Training method and device of hydrogen terminal diamond structure-spectrum-effect relationship prediction model

Abstract

The invention discloses a training method, device and equipment of a hydrogen terminal diamond structure-spectrum-effect relationship prediction model and a readable storage medium, and relates to the technical field of material science. Comprising the following steps: firstly, obtaining structure parameters of the hydrogen terminal diamond, and carrying out first-principle high-throughput calculation on the structure parameters to obtain property parameters of the hydrogen terminal diamond; constructing a relevance data set between the structure parameters and the property parameters; then carrying out infrared spectrum calculation on the structure parameters of the hydrogen terminal diamond to obtain spectrum data of the hydrogen terminal diamond; constructing a mapping data set between the structural parameters and the spectral data; and finally, based on the relevance data set and the mapping data set, training a hydrogen terminal diamond structure-spectrum-effect relationship prediction model. According to the scheme, the high efficiency and accuracy of hydrogen terminal diamond structure-spectrum-effect relationship prediction are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of materials science, and particularly to a training method, device, equipment and readable storage medium for a prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond. Background Art

[0002] Diamond has advantages such as high hole mobility, high thermal conductivity, and breakdown electric field, and has great application potential in the fields of high-frequency high-power devices, optoelectronic devices, quantum computing, etc. as an ultra-wide bandgap semiconductor. However, its key properties such as two-dimensional hole gas are affected by a series of factors such as surface physical and chemical morphology, doping, and heterostructure, resulting in poor stability of device performance. Secondly, the surface physical and chemical properties of diamond are complex, including various crystal planes, chemical terminations, and defects, etc., with numerous independent variables, and it is impossible to establish an accurate structure-property relationship between structural parameters and performance. Traditional experimental methods usually use first-principles calculation methods to establish a relatively accurate structure-property relationship, but it takes a long time, has a high cost, and has a large amount of calculation, making it difficult to meet the needs of large-scale screening. Therefore, it is of great significance to develop an efficient and accurate performance prediction method. Summary of the Invention

[0003] The purpose of the present invention is to provide a training method, device, equipment and readable storage medium for a prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond. First, a correlation dataset between structural parameters and property parameters is constructed, and a mapping dataset between structural parameters and spectral data is constructed. Equivalent to using structural parameters as a bridge, a mapping relationship among structural parameters, property parameters, and spectral data is constructed, that is, according to the spectral data of hydrogen-terminated diamond, its property parameters can be obtained. Then, the prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond trained based on the correlation dataset and the mapping dataset can also realize predicting its property parameters based on the spectral data of hydrogen-terminated diamond, greatly increasing the efficiency and accuracy of property prediction of hydrogen-terminated diamond.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] In a first aspect, the present invention provides a training method for a prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond, and the method includes:

[0006] Obtain the structural parameters of hydrogen-terminated diamond, and perform first-principles high-throughput calculation on the structural parameters to obtain the property parameters of hydrogen-terminated diamond;

[0007] Construct a correlation dataset between structural parameters and property parameters;

[0008] Perform infrared spectrum calculation on the structural parameters of hydrogen-terminated diamond to obtain the spectral data of hydrogen-terminated diamond;

[0009] Construct a mapping data set between structural parameters and spectral data;

[0010] Based on the correlation data set and the mapping data set, train a prediction model for the structure-spectrum-property relationship of hydrogen-terminated diamond.

[0011] In some embodiments, the structural parameters include a first surface structure, a second surface structure, and a third surface structure. Obtain the structural parameters of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the structural parameters to obtain the property parameters of hydrogen-terminated diamond, including:

[0012] Obtain the first surface structure of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the first surface structure to obtain the first parameter of hydrogen-terminated diamond;

[0013] Obtain the second surface structure of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the second surface structure to obtain the second parameter of hydrogen-terminated diamond; the second surface structure is an oxygen-hydrogen-terminated diamond surface structure;

[0014] Obtain the third surface structure of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of hydrogen-terminated diamond; the third surface structure is a hydrogen-terminated diamond heterojunction surface structure;

[0015] Perform data fusion on the first parameter, the second parameter, and the interfacial transfer charge data to obtain the property parameters of hydrogen-terminated diamond.

[0016] In some embodiments, obtain the third surface structure of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of hydrogen-terminated diamond, including:

[0017] Intercept a single layer of hexagonal boron nitride and a bilayer of gold atoms as the electron transport layer to construct the third surface structure of hydrogen-terminated diamond;

[0018] Perform first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of the hydrogen-terminated diamond heterojunction.

[0019] In some embodiments, obtain the second surface structure of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the second surface structure to obtain the second parameter of hydrogen-terminated diamond, including:

[0020] Use oxygen atoms to saturate the dangling bonds of carbon atoms on the surface of hydrogen-terminated diamond to construct the second surface structure of oxygen-hydrogen-terminated diamond;

[0021] Perform first-principles high-throughput calculations on the second surface structure to obtain the second parameter of hydrogen-terminated diamond.

[0022] In some embodiments, based on the correlation dataset and the mapping dataset, training a hydrogen-terminated diamond structure-spectrum-effect relationship prediction model includes:

[0023] Based on the correlation dataset and the mapping dataset, constructing a relationship mapping table between structural parameters, property parameters, and spectral data;

[0024] Based on the relationship mapping table, training a hydrogen-terminated diamond structure-spectrum-effect relationship prediction model.

