Search

Segmenting stroke lesions in Magnetic Resonance Imaging (MRI) is challenging due to diverse clinical imaging domains, with existing models struggling to generalise across different MRI acquisition parameters and sequences. In this work, we propose two novel physics-constrained approaches using synthetic quantitative MRI (qMRI) images to enhance the robustness and generalisability of segmentation models. We trained a qMRI estimation model to predict qMRI maps from MPRAGE images, which were used to simulate diverse MRI sequences for segmentation training. A second approach built upon prior work in synthetic data for stroke lesion segmentation, generating qMRI maps from a dataset of tissue labels. The proposed approaches improved over the baseline nnUNet on a variety of out-of-distribution datasets, with the second approach outperforming the prior synthetic data method.

We investigate the effect of data-driving on flux rope eruptivity in magnetic field simulations by analysing fully data-driven modelling results of active region (AR) 12473 and AR11176, as well as preforming relaxation runs for AR12473 (found to be eruptive). Here, the driving is switched off systematically at different time steps. We analyse the behaviour of fundamental quantities, essential for understanding the eruptivity of magnetic flux ropes (MFRs). The data-driven simulations are carried out with the time-dependent magnetofrictional model (TMFM) for AR12473 and AR11176. For the relaxation runs, we employ the magnetofrictional method (MFM) and a zero-beta magnetohydrodynamic (MHD) model to investigate how significant the differences between the two relaxation procedures are when started from the same initial conditions. To determine the eruptivity of the MFRs, we calculate characteristic geometric properties, such as the cross-section, MFR height along with stability parameters, such as MFR twist and the decay index. For eruptive cases, we investigate the effect of sustained driving beyond the point of eruptivity on the MFR properties. We find that the fully-driven AR12473 MFR is eruptive while the AR11176 MFR is not. For the relaxation runs, we find that the MFM MFRs are eruptive when the driving is stopped around the flare time or later, while the MHD MFRs show eruptive behaviour even if the driving is switched off one and a half days before the flare occurs. We find that characteristic MFR properties can vary greatly even for the eruptive cases of different relaxation simulations. The results suggest that data driving can significantly influence the evolution of the eruption, with differences appearing even when the relaxation time is set to later stages of the simulation when the MFRs have already entered an eruptive phase., Comment: Accepted for publication at A&A

Late-stage infall onto evolved protoplanetary discs is an important source of material and angular momentum replenishment, and disc substructures. In this paper we used 3D smoothed particle hydrodynamics simulations to model streamer-disc interactions for a prograde streamer. The initially parabolic streamer interacts with the disc material to excite disc eccentricity, which can last on the order of $10^5$ years. We found that the spiral arms the streamer excited in the disc can have a variety of pattern speeds, ranging from stationary to super-Keplerian. Spiral arms with various pattern speeds can exist simultaneously, providing a way to diagnose them in observations. Streamer induced spirals appear similar to those generated by a massive outer companion, where the pitch angle of the spiral increases towards the source of the perturbation. Additionally, the spirals arms can show large and sudden pitch angle changes. Streamer induced spirals are long-lived, lasting approximately $3-4\times$ longer than the initial streamer infall timescale ($\sim$$10^4$ years). After the initial interaction with the disc, a long lasting low $m$ azimuthal mode persists in the disc., Comment: 12 pages, 11 figures, submitted to MNRAS, comments welcomed!

Irradiation from the central star controls the temperature structure in protoplanetary discs. Yet simulations of gravitational instability typically use models of stellar irradiation with varying complexity, or ignore it altogether, assuming heat generated by spiral shocks is balanced by cooling, leading to a self-regulated state. In this paper, we perform simulations of irradiated, gravitationally unstable protoplanetary discs using 3D hydrodynamics coupled with live Monte-Carlo radiative transfer. We find that the resulting temperature profile is approximately constant in time, since the thermal effects of the star dominate. Hence, the disc cannot regulate gravitational instabilities by adjusting the temperatures in the disc. In a 0.1 Solar mass disc, the disc instead adjusts by angular momentum transport induced by the spiral arms, leading to steadily decreasing surface density, and hence quenching of the instability. Thus, strong spiral arms caused by self-gravity would not persist for longer than ten thousand years in the absence of fresh infall, although weak spiral structures remain present over longer timescales. Using synthetic images at 1.3mm, we find that spirals formed in irradiated discs are challenging to detect. In higher mass discs, we find that fragmentation is likely because the dominant stellar irradiation overwhelms the stabilising influence of PdV work and shock heating in the spiral arms., Comment: Accepted for publication in MNRAS. 9 pages, 8 figures

