Intelligent precise preparation of target perovskite nanocrystals micro-reaction system and method

Abstract

The application discloses a kind of intelligent precise preparation target perovskite nanocrystal microreaction system and method, belong to micro chemical technology field.The system and method used in the application can realize the quick screening of acid-base ratio, residence time, precursor liquid and anti-solvent ratio by combining microreactor with automatic control, while effectively preventing the generation of impurities in product;And on the basis of the above, the meta-learning algorithm is combined with the hysteresis algorithm, accurate reaction parameters and fluorescence property model can be obtained under a small amount of data points, the synthesis of target emission perovskite nanocrystal is quickly and accurately realized in wide spectrum, and the trial-and-error cost is greatly reduced.In addition, the method used in the application is carried out at room temperature, which further reduces the energy consumption.The system and method described in the application realize the synthesis of target perovskite nanocrystal accurately and efficiently, with low energy consumption, small cost, simple and controllable operation, good product quality, which is beneficial to large-scale production.

Description

Technical Field

[0001] This invention belongs to the field of microchemical technology, specifically relating to an intelligent and precise microreaction system and method for preparing target perovskite nanocrystals. Background Technology

[0002] Perovskite nanocrystals have significant application value in solar cells, LED devices, and photocatalysis due to their unique photoelectric properties. Currently, a series of batch synthesis methods have been developed for perovskite nanocrystal synthesis, such as hot injection, anion exchange, and ligand-assisted reprecipitation (LARP). Hot injection is typically performed under high temperature and inert conditions, which increases experimental difficulty and energy costs. Anion exchange is often used to prepare mixed halide perovskite nanocrystals, requiring the preparation of starting monohalide nanocrystals, which increases experimental complexity. In contrast, LARP involves direct mixing of the precursor and antisolvent at room temperature, making it a more efficient, energy-saving, and potentially large-scale production method. Subtle differences in batch synthesis can significantly affect the size and shape of the product, thus altering the optical properties of the nanocrystals. Furthermore, experimental parameters such as supersaturation and pH ratio greatly influence the nucleation and growth process of nanocrystals; fully utilizing these important experimental parameters can further enable the controllable preparation of nanocrystals (Progress in Materials Science, 2022, 123:100821). However, conventional batch synthesis methods are inefficient, require significant time, raw materials, and manpower for trial and error, and have limitations in optimizing product performance and exploring the synergistic effects of synthesis parameters.

[0003] Microreaction technology can precisely control synthetic reagents and enhance the heat and mass transfer efficiency of the reaction process. It can also be combined with in-situ characterization tools and intelligent algorithm-guided reaction parameter prediction and optimization to achieve rapid screening of reaction parameters and establish accurate models of process parameters and product performance. Over the years, it has shown great potential in the synthesis of various materials as an efficient and reliable technology. At present, most methods for achieving target perovskite nanocrystals based on microreaction technology are based on two-phase flow, which is not conducive to the large-scale production of products. At the same time, high temperature conditions are required in the preparation of precursor liquid or reaction process, which increases experimental energy consumption (Small, 2022, 18(19): 2200740). The combination of the microreaction system and the adopted automatic control and algorithm is too complex, which has high requirements for software, hardware and experimental personnel. It is very difficult to carry out material discovery research in conventional laboratories (Advanced Materials, 2020, 32(30): 2001626). In addition, the algorithm used requires multiple samplings to achieve accurate positioning of the target product, which requires a certain amount of time and raw material costs. Such reaction platforms are not conducive to achieving large-scale production.

