Search

An Early Enrichment Population (EEP) has been theorized to produce the observed metallicity in the intracluster medium (ICM) of galaxy clusters. This population likely existed at high redshifts (z$\sim$10), relics of which we posit exist today as dwarf galaxies. Previous work argues that the initial mass function (IMF) of the EEP must be flatter than those found at lower redshifts, but with considerable uncertainties. In this work, we present a more quantitative model for the EEP and demonstrate how observational constraints can be applied to the IMF using supernova Type Ia (SNIa) rates, delay time distribution (DTD), and the luminosity function (LF) of galaxy clusters. We determine best-fit values for the slope and mass break of the IMF by comparing IMFs from literature with observed DTDs and the low-luminosity component ($M(R)< -12$) of the Coma LF. We derive two best-fit IMFs: (1) $\alpha_{lo} = -0.13 \pm 0.24$ for $0.07 < M/M_\odot < 1.75$ and $\alpha_{hi} = 0.53 \pm 0.01$ for $1.75 < M/M_\odot < 150$, and (2) $\alpha_{lo} = 1.06 \pm 0.11$ for $0.07 < M/M_\odot < 6$ and $\alpha_{hi} = 0.53 \pm 0.01$ for $6 < M/M_\odot < 150$. We also compare these with sl-5 from Loewenstein (2013) with $\alpha=0.5$ for $0.07 < M/M_\odot < 8$ and $\alpha=0.3$ for $8 < M/M_\odot < 150$. This EEP model, along with stars formed at later times, can produce the observed metallicity, is consistent with other observations, and predicts a significant rise in the SNIa rate at increasing redshift.

Electronic health records contain detailed information about the medical condition of patients, but they are difficult for patients to understand even if they have access to them. We explore whether ChatGPT (GPT 4) can help explain multidisciplinary team (MDT) reports to colorectal and prostate cancer patients. These reports are written in dense medical language and assume clinical knowledge, so they are a good test of the ability of ChatGPT to explain complex medical reports to patients. We asked clinicians and lay people (not patients) to review explanations and responses of ChatGPT. We also ran three focus groups (including cancer patients, caregivers, computer scientists, and clinicians) to discuss output of ChatGPT. Our studies highlighted issues with inaccurate information, inappropriate language, limited personalization, AI distrust, and challenges integrating large language models (LLMs) into clinical workflow. These issues will need to be resolved before LLMs can be used to explain complex personal medical information to patients., Comment: under review

Predicting drug efficacy and safety in vivo requires information on biological responses (e.g., cell morphology and gene expression) to small molecule perturbations. However, current molecular representation learning methods do not provide a comprehensive view of cell states under these perturbations and struggle to remove noise, hindering model generalization. We introduce the Information Alignment (InfoAlign) approach to learn molecular representations through the information bottleneck method in cells. We integrate molecules and cellular response data as nodes into a context graph, connecting them with weighted edges based on chemical, biological, and computational criteria. For each molecule in a training batch, InfoAlign optimizes the encoder's latent representation with a minimality objective to discard redundant structural information. A sufficiency objective decodes the representation to align with different feature spaces from the molecule's neighborhood in the context graph. We demonstrate that the proposed sufficiency objective for alignment is tighter than existing encoder-based contrastive methods. Empirically, we validate representations from InfoAlign in two downstream tasks: molecular property prediction against up to 19 baseline methods across four datasets, plus zero-shot molecule-morphology matching., Comment: 21 pages, 8 tables, 7 figures

JWST is uncovering the properties of ever increasing numbers of galaxies at z>6, during the epoch of reionization. Connecting these observed populations to the process of reionization requires understanding how efficiently they produce Lyman continuum (LyC) photons and what fraction (fesc) of these photons escape into the intergalactic medium. By applying the Cox proportional hazards model, a survival analysis technique, to the Low-redshift Lyman Continuum Survey (LzLCS), we develop new, empirical, multivariate predictions for fesc. The models developed from the LzLCS reproduce the observed fesc for z~3 samples, which suggests that LyC emitters may share similar properties at low and high redshift. Our best-performing models for the z~3 galaxies include information about dust attenuation, ionization, and/or morphology. We then apply these models to z$\gtrsim$6 galaxies. For large photometric samples, we find a median predicted fesc=0.047-0.14. For smaller spectroscopic samples, which may include stronger emission line galaxies, we find that $\geq$33% of the galaxies have fesc >0.2, and we identify several candidate extreme leakers with fesc $\geq$0.5. The current samples show no strong trend between predicted fesc and UV magnitude, but limited spectroscopic information makes this result uncertain. Multivariate predictions can give significantly different results from single variable predictions, and the predicted fesc for high-redshift galaxies can differ significantly depending on whether star formation rate surface density or radius is used as a measure of galaxy morphology. We provide all parameters necessary to predict fesc for additional samples of high-redshift galaxies using these models., Comment: Accepted for publication in ApJ. 33 pages, 9 figures, 10 tables, plus appendix