Kinematic studies of protoplanetary discs are a valuable method for uncovering hidden companions. In the first paper of this series, we presented five morphological and kinematic criteria that aid in asserting the binary nature of a protoplanetary disc. In this work we study the kinematic signatures of circumbinary discs in the residuals of their velocity maps. We show that Doppler-flips, spiral arms, eccentric gas motion, fast flows inside of the cavity, and vortex-like kinematic signatures are commonly observed. Unlike in the planetary mass companion case, Doppler-flips in circumbinary discs are not necessarily centred on a companion, and can extend towards the cavity edge. We then compare the kinematic signatures in our simulations with observations and see similarities to the Doppler-flip signal in HD 100546 and the vortex-like kinematic signatures in HD 142527. Our analysis also reveals kinematic evidence for binarity in several protoplantary disks typically regarded as circumstellar rather than circumbinary, including AB Aurigae and HD 100546., Comment: 15 pages, 12 Figures. Submitted to MNRAS, revised version. Comments welcomed and appreciated!

Binary interactions have been proposed to explain a variety of circumstellar structures seen around evolved stars, including asymptotic giant branch (AGB) stars and planetary nebulae. Studies resolving the circumstellar envelopes of AGB stars have revealed spirals, discs and bipolar outflows, with shaping attributed to interactions with a companion. For the first time, we have used a combined chemical and dynamical analysis to reveal a highly eccentric and long-period orbit for W Aquilae, a binary system containing an AGB star and a main sequence companion. Our results are based on anisotropic SiN emission, the first detections of NS and SiC towards an S-type star, and density structures observed in the CO emission. These features are all interpreted as having formed during periastron interactions. Our astrochemistry-based method can yield stringent constraints on the orbital parameters of long-period binaries containing AGB stars, and will be applicable to other systems., Comment: This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature's AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: https://doi.org/10.1038/s41550-023-02154-y

For the past 20 years, our approach to shock capturing in smoothed particle hydrodynamics (SPH) has been to use artificial viscosity and conductivity terms supplemented by switches to control excess dissipation away from shocks (Monaghan 1997; Morris & Monaghan 1997). This approach has been demonstrated to be superior to approximate Riemann solvers in a recent comparison (Puri & Ramachandran 2014). The Cullen & Dehnen (2010) switch is regarded as the state of the art. But are we missing something? I will present a novel approach to shock capturing in SPH that utilises the philosophy of approximate Riemann solvers but provides a direct improvement on the ability to reduce excess dissipation away from shocks while preserving the fidelity of the shock itself., Comment: 6 pages, 12 figures, proceedings of 2019 international SPHERIC workshop, Exeter UK. Despite what I said in these proceedings, I later concluded that switches are indeed better, because one can ensure positive definite entropy increase, which is not the case with reconstruction (see Price & Laibe 2020; arXiv:2005.06562)

We performed numerical simulations of the common envelope (CE) interaction between two intermediate-mass asymptotic giant branch (AGB) stars and their low-mass companions. For the first time, formation and growth of dust in the envelope is calculated explicitly. We find that the first dust grains appear as early as $\sim$1-3 yrs after the onset of the CE, and are smaller than grains formed later. As the simulations progress, a high-opacity dusty shell forms, resulting in the CE photosphere being up to an order of magnitude larger than it would be without the inclusion of dust. At the end of the simulations, the total dust yield is $0.0082~M_{\odot}$ ($0.022~M_{\odot}$) for a CE with a $1.7~M_{\odot}$ ($3.7~M_{\odot}$) AGB star. Dust formation does not substantially lead to more mass unbinding or substantially alter the orbital evolution., Comment: 6 pages, 6 figures, contributed talk to the Proceedings IAU Symposium No. 384

