A method for screening laser glass components and a method for manufacturing laser glass

Abstract

The application discloses a laser glass component screening method and a laser glass manufacturing method, and the component screening method comprises the following steps: dividing a glass forming range based on consistent melting compounds, and obtaining physical properties and spectral data of a plurality of components in the glass forming range; obtaining structural characteristic description operators of the components based on molecular dynamics simulation; constructing a training data set based on the structural characteristic description operators, the physical properties and the spectral data; constructing a component-structure-property model of the glass, training the model based on the training data set, obtaining the trained component-structure-property model of the glass, and predicting corresponding physical properties and spectral data of the glass components based on the trained component-structure-property model of the glass, according to the glass components or the structural characteristic description operators as input variables. The application can efficiently and accurately screen the laser glass components.

Description

Technical Field

[0001] This invention relates to the field of glass material analysis technology, specifically to a method for screening components in laser glass and a method for manufacturing laser glass. Background Technology

[0002] Rare-earth-doped laser glasses have attracted significant attention in high-energy laser fusion devices and fiber optic communication systems. However, the traditional development of laser glasses has relied on costly and time-consuming trial-and-error methods, making it extremely difficult to establish composition-structure-property relationships due to limited experimental data and low efficiency. Furthermore, research on the luminescence behavior of rare-earth-doped glasses is limited by their structure, particularly the localized structure of rare-earth ions. Unlike crystalline materials, the lack of long-range order in glass structures makes it impossible to establish structural models using single experimental techniques. Even if a large amount of structural parameters and property data can be obtained, the analysis process may be overly cumbersome or oversimplified, making it difficult to obtain instructive model rules. Therefore, finding a simple, efficient, and rapid method for screening laser glass components that effectively utilizes computer processing power is crucial. Summary of the Invention

[0003] To overcome the defects and shortcomings of existing technologies, this invention provides a method for screening laser glass components and a method for manufacturing laser glass. This invention partitions the glass-forming region based on a uniformly molten compound, obtains the physical properties and spectral data of multiple components within the glass-forming region, and obtains structural feature descriptors for the components based on molecular dynamics simulations. Based on the structural feature descriptors, physical properties, and spectral data, a training dataset is constructed. A component-structure-property model of the glass is built, and the model is trained based on the training dataset. Based on the trained component-structure-property model of the glass, and using the glass components or structural feature descriptors as input variables, the corresponding physical properties and spectral data of the glass components are predicted, enabling efficient and accurate screening of laser glass components.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] This invention provides a method for screening the components of laser glass, comprising the following steps:

[0006] Based on the uniform melting compound, the glass-forming region is divided into zones to obtain the physical properties and spectral data of multiple components within the glass-forming region;

[0007] Structural characterization operators for components were obtained based on molecular dynamics simulations.

[0008] A training dataset is constructed based on structural feature description operators, physical properties, and spectral data.

[0009] Constructing a composition-structure-property model of glass and training the model based on a training dataset, specifically including:

[0010] Using glass composition as input variable and structural feature descriptor as output variable for model training, the relationship between multiple components and one structural operator is established to obtain the relationship between components and structural feature descriptor.

[0011] Using structural feature descriptors as input variables and glass composition physical properties and spectral data as output variables for model training, the relationship between multiple structural feature descriptors and a glass property is established respectively, and the relationship between structural feature descriptors and glass composition physical properties and spectral data is obtained.

[0012] Based on the trained glass composition-structure-property model, the corresponding glass composition physical properties and spectral data are predicted by using glass composition or structural feature descriptors as input variables.

[0013] As a preferred technical solution, the method for confirming the uniformly melted compound adopts any one of the following methods:

[0014] Confirmed using a binary or ternary phase diagram database;

[0015] Or refer to the phase diagrams of other systems with similar crystallization chemistry conditions to the corresponding system;

[0016] Alternatively, the extreme point component whose properties vary with the composition can be selected as a potential uniformly melting compound;

[0017] Alternatively, starting from first principles and combining with structure search algorithms, we can search for compounds formed under different atomic ratios, and then comprehensively judge whether they are uniformly melted compounds from thermodynamic and kinetic perspectives.

