Construction method of physical characteristics based on self-assembled monolayer interface heat conduction regulation
By constructing physical feature engineering of self-assembled monolayer interfaces, the problem of lack of relevant features for the thermal conductivity of self-assembled monolayer interfaces is solved. This enables the calculation of interface thermal transport properties and dataset preparation under arbitrary structures, thereby improving the fitting effect and robustness of machine learning models.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI JIAOTONG UNIV
- Filing Date
- 2025-05-27
- Publication Date
- 2026-06-26
AI Technical Summary
The lack of relevant physical feature engineering for the thermal conductivity of self-assembled monolayer interfaces in existing technologies has led to a lack of in-depth understanding of their regulation mechanisms among researchers. As a result, machine learning models have shown weak performance and poor robustness, failing to fully tap their data-driven potential.
By constructing a physical feature engineering model based on a self-assembled monolayer interface, including molecular dynamics simulation, data statistical operations, vibrational spectrum analysis, and vibrational spectrum overlap area calculation, a strongly correlated physical feature dataset is obtained for the training and development of machine learning models.
It realizes the calculation of physical characteristics of thermal transport properties at the interface of self-assembled monolayers and the automated preparation of datasets, which improves the fitting accuracy and robustness of machine learning models and guides data-driven research on complex structures.
Smart Images
Figure CN120564863B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of materials genome engineering technology, and in particular relates to a method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers. Background Technology
[0002] From heat dissipation in micro / nano devices, solar thermal evaporation, and nanofluids to nanoparticle-assisted photothermal therapy, micro / nano interface thermal transport is crucial in these applications. When the component length of the interface system is as small as the micro / nano scale, or even comparable to the mean free path of phonons, interfacial thermal resistance plays an important role in heat transfer.
[0003] Therefore, self-assembled monolayers are a mature interfacial material, and using self-assembled monolayers to modify interfaces is a common strategy for improving interfacial thermal conductivity. For example, CN111477744A discloses a metal-self-assembled monolayer-organic semiconductor composite structure and its preparation method, using a sulfur-based monolayer as a medium. By forming chemical bonds with the metal layer and connecting with the organic semiconductor layer, the interfacial thermal conductivity is significantly improved. From a molecular dynamics perspective, the potential relationship between interfacial forces and heat transfer can be explained; stronger interactions can promote more efficient energy transfer at the interface, such as the electrostatic interaction of polar functional groups, which can enhance energy exchange between interfaces. Furthermore, research has shown that vibrational spectral coupling is another important factor affecting interfacial heat transfer. Due to different compositions and bonding properties, the vibrational spectra between interfaces often have significant mismatches, resulting in extremely low interfacial thermal conductivity.
[0004] The Materials Genome Initiative has made data-driven prediction of new material properties a new research paradigm. In this data-driven process, feature engineering, especially physical feature engineering, is crucial for building and developing relevant machine learning models. In recent years, researchers have developed various physical feature engineering methods for different properties in different materials fields for theoretical research and development. For example, CN118821504A discloses a method for predicting the temperature field in additive manufacturing by integrating physical features and physical constraints. It establishes a physically constrained temperature field prediction model for additive manufacturing using a training dataset based on physical feature engineering. CN114021481A discloses a creep fatigue life prediction method based on a fused physical neural network, specifically disclosing a material creep fatigue life model that integrates physical feature engineering to calculate extended features. In conclusion, performance improvement in data-driven research and development is inseparable from the establishment of feature engineering, especially the extraction of data features or model development that are directly physically related to the target (material) properties.
[0005] However, existing experimental and computational data on the interfacial thermal conductivity of self-assembled monolayers are very limited, resulting in a lack of in-depth understanding of their regulatory mechanisms. Furthermore, no related physical feature engineering methods have been publicly reported. In addition, feature engineering for organic molecules (molecular fragments) remains limited to string features, graph features, and general physical features. These features have weak correlations with interfacial thermal transport physics, leading to weak performance and poor robustness of machine learning models. These shortcomings mean that the data-driven potential of self-assembled monolayers in interfacial thermal conductivity regulation has not yet been fully explored.
[0006] Therefore, for the millions of design candidates for self-assembled monolayer molecular structures, there is an urgent need for a physical feature engineering method based on the interfacial thermal conductivity regulation of self-assembled monolayers to assist researchers in carrying out new paradigm research on machine learning. Summary of the Invention
[0007] The purpose of this invention is to overcome the shortcomings of existing methods for constructing physical feature engineering methods based on the control of interfacial thermal conductivity in self-assembled monolayers, thereby providing a method for constructing physical feature engineering methods for the interfacial thermal transport properties of self-assembled monolayers with arbitrary structures. Through this method, the physical feature calculation of interfacial thermal transport properties of self-assembled monolayers with arbitrary structures and the automated preparation of interfacial thermal transport physical datasets can be achieved, thus assisting researchers in conducting data-driven research such as machine learning on self-assembled monolayers with complex structures.
