Publications

An efficient, reliable, and interpretable global solution method, the Deep learning-based algorithm for Heterogeneous Agent Models (DeepHAM), is proposed for solving high dimensional heterogeneous agent models with aggregate shocks. The state distribution is approximately represented by a set of optimal generalized moments. Deep neural networks are used to approximate the value and policy functions, and the objective is optimized over directly simulated paths. In addition to being an accurate global solver, this method has three additional features. First, it is computationally efficient in solving complex heterogeneous agent models, and it does not suffer from the curse of dimensionality. Second, it provides a general and interpretable representation of the distribution over individual states, which is crucial in addressing the classical question of whether and how heterogeneity matters in macroeconomics. Third, it solves the constrained efficiency problem as easily as it solves the competitive equilibrium, which opens up new possibilities for studying optimal monetary and fiscal policies in heterogeneous agent models with aggregate shocks.

State redistribution is an algorithm that stabilizes cut cells for embedded boundary grid methods. This work extends the earlier algorithm in several important ways. First, state redistribution is extended to three spatial dimensions. Second, we discuss several algorithmic changes and improvements motivated by the more complicated cut cell geometries that can occur in higher dimensions. In particular, we introduce a weighted version with less dissipation. Third, we demonstrate that state redistribution can also stabilize a solution update that includes both advective and diffusive contributions. The stabilization algorithm is shown to be effective for incompressible as well as compressible reacting flows. Finally, we discuss the implementation of the algorithm for several exascale-ready simulation codes based on AMReX, demonstrating ease of use in combination with domain decomposition, hybrid parallelism and complex physics.

This paper presents the results of a search for generic short-duration gravitational-wave transients in data from the third observing run of Advanced LIGO and Advanced Virgo. Transients with durations of milliseconds to a few seconds in the 24--4096 Hz frequency band are targeted by the search, with no assumptions made regarding the incoming signal direction, polarization or morphology. Gravitational waves from compact binary coalescences that have been identified by other targeted analyses are detected, but no statistically significant evidence for other gravitational wave bursts is found. Sensitivities to a variety of signals are presented. These include updated upper limits on the source rate-density as a function of the characteristic frequency of the signal, which are roughly an order of magnitude better than previous upper limits. This search is sensitive to sources radiating as little as ∼10−10M⊙c2 in gravitational waves at ∼70 Hz from a distance of 10~kpc, with 50\% detection efficiency at a false alarm rate of one per century. The sensitivity of this search to two plausible astrophysical sources is estimated: neutron star f-modes, which may be excited by pulsar glitches, as well as selected core-collapse supernova models.

In this paper, we consider the reconstruction of the shape and the impedance function of an obstacle from measurements of the scattered field at a collection of receivers outside the object. The data is assumed to be generated by plane waves impinging on the unknown obstacle from multiple directions and at multiple frequencies. This inverse problem can be reformulated as an optimization problem: that of finding band-limited shape and impedance functions which minimize the L2 distance between the computed value of the scattered field at the receivers and the given measurement data. The optimization problem is highly non-linear, non-convex, and ill-posed. Moreover, the objective function is computationally expensive to evaluate (since a large number of Helmholtz boundary value problems need to be solved at every iteration in the optimization loop). The recursive linearization approach (RLA) proposed by Chen has been successful in addressing these issues in the context of recovering the sound speed of an inhomogeneous object or the shape of a sound-soft obstacle. We present an extension of the RLA for the recovery of both the shape and impedance functions of the object. The RLA is, in essence, a continuation method in frequency where a sequence of single frequency inverse problems is solved. At each higher frequency, one attempts to recover incrementally higher resolution features using a step assumed to be small enough to ensure that the initial guess obtained at the preceding frequency lies in the basin of attraction for Newton’s method at the new frequency. We demonstrate the effectiveness of this approach with several numerical examples. Surprisingly, we find that one can recover the shape with high accuracy even when the measurements are generated by sound-hard or sound-soft objects, eliminating the need to know the precise boundary conditions appropriate for modeling the object under consideration. While the method is effective in obtaining high-quality reconstructions for many complicated geometries and impedance functions, a number of interesting open questions remain regarding the convergence behavior of the approach. We present numerical experiments that suggest underlying mechanisms of success and failure, pointing out areas where improvements could help lead to robust and automatic tools for the solution of inverse obstacle scattering problems.

The SDSS-IV Apache Point Observatory Galactic Evolution Experiment (APOGEE) survey has obtained high-resolution spectra for thousands of red giant stars distributed among the massive satellite galaxies of the Milky Way (MW): the Large and Small Magellanic Clouds (LMC/SMC), the Sagittarius Dwarf (Sgr), Fornax (Fnx), and the now fully disrupted \emph{Gaia} Sausage/Enceladus (GSE) system. We present and analyze the APOGEE chemical abundance patterns of each galaxy to draw robust conclusions about their star formation histories, by quantifying the relative abundance trends of multiple elements (C, N, O, Mg, Al, Si, Ca, Fe, Ni, and Ce), as well as by fitting chemical evolution models to the [α/Fe]-[Fe/H] abundance plane for each galaxy. Results show that the chemical signatures of the starburst in the MCs observed by Nidever et al. in the α-element abundances extend to C+N, Al, and Ni, with the major burst in the SMC occurring some 3-4 Gyr before the burst in the LMC. We find that Sgr and Fnx also exhibit chemical abundance patterns suggestive of secondary star formation epochs, but these events were weaker and earlier (∼~5-7 Gyr ago) than those observed in the MCs. There is no chemical evidence of a second starburst in GSE, but this galaxy shows the strongest initial star formation as compared to the other four galaxies. All dwarf galaxies had greater relative contributions of AGB stars to their enrichment than the MW. Comparing and contrasting these chemical patterns highlight the importance of galaxy environment on its chemical evolution.

