Abstract
We extend the method of rescaled Ward identities of Ameur, Kang, and Makarov to study the distribution of eigenvalues close to a bulk singularity, i.e., a point in the interior of the droplet where the density of the classical equilibrium measure vanishes. We prove results to the effect that a certain “dominant part” of the Taylor expansion determines the microscopic properties near a bulk singularity. A description of the distribution is given in terms of the Bergman kernel of a certain Fock-type space of entire functions.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and Main Results
1.1. Consider a system \(\{\zeta _j\}_1^n\) of identical point-charges in the complex plane in the presence of an external field nQ, where Q is a suitable function. The system is assumed to be picked randomly with respect to the Boltzmann–Gibbs probability law at inverse temperature \(\beta =1\),
where \(H_n\) is the weighted energy of the system,
The constant \(Z_n\) in (1.1) is chosen so that the total mass is 1.
It is well known that (with natural conditions on Q) the normalized counting measures \(\mu _n=\frac{1}{n}\sum _{j=1}^n\delta _{\zeta _j}\) converge to Frostman’s equilibrium measure as \(n\rightarrow \infty \). This is a probability measure of the form
where \(\chi _S\) is the indicator function of a certain compact set S called the droplet.
We necessarily have \(\Delta Q\ge 0\) on S. In the papers [4, 5], the method of rescaled Ward identities was introduced and applied to study microscopic properties of the system \(\{\zeta _j\}_1^n\) close to a (moving) point \(p\in S\). The situation in those papers is however restricted by the condition that the point p be “regular” in the sense that \(\Delta Q(p)\ge \mathrm {const.}>0\). In this paper, we extend the method to allow for a “bulk singularity,” i.e., an isolated point p in the interior of S at which \(\Delta Q=0\).
A bulk singularity tends to repel particles away, which means that one must use a relatively coarse scale in order to capture the relevant structure. We prove results to the effect that (in many cases) the dominant terms in the Taylor expansion of \(\Delta Q\) about p determine the microscopic properties of the system in the vicinity of p. Our characterization uses the Bergman kernel for a certain space of entire functions, associated with these dominant terms. In particular, we obtain quite different distributions depending on the degree to which \(\Delta Q\) vanishes at p.
Remark
It is well known that the particles \(\{\zeta _j\}_1^n\) can be identified with eigenvalues of random normal matrices with a suitable weighted distribution. The details of this identification are not important for the present investigation. However, following tradition, we shall sometimes speak of a “configuration of random eigenvalues” instead of a “particle-system.”
Remark
The meaning of the convergence \(\mu _n\rightarrow \sigma \) is that \({{\mathbf {E}}}_n[\mu _n(f)]\rightarrow \sigma (f)\) as \(n\rightarrow \infty \), where f is a suitable test-function, say, continuous and bounded, and \({{\mathbf {E}}}_n\) is expectation with respect to (1.1). In fact, more can be said, see [2].
Notation We write \(\Delta ={\partial }{\bar{\partial }}\) for 1 / 4 of the usual Laplacian and dA for \(1/\pi \) times Lebesgue measure on the plane \({{\mathbb {C}}}\). Here \({\partial }=\frac{1}{2}({\partial }/{\partial }x-i{\partial }/{\partial }y)\) and \({\bar{\partial }}=\frac{1}{2}({\partial }/{\partial }x+i{\partial }/{\partial }y)\) are the usual complex derivatives. We write \({\bar{z}}\) (or occasionally \(z^*\)) for the complex conjugate of a number z. A continuous function h(z, w) will be called Hermitian if \(h(z,w)=h(w,z)^*\). h is called Hermitian-analytic (Hermitian-entire) if h is Hermitian and analytic (entire) in z and \({\bar{w}}\). A Hermitian function c(z, w) is called a cocycle if there is a unimodular function g such that \(c(z,w)=g(z){\bar{g}}(w)\), where for functions we use the notation \({\bar{f}}(z)=f(z)^*\). We write D(p, r) for the open disk with center p and radius r.
1.1 Potential and Equilibrium Measure
The function Q is usually called the “external potential.” This function is assumed to be lower semi-continuous and real-valued, except that it may assume the value \(+\infty \) in portions of the plane. We also assume: (i) the set \(\Sigma _0=\{Q<\infty \}\) has dense interior, (ii) Q is real-analytic in \({\text {Int}}\Sigma _0\), and (iii) Q satisfies the growth condition
For a suitable measure on \({{\mathbb {C}}}\), we define its Q-energy by
The equilibrium measure \(\sigma =\sigma _Q\) is defined as the probability measure that minimizes \(I_Q[\mu ]\) over all compactly supported Borel probability measures \(\mu \). Existence and uniqueness of such a minimizer is well known, see, e.g., [28], where also the explicit expression (1.2) is derived, with \(S={\text {supp}}\,\sigma \).
1.2 Rescaling
Recall that \(\Delta Q\ge 0\) on S. The purpose of the present investigation is to study (isolated) points \(p\in {\text {Int}}S\) at which \(\Delta Q(p)=0\). We refer to such points as bulk singularities. Without loss of generality, we can assume that \(p=0\) is such a point, and we study the microscopic behavior of the system \(\{\zeta _j\}_1^n\) near 0.
By the microscopic scale at \(p=0\), we mean the positive number \(r_n=r_n(p)\) having the property
Intuitively, \(r_n(p)\) means the expected distance from a particle at p to its closest neighbor. If p is a regular bulk point, then, as is easily seen,
which gives the familiar scaling factor used in papers such as [1, 4].
Since the Laplacian \(\Delta Q\) vanishes at 0 and is real-analytic and nonnegative in a neighborhood, there is an integer \(k\ge 1\) such that the Taylor expansion of \(\Delta Q\) about 0 takes the form \(\Delta Q(\zeta )={\tilde{P}}(\zeta )+O(|\,\zeta \,|^{\,2k-1})\), where \({\tilde{P}}(x+iy)=\sum _{j=0}^{2k-2}a_j\,x^{\,j}y^{\,2k-2-j}\) is a positive semi-definite polynomial, homogeneous of degree \(2k-2\).
We refer to the number \(2k-2={\text {degree}}{\tilde{P}}\) as the type of the bulk-singularity at the origin. We shall say that the singularity is non-degenerate if \({\tilde{P}}\) is positive definite, i.e., if there is a positive constant c such that \({\tilde{P}}(\zeta )\ge c\,\left| {\,\zeta \,} \right| ^{\,2k-2}.\) Hereafter, we tacitly assume that this condition is satisfied.
It will be important to have a good grasp of the size of \(r_n=r_n(0)\) as \(n\rightarrow \infty \). For this, we note that
where \(\tau _0=\tau _0[Q,0]\) is the positive constant satisfying
We will call \(\tau _0\) the modulus of the bulk singularity at 0. We have the following lemma; the simple verification is omitted here.
Lemma 1.1
For the microscopic scale \(r_n\) at 0, we have \(r_n=\tau _0\, n^{-1/2k}\,(1+O(n^{-1/2k}))\) as \(n\rightarrow \infty \), where \(\tau _0\) is the modulus (1.4).
Example
For the Mittag–Leffler potential \(Q=\left| {\,\zeta \,} \right| ^{\,2k}\), the droplet is the disk \(\left| {\,\zeta \,} \right| \le k^{-1/2k}\). For \(k=1\), we have the well-known Ginibre potential, cf., e.g., [16, Chapter 15]. For \(k\ge 2\), the Mittag–Leffler potential has a bulk singularity at the origin of type \(2k-2\). It is easy to check that the modulus equals \(\tau _0=k^{-1/2k}\).
Let p be an integer, \(1\le p\le n\). The p-point function of the point-process \(\{\zeta _j\}_1^n\) is the function of p complex variables \(\eta _1,\ldots ,\eta _p\) defined by
The p-point function \({\mathbf {R}}_{n,p}\) should really be understood as the density in the measure \({\mathbf {R}}_{n,p}(\eta _1,\ldots ,\eta _p)\, \mathrm{d}A(\eta _1)\cdots \mathrm{d}A(\eta _p)\). This should be kept in mind when we subject the \(\eta _j\) to various transformations.
A well-known algebraic fact (“Dyson’s determinant formula,” see, e.g., [26] or [28, p. 249]) states that the p-point function takes the form of a determinant,
where \({{\mathbf {K}}}_n\) is a certain Hermitian function called a correlation kernel of the process (Cf. Sect. 2). Of particular importance is the one-point function \({\mathbf {R}}_n={\mathbf {R}}_{n,1}.\)
We now rescale about the origin on the microscopic scale \(r_n\) about the bulk singularity at 0. The rescaled system \(\{z_j\}_1^n\) is taken to be
with the law given by the image of the Boltzmann–Gibbs distribution (1.1) under the scaling (1.5).
It follows that the rescaled system \(\{z_j\}_1^n\) is determinantal with p-point function
where the correlation kernel \(K_n\) for the rescaled system is given by
In particular, the one-point function of the process \(\{z_j\}_1^n\) is \(R_n(z)=K_n(z,z).\)
Clearly, a correlation kernel \(K_n(z,w)\) is only determined up to multiplication by a cocycle \(c_n(z,w)\).