[0018] As a preferred technical solution, the step of dividing the glass-forming region into zones based on a uniformly molten compound and obtaining the physical properties and spectral data of multiple components within the glass-forming region includes the following specific steps:

[0019] After the glass-forming region is divided into zones, each sample component is located within a triangle formed by three nearest neighbor homogeneous molten compounds. Based on the properties of the homogeneous molten compounds at the vertices of the triangles, the physical properties and spectral data of multiple components within the glass-forming region are calculated using the lever principle.

[0020] As a preferred technical solution, the physical properties and spectral data of multiple components within the glass-forming region are calculated based on the lever principle. The principles followed are: selecting from the polygon of the nearest uniformly molten compound of the target component; and connecting uniformly molten compounds with lower melting temperatures is more likely.

[0021] As a preferred technical solution, the calculation formula based on the lever principle is expressed as follows:

[0022] P(X)=P(A)×L A +P(B)×L B +P(C)×L C

[0023] Among them, L A L B and L C P(A), P(B), and P(C) represent the molar contents of uniformly molten compounds A, B, and C of glass with component X, calculated according to the lever principle. P(A), P(B), and P(C) represent the properties of uniformly molten compounds A, B, and C, respectively.

[0024] As a preferred technical solution, the calculation of the properties of the binary glass and one of its nearest homologous molten compounds is also included, specifically expressed as follows:

[0025]

[0026] Where P(Y) represents the property of the homologous glass of the desired molten compound, P(X) represents the property of the binary glass, P(Z) represents the property of the other nearest-neighbor homologous glass of the same composition, and L... Z and L Y These represent the molar contents of the uniformly molten compounds Y and Z, respectively, of glass with component X, calculated using the lever principle.

[0027] As a preferred technical solution, obtaining structural feature descriptors of components based on molecular dynamics simulations includes the following steps:

[0028] Construct the initial box model;

[0029] Energy minimization is performed using gradient descent, conjugate gradient, or steepest descent methods.

[0030] The long-range Coulomb force and the short-range two-body and three-body potential energy are used to simulate the interparticle forces.

[0031] High-temperature melting is performed using NVT, NPT, or NVE ensembles;

[0032] Cooling simulations were performed using the NVT ensemble, NPT ensemble, or NVE ensemble.

[0033] Obtain structural feature descriptors for components.

[0034] As a preferred technical solution, the high-temperature melting is carried out using an NVT ensemble, an NPT ensemble, or an NVE ensemble, with a high-temperature melting temperature of 5000–8000 K and an equilibration time of 300–500 ps.

[0035] Cooling simulations were performed using the NVT, NPT, or NVE ensembles. The cooling simulation rate ranged from 0.1 K / ps to 50 K / ps, and the temperature at which the cooling simulation was completed was 250 to 350 K.

[0036] As a preferred technical solution, the structural feature descriptor includes: radial distribution function, coordination number, angular distribution function, anion bridging degree, and agglomerate polymerization degree.

[0037] The present invention also provides a method for preparing laser glass, wherein laser glass components are obtained according to the above-mentioned laser glass component screening method, and laser glass with corresponding properties is prepared according to the laser glass components.

[0038] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0039] (1) This invention divides the glass-forming region into zones based on a uniformly molten compound, obtains the physical properties and spectral data of multiple components within the glass-forming region, and obtains structural feature descriptors of the components based on molecular dynamics simulation; a training dataset is constructed based on the structural feature descriptors, physical properties, and spectral data; a component-structure-property model of the glass is constructed, and the model is trained based on the training dataset; based on the trained component-structure-property model of the glass, the corresponding physical properties and spectral data of the glass components are predicted according to the glass component or structural feature descriptors as input variables. Compared with the traditional trial-and-error method, this method significantly reduces experimental losses, and the component screening and analysis are more comprehensive and accurate, making the subsequent laser glass manufacturing process more efficient.

