and gambler’s ruin give (23).

For real β>0\beta>0, let S^\widehat{S} be the chain on {0,…,n+1}\{0,\ldots,n+1\} obtained by absorbing SS at both endpoints: S^v=Sv∧τ0∧τn+1\widehat{S}_v=S_{v\wedge\tau_0\wedge\tau_{n+1}}. Its generator Gn\mathcal{G}_n has zero rows at 00 and n+1n+1. For T>0T>0, define the finite-start cutoffs q−T,k(n)(t)q_{-T,k}^{(n)}(t) on [−T,0][-T,0] by the finite system

∂tq−T,k(n)(t)=(Gnq−T(n)(t,⋅))(k)+ikλβ4eβt/4q−T,k(n)(t),q−T,k(n)(−T)=1{k≤n}.\partial_t q_{-T,k}^{(n)}(t)=\bigl(\mathcal{G}_n q_{-T}^{(n)}(t,\cdot)\bigr)(k)+ik\lambda\frac{\beta}{4}e^{\beta t/4}q_{-T,k}^{(n)}(t),\qquad q_{-T,k}^{(n)}(-T)=\mathbf{1}_{\{k\le n\}}.

This finite system has a unique solution, which is bounded on [−T,0][-T,0], with q−T,0(n)≡1q_{-T,0}^{(n)}\equiv1 and q−T,n+1(n)≡0q_{-T,n+1}^{(n)}\equiv0; the equations hold at the endpoints as well as in the interior. Lemma 9, with the time origin shifted to −T-T, gives

q−T,k(n)(t)=Ek[exp⁡(iλβ4eβt/4∫0t+Te−βv/4S~v dv)1{S~t+T≠n+1}].q_{-T,k}^{(n)}(t)=\mathbb{E}_k\left[\exp\left(\mathrm{i}\lambda\frac{\beta}{4}e^{\beta t/4}\int_0^{t+T}e^{-\beta v/4}\widetilde{S}_v\,\mathrm{d}v\right)\mathbf{1}_{\{\widetilde{S}_{t+T}\ne n+1\}}\right].

As T→∞T\to\infty, this converges to

qk,n(β,λeβt/4).q_{k,n}\left(\beta,\lambda e^{\beta t/4}\right).

Indeed, the integrands agree once an endpoint has been reached, so the difference of their expectations is bounded in modulus by 2Pk(τ0∧τn+1>t+T)2P_k(\tau_0\wedge\tau_{n+1}>t+T). Convergence is therefore uniform on compact time intervals. Passing to the integral form of the finite ODE therefore shows that t↦qk,n(β,λeβt/4)t\mapsto q_{k,n}\left(\beta,\lambda e^{\beta t/4}\right) satisfies the same time-dependent system. The change of variables s=λeβt/4s=\lambda e^{\beta t/4} gives (22) for real β>0\beta>0.

To extend the equation, fix 0<a<b0<a<b and write its integral form:

qk,n(β,b)−qk,n(β,a)=∫ab[ikqk,n(β,s)+2k2βs(qk+1,n(β,s)+qk−1,n(β,s)−2qk,n(β,s))]ds.(25)\begin{aligned} q_{k,n}(\beta,b)-q_{k,n}(\beta,a) &=\int_a^b\left[\mathrm{i}kq_{k,n}(\beta,s)\right.\\ &\quad\left.+\frac{2k^2}{\beta s}\left(q_{k+1,n}(\beta,s)+q_{k-1,n}(\beta,s)-2q_{k,n}(\beta,s)\right)\right]\mathrm{d}s. \tag*{(25)} \end{aligned}

Both sides are holomorphic in β\beta; the displayed path bound is uniform for s∈[a,b]s\in[a,b] on compact subsets of the half-plane and justifies the integral by Lemma 11. The identity theorem extends (25) to Re⁡β>0\operatorname{Re}\beta>0. Its integrand is continuous in ss, so the fundamental theorem of calculus now proves differentiability and (22).

