arXiv subjects

Colin Jacobs

Publications and source records attributed to Colin Jacobs.

At least 19 recordsLinked to original sources

arXiv2026

OutThere Survey: Addressing $\mathrm{\xi_{ion}}$ and $\mathrm{f_{esc}}$ with a population of average galaxies at z$\sim$2

Constraining the major contributors to the ionisation of the early universe is an ongoing endeavour of high-redshift galaxy research. We measure the ionising photon production efficiency and Lyman Continuum escape fraction for a sample of 230 intermediate redshift ($1.3 $5. This control sample allows us to verify the correlations between ionising and spectral/physical properties suggested by previous studies. We find no significant correlations between the ionising photon production efficiency ($\mathrm{\xi_{ion}}$) with the UV slope, $\mathrm{M_{UV}}$, M$_*$ or sSFR. We do find that $\mathrm{\xi_{ion}}$ correlates with [OIII]5007\AA\, equivalent width (EW) (Spearman coefficient $\rho$ =0.24; p$< 4\times10^{-4}$) and H$\alpha$ EW ($\rho$ =0.63; p$<< 1\times10^{-6}$) hold even at low EW albeit with more scatter. We also find that our novel approach to determining the ionising photon escape fraction $\mathrm{f_{esc}}$ results in values within theoretical ranges (0-10\%) though vary substantially in comparison to the empirical results (median $\mathrm{f_{esc}} = 0.9\%^{+1.1}_{-0.5}$ including non-detections, median $\mathrm{f_{esc}} = 1.9\%^{+8.9}_{-1.8}$ above a $0.01\%$ threshold). We find that this escape fraction method has consistently significant correlations with the redshift, SFR and M$_{UV}$ and sample-dependent correlations with [OIII]5007\AA\,EW,H$\alpha$ EW and stellar mass.

astro-ph.GA↗

arXiv2026

Designing UNICORN: a Unified Benchmark for Imaging in Computational Pathology, Radiology, and Natural Language

Medical foundation models show promise to learn broadly generalizable features from large, diverse datasets. This could be the base for reliable cross-modality generalization and rapid adaptation to new, task-specific goals, with only a few task-specific examples. Yet, evidence for this is limited by the lack of public, standardized, and reproducible evaluation frameworks, as existing public benchmarks are often fragmented across task-, organ-, or modality-specific settings, limiting assessment of cross-task generalization. We introduce UNICORN, a public benchmark designed to systematically evaluate medical foundation models under a unified protocol. To isolate representation quality, we built the benchmark on a novel two-step framework that decouples model inference from task-specific evaluation based on standardized few-shot adaptation. As a central design choice, we constructed indirectly accessible sequestered test sets derived from clinically relevant cohorts, along with standardized evaluation code and a submission interface on an open benchmarking platform. Performance is aggregated into a single UNICORN Score, a new metric that we introduce to support direct comparison of foundation models across diverse medical domains, modalities, and task types. The UNICORN test dataset includes data from more than 2,400 patients, including over 3,700 vision cases and over 2,400 clinical reports collected from 17 institutions across eight countries. The benchmark spans eight anatomical regions and four imaging modalities. Both task-specific and aggregated leaderboards enable accessible, standardized, and reproducible evaluation. By standardizing multi-task, multi-modality assessment, UNICORN establishes a foundation for reproducible benchmarking of medical foundation models. Data, baseline methods, and the evaluation platform are publicly available via unicorn.grand-challenge.org.

