Search

Neural network potentials (NNPs) offer a fast and accurate alternative to ab-initio methods for molecular dynamics (MD) simulations but are hindered by the high cost of training data from high-fidelity Quantum Mechanics (QM) methods. Our work introduces the Implicit Delta Learning (IDLe) method, which reduces the need for high-fidelity QM data by leveraging cheaper semi-empirical QM computations without compromising NNP accuracy or inference cost. IDLe employs an end-to-end multi-task architecture with fidelity-specific heads that decode energies based on a shared latent representation of the input atomistic system. In various settings, IDLe achieves the same accuracy as single high-fidelity baselines while using up to 50x less high-fidelity data. This result could significantly reduce data generation cost and consequently enhance accuracy and generalization, and expand chemical coverage for NNPs, advancing MD simulations for material science and drug discovery. Additionally, we provide a novel set of 11 million semi-empirical QM calculations to support future multi-fidelity NNP modeling.

Machine Learning Interatomic Potentials (MLIPs) are a highly promising alternative to force-fields for molecular dynamics (MD) simulations, offering precise and rapid energy and force calculations. However, Quantum-Mechanical (QM) datasets, crucial for MLIPs, are fragmented across various repositories, hindering accessibility and model development. We introduce the openQDC package, consolidating 37 QM datasets from over 250 quantum methods and 400 million geometries into a single, accessible resource. These datasets are meticulously preprocessed, and standardized for MLIP training, covering a wide range of chemical elements and interactions relevant in organic chemistry. OpenQDC includes tools for normalization and integration, easily accessible via Python. Experiments with well-known architectures like SchNet, TorchMD-Net, and DimeNet reveal challenges for those architectures and constitute a leaderboard to accelerate benchmarking and guide novel algorithms development. Continuously adding datasets to OpenQDC will democratize QM dataset access, foster more collaboration and innovation, enhance MLIP development, and support their adoption in the MD field.

Recent studies have shown that non-equilibrium optical systems under static electric fields offer a pathway to realize chiral gain, where the non-Hermitian response of a material is controlled by the spin angular momentum of the wave. In this work, we uncover the topological nature of chiral gain and demonstrate how a static electric bias induces topological bandgaps that support unidirectional edge states at the material boundaries. Curiously, in our system, these topological edge states consistently exhibit dissipative properties. We further show that, by operating outside the topological gap, the chiral gain can be leveraged to engineer boundary-confined lasing modes with orbital angular momentum, locked to the orientation of the applied electric field. Our results open new possibilities for loss-compensated photonic waveguides, enabling advanced functionalities such as unidirectional, lossless edge-wave propagation and the generation of structured light with intrinsic orbital angular momentum.

Here, we investigate the effective response of three-dimensional spacetime crystals formed by spherical scatterers under a travelling-wave modulation. We develop an analytical formalism to homogenize the spacetime crystals that extends the renowned Clausius-Mossotti formula to time-varying platforms. Our formalism shows that travelling-wave spacetime crystals can be used to engineer a wide range of classes of nonreciprocal bianisotropic couplings in the long wavelength limit. In particular, our theory reveals the possibility of realizing a purely isotropic Tellegen response in crystals formed by interlaced sub-lattices of scatterers subjected to different modulation velocities. Furthermore, we introduce a class of generalized Minkowskian crystals that displays invariance under arbitrary Lorentz boosts aligned with a fixed spatial direction. We prove that such systems are formed by pseudo-uniaxial materials with the principal axis aligned parallel to the modulation velocity. The electromagnetic response of such generalized Minkowskian crystals is indistinguishable from that of moving photonic crystals.

We present a P-wave scattering image of the volcanic structures under Tenerife Island using the autocorrelation functions of P-wave vertical velocity fluctuations. We have applied cluster analysis to total quality factor attenuation (Q inv t) and scattering quality factor attenuation (Q inv PSc) images to interpret the structures in terms of intrinsic and scattering attenuation variations on a 2D plane, corresponding to a depth of 2000 m, and check the robustness of the scattering imaging. The results show that scattering patterns are similar to total attenuation patterns in the South of the island. There are two main areas where patterns differ: at Ca\~nadas-Teide-Pico Viejo Complex high total attenuation and average-to-low scattering values are observed. We interpret the difference as induced by intrinsic attenuation. In the Santiago Ridge Zone (SRZ) region, high scattering values correspond to average total attenuation. In our interpretation, the anomaly is induced by an extended scatterer, geometrically related to the surficial traces of Garachico and El Chinyero historical eruptions and the area of highest seismic activity during the 2004-2008 seismic crises., Comment: Final reviewed version of the published paper

