Unphysical Behavior Research Articles

A machine learning potential construction based on radial distribution function sampling.

Sampling reference data is crucial in machine learning potential (MLP) construction. Inadequate coverage of local configurations in reference data may lead to unphysical behaviors in MLP-based molecular dynamics (MLP-MD) simulations. To address this problem, this study proposes a new on-the-fly reference data sampling method called radial distribution function (RDF)-based data sampling for MLP construction. This method detects and extracts anomalous structures from the trajectories of MLP-MD simulations by focusing on the shapes of RDFs. The detected structures are added to the reference data to improve the accuracy of the MLP. This method allows us to realize a reasonable MLP construction for liquid water with minimal additional data. We prepare data from an H2O molecular cluster system and verify whether the constructed MLPs are practical for bulk water systems. MLP-MD simulations without RDF-based data sampling show unphysical behaviors, such as atomic collisions. In contrast, after applying this method, we obtain MLP-MD trajectories with features, such as RDF shapes and angle distributions, that are comparable to those of ab initio MD simulations. Our simulation results demonstrate that the RDF-based data sampling approach is useful for constructing MLPs that are robust to extrapolations from molecular cluster systems to bulk systems without any specialized know-how.

The discontinuous strain method: accurately representing fatigue and failure

AbstractFatigue simulation requires accurate modeling of unloading and reloading. However, classical ductile damage models treat deformations after complete failure as irrecoverable—which leads to unphysical behavior during unloading. This unphysical behavior stems from the continued accumulation of plastic strains after failure, resulting in an incorrect stress state at crack closure. As a remedy, we introduce a discontinuous strain in the additive elasto-plastic strain decomposition, which absorbs the excess strain after failure. This allows representing pre- and post-cracking regimes in a fully continuous setting, wherein the transition from the elasto-plastic response to cracking can be triggered at any arbitrary stage in a completely smooth manner. Moreover, the presented methodology does not exhibit the spurious energy release observed in hybrid approaches. In addition, our approach guarantees mesh-independent results by relying on a characteristic length scale—based on the discretization’s resolution. We name this new methodology the discontinuous strain method. The proposed approach requires only minor modifications of conventional plastic-damage routines. To convey the method in a didactic manner, the algorithmic modifications are first discussed for one- and subsequently for two-/three-dimensional implementations. Using a simple ductile constitutive model, the discontinuous strain method is validated against established two-dimensional benchmarks. The method is, however, independent of the employed constitutive model. Elastic, plastic, and damage models may thus be chosen arbitrarily. Furthermore, computational efforts associated with the method are minimal, rendering it advantageous for accurately representing low-cycle fatigue but potentially also for other scenarios requiring a discontinuity representation within a plastic-damage framework. An open-source implementation is provided to make the proposed method accessible.

Ensemble Generalization of the Perdew-Zunger Self-Interaction Correction: A Way Out of Multiple Minima and Symmetry Breaking.

The Perdew-Zunger (PZ) self-interaction correction (SIC) is an established tool to correct unphysical behavior in density functional approximations. Yet, the PZ-SIC is well-known to sometimes break molecular symmetries. An example of this is the benzene molecule, for which the PZ-SIC predicts a symmetry-broken electron density and molecular geometry, since the method does not describe the two possible Kekulé structures on an even footing, leading to local minima [Lehtola et al. J. Chem. Theory Comput. 2016, 12, 3195]. The PZ-SIC is often implemented with Fermi-Löwdin orbitals (FLOs), yielding the FLO-SIC method, which likewise has issues with symmetry breaking and local minima [Trepte et al. J. Chem. Phys. 2021, 155, 224109]. In this work, we propose a generalization of the PZ-SIC─the ensemble PZ-SIC (E-PZ-SIC) method─which shares the asymptotic computational scaling of the PZ-SIC (albeit with an additional prefactor). The E-PZ-SIC is straightforwardly applicable to various molecules, merely requiring one to average the self-interaction correction over all possible Kekulé structures, in line with chemical intuition. We showcase the implementation of the E-PZ-SIC with FLOs, as the resulting E-FLO-SIC method is easy to realize on top of an existing implementation of the FLO-SIC. We show that the E-FLO-SIC indeed eliminates symmetry breaking, reproducing a symmetric electron density and molecular geometry for benzene. The ensemble approach suggested herein could also be employed within approximate or locally scaled variants of the PZ-SIC and its FLO-SIC versions.