A uniform energy estimate

The pathwise bound used above to prove holomorphy grows exponentially with nn. To remove the cutoff, we need a different estimate that is uniform in nn; this is the purpose of the following energy argument. A related weighted energy argument is used for real-parameter uniqueness in Qu and Valkó (2025, Proposition 16); here the estimate must be uniform in the cutoff for complex β\beta.

For z=(z1,…,zn)z=(z_1,\ldots,z_n) put

∥z∥n2=∑k=1n∣zk∣2k2,dn(z)=∣z1∣2+∑k=1n−1∣zk+1−zk∣2+∣zn∣2.\lVert z\rVert_n^2=\sum_{k=1}^{n}\frac{\lvert z_k\rvert^2}{k^2},\qquad d_n(z)=\lvert z_1\rvert^2+\sum_{k=1}^{n-1}\lvert z_{k+1}-z_k\rvert^2+\lvert z_n\rvert^2.

The endpoint values in dnd_n are zero; the boundary value q0,n=1q_{0,n}=1 in the ODE will instead appear as a forcing term.

Lemma 23 (Energy bound). For every n≥1n \ge1, λ>0\lambda> 0, and Re⁡β>0\operatorname{Re}\beta> 0,

∑k=1n∣qk,n(β,λ)∣2k2≤4∣β∣2(Re⁡β)2.(26)\sum_{k=1}^{n}\frac{\lvert q_{k,n}(\beta,\lambda)\rvert^{2}}{k^{2}} \le4\frac{\lvert\beta\rvert^{2}}{(\operatorname{Re}\beta)^{2}}. \tag*{(26)}

Proof. Write qk=qk,n(β,λ)q_k=q_{k,n}(\beta,\lambda), q=(q1,…,qn)q=(q_1,\ldots,q_n), and h(λ)=∥q(λ)∥n2h(\lambda)=\lVert q(\lambda)\rVert_n^2. Multiplying (22) by qk‾/k2\overline{q_k}/k^2, summing, and taking real parts gives, before summation by parts,

λ2h′(λ)=Re⁡[iλ∑k=1n∣qk∣2k+2β∑k=1nqk‾(qk+1+qk−1−2qk)]\frac{\lambda}{2}h'(\lambda)=\operatorname{Re}\left[\mathrm{i}\lambda\sum_{k=1}^{n}\frac{\lvert q_k\rvert^2}{k}+\frac{2}{\beta}\sum_{k=1}^{n}\overline{q_k}(q_{k+1}+q_{k-1}-2q_k)\right]
=2Re⁡[β−1∑k=1nqk‾(qk+1+qk−1−2qk)].=2\operatorname{Re}\left[\beta^{-1}\sum_{k=1}^{n}\overline{q_k}(q_{k+1}+q_{k-1}-2q_k)\right].

The imaginary potential contributes zero: its displayed sum is purely imaginary. Summation by parts, with q0=1q_0=1 and qn+1=0q_{n+1}=0, now yields

λh′=−4a dn(q)+4Re⁡(β−1q1‾),a=Re⁡(β−1)>0.(27)\lambda h'=-4a\,d_n(q)+4\operatorname{Re}(\beta^{-1}\overline{q_1}),\qquad a=\operatorname{Re}(\beta^{-1})>0. \tag*{(27)}

Put b=∣β−1∣b=\lvert\beta^{-1}\rvert. Since ∣q1∣≤dn(q)\lvert q_1\rvert\le\sqrt{d_n(q)},

4Re⁡(β−1q1‾)≤4bdn(q)≤2a dn(q)+2b2a.4\operatorname{Re}(\beta^{-1}\overline{q_1})\le4b\sqrt{d_n(q)}\le2a\,d_n(q)+\frac{2b^2}{a}.

Hardy’s inequality (Lemma 10, equation (15)) gives dn(q)≥h/4d_n(q)\ge h/4, and hence

λh′≤−a2h+2b2a.(28)\lambda h'\le-\frac{a}{2}h+\frac{2b^2}{a}. \tag*{(28)}

By (23),

lim⁡v↓0h(v)=∑k=1n(1−k/(n+1))2k2≤∑k=1∞1k2=π26.\lim_{v\downarrow0}h(v)=\sum_{k=1}^{n}\frac{(1-k/(n+1))^2}{k^2}\le\sum_{k=1}^{\infty}\frac{1}{k^2}=\frac{\pi^2}{6}.

