arXiv subjects

Balsa Terzic

Publications and source records attributed to Balsa Terzic.

14 recordsLinked to original sources

arXiv2023

Porting numerical integration codes from CUDA to oneAPI: a case study

We present our experience in porting optimized CUDA implementations to oneAPI. We focus on the use case of numerical integration, particularly the CUDA implementations of PAGANI and $m$-Cubes. We faced several challenges that caused performance degradation in the oneAPI ports. These include differences in utilized registers per thread, compiler optimizations, and mappings of CUDA library calls to oneAPI equivalents. After addressing those challenges, we tested both the PAGANI and m-Cubes integrators on numerous integrands of various characteristics. To evaluate the quality of the ports, we collected performance metrics of the CUDA and oneAPI implementations on the Nvidia V100 GPU. We found that the oneAPI ports often achieve comparable performance to the CUDA versions, and that they are at most 10% slower.

cs.DC↗

arXiv2023

Hubble Tension and Gravitational Self-Interaction

One of the most important problems vexing the $\Lambda$CDM cosmological model is the Hubble tension. It arises from the fact that measurements of the present value of the Hubble parameter performed with low-redshift quantities, e.g., the Type IA supernova, tend to yield larger values than measurements from quantities originating at high-redshift, e.g., fits of cosmic microwave background radiation. It is becoming likely that the discrepancy, currently standing at $5\sigma$, is not due to systematic errors in the measurements. Here we explore whether the self-interaction of gravitational fields in General Relativity, which are traditionally neglected when studying the evolution of the universe, can explain the tension. We find that with field self-interaction accounted for, both low- and high-redshift data are simultaneously well-fitted, thereby showing that gravitational self-interaction could explain the Hubble tension. Crucially, this is achieved without introducing additional parameters.

astro-ph.CO↗

arXiv2022

m-CUBES An efficient and portable implementation of multi-dimensional integration for gpus

The task of multi-dimensional numerical integration is frequently encountered in physics and other scientific fields, e.g., in modeling the effects of systematic uncertainties in physical systems and in Bayesian parameter estimation. Multi-dimensional integration is often time-prohibitive on CPUs. Efficient implementation on many-core architectures is challenging as the workload across the integration space cannot be predicted a priori. We propose m-Cubes, a novel implementation of the well-known Vegas algorithm for execution on GPUs. Vegas transforms integration variables followed by calculation of a Monte Carlo integral estimate using adaptive partitioning of the resulting space. m-Cubes improves performance on GPUs by maintaining relatively uniform workload across the processors. As a result, our optimized Cuda implementation for Nvidia GPUs outperforms parallelization approaches proposed in past literature. We further demonstrate the efficiency of m-Cubes by evaluating a six-dimensional integral from a cosmology application, achieving significant speedup and greater precision than the CUBA library's CPU implementation of VEGAS. We also evaluate m-Cubes on a standard integrand test suite. m-Cubes outperforms the serial implementations of the Cuba and GSL libraries by orders of magnitude speedup while maintaining comparable accuracy. Our approach yields a speedup of at least 10 when compared against publicly available Monte Carlo based GPU implementations. In summary, m-Cubes can solve integrals that are prohibitively expensive using standard libraries and custom implementations. A modern C++ interface header-only implementation makes m-Cubes portable, allowing its utilization in complicated pipelines with easy to define stateful integrals. Compatibility with non-Nvidia GPUs is achieved with our initial implementation of m-Cubes using the Kokkos framework.

cs.DC↗

arXiv2016

Laser Pulsing in Linear Compton Scattering

Previous work on calculating energy spectra from Compton scattering events has either neglected considering the pulsed structure of the incident laser beam, or has calculated these effects in an approximate way subject to criticism. In this paper, this problem has been reconsidered within a linear plane wave model for the incident laser beam. By performing the proper Lorentz transformation of the Klein-Nishina scattering cross section, a spectrum calculation can be created which allows the electron beam energy spread and emittance effects on the spectrum to be accurately calculated, essentially by summing over the emission of each individual electron. Such an approach has the obvious advantage that it is easily integrated with a particle distribution generated by particle tracking, allowing precise calculations of spectra for realistic particle distributions in collision. The method is used to predict the energy spectrum of radiation passing through an aperture for the proposed Old Dominion University inverse Compton source. Many of the results allow easy scaling estimates to be made of the expected spectrum.