cs.CV↗

arXiv2026

Multi-band Reconstruction of Sixteen Gravitational Lens Systems using PISCO data

Next-generation surveys such as the Euclid survey, the Legacy Survey of Space and Time (LSST), and the China Space Station Telescope (CSST) survey are expected to discover ~10^5 galaxy-galaxy scale strong gravitational lenses. This motivates the development of scalable and robust lens modeling approaches that can efficiently and reliably learn from wide-field survey datasets before high-resolution follow-up. We design a scalable, Bayesian, Lenstronomy-based pipeline and apply it to a sample of sixteen lens candidates observed with the Parallel Imager for Southern Cosmology Observations (PISCO) on the Magellan telescope. PISCO provides four-band imaging (z, i, r, g) with colours, depth and seeing conditions comparable to LSST. To fully exploit the constraining power of this dataset, our pipeline performs simultaneous multi-band modeling, using a common mass profile across all four bands while allowing independent light profiles in each. This approach leverages color information to provide joint constraints on the lens mass and yields reduced uncertainties compared to single-band analyses. Fifteen out of sixteen PISCO lens candidates are successfully recovered with interpretable lensing configurations, including DESJ0533-2536, the first reported hyperbolic-umbilic galaxy-galaxy scale strong lensing candidate. We further assess how much model complexity can be reliably constrained given the resolution and seeing of PISCO-like data. Overall, our results demonstrate that scalable, multi-band lens modeling of ground-based data can extract meaningful constraints on mass and source morphology, providing a practical pathway to maximize the scientific return from large samples in upcoming surveys.

astro-ph.GA↗

arXiv2025

Fairness Evaluation of Risk Estimation Models for Lung Cancer Screening

Lung cancer is the leading cause of cancer-related mortality in adults worldwide. Screening high-risk individuals with annual low-dose CT (LDCT) can support earlier detection and reduce deaths, but widespread implementation may strain the already limited radiology workforce. AI models have shown potential in estimating lung cancer risk from LDCT scans. However, high-risk populations for lung cancer are diverse, and these models' performance across demographic groups remains an open question. In this study, we drew on the considerations on confounding factors and ethically significant biases outlined in the JustEFAB framework to evaluate potential performance disparities and fairness in two deep learning risk estimation models for lung cancer screening: the Sybil lung cancer risk model and the Venkadesh21 nodule risk estimator. We also examined disparities in the PanCan2b logistic regression model recommended in the British Thoracic Society nodule management guideline. Both deep learning models were trained on data from the US-based National Lung Screening Trial (NLST), and assessed on a held-out NLST validation set. We evaluated AUROC, sensitivity, and specificity across demographic subgroups, and explored potential confounding from clinical risk factors. We observed a statistically significant AUROC difference in Sybil's performance between women (0.88, 95% CI: 0.86, 0.90) and men (0.81, 95% CI: 0.78, 0.84, p < .001). At 90% specificity, Venkadesh21 showed lower sensitivity for Black (0.39, 95% CI: 0.23, 0.59) than White participants (0.69, 95% CI: 0.65, 0.73). These differences were not explained by available clinical confounders and thus may be classified as unfair biases according to JustEFAB. Our findings highlight the importance of improving and monitoring model performance across underrepresented subgroups, and further research on algorithmic fairness, in lung cancer screening.

cs.LG↗

arXiv2025

Connecting Environment, Star Formation History, and Morphology of Massive Quiescent Galaxies at $3<z<4$ with JWST