[0025] In some embodiments, the method further includes:

[0026] Obtaining the spectral data of the hydrogen-terminated diamond to be predicted;

[0027] Taking the spectral data of the hydrogen-terminated diamond to be predicted as the eigenvalue input of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model to obtain the property parameters of the hydrogen-terminated diamond to be predicted.

[0028] In a second aspect, the present invention also provides a training device for a hydrogen-terminated diamond structure-spectrum-effect relationship prediction model, and the device includes:

[0029] A parameter calculation module, configured to obtain the structural parameters of the hydrogen-terminated diamond and perform first-principles high-throughput calculation on the structural parameters to obtain the property parameters of the hydrogen-terminated diamond;

[0030] A first construction module, configured to construct a correlation dataset between the structural parameters and the property parameters;

[0031] A spectral calculation module, configured to perform infrared spectral calculation on the structural parameters of the hydrogen-terminated diamond to obtain the spectral data of the hydrogen-terminated diamond;

[0032] A second construction module, configured to construct a mapping dataset between the structural parameters and the spectral data;

[0033] A model training module, configured to train a hydrogen-terminated diamond structure-spectrum-effect relationship prediction model based on the correlation dataset and the mapping dataset.

[0034] In a third aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model provided in the first aspect is implemented.

[0035] In a fourth aspect, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model provided in the first aspect is implemented.

[0036] In a fifth aspect, the present invention also provides a computer program product, including a computer program which, when executed by a processor, implements the training method of the hydrogen-terminated diamond structure-spectrum-property relationship prediction model provided in the first aspect.

[0037] The beneficial effects of the present invention are as follows:

[0038] In the training method of the hydrogen-terminated diamond structure-spectrum-property relationship prediction model of the present invention, first, the structural parameters of the hydrogen-terminated diamond are obtained, and first-principles high-throughput calculations are performed on the structural parameters to obtain the property parameters of the hydrogen-terminated diamond; then, a correlation data set between the structural parameters and the property parameters is constructed; then, infrared spectrum calculations are performed on the structural parameters of the hydrogen-terminated diamond to obtain the spectral data of the hydrogen-terminated diamond; then, a mapping data set between the structural parameters and the spectral data is constructed; finally, based on the correlation data set and the mapping data set, the hydrogen-terminated diamond structure-spectrum-property relationship prediction model is trained. First, a correlation data set between the structural parameters and the property parameters is constructed, and then a mapping data set between the structural parameters and the spectral data is constructed. Equivalent to using the structural parameters as a bridge, a mapping relationship among the structural parameters, the property parameters, and the spectral data is constructed, that is, according to the spectral data of the hydrogen-terminated diamond, its property parameters can be obtained. Then, the hydrogen-terminated diamond structure-spectrum-property relationship prediction model trained based on the correlation data set and the mapping data set can also predict the property parameters based on the spectral data of the hydrogen-terminated diamond, greatly increasing the efficiency and accuracy of the property prediction of the hydrogen-terminated diamond.

[0039] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly and implement it in accordance with the content of the description, the following takes the preferred embodiments of the present invention and describes them in detail with the accompanying drawings as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic flowchart of a training method of a hydrogen-terminated diamond structure-spectrum-property relationship prediction model shown in an embodiment of the present invention;

[0041] Figure 2 It is a schematic diagram of establishing a structure-property relationship between the first surface structure and the first parameter of a hydrogen-terminated diamond shown in an embodiment of the present invention;

[0042] Figure 3 It is a schematic diagram of establishing a structure-property relationship between the second surface structure of an oxygen-hydrogen-terminated diamond and the second parameter shown in an embodiment of the present invention;

[0043] Figure 4 It is a schematic diagram of the relationship between the third surface structure of a hydrogen-terminated diamond and the interfacial transfer charge data shown in an embodiment of the present invention;

[0044] Figure 5 Schematic diagram for establishing the structure-spectrum-effect relationship and prediction result diagram of the prediction model between the first surface structure and the first parameter of hydrogen-terminated diamond shown in an embodiment of the present invention;

[0045] Figure 6 Prediction result diagram of the first prediction model shown in an embodiment of the present invention;

[0046] Figure 7 Prediction result diagram of the second prediction model shown in an embodiment of the present invention;

[0047] Figure 8 Schematic diagram for analyzing the eigenvalue of infrared spectrum data by the structure-spectrum-effect relationship prediction model of hydrogen-terminated diamond shown in an embodiment of the present invention;

[0048] Figure 9 Schematic flow diagram of a training method for another structure-spectrum-effect relationship prediction model of hydrogen-terminated diamond shown in an embodiment of the present invention;

[0049] Figure 10 Schematic structural diagram of a training device for a structure-spectrum-effect relationship prediction model of hydrogen-terminated diamond shown in an embodiment of the present invention;

[0050] Figure 11 Schematic structural diagram of another training device for a structure-spectrum-effect relationship prediction model of hydrogen-terminated diamond shown in an embodiment of the present invention;

[0051] Figure 12 Schematic structural diagram of an electronic device provided in an embodiment of the present application. Detailed implementation manners

[0052] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiment", etc. in this specification mean that the described embodiment may include specific features, structures, or characteristics, but not every embodiment must include these specific features, structures, or characteristics. In addition, such expressions do not refer to the same embodiment. Further, when combining specific features, structures, or characteristics with an embodiment, it is within the knowledge of those skilled in the art to combine such features, structures, or characteristics with other embodiments even without explicit description.