Context: In protoplanetary discs, micron-sized dust grows to form millimetre- to centimetre-sized pebbles but encounters several barriers during its evolution. Collisional fragmentation and radial drift impede further dust growth to planetesimal size. Fluffy grains have been hypothesised to solve these problems. While porosity leads to faster grain growth, the implied porosity values obtained from previous simulations were larger than suggested by observations. Aims: In this paper, we study the influence of porosity on dust evolution taking into account growth, bouncing, fragmentation, compaction, rotational disruption and snow lines, in order to understand their impact on dust evolution. Methods: We develop a module for porosity evolution for the 3D Smoothed Particle Hydrodynamics (SPH) code Phantom that accounts for dust growth and fragmentation. This mono-disperse model is integrated into both a 1D code and the 3D code to capture the overall evolution of dust and gas. Results: We show that porosity helps dust growth and leads to the formation of larger solids than when considering compact grains, as predicted by previous work. Our simulations taking into account compaction during fragmentation show that large millimetre grains are still formed, but are 10 to 100 times more compact. Thus, mm sizes with typical filling factors of ~0.1 match the values measured on comets or via polarimetric observations of protoplanetary discs., Comment: 24 pages, 21 figures, accepted for publication in A&A

V892 Tau is a young binary star surrounded by a circumbinary disc which show hints of interaction with the low-mass nearby star V892 Tau NE. The goal of this paper is to constrain the orbit of V892 Tau NE and to determine the resulting circumbinary disc dynamics. We present new ALMA observations of the V892 Tau circumbinary disc at a twice higher angular and spectral resolution. We model the data with V892 Tau as a triple system and perform a grid of hydrodynamical simulations testing several orbits of the companion. The simulation outputs are then post-processed to build synthetic maps that we compare to the observations. The 12CO emission of the disc shows clear non-Keplerian features such as spiral arms. When comparing the data with our synthetic observations, we interpret these features as ongoing interactions with the companion. Our simulations indicate that an eccentricity of 0.5 of the companion is needed to reproduce the observed disc extent and that a mutual inclination of approximately 60{\deg} with the inner binary reproduces the measured disc tilt. In order to explain most of the features of the circumbinary disc, we propose that V892 Tau NE follows a misaligned eccentric orbit, with an eccentricity between 0.2 and 0.5 and a mutual inclination between 30{\deg} and 60{\deg}. Such a misaligned companion suggests the disc is oscillating and precessing with time, stabilising in an intermediate plane with a non-zero mutual inclination with the inner binary. Given that orbital configuration, we show that the stability of future planets is compromised in the second half of the disc once the gas has dissipated., Comment: Published in A&A, 29 pages, 15 figures

With three-dimensional hydrodynamical simulations we show that the size of the decretion disc and the structure of the accretion flow onto the neutron star in a Be/X-ray binary strongly depends upon the disc aspect ratio, $H/R$. We simulate a Be star disc that is coplanar to the orbit of a circularly or moderately eccentric neutron star companion, thereby maximising the effects of tidal truncation. For low disc aspect ratio, $H/R\lesssim 0.1$, the disc is efficiently tidally truncated by the neutron star. Most material that escapes the Roche lobe of the Be star is accreted by the neutron star through tidal streams. For larger disc aspect ratio, the outflow rate through the Be star disc is higher, tidal truncation becomes inefficient, the disc fills the Roche lobe and extends to the orbit of the companion. Some material escapes the binary as a gas stream that begins near the L2 point. While the accretion rate onto the neutron star is higher, the fraction of the outflow that is accreted by the neutron star is smaller. Low density Be star discs are expected to be approximately isothermal, such that $H/R$ increases with radius. Tidal truncation is therefore weaker for larger separation binaries, and lower mass primaries., Comment: Accepted for publication in MNRAS