We present the morphological properties of 17 spectroscopically confirmed massive quiescent galaxies ($10.2 < \log(M_{\ast}/M_{\odot}) < 11.2$) at $3.0 < z < 4.3$, observed with JWST/NIRSpec and NIRCam. Using S\'ersic profile fits to F277W and F444W imaging, we derive the size-mass relation and find typical sizes of $\sim$0.6--0.8 kpc at $M_{\ast} = 5 \times 10^{10}~M_{\odot}$, consistent with $\sim$7$\times$ growth from $z \sim 4$ to the present, including $\sim$2$\times$ by $z \sim 2$. We find tentative evidence that formation history and morphology jointly influence galaxy sizes: late-forming bulge-dominated galaxies appear more compact by $\sim$0.2-0.3 dex relative to the expected relation, while late-forming disk-dominated galaxies are larger. Using a random forest regressor, we identify local environmental density, quantified by $\log(1+\delta^{\prime}_{3})$ from the three nearest neighbors, as the strongest predictor of bulge-to-total ratio ($B/T$), which spans 0.25-1. In the {\sc IllustrisTNG} simulation, the ex-situ stellar mass fraction ($f_{\ast,\mathrm{ex\text{-}situ}}$) -- a proxy for mergers -- is instead the dominant predictor of $B/T$. Galaxies with high $B/T$ in dense environments show bursty star formation and short quenching timescales ($\lesssim0.4$ Gyr), consistent with bulge growth through merger-driven starbursts; in simulations, such systems exhibit elevated ex-situ fractions ($\sim$20-30%). In contrast, some high-$B/T$ galaxies in intermediate-density environments have low ex-situ fractions, suggesting that additional processes -- such as violent disk instabilities -- also contribute. These results point to multiple bulge growth pathways at high redshift, unified by rapid gas accretion, central starbursts, and AGN feedback, as predicted by cosmological simulations.

astro-ph.GA↗

arXiv2025

The AGEL Survey Data Release 2: A Gravitational Lens Sample for Galaxy Evolution and Cosmology

The ASTRO 3D Galaxy Evolution with Lenses (AGEL) Survey is an ongoing effort to spectroscopically confirm a diverse sample of gravitational lenses with high spatial resolution imaging, to facilitate a broad range of science outcomes. The AGEL systems span single galaxy-scale deflectors to groups and clusters, and include rare targets such as galaxy-scale lenses with multiple sources, lensed quiescent galaxies, and Einstein rings. We build on the 77 systems presented in Tran et al. 2022 (AGEL data release 1) to present a total 138 lenses, and high resolution F140W and F200LP Hubble Space Telescope images for 71 lenses from a completed HST SNAP program. Lens candidates were originally identified by convolutional neural networks in the DES and DECaLS imaging fields, and of the targets with follow-up spectroscopy we find a high (96%) success rate. Compared with other spectroscopic lens samples, AGEL lenses tend to have both higher redshift deflectors and sources. We briefly discuss the common causes of false-positive candidates, and strategies for mitigating false-positives in next generation lens searches. Lastly, we present 6 galaxy-scale double-source plane lenses useful for cosmological analyses. With next-generation telescopes and surveys such as Euclid, Vera Rubin's Legacy Survey of Space and Time, Keck Observatory's KAPA program, and 4MOST's 4SLSLS surveys on the horizon, the AGEL survey represents a pathfinder for refining automated candidate search methods and identifying and triaging candidates for followup based on scientific potential.

astro-ph.GA↗

arXiv2024

The formation histories of massive and quiescent galaxies in the 3 < z < 4.5 Universe

We present the formation histories of 19 massive ($>3X10^{10}M_\odot$) quiescent (sSFR$<0.15Gyr^{-1}$) galaxy candidates at z~3.0-4.5 observed using JWST/NIRSpec. This completes the spectroscopic confirmation of the 24 K-selected quiescent galaxy sample from the ZFOURGE and 3DHST surveys (Schreiber et al. 2018a). Utilizing Prism $1-5\mu m$ spectroscopy, we confirm that all 12 sources that eluded confirmation by ground-based spectroscopy lie at z>3, resulting in a spectroscopically confirmed number density of $\sim1.4X10^{-5}Mpc^{-3}$ between $z\sim3-4$. Rest-frame U-V vs V-J color selections show high effectiveness in identifying quiescent galaxies, with a purity of~90%. Our analysis shows that parametric star-formation histories (SFHs) from FAST++ and binned SFHs from Prospector on average yield consistent results, revealing diverse formation and quenching times. The oldest galaxy formed $~6X10^{10}M_\odot$ by $z\sim10$ and has been quiescent for over 1 Gyr at z~3.2. We detect two galaxies with ongoing star formation and six with active galactic nuclei (AGN). We demonstrate that the choice of stellar population models, stellar libraries, and nebular or AGN contributions does not significantly affect the derived average SFHs of the galaxies. We demonstrate that extending spectral fitting beyond the rest-frame optical regime reduces the inferred average star formation rates in the earliest time bins of the SFH reconstruction. The assumed SFH prior influences the star formation rate at early times, where spectral diagnostic power is limited. Simulated z~3 quiescent galaxies from IllustrisTNG, SHARK, and Magneticum broadly match the average SFHs of the observed sample but struggle to capture the full diversity, particularly at early stages. Our results emphasize the need for mechanisms that rapidly build stellar mass and quench star formation within the first billion years of the Universe.