Quantum nature of spacetime near the black hole singularity

The concept of spacetime loses its usual interpretation at the essential singularity of a black hole. In consequence, all laws of physics must fail at this classical singularity. This unphysical behavior of spacetime at the singularity originates from general relativity. In order to have a consistent description of spacetime, this singularity must disappear in a quantum mechanical description of spacetime which is expected to be given by a quantum theory of gravity. In this paper, we therefore attempt to describe the quantum nature of spacetime in the vicinity of the (classical) singularity of a black hole. We take the Kantowsi–Sachs representation for the interior spacetime of a black hole and include inevitable vacuum fluctuations of matter field in the Klein–Gordon representation. Hence we obtain the Wheeler–DeWitt equation for the black hole interior and solve this equation exactly yielding a general expression for the interior wave function of the black hole. Admissible wave functions consistent with the DeWitt boundary condition implies that the Hilbert space has three nonoverlapping sectors distinguished by the relative character of the eigenvalues. Regular quantum black holes with admissible and well-behaved wave function having no singularity can exist only in two of those sectors. However, the remaining sector does not contain any regular quantum black hole.

Influence of Rolling Resistance and Particle Size Distribution in the Simulation of Sand Infiltration into the Static Gravel Bed

Fine sediment infiltration and subsequent clogging in a gravel bed affect several fluvial, ecological, and biological processes, resulting in the degradation of the river ecosystem. Despite many experimental and a few numerical studies, the process is yet to be entirely understood. We employed a pure Lagrangian framework, called the Discrete Element Method (DEM), to numerically investigate the infiltration process. Special attention is given to tackling the issue of non-spherical and irregular particle shapes and particle size distributions (PSDs) in numerical simulations. Due to computational limitations, these aspects were either not considered or simplified in previous numerical studies. We implicitly included non-spherical and irregular shape effects through rolling resistance models, which do not cause excessive computational overhead. Our study shows that rolling resistance models greatly influence packing and fine sediment infiltration. However, they may also lead to unphysical particle behavior; thus, they should be carefully used in numerical simulations. Oversimplified PSDs do not mirror natural systems, and full PSDs pose computational challenges. Sufficient grain classes are needed to mimic the non-homogeneity and poly-dispersity found in natural fluvial sediments. The infiltrating characteristics of sand concerning PSD and shape effects are linked to size ratio D15,Gravel/D85,Sand, assuring physical and realistic modeling of the infiltration process.

Physics-driven neural networks for nonlinear micromechanics

Micromechanical problems seek full-field solutions in response to external and/or internal thermal-mechanical loads, which have been increasingly encountered in material property and process assessments. Here we present a machine learning based strategy of solving micromechanics at finite strains. Deep neural networks (DNNs) are trained by employing the elastic energy together with other physical constraints to formulate the loss function. In particular, the crossover of material points, which can be a common issue for similar DNNs when applied to compression-dominant problems, is shown to be effectively addressed by including a kinematic penalty term that forces the DNN to maintain structural integrity and stability and avoid unphysical behavior. The generality and accuracy of the proposed physics-driven neural networks (PDNNs) are demonstrated through various micromechanical problems, including single crystal anisotropy elasticity, Eshelby's inclusion problem, buckling, and elastic homogenization of polycrystals. The effect of the choice of optimizers and hyperparameters on the PDNN training are discussed and the computational efficiency is also analyzed. It is shown that this novel PDNN framework can completely remove the need of labeled training data and exhibit improved performances as compared to the conventional physics-informed neural networks (PINNs) in terms of avoiding unphysical solutions and attaining higher computational efficiency. PDNNs can thus serve as a promising tool to solve some critical challenges associated with a wide range of nonlinear micromechanical problems.

Reworking the Tao-Mo exchange-correlation functional. II. De-orbitalization.

