Resource
2025 EN
Andrea Pelissetto · Ettore Vicari
We consider quantum and classical first-order transitions, at equilibrium andunder out-of-equilibrium conditions, mainly focusing on quench and slowquasi-adiabatic protocols. For these phenomena, we review the finite-sizescaling theory appropriate to describe the general features of the large-scale,and long-time for dynamic phenomena, behavior of finite-size systems.
Resource
2025 EN
Emil Horobet · Ettore Teixeira Turatti
Subtracting a critical rank-one approximation from a matrix always results ina matrix with a lower rank. This is not true for tensors in general. Motivatedby this, we ask the question: what is the closure of the set of those tensorsfor which subtracting some of its critical rank-one approximation from it andrepeating the process we will eventually get to zero? In this article, we showhow to construct this variety of tensors and we show how this is connected tothe bottleneck points of the variety of rank-one tensors (and in general to thesingular locus of the hyperdeterminant), and how this variety can be equal toand in some cases be more than (weakly) orthogonally decomposable tensors.
Resource
2025 EN
Claudio Bonati · Haralambos Panagopoulos · Ettore Vicari
We address the out-of-equilibrium critical dynamics of the three-dimensionallattice ${\mathbb Z}_2$ gauge model, and in particular the criticalrelaxational flows arising from instantaneous quenches to the critical point,driven by purely relaxational (single-spin-flip Metropolis) upgradings of thelink ${\mathbb Z}_2$ gauge variables. We monitor the critical relaxationaldynamics by computing the energy density, which is the simplest localgauge-invariant quantity that can be measured in a lattice gauge theory. Thecritical relaxational flow of the three-dimensional lattice ${\mathbb Z}_2$gauge model is analyzed within an out-of-equilibrium finite-size scalingframework, which allows us to compute the dynamic critical exponent $z$associated with the purely relaxational dynamics of the three-dimensional${\mathbb Z}_2$ gauge universality class. We obtain $z=2.610(15)$, whichsignificantly improves earlier results obtained by other methods, in particularthose obtained by analyzing the equilibrium critical dynamics.
Resource
2025 EN
YNU Xiaohui Sun · Radboud Uni Marijke Haverkorn · INAF Ettore Carretti
+6 more
We present data processing and verification of the Southern Twenty-centimetreAll-sky Polarization Survey (STAPS) conducted with Murriyang, the Parkes 64-mtelescope. The survey covers the sky area of -89<Dec<0 and the frequency rangeof 1.3-1.8 GHz split into 1-MHz channels. STAPS was observed commensally withthe S-band Polarization All-Sky Survey (S-PASS). The survey is composed of longazimuth scans, which allows us to absolutely calibrate Stokes Q and U with thedata processing procedure developed for S-PASS. We obtain I, Q, and U maps inboth flux density scale (Jy/beam) and main beam brightness temperature scale(K), for the 301 frequency channels with sufficiently good data. Thetemperature scale is tied to the Global Magneto-ionic Medium Survey (GMIMS)high-band north sky survey conducted with the Dominion Radio AstrophysicalObservatory 26-m telescope. All the STAPS maps are smoothed to a commonresolution of 20 arcmin. The rms noise per channel ranges from about 16 mK to 8mK for I, and from about 8 mK to 5 mK for Q and U at frequencies from 1.3 to1.8 GHz. The rms noise in Q and U varies with declination and reaches minimumat declination of -89 degree. We also run rotation measure (RM) synthesis andRM clean to obtain peak polarized intensity and Faraday depth maps. The wholeSTAPS data processing is validated by comparing flux densities of compactsources, pixel flux density versus pixel flux density for Cen A, pixeltemperature versus pixel temperature for the entire survey area, and RMs ofextragalactic sources between STAPS and other measurements. The uncertainty ofthe flux density scale is less than 10%. STAPS delivers an L-band (20 cm)multi-frequency polarization view of the Galaxy, and will help advance ourunderstanding of the Galactic magnetic field and magnetized interstellarmedium.
Resource
2025 EN
Ettore Aldrovandi · Arash Karimi
We extend McCarthy's stabilization construction to exact $\infty$-categories.This is achieved by constructing, for any functor from exact$\infty$-categories to a fixed stable $\infty$-category $\mathcal{A}$, acoherent chain complex in $\mathcal{A}$ that is an immediate generalization ofMac Lane's cubical $Q$-complex computing the stable homology of abelian groups.
Resource
2025 EN
Emilio Corte · Alberto Bortone · Eleieto Hernández
+6 more
The Ion Beam Induced Charge (IBIC) technique is widely used to characterizethe electronic properties of semiconductor materials and devices. Its mainadvantage over other charge collection microscopies stems in the use of MeV ionprobes, which provide both measurable induced charge signals from single ions,and high spatial resolution, which is maintained along the ion range. It is afact, however, that the use of low-energy ions in the keV range can provide theIBIC technique with complementary analytical capabilities, that are notavailable with MeV ions, for example the higher sensitivity to the status,contamination and morphology of the surface and the fact that the inducedsignal depends on the transport of only one type of charge carrier. This paperoutlines the upgrade that was made at the 100 kV ion implanter of theUniversity of Torino, originally installed for material and surfacemodification, to explore the rather unexplored keV-IBIC field and to assess itspotential to characterize semiconductor devices. Finally, we report the firstIBIC application of our apparatus, which regards the assessment of theradiation damage of a commercially available silicon photodiode, adopting theIAEA experimental protocol and the relevant interpretative model.