astro-ph.GA↗

arXiv2024

Optimization of Approximate Maps for Linear Systems Arising in Discretized PDEs

Generally, discretization of partial differential equations (PDEs) creates a sequence of linear systems $A_k x_k = b_k, k = 0, 1, 2, ..., N$ with well-known and structured sparsity patterns. Preconditioners are often necessary to achieve fast convergence When solving these linear systems using iterative solvers. We can use preconditioner updates for closely related systems instead of computing a preconditioner for each system from scratch. One such preconditioner update is the sparse approximate map (SAM), which is based on the sparse approximate inverse preconditioner using a least squares approximation. A SAM then acts as a map from one matrix in the sequence to another nearby one for which we have an effective preconditioner. To efficiently compute an effective SAM update (i.e., one that facilitates fast convergence of the iterative solver), we seek to compute an optimal sparsity pattern. In this paper, we examine several sparsity patterns for computing the SAM update to characterize optimal or near-optimal sparsity patterns for linear systems arising from discretized PDEs.

math.NA↗

arXiv2024

Impacts of source morphology on the detectability of subhalos in strong lenses

We provide an analysis of a convolutional neural network's ability to identify the lensing signal of single dark matter subhalos in strong galaxy-galaxy lenses in the presence of increasingly complex source light morphology. We simulate a balanced dataset of 800,000 strong lens images both perturbed and unperturbed by a single subhalo ranging in virial mass between $10^{7.5} M_{\odot} - 10^{11}M_{\odot}$ and characterise the source complexity by the number of Sersic clumps present in the source plane ranging from 1 to 5. Using the ResNet50 architecture we train the network to classify images as either perturbed or unperturbed. We find that the network is able to detect subhalos at low enough masses to distinguish between dark matter models even with complex sources and that source complexity has little impact on the accuracy beyond 3 clumps. The model was more confident in its classification when the clumps in the source were compact, but cared little about their spatial distribution. We also tested for the resolution of the data, finding that even in conditions akin to natural seeing the model was still able to achieve an accuracy of 74% in our highest peak signal-to-noise datasets, though this is heavily dominated by the high mass subhalos. It's robustness against resolution is attributed to the model learning the flux ratio anomalies in the perturbed lenses which are conserved in the lower resolution data.

astro-ph.CO↗

arXiv2024

HAYATE: Photometric redshift estimation by hybridising machine learning with template fitting

Machine learning photo-z methods, trained directly on spectroscopic redshifts, provide a viable alternative to traditional template fitting methods but may not generalise well on new data that deviates from that in the training set. In this work, we present a Hybrid Algorithm for WI(Y)de-range photo-z estimation with Artificial neural networks and TEmplate fitting (HAYATE), a novel photo-z method that combines template fitting and data-driven approaches and whose training loss is optimised in terms of both redshift point estimates and probability distributions. We produce artificial training data from low-redshift galaxy SEDs at z<1.3, artificially redshifted up to z=5. We test the model on data from the ZFOURGE surveys, demonstrating that HAYATE can function as a reliable emulator of EAZY for the broad redshift range beyond the region of sufficient spectroscopic completeness. The network achieves precise photo-z estimations with smaller errors ($\sigma_{NMAD}$) than EAZY in the initial low-z region (z<1.3), while being comparable even in the high-z extrapolated regime (1.3<z<5). Meanwhile, it provides more robust photo-z estimations than EAZY with the lower outlier rate ($\eta_{0.2}\lesssim 1\%$) but runs $\sim100$ times faster than the original template fitting method. We also demonstrate HAYATE offers more reliable redshift PDFs, showing a flatter distribution of Probability Integral Transform scores than EAZY. The performance is further improved using transfer learning with spec-z samples. We expect that future large surveys will benefit from our novel methodology applicable to observations over a wide redshift range.

