Committee:
     PRESIDENT: V. GUREVICH, SVETLANA
     SECRETARI: ARTIGAS GARCIA, DAVID
     VOCAL: MAS ARABÍ, CARLOS
Thesis abstract: The aim of this thesis is to explore the effect of non-Hermitian potentials in different nonlinear systems. Specifically, in the different sections of this work we focus on the stabilisation of various states. We consider mainly two systems due to their universality: the Gizburg Landau and the Lugiato Lefever equations. Both models correspond to generalisations of the nonlinear Schrödinger equation. The Ginzburg-Landau equation represents the dissipative extension of physical systems with active media. It has been extensively studied and praised as a foundational model for phenomena such as turbulence, lasers or Bose Einstein condensates. In contrast, the Lugiato-Lefever equation addresses dissipative systems with lossy media, where energy is injected. Recently, this model has gained renewed attention due to its accurate representation of fields in micro-resonators and the rapid advancements in mastering their fabrication. Through analytical and numerical studies we investigate how introducing non-Hermitian potentials leads to the stabilisation of different coherent structures. For laser-like models that exhibit strong and persistent turbulence regimes, this approach results in an effective stabilisation and control of the system. In turn, for micro-resonators, the stabilisation of periodic structures uncovers a novel and hybrid formation mechanism for solitons.

Committee:
     PRESIDENT: FONT GARCÍA, BERNAT
     SECRETARI: VENTOSA MOLINA, JORDI
     VOCAL: ALCÁNTARA ÁVILA, FRANCISCO
Thesis abstract: This thesis presents a suite of methodologies for the dimensionality reduction of turbulent flow data, with a focus on high-fidelity simulations of external aerodynamics in industrial contexts, such as flow around simplified road vehicles. These simulations, typically performed on unstructured meshes with hundreds of millions of degrees of freedom, require scalable tools for analysis and modeling. All developments are implemented in pyLOM, an open-source Python library designed for terabyte-scale reduced-order modeling.The work progresses in four main steps, all of them published and detailed in their corresponding peer-reviewed article. First, classical reduction techniques based on the singular value decomposition (SVD), such as proper orthogonal decomposition (POD), dynamic mode decomposition (DMD), and spectral POD (SPOD), are adapted for high-performance computing by exploiting parallel QR factorization. Strong and weak scalability tests demonstrate high efficiency, with communication during QR factorization identified as the main bottleneck. Such performance gains together with the parallel input/output capabilities of pyLOM helped to apply these algorithms to datasets which occupy several terabytes of data as the direct numerical simulation of the flow in the Stanford diffuser in 80 seconds.Second, a convolutional neural network (CNN) variational autoencoder ($VAE) is developed for nonlinear dimensionality reduction, successfully capturing the temporal dynamics of the Windsor body back pressure with only two latent variables. The model effectively compresses snapshots of back pressure taken at yaw angles of 2.5º, 5º, and 10º. The projection of the mean pressure coefficient to the latent space yields an increasing linear evolution of the two latent variables with the yaw angle. The mean pressure coefficient distribution at yaw angle 7.5º is predicted with a mean error of 3.13% when compared to the results obtained by means of wall-modeled large eddy simulations (WMLES) after computing the values of the latent space with linear interpolation.Both methodologies have been merged to create a novel method named Geometry-Agnostic Variational-autoencoder Integration (GAVI), replacing the SVD step with a VAE operating on QR-factorized data. GAVI provides compact latent spaces without requiring structured grids, achieving high energy recovery across diverse test cases: circular cylinder, Windsor body, and urban flows, with latent spaces of 3–6 variables recovering over 90% of flow energy. The only step involving high performance computing (HPC) in GAVI is the computation of the QR factorization. This operation is done in parallel using pyLOM and its economic cost for all cases tested is lower than 10€. The fit of the VAE to the R matrix can be done with a GPU that fits in a workstation or laptop.Finally, a transformer-based closure strategy is proposed to compensate for energy loss in reduced models. By learning the spatial distribution of unresolved fluctuations, it improves reconstruction accuracy for complex unstructured domains, outperforming approaches based on super resolution generative adversarial networks (SRGAN) in both vehicle wakes and urban wind scenarios. The transformer model learns the probability density function of the missing fluctuations conditioned to the fluctuations already recovered by the model. Adding these fluctuations closes the energy gap between the original and the reconstructed data and improves the accuracy on both the instantaneous fields and the root mean square value of the fluctuations.