physics.acc-ph↗

arXiv2007

Realistic triaxial density--potential--force profiles for stellar systems and dark matter halos

Popular models for describing the luminosity-density profiles of dynamically hot stellar systems (e.g., Jaffe, Hernquist, Dehnen) were constructed to match the deprojected form of de Vaucouleurs' $R^{1/4}$ light-profile. However, we now know that elliptical galaxies and bulges display a mass-dependent range of structural profiles. To compensate this, the model in Terzic & Graham was designed to closely match the deprojected form of Sersic $R^{1/n}$ light-profiles, including deprojected exponential light-profiles and galaxies with partially depleted cores. It is thus applicable for describing bulges in spiral galaxies, dwarf elliptical galaxies, both ``power-law'' and ``core'' elliptical galaxies, also dark matter halos formed from $Λ$CDM cosmological simulations. In this paper, we present a new family of triaxial density-potential-force triplets, which generalizes the spherical model reported in Terzic & Graham to three dimensions. If the (optional) power-law core is present, it is a 5-parameter family, while in the absence of the core it reduces to 3 parameters. The isodensity contours in the new family are stratified on confocal ellipsoids and the potential and forces are expressed in terms of integrals which are easy to evaluate numerically. We provide the community with a suite of numerical routines for orbit integration, which feature: optimized computations of potential and forces for this family; the ability to run simulations on parallel platforms; and modular and easily editable design.

astro-ph↗

arXiv2006

Empirical models for Dark Matter Halos. I. Nonparametric Construction of Density Profiles and Comparison with Parametric Models

We use techniques from nonparametric function estimation theory to extract the density profiles, and their derivatives, from a set of N-body dark matter halos. We consider halos generated from LCDM simulations of gravitational clustering, as well as isolated, spherical collapses. The logarithmic density slopes gamma = d(log rho)/d(log r) of the LCDM halos are found to vary as power-laws in radius, reaching values of gamma ~ -1 at the innermost resolved radii (~0.01 r_virial). This behavior is significantly different from that of broken power-law models like the NFW profile, but similar to that of models like de Vaucouleurs'. Accordingly, we compare the N-body density profiles with various parametric models to find which provide the best fit. We consider an NFW-like model with arbitrary inner slope; Dehnen & McLaughlin's anisotropic model; Einasto's model (identical in functional form to Sersic's model but fit to the space density); and the density model of Prugniel & Simien that was designed to match the deprojected form of Sersic's R^{1/n} law. Overall, the best-fitting model to the LCDM halos is Einasto's, although the Prugniel-Simien and Dehnen-McLaughlin models also perform well. With regard to the spherical collapse halos, both the Prugniel-Simien and Einasto models describe the density profiles well, with an rms scatter some four times smaller than that obtained with either the NFW-like model or the 3-parameter Dehnen-McLaughlin model. Finally, we confirm recent claims of a systematic variation in profile shape with halo mass.

astro-ph↗

arXiv2006

Empirical Models for Dark Matter Halos. II. Inner profile slopes, dynamical profiles, and rho/sigma^3

We have recently shown that both the Prugniel-Simien model and Sersic's function (hereafter referred to as the Einasto model when applied to internal density profiles) describe simulated dark matter halos better than an NFW-like model with equal number of parameters. Here we provide analytical expressions for the logarithmic slopes of these models, and compare them with data from real galaxies. Depending on the Einasto parameters of the dark matter halo, one can expect an extrapolated, inner (0.01--1 kpc), logarithmic profile slope ranging from -0.2 to -1.5, with a typical value at 0.1 kpc around -0.7. Application of this (better fitting) model therefore alleviates some of the past disagreement with observations on this issue. We additionally provide useful expressions for the concentration and assorted scale radii: r_s, r_{-2}, r_e, R_e, r_virial, and r_max -- the radius where the circular velocity profile has its maximum value. We also present the circular velocity profiles and the radial behavior of rho(r)/sigma(r)^3 for both the Einasto and Prugniel-Simien models. We find this representation of the phase-space density profile to be well approximated by a power-law with slope slightly shallower than -2 near r=r_{-2}.

