Edge fluctuations and third-order phase transition in harmonically confined long-range systems. (1st March 2022)

• Record Type: Journal Article
• Authors: Kethepalli, Jitendra; Kulkarni, Manas; Kundu, Anupam; Majumdar, Satya N; Mukamel, David; Schehr, Grégory
• Is Part Of: Journal of statistical mechanics. (2022:Mar.)
• Publication Date: 2022-03-01
• Subjects: extreme value; Statistical mechanics -- Periodicals; Mechanics -- Statistical methods -- Periodicals; 530.1305
• DOI: 10.1088/1742-5468/ac52b2
• Languages: English
• ISSNs: 1742-5468

Abstract

Abstract: We study the distribution of the position of the rightmost particle x max in a N -particle Riesz gas in one dimension confined in a harmonic trap. The particles interact via long-range repulsive potential, of the form r − k with −2 < k < ∞ where r is the inter-particle distance. In equilibrium at temperature O (1), the gas settles on a finite length scale L N that depends on N and k . We numerically observe that the typical fluctuation of y max = x max / L N around its mean is of O ( N − η k ) . Over this length scale, the distribution of the typical fluctuations has a N independent scaling form. We show that the exponent η k obtained from the Hessian theory predicts the scale of typical fluctuations remarkably well. The distribution of atypical fluctuations to the left and right of the mean ⟨ y max ⟩ are governed by the left and right large deviation functions (LDFs), respectively. We compute these LDFs explicitly ∀ k > −2. We also find that these LDFs describe a pulled to pushed type phase transition as observed in Dyson's log-gas ( k → 0) and 1 d one component plasma ( k = −1). Remarkably, we find that the phase transition remains third order for the entire regime. Our results demonstrate the striking universality of the third order transition even in models that fall outside the paradigm of Coulomb systems and the random matrix theory. We numerically verify our analytical expressions of the LDFs via Monte Carlo simulation using an importance sampling algorithm.