Committee:
     PRESIDENT: VENET ZAMBRANO, MICHEL
     SECRETARI: OCHOA GUERRERO, DIEGO ALEJANDRO
     VOCAL: RUBIO MARCOS, FERNANDO
Thesis abstract: In this work, barium titanate (BaTiO3) based lead-free ferroelectric ceramics were synthesized and investigated, viewing their applicability in energy-storage devices. The inclusion of the tin (Sn) ion in the perovskite structure of the BaTiO3 (BT) system has been then considered, in order to study the influence of the doping content on the physical properties of the BaTi1–xSnxO3 (BTS) system, where x = 0.03, 0.06, 0.09, 0.12, 0.15 and 0.20). The samples were sintered from the solid-state reaction method from high-purity chemical grades. The structural Properties, investigated from X-ray diffraction and Raman spectroscopy techniques, confirmed the formation of the perovskite structure for all the cases, revealing a phases’ coexistence depending on the analyzed composition. In fact, samples in the 0.03  x  0.09 compositional range evidenced the coexistence between ferroelectric phases with orthorhombic and tetragonal symmetries. On the other hand, the 0.12 and 0.15 compositions revealed the coexistence of the ferroelectric and paraelectric phases, with rhombohedral and cubic symmetries, respectively, whereas the samples with x = 0.20 has shown to be a single-phase composition with the cubic paraelectric characteristic. Therefore, results reveal a strong influence of the Sn4+ cation on the structural and vibrational characteristics of the BT system. On the other hand, the dielectric properties have shown a decrease in the ferroelectric-paraelectric (high temperature) phases’ transition temperature with the increase in the tin content, whereas the low temperature ferroelectric-ferroelectric phases’ transition temperatures increased as the Sn concentration increases. Such a variation could be associated to the influence of both change and coexistence of the involved phases, with the increase of the doping cation. The increase in the Sn concentration also promoted an enlargement in the maximum permittivity peak, indicating a diffuse-type phases’ transition promoted by cationic disorder at the B-site of the perovskite structure. Moreover, the ferroelectric properties allowed for the estimation of the system’s performance for energy storage and solid-state cooling, revealing an optimized storage value for x = 0.15, with ~45 mJ/cm² of recoverable energy and 77% efficiency, as well as an electrocaloric strength coefficient of 0.40 K mm/kV for the 0.09 composition, with these values being higher than those reported in the literature for other similar lead-free systems.

Committee:
     PRESIDENT: ROVIRA VIRGILI, CARME
     SECRETARI: CANALES GABRIEL, MANEL
     VOCAL: CALERO BORRALLO, CARLOS
Thesis abstract: In this Thesis we have performed all-atom molecular dynamics (MD) and well-tempered metadynamics (WTM) simulations of several benzothiadiazine (DBD)-membrane interaction systems, as well as of Rat Sarcoma (RAS) and RAS-Son of Sevenless (SOS1) solution systems. The first section is devoted to analyze the local structure of DBD and its derivatives embedded in bilayer membranes were analyzed. The most reactive sites of DBD are hydrogens ’H2’ and ’H4’ and oxygens, able to form hydrogen bonding (HB) interactions with the membrane components. The strongest HB interactions corresponded to the association between DBD’s ’H2’ and oxygens of DOPS in the absence of cholesterol. The cholesterol can weaken the DBD-lipid HB connections. Furthermore, the substituents of DBD have strong impact on the DBD-membrane interactions. The electron-donating substituents will enhance the HB interactions between DBD and phosphate groups of lipids, whereas electron-accepting groups will weaken the interactions with phosphate groups and enhance the interactions with ester groups of lipids. Later on, the impact of KRAS-G12 mutations was investigated. There are two new druggable dynamic water pockets located on KRAS-G12D-GDP-Mg2+ ternary complex. Subsequently, based on in silico drug design procedure the new potential KRAS-G12D inhibitor DBD15-21-22 was designed, which was proposed to locking KRAS-G12D in its GDP-bound "off" state. Furthermore, the free energy landscapes of the DBD15-21-22 and its derivatives in aqueous solution also been evaluated by WTM simulations. The proposed potential collective variables for the conformational dynamics investigation of small molecules can help to facilitate the in silico drug design by obtaining the corresponding 2D free energy landscapes. In the following section, the conformational characteristics of NRAS and their underlying atomic-level molecular interactions have been thoroughly analyzed by means of extensive MD simulations and trajectory analysis. The GTP-bound positively charged NRAS Q61 mutations (Q61R and Q61K) can be classified into one category, which have a small targetable pocket between α-helix 3 and switch-II. While the GTP-bound non-polar NRAS Q61L mutation fall into another class and the L61 can be captured by the hydrophobic pocket located on the α-helix 3. The new isomer-sourced structure iteration method was proposed to design the potential inhibitor (HM-387) for