Method for optimizing sublimation and purification parameters of high-purity molybdenum trioxide

Abstract

The invention discloses a high-purity molybdenum trioxide sublimation purification parameter optimization method, and relates to the technical field of high-molecular compound purification and crystal growth. According to the method, through quantitative analysis of a particle size distribution skewness value S and a particle size distribution kurtosis value K, a raw material particle size state factor Gsize is constructed, and digital evaluation of raw material particle size distribution is achieved; whether particle size distribution is concentrated and reasonable or not is effectively recognized before sublimation, abnormal raw materials with too large or too small particle sizes are screened out, and the phenomenon of uneven steam release caused by particle size difference in the sublimation process is promoted to be reduced. Image recognition and a KNN classification model are introduced, the proportion of spherical particles is further screened, aggregates are removed, the proportion of the spherical particles, the flaky particles, the needle-shaped particles and the aggregate particles is calculated, the proportion of the spherical particles is kept at a high level, and meanwhile the proportion of the aggregate particles is controlled to be lower than a preset threshold value. Local disturbance during steam release can be reduced, and the consistency of crystal nucleation in the mass transfer process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of polymer compound purification and crystal growth, and in particular to a method for optimizing purification parameters of high-purity molybdenum trioxide by sublimation. Background Art

[0002] Molybdenum trioxide is an important inorganic compound widely used in optoelectronic devices, catalytic materials, and high-end alloy additives, requiring extremely high purity and crystal structure. Sublimation purification technology, a key process in producing high-purity molybdenum trioxide, utilizes its relatively low sublimation temperature to transform the raw material into a gas-solid phase, effectively removing metallic and non-volatile impurities.

[0003] In the traditional molybdenum trioxide sublimation purification process, the uneven distribution of raw material particle size is one of the important factors affecting the sublimation quality and purification efficiency. Due to the lack of systematic raw material pretreatment and quantitative analysis mechanisms, raw powders of mixed particle sizes are often directly introduced into the sublimation chamber. Particles with larger particle sizes release steam with a delay due to their slow sublimation rate, resulting in local insufficient steam concentration in the sublimation area; while particles with too small particle sizes are easy to vaporize and quickly release large amounts of steam, forming local high-concentration areas. The difference in steam release rate caused by this uneven particle size distribution can easily cause airflow disturbances and temperature fluctuations in the sublimation area, thereby affecting the stable mass transfer conditions between the steam and the crystal plane, resulting in unstable crystal nucleation positions, uneven grain sizes, and disordered crystallization directions.

[0004] Furthermore, if the particle size distribution contains a large number of agglomerates or irregular flake- or needle-shaped particles, it will further disrupt the thermal field distribution and surface sublimation uniformity. In particular, agglomerated particles release uneven vapor during sublimation, even causing "false volatilization" and interfering with the integrity of the subsequent crystallization structure. Consequently, conventional technologies have failed to introduce clear indicators for evaluating particle size status and lack a mechanism to guide the setting of sublimation conditions based on particle size distribution characteristics, severely restricting the consistency of sublimation purification and the ability to reproducibly obtain high-purity products. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide to solve the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented by the following technical solutions: A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide, comprising the following steps: Step 1: Perform a particle size distribution test on the molybdenum trioxide raw material before sublimation, using a laser particle size analyzer, image particle size analysis or SEM image analysis method to obtain the particle size distribution spectrum of the raw material and the particle size distribution skewness value S and the particle size distribution kurtosis value K, and construct an initial particle size state factor Gsize; when Gsize ≥ 0.7 and K ≥ 3, it indicates that the current molybdenum trioxide raw material is pre-monitored and qualified, and is summarized as the first pre-qualified raw material, and proceeds to step 2; Step 2: Photograph the first pre-qualified raw material to obtain a raw material crystal image. After image preprocessing the raw material crystal image, construct a particle feature vector, output a category label through the trained classifier KNN, and calculate the percentage of particles with the kth category label to the total number of particles. Obtain the proportion of spherical particles, the proportion of flaky particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles. When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is considered qualified and proceed to step 3. Step 3: At the beginning of sublimation, the temperature field distribution, gas phase concentration distribution and sublimation wall crystal structure in the sublimation chamber are collected to construct a boundary layer structure model under the initial sublimation conditions; combined with three-dimensional particle tracking technology, the near-wall diffusion rate curve of molybdenum trioxide vapor is obtained to construct the concentration gradient of the sublimation wall boundary layer. and the corresponding steam diffusion flux J to determine whether there is steam accumulation in the boundary layer and a mass transfer bottleneck risk. If so, a first micro-dynamic correction strategy is generated and issued until there is no mass transfer bottleneck risk and then step 4 is entered; Step 4: Use thermal imaging matrix, flow field tracking and sublimation vapor concentration inversion technology to capture the thermal capillary convection path and intensity changes in the sublimation chamber, construct a thermal convection offset vector diagram, and calculate the flow field stability index F TCV ; Preset stable threshold F, when F TCV >F, the system executes the local temperature zone fine-tuning instruction and then enters step 5; Step 5. During the condensation deposition stage, the non-uniform characteristics of sublimation vapor deposition are identified by scanning the condensation surface temperature field and monitoring the morphology evolution map, and the sublimation-condensation asymmetry factor ASC is constructed. The symmetry critical threshold X is preset. When ASC>X, the condensation symmetry is corrected by dynamically adjusting the condensation surface temperature distribution and the crystal plane guidance direction until crystallization occurs.

[0007] Preferably, step one includes: S1. Conduct a particle size distribution test on the molybdenum trioxide raw material before sublimation, using a laser particle size analyzer, image particle size analysis, or SEM image analysis to obtain the particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, and construct an initial particle size state factor Gsize; The specific steps for obtaining the initial particle size state factor Gsize are: S11, particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, to obtain 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K; S121. The method for obtaining the particle size distribution skewness value S is: Where n represents the total number of granularity data points, Indicates the center value of the i-th particle size, the representative particle size of each particle size interval, including 1.2μm and 1.6μm; derived from the particle size analysis equipment; represents the average particle size, represents the standard deviation, Indicates particle size The corresponding volume fraction is the normalized value; S122. Evaluate the particle size distribution skewness value S to obtain a first evaluation result, including: When S = 0, it means that the distribution is completely symmetric with respect to the mean; When S>0, it means that the distribution is right-skewed relative to the mean; When S < 0, it means that the distribution is left-skewed relative to the mean.

[0008] Preferably, step one further comprises: S123. The method for obtaining the particle size distribution kurtosis value K is: The formula means that the standardized value of the fourth-order central moment reflects the peak degree and tail thickness of the distribution; the kurtosis value K of the particle size distribution is the fourth power of the distance. The "extreme values ​​far from the mean" in the distribution have a greater impact, which can be used to determine whether the distribution is peaked or flat; S124. Evaluate the particle size distribution kurtosis value K to obtain a second evaluation result, including: When K = 3, it indicates a moderate normal distribution with moderate tail thickness and no obvious central or dispersed shift; When K>3, it means that the distribution is close to the average value, the concentration is high, the tail is heavier, a peak distribution is formed, and the particle size consistency is good, which is conducive to controlling the sublimation uniformity; When K is less than 3, it means that the distribution is close to the mean value, the concentration is low, the tail is more dispersed, and a flat-top distribution is formed; the distribution is flat, and the mass transfer in the sublimation stage is limited and the fluctuation is large; S125. Combine the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S, and particle size distribution kurtosis value K obtained in S121-S123, and after normalization, obtain the initial particle size state factor Gsize through weighted combination: Where, Indicates the distribution width or particle size range of the particle size, reflecting the dispersion; 、 and Expressed as weight, , , ,and ; S126. Evaluate the initial particle size state factor Gsize. If both of the following judgment conditions are met, including: if Gsize is less than 0.7 and K is less than 3, indicating that the distribution is unqualified and will lead to abnormal risks in the subsequent sublimation stage, trigger the first warning instruction, including: screening the raw materials, adjusting the ratio, ultrasonic vibration or mechanical crushing to promote the deagglomeration of the raw material particles; When Gsize≥0.7 and K≥3, it means that the current molybdenum trioxide raw material is qualified in the pre-monitoring, and is summarized as the first pre-qualified raw material, and enters step 2.

