Quantum chemistry calculation program, quantum chemistry calculation method, and information processing device

The quantum chemical calculation method addresses the challenge of prolonged calculation time in structural optimization by using combined force and energy change rate convergence conditions, efficiently determining stable material structures.

JP2024158044APending Publication Date: 2024-11-08FUJITSU LTD

Patent Information

Application Number
JP2023072880
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-27
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The challenge in structural optimization calculations for determining a stable material structure is the prolonged calculation time due to the difficulty in converging force and energy fluctuations, leading to an increased number of DFT calculations.

Method used

A quantum chemical calculation method that uses convergence conditions involving both the magnitude of force and the rate of energy change to terminate the updating process of atomic positions, reducing the overall calculation time and ensuring accuracy.

Benefits of technology

This approach significantly reduces the amount of calculation required to find a stable structure by employing a logical sum of force and energy change rate convergence conditions, thereby shortening the calculation time without compromising accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024158044000001_ABST
    Figure 2024158044000001_ABST
Patent Text Reader

Abstract

To reduce a calculation amount for obtaining a stable structure of a substance.SOLUTION: An information processing device 10 initially sets positions of multiple atoms to position information 2 indicating the positions of the multiple atoms that constitute a substance and repeatedly updates the positions of the multiple atoms. The information processing device 10 calculates the total energy of the substance when the positions of the multiple atoms are initially set or updated. The information processing device 10 calculates a magnitude of a force acting on each of the multiple atoms when the positions of the multiple atoms are initially set or updated. The information processing device 10 calculates an energy change rate which indicates a degree of change in the total energy of the substance before and after the update when the positions of the multiple atoms are updated. The information processing device 10 terminates processing of updating the positions of the multiple atoms and acquires the position information 2 at the time of termination when at least one of a first convergence condition 3 for the magnitude of the force and a second convergence condition 4 for the energy change rate is satisfied.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a quantum chemical calculation program, a quantum chemical calculation method, and an information processing device. [Background technology]

[0002] In recent years, in the field of materials development, materials informatics (hereinafter referred to as MI) has been progressing worldwide as a data-driven research and development initiative aimed at shortening research and development time and reducing costs. Accumulation of high-quality materials data is important for MI. Therefore, efficient data accumulation through simulations such as quantum chemistry in addition to experiments is expected.

[0003] One of the main calculation methods in quantum chemistry is the density functional theory (DFT). In DFT, the electron density of the target substance is first calculated using the self-consistent field (SCF) method. Then, physical quantities that indicate the state of the substance (such as total energy) are calculated based on the obtained electron density.

[0004] In materials development, DFT is used for structural optimization to find the stable structure of molecules that govern the properties of a material. In structural optimization, DFT is repeated while gradually changing the three-dimensional structure of the molecule to search for the stable structure that minimizes the total energy.

[0005] If the number of atoms is N (N is a natural number), the computational complexity of DFT is O(N 3 ) and the larger the system to be calculated, the greater the amount of calculation required. Therefore, when using DFT to find a stable molecular structure, it is necessary to reduce the amount of calculation as much as possible.

[0006] As a technique for determining the stable structure of a substance, for example, a molecular design support system has been proposed that parallelizes the calculation of atomic coordinate differentials of total energy to determine stable molecular structures at high speed with low memory capacity. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Japanese Patent Application Publication No. 6-187404 Summary of the Invention [Problem to be solved by the invention]

[0008] In structural optimization calculations to find a stable structure of a substance, it may take a long time until the convergence conditions of the calculations are satisfied. For example, in structural optimization, for one or more atoms arranged in space or a coordinate system, the atomic arrangement is modified and the total energy calculation by DFT for that atomic arrangement is performed repeatedly until the forces (distortions) of all atoms converge. In the calculation process of such structural optimization, the total energy may transition to small fluctuations. When the total energy transitions to small fluctuations, it becomes difficult for the forces to converge for some atoms. As a result, the number of iterations of the DFT calculation increases, and the calculation time becomes longer.

[0009] In one aspect, the present invention aims to reduce the amount of calculation required to determine a stable structure of a substance. [Means for solving the problem]

[0010] In one proposal, a quantum chemistry calculation program is provided that causes a computer to execute the following processes. The computer initially sets the positions of the atoms in position information indicating the positions of the atoms constituting the substance, and further repeatedly updates the positions of the atoms in a direction that stabilizes the state of the substance. When the positions of the atoms are initially set or updated, the computer calculates the total energy of the substance based on the position information. When the positions of the atoms are initially set or updated, the computer calculates the magnitude of the force acting on each of the atoms based on the position information. When the positions of the atoms are updated, the computer calculates an energy change rate indicating the degree of change in the total energy of the substance before and after the update. The computer determines whether or not at least one of a first convergence condition for the magnitude of the force and a second convergence condition for the energy change rate is satisfied. When at least one of the first convergence condition and the second convergence condition is satisfied, the computer ends the process of updating the positions of the atoms and obtains the position information at the end point. Effect of the Invention

[0011] According to one aspect, the amount of calculation required to determine a stable structure of a substance is reduced. [Brief description of the drawings]

