A method and system for calculating the steady-state electron distribution of metallic nanostructures

By employing the discontinuous Galerkin method and the Newton-Krylov iterative framework, the multi-scale computational challenge of electron density distribution in nanoscale metallic structures was solved, achieving an efficient and robust solution for electron density distribution, applicable to complex metallic nanostructures and large-scale three-dimensional problems.

CN123120141APending Publication Date: 2026-11-17HUAZHONG UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611315355.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-27
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing techniques for calculating electron density distribution in nanoscale metallic structures suffer from problems such as large computational load, poor convergence, and sensitivity to initial conditions, making it difficult to solve multiscale electron density distribution quickly and robustly.

Method used

The discontinuous Galerkin method is used for mesh discretization. Combined with the fully coupled Newton-Krylov iterative solution framework, the Schrödinger-Poisson equations are calculated independently element by element through LGMRES iteration with block diagonal preconditions, avoiding global matrix construction and supporting unstructured meshes and local refinement.

Benefits of technology

It reduces computational resource consumption, improves computational speed and accuracy, adapts to complex metallic nanostructures and large-scale three-dimensional problems, and supports high-degree-of-freedom nanoscale electronic simulation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120141A_ABST
    Figure CN123120141A_ABST
Patent Text Reader

Abstract

This invention belongs to, but is not limited to, the field of micro-nano photonics technology, and particularly relates to a method and system for calculating the steady-state electron distribution of a metallic nanostructure. The method includes: inputting the geometry, material parameters, boundary conditions, and computational domain settings of the metallic nanostructure; providing initial values ​​based on region labels; discretizing the physical quantities to be solved using polynomial basis functions on an unstructured mesh; discretizing the steady-state Schrödinger-Poisson system using the discontinuous Galerkin method; assembling fully coupled residuals; solving the equations using the Newton-Krylov method; solving the linear correction equations within each Newton step using LGMRES with block diagonal preconditions; finally, determining convergence using the residual and particle number conservation criteria, and verifying the results through the spatial distribution of the physical field.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to, but is not limited to, the field of micro-nano photonics technology, and particularly relates to a method and system for calculating the steady-state electron distribution of metal nanostructures. Background Technology

[0002] The electron density distribution in metallic nanostructures strongly influences their optical response characteristics. The spatial distribution of steady-state electrons forms the basis for subsequent dynamic simulations. Accurately describing the steady-state electron density is crucial for accurately characterizing related non-classical optical processes and effects, including plasmon resonance, optical field localization, optical field enhancement, and light-matter interactions.

[0003] At the deep nanoscale (tens of nanometers and below), the spatial distribution of steady-state electron density is dominated by quantum effects. Theoretically, this density distribution can be obtained by minimizing the energy functional of the system, which is established by the adopted theoretical model. For example, the solution scheme based on the Thomas-Fermi-von Weizsacker (TFvW) model has been well-established. Typically, this energy functional minimization problem can be transformed into solving the steady-state Schrödinger-Poisson equations, and a self-consistent iterative solution to the equations yields the steady-state electron density distribution. It should be noted that at the metallic boundaries and fine structures of nanostructures, the electron density varies drastically within the angstrom scale, exhibiting multi-scale characteristics, which places high demands on numerical discretization solutions. Currently, traditional methods such as the finite difference method or the finite element method have significant limitations in handling such problems, such as high computational cost, poor convergence, or the need for complex techniques to ensure fast convergence. How to construct an appropriate numerical discretization scheme to quickly solve for the electron density distribution of metallic nanostructures is a current challenge that needs to be addressed. The discontinuous Galerkin method has flexible local representation and mesh refinement capabilities, and is theoretically suitable for multi-scale problems, but there is still a lack of systematic solutions for its application and implementation in Schrödinger-Poisson systems.

[0004] Traditional solutions have the following limitations:

[0005] (1) The number of spatially discretized grids is large, resulting in a large amount of computation. The finite difference method relies on structured grids, and discretizing the fine regions of nanostructures will generate a large number of redundant grids. Although the finite element method can use unstructured grids to adapt to complex geometries, it usually requires a gradual transition grid partitioning strategy in the fine regions of nanostructures, which will also introduce additional grid overhead. A large number of grids will increase the computational cost of numerical solution of the model.

[0006] (2) Solving the global matrix requires a large amount of memory and computational overhead, resulting in high computational complexity. The finite element method relies on generating a large-scale global matrix. Although it has sparsity properties, the memory consumption and computation time for matrix solving increase rapidly with the number of degrees of freedom. On the other hand, in multi-scale problems, drastic changes in mesh size across scales can degrade the properties of the matrix, making it difficult to solve quickly and robustly. These two factors together make the finite element method unsuitable for solving multi-scale spatial problems and for extending to large-scale three-dimensional problems.

[0007] (3) Sensitive to initial value, and lacks robustness in computational convergence. Under finite element discretization, the convergence of the traditional Newton iteration method is highly dependent on the choice of initial value, making it difficult for the iteration to converge robustly in the absence of a high-precision reference solution.

