Search

General circulation models (GCMs) are the foundation of weather and climate prediction. GCMs are physics-based simulators which combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine learning (ML) models trained on reanalysis data achieved comparable or better skill than GCMs for deterministic weather forecasting. However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present the first GCM that combines a differentiable solver for atmospheric dynamics with ML components, and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best ML and physics-based methods. NeuralGCM is competitive with ML models for 1-10 day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for 1-15 day forecasts. With prescribed sea surface temperature, NeuralGCM can accurately track climate metrics such as global mean temperature for multiple decades, and climate forecasts with 140 km resolution exhibit emergent phenomena such as realistic frequency and trajectories of tropical cyclones. For both weather and climate, our approach offers orders of magnitude computational savings over conventional GCMs. Our results show that end-to-end deep learning is compatible with tasks performed by conventional GCMs, and can enhance the large-scale physical simulations that are essential for understanding and predicting the Earth system., Comment: 92 pages, 54 figures. Nature (2024)

We present a start-to-end simulation aimed at studying the long-term fate of high mass X-ray binaries and whether a Thorne-\.Zytkow object (T\.ZO) might ultimately be produced. We analyze results from a 3D hydrodynamical simulation that models the eventual fate of LMC X-4, a compact high mass X-ray binary system, after the primary fills its Roche lobe and engulfs the neutron star companion. We discuss the outcome of this engulfment within the standard paradigm of T\.ZO formation. The post-merger angular momentum content of the stellar core is a key ingredient, as even a small amount of rotation can break spherical symmetry and produce a centrifugally supported accretion disk. Our findings suggest the inspiraling neutron star, upon merging with the core, can accrete efficiently via a disk at high rates ($\approx 10^{-2}M_\odot/{\rm s}$), subsequently collapsing into a black hole and triggering a bright transient with a luminosity and duration typical of an ultra-long gamma-ray burst. We propose that the canonical framework for T\.ZO formation via common envelope needs to be revised, as the significant post-merger accretion feedback will unavoidably unbind the vast majority of the surrounding envelope.

The conditions required for massive star formation are debated, particularly whether massive stars must form in conjunction with massive clusters. Some authors have advanced the view that stars of any mass (below the total cluster mass) can form in clusters of any mass with some probability (random sampling). Others pointed out that the scatter in the determinations of the most massive star mass for a given cluster mass was consistent with the measurement error, such that the mass of the most massive star was determined by the total cluster mass (optimal sampling). Here we investigate the relation between cluster mass (M\textsubscript{ecl}) and the maximum stellar mass (M\textsubscript{max}) using a suite of SPH simulations. Varying cloud mass and turbulence random seed results in a range of cluster masses which we compare with their respective maximum star masses. We find that more massive clusters will have, on average, higher mass stars with this trend being steeper at lower cluster masses ($M\textsubscript{max} \propto M\textsubscript{ecl}^{0.31}$ for $M\textsubscript{ecl}<500M\,_{\odot}$) and flattening at higher cluster masses ($M\textsubscript{max} \propto M\textsubscript{ecl}^{0.11}$ for $M\textsubscript{ecl}>500M\,_{\odot}$). This rules out purely stochastic star formation in our simulations. Significant scatter in the maximum masses with identical initial conditions also rules out the possibility that the relation is purely deterministic (that is that a given cluster mass will result in a specific maximum stellar mass). In conclusion our simulations disagree with both random and optimal sampling of the initial mass function., Comment: Accepted in MNRAS

In the dark dimension scenario, which predicts an extra dimension of micron scale, dark gravitons (KK modes) are a natural dark matter candidate. In this paper, we study observable features of this model. In particular, their decay to standard matter fields can distort the CMB and impact other astrophysical signals. Using this we place bounds on the parameters of this model. In particular we find that the natural range of parameters in this scenario is consistent with these constraints and leads to the prediction that the mean mass of the dark matter today is close to a few hundred keV and the effective size of the extra dimension is around $1 - 30 \;\mu\mathrm{m}$., Comment: 20 pages, 5 figures, included updated constraints from related work and added citations