[0012] [Figure 1] FIG. 2 is a diagram illustrating an example of a quantum chemical calculation method according to the first embodiment. [Diagram 2] FIG. 1 illustrates an example of a system configuration. [Diagram 3] FIG. 2 illustrates an example of hardware of a control node. [Figure 4] FIG. 2 illustrates an example of a hardware configuration of a computing node. [Diagram 5] FIG. 13 is a diagram showing the relationship between the magnitude of atomic energy and the ease of convergence of the force. [Figure 6] FIG. 13 is a diagram illustrating an example of the number of iterations of DFT until a convergence condition of Force is satisfied. [Figure 7] FIG. 13 is a diagram showing an example of convergence determination of structural optimization based on the logical sum of two convergence conditions. [Figure 8]FIG. 11 is a diagram showing an example of the number of DFT iterations until two convergence conditions are satisfied. [Figure 9] FIG. 2 is a block diagram showing an example of functions realized by the computing system. [Figure 10] 13 is a flowchart showing an example of a procedure for a structural optimization process. [Figure 11] 13 is a flowchart showing an example of a procedure for a structural optimization process. [Figure 12] FIG. 13 is a diagram showing an example of speeding up a structural optimization calculation by using a convergence condition related to the rate of change of energy. [Figure 13] FIG. 1 is a diagram illustrating an example of a wide area search technology service. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, the present embodiment will be described with reference to the drawings. Note that each embodiment can be implemented in combination with a plurality of other embodiments as long as no contradiction occurs. First Embodiment The first embodiment is a quantum chemical calculation method that reduces the amount of calculation required to obtain a stable structure (ground state) of a substance.

[0014] Fig. 1 is a diagram showing an example of a quantum chemical calculation method according to the first embodiment. Fig. 1 shows an information processing device 10 for implementing the quantum chemical calculation method according to the first embodiment. The information processing device 10 can implement the quantum chemical calculation method according to the first embodiment by, for example, executing a quantum chemical calculation program.

[0015] The information processing device 10 includes a storage unit 11 and a processing unit 12. The storage unit 11 is, for example, a memory or a storage device included in the information processing device 10. The processing unit 12 is, for example, a processor or an arithmetic circuit included in the information processing device 10.

[0016] The storage unit 11 stores, for example, substance information 1 indicating a target substance for which a stable structure is to be obtained. The substance information 1 indicates, for example, the molecular structure of the substance. The processing unit 12 calculates the atomic arrangement when the substance indicated by the substance information 1 has a stable structure. A substance has a stable structure when, for example, the energy of the substance is at a minimum. Such a process for finding a stable structure of a substance is also called structural optimization. For example, the processing unit 12 performs the following process.

[0017] Processing unit 12 initially sets the positions of the atoms in position information 2 indicating the positions of the atoms constituting the substance indicated in substance information 1. Processing unit 12 further repeatedly updates the positions of the atoms in a direction that stabilizes the state of the substance.

[0018] When the positions of the multiple atoms are initially set or updated, the processing unit 12 calculates the total energy of the substance based on the position information 2. For example, the processing unit 12 calculates the electron density in the space in which the substance exists by DFT, and calculates the total energy based on the electron density.

[0019] When the positions of the multiple atoms are initially set or updated, the processing unit 12 calculates the magnitude of the force acting on each of the multiple atoms based on the position information 2. For example, the processing unit 12 calculates, for each of the multiple atoms, the sum of the potential of the Coulomb force between each of the multiple atoms and other atoms based on the distance between each of the multiple atoms. The processing unit 12 calculates the magnitude of the force acting on each atom based on the sum of the potential for each of the multiple atoms.

[0020] When the positions of multiple atoms are updated, the processing unit 12 calculates an energy change rate indicating the degree of change in the total energy of the material before and after the update. For example, the processing unit 12 calculates the difference between a first total energy of the material based on the position information after the nth update (n is a natural number) and a representative value of a second total energy of the material based on the position information after the mth update (m is a natural number equal to or less than n-1) to the n-1th update. The processing unit 12 then divides the calculated difference by the representative value of the second total energy, and the absolute value of the division result is the energy change rate.

[0021] When the magnitude of the force acting on each of the multiple atoms and the energy change rate of the substance are obtained, the processing unit 12 judges whether or not at least one of the first convergence condition 3 for the magnitude of the force and the second convergence condition 4 for the energy change rate is satisfied. This means that the convergence condition of the iterative process is the logical sum of the first convergence condition 3 and the second convergence condition 4.

[0022] For example, the processing unit 12 determines that the first convergence condition 3 is satisfied when the magnitude of the force for all of the atoms is equal to or less than a first allowable value. The processing unit 12 also determines that the second convergence condition 4 is satisfied when the rate of energy change is equal to or less than a second allowable value.

[0023] Then, when at least one of the first convergence condition 3 and the second convergence condition 4 is satisfied, the processing unit 12 ends the process of updating the positions of the multiple atoms and acquires the position information 2 at the end point. The positions of the multiple atoms set in the acquired position information 2 at this time indicate a stable structure of the substance.

[0024] In this way, when at least one of the first convergence condition 3 and the second convergence condition 4 is satisfied, the update of the atomic positions is terminated and the position information 2 at that time is obtained, thereby reducing the amount of calculation required to obtain a stable structure of the substance. In other words, even if the magnitude of the force acting on some atoms exceeds the first allowable value, if the rate of energy change is equal to or less than the second allowable value, the repeated process of updating the atomic positions, calculating the total energy, calculating the force magnitude, and calculating the rate of energy change is terminated. Therefore, convergence can be determined earlier than when the convergence is determined only based on the first convergence condition 3 for the magnitude of the force acting on the atoms, for example, and the amount of calculation is reduced.

