My latest papers

Short summaries of my most recent research articles.

The glow of eternal black holes

An eternal black hole is the idealized, maximally extended Schwarzschild solution. Black holes we observe in the universe are likely different from that solution and originate from the collapse of stellar matter. However, an eternal black hole is a possible mathematical solution and may correspond to an actual object. The full spacetime of an eternal black hole includes a white-hole region connected to a past singularity. In this paper I explore what such an object would look like to an observer. Motivated by Markov's hypothesis of a limiting density of matter, the past singularity is replaced by a “surface of last scattering” that emits thermal radiation. Rays traced backward from the observer that cross the past horizon terminate on this surface. As a result, the shadow of an eternal black hole is not dark but has a bright spot whose radial profile is set by the gravitational frequency shift.

The spot has a remarkably clean interferometric signature. Its visibility decays as a pure exponential in the baseline length, in contrast to the power laws of the accretion disk and the photon ring. The observed 230 GHz fluxes of M87* and Sgr A* then constrain the product of the surface's radius and temperature. The darkness of the observed shadows is thus a test of whether these black holes are eternal.

Explore the spacetime of the paper in the interactive Diagram Zoomer (drag the slider or scroll to change the compactification scale):

Hankel low-rank matrix approximation for gravitational-wave data analysis

N. Geissler, V. Strokov, C. Kümmerle, S. Kushnarev, E. Berti · arXiv:2603.15740 · under review at Physical Review D

Future gravitational-wave detectors such as LISA will record vast numbers of overlapping signals, and separating them from noise and from one another is a central data-analysis challenge. This paper explores a denoising approach based on embedding a time series into a Hankel matrix. A superposition of $N$ damped sinusoids corresponds to a matrix of rank $2N$, which turns signal extraction into a structured low-rank approximation problem. Three algorithms (ESPRIT, Cadzow iterations, and iteratively reweighted least squares) are benchmarked on scenarios that arise in gravitational-wave applications, from monochromatic LISA sources with closely spaced frequencies to black-hole quasinormal modes. All three reach near-optimal performance consistent with Fisher-matrix bounds, and a proof-of-concept application to numerical-relativity waveforms validates the approach on realistic ringdown signals.

Paper on arXiv