[0054] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0055] In some embodiments, as Figure 1 shown, a method for training a prediction model of the structure-spectrum-effect relationship of hydrogen-terminated diamond is provided. The specific method includes:

[0056] S101, obtaining the structural parameters of hydrogen-terminated diamond, and performing first-principles high-throughput calculations on the structural parameters to obtain the property parameters of hydrogen-terminated diamond.

[0057] Among them, hydrogen-terminated diamond is diamond with hydrogen atoms adsorbed on its surface, and the structural parameters include the first surface structure, the second surface structure, and the third surface structure.

[0058] Specifically, obtaining the first surface structure of hydrogen-terminated diamond, performing first-principles high-throughput calculations on the first surface structure to obtain the first parameter of hydrogen-terminated diamond; obtaining the second surface structure of hydrogen-terminated diamond, performing first-principles high-throughput calculations on the second surface structure to obtain the second parameter of hydrogen-terminated diamond, and the second surface structure is the oxygen-hydrogen-terminated diamond surface structure; obtaining the third surface structure of hydrogen-terminated diamond, performing first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of hydrogen-terminated diamond; performing data fusion on the first parameter, the second parameter, and the interfacial transfer charge data to obtain the property parameters of hydrogen-terminated diamond.

[0059] Exemplarily, the method for obtaining the first surface structure of hydrogen-terminated diamond and performing first-principles high-throughput calculations on the first surface structure to obtain the first parameter of hydrogen-terminated diamond can be: first, using a first-principles calculation software based on density functional theory to optimize the diamond lattice constant, and the obtained diamond lattice constant is as Figure 2 (a) shown, intercepting the (100) crystal plane of diamond with a thickness of 10 carbon atom layers, saturating the dangling bonds of the bottom-layer carbon atoms with pseudo-hydrogen atoms, and setting the vacuum layer to Fixing the bottom 4 layers of carbon atoms and pseudo-hydrogen atoms to maintain their bulk structure properties. Establishing a 4×4×1 supercell, the 16 carbon atoms on the surface layer are subjected to 2×1 surface reconstruction, and the carbon atoms on the surface are rearranged to form a stable C-C dimer structure with a bond length of about Each carbon atom adsorbs one hydrogen atom for surface passivation. Develop a structure generation algorithm to traverse all possible first surface structures with a hydrogen atom adsorption concentration > 50%. Develop a structure deduplication algorithm to screen out more than 1000 non-repeated (since multiple first surface structures generated at the same concentration have translational and rotational symmetries, repeated queries and deduplication are required) first surface structures of hydrogen-terminated diamond (the more the types, the more accurate the subsequent predicted model of the structure-spectrum-property relationship of hydrogen-terminated diamond, but the training process will also take longer). Use first-principles calculation software based on density functional theory to perform high-throughput first-principles self-consistent calculations on the first surface structures of hydrogen-terminated diamond after structure optimization, and obtain first parameters (including electronic structure data such as total energy, band gap (absorption spectrum), valence band top, conduction band bottom, Fermi level, and work function, as well as properties such as hydrogen atom adsorption energy and electron affinity).

[0060] It should be noted that different hydrogen atom adsorption concentrations cause complex changes in the first surface structure, such as Figure 2(as shown in (b)). When hydrogen atoms are adsorbed at different concentrations, the adsorption energy shows oscillatory changes due to the uneven distribution of hydrogen atoms on paired carbon atoms; at the same hydrogen atom concentration, different adsorption positions also lead to significant differences in adsorption energy, accompanied by changes in spin states, affecting surface structure and surface property changes. In the case of high-concentration hydrogen atom adsorption (>50%), hydrogen atom adsorption effectively saturates the carbon atoms on the diamond surface, inducing more complex and uniform surface state changes. This effectively reduces surface defects and impurity scattering, helps promote the generation of carriers and reduces the scattering and trapping of carriers during movement, thereby improving the mobility of surface carriers. The bandgap size, the positions of the valence band maximum and the conduction band minimum show different degrees of upward shifts with the increase in the hydrogen atom adsorption concentration. On the contrary, the electron affinity decreases with the increase in the surface hydrogen atom concentration. At the same time, at the same hydrogen atom adsorption concentration, the surface configurations at different adsorption positions show obvious differences, especially in the case of high-concentration (>50%) hydrogen atom adsorption. By extracting the data of the bandgap, valence band maximum, conduction band minimum and electron affinity of the most stable diamond surface configurations at different adsorption concentrations, the high-concentration hydrogen atom adsorption configurations have a greater modulation effect on the bandgap and surface states, resulting in the upward shift of the conduction band minimum to exceed the vacuum level, forming a negative electron affinity. A negative electron affinity means that the energy cost of releasing electrons into the vacuum is negative, that is, electrons can spontaneously escape into the vacuum without external energy, thus significantly increasing the carrier concentration and mobility of the system. With the increase in the hydrogen adsorption concentration, the electron affinity shows an oscillatory downward trend, and this oscillation may be related to the uneven distribution of hydrogen atoms on the surface and the periodic adjustment of the surface electronic structure. With the increase in the hydrogen concentration, the work function of the diamond surface shows a downward trend, which can reduce the work function difference at the diamond-metal contact interface, thereby reducing the contact resistance. A low contact resistance helps improve the carrier injection efficiency and reduce energy loss, which is beneficial to the performance of electronic devices.

