About
I am a theoretical physicist exploring the boundaries of the Standard Model of particle physics and the Lambda cold dark matter model of cosmology. My approach integrates a wide spectrum of methods, ranging from traditional model building in Quantum Field Theory to Quantum Information Theory and interpretable Artificial Intelligence.
Interests
- Beyond the Standard Model: flavour, naturalness, dark matter.
- Beyond the ΛCDM: primordial black holes, modified galaxy formation.
- Phenomenology at/with: future colliders, nuclear facilities, cosmological probes.
- Fusion of Quantum Information-theoretic concepts with High-Energy Physics.
- Interpretable Artificial Intelligence for Physics.
Education
PhD in Theoretical Physics, 2020
University of Zürich
MSc in Physics, 2016
ETH Zürich
BSc in Physics, 2014
Technical University of Dortmund
Experience
Research Associate
Scientific Associate
PostDoc
MIT, IAIFI
PostDoc
SISSA
Recent Publications
For a complete list of publications see my Inspire page.
*
Large language models are increasingly promising on research-level physics reasoning tasks, but it is unclear how researcher-agent interactions affect outcomes. We introduce SCALAR, an Actor-Critic-Judge pipeline for quantum field theory and string theory problems, where an Actor proposes solutions, a Critic gives iterative feedback, and an independent Judge evaluates against reference solutions. Across model families, multi-turn dialogue improves over single-shot attempts, while the value of prompting choices and critique styles depends strongly on the Actor-Critic pairing. SCALAR provides a controlled testbed for identifying interaction structures that help or hinder AI-driven scientific discovery.
The collapse of large-amplitude primordial curvature perturbations into planetary-mass primordial black holes generates a scalar-induced gravitational-wave background in the microhertz range that may be detectable by future Lunar Laser Ranging and Satellite Laser Ranging data. We derive projected constraints on the primordial black hole population from a null detection of such a background, including the impact of the electroweak phase transition, and discuss the connection to recent microlensing observations of the Andromeda Galaxy.
We investigate invisible final states in coherent exclusive electroproduction at the Electron-Ion Collider (EIC), where the characteristic signal is a forward proton with reduced energy and little additional detector activity. Focusing on neutral mesons with highly suppressed invisible Standard Model decays and on invisibly decaying gluon-coupled axion-like particles, we find that the EIC could improve existing bounds on invisible decays by up to four orders of magnitude while probing ALPs in the 0.1-2 GeV mass range.
We study the cosmology of axions in the post-inflationary scenario, where random initial conditions and the ensuing string-domain-wall network generate an isocurvature power spectrum. For subdominant axion fractions, the resulting white-noise fluctuations enhance the growth of structure in the dominant component. Combining ultraviolet luminosity functions of galaxies with Lyman-$\alpha$ and CMB data yields leading cosmological limits on post-inflationary axion dark matter, linking early-universe field dynamics to high-redshift structure formation.
We investigate quantum-information-theoretic observables in electron-proton scattering at the Electron-Ion Collider (EIC). Our analysis focuses on entanglement and magic, two complementary indicators of non-classicality in quantum states. We show that while unpolarized and longitudinally polarized beams yield unentangled separable outcomes, transverse beam polarization enables the generation of entangled and non-stabilizer states. This result holds for both elastic and deep inelastic electron-proton scattering in QED, with the deep inelastic regime linking the strength of quantum correlations to transversity parton distribution functions.
The detection of compact binary mergers with sub-solar masses at gravitational-wave observatories could mark the groundbreaking discovery of primordial black holes. Concurrently, evidence for a nHz stochastic gravitational wave background observed by pulsar timing arrays could suggest a non-astrophysical origin, potentially arising from scalar-induced gravitational waves. In this work, we analyze the connection between the two phenomena in the case where they share a common origin: the collapse of large primordial curvature perturbations in the early universe.
The quantum property of non-stabiliserness, also known as magic, plays a key role in designing quantum computing systems. How to produce, manipulate and enhance magic remains mysterious, such that concrete examples of physical systems that manifest magic behaviour are sought after. In this paper, we study two-particle scattering of gluons and gravitons in Yang–Mills theory and General Relativity, as well as their supersymmetric extensions. This provides an interesting case of two-qubit systems, differing only in the physical spin of the qubits. We show that magic is generically produced in both theories, and also show that magic typically decreases as the spin of the qubits increases.