In Paper I [H. Francisco, A. C. Cancio, and S. B. Trickey, J. Chem. Phys. 159, 214102 (2023)], we gave a regularization of the Tao-Mo exchange functional that removes the order-of-limits problem in the original Tao-Mo form and also eliminates the unphysical behavior introduced by an earlier regularization while essentially preserving compliance with the second-order gradient expansion. The resulting simplified, regularized (sregTM) functional delivers performance on standard molecular and solid state test sets equal to that of the earlier revised, regularized Tao-Mo functional. Here, we address de-orbitalization of that new sregTM into a pure density functional. We summarize the failures of the Mejía-Rodríguez and Trickey de-orbitalization strategy [Phys. Rev. A 96, 052512 (2017)] when used with both versions. We discuss how those failures apparently arise in the so-called z' indicator function and in substitutes for the reduced density Laplacian in the parent functionals. Then, we show that the sregTM functional can be de-orbitalized somewhat well with a rather peculiarly parameterized version of the previously used deorbitalizer. We discuss, briefly, a de-orbitalization that works in the sense of reproducing error patterns but that apparently succeeds by cancelation of major qualitative errors associated with the de-orbitalized indicator functions α and z, hence, is not recommended. We suggest that the same issue underlies the earlier finding of comparatively mediocre performance of the de-orbitalized Tao-Perdew-Staroverov-Scuseri functional. Our work demonstrates that the intricacy of such two-indicator functionals magnifies the errors introduced by the Mejía-Rodríguez and Trickey de-orbitalization approach in ways that are extremely difficult to analyze and correct.

Revised models for evaluating hydrocarbon saturation of tight sandstone formations based on dielectric dispersion logs

AbstractAccurate estimation of hydrocarbon saturation is significant for log interpreters to evaluate tight sandstone formations. It is difficult to use conventional wireline logs to calculate hydrocarbon saturation in complex reservoirs. Dielectric dispersion logs are processed using dielectric models to obtain accurate saturation assessment, especially in high salinity shaly sand formations. However, the dielectric models consist of complex refractive index method model and shaly sand model ignore the impact of clay effects on complex refractive index method model, and the dielectric dispersion logs generated by them appear several unphysical phenomena in the clay‐bearing formations, such as the permittivity data at the highest frequency are greater than those at the second highest frequency and the conductivity data at the highest frequency are lower than those at the second highest frequency. This paper modifies the dielectric models by introducing a combined conductive phase in complex refractive index method model to correct the unphysical behaviours. Revised models are combined with four permittivity and four conductivity logs acquired by dielectric scanner at 24, 102, 360 and 960 MHz to inverse the water saturation, water salinity, textural index for water‐bearing pore and fraction of clay‐related phase. The complex refractive index method model interprets the logs measured at 960 MHz and shaly‐sand model interprets other dielectric dispersion logs. The particle swarm optimization algorithm is used to search inversion parameters. The results of forward simulation show that the revised models can correct unphysical behaviours of dielectric dispersion logs generated by the original models. Moreover, we apply original models and revised models to laboratory core data. The results of experiments show that the mean relative error values of the original models are 19.38% and 23.60% and of the revised models are 5.54% and 6.28% in two cases. It shows that the revised models are more accurate than the original models to interpret dielectric dispersion logs of tight sandstone reservoirs.

Assessment of interfacial turbulence treatment models for free surface flows

The modelling of complex free surface flows is challenging due to the mobility and deformability of the interface and air entrainment characteristics, which are highly affected by turbulence. With the framework of Reynolds averaged Navier–Stokes (RANS) models and the volume of fluid (VOF) method, turbulence quantities at the air–water interface tend to be over-estimated. In this study, interfacial turbulence treatment methods including the buoyancy modification model based on the simple gradient diffusion hypothesis (SGDH) and Egorov’s turbulence damping model are investigated. Furthermore, due to the unconditionally unstable characteristics of the standard k-ε turbulence model, the stabilized k-ε turbulence model is applied as a comparison. The turbulence attenuation performance using different interfacial turbulence treatment methods in the vicinity of the interface is compared and discussed for stratified flows and free overflow weirs for aerated and non-aerated nappe scenarios. The turbulence quantities and free surface profile under different flow conditions are validated against experimental data and an analytical model. The results show that for free surface waves, both the SGDH model and the turbulence damping model give strong improvements in turbulence production compared with the standard k-ε model. The SGDH model augments the turbulence kinetic energy (TKE) in the unstable stratification, leading to unphysical behaviour for the partially dispersed and separated flow.