[0004] Patent CN201811163412.1 discloses an automated device and method for preparing perovskite quantum dot powder. By combining a reaction module, a temperature control module, and a filtration module, it can achieve precise control of process parameters and continuous production. Although each module described in this patent involves automatic control, the actual degree of automation is low. Optimizing product performance still requires significant human intervention, time, and reagent costs. Patent CN201711024415.2 discloses a method for rapid synthesis of full-spectrum perovskite quantum dots via organic-inorganic assisted ion exchange. This method uses CsPbBr3 mother liquor and inorganic salt solution for anion exchange reaction in a microreactor to obtain multicolored perovskite quantum dots. This method requires the prior preparation of CsPbBr3 product before subsequent experiments. Therefore, the quality of the mother liquor determines the quality of the subsequent product, introducing considerable uncertainty and increasing experimental complexity. Furthermore, the process of obtaining multicolored quantum dots requires repeated trial and error, significantly increasing experimental time and raw material costs, and it does not achieve nanoscale, highly precise wavelength positioning.

[0005] In recent years, meta-learning algorithms, as an important part of deep learning, have been proven to solve regression tasks and obtain accurate relational models based on a few datasets. Unlike conventional machine learning algorithms, meta-learning algorithms can autonomously learn important parts of neural networks, such as network parameter update methods, highly generalizable initialization of network parameters, and neural network architecture (Proceedings of the 34th International Conference on Machine Learning, 2017, 70: 1126-1135). Meta-learning algorithms mainly learn network components by acquiring prior knowledge from multiple pre-training tasks independent of real-world data, and then propose a generalizable model that can quickly obtain relational models in real-world tasks with very little training data. For example, in the task of fitting a sine function, a model-independent meta-learning algorithm only needs to sample five data points within a quarter of the target sine function's period to infer the overall shape of the target sine function in the remaining range.

[0006] Therefore, establishing a low-consumption, energy-saving, fast, and accurate micro-reaction system and method for achieving high-quality target perovskite nanocrystals based on microreactor systems and meta-learning algorithms can further accelerate the development of target perovskite nanocrystals. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to propose an intelligent and precise microreaction system and method for preparing target perovskite nanocrystals.

[0008] The present invention provides an intelligent and precise microreactor system for preparing target perovskite nanocrystals. The system includes an automatically controlled feed pump, an automatically controlled multi-channel switching valve, a micromixer, a microreactor, a monitoring device, and a capillary.

[0009] The automatic control feed pump includes a first automatic control feed pump, a second automatic control feed pump, a third automatic control feed pump, and a fourth automatic control feed pump. A Y-type connector is provided between the first automatic control feed pump, the second automatic control feed pump, and the automatic control multi-channel switching valve. The outlet of the first automatic control feed pump is connected to the first inlet of the Y-type connector via a capillary tube. The outlet of the second automatic control feed pump is connected to the second inlet of the Y-type connector via a capillary tube. The outlet of the Y-type connector and the third automatic control feed pump are respectively connected to the inlets of the automatic control multi-channel switching valve via capillary tubes. The outlet of the automatic control multi-channel switching valve is connected to a micro-mixer via a capillary tube. A T-type connector is provided between the microreactor, the micro-mixer, and the fourth automatic control feed pump. The outlet of the micro-mixer is connected to the first inlet of the T-type connector via a capillary tube. The fourth automatic control feed pump is connected to the second inlet of the T-type connector via a capillary tube. The outlet of the T-type connector is connected to the microreactor via a capillary tube.

[0010] The monitoring device includes an optical detection device and a computer; the optical detection device detects the optical properties of the products discharged from the microreactor and transmits the data to the computer for processing. Then, the computer sends instructions to the automatic control feed pump and the automatic control multi-channel switching valve based on the data processing results to control the reaction process.

[0011] On the other hand, the present invention provides an intelligent and precise method for preparing target perovskite nanocrystals based on the aforementioned microreaction system, comprising the following steps:

[0012] 1) The first automatic control feed pump introduces a Br-containing precursor solution, the second automatic control feed pump introduces an I-containing precursor solution, wherein the Br-containing precursor solution is a precursor solution containing CsBr and PbBr2, and the I-containing precursor solution is a precursor solution containing CsI and PbI2; the third automatic control feed pump introduces a solvent, and the fourth automatic control feed pump introduces an anti-solvent.