[0008] The objective of this invention can be achieved through the following technical solutions:
[0009] This invention first provides a method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers, comprising the following steps:
[0010] S1: Construct a simulation system for interfacial thermal transport based on a self-assembled monolayer, a gold substrate, and water molecules;
[0011] S2: Perform molecular dynamics simulations on the interface thermal transport simulation system constructed in S1 to obtain physical dynamic information matrices including interface interaction energy information matrix, vibrational spectrum physical information, energy, and structural thickness.
[0012] S3: Perform data statistical operations on all the physical dynamic information matrices obtained in S2 to obtain the first physical feature dataset;
[0013] Peak-finding operations are performed on the vibrational spectrum physical information of the self-assembled monolayer obtained in S2 to obtain the second physical feature dataset;
[0014] The vibration spectrum physical information obtained from S2 is used to calculate the vibration spectrum overlap area to obtain the third physical feature dataset;
[0015] S4: Summarize the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset obtained in S3, and finally construct a physical feature engineering based on the thermal conductivity regulation of the interface of a self-assembled monolayer.
[0016] Furthermore, in step S1, the monomer chain structure of the self-assembled monolayer in the interfacial thermal transport simulation system is -S(CH2). n -X, the overall structure of the interfacial heat transport simulation system is Au-S(CH2). n -X-H2O; where X is the end group structure.
[0017] Furthermore, the end-group structure is obtained by downloading molecular fragment information from a public database or by designing it independently.
[0018] Furthermore, in step S2, a hybrid force field distribution is used in the molecular dynamics simulation.
[0019] Furthermore, the bond interactions between sulfur atoms in the self-assembled monolayer and gold atoms in the gold substrate were simulated using the Morse force field.
[0020] Furthermore, the self-assembled monolayer molecules were simulated using the PCFF potential, and the water molecules were simulated using the TIP3P potential.
[0021] Furthermore, the non-bonded interactions between the gold atoms, the self-assembled monolayer molecules, and the water molecules were simulated using the LJ potential.
[0022] Furthermore, in step S2, the long-range electrostatic interactions of the simulated system in the molecular dynamics simulation are calculated using the PPPM method.
[0023] Furthermore, in step S2, the time step of the simulation system in the molecular dynamics simulation is set to 0.2-0.3 fs, preferably 0.25 fs.
[0024] Further, in step S2, the gold substrate, the self-assembled monolayer, and water are grouped, and the total energy and force interactions between different groups of atoms at the interface are calculated as descriptors of the interface interactions. The interface interaction energy E in the interface interaction energy information matrix is obtained using formula (1):
[0025]
[0026] In the formula, F ij Δd is the force exerted by atom j in the gold substrate or water molecule on atom i in the self-assembled monolayer. j It is the displacement of atom j; F jiIt is the force exerted by atom i in the self-assembled monolayer on atom j in the gold substrate or water molecule, Δd i It is the displacement of atom i.
[0027] Further, in step S2, the vibration spectrum VDOS in the vibration spectrum physical information is obtained by the Fourier transform of the velocity autocorrelation function (VACF), specifically using formula (2):
[0028]
[0029] In the formula, ω is the vibration wavenumber, v i Here, t represents the atomic velocity, and t0 and t represent the initial time and the associated time, respectively.
[0030] Furthermore, in step S2, the energy includes, but is not limited to, bond energy, angular energy, dihedral angle energy, LJ potential energy, and electrostatic interaction energy, specifically obtained using formula (3):
[0031]
[0032] In the formula, E total E bonds E angles E dihedrals E LJ and E Coulombic These represent the total energy of the system, bond energy, angular energy, dihedral angle energy, LJ potential energy, and electrostatic interaction energy, respectively.
[0033] Further, in step S2, the structural thickness includes the gold substrate thickness, the length of the self-assembled monolayer, and the thickness of the water molecule layer. The dynamic physical information L of the structural thickness is obtained using formula (4):
[0034] L = max(P) z )-min(P z ), (4);
[0035] In the formula, P z It is the position matrix of atoms in the z-direction of the simulated system.
[0036] Furthermore, in step S3, the data statistical operations include seven types of data statistical operations, specifically including average, maximum, minimum, variance, standard deviation, geometric mean, and harmonic mean. The number of physical features in the first physical feature dataset is seven times the number of physical matrices. Different statistical operations can explore the correlation sensitivity between the physical feature data and interface thermal conductivity from different data statistical perspectives, greatly enriching the feature set and facilitating the training of machine learning models.
[0037] Furthermore, in step S3, the second physical feature dataset includes the vibrational frequency and density of states value corresponding to the peak value after the peak-finding operation. The distribution of the vibrational spectrum is a key tool for understanding and predicting the thermal transport behavior of materials. The peak-finding operation can effectively find the spectral positions of the density of states distribution. Therefore, the data information obtained through peak-finding will also help in the training and development of subsequent interface thermal transport machine learning models.
[0038] Furthermore, in step S3, the third physical feature dataset includes the vibrational spectrum overlap area between the self-assembled monolayer and the gold substrate, as well as the vibrational spectrum overlap areas between the self-assembled monolayer and the stretching spectrum, bending spectrum, and vibrational spectrum of water molecules, respectively. Since the materials at the interface are often different (gold and water), they have different phonon vibration modes, i.e., different vibrational spectra (states). The self-assembled layer can effectively achieve vibrational spectrum coupling between different materials, and this coupling can be considered as an increase in the overlap area of the vibrational spectra of the two materials.
