Search

In this article, we highlight physics improvements in the NuWro Monte Carlo event generator. The upcoming version of NuWro will incorporate the integration of the argon spectral function for quasi-elastic scattering, along with the MINER$\nu$A parametrization of the axial form factor. Additionally, the new release will feature the implementation of the Valencia 2020 model for meson exchange current. The previously used simplistic delta resonance model for single-pion production will be replaced by a more accurate Ghent hybrid model in the upcoming version of NuWro. We also discuss the recent advancements made by the Wroclaw Neutrino Group in applying machine-learning techniques to achieve model-independent reconstruction of lepton-nucleus interactions, Comment: 4 pages, 3 figures. This is a conference proceeding. The text may overlap from [1]https://doi.org/10.48550/arXiv.2312.13369 [2] https://doi.org/10.48550/arXiv.2411.11523 [3]https://doi.org/10.48550/arXiv.2405.05212 [4] https://doi.org/10.48550/arXiv.2312.17298 [5] https://doi.org/10.48550/arXiv.2408.09936

In a transition between nonequilibrium steady states, the entropic cost associated with the maintenance of steady-state currents can be distinguished from that arising from the transition itself through the concepts of excess/housekeeping entropy flux and adiabatic/nonadiabatic entropy production. The thermodynamics of this transition is embodied by the Hatano-Sasa relation. In this letter, we show that for a slow transition between quantum nonequilibrium steady states the nonadiabatic entropy production is, to leading order, given by the path action with respect to a Riemannian metric in the parameter space which can be connected to the Kubo-Mori-Bogoliubov quantum Fisher information. We then demonstrate how to obtain minimally dissipative paths by solving the associated geodesic equation and illustrate the procedure with a simple example of a three-level maser. Furthermore, by identifying the quantum Fisher information with respect to time as a metric in state space, we derive an upper bound on the excess entropy flux that holds for arbitrarily fast processes., Comment: 11 pages, 2 figures

Control structure design is an important but tedious step in P&ID development. Generative artificial intelligence (AI) promises to reduce P&ID development time by supporting engineers. Previous research on generative AI in chemical process design mainly represented processes by sequences. However, graphs offer a promising alternative because of their permutation invariance. We propose the Graph-to-SFILES model, a generative AI method to predict control structures from flowsheet topologies. The Graph-to-SFILES model takes the flowsheet topology as a graph input and returns a control-extended flowsheet as a sequence in the SFILES 2.0 notation. We compare four different graph encoder architectures, one of them being a graph neural network (GNN) proposed in this work. The Graph-to-SFILES model achieves a top-5 accuracy of 73.2% when trained on 10,000 flowsheet topologies. In addition, the proposed GNN performs best among the encoder architectures. Compared to a purely sequence-based approach, the Graph-to-SFILES model improves the top-5 accuracy for a relatively small training dataset of 1,000 flowsheets from 0.9% to 28.4%. However, the sequence-based approach performs better on a large-scale dataset of 100,000 flowsheets. These results highlight the potential of graph-based AI models to accelerate P&ID development in small-data regimes but their effectiveness on industry relevant case studies still needs to be investigated.

We present the implementation and results of a new model for the n-particle n-hole ($\it{np-nh}$) contribution in the NuWro event generator, grounded in the theoretical framework established by the Valencia group in 2020. For the $\it{2p2h}$ component, we introduce a novel nucleon sampling function with tunable parameters to approximate correlations in the momenta of outgoing nucleons. These parameters are calibrated by comparing our results to those of the Valencia model across a range of incoming neutrino energies. In addition, our model incorporates a distinct contribution from the $\it{3p3h}$ mechanism. We discuss the differences between the new NuWro implementation, the original Valencia model, and the previous NuWro version, focusing on the distribution of outgoing nucleon momenta. Finally, we assess the impact of the hadronic model on experimental analyses involving hadronic observables., Comment: 17 pages, 11 figures

We have performed a global extraction of the ${\rm ^{12}C}$ longitudinal (${\cal R}_L$) and transverse (${\cal R}_T$) nuclear electromagnetic response functions from an analysis of all available electron scattering data on carbon. The response functions are extracted for energy transfer $\nu$, spanning the nuclear excitation, quasielastic (QE), resonance and inelastic continuum over a large range of the square of the four-momentum transfer ($Q^2$), for fixed values of $Q^2$ and for fixed values of 3-momentum transfer $\bf q$. The data sample consists of approximately 10,000 differential electron scattering and photo-absorption-cross section measurement points for ${\rm ^{12}C}$. In addition, we perform a universal fit to all ${\rm ^{12}C}$ electron scattering data which also provides parmeterizations of ${\cal R}_L$ and ${\cal R}_T$ over a larger kinematic range. Since the extracted response functions and the universal fit cover a large range of $Q^2$ and $\nu$, they can be readily used for comparison to theoretical predictions as well as validating and tuning Monte Carlo generators for electron and neutrino scattering experiments. In this paper we focus on the nuclear excitation, QE, and $\Delta$(1232) regions and compare the measurements to predictions of the following theoretical approaches: ``Energy Dependent-Relativistic Mean Field'' (ED-RMF), ``Green's Function Monte Carlo'' (GFMC), "Short Time Approximation Quantum Monte Carlo" (STA-QMC), "Correlated Fermi Gas" (CFG), as well as the {\textsc{NuWro}}, \ {{\sc{achilles}}}~ and {{\sc{genie}}}~generators. We find that among all the models ED-RMF provides the best description of both the QE and {\it nuclear excitations} response functions over the largest kinematic range $0.01\le Q^2 \le 1.25$ GeV$^2$. The ED-RMF formalism has the added benefit that it should be directly applicable to the same kinematic regions for neutrino scattering., Comment: 34 pages, 23 figures, submitted to Phys. Rev. D