The recent advancement of spatial transcriptomics (ST) allows to characterize spatial gene expression within tissue for discovery research. However, current ST platforms suffer from low resolution, hindering in-depth understanding of spatial gene expression. Super-resolution approaches promise to enhance ST maps by integrating histology images with gene expressions of profiled tissue spots. However, current super-resolution methods are limited by restoration uncertainty and mode collapse. Although diffusion models have shown promise in capturing complex interactions between multi-modal conditions, it remains a challenge to integrate histology images and gene expression for super-resolved ST maps. This paper proposes a cross-modal conditional diffusion model for super-resolving ST maps with the guidance of histology images. Specifically, we design a multi-modal disentangling network with cross-modal adaptive modulation to utilize complementary information from histology images and spatial gene expression. Moreover, we propose a dynamic cross-attention modelling strategy to extract hierarchical cell-to-tissue information from histology images. Lastly, we propose a co-expression-based gene-correlation graph network to model the co-expression relationship of multiple genes. Experiments show that our method outperforms other state-of-the-art methods in ST super-resolution on three public datasets.

Stars falling too close to massive black holes in the centres of galaxies can be torn apart by the strong tidal forces. Simulating the subsequent feeding of the black hole with disrupted material has proved challenging because of the range of timescales involved. Here we report a set of simulations that capture the relativistic disruption of the star, followed by one year of evolution of the returning debris stream. These reveal the formation of an expanding asymmetric bubble of material extending to hundreds of astronomical units -- an outflowing Eddington envelope with an optically thick inner region. Such envelopes have been hypothesised as the reprocessing layer needed to explain optical/UV emission in tidal disruption events, but never produced self-consistently in a simulation. Our model broadly matches the observed light curves with low temperatures, faint luminosities, and line widths of 10,000--20,000 km/s., Comment: 12 pages + appendices, 15 figures, accepted to ApJL. Data+movies on zenodo at https://doi.org/10.5281/zenodo.11438154 . Movies on YouTube... watch this one first: https://www.youtube.com/watch?v=KlffYCtYxFU then: https://www.youtube.com/watch?v=bPurbeQNgvI&t=19s and finally watch the expanding "bubble" with rho and T at photosphere: https://www.youtube.com/watch?v=D4dyZeUbfvM

Current deep learning-based approaches to lesion segmentation in neuroimaging often depend on high-resolution images and extensive annotated data, limiting clinical applicability. This paper introduces a novel synthetic data framework tailored for stroke lesion segmentation, expanding the SynthSeg methodology to incorporate lesion-specific augmentations that simulate diverse pathological features. Using a modified nnUNet architecture, our approach trains models with label maps from healthy and stroke datasets, facilitating segmentation across both normal and pathological tissue without reliance on specific sequence-based training. Evaluation across in-domain and out-of-domain (OOD) datasets reveals that our method matches state-of-the-art performance within the training domain and significantly outperforms existing methods on OOD data. By minimizing dependence on large annotated datasets and allowing for cross-sequence applicability, our framework holds potential to improve clinical neuroimaging workflows, particularly in stroke pathology. PyTorch training code and weights are publicly available at https://github.com/liamchalcroft/SynthStroke, along with an SPM toolbox featuring a plug-and-play model at https://github.com/liamchalcroft/SynthStrokeSPM.

In our Galactic Center, about 10,000 to 100,000 stars are estimated to have survived tidal disruption events, resulting in partially disrupted remnants. These events occur when a supermassive black hole (SMBH) tidally interacts with a star, but not enough to completely disrupt the star. We use the 1D stellar evolution code Kepler and the 3D smoothed particle hydrodynamics code Phantom to model the tidal disruption of 1, 3, and 10 solar mass stars at zero-age (ZAMS), middle-age (MAMS), and terminal-age main-sequence (TAMS). We map the disruption remnants into Kepler in order to understand their post-distribution evolution. We find distinct characteristics in the remnants, including increased radius, rapid core rotation, and differential rotation in the envelope. The remnants undergo composition mixing that affects their stellar evolution. Whereas the remnants formed by disruption of ZAMS models evolve similarly to unperturbed models of the same mass, for MAMS and TAMS stars, the remnants have higher luminosity and effective temperature. Potential observational signatures include peculiarities in nitrogen and carbon abundances, higher luminosity, rapid rotation, faster evolution, and unique tracks in the Hertzsprung-Russell diagram, Comment: 22 pages, 27 figures, submitted to MNRAS, comments welcome

