Information processing device, information processing method, program, and non-transitory computer-readable medium

Abstract

To achieve highly accurate physical property estimation with high versatility and with high efficiency.SOLUTION: An information processing device includes a memory trained by machine learning to store information on a force field model, and a processor. The processor defines a molecular model representing a group of molecules, and a first solid wall model and a second solid wall model representing a solid walls, each of which having a thickness in a third direction and extending in a first direction and a second direction, among the first direction, the second direction crossing the first direction, and the third direction crossing the first direction and the second direction, and fixes the first solid wall model, and executes molecular dynamics simulation obtained by adding force to the second solid wall model by using the force field model in a first model in which the molecular model is sandwiched between the first solid wall mode and the second wall model, and calculates stress of the molecular model to be acquired as a result of the molecular dynamics simulation.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, a program, and a non-transitory computer-readable medium. [Background technology]

[0002] At surfaces and interfaces where various fluids such as aqueous solutions, oils, and greases come into contact with or slide against solids, complex phenomena such as molecular structural changes and chemical reactions occur.One common method for estimating physical properties such as wettability, adsorption, viscosity, coefficient of friction, traction coefficient, and wear coefficient based on the phenomena that occur is to use physical simulations such as molecular dynamics.

[0003] In predicting physical properties using molecular simulations, the accuracy of the force field, which expresses the relationship between interatomic distances and energy, as well as the calculation algorithm, determines the accuracy of the predicted values. Force fields based on classical theory have the advantage of being highly computationally efficient, but they have the problem of being low in accuracy and being unable to handle inexperienced systems. On the other hand, when attempting to construct a force field based on quantum theory, a highly accurate force field can be obtained, but it is tuned for a specific system and has low versatility (generalizability or universality), and it takes a huge amount of calculation time to obtain the force field necessary for predicting physical properties.

[0004] For example, a method for estimating the traction coefficient using a force field based on classical theory has been proposed (Patent Document 1), which estimates the traction coefficient for a specific fluid and shows good agreement with experimental values. However, force fields based on classical theory cannot maintain the accuracy of predictions for new compounds, mixed systems in which two or more compounds are mixed in any ratio, polarizable ionic liquids, etc., resulting in problems with accuracy and versatility. Academia and industry have been calling for the establishment of a method for estimating the physical properties of fluids using a force field that combines computational efficiency, accuracy, and versatility. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2009-301437 Summary of the Invention [Problem to be solved by the invention]

[0006] This disclosure proposes an information processing device that is highly versatile, efficient, and achieves highly accurate physical property predictions by utilizing molecular simulations that use neural network potentials (hereinafter referred to as NNPs) as a force field. [Means for solving the problem]

[0007] According to one embodiment, an information processing device includes a memory that stores information about a force field model trained by machine learning, and a processor, wherein the processor defines a molecular model representing a group of molecules extending in a first direction, a second direction intersecting the first direction, and a third direction intersecting the first and second directions, the molecular model having a thickness in the third direction and extending in the first and second directions, and a first solid wall model and a second solid wall model representing solid walls, fixes the first solid wall model, and performs a molecular dynamics simulation using the force field model in a first model in which the molecular model is sandwiched between the first and second solid wall models and applies a force to the second solid wall model, and calculates a stress of the molecular model obtained as a result of the molecular dynamics simulation. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram showing an outline of an information processing apparatus according to an embodiment. [Figure 2] FIG. 10 is a diagram illustrating an example of model formation according to an embodiment. [Figure 3] 10 is a flowchart showing processing of an information processing apparatus according to an embodiment. [Figure 4]FIG. 1 is a cross-sectional view schematically showing a model according to an embodiment. [Figure 5] FIG. 1 is a cross-sectional view schematically showing a model according to an embodiment. [Figure 6] FIG. 10 is a diagram showing an example of an output of shear stress according to an embodiment. [Figure 7] FIG. 10 is a diagram showing an example of an output of a friction coefficient according to an embodiment. [Figure 8] FIG. 10 is a diagram showing a simulation result of a friction coefficient according to an embodiment. [Figure 9] FIG. 10 is a diagram showing an example of a model of a solid wall according to an embodiment. [Figure 10] 10 is a graph illustrating an example of a friction coefficient output according to an embodiment. [Figure 11] FIG. 10 is a diagram showing an example of a model of a solid wall according to an embodiment. [Figure 12] 10 is a graph illustrating an example of a friction coefficient output according to an embodiment. [Figure 13] FIG. 10 is a diagram showing an example of an input according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] 1 is a block diagram showing an outline of an information processing device according to one embodiment. The information processing device 1 includes a processor 10, a storage unit 12, and an input / output interface (hereinafter referred to as input / output I / F 14), which are appropriately connected to each other via a bus 16. The information processing device 1 acquires predetermined physical properties by inputting the type and position of the element for an atom. Note that the information processing device 1 is equipped with components necessary for appropriately executing information processing in addition to those shown in the figure.