We study the properties of galaxies hosting mid-infrared outbursts in the context of a catalog of five hundred thousand galaxies from the Sloan Digital Sky Survey. We find that nuclear obscuration, as inferred by the surrounding dust mass, does not correlate with host galaxy type, stellar properties (e.g. total mass and mean age), or with the extinction of the host galaxy as estimated by the Balmer decrement. This implies that nuclear obscuration may not be able to explain any over-representation of tidal disruption events in particular host galaxies. We identify a region in the galaxy catalog parameter space that contains all unobscured tidal disruption events but only harbors $\lesssim $ 11\% of the mid-infrared outburst hosts. We find that mid-infrared outburst hosts appear more centrally concentrated and have higher galaxy S\'ersic indices than galaxies hosting active galactic nuclei (AGN) selected using the BPT classification. We thus conclude that the majority of mid-infrared outbursts are not hidden tidal disruption events but are instead consistent with being obscured AGN that are highly variable, such as changing-look AGN., Comment: 17 pages, 9 figures, accepted in ApJ Letters

Priors are essential for reconstructing images from noisy and/or incomplete measurements. The choice of the prior determines both the quality and uncertainty of recovered images. We propose turning score-based diffusion models into principled image priors ("score-based priors") for analyzing a posterior of images given measurements. Previously, probabilistic priors were limited to handcrafted regularizers and simple distributions. In this work, we empirically validate the theoretically-proven probability function of a score-based diffusion model. We show how to sample from resulting posteriors by using this probability function for variational inference. Our results, including experiments on denoising, deblurring, and interferometric imaging, suggest that score-based priors enable principled inference with a sophisticated, data-driven image prior., Comment: ICCV 2023

Star clusters are known to be formed in turbulent molecular clouds. How turbulence is driven in molecular clouds and what effect this has on star formation is still unclear. We compare a simulation setup with turbulent driving everywhere in a periodic box with a setup where turbulence is only driven around the outside of the box. We analyse the resulting gas distribution, kinematics, and the population of stars that are formed from the cloud. Both setups successfully produce a turbulent velocity field with a power law structure function, the externally driven cloud has a more central, monolithic, clump, while the fully driven cloud has many smaller, more dispersed, clumps. The star formation follows the cloud morphology producing large clusters, with high star forming efficiency in the externally driven simulations and sparse individual star formation with much lower star formation efficiency in the fully driven case. We conclude that the externally driven method, which resembles a Global Hierarchical Collapse (GHC) scenario, produces star clusters that more closely match with observations., Comment: 8 pages, 6 figures, accepted by MNRAS

Despite their ubiquity throughout science and engineering, only a handful of partial differential equations (PDEs) have analytical, or closed-form solutions. This motivates a vast amount of classical work on numerical simulation of PDEs and more recently, a whirlwind of research into data-driven techniques leveraging machine learning (ML). A recent line of work indicates that a hybrid of classical numerical techniques and machine learning can offer significant improvements over either approach alone. In this work, we show that the choice of the numerical scheme is crucial when incorporating physics-based priors. We build upon Fourier-based spectral methods, which are known to be more efficient than other numerical schemes for simulating PDEs with smooth and periodic solutions. Specifically, we develop ML-augmented spectral solvers for three common PDEs of fluid dynamics. Our models are more accurate (2-4x) than standard spectral solvers at the same resolution but have longer overall runtimes (~2x), due to the additional runtime cost of the neural network component. We also demonstrate a handful of key design principles for combining machine learning and numerical methods for solving PDEs.