[0039] Furthermore, in step S4, the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset are combined to form a physical feature engineering that can be used for training machine learning models.
[0040] This invention also provides a physical feature engineering construction system based on the interfacial thermal conductivity regulation of self-assembled monolayers, the system specifically comprising:
[0041] The modeling module is used to construct an interfacial thermal transport simulation system based on a self-assembled monolayer, a gold substrate, and water molecules.
[0042] The molecular dynamics simulation module is used to perform molecular dynamics simulations on interfacial thermal transport simulation systems, and to obtain physical dynamic information matrices including interfacial interaction energy information matrix, vibrational spectrum physical information, various types of energy, and the structural thickness of each component.
[0043] The physical feature acquisition module is used to construct a first physical feature dataset, a second physical feature dataset, and a third physical feature dataset based on the physical dynamic information matrix.
[0044] The physical feature engineering construction module is used to summarize the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset to construct physical feature engineering based on the thermal conductivity regulation of the interface of self-assembled monolayer.
[0045] The present invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, is used to perform a physical feature engineering construction method based on the thermal conductivity regulation of a self-assembled monolayer interface.
[0046] Compared with the prior art, the present invention has the following beneficial effects:
[0047] (1) Based on the collection of physical matrix data and the creation of physical feature datasets under molecular dynamics simulation, this invention realizes the efficient construction of physical feature engineering based on the thermal conductivity of self-assembled monolayer interfaces. For the design candidate group of millions of self-assembled monolayer molecular structures, it is of great significance for carrying out data-driven design of the thermal conductivity of self-assembled monolayer interfaces.
[0048] (2) The physical feature engineering of the present invention does not stop at string features, graph features and general physical features, but is based on vibrational spectrum information and data such as interface interaction energy, energy and structural thickness that are strongly related to interface thermal transport. It accurately describes the physical dynamic features of the interface thermal transport process and improves the fitting accuracy of machine learning models in regulating the interface thermal conductivity of self-assembled monolayers.
[0049] (3) This invention solves the problem that there is still no physical feature engineering based on interface thermal conductivity control in the current design of self-assembled monolayer structures. It realizes the calculation of physical features of interface thermal transport properties of self-assembled monolayers under arbitrary structures and the automated preparation of interface thermal transport physical datasets. It can effectively assist researchers in carrying out data-driven research such as machine learning on self-assembled monolayers under complex structures.
[0050] (4) The physical feature engineering construction method of the present invention unifies the process of obtaining physical features in the theoretical calculation of the thermal conductivity of self-assembled monolayer interfaces, and realizes full program automation. It has guiding significance in the prediction of the thermal conductivity of self-assembled monolayer interfaces in the field of materials genome engineering.
[0051] (5) The physical feature engineering construction method of the present invention has strong applicability. In the simulation calculation process, the calculation settings such as time step, number of running steps, cutoff radius, force field parameters, etc. can be changed in accordance with the calculation common sense. The physical information matrix and related physical feature data set associated with the thermal conductivity of the self-assembled monolayer interface can also be obtained. The data set can also be modified according to the design requirements, such as different types of energy dynamic matrix, number of peaks in the vibration spectrum of self-assembled monolayer, etc.
[0052] (6) The physical feature engineering construction method of the present invention can overcome the shortcomings of traditional material research and development methods, realize the physical feature engineering of the thermal conductivity of self-assembled monolayer interfaces and its program automation, greatly improve the fitting effect of the self-assembled monolayer interface heat transport machine learning model, and save costs and time. Attached Figure Description
[0053] Figure 1 This is a flowchart illustrating the physical feature engineering construction method of the present invention.
[0054] Figure 2 This is a schematic diagram of the physical feature engineering construction system of the present invention.
[0055] Figure 3 This is a demonstration diagram of the interfacial heat transport simulation system in Embodiment 2 of the present invention.
[0056] Figure 4 This is a demonstration diagram of the dynamic physical matrix of interface interaction energy in the physical feature engineering of Embodiment 2 of the present invention.
[0057] Figure 5 This is a demonstration diagram of the vibration spectrum calculation results in the physical feature engineering of Embodiment 2 of the present invention.
[0058] Figure 6 This is a demonstration diagram of vibration spectrum peak finding in the physical feature engineering of Embodiment 2 of the present invention.
[0059] Figure 7 This is a demonstration diagram showing the calculation results of the vibrational spectrum overlap area between the self-assembled monolayer and gold in the physical feature engineering of Embodiment 2 of the present invention.
[0060] Figure 8 This is a demonstration diagram showing the calculation results of the vibrational spectrum overlap area between the self-assembled monolayer and water in the physical feature engineering of Embodiment 2 of the present invention.
[0061] Figure 9 This is a comparative demonstration diagram illustrating the improvement of machine learning model training results under physical feature engineering in Embodiment 2 of the present invention.
[0062] Figure 10 This is the training result of the machine learning extreme gradient booster model after using the physical feature engineering set in Embodiment 2 of the present invention.