1.3 Main Structural Lemma
Now suppose that Q has a bulk-singularity of type \(2k-2\) at the origin. It will be useful to single out a canonical “dominant part” of Q near 0. To this end, let \(P(x+iy)\) be the Taylor polynomial of Q of degree 2k about the origin. Let H be the holomorphic polynomial
We will write
We then have the basic decomposition
where \(Q_1(\zeta )=O(|\,\zeta \,|^{\,2k+1})\) as \(\zeta \rightarrow 0\).
The following lemma gives the basic structure of limiting kernels at a singular point (not necessarily in the bulk).
Lemma 1.2
There exists a sequence \(c_n\) of cocycles such that every subsequence of the sequence \(c_nK_n\) has a subsequence converging uniformly on compact subsets to some Hermitian function K. Every limit point K has the structure
where L is a Hermitian-entire function.
Following [4], we refer to a limit point K in Lemma 1.2 as a limiting kernel, whereas L is a limiting holomorphic kernel. We also speak of the limiting 1-point function
Note that R determines K and L by polarization.
Remark
Each limiting one-point function gives rise to a unique limiting point field (or “infinite particle system”) \(\{z_j\}_1^\infty \) with intensity functions
(This follows from Lenard’s theory, see [29] or [4].) It is possible that a limiting point field is trivial in the sense that \(K=0\).
1.4 Universality Results
We will prove universality for two kinds of bulk singularities. Referring to the canonical decomposition \(Q=Q_0+Re\, H+Q_1\) with \(Q_0\) of degree 2k, we say a singularity at 0 is:
-
(i)
homogeneous if \(Q_1=0\) and \(H(z)=c\,z^{\,2k}\) for some constant c,
-
(ii)
dominant radial if \(Q_0\) is radially symmetric, i.e., \(Q_0(z)=Q_0(|\,z\,|)\).
We remark that a homogeneous singularity is necessarily located in the bulk of the droplet; for other types of singularities this must be postulated.
In the following, we denote by \(L_0\) the Bergman kernel of the space of entire functions \(L^2_a(\mu _0)\) associated with the measure
Theorem 1.3
If there is a homogeneous singularity at 0, we have \(L=L_0\) for each limiting holomorphic kernel L.
The next result concerns limiting holomorphic kernels L(z, w) which are rotationally symmetric in the sense that \(L(z,w)=L(ze^{it},we^{it})\) for all real t. Equivalently, L is rotationally symmetric if there is an entire function E such that \(L(z,w)=E(z{\bar{w}})\). (We leave the simple verification of this to the reader.)
Theorem 1.4
If a bulk singularity at 0 is dominant radial, then \(L=L_0\) for each rotationally symmetric limiting kernel.
The result was conjectured in [4, Section 7.3].
We do not know whether or not each limiting kernel at a dominant radial bulk singularity is rotationally symmetric. This question has some similarity to that of deciding the translation invariance of limiting kernels at regular boundary points. See [4] for several comments about this, notably the interpretation in terms of a twisted convolution equation in Section 7.1.
It is natural to conjecture that the kernel in Theorem 1.3 be equal to the limiting kernel in general, regardless of the nature of a (nondegenerate) bulk singularity.
Remark
Note that, as a consequence of the reproducing property of the kernel \(L_0\), we have in the situation of the above theorems the mass-one equation for a limiting kernel K, \(\int _{{\mathbb {C}}}\left| {\,K(z,w)\,} \right| ^{\,2}\, \mathrm{d}A(w)=R(z)\).
Example
For the Mittag–Leffler potential \(Q=|\,\zeta \,|^{\,2k}\), it is possible to calculate the limiting kernel L explicitly, using orthogonal polynomials (see [4, Section 7.3] or[30]). The result is that
where
The function \(M_k\) can be expressed as \(M_k(z)=\tau _0^{\,2}\,k\, E_{1/k,1/k}(\tau _0^{\,2} z)\), where \(E_{a,b}\) is the Mittag–Leffler function (see [18])
Using Theorem 1.3, we can now see that the kernel in (1.7) is universal for potentials of the form \(Q=|\,\zeta \,|^{\,2k}+Re\left( c\,\zeta ^{\,2k}\right) \). (We must insist that \(|\,c\,|<1\) to insure that the growth assumption of Q at infinity is satisfied, see (1.3).) By Theorem 1.4, the universality holds also for all rotationally symmetric limiting kernels \(L(z,w)=E(z{\bar{w}})\) for more general potentials of the form \(Q(\zeta )=\left| {\,\zeta \,} \right| ^{\,2k}+ Re H(\zeta )+Q_1(\zeta )\).
Remark
For \(k=1\) (i.e., when 0 is a “regular” bulk point) the space \(L^2_a(\mu _0)\) becomes the standard Fock space, normed by \(\Vert \,f\,\Vert ^{\,2}=\int _{{\mathbb {C}}}|\,f(z)\,|^{\,2}\mathrm{e}^{-\,|\,z\,|^{\,2}}\, \mathrm{d}A(z)\). In this case, we have \(R=1\) for the limiting 1-point function, by the well-known Ginibre\((\infty )\)-limit. (See, e.g., [4].)
1.5 Further Results
In the following, we consider a potential with canonical decomposition \(Q=Q_0+Re\, H+Q_1\). Following [4], we shall prove auxiliary results that fall into three categories.
Ward’s Equation Let \(R(z)=K(z,z)\) be a limiting kernel in Lemma 1.2. At a point z where \(R>0\), we write
We call B(z, w) a limiting Berezin kernel rooted at z; C(z) is its Cauchy transform.
Theorem 1.5
Let R be a limiting 1-point function.
-
(i)
Zero-one law: Either \(R=0\) identically, or else \(R>0\) everywhere.
-
(ii)
Ward’s equation: If R is nontrivial, we have that
$$\begin{aligned} {\bar{\partial }}C(z)=R(z)-\Delta _z \left[ Q_0(\tau _0z)\right] -\Delta _z\log R(z). \end{aligned}$$(1.10)
As \(n\rightarrow \infty \), it may well happen that \(R_n\rightarrow 0\) locally uniformly (if the singularity at 0 is in the exterior of the droplet).
Apriori Estimates To rule out the possibility of trivial limiting kernels, we shall use the following result.
Theorem 1.6
Let R be any limiting kernel, and let \(R_0(z)=L_0(z,z)\mathrm{e}^{-Q_0(\tau _0z)}\), where \(L_0\) is the Bergman kernel of the space \(L^2_a(\mu _0)\). Then
-
(i)
\(R_0(z)=\Delta _z [Q_0(\tau _0z)]\cdot (1+O(z^{1-k}))\), as \(z\rightarrow \infty \),
-
(ii)
\(R(z)\,\,\,=\Delta _z [Q_0(\tau _0z)]\cdot (1+O(z^{1-k}))\), as \(z\rightarrow \infty \).
Part (i) depends on an estimate of the Bergman kernel for the space \(L^2_a(\mu _0)\). Related estimates valid when \(Q_0\) is a function satisfying uniform estimates of the type \(0<c\le \Delta Q_0\le C\) are found in Lindholm’s paper [24].
In our situation, the function \(\Delta Q_0\) takes on all values between 0 and \(+\infty \), which means that the results from [24] are not directly applicable. It is convenient to include an elementary discussion for the case at hand, following the method of “approximate Bergman projections” in the spirit of [4, Section 5]. This has the advantage that proof of part (ii) follows after relatively simple modifications.
Remark
Part (i) of Theorem 1.6 seems to be of some relevance for the investigation of density conditions for sampling and interpolation in Fock-type spaces \(L^2_a(\mu _0)\); see the recent paper [17, Remark 5.6], cf. [31] for the classical Fock space case. (A very general result of this sort was obtained by different methods in the paper [25], where the hypothesis on the “weight” \(Q_0\) is merely that the Laplacian \(\Delta Q_0\) be a doubling measure.)
Remark
In the case \(Q=|\,z\,|^{\,2\lambda }\), the asymptotic formula in Theorem 1.6 (i) has an alternative proof by more classical methods, using an asymptotic expansion for the function \(M_\lambda (z)\) as \(z\rightarrow \infty \) ([18, Section 4.7]). The formula (i) can be recognized as giving the leading term in that expansion.
Positivity Recall that a Hermitian function K is called a positive matrix if
for all points \(z_j\in {{\mathbb {C}}}\) and all complex scalars \(\alpha _j\). It is clear that each limiting (holomorphic) kernel is a positive matrix.
Theorem 1.7
Let L be a limiting holomorphic kernel. Then L is the Bergman kernel for a Hilbert space \({{\mathcal {H}}}_*\) of entire functions that sits contractively in \(L^2_a(\mu _0)\). Moreover, \(L_0-L\) is a positive matrix.
Here \(L_0\) is the Bergman kernel of \(L^2_a(\mu _0)\).
It may well happen that the space \({{\mathcal {H}}}_*\) degenerates to \(\{0\}\). This is the case when the singularity at 0 is located in the exterior of the droplet.
