Sourav Chatterjee (born November 26, 1979) is an Indian mathematician specializing in probability theory, statistics, and mathematical physics.[1] He is renowned for his groundbreaking contributions to understanding rare events, large deviations in random graphs and matrices, first-passage percolation, and applications spanning social networks, spin glasses, and quantum mechanics.[2] Currently a professor of mathematics and statistics at Stanford University, Chatterjee has earned prestigious honors including the Infosys Prize in Mathematical Sciences (2020), Fellowship of the Royal Society (2023), and membership in the American Academy of Arts and Sciences (2024).[1][3][4]Chatterjee was born in Kolkata, India, and earned his Bachelor of Statistics (2000) and Master of Statistics (2002) from the Indian Statistical Institute in Kolkata.[1] He then pursued a Ph.D. in Statistics at Stanford University, completing it in 2005 under advisor Persi Diaconis.[1] His early career included positions as a Neyman Assistant Professor and later Assistant Professor at the University of California, Berkeley (2005–2009), followed by Associate Professor at the Courant Institute of New York University (2009–2013).[1] Since 2013, he has held a full professorship at Stanford, where his research has produced influential works such as Superconcentration and Related Topics (2014), a monograph on advanced probabilistic phenomena, and papers establishing nonlinear large deviations (2016) and the universal relation in first-passage percolation (2013).[1][5]Throughout his career, Chatterjee has received numerous accolades for his innovative problem-solving in probability, including the Sloan Research Fellowship (2007), Tweedie New Researcher Award (2008), Rollo Davidson Prize (2010), Doeblin Prize (2012), and Loève Prize (2013).[4] He was also an invited speaker at the International Congress of Mathematicians in 2014 and became a Fellow of the Institute of Mathematical Statistics in 2018.[6] His work bridges pure mathematics with applied fields, providing rigorous tools for analyzing complex systems in physics and networks, and he holds U.S. permanent resident status while maintaining Indian citizenship.[2][1]
Early life and education
Early life
Sourav Chatterjee was born on November 26, 1979, in Calcutta (now Kolkata), West Bengal, India, and holds Indian citizenship.[1] He grew up in India during his formative years, immersed in an environment that fostered intellectual curiosity.[2]From a young age, Chatterjee displayed a strong inclination toward mathematics and related activities, crediting his family, along with mentors and friends, for supporting his development.[2] His early hobbies included programming computers and playing chess, pursuits that cultivated his analytical mindset and passion for problem-solving.[2]These interests laid the groundwork for his academic path, leading him to the Indian Statistical Institute in Kolkata for formal education.[2]
Education
Sourav Chatterjee earned his Bachelor of Statistics (B.Stat.) from the Indian Statistical Institute in Kolkata in May 2000.[1] He continued his studies at the same institution, completing a Master of Statistics (M.Stat.) in May 2002.[1][4]Following his M.Stat., Chatterjee enrolled in the Ph.D. program in Statistics at Stanford University.[7] He completed his doctorate in June 2005 under the supervision of Persi Diaconis.[1] His dissertation, titled Concentration Inequalities with Exchangeable Pairs, introduced a version of Stein's method using exchangeable pairs to address problems in measure concentration, with applications to random graphs and statistical physics.[8][9]
In September 2009, Sourav Chatterjee joined the Courant Institute of Mathematical Sciences at New York University as an Associate Professor of Mathematics, marking a transition in his career.[1][7] This appointment was in a center for applied mathematics, where he contributed to interdisciplinary research in probability and statistics.During his tenure at NYU, which lasted until August 2013, Chatterjee maintained an overlapping position as Associate Professor of Mathematics and Statistics at the University of California, Berkeley, from July 2009 to June 2011, after which he committed fully to the Courant Institute.[1] This period of dual affiliation enabled collaboration on probabilistic modeling projects.[7]Chatterjee's involvement in departmental activities at the Courant Institute included editorial responsibilities. He served as an Associate Editor for the Annals of Probability from 2009 to 2014, handling submissions in probability theory.[1][7] He also served as an Associate Editor for Annales de l'Institut Henri Poincaré (B): Probability and Statistics from 2008 to 2013.[1] These roles supported peer review in stochastic processes and statistical inference.