To understand how galaxies reionized the universe, we must determine how the escape fraction of Lyman Continuum (LyC) photons (fesc) depends on galaxy properties. Using the z~0.3 Low-redshift Lyman Continuum Survey (LzLCS), we develop and analyze new multivariate predictors of fesc. These predictions use the Cox proportional hazards model, a survival analysis technique that incorporates both detections and upper limits. Our best model predicts the LzLCS fesc detections with a root-mean-square (RMS) scatter of 0.31 dex, better than single-variable correlations. According to ranking techniques, the most important predictors of fesc are the equivalent width (EW) of Lyman-series absorption lines and the UV dust attenuation, which track line-of-sight absorption due to HI and dust. The HI absorption EW is uniquely crucial for predicting fesc for the strongest LyC emitters, which show properties similar to weaker LyC emitters and whose high fesc may therefore result from favorable orientation. In the absence of HI information, star formation rate surface density ($\Sigma_{\rm SFR}$) and [O III]/[O II] ratio are the most predictive variables and highlight the connection between feedback and fesc. We generate a model suitable for z>6, which uses only the UV slope, $\Sigma_{\rm SFR}$, and [O III]/[O II]. We find that $\Sigma_{\rm SFR}$ is more important in predicting fesc at higher stellar masses, whereas [O III]/[O II] plays a greater role at lower masses. We also analyze predictions for other parameters, such as the ionizing-to-non ionizing flux ratio and Ly=alpha escape fraction. These multivariate models represent a promising tool for predicting fesc at high redshift., Comment: Accepted for publication in ApJ. 34 pages + appendix, 12 figures

Drug-target interaction (DTI) prediction is crucial for identifying new therapeutics and detecting mechanisms of action. While structure-based methods accurately model physical interactions between a drug and its protein target, cell-based assays such as Cell Painting can better capture complex DTI interactions. This paper introduces MOTI$\mathcal{VE}$, a Morphological cOmpound Target Interaction Graph dataset that comprises Cell Painting features for $11,000$ genes and $3,600$ compounds along with their relationships extracted from seven publicly available databases. We provide random, cold-source (new drugs), and cold-target (new genes) data splits to enable rigorous evaluation under realistic use cases. Our benchmark results show that graph neural networks that use Cell Painting features consistently outperform those that learn from graph structure alone, feature-based models, and topological heuristics. MOTI$\mathcal{VE}$ accelerates both graph ML research and drug discovery by promoting the development of more reliable DTI prediction models. MOTI$\mathcal{VE}$ resources are available at https://github.com/carpenter-singh-lab/motive.

We propose a type of phase transition in quantum many-body systems, which occurs in highly excited quantum many-body scar states, while the rest of the spectrum is largely unaffected. Such scar state phase transitions can be realized by embedding a matrix product state, known to undergo a phase transition, as a scar state into the thermal spectrum of a parent Hamiltonian. We find numerically that the mechanism for the scar state phase transition involves the formation or presence of low-entropy states at energies similar to the scar state in the vicinity of the phase transition point., Comment: 5 pages, 2 figures

High-content image-based assays have fueled significant discoveries in the life sciences in the past decade (2013-2023), including novel insights into disease etiology, mechanism of action, new therapeutics, and toxicology predictions. Here, we systematically review the substantial methodological advancements and applications of Cell Painting. Advancements include improvements in the Cell Painting protocol, assay adaptations for different types of perturbations and applications, and improved methodologies for feature extraction, quality control, and batch effect correction. Moreover, machine learning methods recently surpassed classical approaches in their ability to extract biologically useful information from Cell Painting images. Cell Painting data have been used alone or in combination with other -omics data to decipher the mechanism of action of a compound, its toxicity profile, and many other biological effects. Overall, key methodological advances have expanded the ability of Cell Painting to capture cellular responses to various perturbations. Future advances will likely lie in advancing computational and experimental techniques, developing new publicly available datasets, and integrating them with other high-content data types., Comment: Supplementary Table/Code here: https://github.com/srijitseal/CellPainting_SystematicReview