Comments An interesting generalization of our situation is obtained by allowing for a suitably scaled logarithmic singularity at a (regular or singular) bulk point. More precisely, if \({\tilde{Q}}\) is smooth in a neighborhood of 0, we consider a potential of the form \(Q(\zeta )={\tilde{Q}}(\zeta )+2(c/n)\log \left| {\,\zeta \,} \right| \), where \(c<1\) is a constant. Rescaling by \(z=r_n^{-1}\zeta \), where \(c+n\int _{D(0,r_n)}\Delta {\tilde{Q}}\, \mathrm{d}A=1\), we find \(r_n\sim (1-c)^{1/2k}\tau _0n^{-1/2k}\) as \(n\rightarrow \infty \), where \(2k-2\) is the type of \({\tilde{Q}}\) and \(\tau _0=\tau _0[{\tilde{Q}},0]\). It is hence natural to define the dominant part by \(Q_0(z):=c'\,{\tilde{Q}}_0(z)+2c\log |\,z\,|\), where \({\tilde{Q}}_0\) is the dominant part of \({\tilde{Q}}\) and \(c'\) a suitable constant depending on c. In particular, if \(Q(\zeta )=c_1|\,\zeta \,|^{\,2\lambda }+(c_2/n)\log \left| {\,\zeta \,} \right| \) with suitable \(c_1,c_2>0\), the dominant part becomes of the type
for suitable constants \(\lambda \) and \(\mu \). The potential (1.11) was introduced in the paper [3], where all rotationally symmetric solutions to the corresponding Ward equation (1.10) were found. Recently, certain potentials of this form were studied in a context of Riemann surfaces, in a scaling limit about certain types of singular points (conical singularities and branch points), see [22]. We will return to this issue in a forthcoming paper [6].
An interesting occurrence of Mittag–Leffler potentials with logarithmic singularities is found in the paper [11]; the dominant part \(Q_0=|z|^2-(2/M)\log |z|\) is there connected to the eigenvalue density of the product of M complex Gaussian matrices. We are grateful to one of the referees for bringing this to our attention.
As in [4, Section 7.7], we note that it is possible to introduce an “inverse temperature” \(\beta \) into the setting; the case at hand then corresponds to \(\beta =1\). For general \(\beta \), the rescaled process \(\{z_j\}_1^n\) is no longer determinantal, but the rescaled intensity functions \(R_{n,p}^\beta \) make perfect sense. As \(n\rightarrow \infty \), we formally obtain a “Ward’s equation at a bulk singularity” of the form
Here \(C^\beta (z)\) should be understood as the Cauchy transform of the \(\beta \)-Berezin kernel \(B^\beta (z,w)=(R_1^\beta (z)R_1^\beta (w)-R_2^\beta (z,w))/R_1^\beta (z)\). The objects in (1.12) are so far understood mostly on a physical level. We now give a few remarks in this spirit.
First, if 0 is a regular bulk-point, i.e., if \(\Delta Q(0)>0\), then it is believed that \(R^\beta =1\) identically, i.e., the right-hand side in (1.12) should vanish. The equation (1.12) then reflects the fact that the Berezin kernel \(B^\beta (z,w)=b^\beta (r)\) depends only on the distance \(r=|\,z-w\,|\). When \(\beta =1\), one has the well-known identity \(b^1(r)=\mathrm{e}^{-\,r^{\,2}}\). For other \(\beta \) we do not know of an explicit expression, but it was shown by Jancovici in [20] that
where f is a certain explicit function. In the bulk-singular case, the kernel \(B^\beta (z,w)\) will not just depend on \(|\,z-w\,|\), but still it seems natural to expect that we have an expansion of the form
where \(b^1(z,w)=\left| {\,L_0(z,w)\,} \right| ^{\,2}\mathrm{e}^{-Q_0(\tau _0w)}/L_0(z,z)\), \(L_0\) being the Bergman kernel of the space \(L^2_a(\mu _0)\). A natural problem, which will not be taken up here, is to determine the function f(z, w) in (1.13). (A similar investigation at regular boundary points was made recently in the paper [12].)
For boundary points, the term “singular” has a different meaning than for bulk points. Indeed, the singular points p (cusps or double points) studied in the paper [5] all satisfy \(\Delta Q(p)>0\). An example of a situation at which \(\Delta Q=0\) at a boundary point (at 0) is provided by the potential \(Q=|\,\zeta \,|^{\,4}-\sqrt{2}Re(\,\zeta ^{\,2}\,)\). (The boundary of S is here a “figure 8” with 0 at the point of self-intersection, see [9].) A natural question is whether it is possible to define nontrivial scaling limits at (or near) these kinds of singular points, in the spirit of [5].
There is a parallel theory for scaling limits for Hermitian random matrix ensembles. In this situation, the droplet is a union of compact intervals. It is well known that the sine-kernel appears in the scaling limit about a “regular bulk point,” i.e., an interior point where the density of the equilibrium measure is strictly positive. In a generic case, all points are regular, see [21]. Special bulk points where the equilibrium density vanishes may be called “singular”; at such points other types of universality classes appear, see [10, 14, 27]; cf. [15] for a corresponding nonbulk situation.
Finally, we wish to mention that the investigations in this paper were partly motivated by applications to the distribution of Fekete points close to a bulk singularity (see [1]). This issue will be taken up in a later publication.
Addendum After the completion of this paper, at least two relevant papers have appeared. The note [7] generalizes and improves some of our above results, notably Theorem 1.6, in a context allowing for the logarithmic singularities discussed above. The paper [23] investigates logarithmic singularities with respect to properties of associated orthogonal polynomials.
1.6 Plan of the Paper
In Sect. 2, we prove the general structure formula for limiting kernels (Lemma 1.2). We also prove the positivity theorem (Theorem 1.7). In Sect. 3, we prove Ward’s equation and the zero-one law (Theorem 1.5). In Sect. 4, we prove the universality results (Theorems 1.3 and 1.4). Our proof of Theorem 1.4 depends on the apriori estimate from Theorem 1.6, part (ii). In the last two sections, we prove the asymptotics for the functions \(R_0\) and R in Theorem 1.6. For \(R_0\) (part (i)), see Sect. 5; for R (part (ii)), see Sect. 6.
1.7 Convention
Multiplying the potential Q by a suitable constant, we can in the following assume that the modulus \(\tau _0=1\). In fact, the slightly more general assumption that \(\tau _0=1+O(n^{-1/2k})\) as \(n\rightarrow \infty \) will do equally well. This means the microscopic scale about 0 can be taken as \(r_n=n^{-1/2k}\), where \(2k-2\) is the type of the singularity. This will be assumed throughout the rest of this paper.
2 Structure of Limiting Kernels
In this section, we prove Lemma 1.2 on the general structure of limiting kernels and the positivity Theorem 1.7. We shall actually prove a little more: a limiting holomorphic kernel can be written as a subsequential limit of kernels for certain specific Hilbert spaces of entire functions. In later sections, we will use this additional information for our analysis of homogeneous bulk singularities.
2.1 Spaces of Weighted Polynomials
It is well known that we can take for correlation kernel for the process \(\{\zeta _j\}_1^n\) the reproducing kernel for a suitable space of weighted polynomials. Here the “weight” can either be incorporated into the polynomials themselves or into the norm of the polynomials. We will use both these possibilities. In the following, we shall use the symbol “\({\text {Pol}}(n)\)” for the linear space of holomorphic polynomials of degree at most \(n-1\) (without any topology). We write \(\mu _n\) for the measure \(\mathrm{d}\mu _n=\mathrm{e}^{-nQ}\, \mathrm{d}A\).
We let \({{\mathcal {P}}}_n\) denote the space \({\text {Pol}}(n)\) regarded as a subspace of \(L^2(\mu _n)\). The symbol \({{\mathcal {W}}}_n\) will denote the set of weighted polynomials \(f=p\,\mathrm{e}^{-nQ/2}\), (\(p\in {\text {Pol}}(n)\)) regarded as a subspace of \(L^2=L^2(\mathrm{d}A)\). We write \({{\mathbf {k}}}_n\) and \({{\mathbf {K}}}_n\) for the reproducing kernels of \({{\mathcal {P}}}_n\) and \({{\mathcal {W}}}_n\), respectively, and we note that
Now suppose that Q has a bulk singularity at the origin, of type \(2k-2\) and rescale at the microscopic scale by
2.2 Limiting Holomorphic Kernels
Suppose that there is a bulk singularity of type \(2k-2\) at the origin. Consider the canonical decomposition \(Q= Q_0 +Re\, H+ Q_1\) and write \(h=Re\, H\). Thus h is of degree at most 2k, \(Q_0\) is a positive definite homogeneous polynomial of degree 2k, and \(Q_1(\zeta )=O(\left| {\,\zeta \,} \right| ^{\,2k+1})\) as \(\zeta \rightarrow 0\).
Lemma 2.1
For each compact subset \(V\) of \({{\mathbb {C}}}\), there is a constant \(C=C(V)\) such that \(K_n(z,z) \le C\) for \(z \in V\).