A significant fraction of Milky Way (MW) satellites exhibit phase-space properties consistent with a coherent orbital plane. Using tailored N-body simulations of a spherical MW halo that recently captured a massive (1.8 × 1011 M⊙) LMC-like satellite, we identify the physical mechanisms that may enhance the clustering of orbital poles of objects orbiting the MW. The LMC deviates the orbital poles of MW dark matter particles from the present-day random distribution. Instead, the orbital poles of particles beyond R ≈ 50 kpc cluster near the present-day orbital pole of the LMC along a sinusoidal pattern across the sky. The density of orbital poles is enhanced near the LMC by a factor $\delta {\rho }_{\max }$ = 30 percent (50 percent) with respect to underdense regions and δρiso = 15 percent (30 percent) relative to the isolated MW simulation (no LMC) between 50 and 150 kpc (150–300 kpc). The clustering appears after the LMC's pericenter (≈50 Myr ago, 49 kpc) and lasts for at least 1 Gyr. Clustering occurs because of three effects: (1) the LMC shifts the velocity and position of the central density of the MW's halo and disk; (2) the dark matter dynamical friction wake and collective response induced by the LMC change the kinematics of particles; (3) observations of particles selected within spatial planes suffer from a bias, such that measuring orbital poles in a great circle in the sky enhances the probability of their orbital poles being clustered. This scenario should be ubiquitous in hosts that recently captured a massive satellite (at least ≈1:10 mass ratio), causing the clustering of orbital poles of halo tracers.

The competition between thermal fluctuations and potential forces governs the stability of matter in equilibrium, in particular the proliferation and annihilation of topological defects. However, driving matter out of equilibrium allows for a new class of forces that are neither attractive nor repulsive, but rather transverse. The possibility of activating transverse forces raises the question of how they affect basic principles of material self-organization and control. Here we show that transverse forces organize colloidal spinners into odd elastic crystals crisscrossed by motile dislocations. These motile topological defects organize into a polycrystal made of grains with tunable length scale and rotation rate. The self-kneading dynamics drive super-diffusive mass transport, which can be controlled over orders of magnitude by varying the spinning rate. Simulations of both a minimal model and fully resolved hydrodynamics establish the generic nature of this crystal whorl state. Using a continuum theory, we show that both odd and Hall stresses can destabilize odd elastic crystals, giving rise to a generic state of crystalline active matter. Adding rotations to a material’s constituents has far-reaching consequences for continuous control of structures and transport at all scales.

In most animals, the oocyte is the largest cell by volume. The oocyte undergoes a period of large-scale growth during its development, prior to fertilization. At first glance, tissues that support the development of the oocyte in different organisms have diverse cellular characteristics that would seem to prohibit functional comparisons. However, these tissues often act with a common goal of establishing dynamic forms of two-way communication with the oocyte. We propose that this bidirectional communication between oocytes and support cells is a universal phenomenon that can be directly compared across species. Specifically, we highlight fruit fly and mouse oogenesis to demonstrate that similarities and differences in these systems should be used to inform and design future experiments in both models.

Learning appropriate representations of the reward environment is challenging in the real world where there are many options, each with multiple attributes or features. Despite existence of alternative solutions for this challenge, neural mechanisms underlying emergence and adoption of value representations and learning strategies remain unknown. To address this, we measure learning and choice during a multi-dimensional probabilistic learning task in humans and trained recurrent neural networks (RNNs) to capture our experimental observations. We find that human participants estimate stimulus-outcome associations by learning and combining estimates of reward probabilities associated with the informative feature followed by those of informative conjunctions. Through analyzing representations, connectivity, and lesioning of the RNNs, we demonstrate this mixed learning strategy relies on a distributed neural code and opponency between excitatory and inhibitory neurons through value-dependent disinhibition. Together, our results suggest computational and neural mechanisms underlying emergence of complex learning strategies in naturalistic settings.

The circumgalactic medium (CGM) contains information on gas flows around galaxies, such as accretion and supernova-driven winds, which are difficult to constrain from observations alone. Here, we use the high-resolution TNG50 cosmological magnetohydrodynamical simulation to study the properties and kinematics of the CGM around star-forming galaxies in 1011.5−1012M⊙ halos at z≃ 1 using mock MgII absorption lines, which we generate by postprocessing halos to account for photoionization in the presence of a UV background. We find that the MgII gas is a very good tracer of the cold CGM, which is accreting inward at inflow velocities of up to 50 km s−1. For sight lines aligned with the galaxy's major axis, we find that MgII absorption lines are kinematically shifted due to the cold CGM's significant corotation at speeds up to 50% of the virial velocity for impact parameters up to 60 kpc. We compare mock MgII spectra to observations from the MusE GAs FLow and Wind (MEGAFLOW) survey of strong MgII absorbers (EW2796Å0>0.5Å). After matching the equivalent-width (EW) selection, we find that the mock MgII spectra reflect the diversity of observed kinematics and EWs from MEGAFLOW, even though the sight lines probe a very small fraction of the CGM. MgII absorption in higher-mass halos is stronger and broader than in lower-mass halos but has qualitatively similar kinematics. The median-specific angular momentum of the MgII CGM gas in TNG50 is very similar to that of the entire CGM and only differs from non-CGM components of the halo by normalization factors of ≲ 1 dex.