Interplay between perturbative and non-perturbative effects with the ARES method

We present a new semi-numerical method to compute leading hadronisation corrections to two-jet event shapes in e+e− annihilation. The formalism we present utilises the dispersive approach, where the magnitude of power corrections is controlled by suitable moments of an effective strong coupling, but it can be adapted to other methods. We focus on observables where the interplay between perturbative and non-perturbative effects is crucial in determining the power corrections. A naive treatment of power corrections for some of these observables gives rise to an unphysical behaviour in the corresponding distributions for moderate observable values, thus considerably limiting the available range to fit the non-perturbative moments. We present a universal treatment to handle such observables, based on a suitable subtraction procedure, and compare our results to the analytic result in the case of total broadening. Finally, for the first time we present predictions for the thrust major, which cannot be handled with analytic methods.

Characterization of Terfenol-D and Comparison With Predictions of a Thermodynamically Consistent Two-Input, Two-Output Model of Hysteresis

Since its discovery more than forty years ago, Terfenol-D has promised much but has not yet been fully exploited in applications.The main reason for this is that it shows significant hysteresis and other nonlinearities, such as saturation, which need to be modeled correctly in order to maximize the application potential. In this article, we focus on the characterization of a Terfenol-D sample with the intention of developing a predictive two-input (magnetic field and stress) and two-output (magnetic flux density and strain) model of magnetostriction that exhibits hysteresis and saturation and is compatible with the laws of thermodynamics. Compatibility with thermodynamics is essential to ensure that numerical simulations do not exhibit unphysical behavior. The model is based on the Preisach hysteresis operator and its storage function and may be interpreted as a two-input, two-output neural net with elementary hysteresis operators as the neurons. In order to estimate the parameters, it is necessary to collect data over a wide range of magnetic fields (-300 kA/m to 300 kA/m) and compressive stresses (up to 60 MPa). Using the rate-independent memory evolution properties of the Preisach operator, we split the parameter estimation problem into three numerically well-conditioned, linear least squares problems with constraints. We show that the model is able to fit experimental data for strain and magnetization over a wide range of magnetic fields and stress.

A Physics-Informed Automatic Neural Network Generation Framework for Emerging Device Modeling.

With the rapid development of semiconductor technology, traditional equation-based modeling faces challenges in accuracy and development time. To overcome these limitations, neural network (NN)-based modeling methods have been proposed. However, the NN-based compact model encounters two major issues. Firstly, it exhibits unphysical behaviors such as un-smoothness and non-monotonicity, which hinder its practical use. Secondly, finding an appropriate NN structure with high accuracy requires expertise and is time-consuming. In this paper, we propose an Automatic Physical-Informed Neural Network (AutoPINN) generation framework to solve these challenges. The framework consists of two parts: the Physics-Informed Neural Network (PINN) and the two-step Automatic Neural Network (AutoNN). The PINN is introduced to resolve unphysical issues by incorporating physical information. The AutoNN assists the PINN in automatically determining an optimal structure without human involvement. We evaluate the proposed AutoPINN framework on the gate-all-around transistor device. The results demonstrate that AutoPINN achieves an error of less than 0.05%. The generalization of our NN is promising, as validated by the test error and the loss landscape. The results demonstrate smoothness in high-order derivatives, and the monotonicity can be well-preserved. We believe that this work has the potential to accelerate the development and simulation process of emerging devices.

Photoionization of C60 at high energies

Calculations of the photoionization cross section of C60 including all valence and core subshells have been carried out for photon energies up to 1 keV using density functional theory (DFT) and time-dependent DFT (TDDFT) within the framework of a fully molecular model. The high-energy valence subshell cross sections behave rather differently than the results from model potentials where the cross section falls far too rapidly with energy. This unphysical behavior of the model calculations is traced to smearing out of the carbon nuclei in the model potentials which makes it difficult for momentum to be conserved at the higher energies. Comparison with 60 times carbon atomic cross section sheds light on similarities and differences. The high energy behavior of individual valence cross sections reflects the amount of C 2s/2p mixing in the molecular orbitals.