Proof
Let \({\tilde{{{\mathcal {W}}}}}_n\) denote the space of all “rescaled” weighted polynomials \(p \cdot \mathrm{e}^{-{\tilde{Q}}_n/2}\), where \(p\in {\text {Pol}}(n)\) and \({\tilde{Q}}_n(z)= nQ(r_n z)\). Regarding \({\tilde{{{\mathcal {W}}}}}_n\) as a subspace of \(L^2\), we recognize that \(K_n\) is the reproducing kernel of \({\tilde{{{\mathcal {W}}}}}_n\). Hence
Fix a number \(\delta >0\), and let \(V_{\delta }= \{\, z \in {{\mathbb {C}}}\, ;\, {\text {dist}}(z,V)\le \delta \, \}\). We also pick a number \(\alpha > \sup \{\, \Delta Q_0(z)\, \ ; \ z \in V_{\delta }\, \}\). Now let u be an analytic function in a neighborhood of \(V_{\delta }\), and consider the function \(g_n(z)= u(z)\, \mathrm{e}^{-\,{\tilde{Q}}_n (z)/2+\,\alpha \, \left| {\,z\,} \right| ^{\,2}/2}\). Note that
Hence \(\Delta \log |\,g_n(z)\,|^{\,2} \ge - \Delta {\tilde{Q}}_n(z) + \alpha >0\) for all sufficiently large n and all \(z \in V_{\delta }\). Thus \(\left| {\,g_n\,} \right| ^{\,2}\) is subharmonic in \(V_\delta \), so for \(z \in V\),
We obtain
where \(M_{V}=\sup _{z\in V}{|\,z\,|}\). By (2.1) and (2.2), \(K_n(z,z)\) is bounded for \(z\in V\). \(\square \)
We now use the holomorphic polynomial H in the decomposition \(Q=Q_0+Re\, H+Q_1\) to define a Hermitian-entire function (“rescaled holomorphic kernel”) by
Let us write
so that \(nQ(r_nz)=Q_0(z)+Re\, H_n(z)+Q_{1,n}(z)\) and
Define a Hilbert space of entire functions by
equipped with the norm of \(L^2({\tilde{\mu }}_n)\), where
Observe that \(Q_{1,n}=O(r_n)\) as \(n\rightarrow \infty \), where the O-constant is uniform on each given compact subset of \({{\mathbb {C}}}\). In particular, \({\tilde{\mu }}_n\rightarrow \mu _0\) vaguely where \(d\mu _0=\mathrm{e}^{-Q_0}\, \mathrm{d}A.\)
The following result implies Lemma 1.2; it also generalizes [4, Lemma 4.9].
Lemma 2.2
Each subsequence of the kernels \(L_n\) has a further subsequence converging locally uniformly to a Hermitian-entire limit L. Furthermore, \(L_n\) is the reproducing kernel of the space \({{\mathcal {H}}}_n\), and L satisfies the “mass-one inequality,”
Finally, there exists a sequence of cocycles \(c_n\) such that each subsequence of \(c_nK_n\) converges locally uniformly to a Hermitian function K of the type \(K(z,w)=L(z,w)\mathrm{e}^{-Q_0(z)/2-Q_0(w)/2}\).
Proof
Define a function \(E_n(z,w)\) by
Note that \(K_n= L_n E_n\), where \(L_n\) is the Hermitian-entire kernel (2.3). Now, if \(h=Re\, H\), then
We have shown that
where \(o(1)\rightarrow 0\) locally uniformly on \({{\mathbb {C}}}^2\) and \(c_n\) is a cocycle:
On the other hand, for each compact subset \(V\) of \({{\mathbb {C}}}^2\), there is a constant C such that
for sufficiently large n. By Lemma 2.1, the functions \(L_n\) have a uniform bound on \(V\). We have shown that \(\{L_n\}\) is a normal family. We can hence extract a subsequence \(\{ L_{n_\ell } \}\), converging locally uniformly to a Hermitian-entire function L(z, w).
Choosing cocycles \(c_n\) such that \(c_{n} E_n \rightarrow \mathrm{e}^{-Q_0(z)/2-Q_0(w)/2}\) uniformly on compact subsets as \(n\rightarrow \infty \), we now obtain that
The reproducing property \(\int |\,K_{n_\ell }(z,w)\,|^{\,2} \mathrm{d}A(w)= K_{n_\ell }(z,z)\) means that
Letting \(\ell \rightarrow \infty \), we obtain the mass-one inequality (2.5) by Fatou’s lemma.
There remains to prove that \(L_n\) is the reproducing kernel for the space \({{\mathcal {H}}}_n\). For this, we write \(L_{n,w}(z)=L_n(z,w)\) and note that for an element \(f=q\cdot \mathrm{e}^{-H_n/2}\) of \({{\mathcal {H}}}_n\), we have
Noting that \(k_n\) is the reproducing kernel for the space \({\tilde{{{\mathcal {P}}}}}_n\) of polynomials of degree at most \(n-1\) normed by \(\Vert \,p\,\Vert ^{\,2}=\int _{{\mathbb {C}}}\left| {\,p(z)\,} \right| ^{\,2}\mathrm{e}^{-nQ(r_nz)}\, \mathrm{d}A(z)\), we now see that
The proof of the lemma is complete. \(\square \)
2.3 The Positivity Theorem
Let \(\mu _0\) be the measure \(d\mu _0=\mathrm{e}^{-Q_0}\, \mathrm{d}A\), and define \(L_0(z,w)\) to be the Bergman kernel for the Bergman space \(L^2_a(\mu _0)\). Let \(L=\lim L_{n_\ell }\) be a limiting holomorphic kernel at 0.
Recall that the kernel \(L_n\) is the reproducing kernel for a certain subspace \({{\mathcal {H}}}_n\) of \(L^2_a({\tilde{\mu }}_n)\), where \({\tilde{\mu }}_n\rightarrow \mu _0\) in the sense that the densities converge uniformly on compact sets, as \(n\rightarrow \infty \). See Lemma 2.2.
For \(L=\lim L_{n_\ell }\), the assignment \(\langle L_z,L_w\rangle _*=L(w,z)\) defines a positive semi-definite inner product on the linear span \({{\mathcal {M}}}\) of the \(L_z\)’s. In fact, the inner product is either trivial (\(L(z,z)=0\) for all z), or else it is positive definite: this holds by the zero-one law in Theorem 1.5, which will be proved in the next section.
By Fatou’s lemma, we now see that, for all choices of points \(z_j\) and scalars \(\alpha _j\),
This shows that \({{\mathcal {M}}}\) is contained in \(L^2(\mu _0)\) and that the inclusion \(I:{{\mathcal {M}}}\rightarrow L^2(\mu _0)\) is a contraction. Hence the completion \({{\mathcal {H}}}_*\) of \({{\mathcal {M}}}\) can be regarded as a contractively embedded subspace of \(L_a^2(\mu _0)\).
Since the space \(L^2_a(\mu _0)\) has reproducing kernel \(L_0(z,w)\), it follows from a theorem of Aronszajn ([8, p. 355]) that the difference \(L_0-L\) is a positive matrix. The proof of Theorem 1.7 is complete. \(\square \)
3 Ward’s Equation and the Zero-One Law
3.1 Ward’s Equation
Given a limiting kernel K in Lemma 2.2, we recall the definitions
The goal of this section is to prove Theorem 1.5, which we here restate in the following form (the case \(\tau _0=1\)).
Lemma 3.1
If R does not vanish identically, then \(R>0\) everywhere and we have
For the proof of Lemma 3.1, we recall the setting of Ward’s identity from [4].
For a test function \(\psi \in C_0^{\infty }({{\mathbb {C}}})\), we define a function \(W_{n}^{+}[\psi ]\) of n variables by
where
We now regard \(\zeta \) as picked randomly with respect to the Boltzmann–Gibbs distribution (1.1). \(W_n^+[\psi ]\) is then a random variable; the Ward identity proved in [4, Section 4.1] states that its expectation vanishes:
We shall now rescale in Ward’s identity about 0 at the microscopic scale \(r_n=n^{-1/2k}\), given that the basic decomposition \(Q=Q_0+Re H+Q_1\) in (1.6) holds. (We do not need to assume that 0 is in the bulk at this stage.)
To facilitate the calculations, it is convenient to recall a simple algebraic fact (see, e.g., [26]): if f is a function of p complex variables, and if \(f(\zeta _1,\ldots ,\zeta _p)\) is regarded as a random variable on the sample space \(\{\zeta _j\}_1^n\) with respect to the Boltzmann–Gibbs law, then the expectation is
where \(\mathrm{d}V_p(\zeta _1,\ldots ,\zeta _p)=\mathrm{d}A(\zeta _1)\cdots \mathrm{d}A(\zeta _p)\).
We rescale about 0 via \(z=r_n^{-1}\zeta \), \(w=r_n^{-1}\eta \), recalling that the p-point functions transform as densities. We recall that \(R_{n,p}(z)=r_n^{\,2p}\,{\mathbf {R}}_{n,p}(\zeta )\) denotes the rescaled p-point function and use the abbreviation \(R_n=R_{n,1}\) for the one-point function. We also write
Lemma 3.2
We have that
where \(o(1) \rightarrow 0\) uniformly on compact subsets of \({{\mathbb {C}}}\) as \(n\rightarrow \infty \).