For 0<λ0<λ0<\lambda_0<\lambda, Gronwall’s inequality applied in the variable log⁡λ\log\lambda (equivalently, multiplication by the integrating factor λa/2\lambda^{a/2}) gives

h(λ)≤(λ0λ)a/2h(λ0)+4b2a2[1−(λ0λ)a/2].h(\lambda)\le\left(\frac{\lambda_0}{\lambda}\right)^{a/2}h(\lambda_0)+\frac{4b^2}{a^2}\left[1-\left(\frac{\lambda_0}{\lambda}\right)^{a/2}\right].

Let λ0↓0\lambda_0\downarrow0. The first term vanishes because h(λ0)h(\lambda_0) is bounded. Since b/a=∣β∣/Re⁡βb/a=\lvert\beta\rvert/\operatorname{Re}\beta, this proves (26).

Removing the cutoff

Proof of Proposition 5. For each fixed k≥1k \ge1, Lemma 23 gives

∣qk,n(β,λ)∣≤2k∣β∣Re⁡β,n≥k.\lvert q_{k,n}(\beta,\lambda)\rvert\le2k\frac{\lvert\beta\rvert}{\operatorname{Re}\beta}, \qquad n \ge k.

Thus the holomorphic cutoff transforms are locally bounded uniformly in nn. For real β>0\beta> 0, subtracting (7) from (6) and applying the gambler’s ruin formula to the embedded simple random walk ([4 Chapter 4]) gives

∣qk(β,λ)−qk,n(β,λ)∣=∣Ek[eiλAβ1{τn+1<τ0}]∣≤Pk(τn+1<τ0)=kn+1.\lvert q_k(\beta,\lambda)-q_{k,n}(\beta,\lambda)\rvert= \left\lvert\mathbb{E}_k\left[e^{i\lambda A_\beta}\mathbf{1}_{\{\tau_{n+1}<\tau_0\}}\right]\right\rvert\le\mathbb{P}_k(\tau_{n+1}<\tau_0)=\frac{k}{n+1}.

This proves (8) and hence qk,n(β,λ)→qk(β,λ)q_{k,n}(\beta,\lambda)\to q_k(\beta,\lambda). Vitali’s theorem, Lemma 12, supplies a locally uniform holomorphic limit on the whole half-plane. It is the unique extension of the real Fourier coefficient.

For a fixed integer mm, pass to the limit n→∞n\to\infty in the first mm terms of (26), and then let m→∞m\to\infty. We obtain

∑k=1∞∣qk(β,λ)∣2k2≤4∣β∣2(Re⁡β)2.(29)\sum_{k=1}^{\infty}\frac{\lvert q_k(\beta,\lambda)\rvert^2}{k^2}\le4\frac{\lvert\beta\rvert^2}{(\operatorname{Re}\beta)^2}. \tag*{(29)}

For k≥1k \ge1, define

q−k(β,λ)=qk(β‾,λ),q0(β,λ)=1.(30)q_{-k}(\beta,\lambda)=q_k(\overline{\beta},\lambda),\qquad q_0(\beta,\lambda)=1. \tag*{(30)}

These are holomorphic and agree with the negative real Fourier moments. Equation (29) proves (9). In particular,

∣qk(β,λ)∣≤2∣k∣∣β∣Re⁡β,k≠0,\lvert q_k(\beta,\lambda)\rvert\le2\lvert k\rvert\frac{\lvert\beta\rvert}{\operatorname{Re}\beta},\qquad k\ne0,

uniformly in λ>0\lambda>0.

Analytic transfer and completion of the proof

The complex-parameter energy bound and the real stretched-exponential estimate combine through the two-constants theorem as follows.

Lemma 24 (Transfer of stretched-exponential decay). Let fkf_k, k≥1k\ge1, be holomorphic on Re⁡β>0\operatorname{Re}\beta>0. Suppose that for some a≥0a\ge0 and every compact KK in this half-plane,

sup⁡β∈K∣fk(β)∣≤cKka.\sup_{\beta\in K}\lvert f_k(\beta)\rvert\le c_K k^a.