Resource
2025 EN
Adrian Abac · Raul Abramo · Simone Albanesi
+482 more
Einstein Telescope (ET) is the European project for a gravitational-wave (GW)observatory of third-generation. In this paper we present a comprehensivediscussion of its science objectives, providing state-of-the-art predictionsfor the capabilities of ET in both geometries currently under consideration, asingle-site triangular configuration or two L-shaped detectors. We discuss theimpact that ET will have on domains as broad and diverse as fundamentalphysics, cosmology, early Universe, astrophysics of compact objects, physics ofmatter in extreme conditions, and dynamics of stellar collapse. We discuss howthe study of extreme astrophysical events will be enhanced by multi-messengerobservations. We highlight the ET synergies with ground-based and space-borneGW observatories, including multi-band investigations of the same sources,improved parameter estimation, and complementary information on astrophysicalor cosmological mechanisms obtained combining observations from differentfrequency bands. We present advancements in waveform modeling dedicated tothird-generation observatories, along with open tools developed within the ETCollaboration for assessing the scientific potentials of different detectorconfigurations. We finally discuss the data analysis challenges posed bythird-generation observatories, which will enable access to large populationsof sources and provide unprecedented precision.
Resource
2025 EN
Dewei Wang · Wei Zhu · Liyang Ling
+5 more
In the era of LLMs, dense operations such as GEMM and MHA are criticalcomponents. These operations are well-suited for parallel execution using atilebased approach. While traditional GPU programming often relies on low levelinterfaces like CUDA or SYCL, Triton has emerged as a DSL that offers a moreuser-friendly and portable alternative by programming at a higher level. Thecurrent Triton starts at the workgroup (aka threadblock) level, and directlylowers to per-thread level. And then attempt to coalesce and amend through aseries of passes, promoting information from low-level representation. Webelieve this is pre-mature lowering based on the below observations. 1. GPU hasa hierarchical structure both physically and logically. Modern GPUs oftenfeature SIMD units capable of directly operating on tiles on a warp orwarpgroup basis, such as blocked load and blocked MMA. 2. Multi-level graduallowering can make compiler decoupled and clean by separating considerationsinter and intra a logical layer. 3. Kernel developers often need fine controlto get good performance on the latest hardware. FlashAttention2 advocatesexplicit data partition between warps to make a performance boost. In thiscontext, we propose ML-Triton which features multi-level compilation flow andprogramming interface. Our approach begins at the workgroup level andprogressively lowers to the warp and intrinsic level, implementing a multilevellowering align with the hierarchical nature of GPU. Additionally, we extendtriton language to support user-set compiler hint and warp level programming,enabling researchers to get good out-of-the box performance without awaitingcompiler updates. Experimental results demonstrate that our approach achievesperformance above 95% of expert-written kernels on Intel GPU, as measured bythe geometric mean.
Resource
2025 EN
Andrea Pelissetto · Davide Rossini · Ettore Vicari
We study the out-of-equilibrium dynamics of one-dimensional quantum Isingmodels in a transverse field $g$, driven by a time-dependent longitudinal field$h$ across their {\em magnetic} first-order quantum transition at $h=0$, forsufficiently small values of $|g|$. We consider nearest-neighbor Ising chainsof size $L$ with periodic boundary conditions. We focus on theout-of-equilibrium behavior arising from Kibble-Zurek protocols, in which $h$is varied linearly in time with time scale $t_s$, i.e., $h(t)=t/t_s$. Thesystem starts from the ground state at $h_i\equiv h(t_i)0$, where $h_*(t_s)$ decreases with increasing $t_s$, apparently as$h_*\sim 1/\ln t_s$. This suggests the existence of a scaling behavior in termsof the rescaled time $\Omega = t \ln t_s/t_s$. The numerical results also showthat the system converges to a nontrivial stationary state in the large-$t$limit, characterized by an energy significantly larger than that of thecorresponding homogeneously magnetized ground state.
Resource
2025 EN
Ettore Caici · Filippo Caruso
Technological advances in Artificial Intelligence (AI) and Machine Learning(ML) for the healthcare domain are rapidly arising, with a growing discussionregarding the ethical management of their development. In general, MLhealthcare applications crucially require performance, interpretability ofdata, and respect for data privacy. The latter is an increasingly debated topicas commercial cloud computing services become more and more widespread.Recently, dedicated methods are starting to be developed aiming to protect dataprivacy. However, these generally result in a trade-off forcing one to balancethe level of data privacy and the algorithm performance. Here, a Support VectorMachine (SVM) classifier model is proposed whose training is reformulated intoa Quadratic Unconstrained Binary Optimization (QUBO) problem, and adapted to aneutral atom-based Quantum Processing Unit (QPU). Our final model does notrequire anonymization techniques to protect data privacy since the sensitivedata are not needed to be transferred to the cloud-available QPU. Indeed, thelatter is used only during the training phase, hence allowing a future concreteapplication in a real-world scenario. Finally, performance and scaling analyseson a publicly available breast cancer dataset are discussed, both using idealand noisy simulations for the training process, and also successfully tested ona currently available real neutral-atom QPU.