Molecular dynamics algorithm construction and analysis method for adsorbing heavy metals by minerals

Abstract

The invention discloses a molecular dynamics algorithm construction and analysis method for adsorbing heavy metals by minerals. The method comprises the following steps: S1, comprehensively obtaining information of the minerals and the heavy metals from microcosmic to macroscopic; s2, constructing a physical model based on the characterization result of the step S1; s3, multi-scale physical parameters are introduced, and the physical model is converted into a mathematical model; s4, integrating the constructed mathematical model into a software framework, and performing deep customization on the software at the same time; s5, combining environmental factors and coupling multiple physical fields, and executing molecular dynamics simulation; s6, comprehensively analyzing simulation results; according to the method, an accurate model is constructed through multi-level characterization, deep customization simulation is carried out on multiple physical fields and environmental factors, comprehensive analysis is carried out by applying multiple principles and a machine learning algorithm, the model and calculation efficiency are optimized, accurate simulation, deep analysis and efficient research of the process of adsorbing heavy metals by minerals are achieved, and the method is suitable for large-scale popularization and application. And powerful support is provided for practical application and development of related fields.

Description

Technical Field

[0001] The present invention relates to the field of environmental science, and particularly to a method, system and storage medium for constructing and analyzing the kinetic algorithm of mineral adsorption of heavy metals. Background Art

[0002] In the fields of environmental science and materials science, the treatment of heavy metal pollution has always been an important topic of great concern. With the acceleration of the industrialization process, a large amount of heavy metal pollutants are discharged into the environment, posing a serious threat to the ecosystem and human health. Mineral adsorption, as an efficient, economical and environmentally friendly method for treating heavy metal pollution, has been widely studied and applied.

[0003] In the early stage of the research process, researchers mainly focused on exploring the adsorption effect of minerals on heavy metals at the macroscopic level. For example, through conventional chemical analysis methods, the adsorption amounts of different types of minerals on common heavy metal ions were measured under specific experimental conditions. Although this research method can intuitively understand the removal ability of minerals on heavy metals, it only stays at the observation of surface phenomena and cannot deeply understand how the adsorption process occurs and the internal action mechanisms involved.

[0004] With the rapid development of various advanced scientific and technological means, a series of advanced spectroscopic techniques have been gradually applied to the research in this field. Taking X-ray photoelectron spectroscopy as an example, it can accurately measure the chemical valence states of elements on the mineral surface and the changes in elemental composition. By comparing and analyzing the XPS spectra of minerals before and after heavy metal adsorption, it is possible to clearly know whether a chemical reaction has occurred between the surface functional groups of the mineral and heavy metal ions, as well as the changes in the chemical states of elements before and after the reaction. At the same time, electron microscopy technology has also brought new breakthroughs to the research on mineral adsorption of heavy metals, further exploring the internal relationship between the microscopic structure of minerals and their adsorption properties.

[0005] In terms of model construction, the physical models constructed in the early stage were relatively simple and idealized. Only the most basic electrostatic attraction between minerals and heavy metal ions was considered, while the rich and diverse active sites on the mineral surface were ignored. In fact, there are many types of active sites on the mineral surface, and different active sites have significant differences in the adsorption ability and adsorption mode of heavy metal ions. At the same time, the early models did not fully consider the hydrated shell structure of heavy metal ions existing in aqueous solutions, which will have an important impact on the interaction between heavy metal ions and the mineral surface. In addition, for the chemical adsorption and physical adsorption involved in the adsorption process, the early models also failed to describe them comprehensively and accurately.

[0006] In the field of simulation technology, traditional molecular dynamics simulations in the study of the process of minerals adsorbing heavy metals often suffer from relatively simple simulation environment settings. Only the basic interactions between minerals and heavy metal ions are considered, without fully taking into account the influence of the complex and diverse environmental factors in the actual environment on the adsorption process. In real natural environments or industrial wastewater treatment scenarios, factors such as the pH value of the solution, ionic strength, and the presence of an externally applied electric field will all have a significant impact on the process of minerals adsorbing heavy metals. For example, the pH value of the solution will change the charge properties and charge density on the mineral surface, thereby affecting the strength of the electrostatic interaction between the mineral and heavy metal ions; changes in ionic strength may affect the hydration shell structure of heavy metal ions and the migration rate of ions in the solution; the presence of an externally applied electric field may change the movement direction and migration rate of heavy metal ions, thus promoting or inhibiting the adsorption process. Nowadays, with the continuous in-depth study of the process of minerals adsorbing heavy metals, researchers have realized the need to comprehensively consider these complex environmental factors. By coupling multiple physical fields such as electromagnetic fields and temperature fields, a simulation system closer to the actual situation is established to achieve accurate simulation of the process of minerals adsorbing heavy metals under different conditions. However, in the actual operation process, accurately mapping complex physical models into mathematical models and successfully integrating them into a suitable molecular dynamics software framework still faces many technical problems and challenges. For example, how to accurately quantify and parameterize various interactions, how to reasonably introduce multi-scale physical parameters into the mathematical model, and how to ensure that the software framework can efficiently and stably run complex simulation calculations. In addition, for the analysis of simulation results, multi-disciplinary analysis methods also need to be used for comprehensive and in-depth analysis.