[0040] (2) This invention trains the model based on the training dataset of structural feature descriptor operators, physical properties and spectral data, and obtains a glass composition-structure-property model for comprehensive composition screening and analysis. This overcomes the difficulties of traditional methods in establishing composition-structure-performance relationship due to limited experimental data and long-range order loss. It accurately analyzes the luminescence behavior of rare earth doped glass and obtains the correlation model between glass composition and properties, thus improving research efficiency and accuracy.

[0041] (3) The glass composition-structure-property model constructed in this invention has good interpretability and can sort the importance of property data according to the structure operator, which is helpful for the preparation and performance optimization of laser glass.

[0042] (4) This invention has wide applicability and can be used in various types of glass, such as oxide glass, halide glass, chalcogenide glass or metallic glass. It can predict and prepare laser glass with the required properties according to different application requirements. It is widely used in high-energy laser fusion devices such as solid-state lasers, fiber lasers and fiber amplifiers, as well as fiber-optic communication systems. Attached Figure Description

[0043] Figure 1 This is a schematic flowchart of the laser glass component screening method in Example 1.

[0044] Figure 2 This is the glass property database for Example 3;

[0045] Figure 3 This is the molecular dynamics simulation structure diagram of Example 3;

[0046] Figure 4 This is a ranking diagram of the importance of properties for different structural operators in Example 6;

[0047] Figure 5 This is a diagram showing the relationship between the structural operators and property data in Example 6. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] Example 1

[0050] like Figure 1 As shown, this embodiment provides a laser glass component screening method, including the following steps:

[0051] Step S1: The glass-forming region is divided into zones based on the uniformly molten compound. After the zone is divided, each sample component is located within a triangle formed by the three nearest neighbor uniformly molten compounds. According to the properties of the uniformly molten compound at the vertex of the triangle, the physical properties and spectral data of multiple components within the glass-forming region are calculated and obtained according to the lever principle. In this embodiment, glass with the same component of uniformly molten compound is prepared and spectral data is measured.

[0052] In step S1, the method for confirming the uniformly molten compound can be any of the following methods:

[0053] (1) Confirmed through a binary or ternary phase diagram database;

[0054] (2) Refer to the phase diagrams of other systems with similar crystallization chemistry conditions to this system;

[0055] (3) Select the extreme point components whose properties change with the composition as potential uniformly melted compounds;

[0056] (4) Starting from first principles and combining the structure search algorithm, search for compounds formed under different atomic ratios, and then judge whether they are uniformly melted compounds from the perspectives of thermodynamics and kinetics.

[0057] In step S1, the specific method for obtaining the properties of the glass component from the properties of the uniformly molten compound is to use the uniformly molten compound to calculate the triangular partition of the glass phase diagram, and to sum the properties according to the lever principle. The principles to be followed are: select from the polygons formed by the nearest uniformly molten compounds of the target component, and the possibility of connecting uniformly molten compounds with lower melting temperatures is greater.

[0058] In this embodiment, since glass is a product of uniformly molten compounds, its properties can be considered as the sum of the properties of the nearest neighbor uniformly molten compounds. Specifically, the formation region is divided into triangles according to the uniformly molten compounds. The properties of the glass components within the triangles can be calculated using the lever principle based on the properties of the uniformly molten compounds represented by the three vertices. The specific formula is as follows:

[0059] P(X)=P(A)×L A +P(B)×L B +P(C)×L C (1)

[0060] Among them, L A L B and L C P(A), P(B), and P(C) represent the molar contents of uniformly molten compounds A, B, and C, calculated using the lever principle for glass of composition X. P(A), P(B), and P(C) represent the properties of uniformly molten compounds A, B, and C, respectively.

[0061] Furthermore, because glassmaking is a thermodynamic and kinetic process, sometimes a point within the glass-forming range is difficult to measure by producing a large piece of glass. Therefore, it is necessary to use the properties of other component points to infer the glass composition. For homogeneous molten compounds that cannot be prepared into glass, the properties can be calculated from the binary glass and one of the nearest homogeneous molten compounds of the same composition. The specific formula is as follows:

[0062]

[0063] Where P(Y) represents the property of the homologous glass of the desired uniform melting compound, P(X) represents the property of the binary glass, and P(Z) represents the property of the other nearest-neighbor homologous glass of the same composition. L Z and L YThese represent the molar contents of the uniformly molten compounds Y and Z, respectively, of glass with component X, calculated using the lever principle.