[0008] Based on the above analysis, the urgent technical problem to be solved by the existing technology is: to develop an efficient and robust framework for solving the steady-state electron density of metal nanostructures, which is a challenge for multi-scale calculation caused by the drastic changes in electron density at the angstrom level in nanoscale metal structures. Summary of the Invention

[0009] To address the problems of existing technologies, this invention provides a method and system for calculating the steady-state electron distribution of metallic nanostructures. Based on the Schrödinger-Poisson equations and using unstructured mesh discretization, it employs the discontinuous Galerkin (DG) method to establish a fully explicit, low-cost iterative solution algorithm without a global matrix. This invention uses a fully coupled Newton-Krylov iterative framework, solving the linear Jacobian system of the current step through LGMRES iteration with block diagonal preconditions within each Newton step. The computation time and memory overhead of this invention increase approximately linearly with the number of elements (O(N)). No special initial values ​​are required during iteration to ensure convergence, making it suitable for calculating the steady-state electron distribution of multi-scale metallic nanostructures with fine features.

[0010] The present invention is implemented as follows: a method for calculating the steady-state electron distribution of a metal nanostructure, comprising: obtaining geometric and material parameters of the metal nanostructure, establishing the computational domain and boundary conditions, performing unstructured mesh partitioning on the structure, and recording mesh cell information;

[0011] Based on the constructed density functional, the Schrödinger-Poisson equation between the wave function and the electrostatic potential is determined;

[0012] Within the grid cells, polynomial basis functions are used to discretize the wave function and electrostatic potential, respectively. The discretized degrees of freedom are arranged by cell to form a global unknown vector. Then, the Schrödinger-Poisson equation is subjected to a discontinuous Galerkin test to establish a weak form equation. The discretized wave function and electrostatic potential are substituted into the weak form equation to establish a system of algebraic equations about the global unknown vector.

[0013] Based on the Schrödinger-Poisson equation for the aforementioned problem, a residual for the steady-state electron density distribution of metallic nanostructures is established. Discontinuous Galerkin discretization is applied to this residual, and a residual calculation method that eliminates the global matrix and calculates independently for each element is established.

[0014] Based on the above residual calculation method, a Newton method based on Krylov subspace technology, namely the Newton-Krylov method, is constructed to iteratively solve the above Schrödinger-Poisson equation. The initial values ​​of the Newton iteration are first set according to the computational domain and grid conditions.

[0015] In each Newton iteration step, a linear correction equation determined by the Jacobi system of the current step is constructed, and a Jacobi-vector product calculation method that eliminates the global matrix and is calculated independently on a unit-by-unit basis is established under the aforementioned discontinuous Galerkin discrete framework. Combined with the right preconditioning method applied independently on a unit-by-unit basis, the LGMRES algorithm is constructed to iteratively solve the linear correction equation to obtain the Newton update amount and update the Newton iteration solution.

[0016] After each of the above iteration steps, determine whether the solution satisfies the convergence condition. If it does, stop the iteration.

[0017] Output the convergence result based on the converged degrees of freedom data.

[0018] Furthermore, the computational domain is divided into a metal-dominant region and an electron overflow region according to physical properties. Material parameters, background ion density, boundary type, and initial value rules are configured for different regions. The computational domain is partitioned using a multi-scale, non-conformal mesh, and angstrom-scale local refinement is performed on the metal surface and at the location of fine nanostructures where electron density changes drastically. Gaussian integration rules are established for mesh cells and surface elements. The region label, boundary type, cell area, side length, adjacency relationship, and Gaussian integration point information of each mesh cell are recorded, and this information is used as input for subsequent calculations.

[0019] Furthermore, the potential function in the steady-state Schrödinger equation is jointly determined by the electrostatic potential, kinetic potential, exchange-related potential, and chemical potential terms; the Poisson equation is determined by the electrostatic potential Laplace term and the charge term, which includes electrons and background charge.

[0020] Furthermore, the weak-form equations employ numerical flux coupling to the physical fields in different elements. This numerical flux is derived from the Laplace term in the Schrödinger and Poisson equations, where the numerical gradients involved are calculated directly using a direct form. This gradient is determined directly by the average gradients on both sides of the interface and the jump penalty term set along the interface normal vector, without introducing additional variables or equations. The penalty coefficient is determined by the geometric parameters of the elements on both sides of the interface.

[0021] Furthermore, the residuals of the Schrödinger equation are determined by the Laplace term of the wave function, the product of the effective potential and the wave function; the residuals of the Poisson equation are determined by the Laplace term of the electrostatic potential and the charge term, the charge term including electrons and background charge. The calculation method is as follows: after evaluating the weak-form volume fraction and interface integral term within each grid cell, the degree-of-freedom residual of that cell is obtained by the inverse action of the local mass matrix; the residuals of each grid cell are connected through the numerical flux at the interface, without assembling the global matrix in the process.

[0022] Furthermore, before performing the Newton-Krylov iteration, initial values ​​are set based on the region labels of the mesh cells. This method supports stepped initial values, where the initial value of the electron density wavefunction in the main metallic region is 1, and the initial value of the electron density wavefunction in the electron overflow region is 0.