targeting the NRAS positively charged mutation. The HM-387 can tightly bind to GTP-bound NRAS-Q61R and gradually induce the protein conformation transition toward the "off-like" state. Moreover, the collective variables that can accurately characterize conformational changes in NRAS proteins were proposed, and then the accurate local structures of NRAS and its Q61 mutations in aqueous solution have been evaluated by WTM simulations. Finally, the SOS1-catalyzed GDP extraction process and the underlying atomic-level molecular interactions are described. The intrinsically disordered regions 807-818 (IDRs807-818) and αH domains of SOS1 cooperatively participate in the GDP extraction process: αH opens and inserts into the SW-I domain of wild-type KRAS, while the IDRs807-818 interact with the phosphate group of GDP. Among them, residues Lys811, Glu812, Lys939 and Glu942 play the main role. As for the KRAS-G12D mutation, the basic synergy of IDRs807-818 and αH is well-conserved during the GDP extraction steps and it is crucial for both wild-type and KRAS-G12D cases. However with the occurrence of G12D mutation, the molecular interactions of Asp12 with IDRs807-818, especially with residues Ser807, Trp809, Thr810 and Lys811 are enhanced, and increase the GDP extraction rate significantly.

Committee:
     PRESIDENT: CASTELLANO, CLAUDIO
     SECRETARI: SALES PARDO, MARTA
     VOCAL: GARCIA BECERRA, DAVID
Thesis abstract: Like viruses, information, knowledge, and emotions disseminate through a population, and the social interactions between individuals can give rise to a rich tapestry of collective behavior. Usually, we are able to ascertain the way in which individuals interact at a local level. Nevertheless, understanding how behavior emerges as a global feature is much more difficult and requires notions of synergy and complexity. This thesis investigates how social interactions shape collective behavior in social systems, focusing specifically on flocking dynamics in animal groups and opinion polarization in human societies. We employ theoretical frameworks derived from statistical physics and network science to analyze how local interaction rules lead to emergent phenomena, leveraging empirical data to validate our findings.In the context of animal behavior, we use social networks to represent the interactions within the flock. We extend the definition of Vicsek-like models to explore the effects that structural properties of social networks have on flocking stability. Networks with higher heterogeneity exhibit a more resilient ordered state, showing a diverging transition threshold, whereas lower heterogeneity renders the system more susceptible to external perturbations, showing a threshold converging to a finite value. However, if networks are weighted, analytical predictions extracted from the heterogeneous mean-field approximation reveal that the flocking threshold may also approach zero, i.e., a system that is always found in the disordered phase. Numerical simulations confirm these results. Finally, by considering real-world animal social networks, we show that networks with heterogeneous weights tend to exhibit a heightened sensitivity to noise, indicated by lower transition thresholds.Next, in the context of human behavior, we address the dynamics of opinion formation and polarization using a multidimensional perspective. To this aim, we use empirical opinions with respect to many different topics collected by the American National Election Studies. By mapping the opinions within a two-dimensional ideological space, we provide a nuanced analysis of how ideological polarization evolves over time in the United States. We assess the widening ideological gap between Democrats and Republicans over the past 30 years, highlighting an increase in within-party heterogeneity, particularly among Democrats. These findings contradict the partisan sorting hypothesis, which suggests that parties have become more consistent and ideologically more homogeneous in last decades.Lastly, we study the conditions under which polarization or consensus emerges by introducing the multidimensional social compass model. The model incorporates multiple opinion dimensions, each one corresponding to a distinct topic. A phase transition from polarization to consensus is exhibited at a critical threshold of social influence. However, correlations between multidimensional opinions play a pivotal role in the dynamics of depolarization. We demonstrate that if initial opinions are uncorrelated, the transition to consensus can be discontinuous depending on the number of dimensions, whereas if correlated, the transition is always continuous. Moreover, simulations of the model on top of social networks with a heterogeneous pattern of contacts indicate a vanishing threshold in the thermodynamic limit, in full agreement with the perturbation theory. We also explore the role of influential nodes in the depolarization process. We show that the presence of hubs can substantially lower the threshold value, effectively facilitating consensus. Conversely, when they hold divergent opinions, the social influence required to depolarize the system increases significantly.