Stanford University
From September 2012 to August 2013, Chatterjee served as a Visiting Associate Professor of Mathematics and Statistics at Stanford University while affiliated with NYU.[1]In September 2013, Sourav Chatterjee joined Stanford University as a Full Professor with joint appointments in the Departments of Mathematics and Statistics, a position he holds as of 2025.[7][11] This followed his associate professorship at New York University.[6]Chatterjee took a sabbatical as a Member in the School of Mathematics at the Institute for Advanced Study in Princeton for the 2023–2024 academic year, focusing on probability theory and mathematical physics.[6][1]In addition to his teaching and research roles, Chatterjee serves as an associate editor for Communications in Mathematical Physics since 2019.[7][12] He also serves as Associate Editor for Annals of Applied Probability since 2022 and for Proceedings of the London Mathematical Society since 2023.[1]
Research contributions
Probability theory
Sourav Chatterjee has made foundational contributions to concentration inequalities using Stein's method, particularly in settings involving dependence among random variables. In his 2007 paper, he developed a novel version of Stein's method tailored for proving concentration and moment bounds in problems where traditional independence assumptions fail, such as in random structures with complex interactions. This approach leverages exchangeable pairs to derive sharp bounds on deviations, enabling the analysis of fluctuations in dependent systems like those arising in combinatorial probability. For instance, the method yields exponential tail estimates for functions of dependent random variables, improving upon earlier techniques like McDiarmid's inequality by accounting for correlations.[13]Building on this, Chatterjee extended Stein's method to specific applications in random structures, including minimal spanning trees and other graph-theoretic objects, where it provides precise control over variance and higher moments. His 2010 work further refined these tools, demonstrating their utility in establishing concentration for measures on high-dimensional spaces with weak dependence, such as product measures perturbed by local interactions. These advancements have become standard for bounding errors in probabilistic approximations of random structures.Chatterjee's research also encompasses interacting particle systems and large deviations in random graphs. In a 2011 collaboration with Soumik Pal, he provided a combinatorial analysis of interacting diffusions, revealing phase transition behaviors in systems where particles evolve according to rank-based interactions, akin to Atlas models in queueing theory. This work elucidates the emergence of macroscopic order from microscopic dynamics in one-dimensional diffusions. Complementing this, his studies on large deviations for random graphs culminated in the 2017 book Large Deviations for Random Graphs, which presents a comprehensive framework for the large deviation principle in Erdős–Rényi models and related sparse graphs. There, he derives rate functions for graph homomorphisms and subgraph counts, resolving key questions on the probability of atypical graph configurations.[14][15]In the realm of Markov chains, Chatterjee has advanced the understanding of mixing times and cutoff phenomena. His 2020 paper with Persi Diaconis introduces deterministic jumps to accelerate convergence in finite-state Markov chains, reducing mixing times from exponential to polynomial scales in certain reversible cases while preserving the stationary distribution. Additionally, his contributions to cutoff include sharp thresholds for mixtures of permuted Markov chains, where total variation distance drops abruptly after a critical time, as shown in models like the permuted shuffle. More recent work includes enumerative theory for the Tsetlin library (2024), analyzing combinatorial structures in interacting particle systems.[16][17] These results, part of his broader impact on stochastic processes, earned him the inaugural Doeblin Prize in 2012 from the Bernoulli Society for outstanding early-career work in probability.In 2024, Chatterjee established an invariance principle for the one-dimensional Kardar-Parisi-Zhang (KPZ) equation, providing new insights into scaling limits of interface growth models. His 2025 preprint on a Vershik-Kerov theorem for wreath products extends representation-theoretic tools to probability on symmetric groups and their products.[18][19]
Statistics and machine learning