We present an AI model that not only achieves cutting-edge precision for nuclear masses, but does so in an interpretable manner. For example, we find (and explain why) that the most important dimensions of its internal representation form a double helix, where the analog of the hydrogen bonds in DNA here link the number of protons and neutrons found in the most stable nucleus of each isotopic chain. Furthermore, we show that the AI predictions can be factorized and ordered hierarchically, with the most important terms corresponding to well-known symbolic models. Remarkably, the improvement of the AI model over symbolic ones can almost entirely be attributed to an observation made by Jaffe in 1969.
We present updated constraints on the abundance of primordial black holes (PBHs) dark matter from the high-redshift Lyman-$\alpha$ forest data from MIKE/HIRES experiments. Our analysis leverages an effective field theory (EFT) description of the 1D flux power spectrum, allowing us to analytically predict the Lyman-&alpha fluctuations on quasi-linear scales from first principles.
Part of the energy created in deuterium-tritium fusion reactors is carried away from plasma by a high-intensity neutron flux, which is then absorbed by the inner walls of the reactor. The interactions of neutrons with the materials within the walls can result in the production of dark sector particles, feebly interacting light scalars or pseudoscalars, via nuclear transitions. We estimate the potential size of such dark sector flux and consider possible detection methods at current and future thermonuclear fusion reactors.
We speculate on the (unlikely but tantalizing) possibility that minimization of quantum entanglement might be a fundamental principle that determines particle physics input parameters. In a specific limit of electroweak-mediated quark scattering, entanglement is minimized when the CKM matrix is almost, but not exactly, diagonal. Applying the same logic to leptons favors a PMNS matrix with two large angles and a smaller one, together with a hint for suppressed CP violation.
We derive a new metastability bound for the Higgs mass parameter. When the bound for the positive mass parameter is significantly more restrictive than or comparable to its counterpart for the negative, it may offer an explanation for the spontaneous symmetry breaking itself.
Mechanistic Interpretability promises a path toward fully understanding how neural networks make their predictions. We argue that neural networks can learn low-dimensional representations of their training data that are surprisingly faithful to human-derived domain knowledge. As a case study, we extract nuclear physics concepts by studying models trained to reproduce nuclear data.
We introduce Nuclear Co-Learned Representations (NuCLR), a deep learning model that predicts various nuclear observables, including binding and decay energies, and nuclear charge radii. The model is trained using a multi-task approach with shared representations and obtains state-of-the-art performance, achieving levels of precision that are crucial for understanding fundamental phenomena in nuclear (astro)physics.
Recent observations have granted to us two unique insights into the early Universe: the presence of a low-frequency stochastic gravitational wave background detected by the NANOGrav and Pulsar Timing Array experiments and the emergence of unusually massive galaxy candidates at high redshifts reported by the James Webb Space Telescope. We consider the possibility that both observations have a common origin, namely supermassive primordial black holes.
The emission of collinear radiation off an elementary lepton can be factorised from the hard scattering process by introducing Parton Distribution Functions of a Lepton (LePDF), which, contrary to protons, can be derived from first principles. In case of multi-TeV lepton colliders, such as the muon colliders currently being proposed, the complete structure of Standard Model interactions must be taken into account.
News
Quantum Observables for Collider Physics 2026 brought the collider and quantum-information communities together at CERN. The CERN Courier meeting report by Alan Barr and me captures the highlights.
I co-organized PLANCK2026 and the 6th EuCAPT Symposium, which brought particle physics, cosmology and astroparticle theory together at CERN.
I was honored to give the DESY colloquium in Hamburg on axions across cosmology, nuclear physics, and collider phenomenology.
A New Scientist feature explores whether entanglement could be a deeper organising principle in physics, including our result that minimising entanglement reproduces small quark flavour mixing while favouring a more anarchic lepton-sector PMNS pattern.
Contact
If you have any questions, send me a message at:
- sokratis.trifinopoulos@northwestern.edu
- 2145 Sheridan Road, Tech F165, Evanston, Illinois, 60208-3112