The eigenstate thermalization hypothesis describes how isolated many-body quantum systems reach thermal equilibrium. However, quantum many-body scars and Hilbert space fragmentation violate this hypothesis and cause nonthermal behavior. We demonstrate that Hilbert space fragmentation may arise from lattice geometry in a spin-1/2 model that conserves the number of domain walls. We generalize a known, one-dimensional, scarred model to larger dimensions and show that this model displays Hilbert space fragmentation on the Vicsek fractal lattice and the two-dimensional lattice. Using Monte Carlo methods, the model is characterized as strongly fragmented on the Vicsek fractal lattice when the number of domain walls is either small or close to the maximal value. On the two-dimensional lattice, the model is strongly fragmented when the density of domain walls is low and weakly fragmented when the density of domain walls is high. Furthermore, we show that the fragmentation persists at a finite density of domain walls in the thermodynamic limit for the Vicsek fractal lattice and the two-dimensional lattice. We also show that the model displays signatures similar to Hilbert space fragmentation on a section of the second-generation hexaflake fractal lattice and a modified two-dimensional lattice. We study the autocorrelation function of local observables and demonstrate that the model displays nonthermal dynamics., Comment: 16 pages, 9 figures

As the nearest confirmed Lyman continuum (LyC) emitter, Haro 11 is an exceptional laboratory for studying LyC escape processes crucial to cosmic reionization. Our new HST/COS G130M/1055 observations of its three star-forming knots now reveal that the observed LyC originates in Knots B and C, with $903 - 912~\r{A}$ luminosities of $1.9\pm1.5 \times 10^{40}~\rm erg~s^{-1}$ and $0.9\pm0.7 \times 10^{40}~\rm erg~s^{-1}$, respectively. We derive local escape fractions $f_{\rm{esc, 912}} = 3.4\pm2.9\%$ and $5.1\pm4.3\%$ for Knots B and C, respectively. Our Starburst99 modeling shows dominant populations on the order of $\sim1-4$ Myr and $1-2\times10^7 \rm~M_\odot$ in each knot, with the youngest population in Knot B. Thus, the knot with the strongest LyC detection has the highest LyC production. However, LyC escape is likely less efficient in Knot B than in Knot C due to higher neutral gas covering. Our results therefore stress the importance of the intrinsic ionizing luminosity, and not just the escape fraction, for LyC detection. Similarly, the Ly$\alpha$ escape fraction does not consistently correlate with LyC flux, nor do narrow Ly$\alpha$ red peaks. High observed Ly$\alpha$ luminosity and low Ly$\alpha$ peak velocity separation, however, do correlate with higher LyC escape. Another insight comes from the undetected Knot A, which drives the Green Pea properties of Haro 11. Its density-bounded conditions suggest highly anisotropic LyC escape. Finally, both of the LyC-leaking Knots, B and C, host ultra-luminous X-ray sources (ULXs). While stars strongly dominate over the ULXs in LyC emission, this intriguing coincidence underscores the importance of unveiling the role of accretors in LyC escape and reionization., Comment: 22 pages, 5 figures. Accepted for publication in The Astrophysical Journal

Hilbert space fragmentation provides a mechanism to break ergodicity in closed many-body systems. Here, we propose a realistic scheme to comprehensively explore this exotic paradigm on a Rydberg quantum simulator. We show that the Rydberg Ising model in the large detuning regime can be mapped to a generalized folded XXZ model featuring a strongly fragmented Hilbert space. The emergent Hamiltonian, however, displays distinct time scales for the transport of a magnon and a hole excitation. This interesting property facilitates a continuous tuning of the Krylov-subspace ergodicity, from the integrable regime, to the Krylov-restricted thermal phase, and eventually to the statistical bubble localization region. By further introducing nonlocal interactions, we find that both the fragmentation behavior and the ergodicity of the Krylov subspace can be significantly enriched. We also examine the role of atomic position disorders and identify a symmetry-selective many-body localization transition. We demonstrate that these phenomena manifest themselves in quench dynamics, which can be readily probed in state-of-the-art Rydberg array setups., Comment: 13 pages, 12 figures