[0025] For example, in the example in Figure 1, the first tolerance "|F|" is "1.00" and the second tolerance "RCE_TH" is "1.0×10 -5" At this time, the magnitude of the force |F2| acting on the second atom of the substance is equal to or smaller than the first tolerance after the fourth iteration. On the other hand, the magnitude of the force |F1| acting on the first atom of the substance exceeds the first tolerance even after the fifth iteration. Therefore, in the case of only the first convergence condition 3, the iterative process does not end even after the fifth iteration.

[0026] The rate of energy change is equal to or less than the second allowable value at the fifth iteration. Therefore, by also performing a convergence check under the second convergence condition 4, the iterative process ends at the fifth iteration. As a result, the amount of calculation required to identify a stable structure of a substance is reduced.

[0027] It should be noted that a case where it takes a long time for the force acting on any atom to satisfy the first convergence condition 3 may occur, for example, when the energy of some atoms is small. When the energy of an atom is small, the magnitude of the force acting on that atom repeatedly increases and decreases each time the position of the atom is updated, and it may take a long time for the first convergence condition 3 to be satisfied.

[0028] When the magnitude of the force acting on an atom repeatedly increases and decreases, when the magnitude of the force increases, the amount of movement when updating the position of that atom also increases. However, even if the amount of movement of an atom with relatively low energy is somewhat large, the impact on the total energy is small. In other words, even if the magnitude of the force acting on an atom with relatively low energy exceeds the first allowable value, if the rate of energy change is small enough to be equal to or less than the second allowable value, the effect on the total energy of the increase or decrease in the magnitude of the force acting on that atom is negligible. In such a case, it can be determined that the total energy is almost at its minimum value.

[0029] The position of the atom where the total energy is the minimum value indicates the stable structure of the substance. That is, the processing unit 12 judges the convergence of the iterative process by the logical sum of the first convergence condition 3 and the second convergence condition 4, thereby making it possible to reduce the amount of calculation without reducing the accuracy of the finally obtained stable structure of the substance.

[0030] Furthermore, when the rate of energy change used in the second convergence condition 4 is equal to or less than the second allowable value, it is ensured that the value of the total energy is also sufficiently reduced. Therefore, by using the second convergence condition 4 for the rate of energy change, when a user judges the usefulness of a material having a structure obtained by structural optimization based on the total energy of the material, the user can make the judgment appropriately.

[0031] Second Embodiment The second embodiment is a calculation system that reduces the number of DFT iterations when the calculation of the molecular structure of a substance by DFT is performed by HPC (High-performance Computing).

[0032] 2 is a diagram showing an example of a system configuration. A computing system 30 that realizes HPC has a control node 100 and multiple computing nodes 200a, 200b,... The multiple computing nodes 200a, 200b,... are connected to each other via an interconnect 32 that enables high-speed communication. The interconnect 32 is a network that interconnects processors with each other, for example, by a six-dimensional mesh / torus.

[0033] Each of the calculation nodes 200a, 200b,... is also connected to a control network 31. A control node 100 is further connected to the control network 31. The control node 100 is a computer that issues job execution instructions to the calculation nodes 200a, 200b,...

[0034] The control node 100 is connected to a terminal 40 via a network 20. The terminal 40 is a computer that registers information on jobs to be executed by the computation nodes 200a, 200b, ... in the control node 100 based on a user's operation. The control node 100 instructs the computation nodes 200a, 200b, ... to perform computations using DFT according to instructions from the terminal 40.

[0035] For example, a user can obtain physical quantities such as the total energy of an electronic system defined by DFT by having the calculation nodes 200a, 200b, . . . perform calculations of the state of matter using DFT.

[0036] 3 is a diagram showing an example of hardware of a control node. The entire control node 100 is controlled by a processor 101. A memory 102 and a plurality of peripheral devices are connected to the processor 101 via a bus 109. The processor 101 may be a multiprocessor. The processor 101 is, for example, a central processing unit (CPU), a micro processing unit (MPU), or a digital signal processor (DSP). At least a part of the functions realized by the processor 101 executing a program may be realized by an electronic circuit such as an application specific integrated circuit (ASIC) or a programmable logic device (PLD).

[0037] The memory 102 is used as a main storage device of the control node 100. The memory 102 temporarily stores at least a part of the OS (Operating System) program and application programs to be executed by the processor 101. The memory 102 also stores various data used for processing by the processor 101. As the memory 102, for example, a volatile semiconductor storage device such as a RAM (Random Access Memory) is used.

[0038] The peripheral devices connected to the bus 109 include a storage device 103, a GPU (Graphics Processing Unit) 104, an input interface 105, an optical drive device 106, a device connection interface 107, and network interfaces 108a and 108b.

[0039] The storage device 103 electrically or magnetically writes and reads data to and from a built-in recording medium. The storage device 103 is used as an auxiliary storage device for the control node 100. The storage device 103 stores an OS program, application programs, and various data. Note that, for example, a hard disk drive (HDD) or a solid state drive (SSD) can be used as the storage device 103.

