Browse Articles
Discover research articles across all indexed journals
Enhancing student success prediction in higher education with swarm optimized enhanced efficientNet attention mechanism
Predicting student performance is crucial for providing personalized support and enhancing academic performance. Advanced machine-learning approaches are being used to understand student performance variables as educational data grows. A big dataset from several Chinese institutions and high schools is used to develop a credible student performance prediction technique. Moreover, the dataset includes 80 features and 200,000 records, and consequently, it represents one of the most extensive data collections available for educational research. Initially, data is passed through preprocessing to address outliers and missing values. In addition, we developed a novel hybrid feature selection model that combined correlation filtering with mutual information, Cross-Validation (CV) along with Recursive Feature Eliminatio (RFE) (R, and stability selection to identify the most impactful features. Moreover, This study develops the proposed EffiXNet, a more refined version of EfficientNet augmented with self-attention mechanisms, dynamic convolutions, improved normalization methods, and Sparrow Search Optimization Algorithm for hyperparameter optimization. The developed model was tested using an 80/20 train-test split, where 160,000 records were used for training and 40,000 for testing. The results reported, including accuracy, precision, recall, and F1-score, are based on the full test dataset. However, for better visualization, the confusion matrices display only a representative subset of test results. Furthermore, the EffiXNet value of AUC amounting to 0.99, a 25% reduction of logarithmic loss relative to the baseline models, precision of 97.8%, F1-score of 98.1%, and reliable optimization of memory usage. Significantly, the developed model showed a consistently high-performance level demonstrated by various metrics, which indicates that it is proficient in capturing intricate data patterns. The key insights the current research provides are the necessity of early intervention and directed training support in the educational domain. The EffiXNet framework offers a robust, scalable, and efficient solution for predicting student performance, with potential applications in academic institutions worldwide.
Improving a data mining based diagnostic support tool for rare diseases on the example of M. Fabry: Gender differences need to be taken into account
Background Rare diseases often present with a variety of clinical symptoms and therefore are challenging to diagnose. Fabry disease is an x-linked rare metabolic disorder. The severity of symptoms is usually different in men and women. Since therapeutic options for Fabry disease exist, early diagnosis is important. An artificial intelligence (AI)-based diagnosis support algorithm for rare diseases has been developed in preliminary studies. Objective Our aim was to extend and train the questionnaire-based AI, capable of distinguishing patients with from those without rare diseases, to achieve satisfactory sensitivity for the detection of a single rare disease, Fabry disease, taking into account gender differences in disease perception. Methods We collected 33 complete datasets from patients with confirmed Fabry disease. These records contained answered AI questionnaires, general information on disease progression, demographic information and quality of life (QoL) measures. The AI was trained to distinguish patients with Fabry disease from patients with relevant differential diagnoses. Its performance was assayed using stratified eleven-fold cross-validation and ROC curve calculation. Variables influencing the performance of the AI were examined with linear regression and calculation of the coefficient of determination. Result We were able to show that a relatively small sample is sufficient to achieve a sensitivity of 88.12% for the presence of Fabry disease, taking into account gender-specific differences in the disease perception during the pre-diagnostic phase. No confounders of the tool’s performance could be found in the data collected concerning the patients’ quality of life and diagnostic history. Conclusion This study illustrates on the example of Fabry disease that differences between female and male Fabry patients, not only in the expression of symptoms, but also with regard to disease perception, might be relevant influencing variables for improving the performance of AI-based diagnostic support tools for rare diseases.
A Macro-meso damage coupling rock mass damage model based on improved internal crack analysis
The complex geological genesis and different geological environments of rocks can lead to defects such as micropores, microcracks, and joints, resulting in the degradation of their engineering mechanical properties. In order to describe the deformation and failure mechanisms of jointed rock, a rock damage constitutive model that accounts for the geometric parameters and mechanical properties of internal joints within the rock is proposed. By leveraging the principles of damage mechanics and the Lemaitre strain equivalence hypothesis, macroscopic and microscopic damages in rock were combined, ultimately resulting in the development of a damage constitutive model that encompassed both scales of damage. The damage constitutive model was verified by uniaxial compression tests of clay-like rock under different joint dip angles and joint area combinations. The results indicated that the proposed damage constitutive model had clear physical significance and could fully reflect the process of rock failure, which was highly consistent with experimental results. The model parameter m reflected the brittleness of jointed rocks, while F₀ reflected the average strength of jointed rocks. The combination of macroscopic and microscopic analysis methods used in this study is reasonable, and the established damage constitutive model reflects the mechanical behavior of rocks and fits the experimental results well.