Suppose also that there is a nondegenerate compact interval J⊂(0,∞)J \subset(0,\infty) and constants b,c>0b,c>0 such that

sup⁡β∈J∣fk(β)∣≤ce−bk1/3.\sup_{\beta\in J}|f_k(\beta)|\le ce^{-bk^{1/3}}.

Then for every compact K⊂{Re⁡β>0}K\subset\{\operatorname{Re}\beta>0\} there are constants bK,cK′>0b_K,c'_K>0 such that

sup⁡β∈K∣fk(β)∣≤cK′e−bKk1/3.(31)\sup_{\beta\in K}|f_k(\beta)|\le c'_K e^{-b_Kk^{1/3}}. \tag*{(31)}

The conclusion is uniform over an additional index whenever the constants in the two hypotheses are uniform over that index.

Proof. Fix KK. Choose a bounded smooth convex domain DD with K∪J⊂DK\cup J\subset D and D‾⊂{Re⁡β>0}\overline{D}\subset\{\operatorname{Re}\beta>0\}. The slit domain D∖JD\setminus J is connected. Let ω\omega be the harmonic measure from Lemma 13, extended as 11 on JJ, and put ω∗=inf⁡Kω>0\omega_*=\inf_K\omega>0. The bounds on ∂D\partial D and JJ, followed by (16), give

∣fk(β)∣≤(ce−bk1/3)ω(β)(cDka)1−ω(β)≤c1kae−bω∗k1/3,β∈K.|f_k(\beta)|\le\left(ce^{-bk^{1/3}}\right)^{\omega(\beta)}\left(c_Dk^a\right)^{1-\omega(\beta)}\le c_1k^a e^{-b\omega_*k^{1/3}},\qquad\beta\in K.

The same bound holds on K∩JK\cap J directly. Absorbing the polynomial factor gives (31) with bK=bω∗/2b_K=b\omega_*/2. The proof preserves uniformity in an additional index.

Proof of Proposition 6. Fix a compact interval L⊂(0,∞)L\subset(0,\infty) and apply Lemma 24 with fk(β)=qk(β,λ)f_k(\beta)=q_k(\beta,\lambda) and a=1a=1. Proposition 5 supplies the complex bound uniformly in λ>0\lambda>0, and Proposition 15 supplies the real stretched-exponential bound uniformly for λ∈L\lambda\in L. This proves the conclusion for positive indices. For negative indices, use (30) and the positive-index estimate on {β‾:β∈K}\{\overline{\beta}:\beta\in K\}. Enlarging the prefactor covers k=0k=0.

Proof of Theorem 1. Fix λ>0\lambda>0. By Proposition 6,

14π2∑k∈Zqk(β,λ)(32)\frac{1}{4\pi^2}\sum_{k\in\mathbb{Z}}q_k(\beta,\lambda) \tag*{(32)}

converges absolutely and locally uniformly on Re⁡β>0\operatorname{Re}\beta>0. Every summand is holomorphic by Proposition 5, so the sum is holomorphic. For real β>0\beta>0, Theorem 2 identifies (32) with ρβ(0,λ)\rho_\beta(0,\lambda). Uniqueness of the holomorphic extension follows from the identity theorem.

To prove joint continuity, let λn→λ>0\lambda_n\to\lambda>0. By Proposition 6, the holomorphic functions β↦ρβ(0,λn)\beta\mapsto\rho_\beta(0,\lambda_n) are locally bounded uniformly in nn. For real β>0\beta>0, they converge to ρβ(0,λ)\rho_\beta(0,\lambda) by Theorem 2. Vitali’s theorem, Lemma 12, therefore gives this convergence locally uniformly on Re⁡β>0\operatorname{Re}\beta>0. Together with continuity in β\beta at fixed λ\lambda, this proves joint continuity.