Committee:
     PRESIDENT: CAZALILLA GUTIERREZ, MIGUEL
     SECRETARI: MASSIGNAN, PIETRO ALBERTO
     VOCAL: VRANJES MARKIC, LEANDRA
Thesis abstract: This Thesis explores the behavior of impurities embedded in Bose and Fermi systems, addressing both ground-state properties and thermal effects. We investigate these phenomena through advanced computational methods, such as Path Integral Monte Carlo (PIMC) and determinant diagrammatic Monte Carlo algorithms, providing insights into the interplay between impurity dynamics and many-body quantum effects.First, we examine the Bose polaron, i.e., an impurity in a Bose bath, focusing on its behavior near the critical temperature. Using ab-initio PIMC simulations, we study the temperature dependence of the polaron energy, effective mass, and dynamic structure factor for both repulsive and attractive branches. Our results show that quasiparticle characteristics are lost near the critical temperature, consistent with experimental findings. Additionally, we analyze the impact of impurities on the Bose-Einstein condensation and superfluidity of the Bose bath.Next, we address the thermal properties of repulsive impurities in a harmonically trapped Bose gas. At low temperatures, strong impurity-boson repulsion expels the impurity to the trap edges, but increasing temperature induces a miscibility crossover, with the impurity moving to the center of the trap. We identify a temperature-dependent miscibility transition and propose a nondestructive method for temperature measurement based on this phenomenon.We then study two-component repulsive Bose mixtures in a finite box. Our PIMC simulations reveal two distinct thermal behaviors: for phase-separated states at zero temperature, thermal diffusion reduces local population imbalance, while for miscible states, a temperature-induced maximum in local imbalance emerges due to particle bunching and anomalous cross pair distribution behavior.Finally, we analyze a two-dimensional Fermi gas on a lattice with a spin-down impurity interacting attractively with spin-up fermions. Using variational and diagrammatic Monte Carlo methods, we investigate the possibility of a polaron-to-dimeron transition over a range of filling fractions. Our results show that the polaron state remains energetically favorable, maintaining finite quasiparticle residue even at strong interactions.Together, these studies deepen our understanding of impurity physics in quantum systems, shedding light on thermal effects, miscibility, and quasiparticle properties in diverse scenarios.