Quantum dynamics of a two-dimensional model in a microcavity: Polaritonic states of the Hénon–Heiles system
Chemistry under the conditions of vibrational strong coupling has recently attracted major attention from different experimental and theoretical groups alike. In particular, low-dimensional model systems have shed light on the possible formation mechanisms of vibrational polaritons and their dynamics in microcavities subject to the possible influence of external excitations. In the present contribution, the polaritonic states obtained by placing a Hénon–Heiles (HH) 2D model inside a microcavity and their dynamics are theoretically investigated. Oblique coordinates, as introduced by Zuñiga and co-workers [J. Phys. B: At., Mol. Opt. 50, 025101 (2017)], allow the accurate determination of polaritonic eigenstates, assuming the lowest fundamental mode of the HH model to be in resonance with a single cavity mode. Very different regimes for the polariton dynamics are found depending on whether the two vibrational modes of the model are themselves involved in a Fermi resonance or not. In the former case, the flow between photonic and vibrational modes appears mostly regular under molecular times scales and exhibits maximum efficiency near coupling strengths corresponding to avoided crossings in the polaritonic eigenstates. When the two HH modes are in the 1:2 ratio, the dynamics is much less regular but promotes intramolecular vibrational redistribution, although with a greater sensitivity toward initial conditions.
Water ice formed by vapor deposition and liquid aerosol injection: A comparison study using reflectance absorption infrared spectroscopy
Low temperature water ices (90–165 K) were produced via flash freezing during liquid aerosol injection (LAI) or vapor deposition (VD). The infrared spectral shapes of the O–H stretch at 3 μm (3300 cm−1) and the H–O–H bend at 6 μm (1600 cm−1) indicate that the two deposition methods produce different ice structures, with VD producing predominately more ordered structures than LAI at every deposition temperature studied. These different amorphous structures behave similarly with heating but remain spectrally different until crystallization, consistent with previous findings that hyperquenched glassy water and amorphous solid water are structurally different. This work demonstrates the utility of studies with experimental systems capable of directly comparing ice formation methods. To determine the presence of crystalline ice, the shape of the 3-μm feature is most useful, while the intensity of the 6-μm feature is a reliable indicator of amorphous ice and liquid-like behavior of the ice with heating. Liquid-like phases can be produced through LAI at all temperatures studied and through VD at the glass transition temperature of 136 K.
Extending Badger’s rule. II. The relationship between energy and vibrational spectra in hydrogen bonds
We describe the energetic–spectral relationship between the energy of a hydrogen bond (EHB) and the redshift in the vibrational frequency of the covalent hydrogen bond donor (ΔωHB). The relationship was derived by convoluting expressions relating EHB and ΔωHB to the covalent bond distance of the hydrogen bond donor (ΔrHB), while ensuring balanced treatment of the exchange repulsion. We relied on reduced parts of the potential energy surfaces (PESs) of six hydrogen bonded dimers, namely, NH3–NH3, H2O–H2O, HF–HF, H2O–NH3, HF–H2O, and HF–NH3 derived from ab initio electronic structure calculations to fit the parameters of the model and validated its performance for extended parts of the PESs that include non-linear hydrogen bonds. The developed model suggests a novel relationship of a strength of 4.5 kcal/mol per 100 cm−1 redshift in the covalent donor (D–H) vibrational frequency, while non-linear effects become important for redshifts >200 cm−1. The single descriptor (ΔωHB), which is measurable either experimentally via gas-phase spectroscopy or theoretically via electronic structure calculations, was able to predict the experimental or calculated hydrogen bond energies of typical hydrogen bonded dimers using the developed relationship.
Characterization of the hemithioindigo photoswitch and its derivatives with x-ray photoabsorption and photoemission spectroscopies
In this study, we investigate the electronic structure of hemithioindigo–hemistilbene (HTI) photoswitches and their functionalized derivatives, HTI-OMe and HTI-SMe, using x-ray photoemission spectroscopy (XPS) and near-edge x-ray absorption fine structure (NEXAFS) spectroscopy. HTI compounds are known for their high quantum yield, thermal bistability, and rapid photoisomerization, making them promising candidates for applications in molecular motors, optical materials, and photocatalysis. Our analysis, supported by first-principles simulations, reveals how the conjugation of heteroatoms within the π-system affects the core-level chemical shifts and ionization intensities in XPS, while NEXAFS probes the influence of substituents on virtual molecular orbitals and energy transitions. In particular, the comparison between different functionalized HTIs allowed us to evaluate the effect of electronic relaxation following core-level photoionization and photo-excitation. These results provide a detailed understanding of the influence of functionalization on the electron distribution of HTI compounds, providing a robust foundation for the study and control of ultrafast charge transfer and photoswitching mechanisms in these molecular systems.