[0023] Furthermore, the method for constructing the linear correction equation is as follows: using the negative value of the residual vector as the right-hand side term; differentiating the wave function and potential using the weak form Schrödinger-Poisson equation respectively, deriving the Jacobian in the weak form, and deriving its interaction rule with a given vector, thus determining the left-hand side term of the linear correction equation. The Jacobian-vector product calculation method is established in the same way as the aforementioned residual calculation; the interface term of the Jacobian-vector product is processed using the same numerical flux as the aforementioned residual calculation; a global Jacobian matrix is ​​not constructed during the process.

[0024] Furthermore, the linear correction equation is solved iteratively using the restarted Krylov subspace method (LGMRES), with the Jacobian-vector product required during the solution process being evaluated successively according to the method described above. A block diagonal right precondition is applied to the Krylov subspace method on a cell-by-cell basis. The precondition block is determined by the inverse of the local mass matrix of the grid cell, the volume fraction, and the single-sided interface flux term after setting one side of the physical quantities of adjacent grid cells to zero. The precondition block is formed only within the corresponding grid cell and is applied cell-by-cell in the order of cell arrangement during application; a global precondition matrix is ​​not constructed during the process.

[0025] Another object of the present invention is to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the method for calculating the steady-state electron distribution of the metal nanostructure.

[0026] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method for calculating the steady-state electron distribution of the metal nanostructure.

[0027] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0028] To address the problems of large global matrix size, high storage requirements, and low computational efficiency in existing steady-state electron distribution calculations for metallic nanostructures, this invention proposes a method for calculating steady-state electron distribution based on a combination of discontinuous Galerkin discretization and Newton-Krylov iteration. This method constructs a residual calculation process at the unit level, transforming the traditional global matrix assembly process into a unit-by-unit independent calculation process, avoiding the formation and storage of large-scale global matrices and effectively reducing computational resource consumption. Simultaneously, by establishing a unit-by-unit Jacobi-vector product calculation method and combining it with a right-preconditioned LGMRES linear solution strategy, the complexity of matrix operations during iteration is reduced, improving the solution efficiency of the nonlinear Schrödinger-Poisson equation. Compared to traditional electronic structure calculation methods based on global matrices, this invention can adapt to computational scenarios involving complex metallic nanostructures and large-scale unstructured meshes. While ensuring the accuracy of electron density distribution calculation, it reduces memory consumption, increases computational speed, and enhances the ability to handle high-degree-of-freedom nanoscale electronic simulation problems, providing an efficient and reliable numerical calculation technology solution for nanostructure design and electronic device performance analysis.

[0029] This invention employs a discontinuous Galerkin space discretization method, supporting locally refined unstructured meshes. Furthermore, the discontinuous Galerkin method allows for the handling of discontinuous elements at cell interfaces, eliminating the need for asymptotic mesh transitions between adjacent elements and further reducing mesh overhead. This invention avoids the construction of a global matrix and can iteratively calculate the degrees of freedom of physical quantities element by element. Computational cost increases approximately linearly with the number of degrees of freedom, and memory overhead is also linearly related to the number of meshes, making it advantageous for solving large-scale 3D problems and complex geometric models.

[0030] This invention does not require specially constructed initial values. It supports iterative solutions initiated by step-type initial values: that is, even with only step-type initial values ​​that are constant inside and outside the metal, the iteration can still converge to the physical true solution. This invention exhibits fast convergence. It uses a fully coupled Newton-Krylov framework to solve the Schrödinger-Poisson equations; in practical examples, only approximately 10 Newton iterations are needed to reduce the residuals to... Magnitude.

[0031] (1) The expected benefits and commercial value of the technical solution of this invention after transformation are as follows:

[0032] The technical solution of this invention can be transformed into one of the core solution modules of optical response simulation software for metal nanostructures, serving cutting-edge research and industrial applications such as plasmon resonance modulation and nano-optoelectronic device design. Compared with the problem of drastic increases in memory and time overhead when the traditional finite element method reaches scales of millions of degrees of freedom, this invention can perform steady-state electron distribution calculations for large-scale complex geometries on ordinary workstations, significantly reducing hardware costs and computation cycles, and has clear academic promotion value and software industrialization prospects.

[0033] (2) The technical solution of this invention fills a technical gap in the industry both domestically and internationally:

[0034] This invention systematically integrates techniques such as discontinuous Galerkin discretization, direct-form numerical flux of the Laplace operator, fully coupled Newton-Krylov iteration, and block diagonal preconditioning LGMRES to form a new self-consistent numerical scheme that connects the entire chain from discretization to solution, specifically designed for solving the steady-state electron density distribution of metallic nanostructures. During the solution process, core calculations can be performed independently on a unit-by-unit basis, eliminating the need to construct a global matrix. This characteristic ensures that computational cost increases approximately linearly with the number of degrees of freedom, which is beneficial for solving large-scale three-dimensional problems and complex geometric models. Attached Figure Description

[0035] Figure 1 This is a flowchart of the steady-state electron distribution calculation method for metal nanostructures provided in this embodiment of the invention;

[0036] Figure 2 This is a structural diagram of the steady-state electron distribution calculation system for metal nanostructures provided in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of the electron density distribution along a specified cutoff line of the nanowire dimer structure provided in the embodiments of the present invention at different Newton iteration stages;

[0038] Figure 4 This is a convergence field diagram of the effective potential and electron density distribution provided in an embodiment of the present invention;

[0039] Figure 5 This is a schematic diagram illustrating the change in computation time with the number of grid cells provided in an embodiment of the present invention;

[0040] Figure 6 This is a schematic diagram illustrating the variation of peak memory usage with the number of grid cells during the calculation process, provided by an embodiment of the present invention.

