13675 research outputs found
Sort by
FAIR Data Management of the Results from the MulSKIPS Atomistic Simulation Environment for PVD, CVD, and Laser Annealing
No full textThis thesis presents the design and implementation of a FAIR
(Findable, Accessible, Interoperable, Reusable) data management
workflow for atomistic simulations of 3C-SiC growth via Physical
Vapor Deposition (PVD), using the MulSKIPS multiscale Kinetic
Monte Carlo framework [1–3]. The simulation engine is capable
of capturing extended defect formation—including stacking faults
(SFs) and antiphase boundaries (APBs)—with atomistic resolution under experimentally relevant conditions [3, 4].
A central achievement of this work is the development of a
Python-based parser that automates the extraction of simulation
metadata and results, generating NeXus files that conform to
FAIRmat’s contributed definitions NXmicrostructure_imm_config
and NXmicrostructure_imm_results [5]. This enables machineactionable, semantically rich data outputs that are compatible
with the NOMAD repository [6] and the European Open Science
Cloud (EOSC) ecosystem [7].
The simulation–data integration pipeline was validated on
PVD simulations of 3C-SiC substrates, demonstrating reproducibility, robust metadata curation, and automated defect quantification. While the implementation was tested on PVD only, the
modular architecture of the workflow is readily extensible to Chemical Vapor Deposition (CVD) and Pulsed Laser Annealing (PLA)
simulations, supporting the future development of interoperable
digital twins for materials processing [4, 8]
Unraveling Molecular Complexity through Single Cell approaches: From Olfactory Bulb Architecture to Dorsal Root Ganglia Remodeling in Pancreatic Cancer
No full textThe accessory olfactory system and the role of the nervous system in pancreatic cancer represent two critical yet underexplored areas in neuroscience and cancer biology. This thesis employs transcriptomic and spatial profiling techniques to investigate the cellular and molecular diversity of the olfactory bulb (OB), and examine the molecular changes in sensory neurons innervating cancer pancreas.
In the first part of this work, I analyzed single-nucleus RNA sequencing (snRNA-seq) and spatial transcriptomics data to profile the cellular composition of the olfactory bulb (OB), with a particular emphasis on the distinct transcriptomic signatures of cell types in the accessory (AOB) and main olfactory bulb (MOB). By integrating datasets enriched for cells from both regions, I identified unique excitatory neuronal subpopulations in the AOB marked by genes such as Nrp2, Tbx21, and Fst, while Nrp1 specifically marked excitatory neurons in the MOB. Additionally, Trp73 marked cells within the glomerular layer of both regions. These findings reveal previously unrecognized molecular diversity within the AOB, adding to our understanding of the accessory olfactory system and provide insight into the conserved and region-specific features of the AOB and MOB. Spatial transcriptomic analysis not only validated these observations but also identified Cntn6 and Ndst4 as molecular markers for the AOB excitatory neurons.