HD 142527 A is a young and massive Herbig Ae/Be star surrounded by a highly structured disc. The disc shows numerous morphological structures, such as spiral arms, a horseshoe region of dust emission, a set of shadows cast by an inner disc on the outer disc, and a large cavity extending from $\simeq{}$30 au to $\simeq{}$130 au. HD 142527 A also has a lower mass companion, HD 142527 B (M = 0.13 $\pm$ 0.03 $M_\odot{}$), which is thought to be responsible for most of the structures observed in the surrounding disc. We gathered VLTI/GRAVITY observations of HD 142527, either from our own programmes or from the ESO archive. We used this inhomogeneous set of data to extract a total of seven high-precision measurements of the relative astrometry between HD 142527 A and B, spread from mid-2017 to early 2021. Combined with what is available in the literature, we now have 9 yr of astrometric monitoring on HD 142527. We used orbit fitting tools to determine the orbital parameters of HD 142527 B, and used them as inputs for a 3D hydrodynamical model of the disc to determine whether or not the binary is able to create the structures observed in the disc. Our VLTI/GRAVITY astrometry gives excellent constraints on the orbit of HD 142527 B. We show that the secondary is following an orbit of semi-major axis a = 10.80 $\pm$ 0.22 au, with moderate eccentricity (e = 0.47 $\pm$ 0.01). With such a compact orbit, we show that HD 142527 B can only generate a gap and spiral arms of $\sim$30 au in the disc, which is much smaller than what is revealed by observations. Even from a theoretical standpoint, the observed cavity size of $\sim$100 au far exceeds even the most generous predictions for a companion like HD 142527 B on such a compact orbit. Thus, we conclude that the low-mass companion cannot be solely responsible for the observed morphology of the disc surrounding the system., Comment: 11 pages, 6 figures. Accepted for publication in A&A

In many scientific disciplines, the features of interest cannot be observed directly, so must instead be inferred from observed behaviour. Latent variable analyses are increasingly employed to systematise these inferences, and Principal Components Analysis (PCA) is perhaps the simplest and most popular of these methods. Here, we examine how the assumptions that we are prepared to entertain, about the latent variable system, mediate the likelihood that PCA-derived components will capture the true sources of variance underlying data. As expected, we find that this likelihood is excellent in the best case, and robust to empirically reasonable levels of measurement noise, but best-case performance is also: (a) not robust to violations of the method's more prominent assumptions, of linearity and orthogonality; and also (b) requires that other subtler assumptions be made, such as that the latent variables should have varying importance, and that weights relating latent variables to observed data have zero mean. Neither variance explained, nor replication in independent samples, could reliably predict which (if any) PCA-derived components will capture true sources of variance in data. We conclude by describing a procedure to fit these inferences more directly to empirical data, and use it to find that components derived via PCA from two different empirical neuropsychological datasets, are less likely to have meaningful referents in the brain than we hoped., Comment: 26 pages, 3 figures, 3 tables

In a population of multiple protostellar systems with discs, the sub-population of circumbinary discs whose orbital plane is highly misaligned with respect to the binary's orbital plane constrains the initial distribution of orbital parameters of the whole population. We show that by measuring the polar disc fraction and the average orbital eccentricity in the polar discs, one can constrain the distributions of initial eccentricity and mutual inclination in multiple stellar systems at birth., Comment: accepted for publication in A&A