[0064] In step S1, the glass properties obtained include: density, refractive index, Abbe number, molecular volume, polarizability, elastic modulus, Young's modulus, bending resistance, surface hardness, coefficient of thermal expansion, heat capacity, thermal conductivity, thermal shock resistance, characteristic temperature of glass, anti-crystallization stability, chemical stability, absorption spectrum, absorption peak position, absorption cross section, emission spectrum, emission peak position, emission cross section, effective linewidth, fluorescence branching ratio, fluorescence lifetime, and Judd-Ofelt intensity parameters.

[0065] In this embodiment, the glass includes oxide glass, halide glass, chalcogenide glass, or metallic glass;

[0066] Step S2: Obtain the structural feature descriptors of the components based on molecular dynamics simulations;

[0067] In step S2, the steps of the molecular dynamics simulation include, but are not limited to, establishing an initial box model, energy minimization, simulating interparticle forces, high-temperature melting to eliminate the influence of the initial configuration, cooling and collecting structure operators. Specific steps include:

[0068] (1) Use one or more of Materials Studio, Packmol or LAMMPS to create an initial box model with the same density as the actual density;

[0069] (2) Minimize energy using gradient descent, conjugate gradient, or steepest descent methods;

[0070] The energy cutoff tolerance is 10. -4 ~10 -8 The force cutoff tolerance is 10. -6 ~10 -10 The maximum number of iterations is 10. 2 ~10 6 The maximum number of evaluations is 10. 3 ~10 8 ;

[0071] (3) The long-range Coulomb force and the short-range two-body and three-body potential energy are used to simulate the inter-particle forces;

[0072] The potential function chosen for the short-range interaction force between particles is the Buckingham potential function, whose expression is:

[0073]

[0074] Where r represents the distance between atom i and atom j, and A, ρ and C are the Buckingham potential energy parameters between atom i and atom j, respectively.

[0075] (4) High-temperature melting is performed using NVT ensemble, NPT ensemble or NVE ensemble to eliminate the influence of the initial configuration, with a preferred high-temperature range of 5000 to 8000 K;

[0076] Furthermore, the high-temperature melting temperature is 5000–8000K, the ensemble is NVT or NVE, and the equilibration time is 300–500ps;

[0077] (5) Cooling simulation is performed using NVT ensemble, NPT ensemble or NVE ensemble, with a cooling rate of 0.1K / ps to 50K / ps and a temperature of 250 to 350K after the cooling simulation is completed;

[0078] The cooling ensemble is NVT, NVE, or NPT, with a cooling rate of 0.01–0.1 K / step, a step size of 1–2 fs, a target cooling temperature of 250–350 K, and an equilibrium time of 20–50 ps.

[0079] (6) Collect structural feature descriptors within the model. The collection frequency is preferably 100 steps / time to 1000 steps / time. The structural feature descriptors of the collected components include: radial distribution function, coordination number, angular distribution function, anion bridging degree, and polymerization degree of the integrator.

[0080] Step S3: Using the obtained structural feature description operators and glass data, establish a training dataset containing composition, structural operators, physical and spectral properties;

[0081] Step S4: Train the glass composition-structure-property model based on machine learning to obtain the glass composition and the decisive structure operator for the desired performance;

[0082] In step S4, the machine learning model establishes a composition-structure-property model of the glass, including:

[0083] (1) The number of database sample points in the single-phase diagram reaches 30 to 100;

[0084] (2) The number of single-sample structure operators reaches 5 to 50;

[0085] (3) The physical and spectral property data of a single sample reach 1 to 20;

[0086] (4) Perform coherence analysis between structural operators. The analysis methods include Pearson correlation coefficient method, information gain method, and chi-square test.

[0087] (5) Using components as input variables and structural operators as output variables for model training, establish the relationship between multiple components and one structural operator to obtain the component structure relationship. Model training methods include linear regression, logistic regression, support vector machine, random forest, k-nearest neighbor algorithm, and artificial neural network.