New features of the TFmix code: Thermodynamic properties of electrons in mixtures

New features of the TFmix code: Thermodynamic properties of electrons in mixtures

About the applicability of the theory of porous media for the modelling of non‐isothermal material injection into porous structures

AbstractIn this contribution we investigate the relevance of the theory of porous media for the non‐isothermal modelling of material injection into porous structures. In particular, we provide a model describing the injection of cement during percutaneous vertebroplasty, which is derived by consistently following the theory of porous media. We demonstrate numerically that this model elicits unphysical behaviour under local thermal non‐equilibrium conditions. No distinct unphysical behaviour is observed under local thermal equilibrium conditions. We conclude that heuristic modifications of the model equations are necessary and suspect the unphysical behaviour to be caused by contradictory modelling assumptions.

A novel constitutive model for surface elasticity at finite strains suitable across compressibility spectrum

The surface elasticity theory of Gurtin–Murdoch has proven to be remarkably successful in predicting the behavior of materials at the nano scale, which can be attributed to the fact that the surface-to-volume ratio increases as the problem dimension decreases. On the other hand, surface tension can deform soft elastic solids even at the macro scale resulting e.g. in elastocapillary instabilities in soft filaments reminiscent of Plateau–Rayleigh instabilities in fluids. Due to the increasing number of applications involving nanoscale structures and soft solids such as gels, the surface elasticity theory has experienced a prolific growth in the past two decades. Despite the large body of literature on the subject, the constitutive models of surface elasticity theory at large deformations are not suitable to capture the surface behavior from fully compressible to nearly incompressible elasticity, especially from a computational perspective. A physically meaningful and proper decomposition of the surface free energy density in terms of area-preserving and area-varying contributions remains yet to be established. We show that an immediate and intuitive generalization of the small-deformation surface constitutive models does not pass the simple extension test at large deformations and results in unphysical behavior at lower Poisson’s ratios. Thus, the first contribution of the manuscript is to introduce a novel decomposed surface free energy density that recovers surface elasticity across the compressibility spectrum. The second objective of this paper is to formulate an axisymmetric counterpart of the elastocapillary theory methodically derived from its three-dimensional format based on meaningful measures relevant to the proposed surface elasticity model. Various aspects of the problem are elucidated and discussed through numerical examples using the finite element method enhanced with surface elasticity.

Divergence-Preserving MHD Solutions Using a Noniterative Hyperbolic Conservation Law Scheme

The divergence constraint, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\nabla$</tex-math> </inline-formula> <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\cdot$</tex-math> </inline-formula> B <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$=$</tex-math> </inline-formula> 0, presents a central challenging numerical feature of electromagnetic and magnetohydrodynamic solvers. In general, this constraint is not strictly enforced in hyperbolic solvers and tends to accumulate error over the duration of a simulation, polluting the solution with unphysical behavior. In this article, we present a numerical scheme that constructs a system of hyperbolic conservation laws that already satisfy the divergence constraint, starting from the potential form of Maxwell’s equations and basic derivative relationships. The new system consists only of the first-order spatial differentiation, in contrast to the potential equations which are second order; this allows them to be directly integrated into the existing hyperbolic solvers with no additional implementation changes needed. Furthermore, since the system of conservation laws constructed automatically satisfies the divergence criteria, no subiteration is required to resolve the divergence-free solution. We present the results of integrating this new conservation law approach in an existing finite volume plasmadynamic solver that discretizes the coupled Maxwell and Navier–Stokes equations. Numerical tests indicate that the new approach can strongly resolve physical solutions. The divergence error is also measured and quantified for these problems, showing that the new system suppresses divergence error by many orders of magnitude.