[0040] The GPU 104 is an arithmetic device that performs image processing, and is also called a graphics controller. The GPU 104 is connected to a monitor 21. The GPU 104 displays an image on the screen of the monitor 21 in accordance with an instruction from the processor 101. The monitor 21 may be a display device using an organic EL (Electro Luminescence) display device, a liquid crystal display device, or the like.

[0041] The input interface 105 is connected to the keyboard 22 and the mouse 23. The input interface 105 transmits signals sent from the keyboard 22 and the mouse 23 to the processor 101. Note that the mouse 23 is an example of a pointing device, and other pointing devices can also be used. Examples of other pointing devices include a touch panel, a tablet, a touch pad, and a trackball.

[0042] The optical drive device 106 uses a laser beam or the like to read data recorded on the optical disc 24 or write data to the optical disc 24. The optical disc 24 is a portable recording medium on which data is recorded so that it can be read by reflection of light. The optical disc 24 includes a DVD (Digital Versatile Disc), a DVD-RAM, a CD-ROM (Compact Disc Read Only Memory), a CD-R (Recordable) / RW (ReWritable), etc.

[0043] The device connection interface 107 is a communication interface for connecting peripheral devices to the control node 100. For example, a memory device 25 or a memory reader / writer 26 can be connected to the device connection interface 107. The memory device 25 is a recording medium equipped with a communication function with the device connection interface 107. The memory reader / writer 26 is a device that writes data to the memory card 27 or reads data from the memory card 27. The memory card 27 is a card-type recording medium.

[0044] The network interface 108a is connected to the network 20. The network interface 108a transmits and receives data to and from other computers or communication devices, such as the terminal 40, via the network 20. The network interface 108b is connected to the control network 31. The network interface 108b transmits and receives data to and from the computation nodes 200a, 200b, ... via the control network 31.

[0045] The network interfaces 108a and 108b are wired communication interfaces that are connected by cables to wired communication devices such as switches and routers, etc. Alternatively, the network interfaces 108a and 108b may be wireless communication interfaces that are communicatively connected by radio waves to wireless communication devices such as base stations and access points.

[0046] The control node 100 executes a program recorded in, for example, a computer-readable recording medium to realize the processing function of the second embodiment. The program describing the processing contents to be executed by the control node 100 can be recorded in various recording media. For example, the program to be executed by the control node 100 can be stored in the storage device 103. The processor 101 loads at least a part of the program in the storage device 103 into the memory 102 and executes the program. The program to be executed by the control node 100 can also be recorded in a portable recording medium such as the optical disk 24, the memory device 25, or the memory card 27. The program stored in the portable recording medium becomes executable after being installed in the storage device 103 under the control of, for example, the processor 101. The processor 101 can also directly read and execute the program from the portable recording medium.

[0047] 4 is a diagram showing an example of a hardware configuration of a computing node. The computing node 200a has a CPU / memory unit 201 and a router 202. The CPU / memory unit 201 and the router 202 are connected to each other via a plurality of communication interfaces (NICs) 203. The CPU / memory unit 201 is also connected to a NIC 204 for connection to a control network 31.

[0048] The CPU / memory unit 201 includes a CPU 201a having multiple cores and a memory 201b. A process (a processing execution unit) is generated for each core in the CPU / memory unit 201. When the process for each core of the CPU / memory unit 201 performs synchronous processing with a process in another computing node, the process communicates with the other computing nodes 200b, ... via a router 202.

[0049] The router 202 communicates with adjacent computation nodes in each of the three-dimensional directions, for example. The router 202 transmits data to be transmitted from the CPU / memory unit 201 to another computation node to the adjacent computation node in a direction according to the position of the computation node in the interconnect 32. When the router 202 receives data for a process in the CPU / memory unit 201 from an adjacent computation node, it transmits the data to the CPU / memory unit 201. Furthermore, when the data received from the adjacent computation node is destined for another computation node, the router 202 transmits the data to the adjacent computation node in a direction according to the position of the destination computation node in the interconnect 32.

[0050] The other computation nodes 200b,... have the same hardware configuration as the computation node 200a. The computation nodes 200a, 200b,... having the hardware as shown in Fig. 4 are connected in a mesh / torus manner in three dimensions to form a three-dimensional mesh / torus interconnect 32.

[0051] The processing function of the second embodiment can be realized by the above-described hardware control node 100 and the computation nodes 200a, 200b, etc. The information processing device 10 shown in the first embodiment can also be realized by the same hardware as the control node 100 or the computation node 200a.

[0052] The calculation system 30 calculates stable structures of molecules that constitute a substance by structural optimization using DFT. Here, the reason why the amount of calculation increases in structural optimization using DFT will be described.

[0053] In geometry optimization, the atomic positions are updated and DFT calculations are performed repeatedly with the updated atomic positions until the forces of all atoms arranged in space or the coordinate system converge. Force is the magnitude of the force (e.g. Coulomb force) acting on each atom. If the forces for all atoms are sufficiently small, the structure of the molecule at that time is stable. Such convergence criteria can be expressed in the following formula:

[0054] Convergence Criterion: |F i |≦|F| (1) i is the identification number of the atom in the molecule. |F i | is the magnitude of the force of the i-th atom. |F| is the tolerance of the force. The total energy of the molecule when all atoms satisfy the convergence criterion is the total energy of the stable structure.

