Publications

The number of microbial and metagenomic studies has increased drastically due to advancements in next-generation sequencing-based measurement techniques. Statistical analysis and the validity of conclusions drawn from (time series) 16S rRNA and other metagenomic sequencing data is hampered by the presence of significant amount of noise and missing data (sampling zeros). Accounting uncertainty in microbiome data is often challenging due to the difficulty of obtaining biological replicates. Additionally, the compositional nature of current amplicon and metagenomic data differs from many other biological data types adding another challenge to the data analysis. To address these challenges in human microbiome research, we introduce a novel probabilistic approach to explicitly model overdispersion and sampling zeros by considering the temporal correlation between nearby time points using Gaussian Processes. The proposed Temporal Gaussian Process Model for Compositional Data Analysis (TGP-CODA) shows superior modeling performance compared to commonly used Dirichlet-multinomial, multinomial, and non-parametric regression models on real and synthetic data. We demonstrate that the nonreplicative nature of human gut microbiota studies can be partially overcome by our method with proper experimental design of dense temporal sampling. We also show that different modeling approaches have a strong impact on ecological interpretation of the data, such as stationarity, persistence, and environmental noise models. A Stan implementation of the proposed method is available under MIT license at

We present a theoretical investigation of the effects of correlations on the electronic structure of the Mott insulator Sr2IrO4 upon electron doping. A rapid collapse of the Mott gap upon doping is found, and the electronic structure displays a strong momentum-space differentiation at low doping level: The Fermi surface consists of pockets centered around (π/2,π/2), while a pseudogap opens near (π,0). Its physical origin is shown to be related to short-range spin correlations. The pseudogap closes upon increasing doping, but a differentiated regime characterized by a modulation of the spectral intensity along the Fermi surface persists to higher doping levels. These results, obtained within the cellular dynamical mean-field theory framework, are discussed in comparison to recent photoemission experiments and an overall good agreement is found.

Multiple stellar systems are ubiquitous in the Milky Way, but are often unresolved and seen as single objects in spectroscopic, photometric, and astrometric surveys. Yet, modeling them is essential for developing a full understanding of large surveys such as Gaia, and connecting them to stellar and Galactic models. In this paper we address this problem by jointly fitting the Gaia and 2MASS photometric and astrometric data using a data-driven Bayesian hierarchical model that includes populations of binary and trinary systems. This allows us to classify observations into singles, binaries, and trinaries, in a robust and efficient manner, without resorting to external models. We are able to identify multiple systems and, in some cases, make strong predictions for the properties of its unresolved stars. We will be able to compare such predictions with Gaia Data Release 4, which will contain astrometric identification and analysis of binary systems.

The TREX is a recently introduced method for performing sparse high-dimensional regression. Despite its statistical promise as an alternative to the lasso, square-root lasso, and scaled lasso, the TREX is computationally challenging in that it requires solving a nonconvex optimization problem. This article shows a remarkable result: despite the nonconvexity of the TREX problem, there exists a polynomial-time algorithm that is guaranteed to find the global minimum. This result adds the TREX to a very short list of nonconvex optimization problems that can be globally optimized (principal components analysis being a famous example). After deriving and developing this new approach, we demonstrate that (i) the ability of the preexisting TREX heuristic to reach the global minimum is strongly dependent on the difficulty of the underlying statistical problem, (ii) the new polynomial-time algorithm for TREX permits a novel variable ranking and selection scheme, (iii) this scheme can be incorporated into a rule that controls the false discovery rate (FDR) of included features in the model. To achieve this last aim, we provide an extension of the results of Barber and Candes to establish that the knockoff filter framework can be applied to the TREX. This investigation thus provides both a rare case study of a heuristic for nonconvex optimization and a novel way of exploiting nonconvexity for statistical inference.