[0010] The processor 10 executes arithmetic processing in the information processing device 1. The processor 10 includes processing circuits such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 10 may also control the startup and other operations of the information processing device 1. When an accelerator is used, an external processor connected to the processor 10 via an input / output I / F 14 may be used as the accelerator. In other words, the processor 10 for performing arithmetic processing may be provided outside the information processing device 1.

[0011] The memory unit 12 stores data and the like required for starting up the information processing device 1 and estimating physical properties in the information processing device 1. The memory unit 12 includes storage media such as appropriate memory and storage. These storage media include temporary or non-temporary computer-readable media, and the processor 10 executes the required operations by accessing the storage media. When the information processing device 1 specifically realizes information processing by software using hardware resources, the memory unit 12 may store programs, executable files, and the like for executing the information processing by the software. The memory unit 12 may also store data in the process of information processing and information such as physical properties acquired as a result of the information processing.

[0012] The input / output I / F 14 is an interface that connects the processor 10 and the storage unit 12 to the outside. The information processing device 1 receives requests from a user via this input / output I / F 14 and outputs appropriate data to appropriate locations. The input / output I / F 14 can also be used as an interface that connects the processor 10 and the storage unit 12 to an external device when the processor 10 executes calculations using the external device. This connection may be, for example, a wired or wireless network line, or other connection means such as a serial cable or a parallel cable.

[0013] The information processing device 1 executes a molecular dynamics simulation to estimate a physical property value that a user desires to obtain. Molecular dynamics simulations include a classical method, a method based on quantum theory, and a method called NNP (Neural Network Potential) that uses a neural network model trained by machine learning. In the present disclosure, the information processing device 1 executes a molecular dynamics simulation based on NNP to obtain a physical property value.

[0014] This physical property value is, for example, the coefficient of friction at the interface where a solid and a liquid come into contact. The information processing device 1 acquires the mechanical response at the interface between materials (solid and liquid) with high accuracy and in a short time by using NNP, and acquires the friction coefficient from the change in this mechanical response over time. Note that the force field model based on NNP may be a model trained using simulation results based on quantum mechanics as training data.

[0015] 2 shows a non-limiting example of a model to be subjected to molecular dynamics simulation in this embodiment. The processor 10 inputs a set of atoms shown in FIG. 2 as a cell into a force field model based on NNP. Note that the composition of the constituent molecules is shown as a non-limiting example and is not limited thereto.

[0016] The model input to the force field model is a model formed within a cell. In this embodiment, a model is formed between solid walls that have a thickness in the third direction and extend in the first and second directions, and that represents a molecular structure of an object whose physical properties are to be determined for the solid wall.

[0017] The solid wall is formed by two models: a first solid wall model whose position is fixed, and a second solid wall model whose position can be fixed or movable. A model representing an input cell is formed by placing a molecular model that forms the target molecular structure between these two solid walls.