Corollary 25 (Holomorphic extension of the phase density). For every fixed λ>0\lambda> 0, the Fourier series

pβ,λ(θ)=12π∑k∈Zqk(β,λ)e−ikθp_{\beta,\lambda}(\theta)=\frac{1}{2\pi}\sum_{k\in\mathbb{Z}}q_k(\beta,\lambda)e^{-ik\theta}

extends the real phase density to Re⁡β>0\operatorname{Re}\beta> 0. It is holomorphic in β\beta and smooth in θ\theta; the series and every angular derivative converge uniformly for β\beta in compact subsets of the half-plane and θ∈T\theta\in\mathbb{T}.

Proof. For every j≥0j\geq0, the bound in Proposition 6 remains summable after multiplication by ∣k∣j|k|^j. Thus the differentiated series converges locally uniformly, which also proves holomorphy.

Fourier moments and phase density at β=2\beta= 2

We provide an explicit solution to (5) at β=2\beta= 2. The solution shows that the stretched exponential decay exponent 1/31/3 in Proposition 6 cannot be increased beyond 1/21/2. It also yields an explicit phase density, with a simple probabilistic representation at integer multiples of 2π2\pi.

For a>0a>0, b∈Rb\in\mathbb{R}, and Re⁡z>0\operatorname{Re}z>0, Tricomi’s confluent hypergeometric function is defined by

U(a,b,z)=1Γ(a)∫0∞e−ztta−1(1+t)b−a−1 dt.(33)U(a,b,z)=\frac{1}{\Gamma(a)}\int_{0}^{\infty}e^{-zt}t^{a-1}(1+t)^{b-a-1}\,\mathrm{d}t. \tag*{(33)}

We use its analytic continuation to C∖(−∞,0]\mathbb{C}\setminus(-\infty,0], with complex powers on the principal branch, and set U(0,b,z)=1U(0,b,z)=1. See DLMF (2026, (13.4.4)).

Proposition 26 (Fourier moments at β=2\beta= 2). For k≥1k\geq1 and λ>0\lambda> 0,

qk(2,λ)=k![U(k,1,−iλ)−keiλU(k+1,1,−iλ)].(34)q_k(2,\lambda)=k!\left[U(k,1,-i\lambda)-ke^{i\lambda}U(k+1,1,-i\lambda)\right]. \tag*{(34)}

For fixed λ>0\lambda> 0, as k→∞k\to\infty,

∣qk(2,λ)∣∼{2πλ−1/4∣sin⁡(λ/2)∣k3/4e−2λk,λ∉2πZ,πλ1/4k1/4e−2λk,λ∈2πZ.(35)|q_k(2,\lambda)|\sim \begin{cases} 2\sqrt{\pi}\lambda^{-1/4}|\sin(\lambda/2)|k^{3/4}e^{-\sqrt{2\lambda k}}, & \lambda\notin2\pi\mathbb{Z},\\ \sqrt{\pi}\lambda^{1/4}k^{1/4}e^{-\sqrt{2\lambda k}}, & \lambda\in2\pi\mathbb{Z}. \end{cases} \tag*{(35)}

Proof. Let rk(λ)r_k(\lambda) be the right-hand side of (34), with r0=1r_0=1 and r−k=rk‾r_{-k}=\overline{r_k}. The recurrence and derivative identities for UU in DLMF (13.3.7), (13.3.10), and (13.3.22) give

λrk′=ikλrk+k2(rk+1+rk−1−2rk),k≥1.\lambda r_k'=ik\lambda r_k+k^2(r_{k+1}+r_{k-1}-2r_k),\qquad k\geq1.

The negative-index equations follow by conjugation.

For Re⁡z>0\operatorname{Re} z > 0, (33) and a shift in the second integral give

k![U(k,1,z)−ke−zU(k+1,1,z)]=∫0∞e−zthk(t) dt(36)k!\left[U(k,1,z)-ke^{-z}U(k+1,1,z)\right]=\int_{0}^{\infty}e^{-zt}h_k(t)\,dt \tag*{(36)}
hk(t)=k[tk−1(1+t)k−1{t>1}(t−1)ktk+1].h_k(t)=k\left[\frac{t^{k-1}}{(1+t)^k}-\mathbf{1}_{\{t>1\}}\frac{(t-1)^k}{t^{k+1}}\right].