Stars grazing supermassive black holes (SMBHs) on bound orbits may survive tidal disruption, causing periodic flares. Inspired by the recent discovery of the periodic nuclear transient ASASSN-14ko, a promising candidate for a repeating tidal disruption event (TDE), we study the tidal deformation of stars approaching SMBHs on eccentric orbits. With both analytical and hydrodynamics methods, we show the overall tidal deformation of a star is similar to that in a parabolic orbit provided that the eccentricity is above a critical value. This allows one to make use of existing simulation libraries from parabolic encounters to calculate the mass fallback rate in eccentric TDEs. We find the flare structures of eccentric TDEs show a complicated dependence on both the SMBH mass and the orbital period. For stars orbiting SMBHs with relatively short periods, we predict significantly shorter-lived duration flares than those in parabolic TDEs, which can be used to predict repeating events if the mass of the SMBH can be independently measured. Using an adiabatic mass loss model, we study the flare evolution over multiple passages, and show the evolved stars can survive many more passages than main sequence stars. We apply this theoretical framework to the repeating TDE candidate ASASSN-14ko and suggest that its recurrent flares originate from a moderately massive ($M\gtrsim 1 \mathrm{M_\odot}$), extended ($\approx$ a few $\mathrm{R_\odot}$), evolved star on a grazing, bound orbit around the SMBH. Future hydrodynamics simulations of multiple tidal interactions will enable realistic models on the individual flare structure and the evolution over multiple flares., Comment: 19 pages, 9 figures. Updated to accepted version (ApJ)

Ensembles are a straightforward, remarkably effective method for improving the accuracy,calibration, and robustness of models on classification tasks; yet, the reasons that underlie their success remain an active area of research. We build upon the extension to the bias-variance decomposition by Pfau (2013) in order to gain crucial insights into the behavior of ensembles of classifiers. Introducing a dual reparameterization of the bias-variance tradeoff, we first derive generalized laws of total expectation and variance for nonsymmetric losses typical of classification tasks. Comparing conditional and bootstrap bias/variance estimates, we then show that conditional estimates necessarily incur an irreducible error. Next, we show that ensembling in dual space reduces the variance and leaves the bias unchanged, whereas standard ensembling can arbitrarily affect the bias. Empirically, standard ensembling reducesthe bias, leading us to hypothesize that ensembles of classifiers may perform well in part because of this unexpected reduction.We conclude by an empirical analysis of recent deep learning methods that ensemble over hyperparameters, revealing that these techniques indeed favor bias reduction. This suggests that, contrary to classical wisdom, targeting bias reduction may be a promising direction for classifier ensembles.

Modern GPUs come with dedicated hardware to perform ray/triangle intersections and bounding volume hierarchy (BVH) traversal. While the primary use case for this hardware is photorealistic 3D computer graphics, with careful algorithm design scientists can also use this special-purpose hardware to accelerate general-purpose computations such as point containment queries. This article explains the principles behind these techniques and their application to vector field visualization of large simulation data using particle tracing.

This paper builds upon the work of Pfau (2013), which generalized the bias variance tradeoff to any Bregman divergence loss function. Pfau (2013) showed that for Bregman divergences, the bias and variances are defined with respect to a central label, defined as the mean of the label variable, and a central prediction, of a more complex form. We show that, similarly to the label, the central prediction can be interpreted as the mean of a random variable, where the mean operates in a dual space defined by the loss function itself. Viewing the bias-variance tradeoff through operations taken in dual space, we subsequently derive several results of interest. In particular, (a) the variance terms satisfy a generalized law of total variance; (b) if a source of randomness cannot be controlled, its contribution to the bias and variance has a closed form; (c) there exist natural ensembling operations in the label and prediction spaces which reduce the variance and do not affect the bias.