[0063] Explanation of markings in the diagram:
[0064] 1-Modeling module, 2-Molecular dynamics simulation module, 3-Physical feature acquisition module, 4-Physical feature engineering construction module. Detailed Implementation
[0065] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments. Component models, material names, connection structures, control methods, algorithms, and other features not explicitly stated in this technical solution are considered common technical features disclosed in the prior art.
[0066] The purpose of this invention is to address the current lack of physical feature engineering based on interfacial thermal conductivity control in the design of self-assembled monolayer structures. The physical feature engineering construction method of this invention enables the calculation of physical features of the interfacial thermal transport properties of self-assembled monolayers under arbitrary structures and the automated preparation of interfacial thermal transport physical datasets. This assists researchers in conducting data-driven research such as machine learning on self-assembled monolayers with complex structures, and provides theoretical support and experimental basis for related applications.
[0067] The first aspect of this invention provides a method for constructing physical features based on the interfacial thermal conductivity modulation of self-assembled monolayers, such as... Figure 1 As shown, it includes the following steps:
[0068] S1: Construct a simulation system for interfacial thermal transport based on a self-assembled monolayer, a gold substrate, and water molecules;
[0069] S2: Perform molecular dynamics simulations on the interface thermal transport simulation system constructed in S1 to obtain physical dynamic information matrices including interface interaction energy information matrix, vibrational spectrum physical information, energy, and structural thickness.
[0070] S3: Perform data statistical operations on all the physical dynamic information matrices obtained in S2 to obtain the first physical feature dataset;
[0071] Peak-finding operations are performed on the vibrational spectrum physical information of the self-assembled monolayer obtained in S2 to obtain the second physical feature dataset;
[0072] The vibration spectrum physical information obtained from S2 is used to calculate the vibration spectrum overlap area to obtain the third physical feature dataset;
[0073] S4: Summarize the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset obtained in S3, and finally construct a physical feature engineering based on the thermal conductivity regulation of the interface of a self-assembled monolayer.
[0074] In some specific embodiments, in step S1, the monomer chain structure of the self-assembled monolayer in the interfacial thermal transport simulation system is -S(CH2). n -X, the overall structure of the interfacial heat transport simulation system is Au-S(CH2). n -X-H2O; where X is the end group structure.
[0075] In some more specific embodiments, the end-group structure is obtained by downloading molecular fragment information from a public database or by designing it in-house.
[0076] In some specific implementations, in step S2, a mixed force field distribution is used in the molecular dynamics simulation.
[0077] In some more specific embodiments, the bond interactions between sulfur atoms in the self-assembled monolayer and gold atoms in the gold substrate are simulated using Morse force field simulation.
[0078] In some more specific embodiments, the self-assembled monolayer molecules are simulated using the PCFF potential, and the water molecules are simulated using the TIP3P potential.
[0079] In some specific embodiments, the nonbonded interactions between the gold atoms, the self-assembled monolayer molecules, and the water molecules are simulated using the LJ potential.
[0080] In some specific implementations, in step S2, the long-range electrostatic interactions of the simulated system in the molecular dynamics simulation are calculated using the PPM method.
[0081] In some specific implementations, in step S2, the time step of the simulation system in the molecular dynamics simulation is set to 0.2-0.3 fs, preferably 0.25 fs.
[0082] In some specific implementations, in step S2, the interfacial interaction energy E in the interfacial interaction energy information matrix is obtained using formula (1):
[0083]
[0084] In the formula, F ij Δd is the force exerted by atom j in the gold substrate or water molecule on atom i in the self-assembled monolayer. j It is the displacement of atom j; F ji It is the force exerted by atom i in the self-assembled monolayer on atom j in the gold substrate or water molecule, Δd i It is the displacement of atom i.
[0085] In some specific implementations, in step S2, the vibration spectrum VDOS in the vibration spectrum physical information is obtained using formula (2):
[0086]
[0087] In the formula, ω is the vibration wavenumber, v i Here, t represents the atomic velocity, and t0 and t represent the initial time and the associated time, respectively.
[0088] In some specific implementations, the energy in step S2 includes, but is not limited to, bond energy, angular energy, dihedral angle energy, LJ potential energy, and electrostatic interaction energy, specifically obtained using formula (3):
[0089]
[0090] In the formula, Etotal E bonds E angles E dihedrals E LJ and E Coulombic These represent the total energy of the system, bond energy, angular energy, dihedral angle energy, LJ potential energy, and electrostatic interaction energy, respectively.
[0091] In some specific embodiments, in step S2, the structural thickness includes the gold substrate thickness, the length of the self-assembled monolayer, and the thickness of the water molecule layer. The dynamic physical information L of the structural thickness is obtained using formula (4):
[0092] L = max(P) z )-min(P z ), (4);
[0093] In the formula, P z It is the position matrix of atoms in the z-direction of the simulated system.
[0094] In some specific implementations, step S3 includes seven types of data statistical operations, specifically including average, maximum, minimum, variance, standard deviation, geometric mean, and harmonic mean.
[0095] In some specific implementations, in step S3, the number of physical features in the first physical feature dataset is seven times the number of physical matrices.