Proof
We fix a test function \(\psi \in C_{0}^{\infty }({{\mathbb {C}}})\) and let \(\psi _n (\zeta ) = \psi (r_n^{-1} \zeta )\). The change of variables \(z=r_n^{-1}\zeta \) and \(w= r_n^{-1}\eta \) gives that
and
Likewise, changing variables and integrating by parts, we obtain
Hence, by the Ward identity in (3.1), we have
Since \(\psi \) is an arbitrary test function, we have in the sense of distributions,
Dividing through by \(R_{n,1}(z)\) and using the fact that
we obtain
Differentiating with respect to \({\bar{z}}\), we get
Since \(\Delta Q( r_n z) = r_n^{\,2(k-1)} \Delta Q_0(z) + O(r_n^{\,2k-1})\) uniformly on compact subsets of \({{\mathbb {C}}}\) as \(n \rightarrow \infty \) and \(r_n = n^{-1/2k}\), we obtain
where \(o(1) \rightarrow 0\) uniformly on compact subsets of \({{\mathbb {C}}}\) as \(n \rightarrow \infty \). \(\square \)
3.2 The Proof of Theorem 1.5
We will need a few lemmas.
Lemma 3.3
If \(R(z_0)=0\), then there is a real analytic function \({\tilde{R}}\) such that
If R does not vanish identically, then all zeros of R are isolated.
Proof
The assumption gives that the holomorphic kernel L corresponding to R satisfies \(L(z_0, z_0)=0\). Hence \(\int \mathrm{e}^{-Q_0(w)}\,|\,L(z_0,w)\,|^{\,2} \mathrm{d}A(w) \le 0\) by the mass-one inequality (2.5). Thus \(L(z_0,w)=0\) for all \(w\in {{\mathbb {C}}}\). Since L is Hermitian-entire, we can thus write
for some Hermitian-entire function \({\tilde{L}}\). We now have \(R(z)=\left| {\,z-z_0\,} \right| ^{\,2} {\tilde{L}}(z,z)\mathrm{e}^{-Q_0(z)}\).
For the second statement, we assume that R does not vanish identically and there exists a zero \(z_0\) of R which is not isolated. Then, we can take a sequence \(\{z_j\}_{1}^{\infty }\) of distinct zeros of R which converges to \(z_0\), whence by the above argument, for each j we obtain \(L(z_j, w)=0\) for all \(w \in {{\mathbb {C}}}\). If we fix w, then \(L(z,w)=0\) for all \(z\in {{\mathbb {C}}}\) since L(z, w) is holomorphic in z. Hence \(L=0\) identically. \(\square \)
Lemma 3.4
L(z, w) is a positive matrix and \(z \mapsto L(z,z)\) is logarithmically subharmonic.
Proof
It is clear that L is a positive matrix. Now write \(L_z(w) : = L(w,z)\) and define a semi-definite inner product by \(\langle L_z, L_w \rangle _{*} :=L(w,z)\) on the linear span of the functions \(L_z\) for \(z\in {{\mathbb {C}}}\). The completion of this span forms a (perhaps semi-normed) Hilbert space \({{\mathcal {H}}}_{*}\), and L is a reproducing kernel of the space. Now when \(L(z,z)>0\),
Since L(z, w) is Hermitian-entire, we have \({\bar{\partial }}_{z}L_z \in {{\mathcal {H}}}_{*}\), \(\langle {\bar{\partial }}_z L_z, L_z \rangle _{*} = {\bar{\partial }}_z L(z,z)\), and \(\langle {\bar{\partial }}_z L_z , {\bar{\partial }}_z L_z \rangle _{*}= \Delta L(z,z)\). Hence, the numerator of (3.4) can be written as
which is nonnegative by the Cauchy–Schwarz inequality.
At points where \(L(z,z)=0\), \(\log L(z,z)\) satisfies the sub-mean value property since \(\log L(z,z)=-\infty \). Hence the function \(\log L(z,z)\) is subharmonic on \({{\mathbb {C}}}\). \(\square \)
Lemma 3.5
If \(R(z_0)=0\) and \(R(z)=\left| {\,z-z_0\,} \right| ^{\,2}{\tilde{R}}(z)\), then \(\Delta Q_0 + \Delta \log {\tilde{R}} \ge 0\) in a neighborhood of \(z_0\).
Proof
We choose a small disc \(D=D(z_0, \epsilon )\) and consider the function
Observing that \(\Delta _z \log L(z,z) = \Delta Q_0(z) + \Delta \log {\tilde{R}}(z) + \delta _{z_0}\) in the sense of distributions, Lemma 3.4 gives us that \(\Delta S \ge 0\) in the sense of distributions on \(D\backslash \{z_0\}\). If \({\tilde{R}}(z_0)>0\), we extend S analytically to \(z_0\). On the other hand, if \({\tilde{R}}(z_0)=0\), we define \(S(z_0)=-\infty \). In both cases, the extended function S is subharmonic on D. \(\square \)
We now turn to the left-hand side in the rescaled version of Ward’s identity, namely the function \({\bar{\partial }}C_n\), where \(C_n\) is the Cauchy transform of \(B_n\) (see (3.3)).
Lemma 3.6
Suppose that \(R=\lim R_{n_\ell }\) is a limiting 1-point function that does not vanish identically. Let Z be the set of isolated zeros of R, and let \(B(z,w)=\lim B_{n_\ell }(z,w)\) be the corresponding Berezin kernel for \(z\not \in Z\). Then \(C_{n_\ell } \rightarrow C\) locally uniformly on the complement \(Z^{c}={{\mathbb {C}}}\setminus Z\) as \(\ell \rightarrow \infty \), where the function
is bounded on \(Z^{c} \cap V\) for each compact subset \(V\) of \({{\mathbb {C}}}\).
Proof
We have that \(c_{n_\ell }K_{n_\ell } \rightarrow K\) locally uniformly on \({{\mathbb {C}}}^2\), where \(K(z,z)=R(z)>0\) when \(z\not \in Z\). Hence, for fixed \(\epsilon \) with \(0<\epsilon <1\), we can choose N such that if \(\ell \ge N\), then
for all z, w with \(\left| {\,z\,} \right| \le 1/\epsilon \), \(\left| {\,w\,} \right| \le 2/\epsilon \), and \({\text {dist}}(z, Z) \ge \epsilon \). Then, for z with \(\left| {\,z\,} \right| \le 1/\epsilon \) and \({\text {dist}}(z, Z) \ge \epsilon \),
Here, we have used the mass-one inequality for the third inequality. Thus \(C_{n_\ell } \rightarrow C\) uniformly on compact subsets of \(Z^c\).
Now fix a compact subset \(V\) of \({{\mathbb {C}}}\). Then, for all z, w with \(z \in V\setminus Z\) and \({\text {dist}}(w,V)\le 1\),
for some \(M=M_V\) that depends only on \(V\) by Lemma 2.1. Thus, for \(z \in V\setminus Z\),
Hence we obtain \(\left| {\,C(z)\,} \right| \le 2M+1\) for \(z \in V\setminus Z\). \(\square \)
Lemma 3.7
If R does not vanish identically, the Ward’s equation
holds in the sense of distributions.
Proof
The preceding lemmas show that
and that a subsequence \(C_{n_\ell }\) converges to C boundedly and locally uniformly on \({{\mathbb {C}}}\setminus Z\). Since \(Z\cap V\) is a finite set for each compact set \(V\), it follows that \(C_{n_\ell }\rightarrow C\) in the sense of distributions, and hence \({\bar{\partial }}C_{n_\ell }\rightarrow {\bar{\partial }}C\). By Ward’s equation and the locally uniform convergence \(R_{n_\ell }\rightarrow R\), it then follows that \(\Delta \log R_{n_\ell }\rightarrow \Delta \log R\) in the sense of distributions. We can thus pass to the limit as \(n_\ell \rightarrow \infty \) in the rescaled Ward identity (3.5). \(\square \)
Proof of Theorem 1.5
We follow the strategy in [4, Theorem 4.8]. Suppose that \(R(z_0)=0\). We must prove that \(R=0\) identically.
Let D be a small disk centered at \(z_0\), and write \(\chi =\chi _D\) for the characteristic function. Also write \(R(z)=\left| {\,z-z_0\,} \right| ^{\,2}{\tilde{R}}(z)\).