Controlling the evolution of nonequilibrium systems to minimize dissipated heat or work is a key goal for designing nanodevices, both in nanotechnology and biology. Progress in computing optimal protocols has thus far been limited to either simple systems or near-equilibrium evolution. Here, we present an approach for computing optimal protocols based on automatic differentiation. Our methodology is applicable to complex systems and multidimensional protocols and is valid arbitrarily far from equilibrium. We validate our method by reproducing theoretical optimal protocols for a Brownian particle in a time-varying harmonic trap. We also compute departures from near-equilibrium behaviour for magnetization reversal on an Ising lattice and for barrier crossing driven by a harmonic trap, which has been used to represent a range of biological processes including biomolecular unfolding reactions. Algorithms based on automatic differentiation outperform the near-equilibrium theory for far-from-equilibrium magnetization reversal and driven barrier crossing. The optimal protocol for crossing an energy landscape barrier of 10kT is found to hasten the approach to, and slow the departure from, the barrier region compared to the near-equilibrium theoretical protocol., Comment: 14 pages, 3 figures plus supplemental info (12 pages, 10 figures)

Tidal disruption events (TDEs) provide a unique opportunity to probe the stellar populations around supermassive black holes (SMBHs). By combining light curve modeling with spectral line information and knowledge about the stellar populations in the host galaxies, we are able to constrain the properties of the disrupted star for three TDEs. The TDEs in our sample have UV spectra, and measurements of the UV N III to C III line ratios enabled estimates of the nitrogen-to-carbon abundance ratios for these events. We show that the measured nitrogen line widths are consistent with originating from the disrupted stellar material dispersed by the central SMBH. We find that these nitrogen-to-carbon abundance ratios necessitate the disruption of moderately massive stars ($\gtrsim 1 - 2 M_\odot$). We determine that these moderately massive disruptions are over-represented by a factor of $\gtrsim 10^2$ when compared to the overall stellar population of the post-starburst galaxy hosts. This implies that SMBHs are preferentially disrupting higher mass stars, possibly due to ongoing top-heavy star formation in nuclear star clusters or to dynamical mechanisms that preferentially transport higher mass stars to their tidal radii., Comment: 16 pages, 6 figures, 1 table, accepted to ApJ

Bayesian optimization (BO) is a popular paradigm for global optimization of expensive black-box functions, but there are many domains where the function is not completely a black-box. The data may have some known structure (e.g. symmetries) and/or the data generation process may be a composite process that yields useful intermediate or auxiliary information in addition to the value of the optimization objective. However, surrogate models traditionally employed in BO, such as Gaussian Processes (GPs), scale poorly with dataset size and do not easily accommodate known structure. Instead, we use Bayesian neural networks, a class of scalable and flexible surrogate models with inductive biases, to extend BO to complex, structured problems with high dimensionality. We demonstrate BO on a number of realistic problems in physics and chemistry, including topology optimization of photonic crystal materials using convolutional neural networks, and chemical property optimization of molecules using graph neural networks. On these complex tasks, we show that neural networks often outperform GPs as surrogate models for BO in terms of both sampling efficiency and computational cost., Comment: 32 pages, 16 figures; published in TMLR

Variational data assimilation optimizes for an initial state of a dynamical system such that its evolution fits observational data. The physical model can subsequently be evolved into the future to make predictions. This principle is a cornerstone of large scale forecasting applications such as numerical weather prediction. As such, it is implemented in current operational systems of weather forecasting agencies across the globe. However, finding a good initial state poses a difficult optimization problem in part due to the non-invertible relationship between physical states and their corresponding observations. We learn a mapping from observational data to physical states and show how it can be used to improve optimizability. We employ this mapping in two ways: to better initialize the non-convex optimization problem, and to reformulate the objective function in better behaved physics space instead of observation space. Our experimental results for the Lorenz96 model and a two-dimensional turbulent fluid flow demonstrate that this procedure significantly improves forecast quality for chaotic systems., Comment: Published at the International Conference on Machine Learning (ICML) 2021