[0007] Generally speaking, although multi-disciplinary knowledge such as statistical mechanics, thermodynamics, and kinetics principles has been used to analyze the energy changes, kinetic characteristics, and thermodynamic equilibrium of the adsorption process, as well as the structural evolution and stability of the system after adsorption, there is still a large room for improvement in the perfection and accuracy of the analysis methods. Summary of the Invention

[0008] In order to overcome the deficiencies in the above-mentioned prior art, the present invention provides a method for constructing and analyzing a molecular dynamics algorithm for minerals adsorbing heavy metals to solve the above problems.

[0009] To achieve the above object, the technical solution provided by the preferred embodiment of the present invention is as follows: A method for constructing and analyzing a molecular dynamics algorithm for minerals adsorbing heavy metals, the method comprising:

[0010] Step S1: Conduct multi-level characterization of minerals and heavy metals, covering from the microscopic atomic and molecular levels to the macroscopic material property levels. Use spectroscopic techniques to analyze the chemical states of functional groups on the mineral surface and heavy metal ions, and use electron microscopy techniques to observe the microscopic structure and surface morphology of minerals;

[0011] Step S2: Based on the characterization results, design and construct a physical model for the adsorption of heavy metals by minerals. The physical model includes active sites on the mineral surface, the hydration shell structure of heavy metal ions, and the interaction mechanism between the above two;

[0012] Step S3: Based on the constructed physical model, map it to a mathematical model by introducing multi-scale physical parameters, and integrate information at different scales into a unified mathematical framework;

[0013] Step S4: Select a molecular dynamics software framework, integrate the mathematical model into this framework, and at the same time achieve in-depth customization of the software framework by adding functional modules for precise simulation of the adsorption process and handling of special physical phenomena;

[0014] Step S5: Combine environmental factors and couple multiple physical fields to perform molecular dynamics simulations for simulating the adsorption process under different conditions;

[0015] Step S6: Comprehensively analyze the simulation results. The specific analysis methods include statistical mechanics, thermodynamics, and kinetic principles, analyze the energy changes, kinetic characteristics, and thermodynamic equilibrium of the adsorption process, as well as the structural evolution and stability of the system after adsorption.

[0016] In the present invention, in step S2, the interaction mechanism includes electrostatic interaction, chemical adsorption, and physical adsorption;

[0017] Combining step S2 and step S3, in the step of mapping the physical model to the mathematical model, for the electrostatic interaction, the Poisson - Boltzmann equation is used:

[0018]

[0019] where, is the electrostatic potential, ρ is the charge density, and ∈ is the dielectric constant;

[0020] This equation is used to describe the electrostatic interaction between minerals and heavy metal ions, considering the influence of charge distribution on the electrostatic potential.

[0021] In the present invention, after selecting the molecular dynamics software framework in step S4, the Hamiltonian mechanics principle is used for system description, and the Hamiltonian H of the system can be expressed as:

[0022] H = T + V;

[0023] where T is the kinetic energy and V is the potential energy;

[0024] where pi is the momentum of particle i and mi is the mass of particle i;

[0025] The potential energy V includes the intermolecular interaction potential energy, the electrostatic potential energy and the external field potential energy, and its specific form is determined in combination with the physical model.

[0026] In the present invention, when performing molecular dynamics simulation in step S5, for the time evolution of the system, the Nose-Hoover chain thermostat algorithm is adopted, and its equation of motion is:

[0027]

[0028] where pi is the momentum of particle i, qi is the position of particle i, εj is the thermostat variable, Qj is the mass parameter of the thermostat, K B is the Boltzmann constant, T is the temperature, and g is the degree of freedom;

[0029] The above algorithm is used to control the temperature of the simulation system to ensure that the system is in the required thermal equilibrium state.

[0030] In the present invention, when comprehensively analyzing the simulation results in step S6, the Gibbs free energy G of the system is calculated using the thermodynamic relation:

[0031] G = H - TS;

[0032] where H is the enthalpy, T is the temperature, and S is the entropy;

[0033] S is calculated according to the classical statistical mechanics formula: S = -k B ∑ i p i Inp i where p i is the probability of the initial microscopic state i of the system.