We performed numerical simulations of the common envelope (CE) interaction between thermally-pulsing asymptotic giant branch (AGB) stars of 1.7~\Msun and 3.7~\Msun, respectively, and a 0.6~\Msun compact companion. We use tabulated equations of state to take into account recombination energy. For the first time, formation and growth of dust is calculated explicitly, using a carbon dust nucleation network with a C/O abundance ratio of 2.5 (by number). The first dust grains appear within $\sim$1--3~yrs after the onset of the CE, forming an optically thick shell at $\sim$10--20~au, growing in thickness and radius to values of $\sim$400--500~au over $\sim$40~yrs, with temperatures around 400~K. Most dust is formed in unbound material, having little effect on mass ejection or orbital evolution. By the end of the simulations, the total dust yield is $\sim8.4\times10^{-3}$~\Msun and $\sim2.2\times10^{-2}$~\Msun for the CE with a 1.7~\Msun and a 3.7~\Msun AGB star, respectively, corresponding to a nucleation efficiency close to 100\%, if no dust destruction mechanism is considered. Despite comparable dust yields to single AGB stars, \textit{in CE ejections the dust forms a thousand times faster, over tens of years as opposed to tens of thousands of years}. This rapid dust formation may account for the shift in the infrared of the spectral energy distribution of some optical transients known as luminous red novae. Simulated dusty CEs support the idea that extreme carbon stars and "water fountains" may be objects observed after a CE event., Comment: 18 pages, 17 figures, 2 tables

Stars approaching supermassive black holes can be tidally disrupted. Despite being expected to emit X-rays, TDEs have been largely observed in optical bands, which is poorly understood. In this Letter, we simulate the tidal disruption of a $1~M_\odot$ main sequence star on an eccentric ($e=0.95$) orbit with a periapsis distance one or five times smaller than the tidal radius ($\beta = 1$ or $5$) using general relativistic smoothed particle hydrodynamics. We follow the simulation for up to a year post-disruption. We show that accretion disks in eccentric TDEs are masked by unbound material outflowing at $\sim10,000~$km/s. Assuming electron scattering opacity, this material would be visible as a $\sim100~$au photosphere at $\sim10^4~$K, in line with observations of candidate TDEs., Comment: 11 pages, 5 figures, submitted to ApJL

We present a magnetic flux rope (FR) extraction tool for solar coronal magnetic field modelling data, which builds upon the methodology from Wagner et al. (2023). We apply the scheme to magnetic field simulations of active regions AR12473 and AR11176. We compare the method to its predecessor and study the 3D movement of the newly extracted FRs up to heights of 200 and 300 Mm, respectively. The extraction method is based on the twist parameter and a variety of mathematical morphology algorithms, including the opening transform and the morphological gradient. We highlight the differences between the methods by investigating the circularity of the FRs in the plane we extract from. The simulations for the active regions are carried out with a time-dependent data-driven magnetofrictional model (TMFM; Pomoell et al. (2019)). We investigate the FR trajectories by tracking their apex throughout the full simulation time span. We demonstrate that this upgraded methodology provides the user with more tools and less a-priori assumptions about the FR shape that, in turn, leads to a more accurate set of field lines. The propagation analysis yields that the erupting FR from AR12473 showcases stronger dynamics than the AR11176 FR and a significant deflection during its ascent through the domain. The AR11176 FR appears more stable, though there still is a notable deflection. This confirms that at these low coronal heights, FRs do undergo significant changes in the direction of their propagation even for less dynamic cases. The modelling results are also verified with observations, with AR12473 being indeed dynamic and eruptive, while AR11176 only features an eruption outside of our simulation time window., Comment: Accepted for publication in Astronomy & Astrophysics

