Publications

The scattering of cosmic microwave background (CMB) photons off the free-electron gas in galaxies and clusters leaves detectable imprints on high resolution CMB maps: the thermal and kinematic Sunyaev-Zel'dovich effects (tSZ and kSZ respectively). We use combined microwave maps from the Atacama Cosmology Telescope (ACT) DR5 and Planck in combination with the CMASS and LOWZ galaxy catalogs from the Baryon Oscillation Spectroscopic Survey (BOSS DR10 and DR12), to study the gas associated with these galaxy groups. Using individual reconstructed velocities, we perform a stacking analysis and reject the no-kSZ hypothesis at 6.5σ, the highest significance to date. This directly translates into a measurement of the electron number density profile, and thus of the gas density profile. Despite the limited signal to noise, the measurement shows at high significance that the gas density profile is more extended than the dark matter density profile, for any reasonable baryon abundance (formally >90σ for the cosmic baryon abundance). We simultaneously measure the tSZ signal, i.e. the electron thermal pressure profile of the same CMASS objects, and reject the no-tSZ hypothesis at 10σ. We combine tSZ and kSZ measurements to estimate the electron temperature to 20% precision in several aperture bins, and find it comparable to the virial temperature. In a companion paper, we analyze these measurements to constrain the gas thermodynamics and the properties of feedback inside galaxy groups. We present the corresponding LOWZ measurements in this paper, ruling out a null kSZ (tSZ) signal at 2.9 (13.9)σ, and leave their interpretation to future work. Our stacking software ThumbStack is publicly available at \href{https://github.com/EmmanuelSchaan/ThumbStack}{this https URL} and directly applicable to future Simons Observatory and CMB-S4 data.

The thermal and kinematic Sunyaev-Zel'dovich effects (tSZ, kSZ) probe the thermodynamic properties of the circumgalactic and intracluster medium (CGM and ICM) of galaxies, groups, and clusters, since they are proportional, respectively, to the integrated electron pressure and momentum along the line-of-sight. We present constraints on the gas thermodynamics of CMASS galaxies in the Baryon Oscillation Spectroscopic Survey (BOSS) using new measurements of the kSZ and tSZ signals obtained in a companion paper. Combining kSZ and tSZ measurements, we measure within our model the amplitude of energy injection ϵM⋆c2, where M⋆ is the stellar mass, to be ϵ=(40±9)×10−6, and the amplitude of the non-thermal pressure profile to be αNth<0.2 (2σ), indicating that less than 20% of the total pressure within the virial radius is due to a non-thermal component. We estimate the effects of including baryons in the modeling of weak-lensing galaxy cross-correlation measurements using the best-fit density profile from the kSZ measurement. Our estimate reduces the difference between the original theoretical model and the weak-lensing galaxy cross-correlation measurements in arXiv:1611.08606 by half but does not fully reconcile it. Comparing the tSZ measurements to cosmological simulations, we find that simulations underestimate the CGM pressure at large radii while they fare better in comparison with the kSZ measurements. This suggests that the energy injected via feedback models in the simulations that we compared against does not sufficiently heat the gas at these radii. We do not find significant disagreement at smaller radii. These measurements provide novel tests of current and future simulations. This work demonstrates the power of joint, high signal-to-noise kSZ and tSZ observations, upon which future cross-correlation studies will improve.

One of the central features of deep learning is the generalization abilities of neural networks, which seem to improve relentlessly with over-parametrization. In this work, we investigate how properties of data impact the test error as a function of the number of training examples and number of training parameters; in other words, how the structure of data shapes the "generalization phase space". We first focus on the random features model trained in the teacher-student scenario. The synthetic input data is composed of independent blocks, which allow us to tune the saliency of low-dimensional structures and their relevance with respect to the target function. Using methods from statistical physics, we obtain an analytical expression for the train and test errors for both regression and classification tasks in the high-dimensional limit. The derivation allows us to show that noise in the labels and strong anisotropy of the input data play similar roles on the test error. Both promote an asymmetry of the phase space where increasing the number of training examples improves generalization further than increasing the number of training parameters. Our analytical insights are confirmed by numerical experiments involving fully-connected networks trained on MNIST and CIFAR10.

Broken symmetries induce strong even-order nonlinear optical responses in materials and at interfaces. Unlike conventional covalently bonded nonlinear crystals, van der Waals (vdW) heterostructures feature layers that can be stacked at arbitrary angles, giving complete control over the presence or lack of inversion symmetry at a crystal interface. Here, we report highly tunable second harmonic generation (SHG) from nanomechanically rotatable stacks of bulk hexagonal boron nitride (BN) crystals and introduce the term twistoptics to describe studies of optical properties in twistable vdW systems. By suppressing residual bulk effects, we observe SHG intensity modulated by a factor of more than 50, and polarization patterns determined by moiré interface symmetry. Last, we demonstrate greatly enhanced conversion efficiency in vdW vertical superlattice structures with multiple symmetry-broken interfaces. Our study paves the way for compact twistoptics architectures aimed at efficient tunable frequency conversion and demonstrates SHG as a robust probe of buried vdW interfaces.