[0088] (6) Using structural feature descriptors as input variables and glass properties as output variables for model training, establish the relationship between multiple structural feature descriptors and a glass property to obtain the structural property relationship. The model training methods include linear regression, logistic regression, support vector machine, random forest, k-nearest neighbor algorithm, and artificial neural network.

[0089] The composition-structure-property model of glass satisfies any one of the following characteristics:

[0090] (1) Includes structural operators and property data of sample points within the glass phase diagram formation range;

[0091] (2) By using components as input variables, structural operators and property data can be obtained;

[0092] (3) Using structural operators as input variables, components can be deduced and property data obtained;

[0093] (4) Using property data as input variables, the components and structural operators can be deduced in reverse;

[0094] In this embodiment, the glass composition-structure-property model includes the structure operators and property data of sample points within the glass phase diagram formation range. Based on machine learning (such as the random forest algorithm), the importance of the structure operators to the property data is obtained, and the structure operators can be ranked according to their importance to the property data. Using the trained glass composition-structure-property model, the corresponding glass properties can be predicted based on the composition or structure operators as input variables.

[0095] Example 2

[0096] This embodiment provides a method for preparing laser glass. Based on the laser glass component screening method in Embodiment 1 above, the laser glass components are obtained according to the glass component-structure-property model, and then laser glass with corresponding properties is prepared. Based on the properties of the laser glass, it can be applied to solid-state lasers, high-power, single-frequency or high-repetition-rate fiber lasers and fiber amplifiers.

[0097] Example 3

[0098] In this embodiment, as Figure 2As shown, based on the properties of homologous glasses with the same composition Na2O·SiO2, Na2O·2SiO2, SiO2, CaO·SiO2, and Na2O·2CaO·3SiO2, data from 69 sample points within the formation region of the Na2O-CaO-SiO2 (NCS) glass phase diagram were calculated and predicted. Neodymium-doped silicate laser glass was prepared using SiO2 (analytical grade, Aladdin), Na2CO3 (analytical grade, Aladdin), CaCO3 (99.95%, Aladdin), and Nd2O3 (99.99%, Aladdin) as raw materials via a traditional melt-quench process. 20g of raw material was accurately weighed according to the glass composition, transferred to an agate mortar for uniform mixing, placed in a corundum crucible, and held in a glass furnace at 1450-1500℃ for 30min. After melting, homogenization, and clarification, a silicate melt was obtained and poured into a preheated graphite mold for shaping. The sample was then annealed in a muffle furnace at 450°C for 2 hours to release residual stress. The resulting sample was mechanically cut to approximately 2 mm. It was then refined and polished sequentially using 400-mesh, 800-mesh, 1500-mesh, and 2000-mesh diamond grinding wheels and alumina powder with an average particle size of 0.3 μm.

[0099] In this embodiment, the properties include density, coefficient of thermal expansion, thermal conductivity, linear refractive index, nonlinear refractive index, Abbe number, Judd-Ofelt intensity parameter, fluorescence branching ratio, radiative transition lifetime, effective linewidth, absorption cross section, emission cross section, and gain bandwidth.

[0100] In this embodiment, 69 Nd-doped atoms were calculated according to formula (1). 3+ The basic composition, physical and spectroscopic properties of Nd-doped NSC glass (component spacing of 5 mol%) were established. 3+ The CSP database for NSC glass lists glass composition ranges of 49.5–99%, 0–49.5%, 0–49.5%, and 1 mol% for SiO2, CaO, Na2O, and Nd2O3, respectively. These Nd-doped glasses... 3+ The density, thermal expansion coefficient, thermal conductivity, 589 nm refractive index, peak refractive index, nonlinear refractive index, and Abbe number of the NSC glass range from 2.298 to 2.963 g / cm³. 3 23×10 -7 ~233×10 -7 / K, 0.95~1.25W / m·K, 1.4793~1.6327, 1.4671~1.6184, 1.14×10 -13 ~2.24×10 -13 ESU, 48.0~58.4. The ranges for Ω2, Ω4 / Ω6, fluorescence branching ratio, radiative transition lifetime, effective linewidth, absorption cross section, emission cross section, and gain bandwidth are 3.60×10⁻⁶.-20 ~4.93×10 -20 cm 2 , 0.90~1.21, 0.456~0.478, 330~779μs, 46.59~48.63nm, 0.72×10 -20 ~2.05×10 -20 cm 2 0.85×10 -20 ~2.11×10 -20 cm 2 and 4.11×10 -26 ~9.85×10 -26 cm 3 .