Machine learning offers great potential for automated prediction of post-stroke symptoms and their response to rehabilitation. Major challenges for this endeavour include the very high dimensionality of neuroimaging data, the relatively small size of the datasets available for learning, and how to effectively combine neuroimaging and tabular data (e.g. demographic information and clinical characteristics). This paper evaluates several solutions based on two strategies. The first is to use 2D images that summarise MRI scans. The second is to select key features that improve classification accuracy. Additionally, we introduce the novel approach of training a convolutional neural network (CNN) on images that combine regions-of-interest extracted from MRIs, with symbolic representations of tabular data. We evaluate a series of CNN architectures (both 2D and a 3D) that are trained on different representations of MRI and tabular data, to predict whether a composite measure of post-stroke spoken picture description ability is in the aphasic or non-aphasic range. MRI and tabular data were acquired from 758 English speaking stroke survivors who participated in the PLORAS study. The classification accuracy for a baseline logistic regression was 0.678 for lesion size alone, rising to 0.757 and 0.813 when initial symptom severity and recovery time were successively added. The highest classification accuracy 0.854 was observed when 8 regions-of-interest was extracted from each MRI scan and combined with lesion size, initial severity and recovery time in a 2D Residual Neural Network.Our findings demonstrate how imaging and tabular data can be combined for high post-stroke classification accuracy, even when the dataset is small in machine learning terms. We conclude by proposing how the current models could be improved to achieve even higher levels of accuracy using images from hospital scanners.

For many machine learning applications in science, data acquisition, not training, is the bottleneck even when avoiding experiments and relying on computation and simulation. Correspondingly, and in order to reduce cost and carbon footprint, training data efficiency is key. We introduce minimal multilevel machine learning (M3L) which optimizes training data set sizes using a loss function at multiple levels of reference data in order to minimize a combination of prediction error with overall training data acquisition costs (as measured by computational wall-times). Numerical evidence has been obtained for calculated atomization energies and electron affinities of thousands of organic molecules at various levels of theory including HF, MP2, DLPNO-CCSD(T), DFHFCABS, PNOMP2F12, and PNOCCSD(T)F12, and treating tens with basis sets TZ, cc-pVTZ, and AVTZ-F12. Our M3L benchmarks for reaching chemical accuracy in distinct chemical compound sub-spaces indicate substantial computational cost reductions by factors of $\sim$ 1.01, 1.1, 3.8, 13.8 and 25.8 when compared to heuristic sub-optimal multilevel machine learning (M2L) for the data sets QM7b, QM9$^\mathrm{LCCSD(T)}$, EGP, QM9$^\mathrm{CCSD(T)}_\mathrm{AE}$, and QM9$^\mathrm{CCSD(T)}_\mathrm{EA}$, respectively. Furthermore, we use M2L to investigate the performance for 76 density functionals when used within multilevel learning and building on the following levels drawn from the hierarchy of Jacobs Ladder:~LDA, GGA, mGGA, and hybrid functionals. Within M2L and the molecules considered, mGGAs do not provide any noticeable advantage over GGAs. Among the functionals considered and in combination with LDA, the three on average top performing GGA and Hybrid levels for atomization energies on QM9 using M3L correspond respectively to PW91, KT2, B97D, and $\tau$-HCTH, B3LYP$\ast$(VWN5), TPSSH.

Vision transformers are effective deep learning models for vision tasks, including medical image segmentation. However, they lack efficiency and translational invariance, unlike convolutional neural networks (CNNs). To model long-range interactions in 3D brain lesion segmentation, we propose an all-convolutional transformer block variant of the U-Net architecture. We demonstrate that our model provides the greatest compromise in three factors: performance competitive with the state-of-the-art; parameter efficiency of a CNN; and the favourable inductive biases of a transformer. Our public implementation is available at https://github.com/liamchalcroft/MDUNet .

We perform three-dimensional simulations of homogeneous and inhomogeneous cosmologies via the coupling of a numerical relativity code for spacetime evolution and smoothed particle hydrodynamics (SPH) code. Evolution of a flat dust and radiation dominated Friedmann-Lema\^itre-Roberston-Walker (FLRW) spacetime shows an agreement of exact solutions with residuals on the order $10^{-6}$ and $10^{-3}$ respectively, even at low grid resolutions. We demonstrate evolution of linear perturbations of density, velocity and metric quantities to the FLRW with residuals of only $10^{-2}$ compared to exact solutions. Finally, we demonstrate the evolution of non-linear perturbations of the metric past shell-crossing, such that dark matter halo formation is possible. We show that numerical relativistic smoothed particle hydrodynamics is a viable method for understanding non-linear effects in cosmology., Comment: 15 pages, 12 figures, submitted to PRD