astro-ph.IM↗

arXiv2024

Nodule detection and generation on chest X-rays: NODE21 Challenge

Pulmonary nodules may be an early manifestation of lung cancer, the leading cause of cancer-related deaths among both men and women. Numerous studies have established that deep learning methods can yield high-performance levels in the detection of lung nodules in chest X-rays. However, the lack of gold-standard public datasets slows down the progression of the research and prevents benchmarking of methods for this task. To address this, we organized a public research challenge, NODE21, aimed at the detection and generation of lung nodules in chest X-rays. While the detection track assesses state-of-the-art nodule detection systems, the generation track determines the utility of nodule generation algorithms to augment training data and hence improve the performance of the detection systems. This paper summarizes the results of the NODE21 challenge and performs extensive additional experiments to examine the impact of the synthetically generated nodule training images on the detection algorithm performance.

eess.IV↗

arXiv2023

Transfer learning from a sparsely annotated dataset of 3D medical images

Transfer learning leverages pre-trained model features from a large dataset to save time and resources when training new models for various tasks, potentially enhancing performance. Due to the lack of large datasets in the medical imaging domain, transfer learning from one medical imaging model to other medical imaging models has not been widely explored. This study explores the use of transfer learning to improve the performance of deep convolutional neural networks for organ segmentation in medical imaging. A base segmentation model (3D U-Net) was trained on a large and sparsely annotated dataset; its weights were used for transfer learning on four new down-stream segmentation tasks for which a fully annotated dataset was available. We analyzed the training set size's influence to simulate scarce data. The results showed that transfer learning from the base model was beneficial when small datasets were available, providing significant performance improvements; where fine-tuning the base model is more beneficial than updating all the network weights with vanilla transfer learning. Transfer learning with fine-tuning increased the performance by up to 0.129 (+28\%) Dice score than experiments trained from scratch, and on average 23 experiments increased the performance by 0.029 Dice score in the new segmentation tasks. The study also showed that cross-modality transfer learning using CT scans was beneficial. The findings of this study demonstrate the potential of transfer learning to improve the efficiency of annotation and increase the accessibility of accurate organ segmentation in medical imaging, ultimately leading to improved patient care. We made the network definition and weights publicly available to benefit other users and researchers.

eess.IV↗

arXiv2023

Kidney abnormality segmentation in thorax-abdomen CT scans

In this study, we introduce a deep learning approach for segmenting kidney parenchyma and kidney abnormalities to support clinicians in identifying and quantifying renal abnormalities such as cysts, lesions, masses, metastases, and primary tumors. Our end-to-end segmentation method was trained on 215 contrast-enhanced thoracic-abdominal CT scans, with half of these scans containing one or more abnormalities. We began by implementing our own version of the original 3D U-Net network and incorporated four additional components: an end-to-end multi-resolution approach, a set of task-specific data augmentations, a modified loss function using top-$k$, and spatial dropout. Furthermore, we devised a tailored post-processing strategy. Ablation studies demonstrated that each of the four modifications enhanced kidney abnormality segmentation performance, while three out of four improved kidney parenchyma segmentation. Subsequently, we trained the nnUNet framework on our dataset. By ensembling the optimized 3D U-Net and the nnUNet with our specialized post-processing, we achieved marginally superior results. Our best-performing model attained Dice scores of 0.965 and 0.947 for segmenting kidney parenchyma in two test sets (20 scans without abnormalities and 30 with abnormalities), outperforming an independent human observer who scored 0.944 and 0.925, respectively. In segmenting kidney abnormalities within the 30 test scans containing them, the top-performing method achieved a Dice score of 0.585, while an independent second human observer reached a score of 0.664, suggesting potential for further improvement in computerized methods. All training data is available to the research community under a CC-BY 4.0 license on https://doi.org/10.5281/zenodo.8014289