[0041] Figure 7 This is a convergence field diagram of the effective potential and electron density distribution of the complex metal nanostructure provided in the embodiments of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0043] like Figure 1 As shown, the overall flow of the steady-state electron distribution calculation method for metal nanostructures provided in this embodiment of the invention is as follows: input the geometry, material parameters, boundary conditions, and computational domain settings of the metal nanostructure; provide initial values ​​based on region labels; discretize the steady-state Schrödinger-Poisson system on an unstructured mesh using the discontinuous Galerkin method with polynomial basis functions; assemble the fully coupled residual vector; solve the equations using the Newton-Krylov method; solve the linear correction equations within each Newton step using LGMRES with block diagonal preconditions; finally, determine convergence using the residual and particle number conservation criteria, and verify the results through the spatial distribution of the physical field.

[0044] S101: Geometric modeling and generation of computational domain mesh.

[0045] Based on the geometry, size parameters, material parameters, and computational domain of the metal nanodevice to be solved, a practical geometric model is established. Specifically, the computational domain is denoted as... The model is divided into a metallic main region and an electron overflow region based on physical properties. Appropriate material parameters, background ion density, boundary type, and initial value rules are set for different regions. An unstructured mesh is used to partition the geometric model. On the metal surface, in areas with fine nanostructures and drastic changes in electron density, a sub-angstrom scale mesh is used for local refinement, while a coarser mesh is used in other areas. The partitioned elements are denoted as... The interface of the unit is denoted as Then, record the cell grid's label, boundary type, cell area, side length, adjacency relationship, and Gaussian integration point information.

[0046] S102: Establish the steady-state Schrödinger-Poisson physical model to be solved.

[0047] Define electron density as The positive background charge density is The electron density equivalent wave function is Self-consistent electrostatic potential is The electron density and self-consistent electrostatic potential satisfy the steady-state Schrödinger-Poisson equations:

[0048]

[0049]

[0050] in It is the von Weizsäcker parameter. The effective potentials include the Thomas-Fermi local kinetic potential, exchange-related potential, self-consistent electrostatic potential, and chemical potentials. Represents the positive background charge wave function. are the coefficients on the left-hand side of the Poisson equation.

[0051] S103: The weak form of the steady-state Schrödinger-Poisson equation is constructed using the discontinuous Galerkin method.

[0052] Mesh subdivision cells constructed in step one The electron density wavefunction and self-consistent electrostatic potential to be solved are expanded using basis functions:

[0053]

[0054] in Represents the electron density wave function and self-consistent electrostatic potential ; Representation and basis functions The relevant physical quantities and degrees of freedom; Represents the basis functions of the polynomial; This represents the number of polynomial basis functions. Galerkin tests are performed on both sides of the Schrödinger-Poisson equation established in step two using the same basis functions:

[0055]

[0056]

[0057] The For unit The boundary, Let be the out-of-plane normal vector of the element. and Approximate representation of the element boundary and Numerical flux of gradients. A flux method of the following form is used to ensure that physical quantities at element boundaries... The gradient is:

[0058]

[0059] in Indicates from unit External and internal physical quantities Take the average of the extreme values. This represents the jump value of a physical quantity across the boundary from the outside to the inside of the unit. It is the projection of the center point of adjacent cells onto the surface normal vector. The distance above, It is a numerical parameter.

[0060] The above equations are the weak form equations that need to be solved when solving the steady-state electron distribution problem.

[0061] S104: Assemble the fully coupled residual vector.

[0062] Based on the steady-state Schrödinger-Poisson equation constructed in step two, the residual of the steady-state electron density distribution of the metal nanostructure is set. and Their expressions are

[0063]

[0064]

[0065] Define a residual vector , to represent the residual of the numerical solution of the weak form equation of electron density distribution. The degrees of freedom of the equivalent wave function variables and electrostatic potential variables of each element in step three are arranged by element as a global unknown vector. Calculate the fully coupled residual vector The residual vector is evaluated using a cell-by-cell matrix-free method. Based on the weak form expression in step three, the residual vector is evaluated for each cell. ( , The formula for calculating the residual is:

[0066]

[0067]

[0068] in For polynomial basis functions in the unit The local mass matrix on, whose elements are

[0069]

[0070] The residual here contains both Schrödinger equation terms and Poisson equation terms.