[0018] That is, a first model in which a molecular model is sandwiched between two solid wall models is set as a cell model. Of the first model, a model in which a second solid wall model is fixed is set as a second model, and the processor 10 moves the second solid wall model in the first model using this second model as an initial value, thereby obtaining various characteristic values ​​when the second solid wall applies a force to the molecular model or when the second solid wall moves relative to the first solid wall.

[0019] The molecular model is a model that represents a substance that represents, for example, a group of molecules, a liquid, a crystal, or an amorphous solid. However, the molecular model is not limited to these examples, and any substance that exists between solid walls and has an interface with the solid walls, the physical properties of which are to be obtained, may be modeled as a molecular model.

[0020] By defining each of these models, a model within the cell is formed. In this embodiment, the cell input to the force field model is a model having periodic boundary conditions. By inputting this cell model into the force field model, interatomic potentials are calculated and various physical properties are estimated.

[0021] The size of the cell may be any size that allows the simulation to be performed appropriately, but may also be a size that allows the molecular configuration in the molecular model to include a variety of position and posture patterns. Furthermore, the first solid wall model or the second solid wall model may be positioned sufficiently far from the bottom or top surface of the cell in the third direction. By positioning them in this way, the influence of the interatomic potential between the first solid wall model and the second solid wall model can be ignored.

[0022] The following describes the case where the molecular model sandwiched between solid walls is a liquid.

[0023] 3 is a flowchart showing the processing of the information processing device 1 according to an embodiment. A non-limiting example of the processing will be described with reference to FIG.

[0024] First, the processor 10 prepares a cell to be used in the molecular dynamics simulation (S100). Atoms are appropriately arranged within the space of this prepared cell to generate a model to be used in the molecular dynamics simulation.

[0025] Next, the processor 10 forms a molecular model within the cell (S102). As described above, this molecular model is a model that includes liquid molecules sandwiched between solid walls. Note that the molecular model is not limited to a model of a substance that is composed of one type of molecule, but may also be a model of a liquid in which multiple types of molecules are mixed.

[0026] Next, the processor 10 forms a solid wall model within the cell (S104). These solid walls are a first solid wall model and a second solid wall model formed above and below the molecular model, as shown in Fig. 2. The distance between these solid walls may be defined, for example, assuming that the liquid is a lubricant, based on the distance between objects to be lubricated by the lubricant.

[0027] S102 and S104 do not have to be executed in this order, but may be executed in the reverse order, or may be executed in parallel.

[0028] Next, the processor 10 generates a first model in which the first solid wall model is fixed and the second solid wall model is movable on a plane formed by the first direction and the second direction (S106).

[0029] Next, the processor 10 generates a second model (S108) in which the first solid wall model and the second solid wall model are fixed and the molecular model is sandwiched between these models and has an interface that contacts each of the solid wall models. This second model can also be formed as a special case of the first model.

[0030] S106 and S108 may be executed in the reverse order or in parallel, instead of in this order. When executed in the reverse order, the processor 10 may form the first model based on the second model as a special case of the second model.

[0031] Next, the processor 10 forms a force field model (S110). This force field model is a model based on NNP and is a model that has been trained in advance by machine learning. The processor 10 forms a neural network model by, for example, reading out various parameters and hyperparameters stored in the storage unit 12, and uses this neural network model as the force field model.

[0032] Next, the processor 10 executes a molecular dynamics simulation by inputting the generated first model and second model into the force field model (S112). The processor 10 executes the simulation, for example, using the second model as an initial value and controlling the second solid wall model so as to apply pressure, shear force, etc. to the molecular model using the first model.

[0033] 4 is a cross-sectional view schematically showing a second model used in a molecular dynamics simulation, the cross-sectional view being taken from the second direction of the model shown in FIG.

[0034] The area of ​​the cell is the area from the dashed line to the underside of the first solid wall, and each model defined within this cell is input into the force field model to perform a molecular dynamics simulation. Within the cell, models representing the first solid wall, the second solid wall, and the liquid are generated. The liquid is positioned so that it is sandwiched between the first and second solid walls.