Transfer learning (TL) allows a deep neural network (DNN) trained on one type of data to be adapted for new problems with limited information. We propose to use the TL technique in physics. The DNN learns the physics of one process, and after fine-tuning, it makes predictions for related processes. We consider the DNNs, trained on inclusive electron-carbon scattering data, and show that after fine-tuning, they accurately predict cross sections for electron interactions with nuclear targets ranging from lithium to iron. The method works even when the DNN is fine-tuned on a small dataset., Comment: 4 pages, 2 figures

An environment interacting with a quantum system can enhance transport through the suppression of quantum effects responsible for localization. In this paper, we study the interplay between bulk dephasing and a linear potential in a boundary-driven tight-binding chain. A linear potential induces Wannier-Stark localization in the absence of noise, while dephasing induces diffusive transport in the absence of a tilt. We derive an approximate expression for the steady-state current as a function of both dephasing and tilt which closely matches the exact solution for a wide range of parameters. From it, we find that the maximum current occurs for a dephasing rate equal to the period of Bloch oscillations in the Wannier-Stark localized system. We also find that the current displays a maximum as a function of the system size, provided that the total potential tilt across the chain remains constant. Our results can be verified in current experimental platforms and represents a step forward in analytical studies of environment-assisted transport., Comment: 20 pages, 7 figures

The determination of the nuclear spectral function from the measured cross section of the electron-nucleus scattering process $e + A \to e^\prime + p + (A-1)$is discussed, and illustrated for the case of a carbon target. The theoretical model based on the local density approximation, previously employed to derive the spectral function from a combination of accurate theoretical calculations and experimental data, has been developed further by including additional information obtained from measurements performed with high missing energy resolution. The implications for the analysis of $\gamma$-ray emission associated with nuclear deexcitation are considered.

We study the nonequilibrium steady state thermodynamics of a photodevice which can operate as a solar cell or a photoconductor, depending on the degree of asymmetry of the junction. The thermodynamic efficiency is captured by a single coefficient of performance. Using a minimal model, we show that when the electron repulsion energy matches the transport gap of the junction, the photoconductor displays maximal response, performance and signal-to-noise ratio, while the same regime is always detrimental for the solar cell. Nevertheless, we find that electron repulsion is beneficial for the solar cell if it lies below the transport gap. Our work sheds important light on design principles for thermodynamically efficient photodevices in the presence of interactions., Comment: 8 pages, 4 figures

Deep learning (DL)-based methods have achieved state-of-the-art performance for many medical image segmentation tasks. Nevertheless, recent studies show that deep neural networks (DNNs) can be miscalibrated and overconfident, leading to "silent failures" that are risky for clinical applications. Bayesian DL provides an intuitive approach to DL failure detection, based on posterior probability estimation. However, the posterior is intractable for large medical image segmentation DNNs. To tackle this challenge, we propose a Bayesian learning framework using Hamiltonian Monte Carlo (HMC), tempered by cold posterior (CP) to accommodate medical data augmentation, named HMC-CP. For HMC computation, we further propose a cyclical annealing strategy, capturing both local and global geometries of the posterior distribution, enabling highly efficient Bayesian DNN training with the same computational budget as training a single DNN. The resulting Bayesian DNN outputs an ensemble segmentation along with the segmentation uncertainty. We evaluate the proposed HMC-CP extensively on cardiac magnetic resonance image (MRI) segmentation, using in-domain steady-state free precession (SSFP) cine images as well as out-of-domain datasets of quantitative T1 and T2 mapping. Our results show that the proposed method improves both segmentation accuracy and uncertainty estimation for in- and out-of-domain data, compared with well-established baseline methods such as Monte Carlo Dropout and Deep Ensembles. Additionally, we establish a conceptual link between HMC and the commonly known stochastic gradient descent (SGD) and provide general insight into the uncertainty of DL. This uncertainty is implicitly encoded in the training dynamics but often overlooked. With reliable uncertainty estimation, our method provides a promising direction toward trustworthy DL in clinical applications., Comment: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2024:011

A new physics-based compact model, which makes it possible to simulate in a unified way the neutron and proton of cosmic ray induced SEU cross sections, including effects from nuclear reaction products and from direct ionization by low-energy protons, has been proposed and vali-dated. The proposed approach is analytical and based on explicit analytical relationships and approximations with physics-based fitting parameters. GEANT4 or SRIM numerical calculations can be used as an aid to adjust or refine the phenomenological parameters or functions included in the model taking into account real geometrical configurations and chemical compositions of the devices. In particular, explicit energy dependencies of the soft error cross sections for protons and neutrons over a wide range of nucleon energies were obtained and validated. Main application areas of developed model include space physics, accelerator studies high energy physics and nuclear experiments., Comment: 12 pages, 12 figures, 33 references

Machine Learning (ML) techniques are encountered nowadays across disciplines, from social sciences, through natural sciences to engineering. The broad application of ML and the accelerated pace of its evolution lead to an increasing need for dedicated teaching concepts aimed at making the application of this technology more reliable and responsible. However, teaching ML is a daunting task. Aside from the methodological complexity of ML algorithms, both with respect to theory and implementation, the interdisciplinary and empirical nature of the field need to be taken into consideration. This paper introduces the MachineLearnAthon format, an innovative didactic concept designed to be inclusive for students of different disciplines with heterogeneous levels of mathematics, programming and domain expertise. At the heart of the concept lie ML challenges, which make use of industrial data sets to solve real-world problems. These cover the entire ML pipeline, promoting data literacy and practical skills, from data preparation, through deployment, to evaluation.

