davidegerosa.com

Targeting black holes from metal-poor progenitors with next-generation gravitational-wave detectors

2026-06-04T00:00:00+00:00

We have a 3G paper today, led by Federico Leto di Priolo (recent MSc graduate here) and also featuring a new collaboration with Martyna Chruślińska from ESO. A 3G detector will see black holes at high redshift, which means they will come from stars with different properties—namely, stars with low metallicity. Now, how far should you look if you’re specifically interested in the low-metallicity population (that is, what’s your “target redshift”)? And how many events do you need to do astrophysics there?

F. Leto di Priolo, M. Chruślińska, D. Gerosa.
arXiv:2606.03776 [astro-ph.HE].

redfive standing by

2026-05-01T00:00:00+00:00

In the past few months, we’ve been setting up a new computer cluster called redfive (why redfive?…). Right now, it consists of a login node and five compute nodes, totaling about 100 CPUs. It is developed and maintained entirely in-house (mostly by Rodrigo Tenorio and Federico De Santi, while I prepare coffee during their admin sessions!), which is great fun. It’s not meant to substitute a supercomputer, of course, but rather to provide a simple, easy-to-expand, no-queuing setup for mid-sized jobs, as well as a way to introduce students to remote servers. We need a GPU now!

Gravitational-wave astronomy requires population-informed parameter estimation

2026-04-20T00:00:00+00:00

Uninformative priors are good because they’re uninformative, right? Not really. We argue they’re bad because they are uninformative, i.e. they leave out information. Now, that’s the best you can do if you don’t have that information, but if you do and decide to ignore it, then it becomes your fault. This is the case in our beloved GW astronomy. When we analyze a GW event, we do have some information on what it should look like because we’ve seen many other events. If we ignore that information, then it’s no surprise that the results can be crazy biased.

M. Mould, R. Tenorio, D. Gerosa.
arXiv:2604.15885 [gr-qc].

Alice’s (long) phd defense

2026-04-02T00:00:00+00:00

Yeee! Alice Spadaro got her PhD yesterday! She defended her PhD thesis titled “Methods for exploiting gravitational-wave data from massive black-hole binaries.” The thesis was refereed by Laura Blecha and James Alvey, and the defense committee had John Veitch, Enrico Barausse, and Massimo Dotti. They had many, many (many!) questions because Alice’s thesis was just too interesting! Alice will be moving to Toulouse (France) for her postdocs. Thanks Alice for working with me/us!

Alice likes surfing, merging black holes, her purple hoodie… and deals with DINGOs!

A graduation session just for us!

2026-03-27T00:00:00+00:00

A whopping seven students graduated with projects in our group yesterday! It was kind of ridiculous—they had to set up an entire graduation session just for us. The topics were very diverse, offering a great cross-section of our group’s activities as well as our collaboration network.

Simone Abeni earned his BSc degree, co-supervised by Floor Broekgaarden (San Diego), with a project on NS–BH binaries;
Giulia Conti earned her MSc degree, co-supervised by Philippa Cole (Queen Mary University of London), with a project on EMRI parameter estimation in realistic LISA noise;
Alessia Corelli earned her MSc degree, co-supervised by Cecilia Fabbri (Nottingham), with a project on LVK populations (which we submitted already!);
Federico Leto di Priolo earned his MSc degree, co-supervised by Martyna Chruślińska (ESO), with a project on stellar metallicity and 3G detectors;
Oliver Rossi earned his MSc degree, co-supervised by Tristan Bruel (postdoc here), with a project on globular cluster dynamics;
Pietro Zeduri earned his MSc degree, co-supervised by Michele Mancarella (Marseille), with a project on testing GR with 3G detectors.

Congratulations all, and best of luck with everything that comes next!

Accurate and efficient simulation-based inference for massive black-hole binaries with LISA

2026-03-26T00:00:00+00:00

This paper presents the very first application of DINGO to the LISA parameter-estimation problem! DINGO is an Australian animal that eats BILBY, and a powerful new gravitational-wave method for estimating source parameters. It is (at least in my view) the most modern and compelling application of AI to gravitational-wave parameter estimation, and it has already shown tremendous promise for the characterization of LIGO–Virgo sources. Even though it relies on some approximations, this paper demonstrates that the same AI-based approach can also be successfully applied to supermassive binary black holes in LISA.