Consider the measures \(\mu =\chi \cdot (\Delta Q_0+\Delta \log R)\) and \(\nu =\chi \cdot (\Delta Q_0+\Delta \log {\tilde{R}})\). By lemmas 3.4 and 3.5, these measures are positive, and \(\mu =\delta _{z_0}+\nu \). Write \(C^\mu (z)=\int _{{\mathbb {C}}}\frac{1}{z-w}\, d\mu (w)\) for the Cauchy transform of \(\mu \). Clearly,
Also \({\bar{\partial }}C^\nu =\nu \ge 0\). When \(z\in D\), the right-hand side in Ward’s equation equals \(R(z)-\Delta (Q_0+\log R)(z)=R(z)-{\bar{\partial }}C^\mu (z).\) If \(C(z)=\int \frac{B(z,w)}{z-w}\, \mathrm{d}A(w)\), we have, by Ward’s equation, that
Hence, by Weyl’s lemma, \(C(z)=-1/(z-z_0)-C^\nu (z)+v(z)\), where v is smooth near \(z_0\). If \(C^\mu (z)\) were bounded as \(z\rightarrow z_0\), then the measure \(\mu =\nu +\delta _{z_0}\) would place no mass at \(\{z_0\}\), so \(\nu =-\delta _{z_0}+\rho \), where \(\rho (\{z_0\})=0\). This contradicts that \(\nu \ge 0\). The contradiction shows that \(\left| {\,C(z)\,} \right| \rightarrow \infty \) as \(z\rightarrow z_0\). This in turn contradicts that C is bounded (Lemma 3.6), and hence \(R(z_0)=0\) is impossible. Hence \(\Delta \log R\) is a smooth function on \({{\mathbb {C}}}\). Applying Weyl’s lemma to the distributional Ward equation \({\bar{\partial }}C=R-\Delta Q_0-\Delta \log R\) now shows that C(z) is smooth and hence that the equation holds pointwise on \({{\mathbb {C}}}\). \(\square \)
4 Universality Results
In this section, we prove Theorems 1.3 and 1.4. The proof of Theorem 1.4 relies on certain apriori estimates, whose proofs are postponed to Sect. 6.
4.1 Homogeneous Singularities
Assume that Q has a homogeneous singularity of type \(2k-2\) at the origin, i.e., that the canonical decomposition is of the form \(Q=Q_0+Re\, H,\, H=c\,\zeta ^{\,2k}\), where \(Q_0\) is positively homogeneous of degree 2k. As always, we write \(\mu _0\) for the measure \(d\mu _0=\mathrm{e}^{-Q_0}\, \mathrm{d}A\).
We now recall the kernel \(L_n\) (defined in (2.3)):
In the present case, \(L_n(z,w)=k_n(z,w)\mathrm{e}^{-c\,z^{\,2k}/2-{\bar{c}}\,{\bar{w}}^{\,2k}/2}.\) By Lemma 2.2, \(L_n\) is the reproducing kernel for the space
regarded as a subspace of \(L^2(\mu _0)\). (This is because \({\tilde{\mu }}_n=\mu _0\) for the measure \({\tilde{\mu }}_n\) in (2.4).)
Since the spaces \({{\mathcal {H}}}_n\) are increasing, \({{\mathcal {H}}}_n\subset {{\mathcal {H}}}_{n+1},\) where the inclusions are isometric, it follows that a unique limiting holomorphic kernel \(L=\lim L_{n}\) exists. By Theorem 1.7, the kernel L is the reproducing kernel for a contractively embedded subspace \({{\mathcal {H}}}_*\) of \(L^2_a(\mu _0)\), which must contain the dense subset \(U= \bigcup {{\mathcal {H}}}_n\). Furthermore, by the reproducing property of \(L_n\), we have for each element \(f(z)=q(z)\cdot \mathrm{e}^{-cz^{2k}/2}\in U\) that \(\langle f,L_{n,z}\rangle _{L^2(\mu _0)}=f(z)\), whenever \(n>{\text {degree}}q\). It follows that
Since U is dense in \(L^2_a(\mu _0)\), L must equal to the reproducing kernel \(L_0\) of \(L^2_a(\mu _0)\). The proof of Theorem 1.3 is complete. \(\square \)
4.2 Rotational Symmetry
Referring to the canonical decomposition \(Q=Q_0+Re\, H+Q_1\), we now suppose that \(Q_0(z)=Q_0(|\,z\,|)\), and we fix a rotationally symmetric limiting holomorphic kernel
Writing \(E(z)=\sum _0^\infty a_jz^j\), the mass-one inequality
is seen to be equivalent to that
To use Ward’s equation, we first compute the Cauchy transform C(z) as follows:
However, as is shown in [3], we have that
Thus
where
The term A(z) can be written as
which gives
The term B(z) is computed as follows:
Noting that
we infer that Ward’s equation
is equivalent to that \({\bar{\partial }}(B-{\partial }_z\log L(z,z))=0\). This in turn, is equivalent to that the function
be entire. It is easy to check that this is the case if and only if all coefficients in the sum vanish, that is, if and only if for each \(k\ge 1\), we have that
We now apply the growth estimate in Theorem 1.6, part (ii), which says that
We claim that this implies the second alternative in (4.2).
Indeed, (4.3) is clearly not satisfied if E is constant. Next note that the mass-one inequality (4.1) and the zero-one law (Theorem 1.5) imply that \(0<a_0 \le 1/\Vert \,1\,\Vert ^{\,2}_{L^2(\mu _0)}\). Since E(z) is not a polynomial by (4.3), for any k there exists \(N \ge k\) such that \(a_N \ne 0\). By (4.2), we obtain that if \(a_N \ne 0\) but \(a_j=0\) for all j with \(1\le j \le N-1\), then \(N=1\) and \(a_0 = 1/\Vert \,1\,\Vert ^{\,2}_{L^2(\mu _0)}\). By a simple induction, we then have \(a_k=1/\Vert \,z^{\,k}\,\Vert _{L^2(\mu _0)}^{\,2}\) for all \(k \ge 0\). Thus, we have
Since the polynomial \(\phi _j(z)=z^{\,j}/\Vert \,z^{\,j}\,\Vert _{L^2(\mu _0)}\) is the j:th orthonormal polynomial with respect to the measure \(\mu _0\), we have
where \(L_0\) is the Bergman kernel for the space \(L^2_a(\mu _0).\) The proof is complete. \(\square \)
5 Asymptotics for \(L_0(z,z)\)
In this section, we prove part (i) of Theorem 1.6.
To this end, let \(A_0(z,w)\) be the Hermitian polynomial such that \(A_0(z,z)=Q_0(z)\), and write
We write \(L_z^\sharp (w)\) for \(L_0^\sharp (w,z)\) and, for suitable functions u,
Below, we fix a z with |z| large enough; we must estimate \(L_0(z,z)\). We also fix a number \(\delta _0=\delta _0(z)>0\) and write \(\chi _z\) for a fixed \(C^\infty \)-smooth test-function with \(\chi _z(w)=1\) when \(|\,w-z\,|\le \delta _0\) and \(\chi _z(w)=0\) when \(|\,w-z\,|\ge 2\delta _0\).
We will use the following estimate.
Lemma 5.1
If \(|1-w/z|\) is sufficiently small, then \(2Re\, A_0(z,w)\le Q_0(z)+Q_0(w)-c|z|^{2k-2}|w-z|^2\), where c is a positive constant.
Proof
We write \(h=w-z\). By Taylor’s formula, \(A_0(w,z)=Q_0(z)+\sum _1^{2k}\frac{{\partial }^jQ_0(z)}{j!}h^j\). Similarly, \(A_0(w,w)=Q_0(z)+\sum _{i+j\ge 1}\frac{{\partial }^i{\bar{\partial }}^j Q_0(z)}{i!j!}h^i{\bar{h}}^j.\) Hence
However, since \(Q_0\) is homogeneous of degree 2k, the derivative \({\partial }^i{\bar{\partial }}^j Q_0\) is homogeneous of degree \(2k-i-j\). Hence
Thus, if \(i+j\ge 3\) and \(|1-w/z|\) is sufficiently small, then the left-hand side in (5.1) is dominated by an arbitrarily small multiple of \(|z|^{2k-2}|z-w|^2\). On the other hand, by homogeneity and positive definiteness of \(\Delta Q_0\), we have that \(\Delta Q_0(z)|z-w|^2\ge c'|z|^{2k-2}|z-w|^2\), where \(c'\) is a positive constant. The lemma thus follows with any positive constant \(c<c'\) \(\square \)
As always, we write \(d\mu _0=\mathrm{e}^{-Q_0}\, \mathrm{d}A\); \(L^2_a(\mu _0)\) denotes the associated Bergman space of entire functions, and \(L_0\) is the Bergman kernel of that space.