[0101] like Figure 3 As shown, Nd-doped glass can be plotted based on the calculated glass composition properties. 3+ The NCS glass property database.

[0102] Example 4

[0103] In this embodiment, a structural model of 47 sample points in the Na2O-CaO-SiO2 (NCS) glass-forming region was established based on molecular dynamics.

[0104] In this embodiment, Materials Studio is used to establish a box structure using a CVFF force field, with the density consistent with the actual density of the sample points. Optionally, the number of particles is between 8000 and 12000. Optionally, the box is a cube structure, and the box size is... For example, the number of particles could be 8000, 8500, 9000, 9500, 10000, 10500, 11000, 11500, or 12000. Another example is a box with a size of...

[0105] In a specific example, the energy minimization method employs gradient descent, with an energy cutoff margin of 10. -6 The force cutoff tolerance is 10. -8 The maximum number of iterations is 10. 4 The maximum number of evaluations is 10. 6 .

[0106] In a specific example, the potential function chosen for the short-range interaction force between particles is the Buckingham potential function, with the following potential energy parameters:

[0107] Table 1 Buckingham potential parameters in Na2O-CaO-SiO2 (NCS) glass

[0108]

[0109] In a specific example, the high-temperature melting temperature is 8000K, the ensemble is NVT, the equilibrium time is 500ps, then it decreases from 8000K to 6000K at a rate of 0.05K / step with a step size of 1fs, and relaxes for 30ps when it reaches 6000K. Then it decreases from 6000K to 3000K at a rate of 0.05K / step with a step size of 1fs, and relaxes for 30ps when it reaches 3000K. Then it decreases from 3000K to 300K at a rate of 0.05K / step with a step size of 1fs, and relaxes for 30ps when it reaches 300K.

[0110] In a specific example, structural operator data within the model was collected at 300K, with a collection frequency of 1000 steps / time, for a total of 300 times. The collected data included: radial distribution function, coordination number, angular distribution function, anion bridging degree, and agglomerate polymerization degree.

[0111] In one embodiment, the glass structure file output by Lammps can be opened using the visualization software Ovito, such as... Figure 2 As shown in the schematic diagram of the glass structure, the radial distribution function in this embodiment includes the bond length parameters between cations and oxygen ions, as shown in Table 2 below:

[0112] Table 2 shows the bond lengths between cations and oxygen ions in some sample points.

[0113]

[0114] Example 5

[0115] In this embodiment, a structural model of 55 sample points in the K2O-BaO-P2O5 glass-forming region was established based on molecular dynamics.

[0116] In this embodiment, Materials Studio is used to establish a box structure using a CVFF force field, with the density consistent with the actual density of the sample points. Optionally, the number of particles is between 9500 and 11000. Optionally, the box is a cube structure, and the box size is... For example, the number of particles could be 9500, 10000, 10500, or 11000. Another example is a box with a size of...

[0117] In a specific example, the energy minimization method employs gradient descent, with an energy cutoff margin of 10. -6 The force cutoff tolerance is 10. -8 The maximum number of iterations is 10. 5The maximum number of evaluations is 10. 6 .

[0118] In a specific example, the high-temperature melting temperature is 8000K, the ensemble is NVT, the equilibrium time is 400ps, then it decreases from 8000K to 6000K at a rate of 0.1K / step with a step size of 2fs, and relaxes for 50ps when it reaches 6000K. Then it decreases from 6000K to 3000K at a rate of 0.05K / step with a step size of 1fs, and relaxes for 30ps when it reaches 3000K. Then it decreases from 3000K to 300K at a rate of 0.05K / step with a step size of 1fs, and relaxes for 50ps when it reaches 300K.