For t>1t>1 the ratio of the second term in brackets to the first is (1−t−2)k(1-t^{-2})^k, so hk≥0h_k\geq0. Integrating up to R>1R>1 and substituting s=t−1s=t-1 in the second term leaves a boundary term bounded by k/Rk/R; letting R→∞R\to\infty gives

∫0∞hk(t) dt=k∫0∞tk−1(1+t)k+1 dt=1.\int_{0}^{\infty}h_k(t)\,dt=k\int_{0}^{\infty}\frac{t^{k-1}}{(1+t)^{k+1}}\,dt=1.

Letting z=ε−iλz=\varepsilon-i\lambda with ε↓0\varepsilon\downarrow0 in (36), dominated convergence gives rk(λ)=∫0∞eiλthk(t) dtr_k(\lambda)=\int_{0}^{\infty}e^{i\lambda t}h_k(t)\,dt. Hence ∣rk(λ)∣≤1|r_k(\lambda)|\leq1, and Corollary 14 identifies rkr_k with qk(2,λ)q_k(2,\lambda).

The recurrence identity (DLMF (13.3.9))

U(k,1,z)−kU(k+1,1,z)=U(k,0,z)U(k,1,z)-kU(k+1,1,z)=U(k,0,z)

rewrites the left-hand side of (36) as

k![(1−e−z)U(k,1,z)+e−zU(k,0,z)].k!\left[(1-e^{-z})U(k,1,z)+e^{-z}U(k,0,z)\right].

On the principal branch, the large-parameter expansion of UU and the large-argument expansion of the modified Bessel function KνK_\nu of the second kind in DLMF (13.8.11) and (10.40.2) give

k!U(k,b,z)∼πez/2z1/4−b/2k1/4+b/2e−2kz,b∈{0,1},k!U(k,b,z)\sim\sqrt{\pi}e^{z/2}z^{1/4-b/2}k^{1/4+b/2}e^{-2\sqrt{kz}},\qquad b\in\{0,1\},

locally uniformly for z≠0z\ne0 with Re⁡z≥0\operatorname{Re}z\geq0. At z=−iλz=-i\lambda the b=1b=1 term dominates unless eiλ=1e^{i\lambda}=1, when it vanishes. This gives (35). □

The same Fourier coefficients give an explicit phase density. We use the exponential integral

E1(w)=∫1∞e−wtt dt,Re⁡w>0,E_1(w)=\int_{1}^{\infty}\frac{e^{-wt}}{t}\,dt,\qquad\operatorname{Re}w>0,

continued to C∖(−∞,0]\mathbb{C}\setminus(-\infty,0] on its principal branch; see DLMF (2026, §6.2(i)).

Corollary 27 (Phase density and pair correlation at β=2\beta=2). For λ>0\lambda>0 and 0<θ<2π0<\theta<2\pi, put

w=−iλ1−e−iθ=−λ2cot⁡θ2−iλ2.(37)w=-\frac{i\lambda}{1-e^{-i\theta}}=-\frac{\lambda}{2}\cot\frac{\theta}{2}-\frac{i\lambda}{2}. \tag*{(37)}

The density of αλ,2(0)\alpha_{\lambda,2}(0) mod 2π2\pi with respect to dθd\theta is

p2,λ(θ)=1−cos⁡λ+Re⁡([(1+w)(eiλ−1)−iλ]ewE1(w))2π(1−cos⁡θ).(38)p_{2,\lambda}(\theta)=\frac{1-\cos\lambda+\operatorname{Re}\left(\left[(1+w)(e^{i\lambda}-1)-i\lambda\right]e^wE_1(w)\right)}{2\pi(1-\cos\theta)}. \tag*{(38)}

Moreover, we recover the classical sine-kernel formula Dyson, [5]

ρ2(0,λ)=14π2[1−(sin⁡(λ/2)λ/2)2].\rho_2(0,\lambda)=\frac{1}{4\pi^2}\left[1-\left(\frac{\sin(\lambda/2)}{\lambda/2}\right)^2\right].