Employing the neural network framework, we obtain empirical fits to the electron-scattering cross sections for carbon over a broad kinematic region, extending from the quasielastic peak through resonance excitation to the onset of deep-inelastic scattering. We consider two different methods of obtaining such model-independent parametrizations and the corresponding uncertainties: based on the bootstrap approach and the Monte Carlo dropout approach. In our analysis, the $\chi^2$ defines the loss function, including point-to-point and normalization uncertainties for each independent set of measurements. Our statistical approaches lead to fits of comparable quality and similar uncertainties of the order of $7$%. To test these models, we compare their predictions to test datasets excluded from the training process and theoretical predictions obtained within the spectral function approach. The predictions of both models agree with experimental measurements and theoretical calculations. We also perform a comparison to a dataset lying beyond the covered kinematic region, and find that the bootstrap approach shows better interpolation and extrapolation abilities than the one based on the dropout algorithm., Comment: 12 pages, 9 figures; fits and full list of plots are available from repository: https://github.com/bekowal/CarbonElectronNeuralNetwork

The Short-Baseline Neutrino program in Fermilab aims to resolve the nature of the low-energy excess events observed in LSND and MiniBooNE, and analyze with unprecedented precision neutrino interactions with argon. These studies require reliable estimate of neutrino cross sections, in particular for charged current quasielastic scattering (CCQE). Here, we report updates of the NuWro Monte Carlo generator that, most notably, bring the state-of-the-art spectral functions to model the ground state properties of the argon nucleus, and improve the accuracy of the cross sections at low energies by accounting for the effects of the nuclear Coulomb potential. We discuss these developments in the context of electron and neutrino interactions, by comparing updated NuWro predictions to experimental data from Jefferson Laboratory Hall A and MicroBooNE. The MicroBooNE CCQE data are described with the $\chi^2$ per degree of freedom of 0.7, compared with 1.0 in the local Fermi gas model. The largest improvement is observed for the angular distributions of the produced protons, where the $\chi^2$ reduces nearly by half. Being obtained using the axial form factor parametrization from MINERvA, our results indicate a~consistency between the CCQE measurements in MINERvA and MicroBooNE., Comment: 9 pages, 4 figures, matches the published version

Understanding the entropy production of systems strongly coupled to thermal baths is a core problem of both quantum thermodynamics and mesoscopic physics. While there exist many techniques to accurately study entropy production in such systems, they typically require a microscopic description of the baths, which can become numerically intractable to study for large systems. Alternatively an open-systems approach can be employed with all the nuances associated with various levels of approximation. Recently, the mesoscopic leads approach has emerged as a powerful method for studying such quantum systems strongly coupled to multiple thermal baths. In this method, a set of discretised lead modes, each locally damped, provide a Markovian embedding. Here we show that this method proves extremely useful to describe entropy production of a strongly coupled open quantum system. We show numerically, for both non-interacting and interacting setups, that a system coupled to a single bath exhibits a thermal fixed point at the level of the embedding. This allows us to use various results from the thermodynamics of quantum dynamical semi-groups to infer the non-equilibrium thermodynamics of the strongly coupled, non-Markovian central systems. In particular, we show that the entropy production in the transient regime recovers the well established microscopic definitions of entropy production with a correction that can be computed explicitly for both the single- and multiple-lead cases., Comment: 11 pages, 6 figures. Final author version

The process engineering domain widely uses Process Flow Diagrams (PFDs) and Process and Instrumentation Diagrams (P&IDs) to represent process flows and equipment configurations. However, the P&IDs and PFDs, hereafter called flowsheets, can contain errors causing safety hazards, inefficient operation, and unnecessary expenses. Correcting and verifying flowsheets is a tedious, manual process. We propose a novel generative AI methodology for automatically identifying errors in flowsheets and suggesting corrections to the user, i.e., autocorrecting flowsheets. Inspired by the breakthrough of Large Language Models (LLMs) for grammatical autocorrection of human language, we investigate LLMs for the autocorrection of flowsheets. The input to the model is a potentially erroneous flowsheet and the output of the model are suggestions for a corrected flowsheet. We train our autocorrection model on a synthetic dataset in a supervised manner. The model achieves a top-1 accuracy of 80% and a top-5 accuracy of 84% on an independent test dataset of synthetically generated flowsheets. The results suggest that the model can learn to autocorrect the synthetic flowsheets. We envision that flowsheet autocorrection will become a useful tool for chemical engineers.

ReLU neural networks have been modelled as constraints in mixed integer linear programming (MILP), enabling surrogate-based optimisation in various domains and efficient solution of machine learning certification problems. However, previous works are mostly limited to MLPs. Graph neural networks (GNNs) can learn from non-euclidean data structures such as molecular structures efficiently and are thus highly relevant to computer-aided molecular design (CAMD). We propose a bilinear formulation for ReLU Graph Convolutional Neural Networks and a MILP formulation for ReLU GraphSAGE models. These formulations enable solving optimisation problems with trained GNNs embedded to global optimality. We apply our optimization approach to an illustrative CAMD case study where the formulations of the trained GNNs are used to design molecules with optimal boiling points.