[0096] In some specific implementations, in step S3, the second physical feature dataset includes the vibration frequency and density of states corresponding to the peak value after the peak-finding operation.
[0097] In some specific implementations, in step S3, the third physical feature dataset includes the vibrational spectrum overlap area between the self-assembled monolayer and the gold substrate, as well as the vibrational spectrum overlap area between the self-assembled monolayer and the stretching spectrum, bending spectrum, and vibrational spectrum of water molecules, respectively.
[0098] In some specific implementations, in step S4, the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset are aggregated into a dataset to form a physical feature engineering that can be used for training machine learning models.
[0099] In some specific implementations, in step S4, the physical feature engineering can be used for training machine learning models.
[0100] A second aspect of this invention provides a physical feature engineering construction system based on the self-assembled monolayer interface thermal conductivity modulation. For example... Figure 2As shown, the system specifically includes: a modeling module 1, used to construct an interfacial thermal transport simulation system based on a self-assembled monolayer, a gold substrate, and water molecules; a molecular dynamics simulation module 2, used to perform molecular dynamics simulations on the interfacial thermal transport simulation system to obtain a physical dynamic information matrix including an interfacial interaction energy information matrix, vibrational spectrum physical information, various types of energy, and the structural thickness of each component; a physical feature acquisition module 3, used to construct a first physical feature dataset, a second physical feature dataset, and a third physical feature dataset based on the physical dynamic information matrix; and a physical feature engineering construction module 4, used to summarize the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset to construct a physical feature engineering based on the interfacial thermal conductivity regulation of a self-assembled monolayer.
[0101] A third aspect of the present invention provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a physical feature engineering construction method based on the thermal conductivity regulation of a self-assembled monolayer interface.
[0102] The storage medium of this invention can be an electronic medium, magnetic medium, optical medium, electromagnetic medium, infrared medium, or semiconductor system or propagation medium. The storage medium may also include semiconductor or solid-state memory, magnetic tape, removable computer disk, random access memory (RAM), read-only memory (ROM), hard disk, and optical disc. Optical discs may include optical disc-read-only memory (CD-ROM), optical disc-read-write (CD-RW), and DVD.
[0103] Each of the above embodiments can be implemented individually, or in any combination of two or more. The following detailed description of specific examples will further illustrate these embodiments.
[0104] Example 1:
[0105] This embodiment provides a method for constructing physical feature engineering based on the interfacial thermal conductivity regulation of self-assembled monolayers. This method mainly consists of two parts: the collection of physical matrix data under molecular dynamics simulations and the creation of physical feature datasets.
[0106] The physical feature engineering construction method in this embodiment specifically includes the following steps:
[0107] S1: Based on the interface thermal transport simulation system of self-assembled monolayer, gold substrate and water molecules, molecular dynamics simulation is carried out for this interface thermal transport system;
[0108] S2: During the simulation, physical dynamic information such as the interface interaction energy information matrix, vibration spectrum physical information, various types of energy, and the thickness of each component structure is saved;
[0109] S3: Create physical features for various physical dynamic information matrices through seven types of data statistical operations;
[0110] S4: Extract physical features from the physical information of the peak values of the vibration spectrum by peak finding;
[0111] S5: The quantitative characteristics of vibrational spectrum coupling strength are obtained by calculating the ratio of the vibrational spectrum overlap area between the self-assembled monolayer and other interface materials.
[0112] S6: All physical features are collected in the dataset and used to train the machine learning model.
[0113] In step S1, all simulations are automated and integrated into the LAMMPS software to perform molecular dynamics simulations. The self-assembled monolayer chain of the simulated system, collected in the physical matrix data collection, consists of three parts: the head group -S-, the chain -(CH2), and the chain-(CH2). n - and end base X.
[0114] The self-assembled monolayer molecules were simulated using the PCFF potential, while the water molecules were simulated using the TIP3P potential for thermal transport at the solid-liquid interface. Non-bonded interactions between gold atoms, self-assembled monolayer molecules, and water molecules were simulated using the LJ potential. Except for sulfur atoms covalently bonded to gold atoms, the LJ potential parameters were set using a modified universal force field (UFF). Long-range electrostatic interactions of the entire system were calculated using the PPPM method. Considering the presence of relatively light hydrogen atoms in the system, a time step of 0.25 fs was chosen.
[0115] In step S2, the interface interaction energy information matrix is obtained by the following formula (1), where F ij Δd is the force exerted by atom j in the gold substrate or water molecule on atom i in the self-assembled monolayer. j It is the displacement of atom j; F ji It is the force exerted by atom i in a self-assembled monolayer on atom j in a gold substrate or water molecule, Δd i It is the displacement of atom i. The interfacial interactions in the system are also divided into components in three dimensions.
[0116]
[0117] The physical information of the vibration spectrum is obtained by the following formula (2), where ω is the vibration wavenumber and v i Here, t represents the atomic velocity, and t0 and t represent the initial time and the associated time, respectively.
[0118]
[0119] The matrix information for each type of energy is referenced in the following formula (3), where E totalIt is the total energy of the system, which includes bond energy, angle energy, dihedral angle energy, LJ potential energy, electrostatic interaction energy, etc.