Proof. Put η=e−iθ\eta=e^{-i\theta}. For Re⁡z>0\operatorname{Re}z>0, summing the geometric series in (36) and shifting the second integral gives

∑k≥1ηkk![U(k,1,z)−ke−zU(k+1,1,z)]=η∫0∞e−zt[1+(1−e−z)t][1+(1−η)t]2 dt.\sum_{k\geq1}\eta^k k!\left[U(k,1,z)-ke^{-z}U(k+1,1,z)\right] =\eta\int_0^\infty\frac{e^{-zt}\left[1+(1-e^{-z})t\right]}{\left[1+(1-\eta)t\right]^2}\,dt.

The interchange is justified by hk≥0h_k\geq0 and ∑k≥1hk(t)=1+min⁡(t,1)\sum_{k\geq1}h_k(t)=1+\min(t,1). Integrating by parts and using

∫0∞e−zt1+(1−η)t dt=ez/(1−η)E1(z/(1−η))1−η,\int_0^\infty\frac{e^{-zt}}{1+(1-\eta)t}\,dt =\frac{e^{z/(1-\eta)}E_1(z/(1-\eta))}{1-\eta},

we may let z→−iλz\to-i\lambda by the locally uniform decay above to obtain

∑k≥1qk(2,λ)e−ikθ=e−iθ−eiλ+[(1+w)(eiλ−1)−iλ]ewE1(w)2(1−cos⁡θ).\sum_{k\geq1}q_k(2,\lambda)e^{-ik\theta} =\frac{e^{-i\theta}-e^{i\lambda}+\left[(1+w)(e^{i\lambda}-1)-i\lambda\right]e^wE_1(w)}{2(1-\cos\theta)}.

Fourier inversion gives (38), and the decay of the Fourier coefficients gives continuity at θ=0\theta=0.

At angle zero, the identity ∑k≥1hk(t)=1+min⁡(t,1)\sum_{k\geq1}h_k(t)=1+\min(t,1) gives, for Re⁡z>0\operatorname{Re}z>0,

∫0∞e−zt(1+min⁡(t,1)) dt=1z+1−e−zz2.\int_0^\infty e^{-zt}(1+\min(t,1))\,dt=\frac{1}{z}+\frac{1-e^{-z}}{z^2}.

The locally uniform decay in kk allows passage to z=−iλz=-i\lambda, giving

∑k≥1qk(2,λ)=iλ+eiλ−1λ2.\sum_{k\geq1}q_k(2,\lambda)=\frac{i}{\lambda}+\frac{e^{i\lambda}-1}{\lambda^2}.

Substituting into the Fourier formula (32)

ρ2(0,λ)=14π2(1+2Re⁡∑k≥1qk(2,λ))\rho_2(0,\lambda)=\frac{1}{4\pi^2}\left(1+2\operatorname{Re}\sum_{k\geq1}q_k(2,\lambda)\right)

gives the stated sine-kernel formula.

Remark 28 (The phase at integer multiples of 2π2\pi). At λ=2πm\lambda= 2\pi m, m≥1m \ge1 an integer, the density (38) simplifies to the positive real integral

p2,2πm(θ)=m2∫0∞e−πmt1−tsin⁡θ+t2sin⁡2(θ/2) dt,0<θ<2π.(39)p_{2,2\pi m}(\theta) = \frac{m}{2} \int_{0}^{\infty} \frac{e^{-\pi m t}}{1-t\sin\theta+t^2\sin^2(\theta/2)}\,dt,\qquad0<\theta<2\pi. \tag*{(39)}