Development and implementation of a Direct Surface Description method for free surface flows in OpenFOAM

The solution procedure for two phases in OpenFOAM suffers from unphysical velocity oscillations at the free surface between the two phases. It is likely that this problem also exists in other two-phase Computational Fluid Dynamics (CFD) codes. We aim to solve this by imposing boundary conditions directly on the free surface. We have taken the first step towards a new two-phase solution method by first addressing the water phase alone. It is a free surface modelling method based on merging concepts from two existing methods: (1) A single-phase free surface method and (2) the solution method used in OpenFOAM. The underlying motivation is to enable more accurate estimation of wave induced load distributions from wave crest impacts on offshore structures. This first and foremost requires an accurate prediction of the kinematics near the free surface. We present a solution method with boundary conditions directly on the free surface, thereby the name: Direct Surface Description (DSD). Additionally it is the first time that the isoAdvector algorithm is combined with a single phase free surface method. The implementation is made in OpenFOAM, but may also relevant to other codes as well. First a still water level simulation is presented to illustrate the unphysical behavior of the existing solvers and validate the behaviour of the DSD method. The second test case is a moderately steep stream function wave in intermediate water depth. The DSD method is validated and compared to the existing solution methods of OpenFOAM: interFoam and interIsoFoam . We present a detailed comparison of surface elevations and velocity profiles. This is followed by a convergence study including wave height, velocity and phase shift. Additionally the influence of the Courant–Friedrichs–Lewy (CFL) number is studied. The stream function wave case demonstrates that the DSD method accurately predicts the free surface elevation and velocity fields without free surface undulations or oscillatory velocity fields. The convergence study underlines an increased accuracy of the DSD method. Finally, a 2D and a 3D showcase with breaking waves are presented to show that the DSD method is capable of simulating more complex and realistic cases. • The oscillatory velocities at free surface in OpenFOAM is removed by DSD-method. • Presents a rarely seen comparison of velocity profiles below wave crest and trough. • New DSD-method eliminates the air region leading to decreased CPU time. • Implementation in OpenFOAM makes the method ready for unstructured meshes. • Presents convergence study of velocity and wave heights w.r.t. mesh and CFL number.

Intrinsic Pathology of Self-Interacting Vector Fields.

We show that self-interacting vector field theories exhibit unphysical behavior even when they are not coupled to any external field. This means any theory featuring such vectors is in danger of being unphysical, an alarming prospect for many proposals in cosmology, gravity, high energy physics, and beyond. The problem arises when vector fields with healthy configurations naturally reach a point where time evolution is mathematically ill defined. We develop tools to easily identify this issue, and provide a simple and unifying framework to investigate it.

Downhole Pressure Prediction for Deepwater Gas Reservoirs Using Physics-Based and Machine Learning Models

Summary Pressure measurement from permanent downhole gauges (PDHGs) during extended shut ins (SIs) is a key piece of information that is often used for model calibration and reserve estimation in deepwater gas reservoirs. A key challenge in practical operation has been the failure of PDHGs within the first few years of operation. In this work, a physics-based data-driven (PBDD) model and machine learning (ML) models are developed to predict PDHG pressure and temperature measurement from the wellhead and other measurements during well SI events for deepwater dry-gas wells. During SI events, the wellbore cools down, resulting in increased gas density and bottomhole pressure (BHP). In the PBDD model, the temperature profile in the well is modeled with a piece-wise linear model as derived from wellbore simulations. The temperature decline during cooldown is captured using a decline-curve model, with the decline-curve parameters dependent on the location. The dependency of the cooldown effect on past production is captured with a linear model. Model parameters in the PBDD model are calibrated with data. In the ML models, multiple methods are tested, and the best performing method is picked based on cross-validation results. Two use cases are considered in this work. The first case (single well) involves predicting future SI BHP and temperature based on past PDHG measurement of the same well. Both the PBDD model and the ML model show good accuracy in blind tests for this use case. The second case involves predicting SI BHP and temperature of a well based on PDHG measurement of other wells. The PBDD model sees reduced accuracy in temperature prediction but is still reasonably accurate, while unphysical behavior is observed for the ML model even though the cross-validation score is high. It is concluded that, comparing the two types of models, the PBDD model is constrained by physics and thus the result is more interpretable and reasonable even when extrapolating. It also can provide the entire temperature and pressure profile during SIs. However, it does come with a series of assumptions (such as dry gas with no liquid content) and needs to be modified when the problem changes. On the other hand, the ML model is easier to construct and extend to other cases but is not bounded by physics so the result could be unphysical when extrapolation occurs.