Low-mass galaxies can significantly contribute to reionization due to their potentially high Lyman continuum (LyC) escape fraction and relatively high space density. We present a constraint on the LyC escape fraction from low-mass galaxies at z = 1.3 - 3.0. We obtained rest-frame UV continuum imaging with the ACS/SBC and the WFC3/UVIS from the Hubble Space Telescope for eight strongly-lensed galaxies that were identified in the Sloan Giant Arc Survey (SGAS) and the Cluster Lensing And Supernova survey with Hubble (CLASH). The targeted galaxies were selected to be spectroscopically confirmed, highly magnified, and blue in their UV spectral shapes ($\beta<-1.7$). Our targets include intrinsically low luminosity galaxies down to a magnification-corrected absolute UV magnitude of $M_{\rm UV}\sim-14$. We perform custom-defined aperture photometry to place the most reliable upper limits of LyC escape from our sample. From our observations, we report no significant ($>$$2\sigma$) detections of LyC fluxes, placing 1$\sigma$ upper limits on the absolute LyC escape fractions of 3 - 15%. Our observations do not support the expected increased escape fractions of LyC photons from intrinsically UV faint sources. Considering the highly anisotropic geometry of LyC escape, increasing the sample size of faint galaxies in future LyC observations is crucial., Comment: Submitted to ApJ (16 pages, 6 figures, 5 tables)

Image-based or morphological profiling is a rapidly expanding field wherein cells are "profiled" by extracting hundreds to thousands of unbiased, quantitative features from images of cells that have been perturbed by genetic or chemical perturbations. The Cell Painting assay is the most popular imaged-based profiling assay wherein six small-molecule dyes label eight cellular compartments and thousands of measurements are made, describing quantitative traits such as size, shape, intensity, and texture within the nucleus, cytoplasm, and whole cell (Cimini et al., 2023). We have created the Cell Painting Gallery, a publicly available collection of Cell Painting datasets, with granular dataset descriptions and access instructions. It is hosted by AWS on the Registry of Open Data (RODA). As of January 2024, the Cell Painting Gallery holds 656 terabytes (TB) of image and associated numerical data. It includes the largest publicly available Cell Painting dataset, in terms of perturbations tested (Joint Undertaking for Morphological Profiling or JUMP (Chandrasekaran et al., 2023)), along with many other canonical datasets using Cell Painting, close derivatives of Cell Painting (such as LipocyteProfiler (Laber et al., 2023) and Pooled Cell Painting (Ramezani et al., 2023))., Comment: 9 pages, 1 table

The geometry of the neutral gas in and around galaxies is a key regulator of the escape of ionizing photons. We present the first statistical study aiming at linking the neutral and ionized gas distributions to the Lyman continuum (LyC) escape fraction (fesc(LyC)) in a sample of 22 confirmed LyC leakers and non-leakers at z~0.35 using the Keck Cosmic Web Imager (Keck/KCWI) and the Low Resolution Spectrograph 2 (HET/LRS2). Our integral field unit data enable the detection of neutral and low-ionization gas, as traced by Mg II, and ionized gas, as traced by [O II], extending beyond the stellar continuum for 7 and 10 objects, respectively. All but one object with extended Mg II emission also shows extended [O II] emission; in this case, Mg II emission is always more extended than [O II] by a factor 1.3 on average. Most of the galaxies with extended emission are non or weak LyC leakers (fesc(LyC) < 5%), but we find a large diversity of neutral gas configurations around these weakly LyC-emitting galaxies. Conversely, the strongest leakers (fesc(LyC) > 10%) appear uniformly compact in both Mg II and [O II] with exponential scale lengths <1 kpc. We also find a trend between fesc(LyC) and the spatial offsets of the nebular gas and the stellar continuum emission. Moreover, we find significant anti-correlations between the spatial extent of the neutral gas and the [O III]/[O II] ratio, and H$\beta$ equivalent width, as well as positive correlations with metallicity and UV size, suggesting that galaxies with more compact neutral gas sizes are more highly ionized. The observations suggest that strong LyC emitters do not have extended neutral gas halos and ionizing photons may be emitted in many directions. Combined with high ionization diagnostics, we propose the Mg II, and potentially [O II], spatial compactness are indirect indicators of LyC emitting galaxies at high-redshift., Comment: 31 pages, 24 figures, 4 tables, submitted to A&A