astro-ph↗

arXiv2006

Empirical Models for Dark Matter Halos. III. The Kormendy relation and the log(rho_e)-log(R_e) relation

We have recently shown that the 3-parameter density-profile model from Prugniel & Simien provides a better fit to simulated, galaxy- and cluster-sized, dark matter halos than an NFW-like model with arbitrary inner profile slope gamma (Paper I). By construction, the parameters of the Prugniel-Simien model equate to those of the Sersic R^{1/n} function fitted to the projected distribution. Using the Prugniel-Simien model, we are therefore able to show that the location of simulated (10^{12} M_sun) galaxy-sized dark matter halos in the _e-log(R_e) diagram coincides with that of brightest cluster galaxies, i.e., the dark matter halos appear consistent with the Kormendy relation defined by luminous elliptical galaxies. These objects are also seen to define the new, and equally strong, relation log(rho_e) = 0.5 - 2.5log(R_e), in which rho_e is the internal density at r=R_e. Simulated (10^{14.5} M_sun) cluster-sized dark matter halos and the gas component of real galaxy clusters follow the relation log(rho_e) = 2.5[1 - log(R_e)]. Given the shapes of the various density profiles, we are able to conclude that while dwarf elliptical galaxies and galaxy clusters can have dark matter halos with effective radii of comparable size to the effective radii of their baryonic component, luminous elliptical galaxies can not. For increasingly large elliptical galaxies, with increasingly large profile shapes `n', to be dark matter dominated at large radii requires dark matter halos with increasingly large effective radii compared to the effective radii of their stellar component.

astro-ph↗

arXiv2005

Density-potential pairs for spherical stellar systems with Sersic light-profiles and (optional) power-law cores

Popular models for describing the luminosity-density profiles of dynamically hot stellar systems (e.g., Jaffe, Hernquist, Dehnen) were constructed with the desire to match the deprojected form of an R^{1/4} light-profile. Real galaxies, however, are now known to have a range of different light-profile shapes that scale with mass. Consequently, although highly useful, the above models have implicit limitations, and this is illustrated here through their application to a number of real galaxy density profiles. On the other hand, the analytical density profile given by Prugniel & Simien (1997) closely matches the deprojected form of Sersic R^{1/n} light-profiles - including deprojected exponential light-profiles. It is thus applicable for describing bulges in spiral galaxies, dwarf elliptical galaxies, and both ordinary and giant elliptical galaxies. Here we provide simple equations, in terms of elementary and special functions, for the gravitational potential and force associated with this density profile. Furthermore, to match galaxies with partially depleted cores, and better explore the supermassive black hole / galaxy connection, we have added a power-law core to this density profile and derived similar expressions for the potential and force of this hybrid profile. Expressions for the mass and velocity dispersion, assuming isotropy, are also given. These models may also prove appropriate for describing the dark matter distribution in halos formed from LCDM cosmological simulations.

astro-ph↗

arXiv2003

Noise-Induced Phase Space Transport in Time-Periodic Hamiltonian Systems

Orbits in a three-dimensional potential subjected to periodic driving, V(x^i,t)=[1+m_0 sin(omega t) V_0(x^i), divide naturally into two types, regular and chaotic, between which transitions are seemingly impossible. The chaotic orbits divide in turn into two types, apparently separated by entropy barriers, namely `sticky' orbit segments, which are `locked' to the driving frequency and exhibit little systematic energy diffusion, and `wildly' chaotic segments, which are not so locked and can exhibit significant energy diffusion. Attention focuses on how the relative abundance of these different orbit types and the transition rate between sticky and wildly chaotic orbits depends on amplitude m_0, and on the extent to which these quantities can be altered by weak friction and/or noise and by pseudo-random variations in the driving frequency, idealized as an Ornstein-Uhlenbeck process with (in general) nonzero autocorrelation time t_c. When, in the absence of perturbations, there exist large measures of both regular and chaotic orbits, the primary effect of weak noise is to increase the relative measure of chaotic orbits. Alternatively, when almost all the orbits are already chaotic, noise serves primarily to accelerate transitions from sticky to wildly chaotic behavior. The presence or absence of friction is unimportant and the details of the noise seem largely immaterial. In particular, there is only a weak, roughly logarithmic dependence on amplitude, and additive and multiplicative noise typically have virtually identical effects. By contrast, allowing for high frequency `noisy' variations in omega tends to weaken effects of the driving, decreasing the relative measure of chaotic orbits and suppressing large scale energy diffusion.