In this work we have done a 3D joint interpretation of magnetotelluric and seismic tomography models. Previously we have described different techniques to infer the inner structure of the Earth. We have focused on volcanic regions, specifically on Tenerife Island volcano (Canary Islands, Spain). In this area, magnetotelluric and seismic tomography studies have been done separately. The novelty of the present work is the combination of both techniques in Tenerife Island. For this aim we have applied Fuzzy Clusters Method at different depths obtaining several clusters or classes. From the results, a geothermal system has been inferred below Teide volcano, in the center of Tenerife Island. An edifice hydrothermally altered and full of fluids is situated below Teide, ending at 600 m below sea level. From this depth the resistivity and VP values increase downwards. We also observe a clay cap structure, a typical feature in geothermal systems related with low resistivity and low VP values., Comment: Preprint of the published paper without figures

In the field of Machine Learning Interatomic Potentials (MLIPs), understanding the intricate relationship between data biases, specifically conformational and structural diversity, and model generalization is critical in improving the quality of Quantum Mechanics (QM) data generation efforts. We investigate these dynamics through two distinct experiments: a fixed budget one, where the dataset size remains constant, and a fixed molecular set one, which focuses on fixed structural diversity while varying conformational diversity. Our results reveal nuanced patterns in generalization metrics. Notably, for optimal structural and conformational generalization, a careful balance between structural and conformational diversity is required, but existing QM datasets do not meet that trade-off. Additionally, our results highlight the limitation of the MLIP models at generalizing beyond their training distribution, emphasizing the importance of defining applicability domain during model deployment. These findings provide valuable insights and guidelines for QM data generation efforts., Comment: Accepted at NeurIPS 2023 AI4D3 and AI4S workshops

Traditional molecular string representations, such as SMILES, often pose challenges for AI-driven molecular design due to their non-sequential depiction of molecular substructures. To address this issue, we introduce Sequential Attachment-based Fragment Embedding (SAFE), a novel line notation for chemical structures. SAFE reimagines SMILES strings as an unordered sequence of interconnected fragment blocks while maintaining compatibility with existing SMILES parsers. It streamlines complex generative tasks, including scaffold decoration, fragment linking, polymer generation, and scaffold hopping, while facilitating autoregressive generation for fragment-constrained design, thereby eliminating the need for intricate decoding or graph-based models. We demonstrate the effectiveness of SAFE by training an 87-million-parameter GPT2-like model on a dataset containing 1.1 billion SAFE representations. Through targeted experimentation, we show that our SAFE-GPT model exhibits versatile and robust optimization performance. SAFE opens up new avenues for the rapid exploration of chemical space under various constraints, promising breakthroughs in AI-driven molecular design., Comment: Code, data and models available at: https://github.com/datamol-io/safe/

Recently, pre-trained foundation models have enabled significant advancements in multiple fields. In molecular machine learning, however, where datasets are often hand-curated, and hence typically small, the lack of datasets with labeled features, and codebases to manage those datasets, has hindered the development of foundation models. In this work, we present seven novel datasets categorized by size into three distinct categories: ToyMix, LargeMix and UltraLarge. These datasets push the boundaries in both the scale and the diversity of supervised labels for molecular learning. They cover nearly 100 million molecules and over 3000 sparsely defined tasks, totaling more than 13 billion individual labels of both quantum and biological nature. In comparison, our datasets contain 300 times more data points than the widely used OGB-LSC PCQM4Mv2 dataset, and 13 times more than the quantum-only QM1B dataset. In addition, to support the development of foundational models based on our proposed datasets, we present the Graphium graph machine learning library which simplifies the process of building and training molecular machine learning models for multi-task and multi-level molecular datasets. Finally, we present a range of baseline results as a starting point of multi-task and multi-level training on these datasets. Empirically, we observe that performance on low-resource biological datasets show improvement by also training on large amounts of quantum data. This indicates that there may be potential in multi-task and multi-level training of a foundation model and fine-tuning it to resource-constrained downstream tasks.