[0034] In the present invention, in the process of designing and constructing the physical model of mineral adsorption of heavy metals in step S2, the topological structure of the mineral surface is considered. By introducing the fractal geometry theory, the mineral surface is regarded as a fractal object to describe the distribution characteristics of surface roughness and active sites;

[0035] Specifically, the box-counting method is used to determine the fractal dimension D, and the formula is:

[0036] N(∈) ∝ ∈ -D

[0037] where N(∈) is the number of boxes with side length ∈ required to cover the mineral surface;

[0038] Step S2 obtains the fractal dimension D by calculating N(∈) under different ∈ values and performing linear fitting, which is used to accurately characterize the complexity of the mineral surface and the distribution of active sites.

[0039] In the present invention, the parameters related to surface interaction in the physical model are adjusted in combination with the fractal dimension D, so that the physical model can accurately reflect the adsorption characteristics of the actual mineral surface.

[0040] In the present invention, when selecting the molecular dynamics software framework, the parallel computing module of the software is optimized by combining task parallelism and data parallelism;

[0041] For task parallelism, the molecular dynamics simulation process is decomposed into multiple independent subtasks, and the subtasks are assigned to different computing cores, and the message passing interface is used for communication and coordination between tasks;

[0042] For data parallelism, the particle data in the simulation system is divided into multiple data blocks, and the data blocks are assigned to different computing units, and the unified computing device architecture is used for parallel computing;

[0043] According to different hardware architectures and the characteristics of computing tasks, the allocation strategies of tasks and data are dynamically adjusted.

[0044] In the present invention, the multiple physical fields in step S5 include but are not limited to electromagnetic fields and temperature fields;

[0045] The environmental factors include pH value, ionic strength and applied electric field;

[0046] When performing molecular dynamics simulation, for the coupling of different environmental factors, a multi-physical field coupling algorithm is used to truly simulate the adsorption process under the actual environment.

[0047] In the present invention, step S6 uses machine learning algorithms for auxiliary analysis, which specifically includes the following steps:

[0048] Step S6.1: Use the characteristic quantities in the simulation results as input features and the target performance as output labels to construct a data set;

[0049] Step S6.2: Use the TensorFlow deep learning framework to construct a neural network model, select a multi-layer perceptron and a convolutional neural network as the network structure, and select appropriate numbers of layers and neurons according to the size and characteristics of the data set;

[0050] Step S6.3: Use the stochastic gradient descent algorithm as the optimizer to train the model by minimizing the loss function. During the training process, the data set is divided into a training set, a validation set and a test set. The overfitting situation of the model is monitored through the validation set, and the performance of the model is evaluated according to the test set;

[0051] Step S6.4: Use the trained model to predict the adsorption performance under different conditions, and at the same time, by analyzing the weights and activation functions of the neural network, explore the potential physical mechanisms in the adsorption process.

[0052] Beneficial effects:

[0053] In the present invention, through multi-level characterization of minerals and heavy metals, a physical model containing detailed interaction mechanisms is constructed, which is then accurately mapped to a mathematical model. Using the molecular dynamics software framework for in-depth customized simulation, combined with multiple physical fields and environmental factors, the process of mineral adsorption of heavy metals can be simulated extremely accurately. At the same time, by comprehensively analyzing using the principles of statistical mechanics, thermodynamics, and kinetics as well as machine learning algorithms, not only can the energy changes, kinetic characteristics, and thermodynamic equilibrium in the adsorption process be deeply understood, but also potential physical mechanisms can be explored, providing a comprehensive and in-depth analysis perspective for the research on mineral adsorption of heavy metals.

[0054] When constructing the physical model in the present invention, considering the surface topology of minerals, the fractal geometry theory is introduced to accurately characterize the complexity of the mineral surface and the distribution of active sites, and then the parameters of the physical model are adjusted so that the model can accurately reflect the adsorption characteristics of the actual mineral surface. When performing molecular dynamics simulations, a multi-physical field coupling algorithm is used for different environmental factors, and the parallel computing module is optimized when selecting the molecular dynamics software framework. These measures make the algorithm model have stronger adaptability and accuracy, and can better simulate the process of mineral adsorption of heavy metals under different conditions, providing reliable theoretical support for practical applications.

[0055] The present invention also greatly improves the simulation calculation efficiency by selecting a suitable molecular dynamics software framework and adopting an optimization method that combines task parallelism and data parallelism for its parallel computing module, and dynamically adjusting the allocation strategy according to different hardware architectures and computational task characteristics. At the same time, combined with machine learning algorithms for auxiliary analysis, using a neural network model for adsorption performance prediction and potential mechanism exploration further improves the performance of the algorithm. This enables more accurate results to be obtained in a shorter time when studying mineral adsorption of heavy metals, providing strong technical support for the rapid development of related fields.