ChatGPT is a powerful language model from OpenAI that is arguably able to comprehend and generate text. ChatGPT is expected to have a large impact on society, research, and education. An essential step to understand ChatGPT's expected impact is to study its domain-specific answering capabilities. Here, we perform a systematic empirical assessment of its abilities to answer questions across the natural science and engineering domains. We collected 594 questions from 198 faculty members across 5 faculties at Delft University of Technology. After collecting the answers from ChatGPT, the participants assessed the quality of the answers using a systematic scheme. Our results show that the answers from ChatGPT are on average perceived as ``mostly correct''. Two major trends are that the rating of the ChatGPT answers significantly decreases (i) as the complexity level of the question increases and (ii) as we evaluate skills beyond scientific knowledge, e.g., critical attitude.

Microgels are cross-linked, colloidal polymer networks with great potential for stimuli-response release in drug-delivery applications, as their size in the nanometer range allows them to pass human cell boundaries. For applications with specified requirements regarding size, producing tailored microgels in a continuous flow reactor is advantageous because the microgel properties can be controlled tightly. However, no fully-specified mechanistic models are available for continuous microgel synthesis, as the physical properties of the included components are only studied partly. To address this gap and accelerate tailor-made microgel development, we propose a data-driven optimization in a hardware-in-the-loop approach to efficiently synthesize microgels with defined sizes. We optimize the synthesis regarding conflicting objectives (maximum production efficiency, minimum energy consumption, and the desired microgel radius) by applying Bayesian optimization via the solver ``Thompson sampling efficient multi-objective optimization'' (TS-EMO). We validate the optimization using the deterministic global solver ``McCormick-based Algorithm for mixed-integer Nonlinear Global Optimization'' (MAiNGO) and verify three computed Pareto optimal solutions via experiments. The proposed framework can be applied to other desired microgel properties and reactor setups and has the potential of efficient development by minimizing number of experiments and modelling effort needed., Comment: Manuscript: 24 pages, 8 figures; SI: 9 pages, 3 figures

The transformation towards renewable energy and feedstock supply in the chemical industry requires new conceptual process design approaches. Recently, breakthroughs in artificial intelligence offer opportunities to accelerate this transition. Specifically, deep reinforcement learning, a subclass of machine learning, has shown the potential to solve complex decision-making problems and aid sustainable process design. We survey state-of-the-art research in reinforcement learning for process design through three major elements: (i) information representation, (ii) agent architecture, and (iii) environment and reward. Moreover, we discuss perspectives on underlying challenges and promising future works to unfold the full potential of reinforcement learning for process design in chemical engineering.

Artificial intelligence has great potential for accelerating the design and engineering of chemical processes. Recently, we have shown that transformer-based language models can learn to auto-complete chemical process flowsheets using the SFILES 2.0 string notation. Also, we showed that language translation models can be used to translate Process Flow Diagrams (PFDs) into Process and Instrumentation Diagrams (P&IDs). However, artificial intelligence methods require big data and flowsheet data is currently limited. To mitigate this challenge of limited data, we propose a new data augmentation methodology for flowsheet data that is represented in the SFILES 2.0 notation. We show that the proposed data augmentation improves the performance of artificial intelligence-based process design models. In our case study flowsheet data augmentation improved the prediction uncertainty of the flowsheet autocompletion model by 14.7%. In the future, our flowsheet data augmentation can be used for other machine learning algorithms on chemical process flowsheets that are based on SFILES notation., Comment: Submitted to PROCEEDINGS OF THE 33rd European Symposium on Computer Aided Process Engineering (ESCAPE33), June 18-21, 2023, Athens, Greece

Process design is a creative task that is currently performed manually by engineers. Artificial intelligence provides new potential to facilitate process design. Specifically, reinforcement learning (RL) has shown some success in automating process design by integrating data-driven models that learn to build process flowsheets with process simulation in an iterative design process. However, one major challenge in the learning process is that the RL agent demands numerous process simulations in rigorous process simulators, thereby requiring long simulation times and expensive computational power. Therefore, typically short-cut simulation methods are employed to accelerate the learning process. Short-cut methods can, however, lead to inaccurate results. We thus propose to utilize transfer learning for process design with RL in combination with rigorous simulation methods. Transfer learning is an established approach from machine learning that stores knowledge gained while solving one problem and reuses this information on a different target domain. We integrate transfer learning in our RL framework for process design and apply it to an illustrative case study comprising equilibrium reactions, azeotropic separation, and recycles, our method can design economically feasible flowsheets with stable interaction with DWSIM. Our results show that transfer learning enables RL to economically design feasible flowsheets with DWSIM, resulting in a flowsheet with an 8% higher revenue. And the learning time can be reduced by a factor of 2., Comment: Submitted to PROCEEDINGS OF THE 33rd European Symposium on Computer Aided Process Engineering (ESCAPE33), June 18-21, 2023, Athens, Greece

Developing Piping and Instrumentation Diagrams (P&IDs) is a crucial step during the development of chemical processes. Currently, this is a tedious, manual, and time-consuming task. We propose a novel, completely data-driven method for the prediction of control structures. Our methodology is inspired by end-to-end transformer-based human language translation models. We cast the control structure prediction as a translation task where Process Flow Diagrams (PFDs) are translated to P&IDs. To use established transformer-based language translation models, we represent the P&IDs and PFDs as strings using our recently proposed SFILES 2.0 notation. Model training is performed in a transfer learning approach. Firstly, we pre-train our model using generated P&IDs to learn the grammatical structure of the process diagrams. Thereafter, the model is fine-tuned leveraging transfer learning on real P&IDs. The model achieved a top-5 accuracy of 74.8% on 10,000 generated P&IDs and 89.2% on 100,000 generated P&IDs. These promising results show great potential for AI-assisted process engineering. The tests on a dataset of 312 real P&IDs indicate the need of a larger P&IDs dataset for industry applications.