[0035] When the first solid wall and the second solid wall are fixed, the processor 10 can achieve the fixed state by keeping the coordinates of the fixed atoms at the same positions as the first model while the simulation is being performed.

[0036] In this second model, the lower surface of the first solid wall is fixed to generate a first solid wall model, and the upper surface of the second solid wall is fixed to generate a second solid wall model. The processor 10 executes a molecular dynamics simulation using, for example, this state as an initial state.

[0037] The processor 10 executes the process of S108 in accordance with the above.

[0038] 5 is a cross-sectional view schematically showing the first model used in the molecular dynamics simulation, taken from the same perspective as FIG.

[0039] As in Figure 4, the area of ​​the cell extends from the dashed line to the underside of the first solid wall, and each model defined within this cell is input into the force field model to perform a molecular dynamics simulation. Models representing the first solid wall, the second solid wall, and the liquid, etc. are generated within the cell. The liquid, etc. is positioned so as to be sandwiched between the first solid wall and the second solid wall.

[0040] In the first model, the first solid wall is fixed, as in the second model. On the other hand, the second model is not fixed. Therefore, the first model can arbitrarily apply a pressure parallel to the third direction and a shear force parallel to the first direction (a direction parallel to the plane formed by the first and second directions; hereafter referred to as the first direction for simplicity) to the second solid wall. Furthermore, because the second solid wall is not fixed, it is also possible to create a model in which the second solid wall moves with each simulation step.

[0041] When applying pressure parallel to the third direction to the second solid wall, the processor 10 can define the pressure, for example, as follows: The processor 10 acquires information regarding the number N of atoms to which pressure is to be applied, the area A of contact between the second solid wall and the liquid layer, and the pressure P to be applied. In this case, the force F to be applied can be expressed as F = P × A. Then, the processor 10 can realize a predetermined pressure in the force field by applying an equivalent force F / N (= P × A / N) to the atoms to which pressure is to be applied at each step of the simulation.

[0042] When giving the second solid wall a sliding velocity parallel to the first direction, the processor 10 can define the sliding velocity, for example, as follows:

[0043] The processor 10 obtains the sliding velocity v to be applied, multiplies v by the integration time Δt per step used in the simulation, and calculates the displacement per step Δr = v × Δt. The processor 10 then sets the coordinate r of the sliding direction (e.g., the first direction) of all atoms to which a sliding velocity is to be applied per step of the simulation to r + Δr, thereby realizing the predetermined sliding velocity.

[0044] The processor 10 executes the process of S106 in accordance with the above. For example, if the number of steps is n, a first model is generated from the next step of the initial state, pressure and displacement are applied to the second solid wall to form a second solid wall model in the next step, and a first model including this second solid wall model is generated. The processor 10 generates this first model up to the nth step and ends the process of S106.

[0045] The processor 10 may generate (S108) a second model in which a second solid wall model is fixed to such a first model (as a special state of the first model) as an initial state in which pressure and shear force are not applied. As described above, the processor 10 may first generate (S108) a second model, and then generate (S106) a first model in which pressure and shear force are applied to the second solid wall model based on the second model.

[0046] In this way, the fixed state or velocity-dependent displacement state of an atom can be defined based on the position of the atom in the model defined for each time step.

[0047] The processor 10 inputs the generated first and second models into the force field model and executes a molecular dynamics simulation.

[0048] 6 and 7 are diagrams showing examples of shear stress and friction coefficient obtained by the molecular dynamics simulation in this embodiment. For example, it is assumed that the shear stress with respect to time shown in FIG. 6 is obtained by the above simulation.

[0049] The processor 10 divides this shear stress by the pressure applied to the second solid wall in the first model, thereby making it possible to obtain the change in the friction coefficient over time as shown in FIG.