Lemma 5.2
Let \(|z|\ge 1\) and \(\delta _0\) be a positive number with \(\delta _0/|z|\) sufficiently small. Then there is a constant \(C=C(\delta _0)\) such that, for all functions \(u\in L^2_a(\mu _0)\),
Proof
Note that
where
Now fix w. The denominator \(P(z)={\bar{\partial }}_2A_0(w,w)-{\bar{\partial }}_2A_0(z,w)\) is by Taylor’s formula equal to the polynomial
Here the derivative \({\partial }^j\Delta Q_0(w)=\left| {\,w\,} \right| ^{\,2k-2-j} {\partial }^j\Delta Q_0(w/|\,w\,|)\) is positively homogeneous of degree \(2k-2-j\). Set \(c(w)=\Delta Q_0(w/|\,w\,|)\). We then have that
Since also \({\partial }_1{\bar{\partial }}_2A_0(z,w)=c(z)\left| {\,z\,} \right| ^{\,2k-2}(1+O(w-z))\), we have by (5.3),
By the form of F it is also clear that
An integration by parts in (5.2) gives \(\pi ^\sharp [\chi _zu](z)=u(z)+\epsilon _1+\epsilon _2\), where
Inserting the estimates (5.4) and (5.5), using also that \({\bar{\partial }}\chi _z(w)=0\) when \(|\,w-z\,|\le \delta _0\), we find that
To estimate \(\epsilon _1\), we use Lemma 5.1 to get
This gives
To estimate \(\epsilon _2\), we note that (again by (5.6))
The proof is complete. \(\square \)
Let \(\pi _0:L^2(\mu _0)\rightarrow L^2_a(\mu _0)\) be the Bergman projection, \(\pi _0[f](z)=\langle f,L_z\rangle _{L^2(\mu _0)}\), where we write \(L_z(w)\) for \(L_0(w,z)\). Noting that
we see that
If we now choose \(u=L_z\) in Lemma 5.2 and recall that \(\Vert \,L_z\,\Vert _{L^2(\mu _0)}^{\,2}=L_0(z,z)\), we obtain the estimate
Lemma 5.3
There is a constant C such that for all \(|z|\ge 1\) and all \(\delta _0=\delta _0(z)>0\) with \(\delta _0/|z|\) small enough,
Proof
Consider the function \(u_0=\chi _zL_z^\sharp -\pi _0[\chi _zL_z^\sharp ].\) This is the norm-minimal solution in \(L^2(\mu _0)\) to the problem \({\bar{\partial }}u=({\bar{\partial }}\chi _z)\cdot L_z^\sharp \).
Since \(Q_0\) is strictly subharmonic on the support of \(\chi _z\), we can apply the standard Hörmander estimate (e.g., [19, p. 250]) to obtain
where we used homogeneity of \(\Delta Q_0\).
By Taylor’s formula and the estimate (5.6), we have when \(\delta _0\le \left| {\,w-z\,} \right| \le 2\delta _0\),
By the homogeneity of \(\Delta Q_0\), we thus obtain the estimate
We now pick another (small) number \(\delta >0\) and invoke the following pointwise-\(L^2\) estimate (see, e.g., [4, Lemma 3.1] or the proof of the inequality (2.2)):
Combining with (5.8), this gives
Choosing \(\delta _0\) a small multiple of \(|z|^{1/2}\) and then \(\delta \) small enough, we insure that the right-hand side is dominated by \(C|z|^{2k-2}\mathrm{e}^{Q_0(z)}\), as desired. \(\square \)
Proof of Part (i) of Theorem 1.6
By the estimate (5.7) and Lemma 5.3, we have
Writing \(R_0(z)=L_0(z,z)\mathrm{e}^{-Q_0(z)}\), this becomes
We must prove that the left-hand side in (5.10) is dominated by \(M\left| {\,z\,} \right| ^{\,1-k}\Delta Q_0(z)\) for all large \(|\,z\,|\), where M is a suitable constant. If this is false, there are two possibilities. If \(R_0(z)\le (1-M|\,z\,|^{\,1-k})\Delta Q_0(z)\) for arbitrarily large \(|\,z\,|\), then (5.10) implies
and we reach a contradiction for large enough M.
In the remaining case, we have \(R_0(z)\ge (1+M|\,z\,|^{\,1-k})\Delta Q_0(z)\). Then (5.10) gives the estimate \(R_0(z)\ge cM^{\,2}|\,z\,|^{\,2k-2}\) for some \(c>0\). Since \(\Delta Q_0(z) \le c'|\,z\,|^{\,2k-2}\) for some \(c'>0\), we obtain
Choosing M large enough, we obtain \(R_0(z) \ge C_3 M \left| {\,z\,} \right| ^{\,4k-4}\) by (5.10) again. Repeating the above argument gives \(R_0(z) \ge C_p M \left| {\,z\,} \right| ^{\,2p}\) for all sufficiently large \(|\,z\,|\) for some constant \(C_p>0\). On the other hand, we will show that
for all z, which will give the desired contradiction. To see this, note that for functions \(u\in L_a^2(\mu _0)\), the estimate (5.9) gives
Taking \(\delta =|\,z\,|^{1-2k}\), we obtain \(\left| {u(z)} \right| ^2\le C|\,z\,|^{4k-2}\mathrm{e}^{Q_0(z)} \Vert u\Vert _{L^2(\mu _0)}^2\). Since
we now obtain the estimate (5.11). \(\square \)
6 Apriori Estimates for the One-Point Function
In this section, we prove part (ii) of Theorem 1.6.
As before, we write \(Q=Q_0+Re\, H+Q_1\) for the canonical decomposition of Q at 0, and we write \(\mu _0\) for the measure \(d\mu _0=\mathrm{e}^{-Q_0}\, \mathrm{d}A\). In this section, the assumption that 0 is in the bulk of the droplet will become important.
Our arguments below essentially follow by adaptation of the previous section.
Fix a point \(\zeta \) in a small neighborhood of 0 with \(\left| {\,\zeta \,} \right| \ge r_n\). We also fix a number \(\delta _0=\delta _0(\zeta )\ge \mathrm {const.}>0\) with \(\delta _0(\zeta )\cdot r_n/|\zeta |\) uniformly small, and a smooth function \(\psi \) with \(\psi =1\) in \(D(0,\delta _0)\) and \(\psi =0\) outside \(D(0,2\delta _0)\). We define a function \(\chi _\zeta =\chi _{\zeta ,n}\) by
Let \(A(\eta ,\omega )\) be a Hermitian-analytic function in a neighborhood of (0, 0), satisfying \(A(\eta ,\eta )=Q(\eta )\). We shall essentially apply the definition of the approximating kernel (denoted by \(L_0^\sharp \) in the preceding section) with “\(A_0\)” replaced by “nA.” We denote this kernel by \({{\mathbf {L}}}_{n}^\sharp \), viz.
The corresponding “approximate projection” is defined on suitable functions u by
where, for convenience, we write \({{\mathbf {L}}}^\sharp _\zeta \) instead of \({{\mathbf {L}}}^\sharp _{n,\zeta }\).
Lemma 6.1
Suppose that u is holomorphic in a neighborhood of \(\zeta \) and \(\delta _0(\zeta )\cdot r_n/|\zeta |\le {\varepsilon }_0\) (small enough). Then there is a constant \(C=C({\varepsilon }_0)\) such that, when \(r_n\le |\,\zeta \,|\le r_n\log n\),
Proof
It will be sufficient to indicate how the proof of Lemma 5.2 is modified in the present setting. We start as earlier, by writing
where
Here, we may replace “A” by “\(A_0\)” to within negligible terms, for the relevant \(\zeta \) and \(\omega \). More precisely, Taylor’s formula gives that
when \(r_n\le |\,\zeta \,|\le r_n\log n\) and \(\left| {\,\omega -\zeta \,} \right| \le 2\delta _0 r_n\).
From (6.1) and the form of F, we see (as in the proof of Lemma 5.2) that
We continue to write \(\pi _n^\sharp u(\zeta )=u(\zeta )+\epsilon _1+\epsilon _2\), where
To estimate \(\epsilon _1\) and \(\epsilon _2\), we note that there is a positive constant c such that
See Lemma 5.1.
Inserting the estimates in (6.3) and (6.4), using also that \({\bar{\partial }}\chi _\zeta (w)=0\) when \(\left| {\,\zeta -\omega \,} \right| \le \delta _0 r_n\), we find that if \(|\zeta |\ge r_n\),
Using the Cauchy–Schwarz inequality, we find now that
The proof is complete. \(\square \)
Choosing \(u(\eta )={{\mathbf {k}}}_{n}(\eta ,\zeta )\), where \({{\mathbf {k}}}_n\) is the Bergman kernel for the subspace \({{\mathcal {P}}}_n\) of \(L^2(\mu _n)\), we obtain the following estimate, valid when \(r_n \le \left| {\,\zeta \,} \right| \le r_n\log n\):
Here \(\pi _n:L^2(\mu _n)\rightarrow {{\mathcal {P}}}_n\) is the orthogonal projection, \(\pi _n u(\zeta )=\langle u,{{\mathbf {k}}}_{n,\zeta }\rangle _{L^2(\mu _n)}\). (Cf. (5.7) for details on the derivation of equation (6.5) from Lemma 6.1.)
Lemma 6.2
For all \(\zeta \) in the annulus \(r_n\le \left| {\,\zeta \,} \right| \le \log n \cdot r_n\), and for \(\delta _0(\zeta )\cdot r_n\) a small enough multiple of \(|\zeta |\), we have the estimate
Proof
Let \(u_0=\chi _\zeta {{\mathbf {L}}}_\zeta ^\sharp -\pi _n\left[ \chi _\zeta {{\mathbf {L}}}_\zeta ^\sharp \right] \) be the norm-minimal solution in \(L^2(\mu _n)\) to the problem \({\bar{\partial }}u_0={\bar{\partial }}f\), where \(f=\chi _\zeta {{\mathbf {L}}}_\zeta ^\sharp \). We will prove that the problem \({\bar{\partial }}u={\bar{\partial }}f\) has a solution u with \(u-f\in {\text {Pol}}(n)\) and
This is done by a standard device, which now we briefly recall.