Recombination losses contribute to reduce $J_{SC}$, $V_{OC}$ and the fill factor of organic solar cells. Recent advances in non-fullerene organic photovoltaics have shown, nonetheless, that efficient charge generation can occur under small energetic driving forces ($\Delta E_{DA}$) and low recombination losses. To shed light on this issue, we set up a coarse-grained open quantum mechanical model for investigating the charge generation dynamics subject to various energy loss mechanisms. The influence of energetic driving force, Coulomb interaction, vibrational disorder, geminate recombination, temperature and external bias are included in the analysis of the optimal photovoltaic conditions for charge carrier generation. The assessment reveals that the overall energy losses are not only minimized when $\Delta E_{DA}$ approaches the effective reorganization energy at the interface but also become insensitive to temperature and electric field variations. It is also observed that a moderate reverse bias reduces geminate recombination losses significantly at vanishing driving forces, where the charge generation is strongly affected by temperature.

We propose a novel method enabling autocompletion of chemical flowsheets. This idea is inspired by the autocompletion of text. We represent flowsheets as strings using the text-based SFILES 2.0 notation and learn the grammatical structure of the SFILES 2.0 language and common patterns in flowsheets using a transformer-based language model. We pre-train our model on synthetically generated flowsheets to learn the flowsheet language grammar. Then, we fine-tune our model in a transfer learning step on real flowsheet topologies. Finally, we use the trained model for causal language modeling to autocomplete flowsheets. Eventually, the proposed method can provide chemical engineers with recommendations during interactive flowsheet synthesis. The results demonstrate a high potential of this approach for future AI-assisted process synthesis.

Graph neural networks (GNNs) are emerging in chemical engineering for the end-to-end learning of physicochemical properties based on molecular graphs. A key element of GNNs is the pooling function which combines atom feature vectors into molecular fingerprints. Most previous works use a standard pooling function to predict a variety of properties. However, unsuitable pooling functions can lead to unphysical GNNs that poorly generalize. We compare and select meaningful GNN pooling methods based on physical knowledge about the learned properties. The impact of physical pooling functions is demonstrated with molecular properties calculated from quantum mechanical computations. We also compare our results to the recent set2set pooling approach. We recommend using sum pooling for the prediction of properties that depend on molecular size and compare pooling functions for properties that are molecular size-independent. Overall, we show that the use of physical pooling functions significantly enhances generalization.

Trained ML models are commonly embedded in optimization problems. In many cases, this leads to large-scale NLPs that are difficult to solve to global optimality. While ML models frequently lead to large problems, they also exhibit homogeneous structures and repeating patterns (e.g., layers in ANNs). Thus, specialized solution strategies can be used for large problem classes. Recently, there have been some promising works proposing specialized reformulations using mixed-integer programming or reduced space formulations. However, further work is needed to develop more efficient solution approaches and keep up with the rapid development of new ML model architectures.

Molecular property prediction is of crucial importance in many disciplines such as drug discovery, molecular biology, or material and process design. The frequently employed quantitative structure-property/activity relationships (QSPRs/QSARs) characterize molecules by descriptors which are then mapped to the properties of interest via a linear or nonlinear model. In contrast, graph neural networks, a novel machine learning method, directly work on the molecular graph, i.e., a graph representation where atoms correspond to nodes and bonds correspond to edges. GNNs allow to learn properties in an end-to-end fashion, thereby avoiding the need for informative descriptors as in QSPRs/QSARs. GNNs have been shown to achieve state-of-the-art prediction performance on various property predictions tasks and represent an active field of research. We describe the fundamentals of GNNs and demonstrate the application of GNNs via two examples for molecular property prediction.

SFILES is a text-based notation for chemical process flowsheets. It was originally proposed by d'Anterroches (2006) who was inspired by the text-based SMILES notation for molecules. The text-based format has several advantages compared to flowsheet images regarding the storage format, computational accessibility, and eventually for data analysis and processing. However, the original SFILES version cannot describe essential flowsheet configurations unambiguously, such as the distinction between top and bottom products. Neither is it capable of describing the control structure required for the safe and reliable operation of chemical processes. Also, there is no publicly available software for decoding or encoding chemical process topologies to SFILES. We propose the SFILES 2.0 with a complete description of the extended notation and naming conventions. Additionally, we provide open-source software for the automated conversion between flowsheet graphs and SFILES 2.0 strings. This way, we hope to encourage researchers and engineers to publish their flowsheet topologies as SFILES 2.0 strings. The ultimate goal is to set the standards for creating a FAIR database of chemical process flowsheets, which would be of great value for future data analysis and processing.

Process synthesis experiences a disruptive transformation accelerated by digitization and artificial intelligence. We propose a reinforcement learning algorithm for chemical process design based on a state-of-the-art actor-critic logic. Our proposed algorithm represents chemical processes as graphs and uses graph convolutional neural networks to learn from process graphs. In particular, the graph neural networks are implemented within the agent architecture to process the states and make decisions. Moreover, we implement a hierarchical and hybrid decision-making process to generate flowsheets, where unit operations are placed iteratively as discrete decisions and corresponding design variables are selected as continuous decisions. We demonstrate the potential of our method to design economically viable flowsheets in an illustrative case study comprising equilibrium reactions, azeotropic separation, and recycles. The results show quick learning in discrete, continuous, and hybrid action spaces. Due to the flexible architecture of the proposed reinforcement learning agent, the method is predestined to include large action-state spaces and an interface to process simulators in future research.