eess.IV↗

arXiv2023

Emphysema Subtyping on Thoracic Computed Tomography Scans using Deep Neural Networks

Accurate identification of emphysema subtypes and severity is crucial for effective management of COPD and the study of disease heterogeneity. Manual analysis of emphysema subtypes and severity is laborious and subjective. To address this challenge, we present a deep learning-based approach for automating the Fleischner Society's visual score system for emphysema subtyping and severity analysis. We trained and evaluated our algorithm using 9650 subjects from the COPDGene study. Our algorithm achieved the predictive accuracy at 52\%, outperforming a previously published method's accuracy of 45\%. In addition, the agreement between the predicted scores of our method and the visual scores was good, where the previous method obtained only moderate agreement. Our approach employs a regression training strategy to generate categorical labels while simultaneously producing high-resolution localized activation maps for visualizing the network predictions. By leveraging these dense activation maps, our method possesses the capability to compute the percentage of emphysema involvement per lung in addition to categorical severity scores. Furthermore, the proposed method extends its predictive capabilities beyond centrilobular emphysema to include paraseptal emphysema subtypes.

eess.IV↗

arXiv2023

A massive galaxy that formed its stars at $z \sim 11$

The formation of galaxies by gradual hierarchical co-assembly of baryons and cold dark matter halos is a fundamental paradigm underpinning modern astrophysics and predicts a strong decline in the number of massive galaxies at early cosmic times. Extremely massive quiescent galaxies (stellar masses $>10^{11}$ M$_\odot$) have now been observed as early as 1-2 billions years after the Big Bang; these are extremely constraining on theoretical models as they form 300-500 Myr earlier and only some models can form massive galaxies this early. Here we report on the spectroscopic observations with the James Webb Space Telescope of a massive quiescent galaxy ZF-UDS-7329 at redshift 3.205 $\pm$ 0.005 that eluded deep ground-based spectrscopy, is significantly redder than typical and whose spectrum reveals features typical of much older stellar populations. Detailed modelling shows the stellar population formed around 1.5 billion years earlier in time (z ~ 11) at an epoch when dark matter halos of sufficient hosting mass have not yet assembled in the standard scenario. This observation may point to the presence of undetected populations of early galaxies and the possibility of significant gaps in our understanding of early stellar populations, galaxy formation and/or the nature of dark matter.

astro-ph.GA↗

arXiv2022

A population of faint, old, and massive quiescent galaxies at 3 < z < 4 revealed by JWST NIRSpec Spectroscopy

Here we present a sample of 12 massive quiescent galaxy candidates at z~3-4 observed with the James Webb Space Telescope (JWST) Near Infrared Spectrograph (NIRSpec). These galaxies were pre-selected from the Hubble Space Telescope imaging and 10 of our sources were unable to be spectroscopically confirmed by ground based spectroscopy. By combining spectroscopic data from NIRSpec with multi-wavelength imaging data from the JWST Near Infrared Camera (NIRCam), we analyse their stellar populations and their formation histories. We find that all of our galaxies classify as quiescent based on the reconstruction of their star formation histories but show a variety of quenching timescales and ages. All our galaxies are massive ($\sim0.1-1.2 \times 10^{11} M\odot$), with masses comparable to massive galaxies in the local Universe. We find that the oldest galaxy in our sample formed $\sim1.0\times10^{11} M\odot$ of mass within the first few hundred million years of the Universe and has been quenched for more than a billion years by the time of observation at z$\sim$3.2 ($\sim$2 billion years after the Big Bang). Our results point to very early formation of massive galaxies requiring a high conversion rate of baryons to stars in the early Universe.