Numerical simulation of fluids plays an essential role in modeling many physical phenomena, such as weather, climate, aerodynamics and plasma physics. Fluids are well described by the Navier-Stokes equations, but solving these equations at scale remains daunting, limited by the computational cost of resolving the smallest spatiotemporal features. This leads to unfavorable trade-offs between accuracy and tractability. Here we use end-to-end deep learning to improve approximations inside computational fluid dynamics for modeling two-dimensional turbulent flows. For both direct numerical simulation of turbulence and large eddy simulation, our results are as accurate as baseline solvers with 8-10x finer resolution in each spatial dimension, resulting in 40-80x fold computational speedups. Our method remains stable during long simulations, and generalizes to forcing functions and Reynolds numbers outside of the flows where it is trained, in contrast to black box machine learning approaches. Our approach exemplifies how scientific computing can leverage machine learning and hardware accelerators to improve simulations without sacrificing accuracy or generalization., Comment: 13 pages, 9 figures

A binary neutron star merger has been observed in a multi-messenger detection of gravitational wave (GW) and electromagnetic (EM) radiation. Binary neutron stars that merge within a Hubble time, as well as many other compact binaries, are expected to form via common envelope evolution. Yet five decades of research on common envelope evolution have not yet resulted in a satisfactory understanding of the multi-spatial multi-timescale evolution for the systems that lead to compact binaries. In this paper, we report on the first successful simulations of common envelope ejection leading to binary neutron star formation in 3D hydrodynamics. We simulate the dynamical inspiral phase of the interaction between a 12$M_\odot$ red supergiant and a 1.4$M_\odot$ neutron star for different initial separations and initial conditions. For all of our simulations, we find complete envelope ejection and final orbital separations of $a_{\rm f} \approx 1.3$-$5.1 R_\odot$ depending on the simulation and criterion, leading to binary neutron stars that can merge within a Hubble time. We find $\alpha_{\rm CE}$-equivalent efficiencies of $\approx 0.1$-$2.7$ depending on the simulation and criterion, but this may be specific for these extended progenitors. We fully resolve the core of the star to $\lesssim 0.005 R_\odot$ and our 3D hydrodynamics simulations are informed by an adjusted 1D analytic energy formalism and a 2D kinematics study in order to overcome the prohibitive computational cost of simulating these systems. The framework we develop in this paper can be used to simulate a wide variety of interactions between stars, from stellar mergers to common envelope episodes leading to GW sources., Comment: 28 pages, 16 figures. v2: ran some additional simulations to address comments from the community. Included new criterion for calculating a_f and alpha_CE, resulting in new range of values. Added Table 1, updated Fig 3, added Fig 4, updated Fig 12, updated Fig 13, added Fig 14, added Fig 15, added Fig 16. Now submitted to ApJ

We study the properties of the host galaxies of Changing-Look Active Galactic Nuclei (CL AGNs) with the aim of understanding the conditions responsible for triggering CL activity. We find that CL AGN hosts primarily reside in the so-called green valley that is located between spiral-like star-forming galaxies and dead ellipticals, implying that CL AGNs are activated during distinct periods of quenching and galaxy transformation processes. CL AGN hosts have low galaxy asymmetry indicators, suggesting that secular evolutionary processes (the influence of bars and spirals, and possibly minor mergers) might be the primary mechanism for transporting gas to the vicinity of the supermassive black hole (SMBH) rather than major mergers. Similar to tidal disruption events (TDEs) and highly variable AGNs, we find that CL AGN hosts are associated with SMBHs residing in high density pseudo-bulges and appear to overlap most significantly with the population of low-ionization nuclear emission-line region (LINER) galaxies. As such, CL AGN are likely fueled by strong episodic bursts of accretion activity, which appear to take place preferentially as the amount of material accessible for star formation and accretion dwindles. We also identify that CL AGN hosts are characterized by either large S\'ersic indices or high bulge fractions, which suggests a simple metric for identifying candidates for spectroscopic follow-up observations in forthcoming synoptic surveys., Comment: 17 pages, 9 figures. Accepted to ApJ Letters. Revised version includes an expanded discussion on asymmetry measurements and galaxy disturbances