Chatterjee's contributions to statistics and machine learning emphasize robust methods for dependence measurement and theoretical insights into model generalization, particularly in high-dimensional and data-driven settings. His work bridges probabilistic tools with practical statistical inference, enabling better analysis of complex datasets without strong parametric assumptions. These innovations have influenced non-parametric approaches and the understanding of overparameterized models like neural networks.A key advancement is Chatterjee's introduction of a new correlation coefficient, known as Xicor, designed to overcome the shortcomings of classical measures such as Pearson's and Spearman's coefficients. Published in 2021, this rank-based estimator quantifies the strength of dependence between two random variables, achieving a value of 0 under independence and approaching 1 when one variable is a measurable function of the other, even in the presence of noise. Unlike traditional coefficients, Xicor requires no distributional assumptions, making it particularly suitable for high-dimensional data where relationships may be nonlinear or heteroscedastic.[20][21] Its asymptotic distribution under the null hypothesis is simple and tractable, facilitating hypothesis testing, and an open-source R package implements the method for practical use.[20] This coefficient has been adopted in applications requiring robust dependence detection, such as causal inference and graphical models.[22]In machine learning, Chatterjee has addressed the generalization puzzle of neural networks through a 2024 analysis of their behavior on low-complexity data. Co-authored with Timothy Sudijono, the work demonstrates that feedforward ReLU networks, when selected via minimum description length to interpolate independent and identically distributed training samples from a low-complexity distribution, generalize effectively to unseen data with high probability. Low complexity is formalized using a simple programming language to describe data-generating processes, such as primality testing on numbers up to N, where the interpolation error on new samples is bounded by O((\ln N)/n) for n training points.[23] This result explains empirical observations of benign overfitting in overparameterized models by showing that such networks implicitly discover parsimonious, generalizing solutions rather than memorizing noise, even under moderate data corruption.[23] The framework highlights how neural architectures can exploit underlying data structure in non-parametric regimes.Chatterjee's foundational work on empirical processes and non-parametric statistics centers on superconcentration phenomena, which provide refined tools for controlling fluctuations in high-dimensional estimators. His 2014 monograph Superconcentration and Related Topics develops a spectral theory linking superconcentration—exceptionally sharp concentration of measure in random fields—to chaos decomposition and multiple local minima in optimization landscapes.[24] This approach yields improved bounds on empirical process deviations where classical concentration inequalities fall short, enhancing non-parametric methods for tasks like density estimation and goodness-of-fit testing in large samples.[24] By applying these ideas to Gaussian processes and semigroups, the book equips statisticians with techniques to analyze variability in complex, non-i.i.d. data structures, influencing modern empirical risk minimization.[5]In 2024, he published a survey on recent developments in measures of association, building on tools like Xicor. His 2025 collaboration with Trevor Hastie and Robert Tibshirani introduced Univariate-Guided Sparse Regression (UniLasso), a method for high-dimensional regression that uses univariate screening to improve variable selection and prediction accuracy.[25][26]
Mathematical physics
Chatterjee has made significant contributions to the mathematical analysis of spin glass models, particularly in establishing rigorous properties of the Edwards-Anderson model at zero temperature. In this short-range spin glass model, applicable across all dimensions, he demonstrated that the ground state exhibits high sensitivity to small perturbations of the disorder, where even minor noise induces a new ground state nearly orthogonal to the original under the site overlap inner product.[27] Furthermore, he proved that overturning a macroscopic fraction of spins in the ground state incurs an energy cost negligible relative to the boundarysize of the overturned region, a hallmark of glassy behavior distinct from ferromagnetic systems.[27] Additional results include the fractal dimension of the overturned region's boundary exceeding d-1 in dimension d, the expected size of a critical droplet growing at least as a power of the volume, and bond correlations in the ground state decaying no faster than the inverse distance, contrasting with exponential decay in related models like the two-dimensional random field Ising model.[27] These findings provide the first mathematical confirmation of several long-standing physics predictions for short-range spin glasses.[27]In 2024, Chatterjee explored features of spin glasses in the random fieldIsing model, confirming glassy behavior through analysis of overlap distributions and chaos under field perturbations.[28]In the realm of quantum field theories, Chatterjee advanced the rigorous construction of scaling limits for non-Abelian gauge theories coupled with Higgs fields. For the SU(2) lattice Yang-Mills-Higgs model in dimensions d \geq 2, he established a continuum limit as the lattice spacing \varepsilon \to 0, gauge coupling g \to 0, and Higgs length \alpha \to \infty, with \alpha g = c \varepsilon for fixed c > 0 and g = O(\varepsilon^{50d}).[29] Under unitary gauge fixing, the stereographic projection of the gauge field converges in distribution to a scale-invariant massive Gaussian free field, marking the first such scaling limit for non-Abelian lattice Yang-Mills theories beyond dimension 2 and offering the initial rigorous proof of mass generation through the Higgs mechanism in this setting.