[0050] FIG. 8 shows a graph comparing the friction coefficients obtained in this embodiment with experimental values. Values ​​calculated by classical molecular dynamics simulation are also shown as comparative examples. The results according to this embodiment are indicated by circles and solid lines, while the results of the comparative examples are indicated by crosses and dotted lines. Circles and crosses represent a plot of the relationship between the friction coefficients estimated by simulation and those obtained by experiment. The solid and dashed lines are straight lines that approximate the circles and crosses.

[0051] As shown in FIG. 8, the friction coefficient estimated in this embodiment is closer to the experimental value than the comparative example, and it is clear that good results have been obtained.

[0052] 9 and 10 are diagrams showing an example of a case in which no chemical reaction occurs between the liquid and the solid wall in this embodiment. For example, as shown in FIG. 9, the solid wall is formed of diamond, and a hydrogen atom (hereinafter referred to as H) is present at the end on the liquid side. In such a case, no chemical bond occurs at the interface with the liquid. For example, applying pressure and shear force from the state shown in the left diagram will cause a transition to the state shown in the right diagram, but no particular chemical reaction occurs in the simulation.

[0053] As a result, a low friction state is maintained as shown in FIG.

[0054] 11 and 12 are diagrams showing an example of a case where a chemical reaction occurs between a liquid and a solid wall in this embodiment. For example, when a liquid is present as a lubricant, a chemical reaction may occur between the solid walls via the liquid due to pressure. In FIG. 11, the solid wall has a diamond structure without an H at the end. In such a case, unlike FIG. 9, by applying pressure and shear force from the state shown in the left figure, the solid walls undergo a chemical reaction and bond together as shown in the right figure.

[0055] As a result, the coefficient of friction fluctuates significantly depending on the number of chemical bonds, crystal structure, etc., as shown in Figure 12.

[0056] In general, it is difficult to capture such a state using classical molecular dynamics simulations, and it is not possible to obtain an appropriate friction coefficient. In addition, although it is possible to infer such a state using quantum molecular dynamics simulations, it requires enormous amounts of computation time and resources, making it difficult to obtain results in real time.

[0057] As described above, according to this embodiment, simulations can be performed simply by inputting the positions of atoms, the applied pressure, and the like into a trained model and obtaining results. This allows for acquisition in a matter of minutes to hours, significantly faster than quantum mechanical simulations. Furthermore, compared to these methods, as shown in FIG. 8 , highly accurate results can be obtained because the estimated values ​​are closer to the experimental values ​​than the comparative example. This also allows for inferences that can be made even for unknown substances or unknown combinations of substances. In other words, versatility and robustness can be improved. Furthermore, simply by inputting molecular structures and the like, it is possible to automatically determine whether a chemical reaction occurs at an interface and obtain appropriate physical property values.

[0058] For example, if the molecular model is a liquid (fluid), it is possible to infer what substance is optimal as a lubricant between contacting solids by changing the liquid model to various molecules without changing the solid wall model. Also, depending on the purpose, the material of the solid wall may be changed in various ways to infer an appropriate coefficient of friction, etc.

[0059] The processor 10 can input the first and second models into the force field model to obtain the relationship between the interatomic distance and the energy for two or more types of elements and their combinations, and can also store this relationship in a database in advance. By storing the relationship in a database in this way, it is possible to realize even faster calculations using the information registered in the database.

[0060] In addition to the above, various physical properties can be obtained by using this embodiment. For example, the processor 10 may calculate adsorption energy or interaction energy based on energy obtained in a molecular dynamics simulation using a force field model formed by NNP. The processor 10 can calculate the adsorption energy between the molecules constituting the liquid sandwiched between the solid walls and the solid walls, or the interaction energy between the solid walls or between the molecules of the liquid.

[0061] The processor 10 may further acquire information such as time-dependent changes in the positions and velocities of the molecules constituting the liquid based on the energy (and force) information acquired via the NNP. The processor 10 can also acquire physical property values ​​such as the atomic distribution, diffusion coefficient, degree of orientation, viscosity, etc. of the molecular model based on the acquired time-dependent changes in the positions and velocities of the molecules constituting the liquid.