Ionic liquids (ILs) are important solvents for sustainable processes and predicting activity coefficients (ACs) of solutes in ILs is needed. Recently, matrix completion methods (MCMs), transformers, and graph neural networks (GNNs) have shown high accuracy in predicting ACs of binary mixtures, superior to well-established models, e.g., COSMO-RS and UNIFAC. GNNs are particularly promising here as they learn a molecular graph-to-property relationship without pretraining, typically required for transformers, and are, unlike MCMs, applicable to molecules not included in training. For ILs, however, GNN applications are currently missing. Herein, we present a GNN to predict temperature-dependent infinite dilution ACs of solutes in ILs. We train the GNN on a database including more than 40,000 AC values and compare it to a state-of-the-art MCM. The GNN and MCM achieve similar high prediction performance, with the GNN additionally enabling high-quality predictions for ACs of solutions that contain ILs and solutes not considered during training., Comment: 16 pages, 4 figures, 5 tables

Understanding the thermodynamics of driven quantum systems strongly coupled to thermal baths is a central focus of quantum thermodynamics and mesoscopic physics. A variety of different methodological approaches exist in the literature, all with their own advantages and disadvantages. The mesoscopic leads approach was recently generalised to steady state thermal machines and has the ability to replicate Landauer B\"uttiker theory in the non-interacting limit. In this approach a set of discretised lead modes, each locally damped, provide a markovian embedding for the baths. In this work we further generalise this approach to incorporate an arbitrary time dependence in the system Hamiltonian. Following a careful discussion of the calculation of thermodynamic quantities we illustrate the power of our approach by studying several driven mesoscopic examples coupled to finite temperature fermionic baths, replicating known results in various limits. In the case of a driven non interacting quantum dot we show how fast driving can be used to induce heat rectification., Comment: 15 pages, 8 figures. Final author version

Fuels with high-knock resistance enable modern spark-ignition engines to achieve high efficiency and thus low CO2 emissions. Identification of molecules with desired autoignition properties indicated by a high research octane number and a high octane sensitivity is therefore of great practical relevance and can be supported by computer-aided molecular design (CAMD). Recent developments in the field of graph machine learning (graph-ML) provide novel, promising tools for CAMD. We propose a modular graph-ML CAMD framework that integrates generative graph-ML models with graph neural networks and optimization, enabling the design of molecules with desired ignition properties in a continuous molecular space. In particular, we explore the potential of Bayesian optimization and genetic algorithms in combination with generative graph-ML models. The graph-ML CAMD framework successfully identifies well-established high-octane components. It also suggests new candidates, one of which we experimentally investigate and use to illustrate the need for further auto-ignition training data., Comment: manuscript (26 pages, 9 figures, 2 tables), supporting information (12 pages, 8 figures, 1 table)

Innovative membrane technologies optimally integrated into large separation process plants are essential for economical water treatment and disposal. However, the mass transport through membranes is commonly described by nonlinear differential-algebraic mechanistic models at the nano-scale, while the process and its economics range up to large-scale. Thus, the optimal design of membranes in process plants requires decision making across multiple scales, which is not tractable using standard tools. In this work, we embed artificial neural networks~(ANNs) as surrogate models in the deterministic global optimization to bridge the gap of scales. This methodology allows for deterministic global optimization of membrane processes with accurate transport models -- avoiding the utilization of inaccurate approximations through heuristics or short-cut models. The ANNs are trained based on data generated by a one-dimensional extended Nernst-Planck ion transport model and extended to a more accurate two-dimensional distribution of the membrane module, that captures the filtration-related decreasing retention of salt. We simultaneously design the membrane and plant layout yielding optimal membrane module synthesis properties along with the optimal plant design for multiple objectives, feed concentrations, filtration stages, and salt mixtures. The developed process models and the optimization solver are available open-source, enabling computational resource-efficient multi-scale optimization in membrane science.

Non-fullerene acceptor (NFA) materials have posed new paradigms for the design of organic solar cells (OSC), whereby efficient carrier generation is obtained with small driving forces, in order to maximize the open-circuit voltage. In this paper we use a coarse-grained mixed quantum-classical method, that combines Ehrenfest and Redfield theories, to shed light on charge generation process in small energy offset interfaces. We have investigated the influence of the energetic driving force as well as the vibronic effects on the charge generation and photovoltaic energy conversion. By analyzing the effects of the Holstein and Peierls vibrational couplings, we find that vibrational couplings produce an overall effect of improving the charge generation. However, the two vibronic mechanisms play different roles: the Holstein relaxation mechanism decreases the charge generation whereas the Peierls mechanism always assists the charge generation. Moreover, by examining the electron-hole binding energy as a function of time, we evince two distinct regimes for the charge separation: the temperature independent excitonic spread on a sub-100 fs timescale and the complete dissociation of the charge-transfer state that occurs on the timescale of tens to hundreds of picoseconds, depending on the temperature. The quantum dynamics of the system exhibits the three regimes of the Marcus electron transfer kinetics as the energy offset of the interface is varied., Comment: Discussion about role of electric-field in charge dissociation included; general manuscript revision; to appear in JCP

We study dephasing-enhanced transport in boundary-driven quasi-periodic systems. Specifically we consider dephasing modelled by current preserving Lindblad dissipators acting on the non-interacting Aubry-Andr\'e-Harper (AAH) and Fibonacci bulk systems. The former is known to undergo a critical localization transition with a suppression of ballistic transport above a critical value of the potential. At the critical point, the presence of non-ergodic extended states yields anomalous sub-diffusion. The Fibonacci model, on the other hand, yields anomalous transport with a continuously varying exponent depending on the potential strength. By computing the covariance matrix in the non-equilibrium steady-state, we show that sufficiently strong dephasing always renders the transport diffusive. The interplay between dephasing and quasi-periodicity gives rise to a maximum of the diffusion coefficient for finite dephasing, which suggests the combination of quasi-periodic geometries and dephasing can be used to control noise-enhanced transport.

