Fourier dimension of imaginary Gaussian Multiplicative Chaos
Benjamin Bonnefont, Hermanni Rajamäki, Vincent Vargas · 2025
Read Fourier dimension of imaginary Gaussian Multiplicative Chaos by Benjamin Bonnefont, Hermanni Rajamäki, Vincent Vargas (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-19441.txt.
Introduction
Benjamin Bonnefont, Hermanni Rajamäki, and Vincent Vargas measure the Fourier dimension of imaginary Gaussian multiplicative chaos on the unit circle. The chaos is a random, complex-valued distribution built by exponentiating a log-correlated Gaussian field with an imaginary factor. Its dimension captures how fast the average squared Fourier coefficients fade at high frequency.
In the subcritical phase, they prove the dimension is almost surely one minus the square of the coupling. The result covers a broad class of near-logarithmic fields. For the exact field, the chaos just misses a borderline Sobolev space, and rescaled coefficients settle into a complex Gaussian law, jointly independent across neighbors.
Catalog record
| CREATOR | Benjamin Bonnefont, Hermanni Rajamäki, Vincent Vargas |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 76 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-19441
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Fourier dimension of imaginary Gaussian Multiplicative Chaos Benjamin Bonnefont, Hermanni Rajamäki, Vincent Vargas Abstract . We study the Fourier coefficients of imaginary Gaussian multiplicative chaos (GMC) on the unit circle. Under the subcritical phase β ∈ ( 0 , 1 ) \beta\in(0,1) , we show that the Fourier dimension is 1 − β 2 1-\beta^{2} and prove a central limit theorem for the rescaled coefficients. 1. Introduction Gaussian multiplicative chaos (GMC) is a theory of random measures obtained by exponentiating log-correlated Gaussian fields. It was introduced by Kahane [ 10 ] , following earlier ideas of Mandelbrot [ 16 ] on turbulence and Høegh-Krohn [ 8 ] in quantum field theory. Over the last decades, GMC has become a central object in probability and mathematical physics, with connections to turbulence, finance, conformal field theory, random matrix theory and related areas. Since log-correlated fields are random distributions rather than functions, the exponential must be defined through a regularization and renormalization procedure, which yields a canonical multifractal object. Understanding its basic properties remains challenging in general dimensions. In this article we focus on the one-dimensional setting and consider GMC on the unit circle \mathbb{T} . In this case, GMC exhibits numerous symmetries that one can exploit in its study. The underlying field is the trace of the two-dimensional Gaussian free field (GFF) on \mathbb{T} with vanishing mean. More precisely, let (X_{\theta})_{\theta\in\mathbb{T}} be the centered Gaussian field with covariance \mathbb{E}\mathopen{}\mathclose{{\left[X_{\theta}X_{\theta^{\prime}}}}\right]=\log\frac{1}{\mathopen{}\mathclose{{\left|e^{i\theta}-e^{i\theta^{\prime}}}}\right|}, understood as a random distribution on \mathbb{T} . Given a mollification X^{(\varepsilon)} of X and a parameter \gamma\in\mathbb{R} , the GMC measure is defined by the limit \mathrm{M}_{\gamma}\coloneqq\lim_{\varepsilon\to 0}\,\mathrm{M}_{\gamma}^{(\varepsilon)},\qquad\mathrm{M}_{\gamma}^{(\varepsilon)}(d\theta)\coloneqq\exp\mathopen{}\mathclose{{\left(\gamma X^{(\varepsilon)}_{\theta}-\frac{\gamma^{2}}{2}\mathbb{E}\mathopen{}\mathclose{{\left[X^{(\varepsilon)2}_{\theta}}}\right]}}\right)\,\,\mathrm{d}\theta, whenever the limit exists (typically in probability in the space of Radon measures). In the present setting, the limit yields a non-trivial random measure for |\gamma|<\sqrt{2} , the so-called subcritical phase. Moreover, it is almost surely supported on a