[0062] The processor 10 can obtain the adsorption energy by, for example, obtaining the energy when molecules are adsorbed to each other and the energy when molecules are not adsorbed to each other via NNP and calculating the difference between these energies. Because the system becomes stable when molecules are adsorbed to each other, the adsorption energy indicates the degree to which the system is stabilized.

[0063] The processor 10 can estimate the atomic distribution, for example, by dividing the coordinates in the third direction at predetermined intervals to generate meshes. More specifically, the processor 10 can obtain the atomic distribution by acquiring the number of atoms present in each of the generated meshes. In other words, the atomic distribution is obtained based on the position information of the atoms.

[0064] The processor 10 can also acquire the atomic distribution over time. By using this atomic distribution, it is possible to obtain information about the structure, such as the fact that when lubricant molecules contain oxygen atoms, such as alcohols and carboxylic acids, they are densely packed near the solid wall to form a kind of coating.

[0065] The processor 10 can obtain the diffusion coefficient from, for example, MSD (Mean Square Displacement). MSD indicates how much atoms are displaced over a certain period of time. This MSD can be obtained from atomic position information. By using this diffusion coefficient, it is possible to obtain information such as not only the molecular structure of oil but also the degree of fluidity depending on pressure and shear rate.

[0066] The processor 10 can obtain the degree of orientation based on, for example, atomic position information. The degree of orientation indicates, for example, information about the angle at which the lubricant molecules contact (attach) to the solid wall. For example, information about the oil film structure can be obtained, such as the fact that linear molecules such as normal alkanes are oriented relatively parallel, while branched alkanes have a random orientation, and alcohols and carboxylic acids with polar groups are evenly aligned at a predetermined angle. The processor 10 can also determine the distribution of molecules in a third direction based on this information.

[0067] The processor 10 can acquire the viscosity based on, for example, molecular velocity information. More specifically, the processor 10 can acquire the viscosity by calculating the value of the autocorrelation function of the velocity of atoms constituting the molecule in one direction (for example, the first direction or the second direction) and performing mathematical processing.

[0068] As described above, in the present disclosure, various physical property values ​​can be obtained for molecules sandwiched between solid walls by using NNP. However, the physical property values ​​listed above are only examples, and it is possible to obtain physical property values, etc., related to molecules and solid walls in similar situations, such as those affected by the interface between the molecule and the solid wall.

[0069] Although the molecular model is described as a liquid, this liquid may be a pure material composed of one or more types of molecules, or a liquid formed from a material that is a mixture of multiple materials. Furthermore, it is not limited to a liquid. For example, it may be a crystalline or amorphous solid. In these cases, it is also possible to obtain physical properties such as the coefficient of friction. Furthermore, the molecular model may be a gas or solid. In these cases, it is also possible to obtain physical properties such as the coefficient of friction.

[0070] In the above embodiment, the simulation was performed assuming that the second solid wall model moved (i.e., shear force was applied), but this is not limiting. For example, the positions of both the first solid wall model and the second solid wall model may be fixed, and the molecular model may be moved.

[0071] Although the solid wall has been formed as a flat surface, this is not limiting. As a further non-limiting example, it is also possible to appropriately estimate the coefficient of friction between a non-flat solid, such as a bearing, and a fluid by setting a circumferential, cylindrical, and / or spherical solid wall.

[0072] Furthermore, we inferred the state of the second solid wall model relative to the fixed first solid wall model via the molecular model, but by fixing both solid walls in this way and applying force to or moving the molecular model, we can infer the friction coefficient on which the liquid is moving in the area sandwiched between the solid walls and the reaction at the interface. Furthermore, instead of a solid wall model, a model with a cylindrical cavity can be used, and it can also be used to infer the state (e.g., friction coefficient) of a liquid, gas, etc. that is subjected to pressure passing through a straw-shaped object.