[0071] S105: Set initial values.

[0072] Regarding initial values, a stepped equivalent wave function and initial potential are set according to the region labels of the physical model: This method supports setting stepped initial values. The equivalent wave function is set to 1 inside the metal and 0 in the outer region. The initial potential is selected based on the equivalent potential, with the metal interior being chosen as the zero point of the equivalent potential energy. The equivalent potential outside the metal is set as the material work function. Under this specification, the initial potential is calculated using the expression for the equivalent potential. This step enables the solver to start stably without relying on an external high-precision reference field.

[0073] S106: Solve for steady-state electron density distribution using the Newton-Krylov method.

[0074] Configure the parameters of the iterative solver, including the maximum number of iterations, the Krylov subspace size, the Newton / GMRES convergence threshold, and the number of vectors retained upon restart. The solution process consists of two layers: a Newton outer loop and a GMRES inner loop.

[0075] Given the k-th iteration solution in Newton's outer loop Using the formula Update the iterative solution. Wherein... By solving linear problems We obtain the residual vector here. It is obtained from the weak form equation set in step four. For residuals exist The Jacobian matrix at that location.

[0076] The inner loop uses LGMRES iterative solution to solve the linear equations. This process involves the understanding of forms such as The formal Jacobi-vector product is evaluated successively. A weak form of the above differential expression is constructed using the discontinuous Galerkin method. For each element... , ,have

[0077]

[0078]

[0079] in It is with unknown vectors Any vector having the same degree of freedom structure and arrangement order, and It is the effective potential term in the residual of the wave function equation. To each and The partial derivative coefficients are defined as follows: , In the Jacobian-vector product, As the self-coupling coefficient of the wave function equation, It appears as the cross-coupling coefficient of the wave function equation with respect to electric potential. Numerical flux at the interface. , The flux formula is expanded using the exact same form as in step three. In practice, each physical quantity is expressed in each unit. ( , The scalar polynomial basis function is expanded according to step three. The terms on the right-hand side of the above equation are directly evaluated by Gaussian integrals on the unit and its interface without assembling any global matrix.

[0080] To accelerate convergence, a right precondition is applied during this process. The work done in solving linear problems Operations transformed into .in Constructed element-wise, making it a block diagonal matrix formed by matrix blocks on each element. For each element... The corresponding matrix block The construction form is

[0081]

[0082] In the unit The matrix construction form is as follows

[0083]

[0084] in , It is the size of unit array, All matrix sub-blocks A matrix, whose matrix elements are defined by the formula:

[0085]

[0086]

[0087]

[0088]

[0089] in

[0090]

[0091] This is the flux expression obtained by taking the physical quantity on one side of the adjacent unit to zero in the aforementioned flux formula, where the repetition index is specified. Sum.

[0092] This process involves and All are block diagonal matrices, with matrix blocks formed within cells and the global matrix not explicitly constructed or stored.

[0093] S107: Determine if convergence has occurred.

[0094] After each Newton iteration step, the residual norm is calculated as an error measure to determine whether the current solution satisfies the Schrödinger-Poisson equation, and is used to determine whether the current solution has converged. When the residual norm is less than a preset threshold, the current solution is numerically accepted as a convergent solution. Simultaneously, residual norm error information is output at each iteration step, providing a basis for subsequent parameter adjustment and error diagnosis.

[0095] S108: Output the convergence result.

[0096] During or after Newton iteration, the required steady-state physical quantities are reconstructed based on the degree-of-freedom data, including electron density, electron density gradient, self-consistent electrostatic potential, potential gradient, effective potential, background ion density, and particle number conservation error. Output data can be saved in the form of cell centers, integration points, sampling points, or post-processing grids to support the plotting of physics field contour maps, intercept distributions, convergence history curves, error convergence curves, and particle number conservation curves. The output process allows setting output intervals and output variable types, enabling the recording and tracking of intermediate results for each Newton iteration step.

[0097] Electron density

[0098] Positive charge density in the background

[0099] : Equivalent wave function of electron density

[0100] Self-consistent electrostatic potential

[0101] : von Weizsäcker parameters

[0102] Effective potentials include the Thomas-Fermi local kinetic potential, exchange-dependent potential, self-consistent electrostatic potential, and chemical potential terms.

[0103] The coefficients on the left-hand side of the Poisson equation.

[0104] : Represents a general scalar physical field, such as a wave function. and electrostatic potential

[0105] : Represents the basis functions The relevant physical quantities and degrees of freedom

[0106] : Represents polynomial basis functions

[0107] : Represents the number of polynomial basis functions

[0108] , : Represents the flux of the gradient term

[0109] : Represents the out-of-plane normal vector of the element.

[0110] , : Represents the residuals of the steady-state electron density distribution of the metallic nanostructure, which are the residuals of the Schrödinger equation and the Poisson equation, respectively.

[0111] : Represents the global unknown vector representing the degrees of freedom of the wave function variables and electrostatic potential variables arranged in units.

[0112] : Residual vector representing the coupling between the residual degrees of freedom of the Schrödinger equation and the residual degrees of freedom of the Poisson equation.