[0055] In this case, if any of the following conditions are met, the total energy transitions to small fluctuations. The first condition is that there exists an atom that has a small effect on the total energy (an atom with low energy) and that this atom is not yet converged. An atom with low energy is, for example, an atom that only has electrons in the orbit (electron shell) closer to the nucleus. Even if an atom has an electron in an outer electron shell, if there is another atom with which it shares the electron, it will have a stable structure and become an atom with low energy.

[0056] The second condition is that small-energy atomic vibrations (periodic atomic displacements due to atomic forces) are occurring. For atoms that satisfy any of these conditions, the total energy may transition to small vibrations. When the total energy transitions to small fluctuations, atoms whose Force does not satisfy the convergence criteria above remain indefinitely. As a result, the number of iterative DFT calculations in the structure optimization increases, and the calculation time becomes longer.

[0057] 5 is a diagram showing the relationship between the magnitude of atomic energy and the ease of convergence of Force. In the graph 50 shown in FIG. 5, the horizontal axis represents the number of DFT iterations, and the vertical axis represents the Force|F of the atom. iThe solid curve in graph 50 represents the change in force with increasing number of DFT iterations for low energy atoms (e.g., lowest energy atoms). The dashed curve in graph 50 represents the change in force with increasing number of DFT iterations for high energy atoms (e.g., highest energy atoms).

[0058] Graph 50 is an example where only the convergence condition related to force is used as the convergence condition. For atoms with high energy, when DFT is repeated, the force falls below the allowable value at an early stage. In contrast, for atoms with low energy, the force decreases smoothly up to a certain point as the number of DFT iterations increases, but after that, even if the number of DFT iterations increases, the force decreases slowly, fluctuating repeatedly. Therefore, for atoms with low energy, the number of DFT iterations until the force meets the convergence criterion is greater than for atoms with high energy.

[0059] 6 is a diagram showing an example of the number of times DFT is repeated until the convergence condition of Force is satisfied. Force calculation result table 51 shows the calculation results of Force of atoms obtained in each DFT repetition for atoms with low and high energy. The allowable value of Force is "1.00".

[0060] In the Force Calculation Result Table 51, the Force of the atom with high energy is "1.00" below the allowable value after the 4th DFT calculation. On the other hand, the Force of the atom with low energy is "1.00" below the allowable value after the 7th DFT calculation.

[0061] The convergence condition for the structure optimization is that the Force satisfies the convergence criteria for all atoms. If the Force calculation results are as shown in Force Calculation Results Table 51, the DFT calculation will be repeated seven times until the structure optimization process converges.

[0062] In this way, since the ease of convergence of the force varies greatly from atom to atom, the number of DFT iterations required to complete convergence for all atoms increases. Therefore, the calculation system 30 according to the second embodiment determines whether or not the structural optimization has converged by the logical sum of a convergence condition for Force (first convergence condition) and a convergence condition for the degree of change in the total energy (rate of change of energy (RCE)) (second convergence condition).

[0063] The first convergence condition is that the forces for all atoms satisfy the convergence criterion (are less than or equal to the force tolerance "|F|"). The second convergence condition is that the rate of energy change satisfies the convergence criterion (is equal to or less than the allowable value "RCE_TH" of the rate of energy change).

[0064] The energy change rate is a value that represents the degree of change in the current total energy compared to the previous total energy. The energy change rate is given by, for example, the following formula: Rate of change of energy (RCE) = |(current value - previous value) ÷ previous value| (2) The current value is the total energy value obtained from the most recent DFT result. The previous value is the total energy value obtained from the DFT result immediately before the most recent DFT. Note that instead of the previous value, a representative value (e.g., average value) of the total energy obtained from the results of the past several DFTs can be used. For example, if the most recent DFT is the nth DFT, the past several DFTs are the mth (m≦n)th to n-1th DFTs.

[0065] If either one of these two convergence conditions is satisfied, the calculation system 30 determines that the calculation results of the structural optimization have converged. FIG. 7 is a diagram showing an example of convergence determination of structure optimization by the logical sum of two convergence conditions. Graph 52 shows the determination result of the first convergence condition. The horizontal axis of graph 52 is the number of DFT iterations, and the vertical axis is the atomic force "|F iGraph 52 shows the change in force for atoms with low energy as the number of DFT iterations increases. The horizontal line in graph 52 indicates the allowable force value, "|F|".

[0066] Graph 53 shows the result of the determination of the second convergence condition. The horizontal axis of graph 53 is the number of DFT iterations, and the vertical axis is the rate of energy change "RCE." Graph 53 shows a curve that indicates the change in the total energy of the molecule with an increase in the number of DFT iterations. The horizontal line in graph 53 indicates the allowable value "RCE_TH" of the rate of energy change.

[0067] As can be seen by comparing graphs 52 and 53, the energy change rate meets the convergence criterion faster than the atomic force meets the convergence criterion. In other words, by using not only the first convergence condition but also the second convergence condition to judge the convergence of the structure optimization, the number of DFT iterations required until convergence is reduced. As a result, the calculation time for the structure optimization is also shortened.

[0068] In the case shown in FIG. 7, even if the atomic position updating and DFT calculation are repeated until the first convergence condition is satisfied, the structure of the substance (atomic arrangement) does not change significantly from the point when the second convergence condition is satisfied.

[0069] It should be noted that the atomic arrangement obtained by structural optimization may be a locally optimal solution, which is a solution in which the total energy is minimized in a local space, and is not a solution in which the total energy in the entire search space is minimized (globally optimal solution).