Dissipative engineering with strong light–matter coupling for optimized photo-oxidation suppression in organic chromophores
This work addresses the critical challenge of dye molecule oxidation and its impact on device stability by investigating the suppression of photobleaching through engineering the chromophore’s surrounding environment. We induce strong coupling with confined light modes [optical cavities or localized surface plasmons (LSPs)] to reduce the triplet state population, thereby mitigating photo-oxidation. Utilizing the hierarchical-equations-of-motion approach to capture non-Markovian and non-perturbative effects, we analyze both cavity–chromophore and LSP–chromophore systems. Our analysis reveals that the optimal antioxidation performance depends on the competition between cavity–chromophore coupling and cavity–bath (dissipation) interaction. Importantly, in the strong cavity–chromophore coupling regime, increasing cavity dissipation can enhance antioxidation through quantum coherence-induced population transfer. Conversely, in the weak cavity–chromophore coupling regime, increasing cavity dissipation can counterintuitively reduce the antioxidation capability due to a quantum Zeno-like effect. Furthermore, in the LSP–chromophore system, engineering the LSP structure, particularly the LSP dissipation rate, can similarly be used to optimize the antioxidation effect. These findings, complemented by analytical results for finding optimal system parameters, provide practical guidelines for designing photostable organic materials with enhanced performance in various optoelectronic applications.
The role of quantum vibronic effects in the spin polarization of charge transport through molecular junctions
The connection between molecular vibrations and spin polarization in charge transport through molecular junctions is currently a topic of high interest, with important consequences for a variety of phenomena, such as chirality-induced spin selectivity. In this work, we follow this theme by exploring the relationship between vibronic dynamics and the corresponding spin polarization of the nonequilibrium charge current in a molecular junction. We employ the hierarchical equations of motion approach, which, since it is numerically exact and treats the vibrational degrees of freedom quantum mechanically, extends previous analyses of similar models that relied on approximate transport methods. We find significant spin polarization of the charge current in the off-resonant, low-voltage regime, where the vibrations must be treated quantum mechanically. Furthermore, we are able to connect the spin polarization in the charge transport to a corresponding polarization of the vibrational dynamics, which manifests itself in the vibrational angular momentum and excitation. Our analysis covers multiple molecule–lead couplings, temperatures, orbital energies, and spin–orbit couplings, demonstrating that the vibrationally assisted spin polarization is robust across a broad range of parameters.
Hybridization and coherence in subshell differential intercluster plasmonic decay in Na20<i>@</i>C240
We study the ground state structure and aspects of photoionization dynamics of the Na20@C240 endofullerene. The structure shows effects from the electronic coupling between the nested cluster and the fullerene cage. They include the (i) alterations of the overall potential, and thus, the force field, (ii) electron transfer from the cluster to the fullerene forming ionic units, and (iii) hybridization from the admixture of free Na20 occupied levels with experimentally known super-atom molecular orbital (SAMO) type empty levels of C240 accessible in the jellium-density functional theory model. These modifications influence the photoionization dynamics of the endofullerene. For the high energy ionization of Na20-type levels, a significant overall enhancement of the cross section is noted from additional ionizing force that C240 offers. More remarkably, the photoexcited plasmons, both the giant plasmon and the higher energy plasmon, in C240 decay in parts through Na20 ionization continuum via the resonant intercluster Coulombic decay (ICD) process. These lead to dramatic enhancements in the ionization of individual Na20-type levels, resulting in enhancements in the cluster’s total ionization yield. Based on hybridization, this enhancement incorporates a coherent mixing of the ICD and SAMO-induced Auger-decay amplitude, in which the ICD contribution is dominant.
Specific heat anomalies and local symmetry breaking in (anti-)fluorite materials: A machine learning molecular dynamics study
Understanding the high-temperature properties of materials with (anti-)fluorite structures is crucial for their application in nuclear reactors. In this study, we employ machine learning molecular dynamics (MLMD) simulations to investigate the high-temperature thermal properties of thorium dioxide, which has a fluorite structure, and lithium oxide, which has an anti-fluorite structure. Our results show that MLMD simulations effectively reproduce the reported thermal properties of these materials. A central focus of this work is the analysis of specific heat anomalies in these materials at high temperatures, commonly referred to as Bredig, pre-melting, or λ-transitions. We demonstrate that a local order parameter, analogous to those used to describe liquid–liquid transitions in supercooled water and liquid silica, can effectively characterize these specific heat anomalies. The local order parameter identifies two distinct types of defective structures: lattice defect-like and liquid-like local structures. Above the transition temperature, liquid-like local structures predominate and the sub-lattice character of mobile atoms disappears.