[0061] Optionally, the method for obtaining the second surface structure of hydrogen-terminated diamond and performing first-principles high-throughput calculations on the second surface structure to obtain the second parameters of hydrogen-terminated diamond can be: saturating the dangling bonds of the carbon atoms on the surface of hydrogen-terminated diamond with oxygen atoms to construct the second surface structure of oxygen-hydrogen-terminated diamond; performing first-principles high-throughput calculations on the second surface structure to obtain the second parameters of hydrogen-terminated diamond. Performing first-principles high-throughput calculations on the second surface structure to obtain the second parameters of hydrogen-terminated diamond is similar to the above method for obtaining the first parameters and will not be elaborated here.

[0062] However, it should be noted that for the second surface structure of oxygen-hydrogen-terminated diamond, there are dangling bonds on the surface carbon atoms at different hydrogen atom adsorption concentrations. To optimize the surface model and improve the authenticity of the data, the structures with dangling bonds are subjected to saturation treatment of oxygen atom adsorption, such asFigure 3 As shown in (a), the adsorption modes mainly include end-to-end adsorption (on-top) and different forms of bridge adsorption (bridge). The adsorption of oxygen atoms can more significantly affect the change in the electronic structure properties of the diamond surface configuration. For example Figure 3 As shown in (b), when oxygen atoms can saturate almost all the dangling bonds on the diamond surface, it helps to significantly reduce or eliminate surface states. The oxygen atoms form stable C-O bonds with the surface carbon atoms, which not only improves the chemical stability of the surface but also reduces the influence of surface charge carrier scattering or recombination, thereby improving the electronic and optoelectronic properties of the material. On the other hand, if oxygen atoms do not saturate all the dangling bonds, it may lead to the formation of more surface states and intermediate impurity energy levels. These impurity energy levels are located between the conduction band and the valence band and are easily formed as recombination centers for electron-hole pairs, which not only reduces the generation efficiency of carriers but also increases the scattering of carriers during movement, thereby affecting the overall electronic properties of the material. Compared with the system with only hydrogen atom adsorption, the addition of oxygen atoms will further reduce the electron affinity, and even at a relatively low oxygen atom concentration (50%), a negative electron affinity appears. Compared with the system with only hydrogen atom adsorption, the addition of oxygen atoms will cause a significant increase in the work function, which may be due to the surface charge rearrangement caused by the relatively high electronegativity of oxygen atoms, increasing the energy barrier from the metal to the semiconductor. The high-concentration adsorption of hydrogen atoms and the saturation treatment of the oxygen-containing terminal provide an effective way to optimize the key performance attributes of electronic devices by adjusting the bottom of the conduction band and affecting the electron affinity on the diamond surface.

[0063] Optionally, the method for obtaining the third surface structure of hydrogen-terminated diamond and performing first-principles high-throughput calculation on the third surface structure to obtain the interface transfer charge data of hydrogen-terminated diamond can be: intercepting a single layer of hexagonal boron nitride and a double layer of gold atoms as the electron transport layer to construct the third surface structure of hydrogen-terminated diamond; performing first-principles high-throughput calculation on the third surface structure to obtain the interface transfer charge data of the hydrogen-terminated diamond heterojunction.

[0064] Exemplarily, intercepting a single layer of hexagonal boron nitride (lattice constant α = 96.587°, β = 108.446°) and a double layer of gold atoms (lattice constant α = 90°, β = 83.163°), and constructing a surface interface heterojunction (i.e., the third surface structure) with a 4×4×1 hydrogen-terminated diamond surface (lattice constant α = β = 90°), with a lattice mismatch of less than 5%. Use the first-principles calculation software based on density functional theory to optimize the structure of the hydrogen-terminated diamond surface heterojunction at different concentrations, and the optimization parameters are the same as those of the diamond lattice constant. In particular, the lattice mismatch causes the formation of a wrinkled deformation of the single-layer hexagonal boron nitride layer with a heterostructure type, asFigure 4 (as shown in (a)). Using the Bader charge analysis method, the interfacial transfer charge data is quantitatively calculated.

[0065] It should be noted that the interfacial transfer charge data is used to evaluate how hydrogen atoms affect carrier characteristics at different concentrations. As Figure 4 (b) shows, the surface hydrogen atom adsorption concentration and relative position not only affect the charge transfer surface density of hydrogen-terminated diamond, but also cause significant changes in the interfacial structure. In the heterostructure of boron nitride, the wrinkled structure shows stronger stability and greater charge transfer ability compared to the planar structure. At the same time, with the increase in the hydrogen atom adsorption concentration, the maximum transferred charge amount at different hydrogen atom adsorption concentrations shows a trend of first increasing and then decreasing, and reaches the maximum value near 75%.

[0066] S102, construct a correlation dataset between structural parameters and property parameters.