Data-driven models are becoming increasingly popular in engineering, on their own or in combination with mechanistic models. Commonly, the trained models are subsequently used in model-based optimization of design and/or operation of processes. Thus, it is critical to ensure that data-driven models are not evaluated outside their validity domain during process optimization. We propose a method to learn this validity domain and encode it as constraints in process optimization. We first perform a topological data analysis using persistent homology identifying potential holes or separated clusters in the training data. In case clusters or holes are identified, we train a one-class classifier, i.e., a one-class support vector machine, on the training data domain and encode it as constraints in the subsequent process optimization. Otherwise, we construct the convex hull of the data and encode it as constraints. We finally perform deterministic global process optimization with the data-driven models subject to their respective validity constraints. To ensure computational tractability, we develop a reduced-space formulation for trained one-class support vector machines and show that our formulation outperforms common full-space formulations by a factor of over 3,000, making it a viable tool for engineering applications. The method is ready-to-use and available open-source as part of our MeLOn toolbox (https://git.rwth-aachen.de/avt.svt/public/MeLOn).

Precision neutrino oscillation experiments of the future---of which DUNE is a prime example---require reliable event generator tools. The 1--4 GeV energy regime, in which DUNE will operate, is marked by the transition from the low-energy nuclear physics domain to that of perturbative QCD, resulting in rich and highly complex physics. Given this complexity, it is important to establish a validation procedure capable of disentangling the physical processes and testing each of them individually. Here, we demonstrate the utility of this approach by benchmarking the GENIE generator, currently used by all Fermilab-based experiments, against a broad set of inclusive electron-scattering data. This comparison takes advantage of the fact that, while electron-nucleus and neutrino-nucleus processes share a lot of common physics, electron scattering gives one access to precisely known beam energies and scattering kinematics. Exploring the kinematic parameter range relevant to DUNE in this manner, we observe patterns of large discrepancies between the generator and data. These discrepancies are most prominent in the pion-producing regimes and are present not only in medium-sized nuclei, including argon, but also in deuterium and hydrogen targets, indicating mismodeled hadronic physics. Several directions for possible improvement are discussed., Comment: 22 pages, 20 figures, matches the published version

Gaussian processes~(Kriging) are interpolating data-driven models that are frequently applied in various disciplines. Often, Gaussian processes are trained on datasets and are subsequently embedded as surrogate models in optimization problems. These optimization problems are nonconvex and global optimization is desired. However, previous literature observed computational burdens limiting deterministic global optimization to Gaussian processes trained on few data points. We propose a reduced-space formulation for deterministic global optimization with trained Gaussian processes embedded. For optimization, the branch-and-bound solver branches only on the degrees of freedom and McCormick relaxations are propagated through explicit Gaussian process models. The approach also leads to significantly smaller and computationally cheaper subproblems for lower and upper bounding. To further accelerate convergence, we derive envelopes of common covariance functions for GPs and tight relaxations of acquisition functions used in Bayesian optimization including expected improvement, probability of improvement, and lower confidence bound. In total, we reduce computational time by orders of magnitude compared to state-of-the-art methods, thus overcoming previous computational burdens. We demonstrate the performance and scaling of the proposed method and apply it to Bayesian optimization with global optimization of the acquisition function and chance-constrained programming. The Gaussian process models, acquisition functions, and training scripts are available open-source within the "MeLOn - Machine Learning Models for Optimization" toolbox~(https://git.rwth-aachen.de/avt.svt/public/MeLOn).

We point out that the LDMX (Light Dark Matter eXperiment) detector design, conceived to search for sub-GeV dark matter, will also have very advantageous characteristics to pursue electron-nucleus scattering measurements of direct relevance to the neutrino program at DUNE and elsewhere. These characteristics include a 4-GeV electron beam, a precision tracker, electromagnetic and hadronic calorimeters with near 2$\pi$ azimuthal acceptance from the forward beam axis out to $\sim$40$^\circ$ angle, and low reconstruction energy threshold. LDMX thus could provide (semi)exclusive cross section measurements, with detailed information about final-state electrons, pions, protons, and neutrons. We compare the predictions of two widely used neutrino generators (GENIE, GiBUU) in the LDMX region of acceptance to illustrate the large modeling discrepancies in electron-nucleus interactions at DUNE-like kinematics. We argue that discriminating between these predictions is well within the capabilities of the LDMX detector., Comment: 16 pages, 10 figures; updated to match the published version

Artificial neural networks (ANNs) are used in various applications for data-driven black-box modeling and subsequent optimization. Herein, we present an efficient method for deterministic global optimization of ANN embedded optimization problems. The proposed method is based on relaxations of algorithms using McCormick relaxations in a reduced-space [\textit{SIOPT}, 20 (2009), pp. 573-601] including the convex and concave envelopes of the nonlinear activation function of ANNs. The optimization problem is solved using our in-house global deterministic solver MAiNGO. The performance of the proposed method is shown in four optimization examples: an illustrative function, a fermentation process, a compressor plant and a chemical process optimization. The results show that computational solution time is favorable compared to the global general-purpose optimization solver BARON., Comment: J Optim Theory Appl (2018)