[0119] In a specific example, structural operator data within the model was collected at 300K, with a collection frequency of 500 steps / time, for a total of 500 times. The collected data included: radial distribution function, coordination number, angular distribution function, anion bridging degree, and ferropolymerization degree. The structural operator in this embodiment includes the distribution of oxygen, bridging oxygen, and non-bridging oxygen around Nd, as shown in Table 3 below:

[0120] Table 3 shows the distribution of oxygen, bridging oxygen, and non-bridging oxygen around Nd at some sample points.

[0121]

[0122]

[0123] Example 6

[0124] This embodiment establishes a composition-structure-property model of NCS glass based on the structure operators and property data of the NCS glass system using machine learning.

[0125] In this embodiment, the structural operators include: Si-Si bond length, coordination number, Nd-Nd bond length, coordination number, Nd-Si bond length, coordination number, Si-Nd bond length, coordination number, Nd-Na bond length, coordination number, Na-Nd bond length, coordination number, Nd-Ca bond length, coordination number, Ca-Nd bond length, coordination number, Na-Si bond length, coordination number, Si-Na bond length, coordination number, Ca-Si bond length, coordination number, Si-Ca bond length, coordination number, Na-Na bond length, coordination number, Na-Ca Bond length, coordination number, Ca-Na bond length, coordination number, Ca-Ca bond length, coordination number, Si-O bond length, coordination number, Si-BO bond length, coordination number, Si-NBO bond length, coordination number, Nd-O bond length, coordination number, Nd-BO bond length, coordination Number, Nd-NBO bond length, coordination number, Na-O bond length, coordination number, Na-BO bond length, coordination number, Na-NBO bond length, coordination number, Ca-O bond length, coordination number, Ca-BO bond length, coordination number, Ca-NBO bond length, coordination number, Q 0 Q 1 Q 2 Q 3 Q 4 Percentage.

[0126] In this embodiment, the property data include: refractive index, Ω2, Ω4, Ω6, fluorescence branching ratio, radiative transition lifetime, effective linewidth, absorption cross section, and emission cross section.

[0127] In this embodiment, the Pearson correlation coefficient method is used to exclude 23 correlation structure operators, reducing the number of variables from 51 to 28.

[0128] In this embodiment, a random forest model is used, with a training group accounting for 70% and a test group accounting for 30%. Cross-validation is employed to predict glass data and compare it with experiments to verify the model's accuracy. The root mean square errors for refractive index, Ω2, Ω4, Ω6, fluorescence branching ratio, radiative transition lifetime, effective linewidth, absorption cross-section, and emission cross-section are 0.005, 0.058, 0.098, 0.124, 0.001, 16.324, 0.109, 0.062, and 0.065, respectively. 2 The values ​​are 0.97, 0.94, 0.93, 0.94, 0.98, 0.97, 0.92, 0.95, and 0.93, respectively.

[0129] In this embodiment, the importance ranking of structure operators with different properties is obtained, such as Figure 4 As shown, where N Nd-NBO / BO The ratio between the coordination numbers of non-bridging oxygen and bridging oxygen around the rare earth ion Nd is represented by N. Nd-O r represents the coordination number of oxygen ions surrounding the rare earth ion Nd. Na-Si represents the Na-Si bond length, and other represents other structural operators with lower importance. The vertical axis VI represents variable importance, and β represents the fluorescence branching ratio, which is the specific spectral characteristic data in this example.

[0130] In this embodiment, the relationship between key structural operators and glass properties is established as a basis for subsequent glass composition selection. For example... Figure 5 As shown, the vertical axis τ represents the fluorescence lifetime of Nd, and the horizontal axis N Nd-O The coordination number of oxygen ions surrounding Nd ions is a factor, and there is a correlation between the two. To obtain a glass composition with a lifetime of over 500 μs, only the N ion in its structural parameters needs to be increased. Nd-O A value greater than 5.8 is acceptable (as shown by the dotted line).