set of Hausdorff dimension 1-\gamma^{2}/2 . For |\gamma|\geq\sqrt{2} , the limit degenerates (one obtains the zero measure) and a different renormalization is needed, see for instance [ 10 ] and the references therein. One may further generalize the construction to complex parameters \gamma=\alpha+\mathrm{i}\beta\in\mathbb{C} . The range of parameters for which it is non-trivial, in the one-dimensional setting considered here, is given by the subcritical domain \mathcal{D}\coloneqq\mathopen{}\mathclose{{\left\{\,|\gamma|<1\,}}\right\}\;\cup\;\mathopen{}\mathclose{{\left\{\,|\alpha|\in(1,\sqrt{2})\ \text{and}\ |\alpha|+|\beta|<\sqrt{2}\,}}\right\}. A detailed discussion in the general case can be found in [ 12 ] . In particular, \mathcal{D} contains purely imaginary parameters \gamma=\mathrm{i}\beta with |\beta|<1 . In this regime, the limit \mathrm{M}_{\mathrm{i}\beta}\coloneqq\lim_{\varepsilon\to 0}\,\mathrm{M}_{\mathrm{i}\beta}^{(\varepsilon)} takes place in the space H^{s}(\mathbb{T})\coloneqq\mathopen{}\mathclose{{\left\{\varphi\in\mathcal{D}^{\prime}(\mathbb{T}):\mathopen{}\mathclose{{\left\|\varphi}}\right\|_{H^{s}(\mathbb{T})}^{2}=\sum_{n\in\mathbb{Z}}(1+n^{2})^{s}\,|\widehat{\varphi}(n)|^{2}<\infty}}\right\}. for any s<-\beta^{2}/2 where \widehat{\varphi}(n)\coloneqq\int_{\mathbb{T}}e^{\mathrm{i}n\theta}\,\varphi(d\theta) is the Fourier coefficient of \varphi and the integral is understood in the sense of the action of the distribution \varphi on the smooth function \theta\mapsto e^{\mathrm{i}n\theta} . The imaginary case differs markedly from the real one. Although \mathrm{M}_{\mathrm{i}\beta} is no longer a positive measure (it is a complex-valued object), its total mass has finite moments of every order. Beyond this integrability property, the geometry also changes: while real GMC typically exhibits multifractal behavior [ 3 ] , imaginary chaos is monofractal [ 1 ] . For further properties and physical motivation, we refer to the foundational account [ 9 ] and to [ 2 ] for results on absolute continuity. In our setting, the relevant moments can be written as Coulomb-gas integrals on the circle, which can be analyzed through Jack-polynomial expansions as already explored in the physics literature, see e.g. [ 5 ] . Since \mathrm{M}_{\mathrm{i}\beta} lives on the circle, it is natural to investigate its Fourier analytic properties. In the real case, this direction was initiated in [ 7 ] , where it was proved that the Fourier coefficients of \mathrm{M}_{\gamma} tend to 0 almost surely and sharp decay exponents were conjectured. These conjectures were recently established in elegant works for closely related multiplicative cascade models in [ 4 ] and for GMC in [ 14 ] , the first in a series on the harmonic analysis of multiplicative chaos. The goal of the present paper is to continue this harmonic analysis in the purely imaginary setting. We set c_{n}\coloneqq\widehat{\mathrm{M}_{\mathrm{i}\beta}}(n)=\int_{\mathbb{T}}e^{\mathrm{i}n\theta}\,\mathrm{M}_{\mathrm{i}\beta}(d\theta). The Fourier dimension quantifies the decay of the Fourier coefficients of \varphi\in\mathcal{D}^{\prime}(\mathbb{T}) . It is given by \dim_{F}(\varphi)\coloneqq\sup\mathopen{}\mathclose{{\left\{\,s\in[0,1]:|\widehat{\varphi}(n)|^{2}=O(|n|^{-s})\text{ as }|n|\to\infty}}\right\}. Our first result identifies the Fourier dimension of \mathrm{M}_{\mathrm{i}\beta} . Theorem 1.1 (Fourier dimension) . For \beta\in(0,1) , the Fourier dimension of \mathrm{M}_{\mathrm{i}\beta} is 1-\beta^{2} almost surely. An upper bound of the Fourier dimension follows from the regularity results obtained in [ 9 ] . The authors prove that almost surely \mathrm{M}_{\mathrm{i}\beta}\in H^{s}(\mathbb{T}) for s<-\beta^{2}/2 and \mathrm{M}_{\mathrm{i}\beta}\notin H^{s}(\mathbb{T}) for s>-\beta^{2}/2 , which implies \dim_{F}(\mathrm{M}_{\mathrm{i}\beta})\leq 1-\beta^{2} . The content of Theorem 1.1 is to establish the lower bound. The second result is the convergence in law of the rescaled Fourier coefficients. Here, we denote \mathcal{N}_{\mathbb{C}}(0,\sigma^{2}) the Gaussian complex random variable such that \mathbb{E}\,\mathcal{N}_{\mathbb{C}}(0,\sigma^{2})^{2}=0\quad\text{ and }\quad\mathbb{E}\,|\mathcal{N}_{\mathbb{C}}(0,\sigma^{2})|^{2}=\sigma^{2}. Theorem 1.2 (CLT for the rescaled coefficients) . The following convergence in distribution holds: n^{\frac{1-\beta^{2}}{2}}\,c_{n}\longrightarrow\mathcal{N}_{\mathbb{C}}(0,\kappa(\beta)),\qquad\text{ as }n\rightarrow\infty, where \kappa(\beta)=4\pi\Gamma(1-\beta^{2})\sin(\frac{\pi\beta^{2}}{2}) . Additionally, for any fixed integer k\geq 0 , the rescaled process converges n^{\frac{1-\beta^{2}}{2}}\,(c_{n},\dots,c_{n+k})\longrightarrow(\mathcal{N}^{0}_{\mathbb{C}}(0,\kappa(\beta)),\dots,\mathcal{N}^{k}_{\mathbb{C}}(0,\kappa(\beta))),\qquad\text{ as }n\rightarrow\infty, where the Gaussian random variables are independent. As a consequence, we obtain the following result. Corollary 1.3 (Convergence toward a complex white noise) . Let s<-\frac{1}{2} . We have the following convergence in distribution in H^{s}(\mathbb{T}) : n^{\frac{1-\beta^{2}}{2}}\,e^{\mathrm{i}n\theta}\,\mathrm{M}_{\mathrm{i}\beta}\longrightarrow W,\qquad\text{ as }n\rightarrow\infty, where W is a complex white noise of intensity measure \kappa(\beta)\,\mathrm{d}\theta . Motivated by the computation of the correlation dimension as performed in [ 11 , 7 ] , we propose the following conjecture Conjecture 1.4 . For \gamma=\alpha+\mathrm{i}\beta\in\mathcal{D} , the Fourier dimension of \mathrm{M}_{\gamma} is given by \mathrm{dim}_{F}(\mathrm{M}_{\gamma})=\mathrm{dim}_{F}(\mathrm{M}_{\alpha})-\beta^{2}. Strategy. We proceed by the method of moments. The first ingredient is the rotational invariance of the field, which implies that each Fourier mode c_{n} is isotropic in the complex plane. As a consequence, to identify the limiting law of c_{n} it is enough to control the absolute moments. The second ingredient is that these moments admit an explicit integral representation and such integrals can be analyzed by doing an expansion in the basis of Jack symmetric polynomials. This yields exact identities expressing the moments in terms of sums indexed by partitions. A key point in our approach is that the asymptotic study of these sums relies on the arithmetic of Jack polynomials, in particular on Pieri formulas and explicit evaluations. Finally, asymptotics of these expressions provide the moment limits, from which the Gaussian fluctuations and the joint convergence follow. 2. The moments of c_{n} For complex-valued random variables, the convergence in distribution is a consequence of the convergence of the mixed moments (2.1) \mathbb{E}\mathopen{}\mathclose{{\left[Z_{n}^{p}\overline{Z_{n}}^{q}}}\right]\longrightarrow\mathbb{E}\mathopen{}\mathclose{{\left[Z^{p}\overline{Z}^{q}}}\right]\qquad\text{ for }p,q\in\mathbb{N}, as soon as the law of Z is characterized by its moments. This is the case for the complex Gaussian distribution \mathcal{N}_{\mathbb{C}} . Here, an additional symmetry greatly simplifies matters, since the law of c_{n} is isotropic for all n>0 . Recall that a complex-valued random variable Z is isotropic if for every \alpha\in\mathbb{R} , Z has the same law as e^{\mathrm{i}\alpha}Z . It is equivalent to the two following properties: (i) \theta is uniformly distributed on \mathopen{}\mathclose{{\left[0,2\pi}}\right] , (ii) \theta is independent from R . where (R,\theta) is the polar decomposition of Z . For isotropic random variables, the convergence in distribution is equivalent to the convergence in distribution of the modulus