In the second part, I characterized sensory neurons innervating the pancreas in both healthy and KPC (pancreatic cancer) mouse models by analyzing single-cell RNA sequencing data generated after retrograde tracing. Neurofilament (NF) neurons emerged as the predominant DRG type, while non-peptidergic (NP) neurons were selectively reduced in KPC mice. Notably, Pdx1-CreERT2 transcripts were detected specifically in DRGs innervating the KPC pancreas, suggesting potential transfer of tumor model–derived RNA via extracellular vesicles. Gene expression analysis also revealed mitochondrial alterations, particularly in NP and peptidergic (PEP) neurons, indicating cancer-associated metabolic reprogramming. These transcriptomic
findings were supported by immunofluorescence, which confirmed the relative abundance of DRG types innervating the pancreas, and by RT-qPCR, which validated the presence of transgene-derived transcripts in DRGs. Additionally, MitoRed staining combined with IF further validated mitochondrial changes at the protein level, reinforcing the transcriptional evidence of altered metabolic activity in DRGs during pancreatic cancer. Together, these findings underscore the dynamic interplay between cancer cells and the nervous system and suggest new avenues for early detection and therapeutic intervention in pancreatic cancer. Overall, this thesis advances our understanding of the molecular and cellular mechanisms underlying olfactory processing and cancer-neuron interactions
An alternative GPU acceleration for a pseudopotential plane-waves density functional theory code with applications to metallic systems
No full textWe present an alternative GPU acceleration for plane waves pseudopotentials electronic structure codes designed for systems that have small unit cells but require a large number of k points to sample the Brillouin zone as happens, for instance, in metals. We discuss the diagonalization of the Kohn and Sham equations and the solution of the linear system derived in density functional perturbation theory. Both problems take advantage from a rewriting of the routine that applies the Hamiltonian to the Bloch wave-functions to work simultaneously (in parallel on the GPU threads) on the wave-functions with different wave-vectors k, as many as allowed by the GPU memory. Our implementation is written in CUDA Fortran and makes extensive use of kernel routines that run on the GPU (GLOBAL routines) or can be called from inside the GPU kernel (DEVICE routines). We compare our method with the CPUs only calculation and with the approach currently implemented in Quantum ESPRESSO that uses GPU accelerated libraries for the FFT and for the linear algebra tasks such as the matrix-matrix multiplications as well as OpenACC directives for loop parallelization. We show in a realistic example that our method can give a significant improvement in the cases for which it has been designed
The XY model with vision cone: non-reciprocal vs. reciprocal interactions
No full textWe study the behavior of the classical XY model on a two-dimensional square lattice, with interactions occurring within a vision cone of each spin. Via Monte Carlo simulations, we explore one non-reciprocal and two reciprocal implementations of these interactions. The corresponding energy involves couplings that depend non-trivially on the system's configuration, leading to both long-range and quasi-long-range ordered phases at low temperatures. Our results demonstrate that non-reciprocity is not essential for achieving long-range order at low temperatures. Using symmetry arguments, we provide a theoretical framework to explain these findings, and additionally we uncover an unexpected order-by-disorder transition
FAIR data management of quantum-mechanical calculations for the spin states dynamics in SiC materials
No full textThis thesis addresses the application of FAIR data management principles
to quantum-mechanical simulations, using spin state dynamics in Silicon Car
bide (SiC) materials for quantum sensing as a case study.
It begins by ex plaining the importance of FAIR data management, emphasizing significance and benefits of adopting such practices.
Subsequently, a practical research case concerning a quantum enhanced
magnetometer in SiC is presented. This case served as a basis for identifying
the specific types of data and metadata involved in such simulations.
The core of the thesis describes the practical implementation methodology, detailing how FAIR compliance of simulation’s data was achieved using
the NOMAD platform through the development of a specialized plugin for
the QuTiP based calculations.
Finally, the thesis demonstrates a FAIR-by-design workflow through an
example incorporating all types of data and metadata structured by the
plugin’s schemas
Catching a CAPTCHA: the impact of variable input on the processing of emerging orthographic representations
No full textVariability inherent to handwriting has been suggested to help establish more robust letter representations than other methods (e.g., typing). The present study tests whether encoding letter strings from a novel alphabet becomes more resistant to distortion when trained with variable input. Over 5 days, participants learned an 11-character artificial alphabet in a variable handwritten format involving reading, listening and handwriting practice. Another set of 11 artificial characters served as a visual control. Before and after the training, participants completed a masked priming same–different matching task with the novel alphabet letters. The key manipulation was in the primes: the identity/unrelated primes could be presented in a printed or distorted format. Results showed identity priming in both conditions, with a stronger effect for the printed primes. These effects increased post training for experimental and visual control scripts, indicating that exposure to variable input enhances distortion resistance even without explicit training. A second experiment assessed the transposed-letter effect – another marker of orthographic processing – in the novel scripts with an unprimed same–different matching task. Results showed that the transposed-letter effect occurred similarly before and after the training for both scripts. Therefore, letter shape variability when learning to read does not seem to boost orthographic processing
Metric and Probabilistic Aspects of Grassmann and Tensor Geometry
No full textMetric Algebraic Geometry is an emerging field that connects algebraic and differential geometry through the study of metric properties of real algebraic varieties, motivated by applications to the sciences. In this thesis we address three different problems in tensor and Grassmann geometry that showcase the interplay between differential, analytic and algebraic aspects typical of Metric Algebraic Geometry.