[0120]
[0121] The calculation of physical dynamic information such as the structural thickness of each component is based on the reference formula (4), where P z It is the position matrix of atoms in the z-direction under grouping in the simulation system.
[0122] L = max(P) z )-min(P z (4)
[0123] In step S3, seven types of data statistical operations are performed on physical dynamic information such as interface interaction energy information matrix, vibration spectrum physical information, various types of energy, and thickness of each component structure. The operation refers to the following formulas (5) to (11), including average value (ave), maximum value (max), minimum value (min), variance (var), standard deviation (std), geometric mean (gm), and harmonic mean (hm). The number of physical features created is seven times the number of physical matrices.
[0124]
[0125] max(x) = max(x1, x1, x1, ..., x1) n (6)
[0126] min(x) = min(x1, x1, x1, ..., x1) n (7)
[0127]
[0128]
[0129] In step S4, peak searching is performed on the vibrational spectrum physical information of the self-assembled monolayer. The number of peaks to be searched can be set by the user. The vibrational frequency and density of states corresponding to the peaks will be included in the dataset as physical features.
[0130] In step S5, when calculating the vibrational spectrum overlap area between the self-assembled monolayer and other interface materials, the vibrational spectra of the self-assembled monolayer and the other interface material (gold) are calculated separately, and then the vibrational spectrum overlap area is calculated. For the solid-liquid interface thermal transport system, the vibrational spectrum of water is further divided into three regions: the stretching spectrum (95-105 THz), the bending spectrum (48.5-58.5 THz), and the vibrational spectrum (0-30 THz) of hydroxyl groups (-OH). The vibrational spectrum overlap area of these three regions is also calculated separately.
[0131] Example 2:
[0132] This embodiment specifically provides a method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers, including the following steps:
[0133] (1) Self-assembled monolayer solid-liquid interface thermal transport systems with terminal groups X of CC(=O)NC / C(N)=N / O and C1CC[C@@H]2C[C@@H]3CCCC[C@@H]3C[C@H]2C1 were constructed. All simulations were performed using LAMMPS software via an automated program to achieve molecular dynamics simulations.
[0134] like Figure 3 As shown, the self-assembled monolayer chain structure of the solid-liquid interface thermal transport system in this embodiment consists of three parts: the head group -S-, the chain -(CH2), and the chain-(CH2). n - and end group X. Specifically, the self-assembled monolayers are arranged in an 8×8 array on the surface of a gold (111) substrate with a face-centered cubic lattice structure. Each side of the substrate consists of 25 layers of gold atoms, and 64 single chains of the self-assembled monolayers are distributed at the interface, with an interatomic distance of 0.497 nm between sulfur atoms. In addition, 2000 water molecules are compacted in the space between the two gold substrates and the self-assembled monolayers below the solid-liquid interface.
[0135] In the molecular dynamics simulation, a hybrid force field allocation was employed, where the Morse force field was used to simulate the bond interactions between sulfur atoms in the sulfur-based self-assembled layer and gold atoms in the gold substrate. The cutoff radius of the Morse force field was set to [value missing] in the calculation. The PCFF potential was used for self-assembled monolayer molecules. In the calculations of thermal transport properties based on the solid-liquid interface, the TIP3P potential was used for water molecules; its application in the thermal transport of self-assembled monolayers has been reported in previous work. Furthermore, the non-bonded interactions between gold atoms, self-assembled monolayer molecules, and water molecules were simulated using the LJ potential. Except for sulfur atoms covalently bonded to gold atoms, the LJ potential parameters were set to the modified universal force field (UFF), and the cutoff radius of the LJ potential was set to... The long-range electrostatic interactions of the entire system were calculated using the PPM method with an accuracy of 1×10⁻⁶. -5 Considering the presence of relatively light hydrogen atoms in the system, 0.25 fs was chosen as the system time step.
[0136] (2) During the simulation, physical dynamic information such as the interface interaction energy information matrix, vibration spectrum physical information, various types of energy, and the thickness of each component structure is saved. The interface interaction energy information matrix is obtained by formula (1), the vibration spectrum physical information is obtained by formula (2), the matrix information of various types of energy is obtained by formula (3), and the calculation of physical dynamic information such as the thickness of each component structure is obtained by formula (4).
[0137] Specifically, the dynamic physical matrix demonstration of interface interaction energy in the physical feature engineering of this embodiment is as follows: Figure 4 As shown, the interfacial interactions in the system are further divided into three-dimensional components. The figure illustrates the matrix information showing how the interfacial interactions of two different structural segments self-assembled monolayers on the left and right sides of the water surface change with simulation duration.
[0138] The vibration spectrum calculation results in the physical feature engineering of this invention are demonstrated as follows: Figure 5 Figures a, b, c, and d show the convergence (blue) of the self-assembled monolayer VCF of two different structural segments with correlation time and the distribution of VDOS at different frequencies (red), respectively.