[0067] Among them, the correlation dataset is a dataset that can characterize the corresponding relationship between different structural parameters and different property parameters.

[0068] Specifically, since the structural parameters and property parameters are in one-to-one correspondence and have been calculated in the above step S101, the correlation dataset between the structural parameters and property parameters can be directly constructed according to the corresponding relationship between the structural parameters and property parameters.

[0069] S103, perform infrared spectrum calculation on the structural parameters of hydrogen-terminated diamond to obtain the spectral data of hydrogen-terminated diamond.

[0070] Specifically, fix all atoms on the surface configuration of hydrogen-terminated diamond except the surface hydrogen atoms and subsurface carbon atoms, as Figure 5 (a) shows. Secondly, use the finite displacement method of the first-principles calculation software to calculate the vibration modes and Born effective charges of the surface hydrogen atoms and subsurface carbon atoms. The calculation parameters include IBRION = 5, NWRITE = 3, POTIM = 0.015, LEPSILON =.TRUE.. Use the infrared spectrum analysis script IR.sh to extract information and normalize the calculated OUTCAR file in turn to obtain the spectral data of hydrogen-terminated diamond (including the vibration modes and activity data of the infrared spectrum).

[0071] S104, construct a mapping dataset between structural parameters and spectral data.

[0072] Specifically, as Figure 5 (b) shows, different surface configurations of hydrogen-terminated diamond correspond to different infrared spectrum signals, and the mapping dataset of the diamond surface structural parameters and spectral data is constructed.

[0073] S105. Based on the correlation dataset and the mapping dataset, train a prediction model for the structure-spectrum-property relationship of hydrogen-terminated diamond.

[0074] Specifically, a first prediction model between the first surface structure and the first parameter can be constructed first. For example, using graph neural network technology, a first prediction model for the surface properties of hydrogen-terminated diamond is established by using the first surface structure and the first parameter, accurately predicting the key property attributes of the diamond surface, including atomic adsorption energy, absorption spectrum (band gap), and electron affinity, etc. First, use the crystal graph neural network model to establish the structure-property relationship between the surface structure of hydrogen-terminated diamond and related physical properties. The crystal graph neural network model takes the coordinate information of surface hydrogen atoms and subsurface carbon atoms as input, including bond, angle, and crystal periodicity information, etc., and can more comprehensively capture the complexity of the surface chemical environment.

[0075] In addition, as Figure 6 shown, the crystal graph neural network model shows a very high prediction accuracy rate, and the prediction accuracy rates of hydrogen atom adsorption energy, absorption spectrum (band gap), electron affinity, and work function all reach more than 95%. Although the prediction accuracy rates of the valence band top and Fermi level are slightly lower, their R 2 still reach 94.8% and 90.6% respectively, indicating that the algorithm as a whole has good accuracy. In addition, by testing the model containing the first three layers of H-C-C atoms, its prediction results are basically the same as those of the algorithm using only the H-C layer, indicating that adding more levels of atomic coordinate information does not significantly improve the algorithm performance.

[0076] Secondly, a second prediction model between the second surface structure and the second parameter is constructed. For example, for the oxygen-terminated diamond system with higher surface configuration and electronic structure complexity, a graph neural network model based on the Transformer architecture is used to establish the structure-property relationship between the second surface structure and the second parameter. The graph neural network method based on the Transformer architecture uses the principle of equivariance and is invariant to geometric transformations such as rotation and reflection of the input data, and is suitable for processing more complex structural systems, and finally obtains the second prediction model.

[0077] In addition, as Figure 7 shown, using the correlation dataset of the surface structure and property parameters adsorbed with oxygen atoms, the prediction accuracy rates of the graph neural network model based on the Transformer architecture for the band gap, conduction band bottom, electron affinity, and work function are significantly improved, and the R 2 are 91.4%, 97.1%, 96.7%, and 96.5% respectively. These findings emphasize the effectiveness and potential of the crystal graph neural network method in predicting the surface properties of diamond, especially its advantages in accurately simulating the interaction of complex surface atoms.

[0078] Then, construct a third prediction model between the third surface structure and the third parameter. For example, consider the influence of different reconstructed surfaces on the charge transfer of monolayer hexagonal boron nitride heterojunctions. The charge transfer amount of the non-planar structure heterojunction is significantly less than that of the planar structure, and partial inversion occurs at high concentrations. Use the crystal graph neural network machine learning model to train and predict the interfacial transfer charge, establish the structure-activity relationship between the structural parameters and the transfer charge, R 2 can reach 96.7%.

[0079] Then, construct a fourth prediction model between the structural parameters and the spectral data. For example, according to the mapping data set between the structural parameters and the spectral data, construct a fourth prediction model that can reflect the relationship between the structural parameters and the spectral data.

[0080] Finally, fuse the first prediction model, the second prediction model, the third prediction model, and the fourth prediction model to obtain the prediction model of the structure-spectrum-activity relationship of hydrogen-terminated diamond.

[0081] Optionally, it can also be based on the correlation data set and the mapping data set to construct a relationship mapping table between the structural parameters, the property parameters, and the spectral data; based on the relationship mapping table, train the prediction model of the structure-spectrum-activity relationship of hydrogen-terminated diamond.