The first problem deals with estimating the probability that a random real symmetric tensor is close to rank–one tensors. We show how this problem is equivalent to estimating the volume of a neighborhood of the real Veronese variety. We study metric invariants of the Veronese variety and give explicit formulae for its reach and curvature coefficients with respect to the Bombieri–Weyl metric. A description of its second fundamental form is also obtained in terms of matrices from the Gaussian Orthogonal Ensemble. We give a closed formula for the probability of being close to rank–one and show that in the case
of rational normal curves it has an exponential decay with respect to the order of the tensor.
In the second part we study typical ranks of real m×n×l tensors. We introduce a new geometric framework through which we express rank probabilities for Gaussian tensors in terms of the probabilities of having enough points in the intersection between a random linear subspace and the Segre variety of rank–one matrices. We show that in the range (m−1)(n−1)+1 ≤ l ≤ mn typical ranks are contained in {l, l+1} and l is always typical. Based on asymptotic results on the average number of real points in a random slice of the Segre variety with a complementary dimension subspace, we give some heuristics on how the rank probabilities behave. For 3 × 3 × 5 tensors, for which typical ranks are known to be 5 and 6, we relate the rank probabilities to the probability that a random determinantal cubic surface contains only real lines. As a by–product, we obtain bounds on the expected number of real lines on such a cubic.
The last problem constitutes the main project carried out during my PhD studies.
Motivated by the notion of Euclidean Distance Degree, which is a measure of the complexity of the nearest point problem to an algebraic variety in a Euclidean space, we introduce the concept of Grassmann Distance Complexity (GDC) of a subanalytic set in the Grassmannian. This measures the complexity of solving the nearest point problem when the subset is subanalytic and the distance used is the Riemannian one. As this distance is not smooth nor semialgebraic, we employ tools from subgradient theory for locally Lipschitz functions and o–minimial geometry. We prove key properties of the GDC, such as its finiteness, and compute explicit bounds for real algebraic hypersurfaces. In the last part we explicitly solve the nearest point problem for a class of simple Schubert varieties through a non–linear version of Eckart–Young theorem.
The research we present in this thesis can be found in the following publications and preprints:
• A. Cazzaniga, A. Lerario, A. Rosana, What Is the Probability That a Random Symmetric Tensor Is Close to Rank–One?, SIAM Journal on Applied Algebra and Geometry, 8(2):227-258, 2024.
• P. Breiding, S. Eggleston, A. Rosana, Typical Ranks of Random Order–Three Tensors, International Mathematics Research Notices, Volume 2025, Issue 4, 2025.