The Next Generation Deep Extragalactic Exploratory Public (NGDEEP) survey program was designed specifically to include Near Infrared Slitless Spectroscopic observations (NGDEEP-NISS) to detect multiple emission lines in as many galaxies as possible and across a wide redshift range using the Near Infrared Imager and Slitless Spectrograph (NIRISS). We present early results obtained from the the first set of observations (Epoch 1, 50$\%$ of the allocated orbits) of this program (NGDEEP-NISS1). Using a set of independently developed calibration files designed to deal with a complex combination of overlapping spectra, multiple position angles, and multiple cross filters and grisms, in conjunction with a robust and proven algorithm for quantifying contamination from overlapping dispersed spectra, NGDEEP-NISS1 has achieved a 3$\sigma$ sensitivity limit of 2 $\times$ 10$^{-18}$ erg/s/cm$^2$. We demonstrate the power of deep wide field slitless spectroscopy (WFSS) to characterize the star-formation rates, and metallicity ([OIII]/H$\beta$), and dust content, of galaxies at $1

We present new spatially resolved astrometry and photometry of the CD-27 11535 system, a member of the $\beta$ Pictoris moving group consisting of two resolved K-type stars on a $\sim$20-year orbit. We fit an orbit to relative astrometry measured from NIRC2, GPI, and archival NaCo images, in addition to literature measurements. However, the total mass inferred from this orbit is significantly discrepant from that inferred from stellar evolutionary models using the luminosity of the two stars. We explore two hypotheses that could explain this discrepant mass sum; a discrepant parallax measurement from Gaia due to variability, and the presence of an additional unresolved companion to one of the two components. We find that the $\sim$20-year orbit could not bias the parallax measurement, but that variability of the components could produce a large amplitude astrometric motion, an effect which cannot be quantified exactly without the individual Gaia measurements. The discrepancy could also be explained by an additional star in the system. We jointly fit the astrometric and photometric measurements of the system to test different binary and triple architectures for the system. Depending on the set of evolutionary models used, we find an improved goodness of fit for a triple system architecture that includes a low-mass ($M=0.177\pm0.055$\,$M_{\odot}$) companion to the primary star. Further studies of this system will be required in order to resolve this discrepancy, either by refining the parallax measurement with a more complex treatment of variability-induced astrometric motion, or by detecting a third companion., Comment: 12 pages + references and appendix, 12 figures, 5 tables

Advances in high-throughput microscopy have enabled the rapid acquisition of large numbers of high-content microscopy images. Whether by deep learning or classical algorithms, image analysis pipelines then produce single-cell features. To process these single-cells for downstream applications, we present Pycytominer, a user-friendly, open-source python package that implements the bioinformatics steps, known as image-based profiling. We demonstrate Pycytominers usefulness in a machine learning project to predict nuisance compounds that cause undesirable cell injuries., Comment: We updated: Figures (e.g., remove panel from Figure 1) to increase clarity. Consolidated the introduction, results, and discussion into a single section. Added a new analysis to predict compounds that cause undesirable cell injuries. Added three tables including one to highlight image-based profiling software limitations. 14 pages, 2 main figures, 5 supplementary figures, 3 tables