Brillouin-Mandelstam scattering is one of the most accessible nonlinear optical phenomena and has been widely studied since its theoretical discovery one hundred years ago. The scattering mechanism is a three-wave mixing process between two optical fields and one acoustic field and has found a broad range of applications spanning microscopy to ultra-narrow-linewidth lasers. Building on the success of utilizing this nonlinearity at a classical level, a rich avenue is now being opened to explore Brillouin scattering within the paradigm of quantum optics. Here, we take a key step in this direction by employing quantum optical techniques yet to be utilized for Brillouin scattering to characterize the second-order coherence of Stokes scattering across the Brillouin lasing threshold. We use a silica microsphere resonator and single-photon counters to observe the expected transition from bunched statistics of thermal light below the lasing threshold to Poissonian statistics of coherent light above the threshold. Notably, at powers approaching the lasing threshold, we also observe super-thermal statistics, which arise due to instability and a ``flickering'' in and out of lasing as the pump field is transiently depleted. The statistics observed across the transition, including the ``flickering'', are a result of the full nonlinear three-wave mixing process and cannot be captured by a linearized model. These measurements are in good agreement with numerical solutions of the three-wave Langevin equations and are well demarcated by analytical expressions for the instability and the lasing thresholds. These results demonstrate that applying second-order-coherence and photon-counting measurements to Brillouin scattering provides new methods to advance our understanding of Brillouin scattering itself and progress toward quantum-state preparation and characterization of acoustic modes., Comment: Main (8 pages, 2 figures) + Supplementary (6 pages, 1 figures), Submitted

We use numerical simulations of circumplanetary disks to determine the boundary between disks that are radially truncated by the tidal potential, and those where gas escapes the Hill sphere. We consider a model problem, in which a coplanar circumplanetary disk is resupplied with gas at an injection radius smaller than the Hill radius. We evolve the disk using the PHANTOM Smoothed Particle Hydrodynamics code until a steady-state is reached. We find that the most significant dependence of the truncation boundary is on the disk aspect ratio $H/R$. Circumplanetary disks are efficiently truncated for $H/R \lesssim 0.2$. For $H/R \simeq 0.3$, up to about half of the injected mass, depending on the injection radius, flows outwards through the decretion disk and escapes. As expected from analytic arguments, the conditions ($H/R$ and Shakura-Sunyaev $\alpha$) required for tidal truncation are independent of planet mass. A simulation with larger $\alpha=0.1$ shows stronger outflow than one with $\alpha=0.01$, but the dependence on transport efficiency is less important than variations of $H/R$. Our results suggest two distinct classes of circumplanetary disks: tidally truncated thin disks with dust-poor outer regions, and thicker actively decreting disks with enhanced dust-to-gas ratios. Applying our results to the PDS 70c system, we predict a largely truncated circumplanetary disk, but it is possible that enough mass escapes to support an outward flow of dust that could explain the observed disk size., Comment: Accepted for publication in ApJ

We carried out 3D smoothed particle hydrodynamics simulations of the common envelope binary interaction using the approximation of Bowen to calculate the dust opacity in order to investigate the resulting dust-driven accelerations. We have simulated two types of binary star: a 1.7 and a 3.7 $M_{\odot}$ thermally-pulsating, asymptotic giant branch stars with a 0.6 $M_{\odot}$ companion. We carried out simulations using both an ideal gas and a tabulated equations of state, with the latter considering the recombination energy of the envelope. We found that the dust-driven wind leads to a relatively small increase in the unbound gas, with the effect being smaller for the tabulated equation of state simulations and for the more massive primary. Dust acceleration does contribute to envelope expansion with only a slightly elongated morphology, if we believe the results from the tabulated equation of state as more reliable. The Bowen opacities in the outer envelopes of the two models, at late times, are large enough that the photosphere of the post-inspiral object is about ten times larger compared to the same without accounting for the dust opacities. As such, the prediction of the appearance of the transient would change substantially if dust is included., Comment: 14 Pages, 13 Figures. Accepted for publication in MNRAS