[0073] In the above-described embodiment, the generation of a model and the like have been described, but the advantages of using a force field model based on NNP are not limited to calculation time and accuracy. Another advantage of this method is that it can also improve usability. That is, a simulation can be performed without the user having to input detailed information. This is because NNP performs an appropriate simulation based on the positions of atoms.

[0074] FIG. 13 is a diagram showing, as a GUI (Graphical User Interface), an example of information input by a user to realize a simulation according to this embodiment. For example, the material of the solid wall and the material constituting the fluid may be set in a selectable format. The user can cause the processor 10 to execute a simulation by appropriately selecting the set material. The user can also input the thickness of the fluid.

[0075] The processor 10 may define solid walls based on the input material separated by the thickness of the input fluid, and define fluid based on the input material as a molecular model between them.

[0076] The information that can be input is simple, but using only the above information, the processor 10 can obtain the time change in the friction coefficient by changing the pressure and shear force over time. The output result is not limited to the friction coefficient, and may be any of the physical property values ​​listed above. The processor 10 may infer and output an appropriate physical property value based on the user's selection.

[0077] For example, when the shear stress changes as shown in the upper right part of the display in Fig. 13, the change in the friction coefficient over time can be obtained as shown in the lower right part. Furthermore, the information processing device 1 can also obtain and output the steady-state friction coefficient. The steady-state friction coefficient can be obtained as the average value of the friction coefficient in a time domain in which the change in the friction coefficient is within a predetermined range.

[0078] The graph in the upper right corresponds to Fig. 6 and may be automatically generated by an application or may be specified by a user. If specified by a user, the transition may be such that the user can directly input the shear stress (and / or normal pressure). Alternatively, the transition may be automatically generated by the processor 10 based on some conditions specified by the user.

[0079] Based on this change over time, the processor 10 generates a first model for each time (step) and executes a simulation. Then, the information processing device 1 may output the final result in the lower right section.

[0080] Although the GUI is shown, it is not limited to this and may be, for example, an interface for accessing a library such as an API (Application Programming Interface) or a class.

[0081] It may also be an interface that allows input of information to further limit the scope.

[0082] In this way, according to this embodiment, the input information can be simple information such as the material that constitutes the solid wall, the material that constitutes the fluid, and the thickness of the fluid. By using the NNP force field, it becomes possible to infer appropriate physical property values ​​from such simple input.

[0083] The embodiments of the present disclosure can be applied to situations where a solid comes into contact with a molecular structure such as a liquid, for example, in components that make up an automobile engine, a suspension, industrial machinery, an air conditioner, a refrigerator, etc. They can also be applied to space-related technologies and vacuum technologies.

[0084] Although the embodiments of the present disclosure have been described in detail above, the present disclosure is not limited to the individual embodiments described above. Various additions, modifications, substitutions, partial deletions, etc. are possible within the scope of the conceptual idea and spirit of the present invention derived from the content defined in the claims and their equivalents. For example, in all of the above-described embodiments, when numerical values ​​or formulas are used in the explanation, they are shown as examples and are not limited to these. Furthermore, the order of each operation in the embodiments is shown as an example and is not limited to these. [Explanation of symbols]

[0085] 1: Information processing device, 10: Processor, 12: Memory section, 14: Input / output I / F, 16: Bus

Claims

1. a memory and a processor, The memory includes: Store information about the force field model trained by machine learning; The processor: defining a molecular model representing a molecular structure of a target, the molecular model having a thickness in the third direction and extending in the first direction and the second direction, and a first solid wall model and a second solid wall model representing solid walls in a first direction, a second direction intersecting the first direction, and a third direction intersecting the first direction and the second direction; a first model in which the first solid wall model is fixed and the molecular model is sandwiched between the first solid wall model and the second solid wall model, and a molecular dynamics simulation is performed using the force field model in which a force is applied to the second solid wall model; Calculating the stress of the molecular model obtained as a result of the molecular dynamics simulation. An information processing device, The processor: A database of the relationship between interatomic distances and energies for two or more types of elements and their combinations is compiled in advance, and the molecular dynamics simulation is performed. Information processing device.