[29] Analogous convergence holds for the Abelian U(1) case, while non-Gaussian limits remain unresolved.[29] Also in 2024, he constructed a state space for 3D Euclidean Yang-Mills theories, providing a probabilistic framework for non-perturbative analysis.[30]Chatterjee co-authored an accessible introduction to rigorous methods in Liouville quantum field theory, emphasizing developments that provide a direct mathematical foundation for this two-dimensional conformal field theory.[31] The work surveys recent probabilistic and analytic techniques for constructing the theory, including Gaussian multiplicative chaos and regularity structures, aimed at bridging gaps for physicists in quantum field theory.[32] This exposition highlights how these approaches resolve longstanding challenges in defining the Liouville action and its correlation functions on the plane.[31] In 2025, he developed rigorous results for timelike Liouville field theory, including a proof of the DOZZ formula and correlation functions, with applications to quantum gravity models.[33]Chatterjee extended spectral gaptheory to nonreversible Markov chains, defining the gap as the second-smallest singular value of the generator on finite state spaces, which governs the relaxation time for empirical averages.[34] This framework yields Cheeger-type inequalities relating the gap to mixing times and path-based upper bounds, with examples illustrating faster convergence of averages than full mixing in nonreversible settings.[35] In the context of quantum field theories, these stochastic processes facilitate analysis of spectral properties in physical systems modeled by nonreversible dynamics, such as those arising in lattice approximations of field theories.[34] Broader probability tools, like martingale approximations, underpin these applications without altering the physical interpretations.
Awards and honors
Major prizes
Sourav Chatterjee has received several prestigious prizes recognizing his contributions to probability, statistics, and related fields. These awards, awarded during different stages of his career, highlight his innovative approaches and broad impact.In 2007, early in his academic career as an assistant professor at the University of California, Berkeley, Chatterjee was awarded the Sloan Research Fellowship in Mathematics by the Alfred P. Sloan Foundation, which supports exceptional early-career researchers in advancing fundamental knowledge.[7]In 2008, he received the Tweedie New Researcher Award from the Institute of Mathematical Statistics, recognizing his outstanding early contributions to probability and statistics.[36]The 2010 Rollo Davidson Prize, conferred by the University of Cambridge's Rollo Davidson Trustees and shared with Gady Kozma, honored Chatterjee's outstanding early contributions to probability theory, including work on random graphs and interacting particle systems.[37]Chatterjee became the inaugural recipient of the Doeblin Prize in 2012, awarded by the Bernoulli Society for Probability and Mathematical Statistics, for his groundbreaking innovations in probability, particularly in areas like concentration inequalities and statistical mechanics.[38]In 2013, while at New York University, he received the Line and Michel Loève International Prize in Probability from the University of California, Berkeley, which recognizes exceptional achievements by probabilists under 45, specifically for his transformative research bridging probability with statistics and physics.[39]The 2020 Infosys Prize in the Mathematical Sciences, awarded by the Infosys Science Foundation, acknowledged Chatterjee's profound influence on modern probability and statistical inference, including foundational results on high-dimensional data analysis and machine learning foundations.[3]
Fellowships and memberships
Sourav Chatterjee was elected a Fellow of the Institute of Mathematical Statistics in 2018, recognizing his outstanding contributions to the field of probability and statistics.[40] In 2023, he was elected a Fellow of the Royal Society, one of the highest honors for scientists in the United Kingdom, for his fundamental advances in probability theory and its applications.[4] The following year, in 2024, Chatterjee was inducted as a member of the American Academy of Arts and Sciences, acknowledging his influential work across mathematics, statistics, and related disciplines.[41]Chatterjee's recognition extends to prestigious lectureships and speaking invitations that highlight his impact on the mathematical sciences. He delivered the IMS Medallion Lecture in 2012, an honor awarded by the Institute of Mathematical Statistics to distinguished researchers under the age of 40 for exceptional contributions.[42] In 2014, he was selected as an invited speaker at the International Congress of Mathematicians in Seoul, a rare plenary recognition that underscores the broad influence of his research in probability on multiple fields.[4] Additionally, in 2013, he received the Young Researcher Award from the International Indian Statistical Association for his innovative work in theoretical probability.[43]These fellowships, memberships, and honors reflect the high regard in which Chatterjee's interdisciplinary research in probability theory, statistics, and mathematical physics is held within the global academic community.[7]