Let \({\check{Q}}\) be the “obstacle function” pertaining to Q. The main facts about this function to be used here are the following (cf. [28] for details). The obstacle function can be defined as \({\check{Q}}=\gamma -2U^\sigma \), where \(U^\sigma \) is the logarithmic potential of the equilibrium measure and \(\gamma \) is a constant chosen so that \({\check{Q}}=Q\) on S. One has that \({\check{Q}}\) is harmonic outside S, and that its gradient is Lipschitz continuous on \({{\mathbb {C}}}\). Furthermore, \({\check{Q}}(\omega )\) grows like \(2\log \left| {\,\omega \,} \right| +O(1)\) as \(\omega \rightarrow \infty \).
We use the obstacle function to form the strictly subharmonic function \(\phi (\omega )={\check{Q}}(\omega )+n^{-1}\log (1+\left| {\,\omega \,} \right| ^{\,2})\), and we go on to define a measure \(\mu _n'\) by \(d\mu _n'(\omega )=\mathrm{e}^{-n\phi (\omega )}\, \mathrm{d}A(\omega )\). Write \({{\mathcal {P}}}_n'\) for the subspace of \(L^2(\mu _n')\) of holomorphic polynomials of degree at most \(n-1\), and let \(\pi _n'\) be the corresponding orthogonal projection. Finally, we write
Since \(\phi \) is now strictly subharmonic, the standard Hörmander estimate can be applied. It gives
Since \(\chi _\zeta \) is supported in the disk \(D(\zeta ,2\delta _0 r_n)\), and since \(\Delta {\check{Q}}=\Delta Q=\Delta Q_0\cdot (1+o(1))\) there, we see that
Next we use the estimate \(n\phi \le nQ+\mathrm {const.}\) which holds by the growth assumption on Q near infinity. This gives \(\Vert \,v_0\,\Vert _{L^2(\mu _n)}\le C\Vert \,v_0\,\Vert _{L^2(\mu _n')}\), and so we have shown (6.6) with \(u=v_0\).
Since \(n\phi (\omega )=(n+1)\log \left| {\,\omega \,} \right| ^{\,2}+O(1)\) as \(\omega \rightarrow \infty \), we have that \(L^2_a(\mu _n')={\text {Pol}}(n)\). Hence \(u=v_0\) solves, in addition to (6.6), the problem
Using the form of \({\bar{\partial }}f={\bar{\partial }}\chi _\zeta \cdot {{\mathbf {L}}}_\zeta ^\sharp \) and the estimate (6.4), we find that for \(|\omega -\zeta |\le \delta _0 r_n\),
By the homogeneity of \(\Delta Q_0\) and the fact that \({\bar{\partial }}\chi _\zeta =0\) when \(|\omega -\zeta |\le \delta _0r_n\), this gives the estimate
Applying (6.6), we now get
We now pick a small constant \(\delta \) (independent of n) and use the pointwise-\(L^2\) estimate
Choosing \(\delta _0r_n\) as a small multiple of \(|\zeta |\) and then \(\delta \) small enough, we can now use (6.7) to deduce that
finishing the proof. \(\square \)
Proof of Theorem 1.6, part (ii)
Fix \({\varepsilon }>0\) and take \(\zeta \) with \(r_n\le \left| {\,\zeta \,} \right| \le \log n\cdot r_n\). By the estimate (6.5) and Lemma 6.2, we have for all large n that
for some constants \(C_1,C_2\). Multiplying through by \(r_n^{\,2}\) and writing \(R_n(z)=r_n^{\,2}\,{\mathbf {R}}_n(\zeta )\), \(z=r_n^{-1}\zeta \), we get
It follows that each limiting 1-point function R must satisfy
where \(c(z)=\Delta Q_0(z/|\,z\,|)>0\). The proof of part (i) of Theorem 1.6 shows that this is only possible if \(R(z)=\Delta Q_0(z)(1+O(z^{1-k}))\) as \(z\rightarrow \infty \). \(\square \)
References
Ameur, Y.: A density theorem for weighted Fekete sets. Int. Math. Res. Notices (2016). doi:10.1093/imrn/rnw161
Ameur, Y., Hedenmalm, H., Makarov, N.: Ward identities and random normal matrices. Ann. Probab. 43 (2015), 1157–1201. Cf. arxiv:1109.5941v3 for a different version
Ameur, Y., Kang, N.-G.: On a problem for Ward’s equation with a Mittag–Leffler potential. Bull. Sci. Math. 137, 968–975 (2013)
Ameur, Y., Kang, N.-G., Makarov, N.: Rescaling Ward identities in the random normal matrix model. Arxiv preprint, v4. 2015
Ameur, Y., Kang, N.-G., Makarov, N., Wennman, A.: Scaling limits of random normal matrix processes at singular boundary points, arXiv:1510.08723
Ameur, Y., Kang, N.-G., Seo, S.-M.: In preparation
Ameur, Y., Seo, S.-M.: Microscopic densities and Fock–Sobolev spaces. arXiv:1610.10052
Aronszajn, N.: Theory of reproducing kernels. Trans. Am. Math. Soc. 68, 337–404 (1950)
Balogh, F., Grava, T., Merzi, D.: Orthogonal polynomials for a class of measures with discrete rotational symmetries in the complex plane. doi:10.1007/s00365-016-9356-0 (To appear in Constr. Approx.)
Bleher, P., Eynard, B.: Double scaling limit in random matrix models and a nonlinear hierarchy of differential equations. J. Phys. A.: Math. Gen. 36, 3085–3105 (2003)
Burda, Z., Nowak, M.A., Waclaw, B.: Spectrum of the product of independent random Gaussian matrices. Phys. Rev. E 81, 041132 (2010)
Can, T., Forrester, P.J., Téllez, G., Wiegmann, P.: Singular behavior at the edge of Laughlin states. Phys. Rev. B 89, 235137 (2014)
Can, T., Laskin, M., Wiegmann, P.B.: Geometry of quantum Hall states: gravitational anomaly and transport coefficients. Ann. Phys. 362, 752–794 (2015)
Claeys, T.: The birth of a cut in unitary random matrix ensembles. Int. Math. Res. Notices (2008). doi:10.1093/imrn/rnm166
Claeys, T., Kuijlaars, A.B.V., Vanlessen, M.: Multi-critical unitary random matrix ensembles and the general Painlevé II equation. Ann. Math. 167, 601–641 (2008)
Forrester, P.J.: Log-Gases and Random Matrices. Princeton University Press, Princeton (2010)
Führ, H., Gröchenig, K., Haimi, A., Klotz, A., Romero, J. L.: Density of sampling and interpolation in reproducing kernel Hilbert spaces. arXiv:1607.07803
Gorenflo, R., Kilbas, A.A., Mainardi, F., Rogosin, S.V.: Mittag–Leffler Functions, Related Topics and Applications. Springer, Berlin (2014)
Hörmander, L.: Notions of Convexity. Birkhäuser, Boston (1994)
Jancovici, B.: Exact results for the two-dimensional one-component plasma. Phys. Rev. Lett. 46, 386–388 (1981)
Kuijlaars, A.B.J., McLaughlin, K.T.R.: Generic behavior of the density of states in random matrix theory and equilibrium problems in the presence of real analytic external fields. Commun. Pure Appl. Math. 53, 736–785 (2000)
Laskin, M., Chiu, Y.H., Can, T., Wiegmann, P.: Emergent Conformal Symmetry of Quantum Hall States on Singular surfaces. Phys. Rev. Lett. 117, 266803 (2016)
Lee, S.-Y., Yang, M.: Discontinuity in the asymptotic behavior of planar orthogonal polynomials under a perturbation of the Gaussian weight. arXiv:1607.02821
Lindholm, N.: Sampling of weighted \(L^p\)-spaces of entire functions in \(\mathbb{C}^n\) and estimates of the Bergman kernel. J. Funct. Anal. 182, 390–426 (2001)
Marco, N., Massaneda, X., Ortega-Cerdà, J.: Interpolation and sampling sequences for entire functions. Geom. Funct. Anal. 13, 862–914 (2003)
Mehta, M.L.: Random Matrices, 3rd edn. Academic Press, San Diego (2004)
Pastur, L., Shcherbina, M.: Eigenvalue distribution of large random matrices. Math. Surv. Monogr. 171, AMS (2011)
Saff, E.B., Totik, V.: Logarithmic Potentials with External Fields. Springer, Berlin (1997)
Soshnikov, A.: Determinantal random point fields. Russ. Math. Surv. 55, 923–975 (2000)
Veneziani, A.M., Pereira, T., Marchetti, D.H.U.: Conformal deformation of equilibrium measures in normal random ensembles. J. Phys. A 44, 075202 (2011)
Zhu, K.: Analysis on Fock Spaces. Springer, Berlin (2012)
Acknowledgements
For helpful communication: J. O. Lee, S.-Y. Lee, K. Liechty, J. Ortega-Cerdá, J. L. Romero, F. Tellander, A. Wennman, P. Wiegmann.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Peter J. Forrester.
Seong-Mi Seo was supported by Samsung Science and Technology Foundation, SSTF-BA1401-01.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Ameur, Y., Seo, SM. On Bulk Singularities in the Random Normal Matrix Model. Constr Approx 47, 3–37 (2018). https://doi.org/10.1007/s00365-017-9368-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00365-017-9368-4