[0082] Exemplarily, construct a relationship mapping table between the structural parameters, the property parameters, and the spectral data. Then, by analyzing the infrared spectral signals of the system, key information such as the surface hydrogenation degree, chemical bond changes, defects, and vibration modes after hydrogen atom adsorption can be understood. Clear the infrared spectral noise data with a spectral intensity less than 0.01, and align the spectral data according to the frequency size (cm -1 )), where the frequencies of carbon atoms in the low-frequency band and hydrogen atoms in the high-frequency band are preferentially aligned. Use the infrared spectral data as the eigenvalue and the band gap value as the target parameter, and train the spectrum-activity relationship model using different machine learning algorithms. Use R 2 as the evaluation criterion for the model accuracy, select the optimal machine learning algorithm to build the machine learning model of this system through parameter tuning and conduct property prediction. Using the machine learning algorithm, use the processed infrared spectral data as the eigenvalue, and use the valence band top, conduction band bottom, band gap (absorption spectrum), and electron affinity as the target parameters respectively to train the spectrum-activity relationship model and conduct property prediction. 80% of the data set is used as the training set, 20% is used as the test set, and the determination coefficient R 2 and the mean absolute error MAE are used as the evaluation criteria for the model prediction accuracy to train the prediction model of the structure-spectrum-activity relationship of hydrogen-terminated diamond.

[0083] In addition, as shown in Figure 5 (c), in the prediction model of the structure-spectrum-activity relationship of hydrogen-terminated diamond obtained by training, the R of the valence band top and the band gap 2They are 0.87 and 0.88 respectively, and the MAEs are 0.03 and 0.05 respectively; the R of the electron affinity 2 is 0.83 and the MAE is 0.05.

[0084] In order to make the prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond more accurate, the importance analysis of the frequency and intensity of the infrared spectrum can also be carried out by using the trained prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond to obtain the importance parameters of different frequencies and intensities. As Figure 8 shown, the importance analysis results show that the low-frequency vibration spectra (766 cm -1 , 707 cm -1 , 1025 cm -1 ) of the subsurface C atoms and the high-frequency vibration spectra (2977 cm -1 ~3008 cm -1 ) of the surface hydrogen atoms in the diamond surface configuration have the greatest influence on the prediction results of the model properties, and are important spectral feature descriptors for predicting the properties of the hydrogen-terminated diamond surface configuration.

[0085] For the training method of the prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond in the above embodiments, first obtain the structural parameters of hydrogen-terminated diamond, and perform first-principles high-throughput calculations on the structural parameters to obtain the property parameters of hydrogen-terminated diamond; then construct a correlation dataset between the structural parameters and the property parameters; then perform infrared spectrum calculations on the structural parameters of hydrogen-terminated diamond to obtain the spectral data of hydrogen-terminated diamond; then construct a mapping dataset between the structural parameters and the spectral data; finally, based on the correlation dataset and the mapping dataset, train the prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond. First, a correlation dataset between the structural parameters and the property parameters is constructed, and then a mapping dataset between the structural parameters and the spectral data is constructed. It is equivalent to constructing a mapping relationship among the structural parameters, property parameters, and spectral data with the structural parameters as the bridge, that is, according to the spectral data of hydrogen-terminated diamond, its property parameters can be obtained. Then, the prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond trained based on the correlation dataset and the mapping dataset can also predict its property parameters based on the spectral data of hydrogen-terminated diamond, greatly increasing the efficiency and accuracy of predicting the properties of hydrogen-terminated diamond.

[0086] In another embodiment, when it is necessary to obtain the property parameters of the hydrogen-terminated diamond to be predicted, the spectral data of the hydrogen-terminated diamond to be predicted can be obtained; the spectral data of the hydrogen-terminated diamond to be predicted is input as the eigenvalue of the prediction model of the structure-spectrum-property relationship of hydrogen-terminated diamond to obtain the property parameters of the hydrogen-terminated diamond to be predicted.

[0087] Specifically, the hydrogen-terminated diamond to be predicted can be calculated to obtain its spectral data, and then the spectral data is input into the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond, and the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond can output the property parameters of the hydrogen-terminated diamond to be predicted.

[0088] To more comprehensively demonstrate the present solution, an optional method for training the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond is given in this embodiment, as Figure 9 shown:

[0089] S201. Obtain the first surface structure of the hydrogen-terminated diamond, perform first-principles high-throughput calculations on the first surface structure, and obtain the first parameters of the hydrogen-terminated diamond.

[0090] S202. Saturate the dangling bonds of the surface carbon atoms of the hydrogen-terminated diamond with oxygen atoms to construct the second surface structure of the oxygen-hydrogen-terminated diamond.

[0091] S203. Perform first-principles high-throughput calculations on the second surface structure to obtain the second parameters of the hydrogen-terminated diamond.

[0092] S204. Intercept a single-layer hexagonal boron nitride and a double-layer gold atom as the electron transport layer to construct the third surface structure of the hydrogen-terminated diamond.

[0093] S205. Perform first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of the hydrogen-terminated diamond.

[0094] S206. Perform data fusion on the first parameters, the second parameters, and the interfacial transfer charge data to obtain the property parameters of the hydrogen-terminated diamond.

[0095] S207. Construct a correlation dataset between the structural parameters and the property parameters.