[0139] (3) Physical features are created for various physical dynamic information matrices through seven types of data statistical operations. Among them, seven types of data statistical operations are performed on physical dynamic information such as interface interaction energy information matrix, vibration spectrum physical information, various types of energy, and thickness of each component structure. The operation is based on the following formulas (5) to (11), including average value (ave), maximum value (max), minimum value (min), variance (var), standard deviation (std), geometric mean (gm), and harmonic mean (hm). The number of physical features created is seven times the number of physical matrices.
[0140] (4) Physical feature extraction of the physical information of vibrational spectrum peaks through peak finding. Peak finding is performed on the physical information of the vibrational spectrum of the self-assembled monolayer. The number of peaks to be found can be set by the user. The vibrational frequencies and density of states corresponding to the peaks will be included in the dataset as physical features. The vibrational spectrum peak finding demonstration in the physical feature engineering of this embodiment is shown below. Figure 6 Figures a and b show the peak-finding results of the VDOS vibrational spectra of two different structural segments self-assembled into a single layer.
[0141] (5) The quantitative characteristics of vibrational spectral coupling strength are obtained by calculating the ratio of the vibrational spectrum overlap area between the self-assembled monolayer and other interface materials. Specifically, when calculating the vibrational spectrum overlap area between the self-assembled monolayer and gold, the vibrational spectra of the self-assembled monolayer and the other interface material (gold) are calculated separately, and then the vibrational spectrum overlap area is calculated. The calculation results of the vibrational spectrum overlap area in the physical feature engineering of this invention are demonstrated as follows: Figure 7As shown in the figure, the gold line segment represents the vibrational spectrum distribution information of the gold substrate, while the red and green lines represent the VDOS vibrational spectrum information of two different structural segments self-assembled monolayers, respectively. It can be seen from the figure that the overlap area between these two structural segments and the VDOS of the gold substrate is in the range of 0-10 THz. It can be clearly seen that the overlap between the red and gold regions is higher than that between the green and gold regions, and the overlap factors are 0.090 and 0.053, respectively.
[0142] For solid-liquid interface thermal transport systems, the vibrational spectrum of water is further divided into three regions, such as... Figure 8 As shown, the stretching spectrum (95-105 THz), bending spectrum (48.5-58.5 THz), and vibrational spectrum (0-30 THz) of the hydroxyl group (-OH) are displayed, and the overlap area of the vibrational spectra in the three regions is also calculated separately.
[0143] Vibrational spectral coupling is a key characteristic of thermal transport and another important factor affecting interfacial heat transfer. Due to different compositions and bonding properties, the vibrational spectra between interfaces often exhibit significant mismatches, resulting in extremely low interfacial thermal conductivity. Extracting key vibrational spectral coupling indices between interfaces can, to some extent, describe whether there is a significant mismatch in thermal transport between them.
[0144] (6) All acquired physical features are collected in a unified dataset to form a physical feature dataset. The physical feature dataset in this implementation case will be used for training the machine learning model.
[0145] Specifically, the training objective function in this embodiment is selected based on the thermal conductivity of the solid-liquid interface under 300 different end-group modifications. A comparative demonstration of the improvement in training results of the machine learning model under physical feature engineering in this embodiment can be found in [link to example]. Figure 9 The comparison models are Random Forest (RF), Extreme Gradient Booster (XGB), Gradient Boosting (GB) Regression, Decision Tree (DT), and Kernel Ridge (KR) Regression, and the comparison feature sets are Molecule, Mol2vec, MACC, Morgan, and cMorgan.
[0146] from Figure 9As can be seen, the root mean square error (RMSE) of the interfacial thermal conductivity machine learning model under other traditional organic molecule feature engineering sets is represented in green during 10-fold cross-validation; while the training results of each machine learning model with the physical feature engineering set provided by this invention added to the existing feature set are represented in red. Furthermore, all machine learning model results are statistical results obtained after 100 independent machine learning model training iterations to avoid outliers. The results show that the introduction of the physical feature engineering set of this invention significantly reduces the RMSE after machine learning model training, demonstrating the reliability of the physical feature engineering of this invention.
[0147] To better illustrate the training status of a specific model, Figure 10 The training results of a machine learning Extreme Gradient Booster (XGB) model using the physical feature engineering set of this invention are presented. Training employs 10-fold cross-validation, where R... 2 =0.879, RMSE=13.74, which also indicates that the training effect of the model under this physical feature engineering is very ideal.
[0148] Traditional materials research and development methods often involve extremely high costs for experimental preparation and characterization, and the lack of a set of highly correlated physical features frequently leads to poor reliability of data-driven models such as machine learning. The physical feature engineering method of this invention overcomes the shortcomings of traditional materials research and development approaches, enabling the physical feature engineering and automation of the thermal conductivity properties of self-assembled monolayer interfaces. This significantly improves the fitting performance of machine learning models for thermal transport at self-assembled monolayer interfaces, while also saving costs and time.
[0149] The above description of the embodiments is provided to enable those skilled in the art to understand and use the invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the invention should be within the protection scope of the present invention.