• A. Lerario, A. Rosana, The Grassmann distance complexity, arxiv:2411.1658
Theory of Transformers and their application to Neural Quantum States
No full textMy PhD research focused on the Transformer architecture, a powerful deep neural network model that has emerged as a cornerstone for solving complex problems in natural language processing, image analysis, signal processing, and beyond. In particular, we studied the learning dynamics of this architecture, and its application to the representation of many-body wavefunctions, the so-called Neural Quantum States. Initially, we investigated the representational capabilities of Transformers by characterizing the statistical structures that a simplified Transformer layer, utilizing the so-called factored attention, is capable of learning. Building on these results, we utilized factored attention in deep Transformers to develop an accurate ansatz for approximating the ground states of quantum many-body Hamiltonians within the variational Monte Carlo framework. In this specific application, factored attention is crucial for achieving accurate results, demonstrating superior performance compared to the standard attention mechanism used in most of the other applications of the Transformers, and in particular in natural language processing. Alongside the development of an efficient optimization method for large-scale neural networks, we achieved state-of-the-art results on the most popular benchmark in Neural Quantum States and addressed complex physical problems that are subjects of ongoing debate. Finally, we developed a framework to train Foundation Neural Quantum States, which are versatile neural network models that approximate quantum wave functions of multiple systems simultaneously, enabling accurate estimates of challenging quantities such as disorder averages and fidelity susceptibility. We envision numerous future directions for this approach, including its extension to quantum dynamics by explicitly modeling time-dependent variational states, as well as its application to the design of novel materials in fermionic systems
Engineering the Kondo impurity problem with alkaline-earth-atom arrays
No full textWe propose quantum simulation experiments of the Kondo impurity problem using cold alkaline-earth(-like) atoms (AEAs) in a combination of optical lattice and optical tweezer potentials. Within an ab initio model for atomic interactions in the optical potentials, we analyze hallmark signatures of the Kondo effect in a variety of observables accessible in cold-atom quantum simulators. We identify additional terms not part of the textbook Kondo problem, mostly ignored in previous works and giving rise to a direct competition between spin and charge correlations, strongly suppressing Kondo physics. We show that the Kondo effect can be restored by locally adjusting the chemical potential on the impurity site, and we identify realistic parameter regimes and preparation protocols suited to current experiments with AEA arrays. Our work paves the way for quantum simulations of the Kondo problem and offers insights into Kondo physics in unconventional regimes
Nonreciprocal dynamics in soft active structures: From swimming filaments to odd periodic systems
No full textReciprocity is a fundamental property of many physical systems, generally expressing a symmetry relation between two processes in which, for instance, input and output are interchanged. This remarkable and fruitful property underpins powerful theoretical and experimental techniques; at the same time, its tight constraints represent an obstacle to desired behaviours or functionalities. For instance, in low Reynolds number hydrodynamics, reciprocity forbids net thrust or flow generation through time-reversible deformations; microorganisms have hence adapted their motion so as to circumvent this constraint. Likewise, reciprocity in linear elasticity impedes the creation of wave-bearing media where input and output are not interchangeable, prompting recent research efforts aimed at finding means to break loose of this constraint.
This thesis explores reciprocity breaking in soft active structures, focusing on two domains: biomimetic motility in soft robotics and wave propagation in periodic media. Active matter - whether biological or synthetic - negates reciprocity by converting stored or ambient energy into mechanical work. In living systems, this occurs in specialized motor proteins; in synthetic analogues, this is achieved through the use of smart materials, such as stimuli-responsive hydrogels. In these systems, internal activity introduces directed momentum input and disrupts the symmetries underlying classical reciprocity theorems.
Using polyelectrolyte hydrogel filaments as a model system for one dimensional active structures, we show that flutter instability can be harnessed as a new mechanism for undulatory locomotion and periodic beating of active filaments requiring minimal actuation. Far from the detrimental consequences often associated to its advent, here flutter instability emerges from the interaction of the active filament with the fluid environment, thus allowing for the reduction of control complexity; its exploitation can hence be regarded as a form of embodied mechanical intelligence. From another perspective, the responsiveness of polyelectrolyte hydrogels to environmental cues, breaks reciprocity at the constitutive level, leading to direction-dependant wave bearing properties. We explore this phenomenology in a periodic metabeam composed of active rods, demonstrating nonreciprocal transmission in both static and dynamic regimes. The material's ability to harvest energy from external stimuli compensates viscous dissipation, enabling near-lossless unidirectional wave propagation in a fluid environment.
This thesis represents a synthesis between mathematical modelling and physical experiments, in mutual and continuous dialogue. Theory and numerical simulations have informed the design and execution of the experiments, while laboratory observations have provided hints on how to refine the theoretical models and occasionally inspired new research directions