[0113] : indicates that the polynomial basis functions are in the unit Local mass matrix on

[0114] : Represents Jacobi in the Newton-Krylov method

[0115] : Represents the iterative solution correction amount during the iterative process.

[0116] : Represents the relationship between unknown vectors Arbitrary vectors with the same degree of freedom structure and arrangement order

[0117] : Represents the Jacobian-vector product in the Newton-Krylov method

[0118] , : Represents the effective potential term in the residual of the wave function equation To each and partial derivative coefficients

[0119] : Represents the preconditions applied in the Newton-Krylov method

[0120] : Represents the Jacobian-vector product after applying preconditions.

[0121] like Figure 2 As shown, the steady-state electron distribution calculation system for metal nanostructures provided in this embodiment of the invention includes:

[0122] The computational domain network generation module is used to perform geometric modeling and generate computational domain meshes.

[0123] The physical model building module is used to build the steady-state Schrödinger-Poisson physical model to be solved.

[0124] The weak form construction module is used to construct the weak form of the steady-state Schrödinger-Poisson equation using the discontinuous Galerkin method.

[0125] The residual vector assembly module is used to assemble fully coupled residual vectors.

[0126] The condition setting module is used to set boundary conditions and step initial values.

[0127] The equation solving module is used to solve the weak form equations of electron density distribution using the Newton-Krylov method.

[0128] The convergence criterion module is used to determine convergence criteria.

[0129] The output module is used to output steady-state physical quantities and physical field images.

[0130] Evidence related to the technical effects obtained by the embodiments of the present invention.

[0131] To verify the correctness and effectiveness of this invention, the steady-state electron density distribution of a pair of sodium nanowire dimers is analyzed below, and the relationship between computation time and peak memory usage as a function of the number of cells is given.

[0132] The numerical calculations of the method of this invention were all performed on a personal workstation equipped with an Intel Core i9-14900KF processor (32 cores), 64 GB of memory and an NVIDIA GeForce RTX 4090 graphics card, and the operating system was Ubuntu 22.04.4 LTS.

[0133] The steady-state electron density distribution of a pair of sodium nanowire dimers was calculated using the method of this invention. The dimer consists of two nanowires with a cross-sectional diameter of... It is composed of sodium nanowires. A width of [missing information] is formed between two coupled nanowires. The nanometer gaps serve as the core region for the extreme localization of the optical field. Epitaxial growth at the metal boundary The region is defined as the computational domain for the Schrödinger-Poisson equation, used to characterize the quantum mechanical behavior of free electrons. Under the equivalent Schrödinger equation representation, the free electron gas of sodium materials... von Weizsacker parameters The electron density and potential field in this example are set with different initial values ​​inside and outside the metal.

[0134] In practical applications, metallic nanostructures may possess complex geometries. Therefore, verifying the robustness of this invention to complex geometries is crucial. This invention utilizes two highly irregular models (see...). Figure 7 The robustness of the method of the present invention is tested using various challenging geometric features, including sharp corners, narrow slits, and non-convex boundaries. This tests the method's ability to handle complex geometries.

[0135] Figure 3 The electron density distribution of the nanowire dimer structure along a specified intercept at different Newton iteration stages is shown, and high-precision results based on the finite element method using Newton's method are also provided as a reference. The right figure is a logarithmic plot of the electron density distribution in the metal boundary region. Figure 3 The convergent solution matches the reference solution very well. The overall image consists of three parts: the top part is a unified legend for the entire image, representing the reference solution, the initial value of Newton's iterations, and the approximate solution for the first 5 Newton iterations.

[0136] Figure 3 In the left-middle figure, the horizontal axis represents the position in the x-direction of space, and the vertical axis represents the equivalent electron density. The light yellow filled area in the left figure represents the metal nanowire structure, and the multiple colored curves represent the equivalent electron density distribution at each iteration stage. Each light yellow area in the left figure is marked with a " / / " symbol at its left and right boundaries, indicating that the data in the x-direction of space is truncated. A pair of red arrows in the lower right corner of the left figure point to the magnified x-region in the right figure.

[0137] Figure 3 The horizontal axis of the middle right figure represents the x-axis in space relative to the metal boundary. The position of the ordinate is the common logarithm of the equivalent electron density. The light yellow filled area in the right figure represents the metallic region, and the multiple colored curves represent the equivalent electron density distribution at each iteration stage. The blue arrow in the lower left corner of the right figure marks the location of the metallic boundary. The physical model used in this calculation is shown in the right figure, with the model's diameter and gap dimensions labeled. The red dashed arrow on the model indicates the x-axis position.

[0138] Figure 4This is a convergence field plot of the effective potential and electron density distribution. To more intuitively visualize the changes in electron density, a logarithmic plot of the electron density over time is shown here. Figure 4 The black dashed line represents the surface of the metal structure. The relative error of Newton's iteration is less than... This figure illustrates that the solver of this invention can obtain high-quality physical fields. The image as a whole consists of three parts:

[0139] Figure 4 (a) is the convergence field diagram of the effective potential, showing the spatial distribution of the effective potential after multiple Newton iterations. Figure 4 (a) shows the color legend at the top, which represents the correspondence between the intensity of the effective potential and the color. Figure 4 (a) shows the length legend in the upper left corner, indicating the size of the simulated structure.