[0056] In summary, the present invention constructs an accurate model through multi-level characterization, conducts in-depth customized simulation by combining multiple physical fields and environmental factors, comprehensively analyzes using various principles and machine learning algorithms, and at the same time optimizes the model and computational efficiency, realizing accurate simulation, in-depth analysis, and efficient research on the process of mineral adsorption of heavy metals, providing strong support for practical applications and the development of related fields. Description of the Drawings

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0058] Figure 1 It is a step flowchart of a method for constructing and analyzing a mineral adsorption heavy metal molecular dynamics algorithm of the present invention. Detailed implementation manners

[0059] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will further describe this application in detail with reference to the accompanying drawings and specific embodiments.

[0060] As Figure 1 shown, in this embodiment, a method for constructing and analyzing a mineral adsorption heavy metal molecular dynamics algorithm includes the following steps:

[0061] This method includes:

[0062] Step S1: Perform multi-level characterization on minerals and heavy metals, covering from the microscopic atomic and molecular levels to the macroscopic material property level. Use spectroscopic techniques to analyze the chemical states of surface functional groups of minerals and heavy metal ions, and use electron microscopy techniques to observe the microscopic structure and surface morphology of minerals;

[0063] In this embodiment, specifically, X-ray photoelectron spectroscopy is used to analyze the chemical states of surface functional groups of minerals, which can accurately judge the types of chemical active sites on the mineral surface, such as hydroxyl groups, carboxyl groups, etc., and their possible interaction forms with heavy metal ions. At the same time, electron microscopy techniques, including scanning electron microscopy and transmission electron microscopy, are used to observe the microscopic structure of minerals, such as crystal structure defects, crystal plane orientations, and surface morphology, such as roughness, pore structure, etc.; at the macroscopic material property level, the overall property information of minerals can be obtained through means such as specific surface area analysis and density measurement; these characterization results provide a rich data basis for the construction of subsequent physical models.

[0064] Step S2: Design and construct a physical model of mineral adsorption of heavy metals according to the characterization results. The physical model includes the active sites on the mineral surface and the hydrated shell structure of heavy metal ions. The active sites include not only different types of functional groups, but also their spatial distribution and activity degree;

[0065] For heavy metal ions, consider their hydrated shell structure, which involves the arrangement of water molecules around the ions and the formation of hydrogen bond networks, and will affect the activity and diffusion ability of the ions;

[0066] and the interaction mechanism between the above two;

[0067] In this embodiment, the interaction mechanism includes electrostatic interaction, chemical adsorption, and physical adsorption; the electrostatic interaction is based on the charge property of ions, chemical adsorption involves the formation of chemical bonds between ions and surface functional groups of minerals, and physical adsorption considers weak interactions such as van der Waals forces; by comprehensively considering these factors, a physical model that can reflect the actual physical process is constructed;

[0068] In step S2 of this embodiment, during the process of designing and constructing a physical model for mineral adsorption of heavy metals, the topological structure of the mineral surface is considered. By introducing the fractal geometry theory, the mineral surface is regarded as a fractal object to describe the distribution characteristics of surface roughness and active sites;

[0069] Specifically, the box-counting method is used to determine the fractal dimension D, and the formula is:

[0070] N(∈) ∝ ∈ -D

[0071] where N(∈) is the number of boxes with side length ∈ required to cover the mineral surface;

[0072] In step S2, by calculating N(∈) at different ∈ values and performing linear fitting, the fractal dimension D is obtained, which is used to accurately characterize the complexity of the mineral surface and the distribution of active sites;

[0073] Combined with the fractal dimension D, the parameters related to surface interaction in the physical model are adjusted, such as the activity coefficient of adsorption sites; when the fractal dimension D is larger, it indicates that the mineral surface is rougher and the distribution of active sites is more complex. Accordingly, the activity coefficient of adsorption sites is increased to reflect the influence of the actual complex structure of the mineral surface on the adsorption characteristics, so that the physical model accurately reflects the adsorption characteristics of the actual mineral surface;

[0074] Step S3: Based on the constructed physical model, by introducing multi-scale physical parameters, it is mapped to a mathematical model, and information at different scales is integrated into a unified mathematical framework;

[0075] Combining step S2 and step S3, in the step of mapping the physical model to the mathematical model, for electrostatic interaction, the Poisson-Boltzmann equation is used:

[0076]

[0077] where, is the electrostatic potential, ρ is the charge density, and ∈ is the dielectric constant;

[0078] This equation is used to describe the electrostatic interaction between minerals and heavy metal ions, and can accurately describe the electrostatic potential distribution between minerals and heavy metal ions, taking into account the influence of the inhomogeneity of charge distribution on the electrostatic potential. By introducing multi-scale physical parameters, such as charge parameters at the atomic level, bond length and bond angle parameters at the molecular level, and macroscopic dielectric constants, etc., information at different scales is integrated into a unified mathematical framework. Different interactions, such as intermolecular interactions, electrostatic interactions, and external field interactions, are quantified through different mathematical expressions to form a complete mathematical model, providing an accurate mathematical basis for subsequent molecular dynamics simulations.