There have been many attempts to construct de Sitter space-times in string theory. While arguably there have been some successes, this has proven challenging, leading to the de Sitter swampland conjecture: quantum theories of gravity do not admit stable or metastable de Sitter space. Here we explain that, within controlled approximations, one lacks the tools to construct de Sitter space in string theory. Such approximations would require the existence of a set of (arbitrarily) small parameters, subject to severe constraints. But beyond this one also needs an understanding of big-bang and big-crunch singularities that is not currently accessible to standard approximations in string theory. The existence or non-existence of metastable de Sitter space in string theory remains a matter of conjecture., Comment: 20 pages. v2: references added; v3, v4: typos fixed

We present the STARS library, a grid of tidal disruption event (TDE) simulations interpolated to provide the mass fallback rate ($dM/dt$) to the black hole for a main-sequence star of any stellar mass, stellar age, and impact parameter. We use a one-dimensional stellar evolution code to construct stars with accurate stellar structures and chemical abundances, then perform tidal disruption simulations in a three-dimensional adaptive-mesh hydrodynamics code with a Helmholtz equation of state, in unprecedented resolution: from 131 to 524 cells across the diameter of the star. The interpolated library of fallback rates is available on GitHub (https://github.com/jamielaw-smith/STARS_library) and version 1.0.0 is archived on Zenodo; one can query the library for any stellar mass, stellar age, and impact parameter. We provide new fitting formulae for important disruption quantities ($\beta_{\rm crit}, \Delta M, \dot M_{\rm peak}, t_{\rm peak}, n_\infty$) as a function of stellar mass, stellar age, and impact parameter. Each of these quantities vary significantly with stellar mass and stellar age, but we are able to reduce all of our simulations to a single relationship that depends only on stellar structure, characterized by a single parameter $\rho_c/\bar\rho$, and impact parameter $\beta$. We also find that, in general, more centrally concentrated stars have steeper $dM/dt$ rise slopes and shallower decay slopes. For the same $\Delta M$, the $dM/dt$ shape varies significantly with stellar mass, promising the potential determination of stellar properties from the TDE light curve alone. The $dM/dt$ shape depends strongly on stellar structure and to a certain extent stellar mass, meaning that fitting TDEs using this library offers a better opportunity to determine the nature of the disrupted star and the black hole., Comment: 35 pages, 23 figures. Published in ApJ. v2: added Figure 23, references added, minor edits to text; v3: matches published version. Interpolated library of fallback rates is available at https://github.com/jamielaw-smith/STARS_library

Tidal disruption events (TDEs) are among the brightest transients in the optical, ultraviolet, and X-ray sky. These flares are set into motion when a star is torn apart by the tidal field of a massive black hole, triggering a chain of events which is -- so far -- incompletely understood. However, the disruption process has been studied extensively for almost half a century, and unlike the later stages of a TDE, our understanding of the disruption itself is reasonably well converged. In this Chapter, we review both analytical and numerical models for stellar tidal disruption. Starting with relatively simple, order-of-magnitude physics, we review models of increasing sophistication, the semi-analytic ``affine formalism,'' hydrodynamic simulations of the disruption of polytropic stars, and the most recent hydrodynamic results concerning the disruption of realistic stellar models. Our review surveys the immediate aftermath of disruption in both typical and more unusual TDEs, exploring how the fate of the tidal debris changes if one considers non-main sequence stars, deeply penetrating tidal encounters, binary star systems, and sub-parabolic orbits. The stellar tidal disruption process provides the initial conditions needed to model the formation of accretion flows around quiescent massive black holes, and in some cases may also lead to directly observable emission, for example via shock breakout, gravitational waves or runaway nuclear fusion in deeply plunging TDEs., Comment: Review chapter in book: 'The Tidal Disruption of Stars by Massive Black Holes', Space Science Reviews, Springer. Comments welcome