[0140] Figure 4 (b) is the convergence field diagram of the equivalent electron density, showing the spatial distribution of the equivalent electron density after multiple Newton iterations. Figure 4 (b) The top of the graph is a color legend, indicating that the equivalent electron density is taken as the commonly used logarithm ( The correspondence between the value of () and the color. Figure 4 (b) shows the length legend in the upper left corner, indicating the size of the simulated structure.

[0141] Figure 5 This illustrates how computation time varies with the number of grid cells. For this example, the average CPU time for each moment multiplication calculation (left axis, polyline) and the number of moment multiplication calculations required for convergence (right axis, bar chart) are plotted against the number of grid cells. This problem is solved using 10 different cell numbers, roughly forming a range from... arrive The arithmetic sequence is represented by the column height, which indicates the total number of moment multiplication calculations for each grid cell. The total CPU time for each calculation is the product of the average CPU time and the number of moment multiplication evaluations. As shown in the figure, for a single moment multiplication evaluation, the computation time is linearly related to the number of cells. On the other hand, convergence is achieved (the relative residual of the Newton iteration is less than...). The number of moment-vector multiplication evaluations required does not increase significantly with the problem size, indicating that it is mesh-independent.

[0142] Therefore, the overall CPU time of the method of the present invention increases approximately linearly with the problem size. Figure 5The horizontal axis represents the total number of grid cells in the computational model (increasing from left to right). The vertical axis on the left represents the CPU time cost (in seconds) of a single matrix-vector multiplication (m×V), indicating the unit time of the core operation in the algorithm. Here, a single matrix-vector multiplication is defined as all cells in the model performing one matrix-vector multiplication. The vertical axis on the right represents the total number of matrix-vector multiplications required to achieve the preset convergence accuracy (in seconds). The figure shows the iterative complexity during the convergence process. The red line in the graph uses the left ordinate, and the light blue bars use the right ordinate.

[0143] Figure 6 This graph illustrates how peak memory usage changes with the number of grid cells during computation. The horizontal axis represents the total number of grid cells in the computation model (increasing from left to right). The vertical axis represents the peak memory usage during computation (in GB).

[0144] Figure 7 This is a convergence field diagram of the steady-state electron density for a complex structure. The diagram shows the distribution of the initial step values ​​at the start of the calculation, as well as the spatial distribution of the converged steady-state electron density. The image consists of four parts:

[0145] The upper left portion of the figure shows the initial equivalent electron density distribution of the complex structure A before the iteration. In this calculation, the initial electron density distribution of the metallic region is set as follows: The lower left corner of the figure shows the length of structure A, representing the size of the simulated structure. The upper right part of the figure shows the spatial distribution of the steady-state electron density of the complex structure A after multiple Newton iterations and convergence. The right side of the figure shows the color scheme of the equivalent electron density distribution of structure A, indicating the correspondence between the equivalent electron density and the color.

[0146] The lower left figure shows the initial equivalent electron density distribution of the complex structure B before the iteration. In this calculation, the initial electron density distribution of the metallic region is set to... The lower left corner of the figure shows the length of structure B, representing the size of the simulated structure. The lower right section shows the spatial distribution of the steady-state electron density of the complex structure B after multiple Newton iterations and convergence. The right side of the figure shows the color scheme of the equivalent electron density distribution of structure B, indicating the correspondence between the equivalent electron density and the color.

[0147] Table 1 shows the relative stability of the steady-state electron distribution of sodium nanowire dimers obtained by the method of this invention under different mesh sizes. Error (here) (For the second norm). This example uses the smallest unit size as... Based on an unstructured triangular mesh (here) By successively encrypting each triangular unit by dividing it into four equal parts, the size of the smallest unit is halved each time. and And using the high-precision converged solution as the reference solution to calculate the relative error.

[0148] Table 1

[0149]

[0150] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0151] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by any person skilled in the art within the scope of the technology disclosed in the present invention, within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for calculating the steady-state electron distribution of a metallic nanostructure, characterized in that, include: Obtain geometric and material parameters of metallic nanostructures, establish computational domain and boundary conditions, perform unstructured mesh generation on the structure, and record mesh element information; Based on the constructed density functional, the Schrödinger-Poisson equation between the wave function and the electrostatic potential is determined; Within the grid cells, polynomial basis functions are used to discretize the wave function and electrostatic potential respectively. The discretized degrees of freedom are arranged according to the cells to form a global unknown vector. Then, the Schrödinger-Poisson equation is subjected to the discontinuous Galerkin test to establish a weak form equation. The discretized wave function and electrostatic potential are substituted into the weak form equation to establish a system of algebraic equations about the global unknown vector. Based on the Schrödinger-Poisson equation for the problem, the residual of the steady-state electron density distribution of metal nanostructures is established; discontinuous Galerkin discretization is applied to the residual to establish a residual calculation method that eliminates the global matrix and calculates it unit by unit. Based on the above residual calculation method, a Newton method based on Krylov subspace technology, namely the Newton-Krylov method, is constructed to iteratively solve the above Schrödinger-Poisson equation. The initial values ​​of the Newton iteration are first set according to the computational domain and grid conditions. In each Newton iteration step, a linear correction equation determined by the Jacobi system of the current step is constructed, and a Jacobi-vector product calculation method that eliminates the global matrix and is calculated independently on a unit-by-unit basis is established under the aforementioned discontinuous Galerkin discrete framework. Combined with the right preconditioning method applied independently on a unit-by-unit basis, the LGMRES algorithm is constructed to iteratively solve the linear correction equation to obtain the Newton update amount and update the Newton iteration solution. After each of the above iteration steps, determine whether the solution satisfies the convergence condition. If it does, stop the iteration. Output the convergence result based on the converged degrees of freedom data.

2. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, The computational domain is divided into a metal-dominant region and an electron overflow region according to physical properties. Material parameters, background ion density, boundary type, and initial value rules are configured for different regions. The computational domain is partitioned using a multi-scale, non-conformal mesh, and angstrom-scale local refinement is performed on the metal surface and at the location of fine nanostructures where electron density changes drastically. Gaussian integration rules are established for mesh cells and surface elements. The region label, boundary type, cell area, side length, adjacency relationship, and Gaussian integration point information of each mesh cell are recorded, and this information is used as input for subsequent calculations.

3. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, The potential function in the steady-state Schrödinger equation is determined by the terms of electrostatic potential, kinetic potential, exchange-related potential, and chemical potential; the Poisson equation is determined by the Laplace term of electrostatic potential and the charge term, which includes electrons and background charge.

4. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, The weak form equations employ numerical flux coupling of physical fields in different units; the numerical flux is derived from the Laplace term in the Schrödinger and Poisson equations, where the numerical gradients involved are calculated in a direct form, i.e., directly determined by the average term of the gradients on both sides of the interface and the jump penalty term set along the interface normal vector, without introducing additional variables and equations; the penalty coefficient is determined by the geometric parameters of the units on both sides of the interface.

5. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, The residuals of the Schrödinger equation are determined by the Laplace term of the wave function, the effective potential, and the product of the wave function; the residuals of the Poisson equation are determined by the Laplace term of the electrostatic potential and the charge term, the charge term including electrons and background charge; the calculation method is to evaluate the weak form volume integral term and interface integral term within each grid cell, and then obtain the degree of freedom residual of that cell by the inverse action of the local mass matrix; the residuals of each grid cell are connected through the numerical flux at the interface, without assembling the global matrix in the process.

6. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, Before performing the Newton-Krylov iteration, initial values ​​are set according to the region labels of the mesh cells; this method supports stepped initial values, where the initial value of the electron density wave function in the metal body region is 1, and the initial value of the electron density wave function in the electron overflow region is 0.

7. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, The method for constructing the linear correction equation is as follows: using the negative value of the residual vector as the right-hand side term; differentiating the wave function and potential using the weak form Schrödinger-Poisson equation respectively, deriving the Jacobian in the weak form, and deriving its interaction rule with the given vector, thus determining the left-hand side term of the linear correction equation; the Jacobian-vector product calculation method is established in the same way as the aforementioned residual calculation; the interface term of the Jacobian-vector product is processed using the same numerical flux as the aforementioned residual calculation; no global Jacobian matrix is ​​constructed during the process.

8. The method for calculating the steady-state electron distribution of metal nanostructures according to claim 1, characterized in that, The linear correction equation is solved iteratively using the restarted Krylov subspace method (LGMRES), and the Jacobian-vector product required in the solution process is evaluated successively according to the method. A block diagonal right precondition is constructed cell by cell for the Krylov subspace method. The precondition block is determined by the inverse of the local mass matrix of the grid cell, the volume fraction, and the single-sided interface flux term after setting the physical quantity on one side of the adjacent grid cell to zero. The precondition block is formed only within the corresponding grid cell and is applied cell by cell in the order of cell arrangement when applied. No global precondition matrix is ​​constructed in the process.

9. A system for calculating the steady-state electron distribution of a metal nanostructure according to any one of claims 1 to 8, characterized in that, include: The computational domain network generation module is used to perform geometric modeling and generate computational domain meshes; The physical model building module is used to build the steady-state Schrödinger-Poisson physical model to be solved; The weak form construction module is used to construct the weak form of the steady-state Schrödinger-Poisson equation using the discontinuous Galerkin method. The residual vector assembly module is used to assemble fully coupled residual vectors; The condition setting module is used to set boundary conditions and step initial values; The equation solving module is used to solve the weak form equations of electron density distribution using the Newton-Krylov method. The convergence criterion module is used to determine the convergence criteria. The output module is used to output steady-state physical quantities and physical field images.

10. A computer device comprising a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the method for calculating the steady-state electron distribution of a metal nanostructure as described in any one of claims 1 to 8.