astro-ph.GA↗

arXiv2022

A Glimpse of the Stellar Populations and Elemental Abundances of Gravitationally Lensed, Quiescent Galaxies at $z\gtrsim 1$ with Keck Deep Spectroscopy

Gravitational lenses can magnify distant galaxies, allowing us to discover and characterize the stellar populations of intrinsically faint, quiescent galaxies that are otherwise extremely difficult to directly observe at high redshift from ground-based telescopes. Here, we present the spectral analysis of two lensed, quiescent galaxies at $z\gtrsim 1$ discovered by the ASTRO 3D Galaxy Evolution with Lenses survey: AGEL1323 ($M_*\sim 10^{11.1}M_{\odot}$, $z=1.016$, $\mu \sim 14.6$) and AGEL0014 ($M_*\sim 10^{11.5}M_{\odot}$, $z=1.374$, $\mu \sim 4.3$). We measured the age, [Fe/H], and [Mg/Fe] of the two lensed galaxies using deep, rest-frame-optical spectra (S/N $\gtrsim 40$~$\mathring {\mathrm A}$$^{-1}$) obtained on the Keck~I telescope. The ages of AGEL1323 and AGEL0014 are $5.6^{+0.8}_{-0.8}$~Gyr and $3.1^{+0.8}_{-0.3}$~Gyr, respectively, indicating that most of the stars in the galaxies were formed less than 2~Gyr after the Big Bang. Compared to nearby quiescent galaxies of similar masses, the lensed galaxies have lower [Fe/H] and [Mg/H]. Surprisingly, the two galaxies have comparable [Mg/Fe] to similar-mass galaxies at lower redshifts, despite their old ages. Using a simple analytic chemical evolution model connecting the instantaneously recycled element Mg with the mass-loading factors of outflows averaged over the entire star formation history, we found that the lensed galaxies may have experienced enhanced outflows during their star formation compared to lower-redshift galaxies, which may explain why they quenched early.

astro-ph.GA↗

arXiv2022

Resolved velocity profiles of galactic winds at Cosmic Noon

We study the kinematics of the interstellar medium (ISM) viewed "down the barrel" in 20 gravitationally lensed galaxies during Cosmic Noon ($z=1.5 - 3.5$). We use moderate-resolution spectra ($R\sim4000$) from Keck/ESI and Magellan/MagE to spectrally resolve the ISM absorption in these galaxies into $\sim$10 independent elements and use double Gaussian fits to quantify the velocity structure of the gas. We find that the bulk motion of gas in this galaxy sample is outflowing, with average velocity centroid $\left =-141$ km$\,$s$^{-1}$ ($\pm111$ km$\,$s$^{-1}$ scatter) measured with respect to the systemic redshift. 16 out of the 20 galaxies exhibit a clear positive skewness, with a blueshifted tail extending to $\sim -500$ km$\,$s$^{-1}$. We examine scaling relations in outflow velocities with galaxy stellar mass and star formation rate (SFR), finding correlations consistent with a momentum-driven wind scenario. Our measured outflow velocities are also comparable to those reported for FIRE-2 and TNG50 cosmological simulations at similar redshift and galaxy properties. We also consider implications for interpreting results from lower-resolution spectra. We demonstrate that while velocity centroids are accurately recovered, the skewness, velocity width, and probes of high velocity gas (e.g., $v_{95}$) are subject to large scatter and biases at lower resolution. We find that $R\gtrsim1700$ is required for accurate results for the gas kinematics of our sample. This work represents the largest available sample of well-resolved outflow velocity structure at $z>2$, and highlights the need for good spectral resolution to recover accurate properties.

astro-ph.GA↗