We present the multi-wavelength analysis of the tidal disruption event (TDE) AT~2018hyz (ASASSN-18zj). From follow-up optical spectroscopy, we detect the first unambiguous case of resolved double-peaked Balmer emission in a TDE. The distinct line profile can be well-modelled by a low eccentricity ($e\approx0.1$) accretion disk extending out to $\sim$100 $R_{p}$ and a Gaussian component originating from non-disk clouds, though a bipolar outflow origin cannot be completely ruled out. Our analysis indicates that in AT~2018hyz, disk formation took place promptly after the most-bound debris returned to pericenter, which we estimate to be roughly tens of days before the first detection. Redistribution of angular momentum and mass transport, possibly through shocks, must occur on the observed timescale of about a month to create the large \Ha-emitting disk that comprises $\lesssim$5\% of the initial stellar mass. With these new insights from AT~2018hyz, we infer that circularization is efficient in at least some, if not all optically-bright, X-ray faint TDEs. In these efficiently circularized TDEs, the detection of double-peaked emission depends on the disk inclination angle and the relative strength of the disk contribution to the non-disk component, possibly explaining the diversity seen in the current sample., Comment: 24 pages, 8 figures, 6 tables. Accepted for publication in ApJ

Recent studies of Tidal Disruption Events (TDEs) have revealed unexpected correlations between the TDE rate and the large-scale properties of the host galaxies. In this review, we present the host galaxy properties of all TDE candidates known to date and quantify their distributions. We consider throughout the differences between observationally-identified types of TDEs and differences from spectroscopic control samples of galaxies. We focus here on the black hole and stellar masses of TDE host galaxies, their star formation histories and stellar populations, the concentration and morphology of the optical light, the presence of AGN activity, and the extra-galactic environment of the TDE hosts. We summarize the state of several possible explanations for the links between the TDE rate and host galaxy type. We present estimates of the TDE rate for different host galaxy types and quantify the degree to which rate enhancement in some types results in rate suppression in others. We discuss the possibilities for using TDE host galaxies to assist in identifying TDEs in upcoming large transient surveys and possibilities for TDE observations to be used to study their host galaxies., Comment: 55 pages, 13 Figures, 7 Tables. Accepted for publication in Springer Space Science Reviews. Chapter in ISSI review "The Tidal Disruption of Stars by Massive Black Holes" vol. 79

We present the first simulations of the tidal disruption of stars with realistic structures and compositions by massive black holes (BHs). We build stars in the stellar evolution code MESA and simulate their disruption in the 3D adaptive-mesh hydrodynamics code FLASH, using an extended Helmholtz equation of state and tracking 49 elements. We study the disruption of a 1$M_\odot$ star and 3$M_\odot$ star at zero-age main sequence (ZAMS), middle-age, and terminal-age main sequence (TAMS). The maximum BH mass for tidal disruption increases by a factor of $\sim$2 from stellar radius changes due to MS evolution; this is equivalent to varying BH spin from 0 to 0.75. The shape of the mass fallback rate curves is different from the results for polytropes of Guillochon & Ramirez-Ruiz (2013). The peak timescale $t_{\rm peak}$ increases with stellar age, while the peak fallback rate $\dot M_{\rm peak}$ decreases with age, and these effects diminish with increasing impact parameter $\beta$. For a $\beta=1$ disruption of a 1$M_\odot$ star by a $10^6 M_\odot$ BH, from ZAMS to TAMS, $t_{\rm peak}$ increases from 30 to 54 days, while $\dot M_{\rm peak}$ decreases from 0.66 to 0.14 $M_\odot$/yr. Compositional anomalies in nitrogen, helium, and carbon can occur before the peak timescale for disruptions of MS stars, which is in contrast to predictions from the "frozen-in" model. More massive stars can show stronger anomalies at earlier times, meaning that compositional constraints can be key in determining the mass of the disrupted star. The abundance anomalies predicted by these simulations provide a natural explanation for the spectral features and varying line strengths observed in tidal disruption events., Comment: 9 pages, 5 figures. Accepted for publication in ApJ Letters