[0096] Among them, the structural parameters include the first surface structure, the second surface structure, and the third surface structure

[0097] S208. Perform infrared spectrum calculations on the structural parameters of the hydrogen-terminated diamond to obtain the spectral data of the hydrogen-terminated diamond.

[0098] S209. Construct a mapping dataset between the structural parameters and the spectral data.

[0099] S210. Based on the correlation dataset and the mapping dataset, construct a relationship mapping table between the structural parameters, the property parameters, and the spectral data.

[0100] S211. Based on the relationship mapping table, train the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond.

[0101] S212. Obtain the spectral data of the hydrogen-terminated diamond to be predicted.

[0102] S213. Input the spectral data of the hydrogen-terminated diamond to be predicted into the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond to obtain the property parameters of the hydrogen-terminated diamond to be predicted.

[0103] For the specific processes of the above S201 - S213, reference can be made to the description of the method embodiments above. Their implementation principles and technical effects are similar, and will not be elaborated here.

[0104] Based on the same inventive concept, the embodiments of the present application also provide a training device for the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond for implementing the training method of the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond involved above. The implementation solutions provided by this device to solve problems are similar to those recorded in the above method. Therefore, the specific limitations in one or more embodiments of the training device for the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond provided below can refer to the limitations on the training method of the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond in the above text, and will not be elaborated here.

[0105] In one embodiment, as Figure 10 shown, a training device for the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond is provided. The device includes:

[0106] A parameter calculation module 30, configured to obtain the structure parameters of the hydrogen-terminated diamond and perform first-principles high-throughput calculations on the structure parameters to obtain the property parameters of the hydrogen-terminated diamond;

[0107] A first construction module 31, configured to construct a correlation data set between the structure parameters and the property parameters;

[0108] A spectral calculation module 32, configured to perform infrared spectral calculations on the structure parameters of the hydrogen-terminated diamond to obtain the spectral data of the hydrogen-terminated diamond;

[0109] A second construction module 33, configured to construct a mapping data set between the structure parameters and the spectral data;

[0110] A model training module 34, configured to train the structure-spectrum-property relationship prediction model of the hydrogen-terminated diamond based on the correlation data set and the mapping data set.

[0111] In another embodiment, the structure parameters include a first surface structure, a second surface structure, and a third surface structure. The above Figure 10The parameter calculation module 30 therein is specifically configured to: obtain the first surface structure of hydrogen-terminated diamond, perform first-principles high-throughput calculation on the first surface structure to obtain the first parameters of hydrogen-terminated diamond; obtain the second surface structure of hydrogen-terminated diamond, perform first-principles high-throughput calculation on the second surface structure to obtain the second parameters of hydrogen-terminated diamond; the second surface structure is an oxygen-hydrogen-terminated diamond surface structure; obtain the third surface structure of hydrogen-terminated diamond, perform first-principles high-throughput calculation on the third surface structure to obtain the interface transfer charge data of hydrogen-terminated diamond; perform data fusion on the first parameters, the second parameters and the interface transfer charge data to obtain the property parameters of hydrogen-terminated diamond.

[0112] Specifically, use oxygen atoms to saturate the dangling bonds of carbon atoms on the surface of hydrogen-terminated diamond to construct the second surface structure of oxygen-hydrogen-terminated diamond; perform first-principles high-throughput calculation on the second surface structure to obtain the second parameters of hydrogen-terminated diamond. Intercept a single layer of hexagonal boron nitride and a double layer of gold atoms as the electron transport layer to construct the third surface structure of hydrogen-terminated diamond; perform first-principles high-throughput calculation on the third surface structure to obtain the interface transfer charge data of hydrogen-terminated diamond.

[0113] In another embodiment, the above Figure 10 The model training module 34 therein is specifically configured to: based on the correlation dataset and the mapping dataset, construct a relationship mapping table between the structural parameters, the property parameters and the spectral data; based on the relationship mapping table, train the structure-spectrum-property relationship prediction model of hydrogen-terminated diamond.

[0114] In another embodiment, as Figure 11 shown, the training device for the structure-spectrum-property relationship prediction model of hydrogen-terminated diamond further includes:

[0115] The data acquisition module 35 is configured to acquire the spectral data of the hydrogen-terminated diamond to be predicted;

[0116] The property prediction module 36 is configured to input the spectral data of the hydrogen-terminated diamond to be predicted as the eigenvalue of the structure-spectrum-property relationship prediction model of hydrogen-terminated diamond to obtain the property parameters of the hydrogen-terminated diamond to be predicted.

[0117] The embodiments of the present application further provide an electronic device. In some embodiments, refer to Figure 12As shown, the electronic device 700 includes an input unit 710, a memory 720, a processor 730, and an output unit 740. The memory 720 stores program instructions that can be run on the processor 730. By invoking the program instructions, the processor 730 can execute the training method and / or technical solution based on the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model in the foregoing embodiments. The electronic device 700 can be a mobile terminal device such as a mobile phone or a computer, or a high-performance computing server such as a computing cluster.

[0118] In addition, an embodiment of the present application further provides a computer-readable storage medium for storing a computer program for executing the training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model. For example, when the computer program instructions are executed by a computer, through the operation of the computer, the methods and / or technical solutions according to the present application can be invoked or provided. The program instructions for invoking the methods of the present application may be stored in a fixed or removable storage medium, and / or transmitted through a data stream in a broadcast or other signal-bearing medium and / or stored in a storage medium running according to the program instructions.