[0070] In an actual structural optimization, the user checks whether the atomic arrangement when the convergence condition is satisfied is a local optimum solution. For example, if the total energy is higher than that of a similar structural optimization performed in the past, it is considered to be a local optimum solution. If it is a local optimum solution, the user, for example, changes the initial positions of the atoms and causes the calculation system 30 to execute the structural optimization again.

[0071] Since a local optimum may result in this way, even if the structural optimization is continued until the first convergence condition is satisfied when the second convergence condition is satisfied, there is a possibility that the accuracy of the atomic positions resulting in the local optimum solution will merely be improved. If the structural optimization process is terminated when the second convergence condition is satisfied and a process for checking whether or not a local optimum solution has been reached is carried out, recalculation of the structural optimization when a local optimum solution has been reached can be performed early, and the work of searching for a stable structure of a substance will become more efficient.

[0072] 8 is a diagram showing an example of the number of times that the DFT is repeated until two convergence conditions are satisfied. The Force calculation result table 54 shows the calculation results of the Force of atoms obtained in each DFT repetition for atoms with low and high energy. Here, the allowable value of Force is assumed to be "1.00".

[0073] In the Force Calculation Result Table 54, the Force of the atom with high energy is "1.00" which is below the allowable value after the 4th DFT calculation. On the other hand, the Force of the atom with low energy is "1.00" which is below the allowable value after the 7th DFT calculation.

[0074] The RCE calculation result table 55 shows the calculation results of the energy change rate "RCE" obtained in each DFT iteration. Here, the total energy allowable value "RCE_TH" is 1.0×10 -5 It is assumed that.

[0075] In the RCE calculation result table 55, the energy change rate obtained by DFT becomes smaller with each increase in the number of iterations, and by the fifth DFT calculation, the energy change rate is below the allowable value.

[0076] In the example of Fig. 8, the first convergence condition for Force is satisfied in the seventh DFT calculation, but the second convergence condition for the rate of energy change is satisfied in the fifth DFT calculation. By taking the logical sum of the first and second convergence conditions, it is determined that the calculation for structure optimization has converged in the fifth DFT calculation. In other words, the amount of calculation required to obtain a stable molecular structure is reduced.

[0077] Even if there are atoms remaining whose Force exceeds the allowable value, if the rate of energy change is below the allowable value, it is believed that the total energy will not decrease significantly even if the DFT calculation is repeated further. Therefore, even if a second convergence condition is added as a convergence condition for the structure optimization, the total energy finally obtained will be close to the minimum value. In other words, it is possible to identify the stable structure of the molecule with a sufficiently high degree of accuracy.

[0078] 9 is a block diagram showing an example of functions realized by the computation system. The computation system 30 has a storage unit 110, a computation command unit 120, and a structure optimization unit 210. The storage unit 110 and the computation command unit 120 are functions possessed by, for example, the control node 100. The structure optimization unit 210 is a function possessed by, for example, the computation nodes 200a, 200b, ...

[0079] The storage unit 110 stores information about the substance to be calculated (such as the atoms contained therein). When the calculation instruction unit 120 receives an instruction to analyze the stable structure of a predetermined substance from the terminal 40, it instructs the structure optimization unit 210 to execute a structure optimization process for the substance.

[0080] The structure optimization unit 210 uses DFT to calculate the positions of atoms that make the molecule have a stable structure. For example, the structure optimization unit 210 arranges atoms contained in the molecule in the space to be analyzed. The structure optimization unit 210 calculates the electron density in the space when the atoms are at the arranged positions by DFT. In calculating the electron density, the structure optimization unit 210 repeats SCF calculations until the total energy in the current atomic arrangement is minimized. When the structure optimization unit 210 obtains the minimum value of the total energy in the current atomic arrangement, it moves the positions of the atoms so that the energy decreases. The structure optimization unit 210 repeats such DFT calculations and updating of the positions of the atoms. In this way, the atomic arrangement that minimizes the total energy is obtained. The atomic arrangement that minimizes the total energy is a stable structure of the molecule.

[0081] 9 can be realized by, for example, having a computer execute a program module corresponding to the element. For example, the control node 100 executes a program corresponding to the calculation instruction unit 120 to realize the calculation instruction unit 120. The calculation instruction unit 120 realized by the control node 100 instructs one or more of the multiple calculation nodes 200a, 200b, ... to execute a job corresponding to the structure optimization unit 210. The instructed calculation node realizes the structure optimization unit 210 by executing a program corresponding to the specified job.

[0082] Next, the procedure of the structural optimization process will be described. Fig. 10 is a flow chart showing an example of a procedure for a structural optimization process. The process shown in Fig. 10 will be described below in order of step number.

[0083] [Step S101] The calculation instruction unit 120 sets a convergence criterion to be applied to structural optimization. For example, a user inputs an allowable value of Force "|F|" and an allowable value of the rate of energy change "RCE_TH" to the terminal 40. The terminal 40 transmits the input allowable value of Force "|F|" and allowable value of the rate of energy change "RCE_TH" to the control node 100. The calculation instruction unit 120 sets the transmitted allowable value of Force "|F|" and allowable value of the rate of energy change "RCE_TH" as parameters indicating the convergence criterion in the structural optimization processing job.