[0079] Step S4: Select a molecular dynamics software framework, specifically LAMMPS; integrate the mathematical model into this framework, and at the same time deeply customize the software framework by adding functional modules for accurate simulation of the adsorption process and handling of special physical phenomena.

[0080] In this embodiment, after selecting the molecular dynamics software framework in step S4, the Hamiltonian mechanics principle is used to describe the system. The Hamiltonian H of the system can be expressed as:

[0081] H = T + V;

[0082] where T is the kinetic energy and V is the potential energy;

[0083] where pi is the momentum of particle i and mi is the mass of particle i;

[0084] The potential energy V includes intermolecular interaction potential energy, electrostatic potential energy, and external field potential energy, and its specific form is determined in combination with the physical model;

[0085] The intermolecular interaction potential energy can be determined according to the intermolecular force field, such as the Lennard-Jones potential energy. The electrostatic potential energy is obtained according to the calculation result of the above Poisson-Boltzmann equation. The external field potential energy considers the contribution of external factors such as an applied electric field. Integrate the mathematical model into this software framework and at the same time deeply customize the software framework. Add specific functional modules, such as developing a special potential energy function module for the mineral adsorption heavy metal system to handle the complex interactions between different types of minerals and heavy metal ions; add a monitoring module for the structural changes during the adsorption process, which can record in real time the changes in adsorption sites and the changes in adsorption amounts, etc.

[0086] In this embodiment, when selecting the molecular dynamics software framework, the parallel computing module of the software is optimized by combining task parallelism and data parallelism.

[0087] For task parallelism, the molecular dynamics simulation process is decomposed into multiple independent subtasks, such as calculating the forces between particles, updating the particle positions, calculating the total energy of the system, etc. For the calculation of forces, different calculation methods are adopted according to different distance ranges. The direct summation method is used for short-range forces, and the fast multipole method is used for long-range forces, etc. The subtasks are assigned to different computing cores, and the Message Passing Interface is used for communication and coordination between tasks to ensure data synchronization and calculation coordination between cores;

[0088] For data parallelism, the particle data in the simulation system is divided into multiple data blocks according to the distribution of particles in space, and the data blocks are assigned to different computing units, such as GPU threads. For GPU parallel computing, the CUDA (Compute Unified Device Architecture) or OpenCL (Open Computing Language) is used; according to different hardware architectures and characteristics of computing tasks, the allocation strategies of tasks and data are dynamically adjusted; for example, on a GPU with high parallel computing power, more data blocks can be assigned to it for processing to maximize the utilization of computing resources and improve the performance and scalability of the simulation calculation.

[0089] Step S5: Combine environmental factors and couple multiple physical fields. The environmental factors include pH value, ionic strength, and applied electric field, and the multiple physical fields include but are not limited to electromagnetic field and temperature field. Perform molecular dynamics simulation to simulate the adsorption process under different conditions;

[0090] When performing molecular dynamics simulation, for the coupling of different environmental factors, a multi-physics field coupling algorithm is used to truly simulate the adsorption process under actual environments;

[0091] For the coupling of electromagnetic field and molecular dynamics, the Lorentz force F = q(E + v×B) is used, where q is the particle charge, E is the electric field strength, v is the particle velocity, and B is the magnetic induction intensity; update the force on the particle and add this force to the Newton's equation of motion in molecular dynamics F total = F mol + F; where F mol is the intermolecular force. At the same time, the Maxwell's equations and where ρ is the charge density, μ 0 is the vacuum permeability, ∈ 0 is the vacuum permittivity, and J is the current density; are used to describe the evolution of the electromagnetic field and realize the two-way coupling of the electromagnetic field and molecular dynamics. For the coupling of temperature field and molecular dynamics, the energy balance equation where C u is the specific heat capacity at constant volume and q is the heat flux density, is used to consider the evolution of the temperature field. By modifying the energy equation in molecular dynamics E = E mol + E thermal ; where E molis the intermolecular interaction energy, E thermal is the thermal energy. Considering the thermal effect, the coupling of the temperature field and molecular dynamics is realized; in this way, the adsorption process under different temperature, electric field, and magnetic field environments can be simulated, and the influence of environmental factors on the adsorption process can be studied;