A simple and self-consistent approach has been proposed for simulation of the proton-induced soft error rate based on the heavy ion induced single event upset cross-section data and vice versa. The approach relies on the GEANT4 assisted Monte Carlo simulation of the secondary particle LET spectra produced by nuclear interactions. The method has been validated with the relevant in-flight soft error rate data for space protons and heavy ions. An approximate analytical relation is proposed and validated for a fast recalculation between the two types of experimental data., Comment: 14 pages, 8 figures

Martini et al. [Phys. Rev. C 94, 015501 (2016)] recently observed that when the produced-lepton's mass plays an important role, the charged-current quasielastic cross section for muon neutrinos can be higher than that for electron neutrinos. Here I argue that this effect appears solely in the theoretical descriptions of nuclear effects in which nucleon knockout requires the energy and momentum transfers to lie in a narrow range of the kinematically allowed values., Comment: 3 pages, 2 figures, ancillary file, the title and text have been modified to conform to the style of Phys. Rev. C

As the calorimetric method of neutrino-energy reconstruction is generally considered to be largely insensitive to nuclear effects, its application seems to be an effective way for reducing systematic uncertainties in oscillation experiments. To verify the validity of this opinion, we quantitatively study the sensitivity of the calorimetric energy reconstruction to the effect of final-state interactions in an ideal detector and in a realistic scenario. We find that when particles escaping detection carry away a non-negligible fraction of neutrino energy, the calorimetric reconstruction method becomes sensitive to nuclear effects which, in turn, affects the outcome of the oscillation analysis. These findings suggest that the best strategy for reduction of systematic uncertainties in future neutrino-oscillation studies---such as the Deep Underground Neutrino Experiment---is to increase their sensitivity to particles of low energy. The ambitious precision goals appear also to require an extensive development of theoretical models capable of providing an accurate predictions for exclusive cross sections of well-controlled uncertainties., Comment: 5 pages, 2 figures

A self-consistent procedure for the ion-induced soft error rate calculation in space environment taking into account Error Correcting Codes is proposed. The method is based on the partitioning of the multiple cell events into groups with different multiplicities. The proposed partitioning method has been vali-dated for the ground and on-orbit literature data., Comment: 5 pages, 8 figures

A procedure of space proton-induced Soft Error Rate calculation based on heavy ion testing data is proposed. The approach relies on Geant4-assisted Monte Carlo simulation of the secondary particle LET spectra produced by proton interactions., Comment: 2 pages, 3 figures, RADECS Conference abstract

A simple physical model for calculation of the ion-induced soft error rate in space environment has been proposed, based on the phenomenological cross section notion. Proposed numerical procedure is adapted to the multiple cell upset characterization in highly scaled memories. Nonlocality of the ion impact has been revealed as the key concept determining the difference between physical processes in low scaled and highly scaled memories. The model has been validated by comparison between the simulation results and the literature on-board data. It was shown that proposed method provides single-valued prediction results correlating well with on-board data based solely on cross section data and LET spectra without any hidden fitting parameters and procedures., Comment: 8 pages, 10 figures

Future neutrino-oscillation experiments are expected to bring definite answers to the questions of neutrino-mass hierarchy and violation of charge-parity symmetry in the lepton sector. To realize this ambitious program it is necessary to ensure a significant reduction of uncertainties, particularly those related to neutrino-energy reconstruction. In this paper, we discuss different sources of systematic uncertainties, paying special attention to those arising from nuclear effects and detector response. By analyzing nuclear effects we show the importance of developing accurate theoretical models, capable to provide quantitative description of neutrino cross sections, together with the relevance of their implementation in Monte Carlo generators and extensive testing against lepton-scattering data. We also point out the fundamental role of efforts aiming to determine detector responses in test-beam exposures., Comment: 26 pages, 10 figures

We report the results of a study aimed at quantifying the impact on the oscillation analysis of the uncertainties associated with the description of the neutrino-nucleus cross section in the two-particle--two-hole sector. The results of our calculations, based on the kinematic method of energy reconstruction and carried out comparing two data-driven approaches, show that the existing discrepancies in the neutrino cross sections have a sizable effect on the extracted oscillation parameters, particularly in the antineutrino channel., Comment: 7 pages, 5 figures

The hypothesis of the conserved vector current, relating the vector weak and isovector electromagnetic currents, plays a fundamental role in quantitative description of neutrino interactions. Despite being experimentally confirmed with great precision, it is not fully implemented in existing calculations of the cross section for inverse beta decay, the dominant mechanism of antineutrino scattering at energies below a few tens of MeV. In this article, I estimate the corresponding cross section and its uncertainty, ensuring conservation of the vector current. While converging to previous calculations at energies of several MeV, the obtained result is appreciably lower and predicts more directional positron production near the reaction threshold. These findings suggest that in the current estimate of the flux of geologically produced antineutrinos the 232Th and 238U components may be underestimated by 6.1 and 3.7%, respectively. The proposed search for light sterile neutrinos using a 144Ce--144Pr source is predicted to collect the total event rate lower by 3% than previously estimated and to observe a spectral distortion that could be misinterpreted as an oscillation signal. In reactor-antineutrino experiments, together with a re-evaluation of the positron spectra, the predicted event rate should be reduced by 0.9%, diminishing the size of the reported anomaly., Comment: 5 pages, 1 figure