[0084] [Step S102] The calculation instruction unit 120 instructs the structural optimization unit 210 to execute a structural optimization job. The structural optimization unit 210 then executes the structural optimization process. [Step S103] The calculation instruction unit 120 outputs the calculation results of the structure optimization. For example, the calculation instruction unit 120 transmits the analysis results including the coordinates indicating the arrangement of atoms in the stable structure, the total energy in the stable structure, etc. to the terminal 40. The terminal 40 displays the analysis results on, for example, a monitor.

[0085] Next, the procedure of the structural optimization process will be described in detail. Fig. 11 is a flow chart showing an example of a procedure for a structural optimization process. The process shown in Fig. 11 will be described below in order of step numbers.

[0086] [Step S111] The structure optimization unit 210 initializes the atomic positions. For example, the structure optimization unit 210 sets pre-specified initial values ​​to the coordinates indicating the position of each atom included in the molecule.

[0087] [Step S112] The structure optimization unit 210 calculates the electron density in the space to be analyzed in the current atomic arrangement by DFT. For example, the SCF method is used to calculate the electron density.

[0088] [Step S113] The structure optimization unit 210 calculates the total energy of the molecule based on the electron density for each coordinate in space. [Step S114] The structure optimization unit 210 calculates the rate of energy change "RCE" in the second or subsequent DFT calculations.

[0089] [Step S115] The structure optimization unit 210 calculates the Force “|F i Calculate the force of each atom, |F i "|" can be calculated based on the sum of the potentials of interactions with all other atoms. For example, the structure optimization unit 210 calculates the gradient of the potential at the position of each atom. The calculated gradient represents the direction and strength of the force acting on the atom.

[0090] [Step S116] The structure optimization unit 210 calculates the atomic force "|F i Determine whether or not | is equal to or less than the tolerance |F| (|F i |≦|F|). The structure optimization unit 210 performs the optimization for all atoms by using the formula "|F i If |≦|F|" is satisfied, the structure optimization process is terminated. In addition, the structure optimization unit 210 performs a process of i If |≦|F|'' is not satisfied, the process proceeds to step S117.

[0091] [Step S117] The structural optimization unit 210 judges whether the rate of energy change "RCE" is equal to or less than the allowable value "RCE_TH" of the rate of energy change (RCE≦RCE_TH). If "RCE≦RCE_TH" is satisfied, the structural optimization unit 210 ends the structural optimization process. If "RCE≦RCE_TH" is not satisfied, the structural optimization unit 210 advances the process to step S118. Note that the structural optimization unit 210 may execute the processes of steps S116 and S117 in the reverse order.

[0092] [Step S118] The structure optimization unit 210 updates the positions of the atoms so that the total energy decreases. For example, the structure optimization unit 210 moves each atom in the opposite direction of the force acting on the atom by an amount according to the magnitude of the force. The structure optimization unit 210 can also determine the direction and amount of movement of the atom using the quasi-Newton method or the conjugate gradient method. After that, the structure optimization unit 210 advances the process to step S112.

[0093] In this way, when either the first convergence condition regarding Force or the second convergence condition regarding the rate of change of energy is satisfied, it is determined that the structural optimization has converged, and the structural optimization process ends.

[0094] FIG. 12 is a diagram showing an example of speeding up structural optimization calculations by concurrently using a convergence condition related to the rate of energy change. FIG. 12 shows a calculation result 61 when a stable structure of an ammonia catalyst is calculated by DFT using the CP2K program. The convergence conditions for structural optimization are set in two patterns: one where only the first convergence condition related to force is applied, and the other where the logical sum of the first convergence condition related to force and the second convergence condition related to the rate of energy change is applied. The allowable value for the rate of energy change is "1.0×10 -5 "

[0095] When only the first convergence criterion is applied to the convergence criteria, the total energy is -372.06966350 (eV), the number of DFT iterations is 55, and the time to convergence is 633.807 (s). When both the first and second convergence criterion are applied to the convergence criteria, the total energy is -372.04732301 (eV), the number of DFT iterations is 36, the time to convergence is 470.829 (s), and the speedup rate is 1.35.

[0096] Graph 62 shows a speed-up factor of "1" when only the first convergence condition for force is applied, and a speed-up factor of "1.35" when the logical sum of the first convergence condition for force and the second convergence condition for the energy change rate is applied.

[0097] In this way, the calculation speed is increased by using the logical sum of the first convergence condition related to Force and the second convergence condition related to the rate of change of energy as the convergence condition for structural optimization. Moreover, as shown in the calculation result 61, there is no significant difference in the obtained total energy. In other words, the calculation time is reduced without reducing the calculation accuracy.

[0098] When the logical sum of the first and second convergence conditions is applied, the total energy at the time when the structural optimization converges is about 0.2 (eV) higher than when only the first convergence condition is applied. The suitability of a substance as an ammonia catalyst is judged by its adsorption energy, but the quality of the adsorption energy is generally compared in units of 0.5 (eV). Therefore, an error of about 0.2 (eV) in the total energy of the structure obtained by structural optimization is an acceptable error in the search for an ammonia catalyst.

[0099] By using the technology shown in the second embodiment, it is possible to efficiently implement a wide-area search technology service for new materials that combines, for example, HPC and AI (Artificial Intelligence).

[0100] 13 is a diagram showing an example of a wide-area search technology service. For example, the calculation system 30 acquires input data 91 indicating initial conditions for a substance, such as a material to be searched, from a user. The calculation system 30 executes a quantum chemical simulation at high speed based on the input data 91. As a result, high-quality data 92, 94 on the structure of the substance can be easily generated.