[0013] 2) Prepare the desired target CsPb(Br) x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is used as input and fed into the relationship model between the operating parameters of the microreaction system and the wavelength position obtained after training by the meta-learning algorithm. After the computer reads the fluorescence emission wavelength position data, it obtains the flow rates of the first, second, and fourth automatic control feed pumps according to the relationship model between the operating parameters of the microreaction system and the wavelength position, and performs target CsPb(Br) processing. x / I1-x Preparation of 3 nanocrystals; wherein, the Br-containing precursor liquid and the I-containing precursor liquid are mixed in a micro mixer, and then contacted with the antisolvent at the T-joint and enter the microreactor for reaction. The product after reaction is obtained by optical detection device to obtain wavelength position data.

[0014] 3) If the wavelength position data obtained from the experiment does not match the target CsPb(Br) x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is within ±4 nm. The computer uses a hysteresis algorithm to adjust the flow rate ratio of the first and second automatic control feed pumps until the wavelength position data of the reaction product meets the target CsPb(Br) emission wavelength. x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is within the range of ±4 nm.

[0015] 4) When the target CsPb(Br) x / I 1-x After the 3-nano crystals are prepared, the multi-channel switching valve is automatically controlled by computer commands to open the channels of the third and fourth automatic feed pumps, so that the solvent and antisolvent are introduced into the micro-reaction system to flush the fluid channels of the micro-mixer, microreactor and optical detection device.

[0016] The hysteresis algorithm is as follows: If the experimentally obtained wavelength position data is greater than the target wavelength position data + 4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate plus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate minus the minimum adjustment flow rate. If the experimentally obtained wavelength position data is less than the target wavelength position data - 4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate plus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate minus the minimum adjustment flow rate. If the experimentally obtained wavelength position data is greater than the target wavelength position data - 4nm and less than the target wavelength position data + 4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate minus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate plus the minimum adjustment flow rate.

[0017] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0018] (1) Compared with the high-temperature method for synthesizing perovskite nanocrystals, the system and method described in this invention only need to be carried out at room temperature and do not require inert gas protection, which greatly reduces experimental energy consumption, lowers economic costs, and is conducive to large-scale production.

[0019] (2) Compared with the anion exchange method for synthesizing perovskite nanocrystals, the system and method described in this invention do not involve a multi-step synthesis process, which reduces the experimental complexity and is more concise and efficient.

[0020] (3) Compared with other micro-reaction technologies, the present invention uses a spiral inverter microreactor to improve the efficient mixing effect, greatly reduces product adhesion and agglomeration in the microreactor, and greatly improves product quality.

[0021] (4) Compared with other intelligent platforms, the system and method described in this invention are more flexible, less complex, lightweight and intelligent, and more suitable for the optimization and development of target products.

[0022] (5) The system and method described in this invention are all monitored in real time by a self-developed visualization program, which has strong process controllability and high degree of automation, greatly reducing manpower, time and raw material costs. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the structure of the system of the present invention.

[0024] Figure 2 This is a flowchart illustrating the method of the present invention.

[0025] Figure 3 The fluorescence emission spectrum of CsPbBr3 nanocrystals prepared in Example 1.

[0026] Figure 4 The ultraviolet absorption spectrum of CsPbBr3 nanocrystals prepared in Example 1.

[0027] Figure 5 The fluorescence emission spectrum of CsPbBr3 nanocrystals prepared in Example 2.

[0028] Figure 6 The ultraviolet absorption spectrum of CsPbBr3 nanocrystals prepared in Example 2.

[0029] Figure 7 The fluorescence emission spectrum of CsPbBr3 nanocrystals prepared in Example 3.

[0030] Figure 8 The ultraviolet absorption spectrum of CsPbBr3 nanocrystals prepared in Example 3.

[0031] Figure 9 CsPb(Br) prepared in Example 4 x / I 1-x 3 nanocrystal fluorescence emission spectrum.

[0032] Figure 10 CsPb(Br) prepared in Example 5 x / I 1-x 3 nanocrystal fluorescence emission spectrum.

[0033] Figure 11 This is the meta-learning model obtained in Example 6.