2. The force field model is a model based on NNP (Neural Network Potential). The information processing device according to claim 1.

3. The molecular dynamics simulation is performed using an interatomic potential calculated based on the NNP. The information processing device according to claim 2.

4. The molecular model is a model representing a group of molecules, a liquid, a crystal, or an amorphous solid; The information processing device according to claim 3.

5. The processor: performing a molecular dynamics simulation on a second model in which both the first solid wall model and the second solid wall model are fixed; Execute a calculation in the first model using the result obtained in the second model as an initial state.

5. The information processing device according to claim 3 or claim 4.

6. The processor: performing a molecular dynamics simulation on the first model by applying a pressure along the third direction and a shear force along a direction obtained by combining the first direction and the second direction to the second solid wall model, thereby obtaining a shear stress; 6. The information processing device according to claim 3.

7. The processor: Dividing the shear stress by the pressure to estimate the coefficient of friction; The information processing device according to claim 6.

8. The processor: calculating time evolution of positions and velocities of atoms constituting the molecular model based on the energy obtained by using the molecular dynamics simulation; calculating at least one of an atomic distribution, a diffusion coefficient, an orientation degree, or a viscosity of the molecular model based on the time evolution of the atomic positions and velocities obtained using the molecular dynamics simulation; The information processing device according to claim 7.

9. The processor: performing a molecular dynamics simulation based on the input types and regions of atoms or molecules forming the molecular model, the first solid wall model, and the second solid wall model; 9. The information processing device according to claim 1.

10. The processor: receiving information on only the molecules constituting the first solid wall model, the second solid wall model and the molecular model, and the thickness of the molecular model in a third direction; Execute a molecular dynamics simulation based on the received information.

10. The information processing device according to claim 1.

11. The processor defining a molecular model representing a group of molecules having a thickness in a first direction, a second direction intersecting the first direction, and a third direction intersecting the first direction and the second direction, and a first solid wall model and a second solid wall model representing solid walls; a first model in which the first solid wall model is fixed and the molecular model is sandwiched between the first solid wall model and the second solid wall model, and a molecular dynamics simulation is performed using a force field model trained by machine learning, in which a force is applied to the second solid wall model; Calculating the stress of the molecular model obtained as a result of the molecular dynamics simulation. An information processing method, comprising: The processor: A database of the relationship between interatomic distances and energies for two or more types of elements and their combinations is compiled in advance, and the molecular dynamics simulation is performed. Information processing methods.

12. The processor defining a molecular model representing a group of molecules having a thickness in a first direction, a second direction intersecting the first direction, and a third direction intersecting the first direction and the second direction, and a first solid wall model and a second solid wall model representing solid walls; a first model in which the first solid wall model is fixed and the molecular model is sandwiched between the first solid wall model and the second solid wall model, and a molecular dynamics simulation is performed using a force field model trained by machine learning, in which a force is applied to the second solid wall model; Calculating the stress of the molecular model obtained as a result of the molecular dynamics simulation. A program for performing an operation, the processor, A database of the relationship between interatomic distances and energies for two or more types of elements and their combinations is compiled in advance, and the molecular dynamics simulation is performed. A program that performs an action.

13. The processor defining a molecular model representing a group of molecules having a thickness in a first direction, a second direction intersecting the first direction, and a third direction intersecting the first direction and the second direction, and a first solid wall model and a second solid wall model representing solid walls; a first model in which the first solid wall model is fixed and the molecular model is sandwiched between the first solid wall model and the second solid wall model, and a molecular dynamics simulation is performed using a force field model trained by machine learning, in which a force is applied to the second solid wall model; Calculating the stress of the molecular model obtained as a result of the molecular dynamics simulation. A program for performing an operation, the processor, A database of the relationship between interatomic distances and energies for two or more types of elements and their combinations is compiled in advance, and the molecular dynamics simulation is performed. A temporary computer-readable medium that stores a program for performing operations.

Images