[0101] The data 92 generated by the calculation system 30 can be analyzed for characteristics, for example, by AI simulation using technology such as GNN (Graph Neural Network). By using the data 93 generated by the AI ​​simulation and the data 94 generated by the calculation system 30, it is possible to analyze causal relationships from a large number of combinations of candidate materials and the physical properties of those materials, which cannot be analyzed by humans, for example, using a causal discovery AI. As a result, it becomes possible to make new discoveries and discoveries.

[0102] The results of the causal relationship analysis can be used to narrow the search range by, for example, reflecting them in the search conditions of a quantum chemical simulation. By appropriately narrowing the search range, for example, a material with the properties desired by the user can be found quickly.

[0103] Other embodiments In the second embodiment, an example has been shown in which the disclosed technology is applied to a wide-area search service for new materials using AI, but it can also be applied to other information processing services.

[0104] In the second embodiment, the DFT calculation is performed by HPC. However, if the number of atoms in the substance to be analyzed is small, the structural optimization shown in the second embodiment can also be performed by a computer other than HPC.

[0105] Although the embodiments have been described above, the configurations of the parts shown in the embodiments can be replaced with other parts having similar functions. Any other components or steps may be added. Furthermore, any two or more configurations (features) of the above-mentioned embodiments may be combined. [Explanation of symbols]

[0106] 1 Substance information 2 Location information 3 First convergence condition 4 Second convergence condition 10. Information processing device 11 Storage section 12 Processing section

Claims

1. initializing the positions of the plurality of atoms in position information indicating the positions of the plurality of atoms constituting the substance, and repeatedly updating the positions of the plurality of atoms in a direction toward stabilizing the state of the substance; Once the positions of the plurality of atoms are initially set or updated, calculating a total energy of the material based on the position information; When the positions of the plurality of atoms are initially set or updated, a magnitude of a force acting on each of the plurality of atoms is calculated based on the position information; When the positions of the plurality of atoms are updated, an energy change rate is calculated that indicates a degree of change in the total energy of the material before and after the update; determining whether or not at least one of a first convergence condition for the magnitude of the force and a second convergence condition for the rate of change of energy is satisfied; when at least one of the first convergence condition and the second convergence condition is satisfied, the process of updating the positions of the plurality of atoms is terminated, and the position information at the end point is obtained. A quantum chemistry calculation program that causes a computer to carry out the processing.

2. In the process of determining whether or not at least one of the first convergence condition and the second convergence condition is satisfied, it is determined that the first convergence condition is satisfied when the magnitude of the force for all of the plurality of atoms is equal to or less than a first allowable value. The quantum chemical calculation program according to claim 1.

3. In the process of determining whether or not at least one of the first convergence condition and the second convergence condition is satisfied, it is determined that the second convergence condition is satisfied when the energy change rate is equal to or less than a second allowable value. The quantum chemical calculation program according to claim 1.

4. In the process of calculating the energy change rate, a difference between a first total energy of the material based on the position information after an n-th update (n is a natural number) and a representative value of a second total energy of the material based on the position information after an m-th update (m is a natural number equal to or less than n-1) to an n-1-th update is divided by the representative value of the second total energy, and an absolute value of the division result is set as the energy change rate. The quantum chemical calculation program according to any one of claims 1 to 3.

5. initializing the positions of the plurality of atoms in position information indicating the positions of the plurality of atoms constituting the substance, and repeatedly updating the positions of the plurality of atoms in a direction toward stabilizing the state of the substance; Once the positions of the plurality of atoms are initially set or updated, calculating a total energy of the material based on the position information; When the positions of the plurality of atoms are initially set or updated, a magnitude of a force acting on each of the plurality of atoms is calculated based on the position information; When the positions of the plurality of atoms are updated, an energy change rate is calculated that indicates a degree of change in the total energy of the material before and after the update; determining whether or not at least one of a first convergence condition for the magnitude of the force and a second convergence condition for the rate of change of energy is satisfied; when at least one of the first convergence condition and the second convergence condition is satisfied, the process of updating the positions of the plurality of atoms is terminated, and the position information at the end point is obtained. A quantum chemical calculation method in which processing is performed by a computer.

6. a processing unit which initially sets positions of a plurality of atoms in position information indicating positions of the plurality of atoms constituting a substance, and further repeatedly updates the positions of the plurality of atoms in a direction in which a state of the substance becomes stable, and when the positions of the plurality of atoms are initially set or updated, calculates a total energy of the substance based on the position information, and when the positions of the plurality of atoms are initially set or updated, calculates a magnitude of a force acting on each of the plurality of atoms based on the position information, and when the positions of the plurality of atoms are updated, calculates an energy change rate indicating a degree of change in the total energy of the substance before and after the update, and determines whether or not at least one of a first convergence condition for the magnitude of the force and a second convergence condition for the energy change rate is satisfied, and when at least one of the first convergence condition and the second convergence condition is satisfied, ends the process of updating the positions of the plurality of atoms and acquires the position information at the end point; An information processing device having the above configuration.

Citation Information

Patent Citations

  • Method and device for molecule design support

    JP1994187404A

Cited By

  • Molecular energy calculation program, molecular energy calculation device, and molecular energy calculation method

    JP2026144891A