[0034] Figure 12 CsPb(Br) prepared in Example 6 x / I 1-x 3 nanocrystal fluorescence emission spectrum.

[0035] Figure 13 Example 6 shows the deviation between the predictions and actual results of the meta-learning model.

[0036] Figure 14 Transmission electron microscopy image of CsPbBr3 nanocrystals prepared in Example 3.

[0037] Figure 15 CsPb(Br) prepared in Example 4 x / I 1-x Transmission electron microscopy image of 3 nanocrystals.

[0038] Figure 16 This is the visual program interface of the present invention.

[0039] The above Figure 1 In the middle: 1-Precursor liquid at the first automatic control feed pump, 2-Precursor liquid at the second automatic control feed pump, 3-Solvent, 4-First automatic control feed pump, 5-Second automatic control feed pump, 6-Third automatic control feed pump, 7-Automatic control multi-channel switching valve, 8-Micro mixer, 9-Micro reactor, 10-Optical detection bracket, 11-Laser, 12-Fourth automatic control feed pump, 13-Spectrometer, 14-Fiber optic cable, 15-Antisolvent, 16-Product collection bottle. Detailed Implementation

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Figure 1 This is a schematic diagram of the microreactor system used in this invention. An intelligent and precise microreactor system for preparing target perovskite nanocrystals includes an automatically controlled feed pump, an automatically controlled multi-channel switching valve, a micromixer, a microreactor, a monitoring device, and a capillary.

[0042] The automatic control multi-channel switching valve is used to switch between the precursor liquid channel and the solvent channel; the monitoring device includes an optical detection device and a computer.

[0043] The aforementioned microreactor system includes a first automatic control feed pump, a second automatic control feed pump, a third automatic control feed pump, and a fourth automatic control feed pump. A Y-type connector is provided between the first automatic control feed pump, the second automatic control feed pump, and the automatic control multi-channel switching valve. The outlet of the first automatic control feed pump is connected to the first inlet of the Y-type connector through a capillary tube, and the outlet of the second automatic control feed pump is connected to the second inlet of the Y-type connector through a capillary tube. The outlet of the Y-type connector and the third automatic control feed pump are respectively connected to the inlets of the automatic control multi-channel switching valve through capillary tubes. The automatic control multi-channel switching valve is connected to a micro-mixer through a capillary tube. A T-type connector is provided between the microreactor, the micro-mixer, and the fourth automatic control feed pump. The outlet of the micro-mixer is connected to the first inlet of the T-type connector through a capillary tube, and the fourth automatic control feed pump is connected to the second inlet of the T-type connector through a capillary tube. The outlet of the T-type connector is connected to the microreactor through a capillary tube. The product discharged from the microreactor is tested for optical properties by an optical detection device, and then the data is transmitted to a computer for processing. The computer sends instructions to the automatic control feed pump and the automatic control multi-channel switching valve based on the data processing results to control the reaction process.

[0044] The capillary materials mentioned above are polytetrafluoroethylene (PTFE) and soluble polytetrafluoroethylene (PFA).

[0045] The aforementioned micromixer includes a helical tube and its 3D support structure. The helical height of the helical tube is controlled at 2-5 mm, the helical diameter at 7-16 mm, and the helical length at 100-150 cm. The micromixer is used to mix the precursor liquid introduced by a first automatically controlled feed pump and a second automatically controlled feed pump; the first automatically controlled feed pump introduces a precursor liquid containing Br; the second automatically controlled feed pump introduces a precursor liquid containing I.

[0046] A T-connector is provided between the micro mixer and the antisolvent automatic control feed pump. After the precursor liquid and antisolvent come into contact at the T-connector, they flow into the microreactor for reaction.

[0047] The aforementioned microreactor includes a fluid reversing helical tube and its 3D support structure. The helix height of the fluid reversing helical tube is controlled at 3-5 mm, the helix diameter is controlled at 8-14 mm, and the helix length is controlled at 300-500 cm. The microreactor achieves fluid reversal by winding around a spiral grooved 3D support structure.