[0009] Preferably, step 2 includes: S21, photographing the first pre-qualified raw material to obtain a raw material crystal image, and performing image preprocessing on the raw material crystal image, including: grayscale and binarization of the raw material crystal image, removing background noise and enhancing edge contrast, extracting single particle boundary contours, and completing particle segmentation; and closing each particle contour to eliminate non-target areas; The roundness feature, aspect ratio feature, boundary irregularity feature and area ratio feature of each particle contour are extracted to obtain the roundness of the particle. , aspect ratio , boundary irregularity And the ratio of the projected area to the convex hull area ; S2101, roundness of particles The calculation expression is: Among them, A is the projected area of ​​the particle, P is the boundary perimeter of the particle, and the closer the roundness value is to 1, the more regular and spherical the particle is; S2102, aspect ratio of particles The calculation expression is: Among them, L and W are the major axis length and minor axis length of the particle fitting circumscribed rectangle, respectively, reflecting the stretching degree of the particle; S2103, particle boundary irregularity The calculation expression is: Among them, the larger the value, the more complex the boundary and the more irregular the morphology; S2104, the ratio of the particle's projected area to its convex hull area The calculation expression is: in, Indicates the convex hull area of ​​the particle, which is used to determine whether there is agglomeration or adhesion; S22, the roundness of the particles , aspect ratio , boundary irregularity And the ratio of the projected area to the convex hull area Compose the eigenvectors of the particles in order: The characteristic vector of each particle Input the trained classifier KNN and calculate the feature vector of the particle through the trained classifier KNN The Euclidean distance between the feature vectors of all training samples is used to find the first k neighbors with the closest distance, and count which category appears most among these k neighbors to obtain the corresponding category and output the category label; Category labels include: 001 spherical, 002 flake, 003 needle / columnar, and 004 agglomerated; Counting the total number of particles of the kth category label in the first pre-qualified raw material, and calculating the percentage of particles of the kth category label in the total number of particles, to obtain the proportion of spherical particles, the proportion of flake particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles; When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is qualified and enters step 3; if the conditions are not met, it is unqualified, that is, "insufficient sphericity" or "too much agglomeration", which means that it needs to be returned and re-screened.

[0010] Preferably, step three includes: S311. At the initial stage of sublimation, using a temperature sensor array and an infrared thermal imaging system to collect the temperature field distribution in the sublimation chamber; S312, using a gas phase concentration sensor or a Raman / infrared spectrometer to collect gas phase molybdenum trioxide concentration distribution in the sublimation chamber; S313, obtaining the sublimation wall crystal structure by in-situ surface diffraction or reflection electron diffraction; S314. Combined with 3D-PTV, the motion trajectory of molybdenum trioxide vapor particles during the sublimation process is tracked to extract the velocity field, diffusion path and time series of boundary layer vapor particles. S315. During the sublimation of molybdenum trioxide, the raw material solid is directly converted into vapor at high temperature. The thin layer area between the sublimation wall and the main airflow is marked as the boundary layer. In the boundary layer, the concentration, temperature and flow rate will change significantly. It is also the place where the diffusion of gas phase substances is most "difficult"; S316, constructing the vapor diffusion rate curve of the boundary layer, and calculating the concentration gradient of the sublimation wall boundary layer based on the vapor diffusion rate curve of the boundary layer : Where C is the concentration of molybdenum trioxide vapor in the boundary layer, and x is the direction perpendicular to the sublimation wall; It represents the derivative of concentration with respect to position, reflecting "how fast the vapor concentration decreases from the wall to the outside"; The value representing the concentration gradient of the sublimation wall boundary layer.

[0011] Preferably, step three further includes: S317. According to Fick's first law, the vapor diffusion flux J is proportional to the concentration gradient of the sublimation wall boundary layer, and the expression is as follows: Where D represents the diffusion coefficient, including: Under atmospheric pressure air sublimation conditions, the sublimation temperature range of molybdenum trioxide is 700-850°C, corresponding to a temperature of 973–1123K. When the pressure is 1.0 atm, the diffusion coefficient D under these conditions is: 0.08-0.12 cm² / s; Under the conditions of sublimation in reduced pressure air, the sublimation temperature range of molybdenum trioxide is 600-750°C, corresponding to a temperature of 873–1023. When the pressure is 0.1atm, the diffusion coefficient D under this condition is: 0.60-0.90cm² / s; Under Ar or N2 inert gas protection sublimation conditions, the sublimation temperature range of molybdenum trioxide is 650-800°C, corresponding to a temperature of 923–1073, and a pressure of 1.0 atm. The diffusion coefficient D under these conditions is: 0.10-0.15 cm² / s; Under vacuum sublimation conditions, the sublimation temperature range of molybdenum trioxide is 550–700°C, corresponding to a temperature of 823–973K and a pressure of 10 -4 atm, the diffusion coefficient D under this condition is: 10–25 cm² / s, and increases with increasing temperature; The negative sign indicates that the substance moves from areas of high concentration to areas of low concentration.

[0012] Preferably, step three further includes: S318, preset critical threshold G1 and diffusion flux efficiency threshold J1, when the concentration gradient of the sublimation wall boundary layer When the critical threshold G1 is set, and the vapor diffusion flux J < diffusion flux efficiency threshold J1, it indicates that the vapor is accumulated in the boundary layer, and there is a risk of mass transfer bottleneck, and a first micro dynamic correction strategy is generated, including: the change amplitude of the temperature gradient is adjusted by 0.5°C-2°C each time, and the vibration amplitude of the boundary layer is increased by 1 µm-2 µm each time; After the first micro dynamic correction strategy is executed, the concentration gradient of the sublimation wall boundary layer and the vapor diffusion flux are re-detected after stabilization, until ≤G1, and the vapor diffusion flux J≥J1, then step four is entered; when ≤G1, and the vapor diffusion flux J≥J1, it indicates that the mass transfer condition is normal, and the vapor diffusion flux efficiency meets the standard, and the sublimation state is continuously monitored.

[0013] Preferably, the step four includes: S41, using thermal imaging and airflow sensors, monitoring the temperature and airflow changes in the sublimation cavity to obtain thermal field data and flow field data in the cavity; S42, according to the obtained thermal field data and flow field data in the cavity, combining with the sublimation vapor concentration inversion information, constructing a thermal convection displacement vector map in the sublimation cavity, and calculating a flow field stability index F TCV : Wherein, N represents the total number of sampling points, represents the airflow velocity vector of the jth point, represents the average velocity vector, represents the temperature of the jth point, represents the average temperature, represents the weight coefficient of temperature fluctuation; The setting range of includes: In the initial coarse adjustment stage, α=1.0, considering the speed and temperature fluctuation; In the fine monitoring stage: α=1.5-2.0, the temperature disturbance detection ability is strengthened; In the local condensation and re-crystallization stage, α=2.5; represents the dimensionless relative speed fluctuation, represents the dimensionless relative temperature fluctuation; S43, a stability threshold F is preset, and whether the flow field stability index F TCV is beyond the set stability threshold F, if yes→triggering a local temperature zone fine adjustment instruction to suppress the abnormal thermal convection, the local temperature zone fine adjustment instruction includes: arranging a cold source on the back or side of the temperature zone for directional heat adjustment by 0.2-0.8°C, and according to the local airflow direction disturbance, triggering the control of the guide vane adjustment to close the direction by ±5°-20°. If the flow field stability index F TCV If the stability threshold F is not exceeded, the current parameters are maintained, monitoring is continued, and the process proceeds to step 5.

[0014] Preferably, the step five includes: S51. Use a high-resolution infrared thermal imager and a morphology evolution map acquisition device to continuously scan the condensation surface to obtain the following parameters and construct a condensation surface deposition uniformity dataset, including: condensation surface temperature distribution standard deviation, condensation crystal thickness distribution, and condensation crystal deposition morphology image; S52. Based on the uniform deposition data set on the condensation surface, the sublimation-condensation asymmetry factor ASC is constructed to quantify the degree of imbalance between the deposition distribution and the thermal field distribution on the condensation surface. The expression is: Where, represents the standard deviation of crystal thickness, represents the average value of the crystal thickness, represents the standard deviation of the condensation surface temperature distribution, represents the average value of the temperature field during the condensation process, Indicates the symmetry deviation of the topographic image; 、 and Expressed as weight, , , ,and ; in, The way to obtain is: For the condensed crystal deposition morphology image, mirror it left and right or up and down according to the central axis of the condensation surface, and use the mean square error (MSE) method to calculate the image difference and obtain the symmetry deviation of the morphology image. : Where M represents the total number of pixels in the condensed crystal deposition morphology image, represents the pixel value of the kth pixel in the condensation crystal deposition morphology image, It represents the pixel value corresponding to the kth pixel after being mirrored left and right or up and down along the central axis of the condensation surface; If the two images are completely symmetrical and the pixel values ​​at the symmetrical positions are exactly the same, then all ,so , indicating perfect symmetry. If the difference in symmetrical pixel values ​​is large, then The larger the value, the more asymmetric the morphology.

[0015] Preferably, step five further includes: S53, preset symmetric critical threshold X, when >X, it indicates that there is an asymmetric risk of condensation deposition, triggering the second warning instruction; when If ≤X, it indicates that the current condensation deposition process is qualified, and the crystal structure is uniform, with no signs of reflux, and the current batch of crystallization process is marked as qualified and continuously monitored; S54. After receiving the second warning instruction, generate a second slight dynamic correction strategy, reduce the temperature in the asymmetric area by 0.5 to 2°C, promote the migration of deposition to the symmetric area, and guide the crystal to grow along a more stable symmetric direction by rotating or shifting the direction of the seed crystal. After each adjustment, recalculate the sublimation-condensation asymmetry factor ASC to determine whether it falls back to ≤X. If not, continue the second slight dynamic correction strategy until the deposition returns to a symmetric state.