Evaluation of the electron affinity of substituted 9,10-anthraquinones using molecular anion lifetime data
The dissociative electron attachment (DEA) of anthraquinone (AQ) derivatives, namely, 9,10-AQ (I), 1,4-AQ (II), 1-(methylamino)-9,10-AQ (III), 1-(3-methoxyphenoxy)-9,10-AQ (IV), and 1,8-bis(benzamido)-9,10-anthraquinone (V) has been studied in the gas phase. The most intense negative ion in all DEA spectra is the molecular anion M−, which is observed in three and even four resonance states. In addition, some low-intensity decay channels are observed at higher electron energies: [M-H]−, [M-CH3]−, and [M-OCH3]−, depending on the type of substituent. To complete the picture, the molecules of naphthalene (VI), parabenzoquinone (VII), fluoranyl (VIII), and chloranyl (IX), for which the adiabatic electron affinity (EAa) values are known and the molecular anion lifetimes τa have been measured, have also been considered. The correlation between the EAa values (EALT) estimated in the simple Arrhenius approximation from molecular anion lifetime data and those (EADFT) calculated in the DFT approximation CAM-B3LYP/6-311+G(d,p) has been analyzed.
Sum-frequency vibrational specstrocopy of the bending mode of water at interfaces
We apply the mixed quantum/classical method and neural network-based molecular dynamics simulations to investigate sum-frequency vibrational spectroscopy (SFVS) of the bending vibration of the interfacial water. By analyzing the contributions of the electric dipole, electric quadrupole, magnetic dipole, and non-resonant mechanisms in SFVS, we demonstrate that electric quadrupole interactions play a dominant role in both the imaginary part of the SSP SFVS and the real and imaginary parts of the PPP SFVS. Conversely, the non-resonant contribution primarily influences the real part of the SSP SFVS signal. These findings significantly enhance our understanding of SFVS data interpretation by elucidating the underlying mechanisms governing the bending vibration at liquid interfaces. The results provide critical insights into resolving existing contradictions observed in SFVS studies, particularly concerning bending vibrational characteristics.
Effect of hard confinement on the phase state and dynamics of 1-propanol/water mixtures
1-propanol/water mixtures are structurally and dynamically heterogeneous over several length and time scales. Their phase diagram comprises liquid droplet phases of 1-propanol or water, as well as crystalline phases (hexagonal ice and different hydrates). We study the effect of hard confinement on the phase state and dynamics of 1-propanol/water mixtures. The mixtures were confined within self-ordered nanoporous alumina templates having long cylindrical nanopores. We show that the remote hard interface influences the phase state and the molecular dynamics. In the propanol-rich regime, where water exists in droplets, water could be supercooled below the homogeneous nucleation temperature and well within “no man’s land” where water crystallizes rapidly. Within the water-rich regime, the remote interface breaks the 1-propanol chain-like hydrogen bonded assemblies as reflected in the suppression of the Debye process. Calculation of the Hamaker constants provided quantitative information on distant intermolecular forces and their role in the structure and dynamics.
An efficient method for coupled rotation and torsion of oriented AXn1–BYn2 molecules with application to BF2BCl2
A new and concise method to obtain the eigenvalues and eigenfunctions of a two-dimensional rotational–torsional Hamiltonian with correct boundary conditions is proposed for oriented AXn1-BYn2 type molecules. The corresponding eigenfunctions are not only eigenfunctions of the Hamiltonian but also common eigenfunctions of several symmetry operators. Numerical calculations for the lowest 1000 rotational–torsional eigenstates of BF2BCl2 demonstrate the advantages and applicability of our method.
Performance of a Brownian information engine through potential profiling: Optimum output requisites, heating-to-refrigeration transition, and their re-entrance
Brownian information engine (BIE) harnesses the energy from a fluctuating environment by utilizing the associated information change in the presence of a single heat bath. The engine operates in a space-dependent confining potential and requires an appropriate feedback control mechanism. BIE utilizes the overall information (surprise) gained during the feedback cycle for the energy output. The feedback step is related to a sudden change in the potential energy and hence the information that is essential for a non-zero work output. The net (available) information, therefore, can be monitored by tuning the feedback controller and the shape of the confining potential. In this paper, we explore the effect of the shape modulation of the confining potential, which may have multiple stable valleys and unstable hills, on the available information and, hence, the performance of a BIE that operates under an asymmetric feedback protocol. For monostable trapping, a concave shape in the confining potential results in a higher work output than a convex one. We also find that hills and valleys in the confining potential may lead to multiple good operating conditions. An appropriate shape modulation can create a heater–refrigerator transition and its re-entrance due to non-trivial changes in information loss during the relaxation process.