Sources that leak Lyman-continuum (LyC) photons and lead to the reionisation of the universe are intensely studied using multiple observing facilities. Recently, the Low-redshift LyC Survey (LzLCS) has found the first large sample of LyC emitting galaxies at low redshift ($z\sim 0.3$) with the Hubble Space Telescope/Cosmic Origins Spectrograph. The LzLCS sample contains a robust estimate of the LyC escape fraction ($f_\mathrm{esc}^\mathrm{LyC}$) for 66 galaxies spanning a wide range of $f_\mathrm{esc}^\mathrm{LyC}$. Here we, for the first time, aim to study the radio continuum (RC) properties of LzLCS sources and their dependence on $f_\mathrm{esc}^\mathrm{LyC}$. We present Karl G. Jansky Very Large Array RC observations at C (4-8 GHz), S (2-4 GHz) and L (1-2 GHz) bands for a sub-sample of the LzLCS sources. The radio spectral index ($\alpha^{\mathrm{3GHz}}_\mathrm{6GHz}$) spans a wide range from being flat ( $\geq -0.1$) to very steep ($\leq -1.0$). The strongest leakers in our sample show flat $\alpha^{\mathrm{3GHz}}_\mathrm{6GHz}$, weak leakers have $\alpha^{\mathrm{3GHz}}_\mathrm{6GHz}$ close to normal star-forming galaxies, and non-leakers are characterized by steep $\alpha^{\mathrm{3GHz}}_\mathrm{6GHz}$. We argue that a combination of young ages, free-free absorption, and a flat cosmic-ray energy spectrum can altogether lead to a flat $\alpha^{\mathrm{3GHz}}_\mathrm{6GHz}$ for strong leakers. Non-leakers are characterized by steep spectra which can arise due to break/cutoff at high frequencies. Such a cutoff in the spectrum can arise in a single injection model of CRs characteristic of galaxies which have recently stopped star formation. Such a relation between $\alpha^{\mathrm{3GHz}}_\mathrm{6GHz}$ and $f_\mathrm{esc}^\mathrm{LyC}$ hints at the interesting role of supernovae, CRs, and magnetic fields in facilitating the escape ( and/or the lack) of LyC photons. (Abridged), Comment: 20 pages, 17 figures, 4 tables, 8 appendix pages, Astronomy & Astrophysics in press

We use the Hubble Space Telescope ACS camera to obtain the first spatially resolved, nebular imaging in the light of C IV 1548,1551 by using the F150LP and F165LP filters. These observations of the local starburst Mrk 71 in NGC 2366 show emission apparently originating within the interior cavity around the dominant super star cluster (SSC), Knot A. Together with imaging in He II 4686 and supporting STIS FUV spectroscopy, the morphology and intensity of the C IV nebular surface brightness and the C IV / He II ratio map provide direct evidence that the mechanical feedback is likely dominated by catastrophic radiative cooling, which strongly disrupts adiabatic superbubble evolution. The implied extreme mass loading and low kinetic efficiency of the cluster wind are reasonably consistent with the wind energy budget, which is probably enhanced by radiation pressure. In contrast, the Knot B SSC lies within a well-defined superbubble with associated soft X-rays and He II 1640 emission, which are signatures of adiabatic, energy-driven feedback from a supernova-driven outflow. This system lacks clear evidence of C IV from the limb-brightened shell, as expected for this model, but the observations may not be deep enough to confirm its presence. We also detect a small C IV-emitting object that is likely an embedded compact H II region. Its C IV emission may indicate the presence of very massive stars (> 100 M_sun) or strongly pressure-confined stellar feedback., Comment: 13 pages, 8 figures. Accepted to ApJ Letters

Mrk 71 is a low metallicity (Z = 0.16 Z_sun) starburst region in the local dwarf galaxy NGC 2366, hosting two super star clusters (SSCs A and B), and is recognized as a Green Pea (GP) analog with SSC A responsible for the GP properties. We present STIS and FOS far-ultraviolet (FUV) spectra of the embedded SSC Mrk 71-A obtained with the Hubble Space Telescope (HST). The STIS FUV spectrum shows the characteristic features of very massive stars (VMS, masses > 100 M_sun) and we derive an age of 1+/-1 Myr by comparison with the Charlot & Bruzual suite of spectral population synthesis models with upper mass limits of 300 and 600 M_sun. We compare the STIS spectrum with all known SSC spectra exhibiting VMS signatures: NGC 5253-5, R136a, NGC 3125-A1 and the z = 2.37 Sunburst cluster. We find that the cluster mass-loss rates and wind velocities, as characterized by the C IV P Cygni profiles and the He II emission line strengths, are very similar over Z = 0.16 to 0.4 Z_sun. This agrees with predictions that the optically thick winds of VMS will be enhanced near the Eddington limit and show little metallicity dependence. We find very strong damped Lyman-alpha absorption with log N(H I) = 22.2 cm-2 associated with Mrk 71-A. We discuss the natal environment of this young SSC in terms of radiatively-driven winds, catastrophic cooling and recent models where the cluster is surrounded by highly pressurized clouds with large neutral columns., Comment: 15 pages accepted for publication in the Astrophysical Journal