arXiv subjects
Publications and source records attributed to Gaurav Khanna.
We present a detailed approach for making use of two new computer hardware architectures -- CBEA and CUDA -- for accelerating a scientific data-analysis application (Einstein@Home). Our results suggest that both the architectures suit the application quite well and the achievable performance in the same software developmental time-frame, is nearly identical.
This is the second in a series of papers whose aim is to generate ``adiabatic'' gravitational waveforms from the inspiral of stellar-mass compact objects into massive black holes. In earlier work, we presented an accurate (2+1)D finite-difference time-domain code to solve the Teukolsky equation, which evolves curvature perturbations near rotating (Kerr) black holes. The key new ingredient there was a simple but accurate model of the singular source term based on a discrete representation of the Dirac-delta function and its derivatives. Our earlier work was intended as a proof of concept, using simple circular, equatorial geodesic orbits as a testbed. Such a source is effectively static, in that the smaller body remains at the same coordinate radius and orbital inclination over an orbit. (It of course moves through axial angle, but we separate that degree of freedom from the problem. Our numerical grid has only radial, polar, and time coordinates.) We now extend the time-domain code so that it can accommodate dynamic sources that move on a variety of physically interesting world lines. We validate the code with extensive comparison to frequency-domain waveforms for cases in which the source moves along generic (inclined and eccentric) bound geodesic orbits. We also demonstrate the ability of the time-domain code to accommodate sources moving on interesting non-geodesic worldlines. We do this by computing the waveform produced by a test mass following a ``kludged'' inspiral trajectory, made of bound geodesic segments driven toward merger by an approximate radiation loss formula.
Gravitational waveforms and fluxes from extreme mass--ratio inspirals can be computed using time--domain methods with accuracy that is fast approaching that of frequency--domain methods. We study in detail the computational efficiency of these methods for equatorial orbits of fast spinning Kerr black holes, and find the number of modes needed in either method --as functions of the orbital parameters-- in order to achieve a desired accuracy level. We then estimate the total computation time and argue that for high eccentricity orbits the time--domain approach is more efficient computationally. We suggest that in practice low--$m$ modes are computed using the frequency--domain approach, and high--$m$ modes are computed using the time--domain approach, where $m$ is the azimuthal mode number.
The decay rate of late time tails in the Kerr spacetime have been the cause of numerous conflicting results, both analytical and numerical. In particular, there is much disagreement on whether the decay rate of an initially pure multipole moment ${\ell}$ is according to $t^{-(2{\bar\ell}+3)}$, where ${\bar\ell}$ is the least multipole moment whose excitation is not disallowed, or whether the decay rate is according to $t^{-n}$, where $n=n({\ell})$. We do careful 2+1D numerical simulations, and explain the various results. In particular, we show that pure multipole outgoing initial data in either Boyer--Lindquist on ingoing Kerr coordinates on the corresponding slices lead to the same late time tail behavior. We also show that similar initial data specified in terms of the Poisson spherical coordinates lead to the simpler $t^{-(2{\bar\ell}+3)}$ late time tail. We generalize the rule $n=n({\ell})$ to subdominant modes, and also study the behavior of non--axisymmetric initial data. We discuss some of the causes for possible errors in 2+1D simulations, demonstrate that our simulations are free of those errors, and argue that some conflicting past results may be attributed to them.
In this article, we develop an intuitive and efficient, numerical technique to solve the quantum evolution equation of generic lattice-refined models in loop quantum cosmology. As an application of this method, we extensively study the solutions of the recently introduced, lattice-refined anisotropic model of the Schwarzschild interior geometry. Our calculations suggest that the results obtained from the approach are accurate, robust and are in complete agreement with the expectations from the von Neumann stability analysis of the model.
We revisit the problem of the emission of gravitational waves from a test mass orbiting and thus perturbing a Kerr black hole. The source term of the Teukolsky perturbation equation contains a Dirac delta function which represents a point particle. We present a technique to effectively model the delta function and its derivatives using as few as four points on a numerical grid. The source term is then incorporated into a code that evolves the Teukolsky equation in the time domain as a (2+1) dimensional PDE. The waveforms and energy fluxes are extracted far from the black hole. Our comparisons with earlier work show an order of magnitude gain in performance (speed) and numerical errors less than 1% for a large fraction of parameter space. As a first application of this code, we analyze the effect of finite extraction radius on the energy fluxes. This paper is the first in a series whose goal is to develop adiabatic waveforms describing the inspiral of a small compact body into a massive Kerr black hole.
A general class of loop quantizations for anisotropic models is introduced and discussed, which enhances loop quantum cosmology by relevant features seen in inhomogeneous situations. The main new effect is an underlying lattice which is being refined during dynamical changes of the volume. In general, this leads to a new feature of dynamical difference equations which may not have constant step-size, posing new mathematical problems. It is discussed how such models can be evaluated and what lattice refinements imply for semiclassical behavior. Two detailed examples illustrate that stability conditions can put strong constraints on suitable refinement models, even in the absence of a fundamental Hamiltonian which defines changes of the underlying lattice. Thus, a large class of consistency tests of loop quantum gravity becomes available. In this context, it will also be seen that quantum corrections due to inverse powers of metric components in a constraint are much larger than they appeared recently in more special treatments of isotropic, free scalar models where they were artificially suppressed.