A resonantly excited coherent phonon leads to a periodic oscillation of the atomic lattice in a crystal structure bringing the material into a non-equilibrium electronic configuration. Periodically oscillating quantum systems can be understood in terms of Floquet theory and we show these concepts can be applied to coherent lattice vibrations reflecting the underlying coupling mechanism between electrons and bosonic modes. This coupling leads to dressed quasi-particles imprinting specific signatures in the spectrum of the electronic structure. Taking graphene as a paradigmatic material we show how the phonon-dressed states display an intricate sideband structure revealing electron-phonon coupling and topological ordering. This work establishes that the recently demonstrated concept of light-induced non-equilibrium Floquet phases can also be applied when using coherent phonon modes for the dynamical control of material properties. The present results are generic for bosonic time-dependent perturbations and similar phenomena can be observed for plasmon, magnon or exciton driven materials.

Local spin fluctuations provide the glue for orbital-singlet spin-triplet pairing in the doped Mott insulating regime of multi-orbital Hubbard models. At large Hubbard repulsion U, the pairing susceptibility is nevertheless very low, because the pairing interaction cannot overcome the suppression of charge fluctuations. Using nonequilibrium dynamical mean field simulations of the two-orbital Hubbard model, we show that out of equilibrium the pairing susceptibility in this large-U regime can be strongly enhanced by creating a photo-induced population of the relevant charge states, and that this susceptibility correlates with the local spin susceptibility. Since a strong enhancement of the pairing requires a low kinetic energy of the charge carriers, the phenomenon is supported by the ultra-fast cooling of the photo-doped carriers through the creation of local spin excitations.

We extend the self-energy functional theory (SFT) to the case of interacting lattice bosons in the presence of symmetry breaking and quenched disorder. The self-energy functional we derive depends only on the self-energies of the disorder-averaged propagators, allowing for the construction of general non-perturbative approximations. Using a simple single-site reference system with only three variational parameters, we are able to reproduce numerically exact quantum Monte Carlo results in the Bose-Hubbard model with box disorder with high accuracy. Deep in the strongly-disordered weakly-interacting regime, the simple reference system employed is insufficient and no stationary solutions can be found within its restricted variational subspace. By systematically analyzing thermodynamical observables and the spectral function, we find that the strongly-interacting Bose glass is characterized by different regimes, depending on which local occupations are activated as a function of the disorder strength. We find that the particles delocalize into isolated superfluid lakes over a strongly localized background around maximally-occupied sites whenever these sites are particularly rare. Our results indicate that the transition from the Bose glass to the superfluid phase around unit filling at strong interactions is driven by the percolation of superfluid lakes which form around doubly occupied sites.

We consider a one-dimensional interacting spinless fermion model, which displays the well-known Luttinger liquid (LL) to charge density wave (CDW) transition as a function of the ratio between the strength of the interaction, U, and the hopping, J. We subject this system to a spatially uniform drive which is ramped up over a finite time interval and becomes time-periodic in the long time limit. We show that by using a density matrix renormalization group (DMRG) approach formulated for infinite system sizes, we can access the large-time limit even when the drive induces finite heating. When both the initial and long-time states are in the gapless (LL) phase, the final state has power law correlations for all ramp speeds. However, when the initial and final state are gapped (CDW phase), we find a pseudothermal state with an effective temperature that depends on the ramp rate, both for the Magnus regime in which the drive frequency is very large compared to other scales in the system and in the opposite limit where the drive frequency is less than the gap. Remarkably, quantum defects (instantons) appear when the drive tunes the system through the quantum critical point, in a realization of the Kibble-Zurek mechanism.

Many materials exhibit metal-insulator transitions that are driven by electron correlation effects but also involve structural changes. This paper uses density functional, dynamical mean field and Landau-theory methods to elucidate the interplay of electronic and structural energetics in Ca2RuO4. We find that the change in lattice energies across the metal-insulator transition is comparable to the change in electronic energies. Important consequences are the strongly first order nature of the transition, a sensitive dependence on pressure, and that imposition of geometrical constraints (for example via epitaxial growth on a substrate) can change the lattice energetics enough to eliminate the metal-insulator transition entirely. A comparison to recent data is presented.