A. Spadaro, J. Gair, D. Gerosa, S. R. Green, R. Buscicchio, N. Gupte, R. Tenorio, S. Clyne, M. Purrer, N. Korsakova.
arXiv:2603.20431 [astro-ph.HE].

Joint PhD position in gravitational-wave astronomy, Milan (Italy) and Marseille (France)

2026-03-19T00:00:00+00:00

The University of Milano-Bicocca (Italy) and Aix-Marseille University (France) welcome applications for a joint PhD scholarship in gravitational-wave astronomy. The application deadline is April 17th, 2026 (noon CEST) for positions starting in Fall 2026:

https://en.unimib.it/study/doctoral-research-phd-programmes/applying-doctorate/calls-application

We are seeking highly motivated candidates to join our research in black-hole binary dynamics, gravitational-wave data analysis, and cosmology. Both Milano-Bicocca and Aix-Marseille host large gravitational-wave groups; candidates will benefit from both research environments as well as their broad network of external collaborators.

This joint position will be supervised by Davide Gerosa at Milano-Bicocca and Michele Mancarella at Aix-Marseille. The selected candidate is expected to spend approximately two years in Milan and one year in Marseille, although this arrangement is flexible.

Applications should be submitted through Milano-Bicocca using the link above. The PhD admission program includes several “open scholarships”, covering all research areas in the department (including ours), for which all applicants are automatically considered. Candidates interested in this joint position should explicitly indicate their interest by selecting the “project scholarship” number PROG.5, titled “Interpreting gravitational-wave data with modern inference techniques.”

For more information on our research groups, see www.davidegerosa.com/group and www.cpt.univ-mrs.fr/~mancarella.

We are committed to building an inclusive research environment and welcome applications from all interested candidates. For informal inquiries or expressions of interest, please contact davide.gerosa@unimib.it and michele.mancarella@cpt.univ-mrs.fr.

Teaching this year

2026-03-02T00:00:00+00:00

Our teaching term started today, and I’ll be busy with a few classes here and there. Let’s recap my year of teaching:

For PhD students, “Scientific computing with python” took place in the winter of 2025. github.com/dgerosa/scientificcomputing_bicocca_2025
For MSc students, it’s time for another edition of my flagship class on “Astrostatistics and machine learning”. github.com/dgerosa/astrostatistics_bicocca_2026
…and for BSc students in the AI degree, you’ll see me for my lab titles “Machine learning for physics and astronomy”. dgerosa/machinelearning4physics_bicocca_2026
But there’s a bit more this year. I was invited to teach a one-day mini-course in Lisbon, on “Bayesian inference and stochastic sampling in (astro)physics”. github.com/dgerosa/bayesianinference_lisbon_2026
I will also deliver a class on “Gravitational waves from compact binarie” at this PhD school in Florence. github.com/dgerosa/gwbinaries_florence_2026

Variance of gravitational-wave populations

2026-02-28T00:00:00+00:00

Say we live in another Universe, and in that Universe of course we are still collecting gravitational-wave data. Those data will kind of look like ours, but not quite, they’ll be another realization of the same statistical process. What I’m trying to say is that the astrophysics we infer from GW data is, of course, constrained by the actual data we are collecting; this is equivalent to saying that there is an intrinsic variance due to the finite size of our event catalog. (If I could have infinitely many events, GW data from all parallel Universes would look the same!)

Here we try to quantify this variance using statistical bootstrapping, and it turns out it’s not negligible at all. Some technical details include bootstrapping over both events and sensitivity injections to avoid messing up selection effects. This is work by Alessia Corelli, a current MSc student with me (who managed to submit a paper even before graduating!) together with our friends in Nottingham, Matt and Cecilia.

A. Corelli, D. Gerosa, M. Mould, C. M. Fabbri.
arXiv:2603.00239 [astro-ph.HE].

Postdoc opportunity from the Cariplo Foundation

2026-02-26T00:00:00+00:00

Happy to share a postdoc opportunity from the Cariplo Foundation, which is a private trust that operates in northern Italy:

www.fondazionecariplo.it/wp-content/uploads/2026/02/Young-researchers-2026.pdf

It’s an independent fellowship for early-career researchers, with a duration of 3+1 years and a total budget of 200k EUR. Their eligibility criteria are kind of specific though; they only want people who got their PhD between April 14th, 2024 and April 14th, 2026 (which is the submission deadline). The starting date is flexible, sometime between January 2027 and September 2027.

If anyone here is interested in applying with us at Milano-Bicocca please shoot us an email!