Committee:
     PRESIDENT: CAMPINS MARTÍ, MAGDA
     SECRETARI: ECHEBARRIA DOMINGUEZ, BLAS
     VOCAL: DIEZ DOMINGO, JAVIER
Thesis abstract: This PhD thesis focused on the characterisation and understanding of the dynamics of SARS-CoV-2, influenza, and respiratory syncytial virus (RSV) in children in Catalonia over the last decade. Through a combination of empirical epidemiological analyses and dynamic modelling approaches, the study explored how the COVID-19 pandemic disrupted the seasonality and interaction patterns of respiratory viruses, altering their epidemiological behaviour and public health impact.We described epidemiological changes before, during and after the pandemic, highlighting the distinct post-pandemic behaviour of SARS-CoV-2 in children, that after the emergence of the Omicron variant and together with vaccination, has ceased to produce severe outcomes and predictable epidemic waves like influenza and RSV. Influenza and RSV, in turn, showed a progressive return to their pre-pandemic epidemic patterns, but they are not in a stationary state yet.The thesis introduced new epidemiological indicators, such as the epidemic potential growth (EPG), risk diagrams and risk panels, which were successfully applied for real-time epidemic surveillance. Additionally, empirical Gompertz models were employed to retrospectively analyse pre-pandemic epidemic dynamics of influenza and RSV, providing a quantitative framework for comparing the two viruses’ epidemic curves and identifying consistent patterns in their magnitude and timing and to verify post pandemic changes. These models served as well for predicting pre- and post-pandemic (when possible) epidemic peaks almost a month in advance with a maximum of a one-week error margin and magnitude falling within the 95% CI of the models, while human supervision and intervention enhanced their performance.To better understand the factors modulating influenza transmission, a SEIR compartmental model was developed, using a two-step transmissivity function that was ultimately related to meteorological factors. This model revealed a significant association between temperature and influenza transmissivity, following a two-week delayed effect, offering valuable insights into the environmental drivers of seasonal influenza epidemics.We also evaluated the real-world effectiveness of preventive strategies against RSV, particularly the monoclonal antibody nirsevimab, which demonstrated high effectiveness in reducing hospital admissions and paediatric intensive care unit (PICU) admissions by more than 80%. Furthermore, this work questioned long-standing assumptions regarding the role of children as primary drivers of respiratory virus epidemics. While confirming their value as a sentinel population for early epidemic detection given their frequent contact rates and active surveillance, the results suggest that children may not necessarily initiate epidemics, especially for viruses like influenza. This finding invites a reconsideration of public health strategies and control measures traditionally directed at the paediatric population.In conclusion, this thesis contributed to a deeper understanding of paediatric respiratory virus epidemiology in Catalonia, validated new modelling tools and surveillance indicators, and laid the groundwork for improving epidemic forecasting and prevention strategies in the post-pandemic context: placing children in the spotlight but safeguarding them from assuming responsibilities that belong to all of us.

Committee:
     PRESIDENT: ZAPPERI, STEFANO
     SECRETARI: ALONSO MUÑOZ, SERGIO
     VOCAL: REINA, ANDREAGIOVANNI
Thesis abstract: Moving animal groups, such as flocks of birds, schools of fish, insect swarms, herds of mammals, and human crowds, exhibit intricate and highly coordinated behaviors. This thesis explores the underlying mechanisms of collective behavior by analyzing extensive, high-resolution experimental data of trajectories of schooling fish in a controlled environment.We first introduce a force map technique to investigate the alignment forces experienced by individual fish. After validating the method with an agent-based model, we apply it to experimental data and find evidence of effective alignment with faster neighbors, alongside an unexpected anti-alignment with slower ones. Instead of interpreting this as explicit anti-alignment behavior, we suggest it arises from a selective attention mechanism, where fish are less attentive to slower neighbors. This implies the existence of temporal leadership dynamics based on relative speeds, a hypothesis we support through both agent-based modeling and analysis of leader-follower relationships in the experimental data.Next, we examine how variations in perceived risk influence the behavior and interactions of fish within a group. By merging two subgroups of fish, each with different habituation levels to the experimental tank, we observe that individuals with higher perceived risk tend to occupy more central positions and show increased group cohesion, aligning with the selfish herd hypothesis. Beyond spatial positioning, we note behavioral adjustments in coordination, leader-follower dynamics, and burst-and-coast movement patterns. To further validate these findings, we develop a machine learning tool capable of identifying the subgroup identity of individual fish based on their behavior.We also investigate spontaneous behavioral cascades in schooling fish, focusing on turning avalanches, where large directional shifts are propagated through the group. Using tools from condensed matter physics and seismology, we analyze various avalanche metrics and reveal scale-free behaviour through power-law distributions, data collapses and scaling relationships. This confirms a necessary condition for criticality in fish schools. We also explore the biological role of turning avalanches, linking them to collective decision-making processes that govern changes in the school’s direction. Boundary effects from interactions with the tank walls, as well as the influential roles of boundary individuals, are also identified. Furthermore, by applying the nearest-neighbor declustering method from seismology, which is based on identifying main events and aftershocks, we uncover spatiotemporal correlations in avalanche events and an Omori-Utsu law with a faster decay rate compared to typical earthquake dynamics.Finally, we explore and extend the nearest-neighbor declustering technique for analyzing aftershock correlations in point processes. Using earthquake catalogs, we show that this method can in some cases produce spurious Omori-Utsu and productivity relationships in the absence of true correlations. However, by assessing the robustness of these results under variations in newly introduced parameters, we demonstrate that the method can reliably distinguish genuine aftershocks.Overall, this thesis provides novel insights into the mechanisms driving collective behavior in animal groups, emphasizing the role of selective social interactions and critical dynamics. Our empirical, data-driven approach highlights the complexity of animal collectives and lays a foundation for future studies on collective decision-making, leadership, and information transfer in biological systems.