Nuclear quantum effects at the liquid/vapor interface from neural-network based path integral molecular dynamics simulations
Nuclear quantum effects (NQEs) significantly influence the properties of water, including its structure, dynamics, and phase behavior. While their impact on bulk water has been extensively studied, their role at the liquid–vapor interface remains largely unexplored. In this work, we employ machine-learned neural network potentials trained on ab initio data to conduct large-scale path-integral molecular dynamics simulations at the RPBE-D3 level. Our results reveal that NQEs increase the surface tension, albeit marginally, shift the critical point to higher temperatures, and alter the orientational preferences of interfacial water molecules. This study provides the first direct quantification of the effect of NQEs on the surface tension of water. These findings highlight the fundamental role of quantum fluctuations in interfacial physics and underscore the necessity of including NQEs in accurate simulations of aqueous systems.
Particle dynamics in biconical cavities: First-passage, direct-transit, and looping time distributions
Earlier, we analyzed the effects of monotonically changing entropy potentials imposed by expanding or narrowing tubes on particle diffusion in such tubes [Berezhkovskii et al., J. Chem. Phys. 147, 134104 (2017)]. In the present study, we examine particle dynamics in biconical cavities, wherein particle motion is influenced by either an entropy potential well, as in a cavity composed of first expanding and then narrowing cones, or an entropy potential barrier, as in a cavity made up of first narrowing and then expanding identical cones. Both types of cavities are relevant to multiple technological and biological problems, where examples of such structures can be found at the micro- and nanoscales. We derive analytical expressions for the Laplace transforms of the distributions for the first-passage, direct-transit, and looping times in such structures. We find that not only the average values but also the distributions of the first-passage times in both cavities are indeed identical. However, the direct-transit and looping time distributions are drastically different. In particular, the mean direct-transit time for the expanding–narrowing cavity (entropy potential well) approaches a constant value with the increasing ratio of the cavity’s largest radius to the radius of its opening. In contrast, it goes to infinity in the case of the narrowing–expanding cavity (entropy potential barrier).
Phase transition of a chiral molecular model over the complete graph
This paper presents a mean-field approximation of the two-dimensional lattice chiral molecular model introduced by Lombardo et al., Proc. Natl. Acad. Sci. U. S. A. 106(36), 15131–15135 (2009) and simplified by Cruz-Simbron et al., J. Chem. Phys. 160, 084502 (2024), obtained by defining a chiral system over a complete graph. Using thermodynamic formalism, we derive a closed-form expression for the free energy of the system fβ; we establish its analyticity for all β &gt; 0 and its convexity over an interval depending on the Hamiltonian of the system. Our findings confirm the existence of first- and second-order phase transitions within this mean-field framework, confirming the first-order phase transition previously proposed by Lombardo et al., Proc. Natl. Acad. Sci. U. S. A. 106(36), 15131–15135 (2009) in the mean-field approximation. Furthermore, we construct a comprehensive phase diagram for the proposed model, providing a deeper understanding of its thermodynamic behavior.
Tensor hypercontraction for self-consistent vertex corrected GW with static and dynamic screening; applications to molecules and solids with superexchange
For molecules and solids, we developed efficient MPI-parallel algorithms for evaluating the second-order exchange (SOX) term with bare, statically screened, and dynamically screened interactions. We employ the resulting term in a fully self-consistent manner together with self-consistent GW (scGW), resulting in the following vertex-corrected scGW schemes: scGWSOX, scGWSOSEX, scGW2SOSEX, and scG3W2 theories. We show that for the vertex evaluation, the reduction of scaling by tensor hypercontraction has two limiting execution regimes. We used the resulting code to perform the largest (by the number of orbitals) fully self-consistent calculations with the SOX term. We demonstrate that our procedure allows for a reliable evaluation of even small energy differences. Utilizing a broken-symmetry approach, we explore the influence of the SOX term on the effective magnetic exchange couplings. We show that the treatment of SOX has a significant impact on the obtained values of the effective exchange constants, which we explain through a self-energy dependence on an effective dielectric constant. We confirm this explanation by analyzing natural orbitals and local changes in charge transfer, quantifying superexchange. Our analysis explains the structure of weak electron correlation responsible for the modulation of superexchange in both molecules and solids. Finally, for solids, we evaluate Néel temperatures utilizing the high-temperature expansion and compare the results obtained with experimental measurements. In addition, we prove a lack of Φ-derivability of the considered theories.