astro-ph↗

arXiv2003

Orbital structure in oscillating galactic potentials

This paper focuses on symmetric potentials subjected to periodic driving. Four unperturbed potentials V_0(r) were considered, namely the Plummer potential and Dehnen potentials with γ=0.0, 0.5, and 1.0, each subjected to a time-dependence of the form V(r,t)=V_0(r)(1+m_0\sin(ωt)). In each case, the orbits divide clearly into regular and chaotic, distinctions which appear absolute. In particular, transitions from regularity to chaos are seemingly impossible. Over finite time intervals, chaotic orbits subdivide into what can be termed `sticky' chaotic orbits, which exhibit no large scale secular changes in energy and remain trapped in the phase space region where they started; and `wildly chaotic' orbits, which do exhibit systematic drifts in energy as the orbits diffuse to inhabit different phase space regions. This latter distinction is not absolute, apparently corresponding instead to orbits penetrating a `leaky' phase space barrier. The three different orbit types can be identified simply in terms of the frequencies for which their Fourier spectra have the most power. An examination of the statistical properties of orbit ensembles as a function of driving frequency ωallows one to identify the specific resonances that determine orbital structure. Attention focuses also on how, for fixed amplitude m_0, such quantities as the mean energy shift, the relative measure of chaotic orbits, and the mean value of the largest Lyapunov exponent vary with driving frequency ωand how, for fixed ω, the same quantities depend on m_0.

astro-ph↗

arXiv2003

Supermassive Black Hole Binaries as Galactic Blenders

This paper focuses on the dynamical implications of close supermassive black hole binaries both as an example of resonant phase mixing and as a potential explanation of inversions and other anomalous features observed in the luminosity profiles of some elliptical galaxies. The presence of a binary comprised of black holes executing nearly periodic orbits leads to the possibility of a broad resonant coupling between the black holes and various stars in the galaxy. This can result in efficient chaotic phase mixing and, in many cases, systematic increases in the energies of stars and their consequent transport towards larger radii. Allowing for the presence of a supermassive black hole binary with plausible parameter values near the center of a spherical, or nearly spherical, galaxy characterised initially by a Nuker density profile enables one to reproduce in considerable detail the central surface brightness distributions of such galaxies as NGC 3706.

astro-ph↗

arXiv2003

Building Self-Consistent Triaxial Galaxy Models Using Schwarzschild's Method

We use Schwarzschild's orbit superposition method to build self-consistent models of elliptical galaxies with scale-free potentials. Our exhaustive study of all physical shapes and central densities for galaxies with scale-free potentials establishes a relationship between the presence of chaos and self-consistency. The extent and the onset of chaos is studied by varying parameters of the model which are believed to be its inducers, such as the steepness of the central density cusp, flatness and triaxiality. We show that gravitational scattering of the central density cusp plays a dominant role in restricting the shapes of elliptical galaxies.

astro-ph↗

arXiv2003

Semi-Analytic Estimates of Lyapunov Exponents in Lower-Dimensional Systems

Recent work has shown that statistical arguments, seemingly well-justified in higher dimensions, can also be used to derive reasonable, albeit less accurate, estimates of the largest Lyapunov exponent $χ$ in lower-dimensional Hamiltonian systems. This letter explores the detailed assumptions incorporated into these arguments. The predicted values of $χ$ are insensitive to most of these details, which can in any event be relaxed straightforwardly, but {\em can} depend sensitively on the nongeneric form of the auto-correlation function characterising the time-dependence of an orbit. This dependence on dynamics implies a fundamental limitation to the application of thermodynamic arguments to such lower-dimensional systems.

astro-ph↗