[0092] In step S5 of this embodiment, when performing molecular dynamics simulation, for the time evolution of the system, the Nose-Hoover chain thermostat algorithm is adopted, and its equation of motion is:

[0093]

[0094] where pi is the momentum of particle i, qi is the position of particle i, εj is the thermostat variable, Qj is the mass parameter of the thermostat, K B is the Boltzmann constant, T is the temperature, and g is the degree of freedom;

[0095] The above algorithm can accurately control the temperature of the simulation system by adjusting the thermostat variable to make the system in the required thermal equilibrium state, ensuring the accuracy and reliability of the simulation results;

[0096] Step S6: Comprehensively analyze the simulation results. The specific analysis methods include the principles of statistical mechanics, thermodynamics, and kinetics, analyze the energy change, kinetic characteristics, and thermodynamic equilibrium of the adsorption process, as well as the structural evolution and stability of the system after adsorption;

[0097] When comprehensively analyzing the simulation results in step S6 of this embodiment, calculate the Gibbs free energy G of the system and use the thermodynamic relation:

[0098] G = H - TS;

[0099] where H is the enthalpy, T is the temperature, and S is the entropy;

[0100] S is calculated according to the classical statistical mechanics formula: S = -k B ∑ i p i Inp i where p i is the probability of the initial microscopic state i of the system; by calculating the Gibbs free energy, the spontaneity and thermodynamic feasibility of the adsorption process are judged, providing a key index for the energy analysis of the adsorption process.

[0101] Analyze the kinetic characteristics of the adsorption process. For example, by calculating the mean square displacement (MSD), where: to study the diffusion behavior of particles, and by calculating the velocity autocorrelation function C u (t) = <V (t) *V (0) > to analyze the correlation of particle motion, and then infer the adsorption kinetic process.

[0102] Step S6 uses a machine learning algorithm for assisted analysis, which specifically includes the following steps:

[0103] Step S6.1: Use the characteristic quantities in the simulation results as input features, where the characteristic quantities include adsorption energy, adsorption capacity, particle position, velocity, etc.; use the target performance (such as adsorption rate, adsorption equilibrium time, adsorption stability, etc.) as the output label to construct a data set;

[0104] Step S6.2: Use the TensorFlow deep learning framework to build a neural network model. According to the complexity of the input features and output labels, select a multi-layer perceptron and a convolutional neural network as the network structure, and select appropriate numbers of layers and neurons according to the size and characteristics of the data set;

[0105] Step S6.3: Use the stochastic gradient descent algorithm as the optimizer to train the model by minimizing the loss function. During the training process, divide the data set into a training set (70%), a validation set (20%), and a test set (10%); during the training process, divide the data set into a training set, a validation set, and a test set, monitor the overfitting situation of the model through the validation set, and evaluate the performance of the model according to the test set;

[0106] Step S6.4: Use the trained model to predict the adsorption performance under different conditions, which can not only provide a reference for experimental design, but also, by analyzing the weights and activation functions of the neural network, explore the potential physical mechanisms in the adsorption process and provide a basis for further optimizing the physical and mathematical models.

[0107] Embodiment

[0108] Preparation of experimental materials and equipment:

[0109] Mineral materials: Select illite as the adsorption mineral, grind the illite into powder, and sieve it to obtain particles with uniform particle size, with an average particle size of about 100 μm.

[0110] Heavy metal ions: Use lead ions (Pg 2+ ) as the target heavy metal ions, dissolve them in an aqueous solution with a certain concentration to form a lead ion solution with a concentration of 0.1 mol / L

[0111] Instrumentation:

[0112] Spectral analysis equipment: X-ray photoelectron spectrometer (XPS) and Fourier transform infrared spectrometer (FTIR).

[0113] Electron microscopes: Scanning electron microscope (SEM) and transmission electron microscope (TEM).

[0114] Molecular dynamics simulation software, Intel Xeon E5-2690 processors, each processor having 10 cores and a GPU acceleration card, e.g., NVIDIA Tesla V100, to meet the high-performance computing requirements.

[0115] Data analysis and machine learning software, using the Python programming language and the TensorFlow algorithm.

