Talks

The Kawai–Lewellen–Tye (KLT) relations are a striking example of how string theory can reveal hidden structures and connect a web of, in principle, very different theories. Originally, they state that (tree-level) closed-string amplitudes can be expressed as quadratic combinations of open-string amplitudes. In the field theory limit, this gives rise to the celebrated double-copy between gluons and gravitons. In this talk, I will discuss how much of this elegant structure survives away from flat space, focusing on AdS. As a first step, I will propose the fundamental building blocks of tree-level string amplitudes in AdS and then show how they are related via an AdS version of the KLT relations, whose kernel can be computed exactly. This structure not only significantly simplifies explicit calculations but also points to deeper algebraic and geometric structures to be uncovered.

This talk will be a brief and interactive introduction to my current research and the cosmological concepts that go along with it. I will assume limited prior knowledge of GR and cosmology so that even students in unrelated fields can learn something interesting.

When dealing with QFTs or with quantum mechanical systems in the thermodynamic limit, one encounters uncountable Hilbert spaces. These are mathematically wilder and physically less realistic than their countable counterparts. In this talk, we will explore how vacua encode a choice to reduce the Hilbert Space in size. In spite of the mathematically sounding intro, we’ll see this has interesting physical implications.

Current waveform models of binary black hole mergers incorporate a large amount of analytical information during the inspiral phase. In contrast, post-merger descriptions typically rely on phenomenological ansätze informed by numerical relativity, with the quasi-normal mode frequencies providing the only direct analytical input. In this seminar, I will first give a gentle introduction to black hole perturbation theory and the Green’s function approach, which is useful for describing the time-domain response of a black hole to a dynamical source, e.g., an infalling test particle. I will then present an analytical framework to study the dynamical excitation of quasi-normal modes during the plunge–merger–ringdown stages for generic orbits, including highly eccentric ones, and discuss its implications for ringdown modeling. I will conclude by presenting novel insights into the modelling of the prompt response, fundamental to correctly capture the signal emitted during the merger at the linear level. 

As a class of generative artificial intelligence frameworks inspired by statistical physics, diffusion models have shown extraordinary performance in synthesizing complicated data distributions through a denoising process gradually guided by score functions. Real-life data, like images, is often spatially structured in low-dimensional spaces. However, ordinary diffusion models ignore this local structure and learn spatially global score functions, which are often computationally expensive. In this work, motivated by recent advances in statistical physics, we develop a generic framework for defining phases of data distributions and use it to analyze the locality requirements of denoisers in diffusion models. We define two distributions as belonging to the same data distribution phase if they can be mutually connected via spatially local operations such as local denoisers. We demonstrate that the reverse denoising process consists of an early trivial phase and a late data phase, sandwiching a rapid phase transition where local denoisers must fail. We further demonstrate that the performance of local denoisers is closely tied to spatial Markovianity, which provides an operational criterion for diagnosing such phase transitions. We validate this criterion through numerical experiments on real-world datasets. Our work suggests guidance for simpler and more efficient architectures of diffusion models: far from the phase transition point, we can use small local neural networks to compute the score function; global neural networks are only necessary around the narrow time interval of phase transitions. This result also opens up new directions for studying phases of data distributions, the broader science of generative artificial intelligence, and guiding the design of neural networks inspired by physics concepts.