[0016] The present invention provides a method for optimizing the parameters of sublimation purification of high-purity molybdenum trioxide. It has the following beneficial effects: (1) This method is a parameter optimization method for the sublimation purification of high-purity molybdenum trioxide. By quantitatively analyzing the particle size distribution skewness value S and the particle size distribution peak value K, the raw material particle size state factor Gsize is constructed, and the digital evaluation of the raw material particle size distribution is realized. It can effectively identify whether the particle size distribution is concentrated and reasonable before sublimation, screen out abnormal raw materials with too large or too small particle sizes, significantly reduce the uneven steam release caused by particle size differences during sublimation, and improve sublimation stability. By introducing image recognition and KNN classification models, the proportion of spherical particles is further screened and agglomerates are eliminated. By calculating the proportion of spherical particles, flaky particles, needle-shaped particles and agglomerated particles, the proportion of spherical particles is maintained at a high level. At the same time, the proportion of agglomerated particles is controlled below the preset threshold, which can reduce local disturbances during steam release and improve the stability of the mass transfer process and the consistency of crystal nucleation.

[0017] (2) This method for optimizing the parameters of high-purity molybdenum trioxide sublimation purification can effectively identify the accumulation trend of steam in the boundary layer by monitoring the vapor concentration gradient and vapor diffusion flux J near the wall in the initial stage of sublimation, and adjust the sublimation conditions in time to avoid the reduction of sublimation rate and condensation caused by mass transfer bottlenecks. During the sublimation process, the changes in the thermal convection path and the flow field stability index F are monitored. TCV The sublimation zone temperature can be fine-tuned before thermal convection drifts or becomes turbulent, ensuring the stability of the vapor atmosphere and reducing the risk of defects during the crystallization process. During the condensation stage, the sublimation-condensation asymmetry factor (ASC) is quantified by analyzing the spatial distribution of sublimation vapor deposition and the temperature gradient of the condensation surface. When the ASC exceeds a critical value, the condensation surface temperature and crystal orientation are corrected, effectively improving grain size consistency and structural integrity.

[0018] (3) The present invention further introduces a KNN (K-nearest neighbor) classifier as an intelligent classification tool for condensation surface deposition symmetry images. KNN, as a supervised learning algorithm based on sample similarity, is a basic classification model in the field of artificial intelligence. Without the need for complex model training, it can output category labels through a trained KNN classifier and calculate the percentage of particles with the kth category label in the total number of particles, thereby obtaining the proportion of spherical particles, the proportion of flaky particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles. This significantly improves the accuracy and efficiency of particle morphology recognition and reduces the probability of human intervention and misjudgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic diagram of the steps of the artificial intelligence high-purity molybdenum trioxide sublimation purification parameter optimization method of the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] Secondly, during the sublimation process, steam molecules form a high-concentration boundary layer as they approach the sublimation wall. If this vapor accumulation in the boundary layer is not promptly identified and addressed, a "mass transfer bottleneck" can easily occur, leading to a decrease in the gas-phase concentration gradient, a drop in the sublimation rate, and the induction of localized recondensation or crystallization defects. This issue is often overlooked in traditional processes, which lack quantitative monitoring of the boundary layer concentration distribution and diffusion flux.

[0022] Furthermore, temperature differences often drive complex capillary convection within the sublimation chamber. Especially when vapor concentrations are high and air viscosity fluctuates significantly, localized convection paths can drift or reverse, disrupting the stable atmosphere for crystal growth and reducing crystal quality. Traditional sublimation systems often rely on static thermal field control, making it difficult to sense and respond to dynamic changes in the convection structure in real time.

[0023] Finally, during the condensation deposition stage, the deposition behavior of sublimated vapor is significantly affected by the temperature distribution of the condensation surface and the control of crystal orientation. If the condensation conditions are asymmetric, it will lead to non-uniform grain size, unbalanced stress in the deposited layer, and even cause "reflow" phenomenon, which will destroy the homogeneity of crystallization. However, existing technologies lack effective characterization and feedback control methods for condensation symmetry.

[0024] Example 1 A method for optimizing the parameters of high-purity molybdenum trioxide sublimation purification comprises the following steps: Step 1: Perform a particle size distribution test on the molybdenum trioxide raw material before sublimation, using a laser particle size analyzer, image particle size analysis or SEM image analysis method to obtain the particle size distribution spectrum of the raw material and the particle size distribution skewness value S and the particle size distribution kurtosis value K, and construct an initial particle size state factor Gsize; when Gsize ≥ 0.7 and K ≥ 3, it indicates that the current molybdenum trioxide raw material is pre-monitored and qualified, and is summarized as the first pre-qualified raw material, and proceeds to step 2; Step 2: Photograph the first pre-qualified raw material to obtain a raw material crystal image. After image preprocessing the raw material crystal image, construct a particle feature vector, output a category label through the trained classifier KNN, and calculate the percentage of particles with the kth category label to the total number of particles. Obtain the proportion of spherical particles, the proportion of flaky particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles. When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is considered qualified and proceed to step 3. Step 3: At the beginning of sublimation, the temperature field distribution, gas phase concentration distribution and sublimation wall crystal structure in the sublimation chamber are collected to construct a boundary layer structure model under the initial sublimation conditions; combined with three-dimensional particle tracking technology, the near-wall diffusion rate curve of molybdenum trioxide vapor is obtained to construct the concentration gradient of the sublimation wall boundary layer. and the corresponding steam diffusion flux J to determine whether there is steam accumulation in the boundary layer and a mass transfer bottleneck risk. If so, a first micro-dynamic correction strategy is generated and issued until there is no mass transfer bottleneck risk and then step 4 is entered; Step 4: Use thermal imaging matrix, flow field tracking and sublimation vapor concentration inversion technology to capture the thermal capillary convection path and intensity changes in the sublimation chamber, construct a thermal convection offset vector diagram, and calculate the flow field stability index F TCV ; Preset stable threshold F, when F TCV >F, the system executes the local temperature zone fine-tuning instruction and then enters step 5; Step 5. During the condensation deposition stage, the non-uniform characteristics of sublimation vapor deposition are identified by scanning the condensation surface temperature field and monitoring the morphology evolution map, and the sublimation-condensation asymmetry factor ASC is constructed. The symmetry critical threshold X is preset. When ASC>X, the condensation symmetry is corrected by dynamically adjusting the condensation surface temperature distribution and the crystal plane guidance direction until crystallization occurs.

[0025] In this embodiment, this method constructs the raw material particle size state factor Gsize through quantitative analysis of the particle size distribution skewness value S and the particle size distribution kurtosis value K, achieving digital evaluation of the raw material particle size distribution. It can effectively identify whether the particle size distribution is concentrated and reasonable before sublimation, screen out abnormal raw materials with excessively large or small particle sizes, significantly reduce the uneven steam release caused by particle size differences during sublimation, and improve sublimation stability. By introducing image recognition and KNN classification models, the proportion of spherical particles is further screened and agglomerates are eliminated. By calculating the proportions of spherical particles, flaky particles, needle-shaped particles, and agglomerated particles, the proportion of spherical particles is maintained at a high level. At the same time, the proportion of agglomerated particles is controlled below a preset threshold. This can reduce local disturbances during steam release and improve the stability of the mass transfer process and the consistency of crystal nucleation.

[0026] In the early stage of sublimation, by monitoring the vapor concentration gradient and vapor diffusion flux J near the wall, the accumulation trend of vapor in the boundary layer can be effectively identified, and the sublimation conditions can be adjusted in time to avoid the reduction of sublimation rate and condensation caused by mass transfer bottlenecks. During the sublimation process, by monitoring the changes in the heat convection path and the flow field stability index F TCV The sublimation zone temperature can be fine-tuned before thermal convection drifts or becomes turbulent, ensuring the stability of the vapor atmosphere and reducing the risk of defects during the crystallization process. During the condensation stage, the sublimation-condensation asymmetry factor (ASC) is quantified by analyzing the spatial distribution of sublimation vapor deposition and the temperature gradient of the condensation surface. When the ASC exceeds a critical value, the condensation surface temperature and crystal orientation are corrected, effectively improving grain size consistency and structural integrity.

[0027] Example 2 This embodiment is explained in Example 1. Specifically, step 1 includes: S1. Conduct a particle size distribution test on the molybdenum trioxide raw material before sublimation, using a laser particle size analyzer, image particle size analysis, or SEM image analysis to obtain the particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, and construct an initial particle size state factor Gsize; The specific steps for obtaining the initial particle size state factor Gsize are: S11, particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, to obtain 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K; S121. The method for obtaining the particle size distribution skewness value S is: Where n represents the total number of granularity data points, Indicates the center value of the i-th particle size, the representative particle size of each particle size interval, including 1.2μm and 1.6μm; derived from the particle size analysis equipment; represents the average particle size, represents the standard deviation, Indicates particle size The corresponding volume fraction is a normalized value, for example, 1.2 μm particles account for 5.3% of the total volume → f i =0.053; if the volume fraction of 1.6 μm particles is 8.2% → f i =0.082; S122. Evaluate the particle size distribution skewness value S to obtain a first evaluation result, including: When S = 0, it means that the distribution is completely symmetrical with respect to the mean, with the left and right sides mirrored; When S>0, it means that the distribution is right-skewed relative to the mean, the probability of large particle size is high, and there is a "tail" on the right side; When S<0, it means that the distribution is left-skewed relative to the mean, with a large proportion of small particles and a "tail" on the left; Since the third central moment of the normal distribution is 0, 0 is used as the symmetry reference.

[0028] In this example, the introduction of the particle size distribution skewness value S allows for quantitative assessment of the asymmetry of the raw material particle size distribution, effectively identifying whether small or large particles dominate the raw material. This helps avoid imbalances in vapor release rates caused by extreme particle size deviations during the sublimation process. By introducing the skewness value S, a third-order statistic, the accuracy of particle size assessment is further improved beyond the traditional approach of examining only the D50 value. This allows for screening sublimation raw materials based on their distribution morphology, minimizing the impact of abnormal distribution on the subsequent sublimation thermal and gas fields, and improving crystallization consistency.

[0029] Example 3 This embodiment is explained in Example 2. Specifically, step 1 also includes: S123. The method for obtaining the particle size distribution kurtosis value K is: The formula means that the standardized value of the fourth-order central moment reflects the peak degree and tail thickness of the distribution; the kurtosis value K of the particle size distribution is the fourth power of the distance. The "extreme values ​​far from the mean" in the distribution have a greater impact, which can be used to determine whether the distribution is peaked or flat; S124. Evaluate the particle size distribution kurtosis value K to obtain a second evaluation result, including: The kurtosis value K of the particle size distribution is the fourth power of the distance. The "extreme values ​​far from the mean" in the distribution have a greater impact. It can be used to determine whether the distribution is peaked or flat. The normal distribution (i.e., the ideal bell curve) is used as a reference. Its kurtosis value is fixed at 3, which is also the standard distribution reference line commonly used in the industry. When K = 3, it indicates a moderate normal distribution with moderate tail thickness and no obvious central or dispersed shift; When K>3, it means that the distribution is close to the average value, the concentration is high, the tail is heavier, a peak distribution is formed, and the particle size consistency is good, which is conducive to controlling the sublimation uniformity; When K is less than 3, it means that the distribution is close to the mean value, the concentration is low, the tail is more dispersed, and a flat-top distribution is formed; the distribution is flat, and the mass transfer in the sublimation stage is limited and the fluctuation is large; The particle size distribution of order 1 to order 4 is shown in Table 1: S125. Combine the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S, and particle size distribution kurtosis value K obtained in S121-S123, and after normalization, obtain the initial particle size state factor Gsize through weighted combination: Where, Indicates the distribution width or particle size range of the particle size, reflecting the dispersion; 、 and Expressed as weight, , , ,and ; S126. Evaluate the initial particle size state factor Gsize. If both of the following judgment conditions are met, including: if Gsize is less than 0.7 and K is less than 3, indicating that the distribution is unqualified and will lead to abnormal risks in the subsequent sublimation stage, trigger the first warning instruction, including: screening the raw materials, adjusting the ratio, ultrasonic vibration or mechanical crushing to promote the deagglomeration of the raw material particles; When Gsize≥0.7, and K≥3, it means that the current molybdenum trioxide raw material pre-monitoring is qualified, and is summarized as the first pre-qualified raw material, and enters step two. K=3 is the standard kurtosis of normal distribution, which is a general reference value in statistics and particle engineering, representing moderate particle size concentration, moderate tail thickness, and no skewness influence. When: K>3: particle size is highly concentrated, with a sharp peak distribution, indicating that most particle sizes are close to the average value. This distribution is more conducive to sublimation uniformity and mass transfer stability; K<3: indicates that the particle size distribution is flat, with many extreme values far from the mean, indicating that the raw material dispersion is too high, and the particles are inconsistent, which is not conducive to the consistency of gasification in the sublimation process, and is prone to "false sublimation" or local incomplete sublimation phenomenon.

[0030] Gsize is a normalized index obtained by weighted combination of D10, D50, D90, skewness S, and kurtosis K, which represents the overall rationality and concentration of particle size distribution. Set 0.7 as the judgment line, the value range of Gsize after normalization is generally [0, 1] or close to the range, 0.7 is a high intermediate critical point, indicating that the particle size structure is overall reasonable and has a certain concentration; Engineering experience value and test verification results show that when Gsize is less than 0.7, the particle size tripartition interval (D90-D10) is usually large, the distribution width is high, indicating that the raw material particle size span is large and the particle distribution is chaotic; below 0.7 usually means that the raw material particles have obvious agglomeration, uneven mixing or incomplete crushing during preparation, which will directly affect the steam release rate and gasification interface stability in the sublimation process; 0.7 as a fault tolerance boundary can be adjusted according to different equipment tolerance, but in most industrial applications, Gsize≥0.7 is the basic sublimable raw material particle size requirement threshold.

[0031] In this embodiment, by introducing the joint evaluation mechanism of particle size distribution skewness S and kurtosis K, the symmetry and concentration of particle size distribution in the raw material can be quantitatively identified, avoiding the steam release unevenness caused by particle size right skew (S>0) or flat distribution (K<3), and effectively improving the steam stability of the sublimation area. By setting the pre-screening threshold of Gsize≥0.7 and K≥3, the quantitative grading of raw material sublimation adaptability can be realized, and the raw material batches with extremely uneven particle size distribution can be removed in advance, ensuring the controllability and crystallization uniformity of the mass transfer process in the sublimation process. Normalize skewness S, kurtosis K, and particle size tripartition D10 / D50 / D90 to construct the initial particle size state factor Gsize, which can systematically reflect the overall state of particle size distribution, replace the single particle size judgment of traditional experience method, and improve the intelligence and consistency of the quality control link before sublimation. When Gsize<0.7 and K<3, the first warning instruction is triggered, and operations such as ultrasonic vibration and mechanical crushing can be taken to scatter the agglomerates and improve the particle size structure, avoiding the "false volatilization" phenomenon caused by large or irregular agglomeration of particle size, and improving the gasification efficiency in the sublimation stage.

[0032] Example 4 This embodiment is explained in Example 1. Specifically, step 2 includes: S21, photographing the first pre-qualified raw material to obtain a raw material crystal image, and performing image preprocessing on the raw material crystal image, including: grayscale and binarization of the raw material crystal image, removing background noise and enhancing edge contrast, extracting single particle boundary contours, and completing particle segmentation; and closing each particle contour to eliminate non-target areas; The roundness feature, aspect ratio feature, boundary irregularity feature and area ratio feature of each particle contour are extracted to obtain the roundness of the particle. , aspect ratio , boundary irregularity And the ratio of the projected area to the convex hull area ; S2101, roundness of particles The calculation expression is: Among them, A is the projected area of ​​the particle, P is the boundary perimeter of the particle, and the closer the roundness value is to 1, the more regular and spherical the particle is; S2102, aspect ratio of particles The calculation expression is: Among them, L and W are the major axis length and minor axis length of the particle fitting circumscribed rectangle, respectively, reflecting the stretching degree of the particle; S2103, particle boundary irregularity The calculation expression is: Among them, the larger the value, the more complex the boundary and the more irregular the morphology; S2104, the ratio of the particle's projected area to its convex hull area The calculation expression is: in, Indicates the convex hull area of ​​the particle, which is used to determine whether there is agglomeration or adhesion; S22. For example, the four characteristic values ​​of a particle are: roundness = 0.95; Aspect ratio = 1.2; Boundary irregularity = 1.3; The ratio of the particle's projected area to its convex hull area = 0.92; Then the characteristic vector of the particle : The characteristic vector of each particle Input the trained classifier KNN and calculate the feature vector of the particle through the trained classifier KNN The Euclidean distance between the feature vectors of all training samples is used to find the first k neighbors with the closest distance, and count which category appears most among these k neighbors to obtain the corresponding category and output the category label; The characteristic vector of each particle Input the trained KNN classifier. The trained KNN classifier has training samples of known categories. Each sample has a corresponding feature vector, including: Sample 1: , category = spherical; sample 2: , category=flaky; Sample 3: , category = needle-shaped / columnar; sample 4: , category = agglomerate; Calculate the feature vector of the particle through the trained classifier KNN The Euclidean distance between the feature vectors of all training samples is used to find the first k neighbors with the closest distance, and count which category appears most among these k neighbors to obtain the corresponding category and output the category label; Category labels include: 001 spherical, 002 flake, 003 needle / columnar, and 004 agglomerated; Counting the total number of particles of the kth category label in the first pre-qualified raw material, and calculating the percentage of particles of the kth category label in the total number of particles, to obtain the proportion of spherical particles, the proportion of flake particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles; When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is qualified and enters step 3; if the conditions are not met, it is unqualified, that is, "insufficient sphericity" or "too much agglomeration", which means that it needs to be returned and re-screened.

[0033] Spherical particles have the smallest specific surface area and optimal spatial packing properties, forming a dense and uniform raw material bed during the packing process. This promotes uniform heat transfer within the bed and reduces instabilities such as crystal crusting or cracking caused by localized overheating. Setting a spherical particle ratio of ≥50% helps ensure a smooth and stable sublimation interface, improving sublimation yield and crystal quality. Agglomerated particles often have large volumes and complex morphologies, which can lead to incomplete internal sublimation, uneven surface vaporization, or structural collapse during the sublimation process. In severe cases, this can lead to blockage of the sublimation path or localized overheating. Setting an agglomerated particle ratio of <10% is intended to control the consistency of the raw material during mass transfer and vaporization, reduce the risk of abnormal sublimation, and ensure a continuous and stable sublimation process. Statistical analysis of morphology distribution data for sublimation materials such as molybdenum trioxide shows that the spherical particle ratio of general production-grade raw materials after primary screening and particle size control is typically between 40% and 70%. Setting a ratio of ≥50% ensures that the median level of morphology deviation remains within a controllable range, rather than experiencing sporadic or localized improvements. Comparative experiments on sublimation of multiple batches of raw materials show that when the proportion of spherical particles exceeds 50%, the sublimation temperature control curve becomes more stable, the fluctuation in the average particle size of the finished crystals decreases by 10% to 15%, and the sublimation yield increases by approximately 5% on average. However, when the proportion of agglomerated particles exceeds 10%, abnormalities such as bed deformation and product contamination are more likely to occur in the middle and late stages of sublimation. Therefore, this determination method, derived from the results of multiple rounds of process validation, is an effective strategy for quantitative pre-sublimation morphology screening.

[0034] The following is an example of the output category labels for particle sample numbers, as shown in Table 2: Spherical samples have a high roundness (close to 1), an aspect ratio close to 1, low boundary irregularity, and a high area percentage. Flake samples have a medium to low roundness, a slightly larger aspect ratio, medium boundary irregularity, and a low area percentage. Needle / columnar samples have a high roundness but a significantly larger aspect ratio, high boundary irregularity, and a moderate area percentage. Agglomerated samples have a low roundness, a moderate aspect ratio, but the highest boundary irregularity and the lowest area percentage.

[0035] In this embodiment, the image processing and feature extraction methods are used to avoid the subjective errors caused by manual judgment, and can objectively and accurately identify typical particle morphologies such as spherical, flaky, needle-shaped and agglomerated. By calculating the proportion of particles with different morphologies, a standardized morphology evaluation mechanism is established, especially setting dual constraints for the proportion of spherical particles ≥50% and the proportion of agglomerated particles <10%, to ensure that the raw materials have good morphological uniformity and dispersibility during the sublimation process, which helps to control the stability of the sublimation interface and the uniformity of the sublimation rate. If the particle morphology does not meet the standards, such as "insufficient sphericity" or "too much agglomeration", the screening or deagglomeration process can be triggered before sublimation to avoid the abnormalities caused by the raw material morphology. This method establishes a traceable "raw material morphology → sublimation behavior" path for the subsequent sublimation process, and provides a quantitative basis for the setting of key process parameters and quality control. Problems such as local sintering, crusting or uneven sublimation in the sublimation process.

[0036] Example 5 Secondly, during the sublimation process, steam molecules form a high-concentration boundary layer as they approach the sublimation wall. If this vapor accumulation in the boundary layer is not promptly identified and addressed, a "mass transfer bottleneck" can easily occur, leading to a decrease in the gas-phase concentration gradient, a drop in the sublimation rate, and the induction of localized recondensation or crystallization defects. This issue is often overlooked in traditional processes, which lack quantitative monitoring of the boundary layer concentration distribution and diffusion flux.

[0037] This embodiment is explained in Example 1. Specifically, step three includes: S311. At the initial stage of sublimation, using a temperature sensor array and an infrared thermal imaging system to collect the temperature field distribution in the sublimation chamber; S312, using a gas phase concentration sensor or a Raman / infrared spectrometer to collect gas phase molybdenum trioxide concentration distribution in the sublimation chamber; S313, obtaining the sublimation wall crystal structure by in-situ surface diffraction or reflection electron diffraction; S314. Combined with 3D-PTV, the motion trajectory of molybdenum trioxide vapor particles during the sublimation process is tracked to extract the velocity field, diffusion path and time series of boundary layer vapor particles. S315. During the sublimation of molybdenum trioxide, the raw material solid is directly converted into vapor at high temperature. The thin layer area between the sublimation wall and the main airflow is marked as the boundary layer. In the boundary layer, the concentration, temperature and flow rate will change significantly. It is also the place where the diffusion of gas phase substances is most "difficult"; S316, constructing the vapor diffusion rate curve of the boundary layer, and calculating the concentration gradient of the sublimation wall boundary layer based on the vapor diffusion rate curve of the boundary layer : Where C is the concentration of molybdenum trioxide vapor in the boundary layer, and x is the direction perpendicular to the sublimation wall; It represents the derivative of concentration with respect to position, reflecting "how fast the vapor concentration decreases from the wall to the outside"; The value representing the concentration gradient of the sublimation wall boundary layer.

[0038] S317. According to Fick's first law, the vapor diffusion flux J is proportional to the concentration gradient of the sublimation wall boundary layer, and the expression is as follows: Where D represents the diffusion coefficient, including: Under atmospheric pressure air sublimation conditions, the sublimation temperature range of molybdenum trioxide is 700-850°C, corresponding to a temperature of 973–1123K. When the pressure is 1.0 atm, the diffusion coefficient D under these conditions is: 0.08-0.12 cm² / s; Under the conditions of sublimation in reduced pressure air, the sublimation temperature range of molybdenum trioxide is 600-750°C, corresponding to a temperature of 873–1023. When the pressure is 0.1atm, the diffusion coefficient D under this condition is: 0.60-0.90cm² / s; Under Ar or N2 inert gas protection sublimation conditions, the sublimation temperature range of molybdenum trioxide is 650-800°C, corresponding to a temperature of 923–1073, and a pressure of 1.0 atm. The diffusion coefficient D under these conditions is: 0.10-0.15 cm² / s; Under vacuum sublimation conditions, the sublimation temperature range of molybdenum trioxide is 550–700°C, corresponding to a temperature of 823–973K and a pressure of 10 -4 atm, the diffusion coefficient D under this condition is: 10–25 cm² / s, and increases with increasing temperature; The negative sign indicates that the substance moves from areas of high concentration to areas of low concentration.

[0039] S318, preset critical threshold G1 and diffusion flux efficiency threshold J1, when the concentration gradient of the sublimation wall boundary layer When the critical threshold G1 is set and the vapor diffusion flux J is less than the diffusion flux efficiency threshold J1, it means that the vapor accumulates in the boundary layer and there is a risk of mass transfer bottleneck. The first micro-dynamic correction strategy is generated, including: adjusting the temperature gradient by 0.5℃~2℃ each time and increasing the vibration amplitude of the boundary layer by 1µm~2µm each time, increasing the local surface temperature to enhance the volatilization rate and diffusion rate of the vapor, or adjusting the temperature distribution uniformity to reduce the accumulation of local high concentrations; the purpose of vibration is to disturb the boundary layer gas, reduce concentration accumulation, and promote gas diffusion and ventilation; and re-test the concentration gradient and vapor diffusion flux of the sublimation wall boundary layer after stabilization until ≤G1, and the vapor diffusion flux J≥J1, then enter step 4; when When ≤G1 and the vapor diffusion flux J≥J1, it means that the mass transfer conditions are normal, the vapor diffusion flux efficiency meets the standard, and the sublimation state is continuously monitored.

[0040] Setting the critical upper limit of the concentration gradient G1 means that if this value is exceeded, gas accumulation or uneven sublimation may occur in the system. If the concentration gradient in the boundary layer is too large, it means that the vapor concentration changes drastically per unit length, and vapor accumulates in the boundary layer, posing a risk of mass transfer "blockage". After statistical analysis of multiple data sets, it was found that: Example: G1 = 0.25-0.35 mol / cm 4 , crystal segregation, agglomeration, and irregular growth often occur, indicating that gas diffusion is restricted and accumulation is obvious; Under different atmosphere conditions (vacuum, inert gas, atmospheric pressure), combined with the typical range of diffusion coefficient D (e.g., 0.1–25 cm² / s), simulate / measure the J value under various operating conditions. Simultaneously, evaluate the corresponding product crystallization conditions (whether there are defects, agglomeration, wall sticking, etc.) to determine the "minimum effective diffusion flux." For example, under vacuum sublimation conditions (where D is large), the corresponding minimum effective diffusion flux J1 is ≈0.02-0.05 mol / (cm²⋅s). Under atmospheric pressure inert gas conditions, where D is small, the effective J1 will be even lower. In this embodiment, during the initial heating phase of sublimation, heating causes the molybdenum trioxide solid to sublime into vapor, which then migrates toward the condensation zone. Using a temperature sensor array, infrared thermal imaging system, gas-phase concentration sensor, and in-situ crystal plane detection, the temperature field, gas-phase concentration field, and crystal plane structure distribution within the sublimation chamber are systematically constructed. Combined with three-dimensional particle tracking velocimetry (3D-PTV), dynamic analysis of the vapor diffusion path and velocity field is achieved, enabling precise extraction of the concentration gradient in the boundary layer at the sublimation wall. This enables the system to promptly identify mass transfer bottleneck risks caused by vapor accumulation, resolving the uncontrollable "black box" nature of the boundary layer state in traditional sublimation processes. Using a concentration gradient threshold (G1) and a diffusion flux efficiency threshold (J1) as dual criteria for mass transfer efficiency evaluation, combined with Fick's first law for dynamic judgment, it effectively identifies non-ideal diffusion states characterized by high concentration gradients and insufficient diffusion flux. This mechanism avoids the misjudgment associated with single-parameter evaluation, making the assessment of mass transfer states more scientific and stable. When the system detects a mass transfer bottleneck in the boundary layer, the first micro-dynamic correction strategy is generated. By slightly adjusting the temperature gradient of the sublimation wall, including adjusting the temperature gradient by 0.5℃~2℃ each time and increasing the vibration amplitude of the boundary layer by 1µm~2µm each time, it effectively promotes the diffusion of gas molecules and the ventilation efficiency of the boundary layer without destroying the overall sublimation thermal field structure. After the implementation of the first micro-dynamic correction strategy, a complete closed-loop control logic is formed by re-collecting and judging the concentration gradient and diffusion flux, ensuring that each correction has a monitoring basis and stability confirmation, thereby effectively preventing process deviations caused by excessive correction or misjudgment. Compared with traditional large-scale temperature jumps or mechanical vibration methods, this strategy has the advantages of a small disturbance range, high control accuracy, and is not prone to causing new thermal unevenness or crystal segregation.

[0041] The diffusion coefficient D constructed in the present invention covers a variety of typical sublimation environments, including normal pressure, reduced pressure, inert gas protection, and vacuum. By setting different diffusion coefficient ranges D under different atmospheric conditions, the system's ability to adapt to different sublimation process scenarios is improved, and it has wide engineering applicability and versatility.

[0042] Example 6 Furthermore, temperature differences often drive complex capillary convection within the sublimation chamber. Especially when vapor concentrations are high and air viscosity fluctuates significantly, localized convection paths can drift or reverse, disrupting the stable atmosphere for crystal growth and reducing crystal quality. Traditional sublimation systems often rely on static thermal field control, making it difficult to sense and respond to dynamic changes in the convection structure in real time.

[0043] This embodiment is explained in Example 1. Specifically, the fourth step includes: S41. Using thermal imaging and airflow sensors, monitor the temperature and airflow changes in the sublimation chamber to obtain thermal field data and flow field data in the chamber; S42. Based on the obtained thermal field data and flow field data in the cavity and combined with the sublimation vapor concentration inversion information, a heat convection offset vector diagram in the sublimation cavity is constructed, and the flow field stability index F is calculated. TCV : Where N represents the total number of sampling points, represents the airflow velocity vector at the jth point, represents the average velocity vector, represents the temperature of the jth point, represents the average temperature, Represents the weight coefficient of temperature fluctuation; The settings range includes: In the initial coarse adjustment stage, α=1.0, taking into account both speed and temperature fluctuations; Fine-tuning monitoring stage: α = 1.5 to 2.0, enhancing the temperature disturbance detection capability; Local condensation and re-crystallization stage, α = 2.5; represents the dimensionless relative velocity fluctuation, represents the dimensionless relative temperature fluctuation; S43, preset stability threshold F, judge flow field stability index F TCV Does it exceed the set stability threshold F? If so, a local temperature zone fine-tuning instruction is triggered to suppress abnormal thermocapillary convection. The local temperature zone fine-tuning instruction includes: placing a cooling source on the back or side of the temperature zone to perform directional thermal adjustment of 0.2-0.8°C, and triggering the control guide vane to adjust the closing direction by ±5°-20° based on the local airflow direction disturbance in the temperature zone circle to interrupt the local circulation or deviation flow path and restore the symmetry and straightness of the airflow; If the flow field stability index F TCV If the stability threshold F is not exceeded, the current parameters are maintained, monitoring is continued, and the process proceeds to step 5.

[0044] When using tools such as Ansys Fluent and COMSOL to perform CFD numerical simulation on the sublimation cavity, the following rules can be statistically analyzed by running multiple sets of thermal-fluid coupling models: F TCV When the stability threshold F is less than or equal to the threshold value, the flow field is stable and symmetrical, and the sublimation vapor is evenly distributed; F TCVWhen the stability threshold F is greater than the threshold, obvious thermocapillary convection and boundary layer disturbances will occur in the cavity, which may easily induce condensation, deviation or re-crystallization. Under the conditions of typical sublimation cavity dimensions (such as 100mm diameter, 200mm height), center temperature of 750℃ and wall temperature of 700℃, the stability threshold F is ≈ 0.35~0.45; F TCV Exceeding the threshold F indicates that the local flow field in the cavity is unstable. To avoid steam accumulation, deviation, and re-crystallization, the local temperature zone fine-tuning instructions are triggered: thermal regulation (0.2~0.8℃): slowly interrupting the thermocapillary driven flow; guide vane adjustment (±5°~20°): reshaping the airflow direction, guiding its symmetrical backflow, and suppressing vortexes or short-circuit paths; cooling source control in the cooling zone: forming a temperature difference wall to block the convection path.

[0045] In this embodiment, it is the stage of thermocapillary convection and flow field stability: there is convection caused by temperature difference in the cavity, which affects the flow and distribution of steam. The flow field stability index F based on the coupled analysis of thermal field and flow field is TCV By introducing the temperature fluctuation weight control mechanism, the real-time discrimination of the thermal capillary convection and bias flow trend of the sublimation cavity is realized. TCV When the stability threshold F is exceeded, the fine-tuning command is automatically triggered to perform local thermal regulation and flow path correction, effectively suppressing the problems of steam deviation and condensation re-crystallization, and significantly improving the mass transfer efficiency and crystal morphology uniformity of the sublimation process.

[0046] Example 7 Finally, during the condensation deposition stage, the deposition behavior of sublimated vapor is significantly affected by the temperature distribution of the condensation surface and the control of crystal orientation. If the condensation conditions are asymmetric, it will lead to non-uniform grain size, unbalanced stress in the deposited layer, and even cause "reflow" phenomenon, which will destroy the homogeneity of crystallization. However, existing technologies lack effective characterization and feedback control methods for condensation symmetry.

[0047] This embodiment is explained in Example 1. Specifically, the step 5 includes: S51. Use a high-resolution infrared thermal imager and a morphology evolution map acquisition device to continuously scan the condensation surface to obtain the following parameters and construct a condensation surface deposition uniformity dataset, including: condensation surface temperature distribution standard deviation, condensation crystal thickness distribution, and condensation crystal deposition morphology image; S52. Based on the uniform deposition data set on the condensation surface, the sublimation-condensation asymmetry factor ASC is constructed to quantify the degree of imbalance between the deposition distribution and the thermal field distribution on the condensation surface. The expression is: Where, represents the standard deviation of crystal thickness, represents the average value of the crystal thickness, represents the standard deviation of the condensation surface temperature distribution, represents the average value of the temperature field during the condensation process, Indicates the symmetry deviation of the topographic image; in, The way to obtain is: For the condensed crystal deposition morphology image, mirror it left and right or up and down according to the central axis of the condensation surface, and use the mean square error (MSE) method to calculate the image difference and obtain the symmetry deviation of the morphology image. : Where M represents the total number of pixels in the condensed crystal deposition morphology image, represents the pixel value of the kth pixel in the condensation crystal deposition morphology image, It represents the pixel value corresponding to the kth pixel after being mirrored left and right or up and down along the central axis of the condensation surface; 、 and Expressed as weight, , , ,and ; If the two images are completely symmetrical and the pixel values ​​at the symmetrical positions are exactly the same, then all ,so , indicating perfect symmetry. If the difference in symmetrical pixel values ​​is large, then The larger the value, the more asymmetric the morphology.

[0048] S53, preset symmetric critical threshold X, when >X, it indicates that there is an asymmetric risk of condensation deposition, triggering the second warning instruction; when If ≤X, it indicates that the current condensation deposition process is qualified, and the crystal structure is uniform, with no signs of reflux, and the current batch of crystallization process is marked as qualified and continuously monitored; By analyzing more than 200 sets of crystal deposition process data, combined with measured parameters and failure analysis of crystal products, we found that: When ASC≤0.15, the grains are uniform, stable and highly symmetrical, and the product is excellent and can be passed in batches; When 0.15<ASC≤0.25, there is local grain inhomogeneity, slight stress, and moderate symmetry deviation; When ASC>0.25, the grains are obviously recondensed, the structure is distorted, and the symmetry is seriously unbalanced, which is judged as an unqualified batch; Therefore, the crystal structure instability and recondensation phenomena are mainly concentrated in the case of ASC>0.25; Considering that in actual industrial applications, an overly sensitive threshold (such as X=0.15) will lead to frequent false alarms and affect process efficiency, a compromise principle between stability and response sensitivity is adopted; Set X=0.20~0.25 as the risk control critical value interval.

[0049] S54. After receiving the second warning instruction, generate a second slight dynamic correction strategy, reduce the temperature in the asymmetric area by 0.5 to 2°C, promote the migration of deposition to the symmetric area, and guide the crystal to grow along a more stable symmetric direction by rotating or shifting the direction of the seed crystal. After each adjustment, recalculate the sublimation-condensation asymmetry factor ASC to determine whether it falls back to ≤X. If not, continue the second slight dynamic correction strategy until the deposition returns to a symmetric state.

[0050] In this embodiment, it is during the condensation and deposition stage: the vapor cools and deposits into a solid on the condensation surface, forming grains. The sublimation-condensation asymmetry factor ASC, as a comprehensive indicator to measure the symmetry of the heat-mass distribution on the condensation surface, can effectively identify symmetry problems caused by thermal field deviation and uneven crystal morphology during crystal growth. When the ASC exceeds the preset critical value X, the second early warning instruction is automatically triggered, and the condensation behavior is dynamically repaired through local cooling and seed crystal orientation fine-tuning strategies, thereby significantly improving the uniformity and structural stability of crystal deposition and reducing the occurrence of abnormal conditions such as "re-condensation reflux".

[0051] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technicians in this field for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0052] The above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The coefficients in the formula are set by those skilled in the art according to actual conditions. The above is only a preferred specific implementation method of the present invention, but the protection scope of the present invention is not limited to this. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, can make equivalent replacements or changes based on the technical solution and inventive concept of the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide, characterized in that: The following steps are involved: Step 1: Perform a particle size distribution test on the molybdenum trioxide raw material before sublimation, using a laser particle size analyzer, image particle size analysis or SEM image analysis method to obtain the particle size distribution spectrum of the raw material and the particle size distribution skewness value S and the particle size distribution kurtosis value K, and construct an initial particle size state factor Gsize; when Gsize ≥ 0.7 and K ≥ 3, it indicates that the current molybdenum trioxide raw material is pre-monitored and qualified, and is summarized as the first pre-qualified raw material, and proceeds to step 2; Step 2: Photograph the first pre-qualified raw material to obtain a raw material crystal image. After image preprocessing the raw material crystal image, construct a particle feature vector, output a category label through the trained classifier KNN, and calculate the percentage of particles with the kth category label to the total number of particles. Obtain the proportion of spherical particles, the proportion of flaky particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles. When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is considered qualified and proceed to step 3. Step 3: At the beginning of sublimation, the temperature field distribution, gas phase concentration distribution and sublimation wall crystal structure in the sublimation chamber are collected to construct a boundary layer structure model under the initial sublimation conditions; combined with three-dimensional particle tracking technology, the near-wall diffusion rate curve of molybdenum trioxide vapor is obtained to construct the concentration gradient of the sublimation wall boundary layer. and the corresponding steam diffusion flux J to determine whether there is steam accumulation in the boundary layer and a mass transfer bottleneck risk. If so, a first micro-dynamic correction strategy is generated and issued until there is no mass transfer bottleneck risk and then step 4 is entered; Step 4: Use thermal imaging matrix, flow field tracking and sublimation vapor concentration inversion technology to capture the thermal capillary convection path and intensity changes in the sublimation chamber, construct a thermal convection offset vector diagram, and calculate the flow field stability index F TCV ; Preset stable threshold F, when F TCV >F, the system executes the local temperature zone fine-tuning instruction and then enters step 5; Step 5. During the condensation deposition stage, the non-uniform characteristics of sublimation vapor deposition are identified by scanning the condensation surface temperature field and monitoring the morphology evolution map, and the sublimation-condensation asymmetry factor ASC is constructed. The symmetry critical threshold X is preset. When ASC>X, the condensation symmetry is corrected by dynamically adjusting the condensation surface temperature distribution and the crystal plane guidance direction until crystallization occurs.

2. A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide according to claim 1, characterized in that: Step one includes: S1. Conduct a particle size distribution test on the molybdenum trioxide raw material before sublimation, using a laser particle size analyzer, image particle size analysis, or SEM image analysis to obtain the particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, and construct an initial particle size state factor Gsize; The specific steps for obtaining the initial particle size state factor Gsize are: S11, particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, to obtain 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K; S121. The method for obtaining the particle size distribution skewness value S is: Where n represents the total number of granularity data points, Indicates the center value of the i-th particle size, the representative particle size of each particle size interval, including 1.2μm and 1.6μm; derived from the particle size analysis equipment; represents the average particle size, represents the standard deviation, Indicates particle size The corresponding volume fraction is the normalized value; S122. Evaluate the particle size distribution skewness value S to obtain a first evaluation result, including: When S = 0, it means that the distribution is completely symmetric with respect to the mean; When S>0, it means that the distribution is right-skewed relative to the mean; When S < 0, it means that the distribution is left-skewed relative to the mean.

3. A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide according to claim 2, characterized in that: Step 1 also includes: S123. The method for obtaining the particle size distribution kurtosis value K is: The formula means that the standardized value of the fourth-order central moment reflects the peak degree and tail thickness of the distribution; the kurtosis value K of the particle size distribution is the fourth power of the distance, and the "extreme values ​​far from the mean" in the distribution have a greater impact, which is used to determine whether the distribution is peaked or flat; S124. Evaluate the particle size distribution kurtosis value K to obtain a second evaluation result, including: When K = 3, it indicates a moderate normal distribution with moderate tail thickness and no central or dispersed bias; When K>3, it means that the distribution is close to the average value, the concentration is high, the tail is heavier, a peak distribution is formed, and the particle size consistency is qualified, which is conducive to controlling the sublimation uniformity; When K is less than 3, it means that the distribution is close to the mean value, the concentration is low, the tail is more dispersed, and a flat-top distribution is formed; the distribution is flat, and the mass transfer in the sublimation stage is limited and the fluctuation is large; S125. Combine the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S, and particle size distribution kurtosis value K obtained in S121-S123, and after normalization, obtain the initial particle size state factor Gsize through weighted combination: Where, Indicates the distribution width or particle size range of the particle size, reflecting the dispersion; 、 and Expressed as weight, , , ,and ; S126. Evaluate the initial particle size state factor Gsize. If both of the following judgment conditions are met, including: if Gsize is less than 0.7 and K is less than 3, indicating that the distribution is unqualified and will lead to abnormal risks in the subsequent sublimation stage, trigger the first warning instruction, including: screening the raw materials, adjusting the ratio, ultrasonic vibration or mechanical crushing to promote the deagglomeration of the raw material particles; When Gsize≥0.7 and K≥3, it means that the current molybdenum trioxide raw material is qualified in the pre-monitoring, and is summarized as the first pre-qualified raw material, and enters step 2.

4. A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide according to claim 3, characterized in that: Step 2 includes: S21, photographing the first pre-qualified raw material to obtain a raw material crystal image, and performing image preprocessing on the raw material crystal image, including: grayscale and binarization of the raw material crystal image, removing background noise and enhancing edge contrast, extracting single particle boundary contours, and completing particle segmentation; and closing each particle contour to eliminate non-target areas; The roundness feature, aspect ratio feature, boundary irregularity feature and area ratio feature of each particle contour are extracted to obtain the roundness of the particle. , aspect ratio , boundary irregularity And the ratio of the projected area to the convex hull area ; S2101, roundness of particles The calculation expression is: Among them, A is the projected area of ​​the particle, P is the boundary perimeter of the particle, and the closer the roundness value is to 1, the more regular and spherical the particle is; S2102, aspect ratio of particles The calculation expression is: Among them, L and W are the major axis length and minor axis length of the particle fitting circumscribed rectangle, respectively, reflecting the stretching degree of the particle; S2103, particle boundary irregularity The calculation expression is: S2104, the ratio of the particle's projected area to its convex hull area The calculation expression is: in, Indicates the convex hull area of ​​the particle, which is used to determine whether there is agglomeration or adhesion; S22, the roundness of the particles , aspect ratio , boundary irregularity And the ratio of the projected area to the convex hull area Compose the eigenvectors of the particles in order: The characteristic vector of each particle Input the trained classifier KNN and calculate the feature vector of the particle through the trained classifier KNN The Euclidean distance between the feature vectors of all training samples is used to find the first k neighbors with the closest distance, and count which category appears most among these k neighbors to obtain the corresponding category and output the category label; Category labels include: 001 spherical, 002 flake, 003 needle / columnar, and 004 agglomerated; Counting the total number of particles of the kth category label in the first pre-qualified raw material, and calculating the percentage of particles of the kth category label in the total number of particles, to obtain the proportion of spherical particles, the proportion of flake particles, the proportion of needle-shaped / columnar particles, and the proportion of agglomerated particles; When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is qualified and enters step 3; if the conditions are not met, it is unqualified, that is, "insufficient sphericity" or "too much agglomeration", indicating that it needs to be returned and re-screened.

5. A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide according to claim 4, characterized in that: Step three includes: S311. At the initial stage of sublimation, using a temperature sensor array and an infrared thermal imaging system to collect the temperature field distribution in the sublimation chamber; S312, using a gas phase concentration sensor or a Raman / infrared spectrometer to collect gas phase molybdenum trioxide concentration distribution in the sublimation chamber; S313, obtaining the sublimation wall crystal structure by in-situ surface diffraction or reflection electron diffraction; S314. Combined with 3D-PTV, the motion trajectory of molybdenum trioxide vapor particles during the sublimation process is tracked to extract the velocity field, diffusion path and time series of boundary layer vapor particles. S315, during the sublimation process of molybdenum trioxide, the raw material solid is directly converted into vapor at high temperature, and the thin layer area between the sublimation wall and the main air flow is marked as the boundary layer; S316, constructing the vapor diffusion rate curve of the boundary layer, and calculating the concentration gradient of the sublimation wall boundary layer based on the vapor diffusion rate curve of the boundary layer : Where C is the concentration of molybdenum trioxide vapor in the boundary layer, and x is the direction perpendicular to the sublimation wall; It represents the derivative of concentration with respect to position, reflecting "how fast the vapor concentration decreases from the wall to the outside"; The value representing the concentration gradient of the sublimation wall boundary layer.

6. A method for optimizing the sublimation purification parameters of high-purity molybdenum trioxide according to claim 5, characterized in that: Step three also includes: S317. According to Fick's first law, the vapor diffusion flux J is proportional to the concentration gradient of the sublimation wall boundary layer, and the expression is as follows: Where D represents the diffusion coefficient, including: Under atmospheric pressure air sublimation conditions, the sublimation temperature range of molybdenum trioxide is 700-850°C, corresponding to a temperature of 973–1123K. When the pressure is 1.0 atm, the diffusion coefficient D under these conditions is: 0.08-0.12 cm² / s; Under the conditions of sublimation in reduced pressure air, the sublimation temperature range of molybdenum trioxide is 600-750°C, corresponding to a temperature of 873–1023. When the pressure is 0.1atm, the diffusion coefficient D under this condition is: 0.60-0.90cm² / s; Under Ar or N2 inert gas protection sublimation conditions, the sublimation temperature range of molybdenum trioxide is 650-800°C, corresponding to a temperature of 923–1073, and a pressure of 1.0 atm. The diffusion coefficient D under these conditions is: 0.10-0.15 cm² / s; Under vacuum sublimation conditions, the sublimation temperature range of molybdenum trioxide is 550–700°C, corresponding to a temperature of 823–973K and a pressure of 10 -4 atm, the diffusion coefficient D under this condition is: 10–25 cm² / s, and increases with increasing temperature; The negative sign indicates that the substance moves from areas of high concentration to areas of low concentration.

7. A method for optimizing parameters of sublimation purification of high-purity molybdenum trioxide according to claim 6, characterized in that: Step three also includes: S318, preset critical threshold G1 and diffusion flux efficiency threshold J1, when the concentration gradient of the sublimation wall boundary layer When the critical threshold G1 is set and the steam diffusion flux J is less than the diffusion flux efficiency threshold J1, it means that steam accumulates in the boundary layer and there is a risk of mass transfer bottleneck. The first micro-dynamic correction strategy is generated, including: adjusting the temperature gradient by 0.5℃~2℃ each time and increasing the vibration amplitude of the boundary layer by 1µm~2µm each time; After the first micro-dynamic correction strategy is executed, the concentration gradient and vapor diffusion flux of the sublimation wall boundary layer are re-detected after stabilization until ≤G1, and the vapor diffusion flux J≥J1, then enter step 4; when When ≤G1 and the vapor diffusion flux J≥J1, it means that the mass transfer conditions are normal, the vapor diffusion flux efficiency meets the standard, and the sublimation state is continuously monitored.

8. A method for optimizing parameters of sublimation purification of high-purity molybdenum trioxide according to claim 7, characterized in that: The fourth step includes: S41. Using thermal imaging and airflow sensors, monitor the temperature and airflow changes in the sublimation chamber to obtain thermal field data and flow field data in the chamber; S42. Based on the obtained thermal field data and flow field data in the cavity and combined with the sublimation vapor concentration inversion information, a heat convection offset vector diagram in the sublimation cavity is constructed, and the flow field stability index F is calculated. TCV : Where N represents the total number of sampling points, represents the airflow velocity vector at the jth point, represents the average velocity vector, represents the temperature of the jth point, represents the average temperature, Represents the weight coefficient of temperature fluctuation; The settings range includes: represents the dimensionless relative velocity fluctuation, represents the dimensionless relative temperature fluctuation; S43, preset stability threshold F, judge flow field stability index F TCV Whether it exceeds the set stability threshold F. If so, a local temperature zone fine-tuning instruction is triggered to suppress abnormal thermal capillary convection. The local temperature zone fine-tuning instruction includes: placing a cold source on the back or side of the temperature zone to perform directional thermal adjustment of 0.2-0.8°C, and triggering the control of the guide vane to adjust the closing direction by ±5°-20° based on the local airflow direction disturbance in the temperature zone circle; If the flow field stability index F TCV If the stability threshold F is not exceeded, the current parameters are maintained, monitoring is continued, and the process proceeds to step 5.

9. A method for optimizing parameters of sublimation purification of high-purity molybdenum trioxide according to claim 8, characterized in that: The step five includes: S51. Use a high-resolution infrared thermal imager and a morphology evolution map acquisition device to continuously scan the condensation surface to obtain the following parameters and construct a condensation surface deposition uniformity dataset, including: condensation surface temperature distribution standard deviation, condensation crystal thickness distribution, and condensation crystal deposition morphology image; S52. Based on the uniform deposition data set on the condensation surface, the sublimation-condensation asymmetry factor ASC is constructed to quantify the degree of imbalance between the deposition distribution and the thermal field distribution on the condensation surface. The expression is: Where, represents the standard deviation of crystal thickness, represents the average value of the crystal thickness, represents the standard deviation of the condensation surface temperature distribution, represents the average value of the temperature field during the condensation process, Indicates the symmetry deviation of the topographic image; 、 and Expressed as weight, , , ,and ; in, The way to obtain is: For the condensed crystal deposition morphology image, mirror it left and right or up and down according to the central axis of the condensation surface, and use the mean square error (MSE) method to calculate the image difference and obtain the symmetry deviation of the morphology image. : Where M represents the total number of pixels in the condensed crystal deposition morphology image, represents the pixel value of the kth pixel in the condensation crystal deposition morphology image, It represents the pixel value corresponding to the kth pixel after being mirrored left and right or up and down along the central axis of the condensation surface; If the two images are completely symmetrical and the pixel values ​​at the symmetrical positions are exactly the same, then all ,so , indicating perfect symmetry. If the difference in symmetrical pixel values ​​is large, then The larger the value, the more asymmetric the morphology.

10. A method for optimizing parameters of sublimation purification of high-purity molybdenum trioxide according to claim 9, characterized in that: Step five also includes: S53, preset symmetric critical threshold X, when >X, it indicates that there is an asymmetric risk of condensation deposition, triggering the second warning instruction; when If ≤X, it indicates that the current condensation deposition process is qualified, and the crystal structure is uniform, with no signs of reflux, and the current batch of crystallization process is marked as qualified and continuously monitored; S54. After receiving the second warning instruction, generate a second slight dynamic correction strategy, reduce the temperature in the asymmetric area by 0.5 to 2°C, promote the migration of deposition to the symmetric area, and guide the crystal to grow along a more stable symmetric direction by rotating or shifting the direction of the seed crystal. After each adjustment, recalculate the sublimation-condensation asymmetry factor ASC to determine whether it falls back to ≤X. If not, continue the second slight dynamic correction strategy until the deposition returns to a symmetric state.

Images