The accuracy of time-domain solutions of the inhomogeneous Teukolsky equation is improved significantly. Comparing energy fluxes in gravitational waves with highly accurate frequency-domain results for circular equatorial orbits in Schwarzschild and Kerr, we find agreement to within 1% or better, which we believe can be even further improved. We apply our method to orbits for which frequency-domain calculations have a relative disadvantage, specifically high-eccentricity (elliptical and parabolic) "zoom-whirl" orbits, and find the energy fluxes, waveforms, and characteristic strain in gravitational waves.
In this article we perform von Neumann analysis of the difference equations that arise as a result of loop quantum gravity being applied to models of cosmology and black holes. In particular, we study the numerical stability of Bianchi I LRS (symmetric and non-symmetric constraint) and Schwarzschild interior (symmetric constraint) models, and find that there exist domains over which there are instabilities, generically. We also present explicit evolutions of wave-packets in these models and clearly demonstrate the presence of these instabilities.
Using the Hamiltonian constraint derived by Ashtekar and Bojowald, we look for pre-classical wave functions in the Schwarzschild interior. In particular, when solving this difference equation by separation of variables, an inequality is obtained relating the Immirzi parameter $γ$ to the quantum ambiguity $δ$ appearing in the model. This bound is violated when we use a natural value for $δ$ based on loop quantum gravity together with a recent proposal for $γ$. We also present numerical solutions of the constraint.
In this paper, we discuss the properties of one-parameter sequences that arise when solving the Hamiltonian constraint in Bianchi I loop quantum cosmology using a separation of variables method. In particular, we focus on finding an expression for the sequence for all real values of the parameter, and discuss the pre-classicality of this function. We find that the behavior of these preclassical sequences imply time asymmetry on either side of the classical singularity in Bianchi I cosmology.
In this paper we introduce a numerical approximation technique to obtain pre-classical solutions to models of loop quantum gravity. In particular, we apply the technique to vacuum Bianchi I cosmological models and recover known solutions. We also present a pre-classical solution to the Bianchi I LRS model with cosmological constant, which has not appeared elsewhere.
Loop quantum cosmology, the symmetry reduction of quantum geometry for the study of various cosmological situations, leads to a difference equation for its quantum evolution equation. To ensure that solutions of this equation act in the expected classical manner far from singularities, additional restrictions are imposed on the solution. In this paper, we consider the Bianchi I model, both the vacuum case and the addition of a cosmological constant, and show using generating function techniques that only the zero solution satisfies these constraints. This implies either that there are technical difficulties with the current method of quantizing the evolution equation, or else loop quantum gravity imposes strong restrictions on the physically allowed solutions.
We study the Klein-Gordon equation for a massive scalar field in Kerr spacetime in the time-domain. We demonstrate that under conditions of super-radiance, the scalar field becomes unstable and its amplitude grows without bound. We also estimate the growth rate of this instability.
Loop quantum cosmology leads to a difference equation for the wave function of a universe, which in general has solutions changing rapidly even when the volume changes only slightly. For a semiclassical regime such small-scale oscillations must be suppressed, by choosing the parameters of the solution appropriately. For anisotropic models this is not possible to do numerically by trial and error; instead, it is shown here for the Bianchi I LRS model how this can be done analytically, using generating function techniques. Those techniques can also be applied to more complicated models, and the results gained allow conclusions about initial value problems for other systems.
We extend the treatment of the problem of the gravitational waves produced by a particle of negligible mass orbiting a Kerr black hole using black hole perturbation theory in the time domain, to elliptic and inclined orbits. We model the particle by smearing the singularities in the source term using narrow Gaussian distributions. We compare results (energy and angular momentum fluxes) for such orbits with those computed using the frequency domain formalism.
Treating the Teukolsky perturbation equation numerically as a 2+1 PDE and smearing the singularities in the particle source term by the use of narrow Gaussian distributions, we have been able to reproduce earlier results for equatorial circular orbits that were computed using the frequency domain formalism. A time domain prescription for a more general evolution of nearly geodesic orbits under the effects of radiation reaction is presented. This approach can be useful when tackling the more realistic problem of a stellar-mass black hole moving on a generic orbit around a supermassive black hole under the influence of radiation reaction forces.
We study numerically the late-time tails of linearized fields with any spin $s$ in the background of a spinning black hole. Our code is based on the ingoing Kerr coordinates, which allow us to penetrate through the event horizon. The late time tails are dominated by the mode with the least multipole moment $\ell$ which is consistent with the equatorial symmetry of the initial data and is equal to or greater than the least radiative mode with $s$ and the azimuthal number $m$.