[0119] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present application can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple of them can be fabricated into a single integrated circuit module. In this way, the present application is not limited to any specific combination of hardware and software.

[0120] The technical features of the above embodiments can be arbitrarily integrated. For the sake of brevity of description, not all possible integrations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the integration of these technical features, it should be considered to be within the scope described in this specification.

[0121] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent should be subject to the appended claims.

Claims

1. A training method for a prediction model of the structure-spectrum-effect relationship of hydrogen-terminated diamond, characterized in that, The method includes: Obtaining the structural parameters of hydrogen-terminated diamond, and performing first-principles high-throughput calculations on the structural parameters to obtain the property parameters of the hydrogen-terminated diamond; Constructing a correlation data set between the structural parameters and the property parameters; Performing infrared spectrum calculations on the structural parameters of the hydrogen-terminated diamond to obtain the spectral data of the hydrogen-terminated diamond; Constructing a mapping data set between the structural parameters and the spectral data; Training a structure-spectrum-property relationship prediction model for hydrogen-terminated diamond based on the correlation data set and the mapping data set.

2. The training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model according to claim 1, wherein The structural parameters include a first surface structure, a second surface structure, and a third surface structure. Obtaining the structural parameters of hydrogen-terminated diamond and performing first-principles high-throughput calculations on the structural parameters to obtain the property parameters of the hydrogen-terminated diamond includes: Obtaining the first surface structure of the hydrogen-terminated diamond and performing first-principles high-throughput calculations on the first surface structure to obtain the first parameter of the hydrogen-terminated diamond; Obtaining the second surface structure of the hydrogen-terminated diamond and performing first-principles high-throughput calculations on the second surface structure to obtain the second parameter of the hydrogen-terminated diamond; the second surface structure is an oxygen-hydrogen-terminated diamond surface structure; Obtaining the third surface structure of the hydrogen-terminated diamond and performing first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of the hydrogen-terminated diamond; the third surface structure is a hydrogen-terminated diamond heterojunction surface structure; Performing data fusion on the first parameter, the second parameter, and the interfacial transfer charge data to obtain the property parameters of the hydrogen-terminated diamond.

3. The training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model according to claim 2, wherein, Obtaining the third surface structure of the hydrogen-terminated diamond and performing first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of the hydrogen-terminated diamond includes: Intercepting a single layer of hexagonal boron nitride and a double layer of gold atoms as an electron transport layer to construct the third surface structure of the hydrogen-terminated diamond; Performing first-principles high-throughput calculations on the third surface structure to obtain the interfacial transfer charge data of the hydrogen-terminated diamond.

4. The training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model according to claim 2, characterized in that, Obtaining the second surface structure of the hydrogen-terminated diamond and performing first-principles high-throughput calculations on the second surface structure to obtain the second parameter of the hydrogen-terminated diamond includes: Using oxygen atoms to saturate the dangling bonds of surface carbon atoms of the hydrogen-terminated diamond to construct the second surface structure of the oxygen-hydrogen-terminated diamond; Performing first-principles high-throughput calculations on the second surface structure to obtain the second parameter of the hydrogen-terminated diamond.

5. The training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model according to claim 1, characterized in that Training a structure-spectrum-property relationship prediction model for hydrogen-terminated diamond based on the correlation data set and the mapping data set includes: Based on the correlation data set and the mapping data set, constructing a relationship mapping table between the structural parameters, the property parameters, and the spectral data; Based on the relationship mapping table, training the structure-spectrum-property relationship prediction model for hydrogen-terminated diamond.

6. The training method of the hydrogen-terminated diamond structure-spectrum-effect relationship prediction model according to any one of claims 1-5, characterized in that, The method further includes: Obtaining the spectral data of the hydrogen-terminated diamond to be predicted; Input the spectral data of the diamond with hydrogen terminals to be predicted as the eigenvalue into the structure-spectrum-property relationship prediction model of the diamond with hydrogen terminals, and obtain the property parameters of the diamond with hydrogen terminals to be predicted.

7. A training device for a prediction model of the structure-spectrum-effect relationship of hydrogen-terminated diamond, characterized in that, The device includes: A parameter calculation module, configured to obtain the structural parameters of the diamond with hydrogen terminals, and perform first-principles high-throughput calculation on the structural parameters to obtain the property parameters of the diamond with hydrogen terminals; A first construction module, configured to construct a correlation data set between the structural parameters and the property parameters; A spectral calculation module, configured to perform infrared spectral calculation on the structural parameters of the diamond with hydrogen terminals to obtain the spectral data of the diamond with hydrogen terminals; A second construction module, configured to construct a mapping data set between the structural parameters and the spectral data; A model training module, configured to train a structure-spectrum-property relationship prediction model of the diamond with hydrogen terminals based on the correlation data set and the mapping data set.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, the training method of the structure-spectrum-property relationship prediction model of the diamond with hydrogen terminals according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, the training method of the structure-spectrum-property relationship prediction model of the diamond with hydrogen terminals according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, the training method of the structure-spectrum-property relationship prediction model of the diamond with hydrogen terminals according to any one of claims 1 to 6 is implemented.

Images