Suppose that females choose males based on attributes that do not signal any genetic quality that is not related to the choice itself. Can being choosy confer selective advantage in this situation? We introduce correlated strategies, which means that females, when making their choice, may take into consideration external and independent random factors that are known to be observable by all. Individual-based simulation is used to show that, in this case, choosiness can emerge against the cost of over 25\% when pitted against randomly mating females. Moreover, after being established in the population, it can sustain costs of over 35\% . While such costs are not biologically plausible, they demonstrate unequivocally that sexual choice is a strong evolutionary force. Thus, correlated strategies are shown to be an evolutionary tool that channels randomness from the environment into genetic diversity. In addition, it turns out that a higher number of attributes in the ornament makes the choice more advantageous, which may result in a runaway complexity of sexual traits. Implications for the evolution of (female) cognitive abilities and speciation are discussed.

Item response theory aims to estimate respondent's latent skills from their responses in tests composed of items with different levels of difficulty. Several models of item response theory have been proposed for different types of tasks, such as binary or probabilistic responses, response time, multiple responses, among others. In this paper, we propose a new version of $\beta^3$-IRT, called $\beta^{4}$-IRT, which uses the gradient descent method to estimate the model parameters. In $\beta^3$-IRT, abilities and difficulties are bounded, thus we employ link functions in order to turn $\beta^{4}$-IRT into an unconstrained gradient descent process. The original $\beta^3$-IRT had a symmetry problem, meaning that, if an item was initialised with a discrimination value with the wrong sign, e.g. negative when the actual discrimination should be positive, the fitting process could be unable to recover the correct discrimination and difficulty values for the item. In order to tackle this limitation, we modelled the discrimination parameter as the product of two new parameters, one corresponding to the sign and the second associated to the magnitude. We also proposed sensible priors for all parameters. We performed experiments to compare $\beta^{4}$-IRT and $\beta^3$-IRT regarding parameter recovery and our new version outperformed the original $\beta^3$-IRT. Finally, we made $\beta^{4}$-IRT publicly available as a Python package, along with the implementation of $\beta^3$-IRT used in our experiments.

In recent years there has been a great interest in topological photonics and protected edge states. Here, we present a first principles method to compute topological invariants of three-dimensional gapless phases. Our approach allows to calculate the topological charges of Weyl points through the efficient numerical computation of gap Chern numbers, which relies solely on the photonic Green's function of the system. We particularize the framework to the Weyl points that are found to emerge in a magnetized plasma due to the breaking of time reversal symmetry. We discuss the relevance of modelling nonlocality when considering the topological properties of continuous media such as the magnetized plasma. We find that for some of the considered material models the charge of the Weyl point can be expressed in terms of a difference of the gap Chern numbers of two-dimensional material subcomponents. Our theory may be extended to other three-dimensional topological phases, or to Floquet systems., Comment: 13 pages, 11 figures

Here, we show that isorefractive spacetime crystals with a travelling-wave modulation may mimic rigorously the response of moving material systems. While generic spacetime crystals are characterized by a bi-anisotropic coupling in the co-moving frame, isorefractive crystals have a response that is observer independent, which leads to isotropic constitutive relations free of bianisotropy. We show how to take advantage of this property in the calculation of the band diagrams of isorefractive spacetime crystals in the laboratory frame and in the study of the synthetic Fresnel drag. Furthermore, we discuss the impact of considering either a Galilean or a Lorentz transformation in the homogenization of spacetime crystals, showing that the effective response is independent of the considered transformation.

Intelligent systems that use Machine Learning classification algorithms are increasingly common in everyday society. However, many systems use black-box models that do not have characteristics that allow for self-explanation of their predictions. This situation leads researchers in the field and society to the following question: How can I trust the prediction of a model I cannot understand? In this sense, XAI emerges as a field of AI that aims to create techniques capable of explaining the decisions of the classifier to the end-user. As a result, several techniques have emerged, such as Explanation-by-Example, which has a few initiatives consolidated by the community currently working with XAI. This research explores the Item Response Theory (IRT) as a tool to explaining the models and measuring the level of reliability of the Explanation-by-Example approach. To this end, four datasets with different levels of complexity were used, and the Random Forest model was used as a hypothesis test. From the test set, 83.8% of the errors are from instances in which the IRT points out the model as unreliable., Comment: 15 pages, 5 figures, 3 tables, submitted for the BRACIS'22 conference

Here, we show that spacetime modulations provide an exciting route to realize complex nonreciprocal couplings, and in particular the elusive axion response. We develop an analytical formalism to homogenize anisotropic spacetime crystals in the long wavelength limit. It is found that spacetime crystals with suitable glide-rotation symmetry can have a giant axion-type response, several orders of magnitude larger than in natural materials. The nonreciprocal axion response may have interesting applications in optics, for example in electromagnetic isolation, and in addition may enable exciting new forms of light-wave interactions.