Committee:
     PRESIDENT: STERN TAULATS, ENRIC
     SECRETARI: MACOVEZ, ROBERTO
     VOCAL: BERMÚDEZ GARCÍA, JUAN MANUEL
Thesis abstract: Artificial refrigeration is one of the greatest technological achievements, essential in both industrial and domestic sectors for the preservation of food, medicines, and space cooling. It is expected that the demand for refrigeration will increase due to the growth of the population, global wealth, and climate change. However, current refrigeration technology has moderate efficiency and relies on fluids that harm the environment, presenting a serious scientific and technological challenge. Moreover, these systems are increasingly used in heat pumps replacing inefficient gas burners.Thermal effects induced by external fields in solid materials, such as barocaloric effects, emerge as a promising solution for more sustainable refrigeration. Despite recent advancements, challenges still persist. Plastic crystals have emerged as environmentally friendly barocaloric materials, very promising for applications in solid-state refrigeration and heating, thanks to their large isothermal entropy changes and temperature variations induced by pressure. Additionally, many of these plastic crystals, due to their high latent heat, could be used for thermal energy storage in waste heat management.This thesis investigates barocaloric effects in two types of compounds: (i) organic plastic crystals derived from adamantane and neopentane families, such as 1-adamantanol, 2-adamantanol, 2-methyl-2-adamantanol, 2-Br-adamantane, and 2-methyl-2-nitro-1-propanol; and (ii) ferroelectric plastic crystals, such as quinuclidine perrenate and tetramethylammonium tetrachloroferrate. Experimental techniques have been employed to characterize the barocaloric properties of these materials, such as conventional calorimetry, calorimetry, and dilatometry at high pressure, and dielectric spectroscopy under pressure and electric field, complemented by diffraction measurements carried out by other research group members or external collaborators.The objective of this thesis is to expand the range of solid compounds with applications in both industrial and domestic refrigeration and characterize their thermodynamic properties and barocaloric response. It also aims to understand the origin of latent heats and their relationship with the disorder that arises in each transition. Regarding suitable compounds for reversible refrigeration cycles, 2-Br-adamantane is found to be ideal for domestic refrigeration near ambient temperature. 1-adamantanol, 2-adamantanol, and 2-methyl-2-adamantanol are suitable for applications at temperatures higher than ambient, likely in the industrial sector, and have great potential for waste heat management due to their endotermic transitions. On the other hand, 2-methyl-2-nitro-1-propanol presents colossal irreversible barocaloric effects, but in its plastic phase, it can offer reversible effects, suggesting its use in reversible refrigeration cycles. This compound could also be useful in waste heat management at temperatures above ambient due to its high transition enthalpy.Ferroelectric plastic crystals, which combine a plastic phase at high temperature with a ferroelectric phase at low temperature, are a novelty. This work is the first to study their barocaloric effects, observing large effects, although not comparable to those of the studied organic plastic crystals. However, these materials open the door to possible multicaloric effects under pressure and electric field. However, large electric fields are required to observe electrocaloric effects in the studied cases.