[0131] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for screening laser glass components, characterized in that, Includes the following steps: Based on the uniform melting compound, the glass-forming region is divided into zones to obtain the physical properties and spectral data of multiple components within the glass-forming region; Obtaining structural characterization operators for components based on molecular dynamics simulations includes the following steps: Construct the initial box model; Energy minimization is performed using gradient descent, conjugate gradient, or steepest descent methods. The long-range Coulomb force and the short-range two-body and three-body potential energy are used to simulate the interparticle forces. High-temperature melting is performed using NVT, NPT, or NVE ensembles. The high-temperature melting temperature is 5000~8000 K, and the equilibrium time is 300~500 ps. Cooling simulations were performed using NVT, NPT, or NVE ensembles. The cooling simulation rate ranged from 0.1 K / ps to 50 K / ps, and the temperature at which the cooling simulation was completed was 250 to 350 K. Obtain structural feature descriptors for the components, including: radial distribution function, coordination number, angular distribution function, anion bridging degree, and degree of polymerization of the forming organism; A training dataset is constructed based on structural feature description operators, physical properties, and spectral data. A composition-structure-property model of glass is constructed. The model is trained using a training dataset to obtain the trained composition-structure-property model of glass, specifically including: Using glass composition as input variable and structural feature descriptor as output variable for model training, the relationship between multiple components and one structural operator is established to obtain the relationship between components and structural feature descriptor. Using structural feature descriptors as input variables and glass composition physical properties and spectral data as output variables for model training, the relationship between multiple structural feature descriptors and a glass property is established respectively, and the relationship between structural feature descriptors and glass composition physical properties and spectral data is obtained. Based on the trained glass composition-structure-property model, the corresponding glass composition physical properties and spectral data are predicted by using glass composition or structural feature descriptors as input variables.

2. The laser glass component screening method according to claim 1, characterized in that, The method for confirming a uniformly melting compound is any one of the following: Confirmed using a binary or ternary phase diagram database; Or refer to the phase diagrams of other systems with similar crystallization chemistry conditions to the corresponding system; Alternatively, the extreme point component whose properties vary with the composition can be selected as a potential uniformly melting compound; Alternatively, starting from first principles and combining with structure search algorithms, we can search for compounds formed under different atomic ratios, and then comprehensively judge whether they are uniformly melted compounds from thermodynamic and kinetic perspectives.

3. The laser glass component screening method according to claim 1, characterized in that, The step of partitioning the glass-forming region based on a uniformly molten compound and obtaining the physical properties and spectral data of multiple components within the glass-forming region includes: After the glass-forming region is divided into zones, each sample component is located within a triangle formed by three nearest neighbor homogeneous molten compounds. Based on the properties of the homogeneous molten compounds at the vertices of the triangles, the physical properties and spectral data of multiple components within the glass-forming region are calculated using the lever principle.

4. The laser glass component screening method according to claim 3, characterized in that, The physical properties and spectral data of multiple components within the glass-forming region are calculated based on the lever principle. The principles followed are: selecting from the polygon of the nearest homogeneous molten compound of the target component; and connecting homogeneous molten compounds with lower melting temperatures is more likely.

5. The laser glass component screening method according to claim 3, characterized in that, The formula for calculating the lever principle is expressed as follows: ； Among them, L A L B and L C P(A), P(B), and P(C) represent the molar contents of uniformly molten compounds A, B, and C of glass with component X, calculated according to the lever principle. P(A), P(B), and P(C) represent the properties of uniformly molten compounds A, B, and C, respectively.

6. The laser glass component screening method according to claim 1, characterized in that, This also includes calculations of the properties of binary glasses and one of their nearest congruent, homologous molten compounds, specifically expressed as follows: ； Where P(Y) represents the property of the homologous glass of the desired molten compound, P(X) represents the property of the binary glass, P(Z) represents the property of the other nearest-neighbor homologous glass of the same composition, and L... Z and L Y These represent the molar contents of the uniformly molten compounds Y and Z, respectively, of glass with component X, calculated using the lever principle.

7. A method for preparing laser glass, characterized in that, According to the laser glass component screening method of any one of claims 1-6, a laser glass component is obtained, and laser glass with corresponding properties is prepared based on the laser glass component.

Images