Claims
1. A method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers, characterized in that, Includes the following steps: S1: Construct a simulation system for interfacial thermal transport based on a self-assembled monolayer, a gold substrate, and water molecules; The monomer chain structure of the self-assembled monolayer in the interfacial thermal transport simulation system is -S(CH2). n -X, the overall structure of the interfacial heat transport simulation system is Au-S(CH2). n -X-H2O, where X is the end group structure; S2: Perform molecular dynamics simulations on the interface thermal transport simulation system constructed in S1 to obtain physical dynamic information matrices including interface interaction energy information matrix, vibrational spectrum physical information, energy, and structural thickness. The interfacial interaction energy E in the interfacial interaction energy information matrix is obtained using formula (1): (1)； In the formula, F ij It is an atom in a gold substrate or water molecules. j For atoms in self-assembled monolayers i The applied force, Δ d j It is an atom j The displacement; F ji It is the atoms in a self-assembled monolayer i Atoms in gold substrate or water molecules j The applied force, Δ d i It is an atom i The displacement; The vibration spectrum in the physical information of the vibration spectrum VDOS Using formula (2), we get: (2)； In the formula, ω The vibration wave number, v i For atomic velocity, t 0 and t These are the initial time and the associated time, respectively; The energy includes, but is not limited to, bond energy, angular energy, dihedral angle energy, LJ potential energy, and electrostatic interaction energy, which are specifically obtained using formula (3): (3)； In the formula, E total , E bonds , E angles , E dihedrals , E LJ as well as E Coulombic These represent the total energy, bond energy, angle energy, dihedral angle energy, LJ potential energy, and electrostatic interaction energy of the system, respectively. The structural thickness includes the gold substrate thickness, the length of the self-assembled monolayer, and the thickness of the water molecule layer; dynamic physical information of the structural thickness. L Using formula (4), we get: (4)； In the formula, P z It is the position matrix of atoms in the z-direction of the simulated system; S3: Perform data statistical operations on all physical dynamic information matrices obtained in S2 to obtain the first physical feature dataset; perform peak finding operations on the vibrational spectrum physical information of the self-assembled monolayer obtained in S2 to obtain the second physical feature dataset; calculate the vibrational spectrum overlap area on the vibrational spectrum physical information obtained in S2 to obtain the third physical feature dataset. S4: Summarize the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset obtained in S3, and finally construct a physical feature engineering based on the thermal conductivity regulation of the interface of a self-assembled monolayer.
2. The method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers according to claim 1, characterized in that, In step S1, the end-group structure is obtained by downloading molecular fragment information from a public database or by designing it independently.
3. The method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers according to claim 1, characterized in that, In step S2, a hybrid force field distribution is used in the molecular dynamics simulation; The bonding interactions between sulfur atoms in the self-assembled monolayer and gold atoms in the gold substrate were simulated using the Morse force field; the molecules in the self-assembled monolayer were simulated using the PCFF potential, and the water molecules were simulated using the TIP3P potential; the non-bonding interactions between the gold atoms, the self-assembled monolayer molecules, and the water molecules were simulated using the LJ potential. In the molecular dynamics simulation, the long-range electrostatic interactions of the simulated system were calculated using the PPPM method, and the time step of the simulation system was set to 0.2-0.3 fs.
4. The method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers according to claim 1, characterized in that, In step S3, the data statistical operation includes seven types of data statistical operations, specifically including average, maximum, minimum, variance, standard deviation, geometric mean, and harmonic mean.
5. The method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers according to claim 1, characterized in that, In step S3, the second physical feature dataset includes the vibration frequency and density of states corresponding to the peak value after the peak-finding operation.
6. The method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers according to claim 1, characterized in that, In step S3, the third physical feature dataset includes the vibrational spectrum overlap area between the self-assembled monolayer and the gold substrate, as well as the vibrational spectrum overlap area between the self-assembled monolayer and the stretching spectrum, bending spectrum, and vibrational spectrum of water molecules, respectively.
7. The method for constructing physical features based on the interfacial thermal conductivity regulation of self-assembled monolayers according to claim 1, characterized in that, In step S4, the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset are combined to form a physical feature engineering that can be used for training machine learning models.
8. A physical feature engineering construction system based on the interfacial thermal conductivity regulation of a self-assembled monolayer, characterized in that, The construction method based on any one of claims 1-7 specifically includes: The modeling module is used to construct an interfacial thermal transport simulation system based on a self-assembled monolayer, a gold substrate, and water molecules. The molecular dynamics simulation module is used to perform molecular dynamics simulations on interfacial thermal transport simulation systems, and to obtain physical dynamic information matrices including interfacial interaction energy information matrix, vibrational spectrum physical information, various types of energy, and the structural thickness of each component. The physical feature acquisition module is used to construct a first physical feature dataset, a second physical feature dataset, and a third physical feature dataset based on the physical dynamic information matrix. The physical feature engineering construction module is used to summarize the first physical feature dataset, the second physical feature dataset, and the third physical feature dataset to construct physical feature engineering based on the thermal conductivity regulation of the interface of self-assembled monolayer.
9. A storage medium containing computer-executable instructions, characterized in that, When executed by a computer processor, the storage medium of the computer-executable instructions is used to perform the physical feature engineering construction method based on the self-assembled monolayer interface thermal conductivity regulation as described in any one of claims 1-7.