Equivalently,

α2πm,2(0) mod⁡ 2π=d2arccot⁡(X+Y) mod⁡ 2π,(40)\alpha_{2\pi m,2}(0)\ \operatorname{mod}\ 2\pi\mathrel{\overset{d}{=}} 2\operatorname{arccot}(X+Y)\ \operatorname{mod}\ 2\pi, \tag*{(40)}

where XX is standard Cauchy and YY is independent exponential with rate πm\pi m. To see this, substitute eiλ=1e^{i\lambda}=1 in (38) and use

ewE1(w)=∫0∞e−ss+w ds,w∉(−∞,0],e^w\mathrm{E}_1(w)=\int_{0}^{\infty}\frac{e^{-s}}{s+w}\,ds,\qquad w\notin(-\infty,0],

the analytic continuation of DLMF (6.7.1). Evaluating the real part and setting s=πmts=\pi mt gives (39). Under x=cot⁡(θ/2)x=\cot(\theta/2), the density becomes the convolution of the exponential density with rate πm\pi m and the standard Cauchy density, proving (40).

AI use. ChatGPT was used extensively to develop the random walk coupling, the pair correlation identity, and the analytic continuation argument, and to check calculations, locate references, and revise the manuscript. The authors take full responsibility for the mathematics and references.

Acknowledgments. Benedek Valkó was partially supported by the National Science Foundation award DMS-2246435. Bálint Virág was partially supported by an NSERC Discovery Grant.

Yun Li. School of Mathematical Sciences, Nankai University, Tianjin, China. [email protected]

Yahui Qu. Department of Mathematics, University of Wisconsin–Madison, Madison, USA. [email protected]

Benedek Valkó. Department of Mathematics, University of Wisconsin–Madison, Madison, USA. [email protected]

Bálint Virág. Department of Mathematics, University of Toronto, Toronto, Canada. [email protected]

References

  1. [1]T. Assiotis and J. Najnudel (2026). Moments of CβE field partition function, Sineβ correlations and stochastic zeta. arXiv:2602.08739.
  2. [4]R. Durrett (2019). Probability: Theory and Examples, fifth edition. Cambridge University Press.
  3. [5]F. J. Dyson (1962). Statistical theory of the energy levels of complex systems. III. J. Math. Phys. 3, 166–175.DOI
  4. [6]S. N. Ethier and T. G. Kurtz (1986). Markov Processes: Characterization and Convergence. John Wiley & Sons.
  5. [8]P. J. Forrester (1992). Selberg correlation integrals and the 1/r² quantum many-body system. Nuclear Physics B 388(3), 671–699.
  6. [9]P. J. Forrester (1994). Addendum to “Selberg correlation integrals and the 1/r² quantum many-body system”. Nuclear Physics B 416(1), 377–385.DOI
  7. [10]P. J. Forrester (2010). Log-Gases and Random Matrices. London Mathematical Society Monographs 34, Princeton University Press.DOI
  8. [12]Z. N. C. Ha (1995). Fractional statistics in one dimension: view from an exactly solvable model. Nuclear Physics B 435(3), 604–636. doi:10.1016/0550-3213(94)00537-O.DOI
  9. [14]B. Jancovici (1981). Exact results for the two-dimensional one-component plasma. Physical Review Letters 46(6), 386–388. doi:10.1103/PhysRevLett.46.386.DOI
  10. [15]Y. Katznelson (2004). An Introduction to Harmonic Analysis, third edition. Cambridge University Press.
  11. [16]R. Killip and M. Stoiciu (2009). Eigenvalue statistics for CMV matrices: From Poisson to clock via random matrix ensembles. Duke Mathematical Journal 146, 361–399.DOI
  12. [17]Y. Li and B. Valkó (2022). Operator level limit of the circular Jacobi β-ensemble. Random Matrices: Theory and Applications 11(4), 2250043. doi:10.1142/S2010326322500435.
  13. [18]F. Nakano (2014). Level statistics for one-dimensional Schrödinger operators and Gaussian beta ensemble. Journal of Statistical Physics 156, 66–93. doi:10.1007/s10955-014-0987-x.DOI
  14. [19]A. Okounkov (1997). On n-point correlations in the log-gas at rational temperature. arXiv:hep-th/9702001.arxiv.org/abs/hep-th/9702001
  15. [21]T. Ransford (1995). Potential Theory in the Complex Plane. London Mathematical Society Student Texts 28, Cambridge University Press.DOI

Paper details

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

views · 0

downloads · 0

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

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents