A high-throughput optimization method and system for REBCO superconducting thin films

Abstract

This invention discloses a high-throughput optimization method and system for REBCO superconducting thin films, achieving efficient and continuous optimization of the types and proportions of multi-component rare earth elements. It overcomes the shortcomings of traditional trial-and-error methods that rely on human experience, significantly shortening the R&D cycle of high-performance superconducting tapes. The method includes: collecting publicly available literature and historical experimental data on REBCO superconducting thin films to establish an initial database; extracting the intrinsic physical characteristics of rare earth ions as input feature vectors for machine learning; using the initial database to train an algorithm and establish a mapping model between the input feature vectors and macroscopic superconducting performance indicators; performing global optimization based on the mapping model within preset parameter boundaries and outputting candidate parameter combinations; driving a high-throughput device to fabricate a thin film array and perform performance characterization; and providing closed-loop feedback of the actual characterization results to iteratively update the mapping model.

Description

Technical Field

[0001] This invention relates to the field of high-throughput optimization technology for superconducting thin films, specifically to a high-throughput optimization method and system for REBCO superconducting thin films. Background Technology

[0002] Rare earth barium copper oxide (REBa2Cu3O) 7-δ REBCO (Reinforced Biofilm Electrode) coated conductors are known for their excellent critical transition temperature (CBT). T c ) and high-field critical current density ( J c It has significant application value in fields such as strong electromagnetic fields and controlled nuclear fusion. To further enhance its flux pinning force under complex magnetic fields... F p Researchers typically employ strategies such as mixing different rare earth elements (e.g., Y, Gd, Sm) and introducing artificial pinned centers (APCs). Among the many methods for preparing REBCO superconducting thin films, fluorine-free metal-organic deposition (FF-MOD) has become a highly promising, efficient, and environmentally friendly preparation technology due to its advantages such as environmental friendliness and fast deposition rate.

[0003] In complex magnetic field environments, the superconducting properties of REBCO thin films are influenced by a variety of parameters exhibiting nonlinear coupling relationships, including the type and ratio of rare earth elements, the doping amount of APC, the phase formation temperature, and the oxygen partial pressure. Traditional research and development heavily relies on manual trial and error, conducting experiments through the control of a single variable. This approach not only involves lengthy development cycles and enormous material and time costs, but also easily gets trapped in local optima, failing to find the truly globally optimal process within the vast multidimensional parameter space.

[0004] In recent years, artificial intelligence technologies such as machine learning have been widely applied in the field of materials science. By constructing predictive models between material properties and preparation parameters, rapid optimization of material composition and process parameters can be achieved. However, current technologies lack methods for systematically applying machine learning combined with high-throughput experiments to the preparation of rare-earth-barium copper oxide superconducting tapes. In particular, there are currently no mature technical solutions for how to achieve continuous variation of multi-component rare-earth elements in a single preparation process, and how to use machine learning to guide the synergistic optimization of precursor solution formulation and low-temperature thermal decomposition process, thereby achieving high-throughput preparation and rapid screening of multi-component REBCO superconducting thin films. Summary of the Invention

[0005] To address the problems in the existing technology, this invention provides a high-throughput optimization method and system for REBCO superconducting thin films, which achieves efficient and continuous optimization of the types and proportions of multi-component rare earth elements, overcomes the shortcomings of traditional trial-and-error methods that rely on human experience, and significantly shortens the R&D cycle of high-performance superconducting tapes.

[0006] To achieve the above objectives, the present invention provides a high-throughput optimization method for REBCO superconducting thin films, comprising the following steps: Step S1: Collect publicly available literature and historical experimental data on REBCO superconducting thin films to establish an initial database; based on the environment of rare earth ions in the mixed REBCO system under a trivalent, eight-coordinate crystal field, calculate and extract the eight-coordinate average ionic radius, local size mismatch, and average ionic potential of rare earth ions as intrinsic physical features, and convert them into input feature vectors for machine learning. Step S2: Train the machine learning algorithm using the initial database to establish a mapping model between the input feature vector and the superconducting performance index of REBCO superconducting thin film; then use a global optimization algorithm to perform optimization calculations based on the mapping model within the boundaries of the preset composition and preset heat treatment process parameters, and output candidate parameter combinations. Step S3: Drive the high-throughput fabrication module to fabricate a hybrid REBCO superconducting thin film array with composition or process gradient in parallel according to the candidate parameter combination; characterize the hybrid REBCO superconducting thin film array to obtain actual superconducting performance indicators, and feed the actual superconducting performance indicators obtained by characterization back to step S2 as new samples to update the mapping model; repeat steps S2 and S3 until the actual superconducting performance indicators reach the preset convergence condition.

[0007] Furthermore, the eight-coordinate average ion radius in step S1 Local size mismatch The formula for calculating the average ionic potential Φ8 is as follows:

[0008]

[0009]

[0010] Where n is the total number of rare earth elements in the mixture; c i Let be the mole fraction of the i-th rare earth element in the total rare earth composition; Z represents the standard ionic radius of the i-th rare earth element when it is in the trivalent state and has a coordination number of 8; i Let Z be the valence of the i-th rare earth element and Z i =3.

[0011] Furthermore, the input feature vector in step S1 includes average electronegativity. Electronegativity mismatch And the octetal coordinate bond valence and V i The calculation formula is as follows:

[0012]

[0013]

[0014] in, The electronegativity of the i-th rare earth element; For a specific rare earth-oxygen bond, the empirical constant is used. denoted as , where is the actual bond length from the central rare earth ion to the j-th oxygen atom in its coordination polyhedron, and j takes values ​​from 1 to 8; B is an empirical constant.

[0015] Furthermore, the main phase chemical formula of the hybrid REBCO system is REBa2Cu3O. 7-δ The rare earth element RE is selected from at least two of Y, Gd, Sm, Nd and Eu, and the total molar ratio of the mixture is 1.

[0016] Furthermore, the hybrid REBCO system also includes artificial pinning centers selected from BaZrO3, BaSnO3, or BaHfO3, with a doping amount of 0.1 mol% to 2 mol% relative to the rare earth element RE.

[0017] Furthermore, the preset heat treatment process parameter boundaries in step S2 include: the optimal boundary for the phase formation crystallization temperature is 760 ℃ to 830 ℃, and the optimal boundary for the oxygen partial pressure in the phase formation atmosphere is 10 ppm to 1000 ppm; the superconducting performance indicators of the REBCO superconducting thin film include the critical transition temperature, critical current density, and magnetic flux pinning force; the global optimization algorithm adopts the Bayesian optimization algorithm.

[0018] Furthermore, in step S3, the high-throughput preparation module uses a fluorine-free metal-organic deposition combined with liquid phase component mixing method to prepare a hybrid REBCO superconducting thin film array.

[0019] Furthermore, the fluorine-free precursor solution used in the fluorine-free metal-organic deposition uses fluorine-free acetic acid or propionate as the metal source, and the total metal ion concentration is 0.1 mol / L to 2.5 mol / L.

[0020] Furthermore, the liquid phase component mixing is performed using a multi-channel inkjet printing system. After receiving the candidate parameter combination, the multi-channel inkjet printing system performs in-situ mixing and deposition in the substrate micro-region by adjusting the jetting frequency and droplet overlap rate of each nozzle to form a composition gradient array film layer. After heat treatment, a mixed REBCO superconducting thin film array with composition or process gradient is prepared.

[0021] Another aspect of the present invention provides a high-throughput optimization system for REBCO superconducting thin films, used to implement the above-described high-throughput optimization method for REBCO superconducting thin films, comprising: The data and feature extraction module is used to establish an initial database, calculate and extract the intrinsic physical features of rare earth ions in the mixed REBCO system under the trivalent and octagonal crystal field environment, and convert them into input feature vectors for machine learning. The prediction and optimization module is used to train the machine learning algorithm based on the initial database, establish a mapping model between the input feature vector and the superconducting performance index of REBCO superconducting thin film, and perform optimization calculations based on the mapping model within the boundaries of preset composition and preset heat treatment process parameters to output candidate parameter combinations. A high-throughput preparation module is used to receive the candidate parameter combinations and prepare a mixed REBCO superconducting thin film array with composition or process gradient in parallel using a fluorine-free precursor solution. The system also includes a characterization and feedback module, which characterizes the actual superconducting performance indicators of the hybrid REBCO superconducting thin film array and feeds the obtained actual superconducting performance indicators as new samples back to the prediction and optimization module to update the mapping model.

[0022] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention incorporates fundamental condensed matter physics characteristics such as the eight-coordinate average radius, local size mismatch, and average ion potential of rare earth ions as rigid constraints into machine learning feature engineering. This "white-box" strategy endows the prediction model with a clear physical core, enabling it to deeply understand the mapping relationship between lattice strain and superconductivity, achieving prediction accuracy far exceeding that of traditional black-box algorithms in a vast multidimensional parameter space.

[0023] 2. This invention constructs an initial database by extracting publicly available literature and historical experimental data for the basic training of the algorithm. This mechanism effectively introduces prior knowledge in the field of superconductivity, enabling high-throughput experiments to avoid inefficient blind spots in the first round of global optimization, rapidly approaching the optimal parameter combination with a very small number of physical experiment batches, and shortening the long traditional trial-and-error cycle to several weeks.

[0024] 3. This invention seamlessly integrates the entire process of "physical feature extraction - intelligent prediction and optimization - high-throughput synthesis - non-destructive characterization feedback." The actual characterization obtained... T c , J c Performance indicators are automatically fed back, driving the prediction model to perform adaptive iterations. This closed-loop correction mechanism eliminates the lag of human intervention, providing an efficient and automated path for the low-cost, customized development of novel second-generation high-temperature superconducting tapes. Attached Figure Description

[0025] Figure 1 This is a process thermodynamic correlation diagram of the binary YGdBCO mixture in Example 1; Figure 2a It is the binary mixture of YGdBCO in Example 1 J c Curve showing the change in Y content; Figure 2b It is the ternary mixture YGdSmBCO in Example 4 J c Distribution diagram of Y-Gd content variation; Figure 3 It is a high-flux REBCO superconducting thin film. J c Characterization results diagram. Detailed Implementation

[0026] The present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] This invention addresses the problems encountered in preparing rare-earth-barium-copper oxide superconducting thin films using fluorine-free metal acetates, improving the flux pinning capability of these films. Firstly, it provides a high-throughput optimization method for hybrid REBCO superconducting thin films based on intrinsic physical characteristics and machine learning. The method mainly includes: collecting publicly available literature and historical experimental data on REBCO superconducting thin films to establish an initial database; extracting the intrinsic physical characteristics of rare-earth ions as input feature vectors for machine learning; training an algorithm using the initial database to establish a mapping model between the input feature vectors and macroscopic superconducting performance indicators; performing global optimization based on the mapping model within preset parameter boundaries, outputting candidate parameter combinations; driving a high-throughput preparation module to prepare thin film arrays and characterize their performance; and providing closed-loop feedback of the actual characterization results to iteratively update the mapping model. This invention achieves efficient and continuous optimization of the types and proportions of multi-component rare-earth elements, overcoming the shortcomings of traditional trial-and-error methods that rely on human experience, and significantly shortening the development cycle of high-performance superconducting tapes.

[0028] This invention also provides a high-throughput optimization system for REBCO superconducting thin films to implement the above method, comprising: a data and feature extraction module for establishing an initial database, calculating and extracting the intrinsic physical features of rare earth ions in a trivalent octagonal crystal field environment and converting them into input feature vectors for machine learning; a prediction and optimization module for training and establishing a mapping model based on the initial database and performing optimization calculations within preset parameter boundaries to output candidate parameter combinations; and a high-throughput preparation module for receiving candidate parameter combinations and using a fluorine-free precursor solution to prepare mixed REBa2Cu3O with compositional or process gradients in parallel. 7-δ A superconducting thin film array; and a characterization and feedback module, used to obtain the actual superconducting performance indicators of the superconducting thin film array and feed them back to the prediction and optimization module to update the mapping model.

[0029] All or part of the steps in the method embodiment can be completed by program instructions and related hardware. The aforementioned program can be stored in a computer-readable storage medium. The storage medium can include the built-in storage medium in the smart terminal, or it can include the extended storage medium supported by the smart terminal. The smart terminal includes the aforementioned storage medium and system. The smart terminal can run program instructions that can be read and stored by the storage medium to realize the high-throughput optimization method for REBCO superconducting thin films as described above.

[0030] The present invention will now be described in detail with reference to specific embodiments. It should be noted that these embodiments are merely descriptive and do not limit the invention in any way. Unless otherwise specified, the raw materials used in the following embodiments and comparative examples are all common commercially available products that can be directly purchased in the art or prepared according to conventional methods existing in the art.

[0031] Example 1 Taking the high-throughput optimization process of Y-Gd binary hybrid REBCO superconducting thin films as an example, the specific implementation of the method disclosed in this invention is illustrated.

[0032] (1) Collect publicly available literature on YBa2Cu3O 7-δ GdBa2Cu3O 7-δ and Y x Gd （1-x） Ba2Cu3O 7-δ Data related to superconducting thin films, including rare earth element ratios, phase formation and crystallization temperatures, oxygen partial pressures, and other process parameters for each sample, as well as corresponding superconducting performance indicators such as critical transition temperature, critical current density, and magnetic flux pinning force, are entered into the system to establish an initial database.

[0033] (2) The intrinsic physical characteristics of rare earth ions under a trivalent, eight-coordinate crystal field in the mixed REBCO system are calculated and extracted. For Y x Gd（1-x） BCO system, total number of rare earth elements n=2, Y 3+ The radius of the eight-coordinate ion is Gd 3+ The radius of the eight-coordinate ion is Calculate according to the following formula: Eight-coordinate average ionic radius:

[0034] Local size adaptation:

[0035] Average ionic potential:

[0036] The average electronegativity (Y electronegativity 1.22, Gd electronegativity 1.20), electronegativity fitness, and octetal bond valence sum are further calculated. These features are then combined to form the input feature vector.

[0037] (3) Five-fold cross-validation was used to evaluate the model performance. The optimization boundaries for the components and process parameters were set as follows: Y mole fraction 0.1–0.9, phase formation and crystallization temperature 760–830℃, and oxygen partial pressure 10–1000 ppm. A Bayesian optimization algorithm was used, with the expected improvement (EI) as the acquisition function for optimization calculation. After 30 iterations, the candidate parameter combinations were output as follows: Y mole fraction 0.5–0.7, phase formation and crystallization temperature 790–800℃, and oxygen partial pressure 50 ppm.

[0038] (4) Using the fluorine-free organometallic salt preparation technology, propionate precursor solutions of rare earth, barium and copper were prepared respectively, with the cation concentration in the precursor solution being 1.5 mol / L.

[0039] (5) Gradient array fabrication was performed using a multi-channel inkjet printing system. The system was equipped with four independently controlled printheads, each loaded with a Y source, a Gd source, a Ba source, and a Cu source precursor solution, respectively. In-situ mixing deposition was performed on a 2-inch × 2-inch metal substrate, which was a metal substrate with a biaxially textured CeO2 buffer layer deposited on a nickel-based strip.

[0040] (6) The specific program control process is as follows: From the starting point to the ending point of the metal substrate, the spray frequency ratio of the Y source and the Gd source nozzles is adjusted in real time to form a composition gradient array film layer with a continuous Y mole fraction from 0.2 to 0.5 in the X direction of the substrate. The distance between the nozzle and the substrate is 100 μm, the spray frequency is 5 to 20 kHz, and the scanning speed is 80 mm / s. After deposition, the film layer is allowed to stand at room temperature for 5 minutes to level out, resulting in a wet gel film.

[0041] (7) The substrate coated with the wet gel film was placed in a tube furnace and subjected to low-temperature heat treatment in a humid oxygen atmosphere. The humid oxygen was formed by passing oxygen into a gas washing bottle containing deionized water at a flow rate of 0.8 L / min. The temperature was increased to 510 °C at a heating rate of 1 °C / min to fully pyrolyze the precursor and form a rare earth-barium copper oxide pyrolysis film. The pyrolysis film was subjected to high-temperature phase formation heat treatment in a nitrogen-oxygen mixed gas atmosphere with an oxygen content of 50 ppm, a total flow rate of 0.5 L / min, and a heating rate of 20 °C / min to 790 °C. Finally, the superconducting thin film array was obtained by heat treatment in a pure oxygen atmosphere at 500 °C for 180 min.

[0042] (8) A non-contact inductive critical current density testing system was used to scan and test different positions of the array to obtain the critical current density of each micro-region. J c The critical transition temperature was measured using a Power Property Measurement System (PPMS). T c The measured performance data of each micro-region are correlated with the corresponding components, process parameters, and intrinsic physical characteristics to form new samples, which are then fed back into the database.

[0043] in, Figure 1 This is a thermodynamic correlation diagram of the YGdBCO process in Example 1. The characteristic correlation thermogram of the YGdBCO dataset is plotted based on the Pearson correlation coefficient to characterize the linear correlation between different variables in the synthesized YGdBCO sample. The horizontal and vertical axes represent the variables involved in the analysis, including Y, Gd, Y_fraction, REBCO005_2theta, REBCO005_Peak Area, FWHM, c_axis, Texture Score, Jc77K0T, Deposition Temperature, and Oxygen Pressure. The value in each cell of the graph represents the correlation coefficient between the corresponding two variables, with the correlation coefficient ranging from -1 to 1; a correlation coefficient close to 1 indicates a strong positive correlation, close to -1 indicates a strong negative correlation, and close to 0 indicates a weak linear correlation. Figure 1It can be seen that Y is significantly positively correlated with Y_fraction, and Gd is significantly negatively correlated with Y_fraction, indicating that the Y / Gd composition ratio has a direct correlation with the rare earth element content; REBCO005_2theta is positively correlated with Y_fraction, while c_axis is negatively correlated with Y_fraction, indicating that changes in composition affect the thin film crystal structure parameters; REBCO005_Peak Area is positively correlated with Texture Score and Jc77K0, and FWHM is positively correlated with Texture Score and Jc77K0. J c The negative correlation between 77K0T and the critical current density indicates that the crystallinity quality and texture characteristics of the thin film are related to the critical current density. Therefore, this correlation thermogram can be used to identify the correlations between variables in the YGdBCO dataset, providing a basis for feature selection, structure-performance relationship analysis, process parameter optimization, and the construction of superconducting performance prediction models.

[0044] Figure 2a In Example 1, the Y content in the binary YGdBCO mixture varies with... J c The variation curves represent the composition optimization results of the YGdBCO binary system. The horizontal axis represents the mole fraction of Y in the YGdBCO material, and the vertical axis represents the critical current density under self-field conditions at 77 K. J c The unit is MA / cm². The solid line in the figure represents... J c The curves showing the relationship between the γ mole fraction and the Y mole fraction are shown, with the dashed line representing the optimal Y mole fraction position and the circular markers indicating the optimal composition points determined by this invention. Figure 2a It can be seen that as the mole fraction of Y increases, the YGdBCO material... J c It shows a trend of first increasing and then decreasing. When the mole fraction of Y is approximately 0.28, J c The peak value of approximately 5.5 MA / cm² was reached, indicating that superior superconducting current-carrying performance can be achieved in this composition. Further increasing the molar fraction of Y... J c The significant decrease indicates that excessively high Y content is detrimental to further improving the critical current density. Therefore, this figure shows that the Y / Gd composition ratio has a significant impact on the superconducting properties of YGdBCO materials, and a Y molar fraction of approximately 0.28 can be considered as a preferred composition parameter in the embodiments of this invention. This result can be used to guide the composition design and fabrication process optimization of YGdBCO superconducting materials, thereby improving the critical current density of the material under self-field conditions at 77 K.

[0045] Example 2 Taking the closed-loop iterative optimization process as an example, the specific implementation of the method disclosed in this invention is illustrated.

[0046] This embodiment is based on Embodiment 1. The newly added samples (measured data from 36 different component points) obtained in Embodiment 1 are added to the initial database to update the mapping model. Bayesian optimization is then performed again using the updated model to output the second round of candidate parameter combinations.

[0047] The specific program control process is as follows: The prediction and optimization module receives the updated database, retrains the model, and performs Bayesian optimization within the same parameter boundaries. After 20 iterations, it outputs the second round of candidate parameter combinations: Y molar fraction 0.28, phase formation and crystallization temperature 800 ℃, and oxygen partial pressure 75 ppm. After receiving this combination, the high-throughput preparation module performs the second round of gradient array preparation and characterization according to the inkjet printing and heat treatment process in Example 1.

[0048] Repeating the above preparation, characterization, and feedback process, after 5 iterations, the actual superconducting performance index reached the preset convergence condition. J c ≥5.5 MA / cm² 77 K self-field), the optimization process terminates. The final obtained Y 0.28 Gd 0.72 Ba2Cu3O 7-δ The critical current density of the superconducting thin film reached 5.8 MA / cm² under a self-field at 77 K, which is about 18% higher than the best sample in the initial database.

[0049] Example 3 Taking a mixed REBCO system containing artificial pinning centers as an example, this invention illustrates a specific implementation of its method. This embodiment, based on Example 1, further introduces artificial pinning centers BaZrO3 (BZO). The BZO precursor (in propionate form) is added as a fourth component to the precursor solution system, with a doping amount set at 0.5 mol% (relative to RE elements). During intrinsic physical feature extraction, the local distortion effect caused by the introduction of BZO is reflected by correcting bond valences and calculating the coordination environment.

[0050] The optimization boundaries were adjusted to: Y mole fraction 0.2–0.5, BZO doping amount 0.1–2.0 mol%, phase formation and crystallization temperature 780–820℃, and oxygen partial pressure 10–500 ppm. A Bayesian optimization algorithm was used for optimization, with the desired improvement (EI) function as the acquisition function.

[0051] The specific program control process is as follows: The multi-channel inkjet printing system is equipped with five independent printheads, each loaded with a Y source, a Gd source, a Ba source, a Cu source, and a BZO source precursor solution, respectively. From the starting point to the end point of the metal substrate, the ejection frequency ratio of each printhead is adjusted in real time to form a dual gradient distribution of Y molar fraction and BZO doping amount in the X direction of the substrate. The remaining preparation and characterization processes are the same as in Example 1.

[0052] After six rounds of closed-loop iterations, the optimal parameter combination was obtained: Y molar fraction 0.31, BZO doping amount 0.8 mol%, phase formation and crystallization temperature 812 ℃, and oxygen partial pressure 65 ppm. The prepared superconducting thin film achieved a critical current density of 1.2 MA / cm² under a magnetic field of 77 K and 1 T, which is about 35% higher than that of the undoped sample, indicating that the artificial pinning centers of BZO effectively enhance the flux pinning performance under high fields.

[0053] Example 4 Taking the high-throughput optimization process of Y-Gd-Sm ternary hybrid REBCO superconducting thin films as an example, the specific implementation of the method disclosed in this invention is illustrated.

[0054] (1) Collect publicly available literature on YBa2Cu3O 7-δ GdBa2Cu3O 7-δ and SmBa2Cu3O 7-δ Data related to superconducting thin films, including rare earth element ratios, phase formation and crystallization temperatures, oxygen partial pressures, and other process parameters for each sample, as well as corresponding superconducting performance indicators such as critical transition temperature, critical current density, and magnetic flux pinning force, are entered into the system and a ternary hybrid database is established based on binary hybridization.

[0055] (2) For the Y-Gd-Sm ternary mixed REBCO system, the total number of rare earth elements is n=3. 3+ Gd 3+ 、Sm 3+ The radii of the trivalent and octagonal ions are respectively , , The electronegativity of the three rare earth elements are as follows: , 20. .

[0056] (3) Let the mole fraction of Y be x, the mole fraction of Gd be y, and the mole fraction of Sm be z = 1 - xy, where x, y, and z are all greater than 0 and satisfy x + y + z = 1. Calculate the intrinsic physical characteristics according to the formula above: Eight-coordinate average ionic radius:

[0057] Local size mismatch:

[0058] Average ionic potential:

[0059] Average electronegativity

[0060] Electronegativity mismatch

[0061] The above six intrinsic physical properties are combined to form the input feature vector.

[0062] (4) A Bayesian optimization algorithm is used, with the input feature vector and critical current density as the basis. J c A mapping model was established for the target variable. Due to the limited number of ternary system samples in the initial database, transfer learning was employed: the model was pre-trained on binary system data and then fine-tuned using existing ternary system data to improve the model's generalization ability under small sample conditions. Five-fold cross-validation was used to evaluate model performance during training.

[0063] (5) The optimization boundary for components and process parameters is set as follows: Y mole fraction x: 0.1~0.6, Gd mole fraction y: 0.1~0.6, Sm mole fraction z: 0.1~0.6 (satisfying x+y+z=1), sintering temperature: 780~810 ℃, oxygen partial pressure: 50 ppm.

[0064] (6) A Bayesian optimization algorithm was adopted, with the desired improvement (EI) as the acquisition function, and optimization calculations were performed based on a mapping model within a preset boundary. To balance exploration and utilization, 20 initial exploration points were set, and 40 subsequent iterations were performed. After 60 iterations, the candidate parameter combination was output as follows: Y mole fraction 0.25, Gd mole fraction 0.35, Sm mole fraction 0.40, phase formation and crystallization temperature 810 ℃, and oxygen partial pressure 50 ppm.

[0065] (7) Using the fluorine-free organometallic salt preparation technology, rare earth, barium and copper propionates were prepared as precursor solutions, with a cation concentration of 1.5 mol / L for each precursor solution.

[0066] (8) To further investigate the influence of ternary component space on superconducting performance, this embodiment uses a multi-channel inkjet printing system to prepare a ternary component gradient array. The system is equipped with five independently controlled printheads, each loaded with a Y source, a Gd source, a Sm source, a Ba source, and a Cu source precursor solution. In-situ mixing deposition was performed on a 3-inch × 3-inch metal substrate, which was a metal substrate with a biaxially textured CeO2 buffer layer deposited on a nickel-based strip.

[0067] (9) The specific program control process is as follows: From the starting point to the ending point of the metal substrate, the jetting frequency ratio of the three nozzles (Y source, Gd source, and Sm source) is adjusted in real time by the program to form a ternary composition gradient array film layer with continuously changing components in the XY two-dimensional direction of the substrate. Specifically, the molar fraction of Y is set to linearly change from 0.15 to 0.35 along the X direction, the molar fraction of Gd is set to linearly change from 0.25 to 0.45 along the Y direction, and the molar fraction of Sm is automatically determined by x+y+z=1. The distance between the nozzle and the substrate is 100 μm, the jetting frequency of each nozzle is independently adjustable from 5 to 25 kHz, the scanning speed is 80 mm / s, and the droplet volume is controlled within the range of 5 to 30 p1. After deposition, a wet gel film is obtained.

[0068] (10) The substrate coated with the wet gel film was placed in a tube furnace and subjected to low-temperature heat treatment in a humid oxygen atmosphere. The humid oxygen atmosphere was formed by passing oxygen into a gas washing bottle containing deionized water at a flow rate of 0.8 L / min. The temperature was increased to 510 °C at a heating rate of 1 °C / min to fully pyrolyze the precursor and form a rare earth-barium copper oxide pyrolysis film. Finally, a high-temperature phase-forming heat treatment was performed in a nitrogen-oxygen mixed gas atmosphere. The oxygen content in the mixed gas was 50 ppm and the total flow rate was 0.5 L / min. The temperature was increased to 810 °C at a heating rate of 20 °C / min and held for 30 min to obtain a superconducting thin film array. The array contained microregions with approximately 100 different composition points and a film thickness between 200 and 300 nm.

[0069] (11) A non-contact inductive critical current density testing system was used to automatically scan and test different positions of the array to obtain the critical current density of each micro-region under a 77K self-field. J c The critical transition temperature was measured using a comprehensive physical property measurement system (PPMS) at typical component points. T c The magnetic flux pinning force under different magnetic fields was also studied. The microstructure and crystal orientation were analyzed using scanning electron microscopy and X-ray diffraction. The measured performance data of each microregion were correlated with the corresponding composition, process parameters, and intrinsic physical characteristics to form new samples, which were automatically fed back into the database.

[0070] (12) Add the newly obtained samples (a total of 96 valid data points) from the previous step to the initial database and update the mapping model. Use the updated model to perform Bayesian optimization again and output the second round of candidate parameter combinations. In this embodiment, the second round of optimization results converged to Y molar fraction 0.22, Gd molar fraction 0.38, Sm molar fraction 0.40, phase formation crystallization temperature 810 ℃, and oxygen partial pressure 75ppm.

[0071] (13) Repeat the above preparation, characterization and feedback process. After 4 rounds of closed-loop iteration, the actual superconducting performance index reaches the preset convergence condition ( J c ≥6.0 MA / cm² 77 K self-field), the optimization process terminates. The final obtained Y 0.22 Gd 0.38 Sm 0.40 Ba2Cu3O 7-δ The critical current density of the superconducting thin film reached 6.3 MA / cm² under a self-field at 77K, and 1.5 MA / cm² under a magnetic field of 77K and 1T, which are about 25% and 40% higher than the best binary system sample in the initial database, respectively.

[0072] in Figure 2b In Example 4, the ternary mixture YGdSmBCO J c The graph shows the distribution of Y-Gd content and the composition optimization contour plot of the Y-Gd-Sm ternary system. The horizontal axis represents the mole fraction of Y, and the vertical axis represents the mole fraction of Gd. The mole fraction of Sm satisfies Sm = 1 - Y - Gd. The shades of gray in the graph represent the critical current density under different composition conditions. J c The unit is MA / cm², and the darker the grayscale, the more grayscale it represents. J c The higher the value, the more the star-shaped marker indicates the optimal composition position determined by this invention. Figure 2b It can be seen that in the Y-Gd-Sm ternary system, J c The curve exhibits local peaks as the mole fractions of Y and Gd change. When the mole fraction of Y is approximately 0.22 and the mole fraction of Gd is approximately 0.38, J c It reaches a relatively high level, approximately 5.6 MA / cm²; calculated according to Sm = 1 - Y - Gd, the Sm mole fraction at this point is approximately 0.40. As the compositional ratio deviates from this region, J cThe concentration gradually decreases. Therefore, this figure shows that the relative contents of the three rare earth elements, Y, Gd, and Sm, have a significant impact on the critical current density of superconducting materials, and that an optimal composition can be obtained by adjusting the proportion of the ternary rare earth elements, thereby improving the superconducting current-carrying performance of the material under self-field conditions at 77 K. This result can provide a basis for the composition design, performance optimization, and selection of preparation parameters for REBCO superconducting materials.

[0073] Figure 3 For high-flux REBCO superconducting thin films J c The characterization results are shown in the figure, which displays the critical current density of the high-flux REBCO superconducting thin film in different test regions in the form of a two-dimensional matrix. J c Each rectangular cell in the figure corresponds to a test point, test area, or high-throughput component unit on the thin film sample. The value in the cell represents the measurement result at the corresponding area. J c Value, representing the shade of gray. J c The relative magnitude of the values. Figure 3 It can be seen that high-flux REBCO superconducting thin films exhibit variations in different regions. J c The significant differences indicate that the superconducting current-carrying properties of the thin film vary with the test location, local composition, or process conditions. Specifically, some regions exhibit higher [superconducting current-carrying capacity]. J c The value can reach up to approximately 5.44 MA / cm², indicating that this region has excellent superconducting current carrying capacity; some regions J c The value is relatively low, approximately 1.43 MA / cm², indicating relatively weak superconductivity in this region. Therefore, this figure can intuitively reflect the performance of high-flux REBCO superconducting films. J c Distribution characteristics, performance gradients, and local differences can be used to quickly screen high-performance samples. J c This study identifies key regions or preferred compositional regions and provides experimental basis for subsequent composition optimization, deposition process optimization, and performance improvement of REBCO superconducting thin films.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-throughput optimization method for REBCO superconducting thin films, characterized in that, Includes the following steps: Step S1: Collect publicly available literature and historical experimental data on REBCO superconducting thin films to establish an initial database; based on the environment of rare earth ions in the mixed REBCO system under a trivalent, eight-coordinate crystal field, calculate and extract the eight-coordinate average ionic radius, local size mismatch, and average ionic potential of rare earth ions as intrinsic physical features, and convert them into input feature vectors for machine learning. Step S2: Train the machine learning algorithm using the initial database to establish a mapping model between the input feature vector and the superconducting performance index of REBCO superconducting thin film; then use a global optimization algorithm to perform optimization calculations based on the mapping model within the boundaries of the preset composition and preset heat treatment process parameters, and output candidate parameter combinations. Step S3: Drive the high-throughput fabrication module to fabricate hybrid REBCO superconducting thin film arrays with composition or process gradients in parallel according to the candidate parameter combination; The hybrid REBCO superconducting thin film array is characterized to obtain actual superconducting performance indicators, and the actual superconducting performance indicators obtained by characterization are fed back as new samples to step S2 to update the mapping model; Repeat steps S2 and S3 until the actual superconducting performance index reaches the preset convergence condition.

2. The high-throughput optimization method for REBCO superconducting thin films according to claim 1, characterized in that, The eight-coordinate average ion radius in step S1 Local size mismatch The formula for calculating the average ionic potential Φ8 is as follows: Where n is the total number of rare earth elements in the mixture; c i denoted as the mole fraction of the i-th rare earth element in the total rare earth composition; Z represents the standard ionic radius of the i-th rare earth element when it is in the trivalent state and has a coordination number of 8; i Let Z be the valence of the i-th rare earth element and Z i =3.

3. A high-throughput optimization method for REBCO superconducting thin films according to claim 1 or 2, characterized in that, The input feature vector in step S1 includes average electronegativity. Electronegativity mismatch And the octetal coordinate bond valence and V i The calculation formula is as follows: in, The electronegativity of the i-th rare earth element; For a specific rare earth-oxygen bond, the empirical constant is used. denoted as , where is the actual bond length from the central rare earth ion to the j-th oxygen atom in its coordination polyhedron, and j ranges from 1 to 8; B is an empirical constant.

4. The high-throughput optimization method for REBCO superconducting thin films according to claim 1, characterized in that, The main phase chemical formula of the mixed REBCO system is REBa2Cu3O. 7-δ The rare earth element RE is selected from at least two of Y, Gd, Sm, Nd and Eu, and the total molar ratio of the mixture is 1.

5. A high-throughput optimization method for REBCO superconducting thin films according to claim 1 or 4, characterized in that, The hybrid REBCO system also includes artificial pinning centers selected from BaZrO3, BaSnO3 or BaHfO3, with a doping amount of 0.1 mol% to 2 mol% relative to the rare earth element RE.

6. The high-throughput optimization method for REBCO superconducting thin films according to claim 1, characterized in that, The preset heat treatment process parameter boundaries in step S2 include: the optimal boundary for the phase formation crystallization temperature is 760 ℃ to 830 ℃, and the optimal boundary for the oxygen partial pressure in the phase formation atmosphere is 10 ppm to 1000 ppm; the superconducting performance indicators of the REBCO superconducting thin film include the critical transition temperature, critical current density, and magnetic flux pinning force; the global optimization algorithm adopts the Bayesian optimization algorithm.

7. The high-throughput optimization method for REBCO superconducting thin films according to claim 1, characterized in that, In step S3, the high-throughput preparation module uses a fluorine-free metal-organic deposition combined with liquid phase component mixing method to prepare a hybrid REBCO superconducting thin film array.

8. The high-throughput optimization method for REBCO superconducting thin films according to claim 7, characterized in that, The fluorine-free metal-organic deposition uses a fluorine-free precursor solution with fluoroacetic acid or propionate as the metal source, and the total metal ion concentration is 0.1 mol / L to 2.5 mol / L.

9. A high-throughput optimization method for REBCO superconducting thin films according to claim 7 or 8, characterized in that, The liquid phase component mixing is performed using a multi-channel inkjet printing system. After receiving the candidate parameter combination, the multi-channel inkjet printing system adjusts the jetting frequency and droplet overlap rate of each nozzle to perform in-situ mixing and deposition in the substrate micro-region, forming a composition gradient array film layer. After heat treatment, a mixed REBCO superconducting thin film array with composition or process gradient is prepared.

10. A high-throughput optimization system for REBCO superconducting thin films, characterized in that, A high-throughput optimization method for implementing a REBCO superconducting thin film as described in any one of claims 1 to 9 includes: The data and feature extraction module is used to establish an initial database, calculate and extract the intrinsic physical features of rare earth ions in the mixed REBCO system under the trivalent and octagonal crystal field environment, and convert them into input feature vectors for machine learning. The prediction and optimization module is used to train the machine learning algorithm based on the initial database, establish a mapping model between the input feature vector and the superconducting performance index of REBCO superconducting thin film, and perform optimization calculations based on the mapping model within the boundaries of preset composition and preset heat treatment process parameters to output candidate parameter combinations. A high-throughput preparation module is used to receive the candidate parameter combinations and prepare a mixed REBCO superconducting thin film array with composition or process gradient in parallel using a fluorine-free precursor solution. The system also includes a characterization and feedback module, which characterizes the actual superconducting performance indicators of the hybrid REBCO superconducting thin film array and feeds the obtained actual superconducting performance indicators as new samples back to the prediction and optimization module to update the mapping model.

Images