We present an analysis of 1524 spectra of Vega spanning 10 years, in which we search for periodic radial velocity variations. A signal with a periodicity of 0.676 days and a semi-amplitude of ~10 m/s is consistent with the rotation period measured over much shorter time spans by previous spectroscopic and spectropolarimetric studies, confirming the presence of surface features on this A0 star. The timescale of evolution of these features can provide insight into the mechanism that sustains the weak magnetic fields in normal A type stars. Modeling the radial velocities with a Gaussian process using a quasi-periodic kernel suggests that the characteristic spot evolution timescale is ~180 days, though we cannot exclude the possibility that it is much longer. Such long timescales may indicate the presence of failed fossil magnetic fields on Vega. TESS data reveal Vega's photometric rotational modulation for the first time, with a total amplitude of only 10 ppm, and a comparison of the spectroscopic and photometric amplitudes suggest the surface features may be dominated by bright plages rather than dark spots. For the shortest orbital periods, transit and radial velocity injection recovery tests exclude the presence of transiting planets larger than 2 Earth radii and most non-transiting giant planets. At long periods, we combine our radial velocities with direct imaging from the literature to produce detection limits for Vegan planets and brown dwarfs out to distances of 15 au. Finally, we detect a candidate radial velocity signal with a period of 2.43 days and a semi-amplitude of 6 m/s. If caused by an orbiting companion, its minimum mass would be ~20 Earth masses; because of Vega's pole-on orientation, this would correspond to a Jovian planet if the orbit is aligned with the stellar spin. We discuss the prospects for confirmation of this candidate planet.

We introduce a new general-purpose time-dependent ionisation network (IN) and a radiation transport (RT) module in the magneto-hydrodynamic (MHD) code PLUTO. Our ionisation network is reliable for temperatures ranging from 5e3 to 3e8 K, and includes all ionisation states of H, He, C, N, O, Ne, Mg, Si, S and Fe making it suitable for studying a variety of astrophysical scenarios. Radiation loss for each ion-electron pair is calculated using CLOUDY-17 data on-the-fly. Photo-ionisation and charge exchange are the chemical heating mechanisms. The IN is fully coupled to the radiation transport module over a very large range of opacities at different frequencies. The RT module employs a method of short characteristics assuming spherical symmetry. The radiation module requires the assumption of spherical symmetry, while the IN is compatible with full 3D. We also include a simple prescription for dust opacity, grain destruction, and the dust contribution to radiation pressure. We present numerical tests to show the reliability and limitations of the new modules. We also present a post-processing tool to calculate projected column densities and emission spectra.

The most popular algorithm for the nonuniform fast Fourier transform (NUFFT) uses the dilation of a kernel $\phi$ to spread (or interpolate) between given nonuniform points and a uniform upsampled grid, combined with an FFT and diagonal scaling (deconvolution) in frequency space. The high performance of the recent FINUFFT library is in part due to its use of a new ``exponential of semicircle'' kernel $\phi(z)=e^{\beta \sqrt{1-z^2}}$, for $z\in[-1,1]$, zero otherwise, whose Fourier transform $\hat\phi$ is unknown analytically. We place this kernel on a rigorous footing by proving an aliasing error estimate which bounds the error of the one-dimensional NUFFT of types 1 and 2 in exact arithmetic. Asymptotically in the kernel width measured in upsampled grid points, the error is shown to decrease with an exponential rate arbitrarily close to that of the popular Kaiser--Bessel kernel. This requires controlling a conditionally-convergent sum over the tails of $\hat\phi$, using steepest descent, other classical estimates on contour integrals, and a phased sinc sum. We also draw new connections between the above kernel, Kaiser--Bessel, and prolate spheroidal wavefunctions of order zero, which all appear to share an optimal exponential convergence rate.

We present an extension of constrained-path auxiliary-field quantum Monte Carlo (CP-AFQMC) for the treatment of correlated electronic systems coupled to phonons. The algorithm follows the standard CP-AFQMC approach for description of the electronic degrees of freedom while phonons are described in first quantization and propagated via a diffusion Monte Carlo approach. Our method is tested on the one- and two-dimensional Holstein and Hubbard-Holstein models. With a simple semiclassical trial wavefunction, our approach is remarkably accurate for ω/(2dtλ) < 1 for all parameters in the Holstein model considered in this study. In addition, we empirically show that the autocorrelation time scales as 1/ω for ω/t 1, which is an improvement over the 1/ω

Convolutional neural networks (CNN) have become a standard approach for modeling genomic sequences. CNNs can be effectively built by Neural Architecture Search (NAS) by trading computing power for accurate neural architectures. Yet, the consumption of immense computing power is a major practical, financial, and environmental issue for deep learning. Here, we present a novel NAS framework,
AMBIENT, that generates highly accurate CNN architectures for biological sequences of diverse functions, while substantially reducing the computing cost of conventional NAS.