PRIN 2022 Projects at INFN Padova

AxionOrigins: towards a complete theory for the origin of the axion – PRIN 2022K4B58X
P.I. and INFN Scientific coordinator: Di Luzio Luca
Total budget € 105.136,00 – MUR funding € 98.607,00
Start: 28/09/2023 – End: 28/02/2026

The AxionOrigins project aims to develop complete theories of the axion that address the origin and quality of the Peccei-Quinn (PQ) symmetry, exploring connections with the Standard Model (SM) flavor puzzle. The project seeks to identify new SM extensions, based on horizontal gauge symmetries, composite dynamics, and grand-unified theories, capable of selecting theoretically motivated ranges for the axion mass and coupling parameters. Expected outcomes include linking axion physics with low-energy observables and deriving new constraints on axion dark matter. So far, connections between horizontal symmetries and PQ have been established, with implications for low-energy axion phenomenology.

Cold paramagnetic polar molecules: from particle physics to quantum technology – PRIN 20227F5W4N
INFN Scientific coordinator: Carugno Giovanni
Total budget totale € 70.873,00 – MUR funding € 36.577,00
Start: 28/09/2023 – End: 28/02/2026

The electric dipole moment (EDM) of elementary particles has become a fundamental parameter to be measured. In fact, a permanent EDM of a particle with spin angular momentum would directly violate T symmetry, thus becoming a strong indication for the existence of new physics Beyond the Standard Model (BSM). Moreover, following Sakharov criteria, a “T” violation is mandatory to explain the macroscopic matter-antimatter imbalance in our Universe. In a BSM theory, the EDM of elementary particles could provide constraints on the mass of BSM particles higher than TeV scale, much larger of the LHC accessible energy scale. In the last years, the use of cold beams of reactive molecules has led to a limit of 10-30 e cm for the electron EDM. New proposals, mostly based on atomic, molecular and optical techniques, have recently appeared with increasing electron EDM sensitivity. Among such proposals, a significant improvement is expected from systems based on polar molecules embedded in an inert gas cold matrix. This project aimed to set-up an experimental apparatus doping para-hydrogen (pH2) cryo cristals with dipolar paramagnetic molecules.
During the last year we focused our experimental efforts growing para-Hydrogen cryogenic crystals at few kelvin temperature doing spectroscopy studies on such matrix so to characterize the ortho-Hydrogen content of such crystals. A different set-up has been used to produce BaF molecules in flight via ablation approach resulting in a high yield production.

Unveiling the role of low dimensional activity manifolds in biological and artificial neural networks – PRIN 2022HSKLK9
INFN Scientific coordinator: Zucchetta Alberto
Total budget totale € 84.800,00 – MUR funding € 84.800,00
Start: 28/09/2023 – End: 28/02/2026

The research project “Unveiling the role of low dimensional activity manifolds in biological and artificial neural networks” focused on characterizing the collective properties of neuronal activity. The fundamental premise of this research is that the brain does not function merely as a collection of isolated units, but rather it organizes its activity along low-dimensional geometric structures known as “neural manifolds.” Identifying the intrinsic dimension of
these structures is essential for determining how many parameters are effectively required to describe neural computation and how information is encoded and processed by the system.

From a methodological perspective, the project achieved a significant breakthrough with the development of the local Full Correlation Integral (lFCI) algorithm. Previous dimensionality estimators suffered from structural limitations, tending to overestimate complexity in the presence of curved manifolds or drastically underestimating it when the ground-truth dimension exceeded ten.
The new lFCI pipeline overcomes these issues by combining local density estimates with curvature indicators, thereby providing a reliable global measure even for high-dimensional systems. This tool has been validated on both synthetic and biological datasets and has been released as open-source software. 
The application of these techniques to experimental data, specifically calcium imaging of zebrafish larvae and electrophysiological recordings from rodent models, clarified the link between neural complexity and behavior. The results indicate that activity dimensionality is not a fixed property but is highly context-dependent. In the presence of simple sensory stimuli or structured tasks, the brain compresses information into low-dimensional manifolds, ensuring highly efficient processing.
Conversely, during spontaneous activity or at rest, the system explores a much broader and more complex geometry, characterized by high dimensionality and correlations reminiscent of physical systems approaching a phase transition. 
In parallel, the study of Artificial Neural Networks (ANNs) provided a robust theoretical framework for interpreting biological data. By training artificial networks on batteries of cognitive tasks, the spontaneous emergence of geometric representations analogous to those found in vivo has been observed.
The project further investigated learning mechanisms through the “kernel renormalization” theory, explaining how the training process shapes the internal features of networks to optimize performance. This integrated approach confirms that artificial networks are not merely computational tools but are reliable models for testing hypotheses
regarding the organizational principles of the biological brain.