[0116] The specific steps are as follows:

[0117] Step S1: Use XPS to analyze the surface of illite and find that there are a large number of hydroxyl groups (-OH) and silicon-oxygen tetrahedral structures on its surface. Through FTIR spectroscopy, the vibration modes of functional groups are further confirmed, indicating the presence of chemical bonds such as Si-O-Si and Al-O;

[0118] Use TEM to observe the microstructure of illite, showing a typical layered structure with a layer spacing of about Observe its surface morphology through SEM and find that it has a certain roughness;

[0119] Use a specific surface area analyzer to measure the specific surface area of illite, and obtain a specific surface area of about 300 m 2 / g, and determine the density of illite to be about 2.5 g / cm 3 ;

[0120] Step S2: According to the characterization results, construct a physical model for the adsorption of lead ions by illite. Considering the layered structure and surface functional groups of illite, place the lead ions in a solution environment and consider its hydration shell structure; for the electrostatic interaction, the charge of the lead ions will produce an electrostatic attraction with the negative charge sites on the surface of illite; chemical adsorption may occur between the hydroxyl groups and the lead ions to form coordination bonds. Physical adsorption is based on van der Waals forces, considering the short-range attraction between the lead ions and the surface of illite;

[0121] Introduce fractal geometry theory to analyze the surface of illite. Determine the fractal dimension through the box-counting method. Divide the surface of illite into grids with different side lengths ∈, and count the number of boxes N(∈) required to cover its surface. In the case where the value range of ∈ is from to , by measuring N(∈) under different conditions multiple times and performing linear fitting, the fractal dimension is obtained to be about 2.3; according to this fractal dimension, the adsorption site activity coefficient is increased to reflect the influence of the surface roughness of illite on adsorption;

[0122] Step S3: According to the charge distribution on the mineral surface and the charge distribution of lead ions, determine the charge density to be 1.5×10 - 6 C / m 3, and according to the dielectric constant of 80 of the aqueous solution, the distribution of the electrostatic potential is obtained by solving the equation; the Hamiltonian of the system is calculated based on the mass of lead ions of 200.59 amu, the mass of illite atoms, and the initial velocity of 80 m / s;

[0123] Step S4: Integrate the above mathematical model into LAMMPS, develop a special potential energy function module for the illite-lead ion system, and add a structure change monitoring module to record in real time the distribution of adsorption sites of lead ions on the illite surface and the change of adsorption amount over time;

[0124] Step S5: Set the initial temperature to 300 K, use the Nose-Hoover chain thermostat algorithm to control the temperature, and set the thermostat mass parameter to 1.0;

[0125] Set the initial ionic strength to 0.1 mol / L, place the lead ion solution and illite particles in a simulation box under periodic boundary conditions, and set the size of the simulation box to

[0126] Set the applied electric field strength to 100 V / m and the magnetic field strength to 0.1 T, and realize the coupling of the electromagnetic field and molecular dynamics through the Lorentz force and Maxwell's equations;

[0127] Set the initial pH value to 7, which is achieved by adjusting the proton concentration, and consider the changes in the surface charge of illite and the chemical form of lead ions at different pH values;

[0128] Start the molecular dynamics simulation, set the time step to 1 fs, and run the simulation duration for 100 ps; according to the motion equation and temperature control of the Nose-Hoover chain thermostat algorithm, ensure that the system is in a thermal equilibrium state; at the same time, through the coupled energy balance equation and the modified energy equation (assumed to be C u is 1 J / (mol·K)), consider the coupling of the temperature field and molecular dynamics, and update the energy and temperature of the system in real time;

[0129] Step S6: The Gibbs free energy results calculated at different simulation time points show that the G value gradually decreases during the adsorption process, indicating that the adsorption process is spontaneous thermodynamically;

[0130] Analyze the diffusion behavior of lead ions on the illite surface and find that as time goes by, the MSD of lead ions gradually decreases, indicating that its movement is restricted and tends to the adsorption state; at the same time, calculate the velocity autocorrelation function, which shows the change of the correlation of lead ion movement and reflects the transformation of the kinetic behavior during the adsorption process.

[0131] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A molecular dynamics algorithm construction and analysis method for heavy metal adsorption by minerals, characterized in that: The method includes: Step S1, characterizing minerals and heavy metals at multiple levels, from the microscopic atomic and molecular level to the macroscopic material property level, using spectroscopy to analyze the chemical state of mineral surface functional groups and heavy metal ions, and using electron microscopy to observe the microstructure and surface morphology of minerals; Step S2: designing and constructing a physical model of heavy metal adsorption by minerals based on the characterization results, wherein the physical model includes active sites on the surface of the minerals, the hydration shell structure of the heavy metal ions, and the interaction mechanism between the active sites on the surface of the minerals and the hydration shell structure of the heavy metal ions; Step S3, mapping the physical model into a mathematical model in combination with multi-scale physical parameters, integrating information at different scales into a unified mathematical framework; Step S4: Select a molecular dynamics software framework, integrate the mathematical model into the framework, and add functional modules to the software framework to achieve deep customization; Step S5, combining environmental factors and coupling multiple physical fields to perform a molecular dynamics simulation process to simulate the adsorption process under different conditions; Step S6: Comprehensively analyze the simulation results. The specific analysis methods include statistical mechanics, thermodynamics and kinetics principles to analyze the energy changes, kinetic characteristics and thermodynamic equilibrium of the adsorption process, as well as the structural evolution and stability of the system after adsorption.

2. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: In step S2, the interaction mechanism includes electrostatic effect, chemical adsorption and physical adsorption; Combining step S2 and step S3, in the step of mapping the physical model to the mathematical model, the Poisson-Boltzmann equation is used for the electrostatic interaction: in, is the electrostatic potential, ρ is the charge density, and ∈ is the dielectric constant; This equation is used to describe the electrostatic interaction between minerals and heavy metal ions, taking into account the effect of charge distribution on the electrostatic potential.

3. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: After the molecular dynamics software framework is selected in step S4, the system is described using the Hamiltonian mechanics principle, and the Hamiltonian H of the system can be expressed as: H = T + V; Among them, T is kinetic energy and V is potential energy; Among them, pi is the momentum of particle i, mi is the mass of particle i; The potential energy V includes intermolecular interaction potential energy, electrostatic potential energy and external field potential energy, and its specific form is determined in combination with the physical model.

4. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: In step S5, when performing molecular dynamics simulation, the Nose-Hoover chain heat bath algorithm is used for the time evolution of the system, and its motion equation is: Where pi is the momentum of particle i, qi is the position of particle i, εj is the heat bath variable, Qj is the mass parameter of the heat bath, and K B is the Boltzmann constant, T is the temperature, and g is the degree of freedom; The above algorithm is used to control the temperature of the simulation system to ensure that the system is in the desired thermal equilibrium state.

5. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: When the simulation results are comprehensively analyzed in step S6, the Gibbs free energy G of the system is calculated using the thermodynamic relationship: G = H-TS; Among them, H is enthalpy, T is temperature, and S is entropy; S is based on the classical statistical mechanics formula: S = -k B ∑ i p i Inp i Calculate, where p i is the probability of the initial microscopic state i of the system.

6. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: In the process of designing and constructing the physical model of mineral adsorption of heavy metals, the topological structure of the mineral surface is considered in step S2. By introducing the fractal geometry theory, the mineral surface is regarded as a fractal object to describe the surface roughness and the distribution characteristics of active sites; The box counting method is used to determine the fractal dimension D, and the formula is: N(∈)∝∈ -D ； Where N(∈) is the number of boxes with side length ∈ required to cover the mineral surface; In step S2, N(∈) is calculated under different ∈ conditions and linear fitting is performed to obtain a fractal dimension D, which is used to accurately characterize the complexity of the mineral surface and the distribution of active sites.

7. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 6, characterized in that: The parameters related to surface interaction in the physical model are adjusted in combination with the fractal dimension D, so that the physical model can accurately reflect the adsorption characteristics of the actual mineral surface.

8. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: When selecting a molecular dynamics software framework, the parallel computing module of the software is optimized by combining task parallelism and data parallelism; For task parallelism, the molecular dynamics simulation process is decomposed into multiple independent subtasks, and the subtasks are assigned to different computing cores, using the message passing interface for communication and coordination between tasks; For data parallelism, the particle data in the simulation system is divided into multiple data blocks, and the data blocks are assigned to different computing units, and parallel computing is performed using a unified computing device architecture; Dynamically adjust the task and data allocation strategy according to different hardware architectures and the characteristics of computing tasks.

9. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: The multiple physical fields in step S5 include but are not limited to electromagnetic fields and temperature fields; The environmental factors include pH value, ionic strength and external electric field; When performing molecular dynamics simulations, multi-physics field coupling algorithms are used to couple different environmental factors and truly simulate the adsorption process in actual environments.

10. The method for constructing and analyzing a molecular dynamics algorithm for heavy metal adsorption by minerals according to claim 1, characterized in that: The step S6 uses a machine learning algorithm to assist in the analysis, and specifically includes the following steps: Step S6.1, using the feature quantities in the simulation results as input features and the target performance as output labels to construct a data set; Step S6.2, use the TensorFlow deep learning framework to build a neural network model, select multi-layer perceptron and convolutional neural network as the network structure, and select the appropriate number of layers and neurons according to the size and characteristics of the data set; Step S6.3, using the stochastic gradient descent algorithm as the optimizer, training the model by minimizing the loss function. During the training process, the data set is divided into a training set, a validation set, and a test set. The overfitting of the model is monitored by the validation set, and the performance of the model is evaluated based on the test set. Step S6.4: Use the trained model to predict the adsorption performance under different conditions, and at the same time, explore the potential physical mechanism of the adsorption process by analyzing the weights and activation functions of the neural network.

Images