[0048] The aforementioned optical detection device consists of a detection bracket, optical fiber, spectrometer, and laser. The detection bracket includes a detection tube and its 3D support structure. The detection tube can be made of transparent polytetrafluoroethylene or quartz. The inner diameter of the detection tube is controlled at 1-5 mm, the outer diameter at 2-7 mm, and the length at 10 cm.

[0049] A method for intelligent and precise preparation of target perovskite nanocrystals includes the following steps:

[0050] 1) The first automatic control feed pump introduces a Br-containing precursor solution, the second automatic control feed pump introduces an I-containing precursor solution, wherein the Br-containing precursor solution is a precursor solution containing CsBr and PbBr2, and the I-containing precursor solution is a precursor solution containing CsI and PbI2; the third automatic control feed pump introduces a solvent, and the fourth automatic control feed pump introduces an anti-solvent.

[0051] 2) Prepare the desired target CsPb(Br) x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is used as input and fed into the relationship model between the operating parameters of the microreaction system and the wavelength position obtained after training by the meta-learning algorithm. After the computer reads the fluorescence emission wavelength position data, it obtains the flow rates of the first, second, and fourth automatic control feed pumps according to the relationship model between the operating parameters of the microreaction system and the wavelength position, and performs target CsPb(Br) processing. x / I 1-x Preparation of 3 nanocrystals; wherein, the Br-containing precursor liquid and the I-containing precursor liquid are mixed in a micro mixer, and then contacted with the antisolvent at the T-joint and enter the microreactor for reaction. The product after reaction is obtained by optical detection device to obtain wavelength position data.

[0052] 3) If the wavelength position data obtained from the experiment does not match the target CsPb(Br) x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is within ±4 nm. The computer uses a hysteresis algorithm to adjust the flow rate ratio of the first and second automatic control feed pumps until the wavelength position data of the reaction product meets the target CsPb(Br) emission wavelength. x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is within the range of ±4 nm.

[0053] 4) Target CsPb(Br) x / I 1-x After the 3-nano crystals are prepared, the multi-channel switching valve is automatically controlled by computer commands to open the channels of the third and fourth automatic feed pumps, so that the solvent and antisolvent are introduced into the micro-reaction system to flush the fluid channels of the micro-mixer, microreactor and optical detection device.

[0054] In step 2), the training process of the micro-reaction system operating parameters and wavelength position relationship model obtained after training by the meta-learning algorithm includes the following steps:

[0055] a) The initialized meta-learning algorithm neural network model f θ Pre-training is performed on a meta-learning algorithm neural network model that is a single-input, single-output multilayer perceptron with two hidden layers (40 nodes). The support and query sets for pre-training are sampled using a sine function, in the following form:

[0056] y (j) =asin(x) (j) +b) (1)

[0057] The amplitude *a* varies within the range [0.1, 7.0], and the phase *b* varies within the range [0, 2π]. Input *x* (j) and output y (j) The dimension is one. For pre-training, the training task set contains 160,000 sinusoidal tasks T. For each sinusoidal task T... i 14 pairs (x) are uniformly sampled from [-7,7]. (j) ,y (j) The support set is composed of the mean squared error. The mean squared error is used as the loss function and takes the following form:

[0058]

[0059] Based on the i-th sine task T i The support is concentrated in 14 pairs (x (j) ,y (j) The mean squared error loss is calculated using a one-step gradient descent method to convert the initialized neural network model parameters θ into θmin. i (The original model parameters θ are not updated here; only the new model parameters θ are calculated.) i Here, stochastic gradient descent with a learning rate of α = 0.01 is used as the optimization algorithm. The computation graph is preserved in this step. The parameters are calculated according to the following form:

[0060]

[0061] in, loss function The gradient.

[0062] In addition, 14 new pairs of data were extracted to form the sine task T. i The query set. The new mean squared error loss is based on θ. i The model was calculated using 14 new data points. The Adam optimization algorithm was used to update the model's original parameters θ, with a learning rate β of 0.001. The update followed this form:

[0063]

[0064] Based on the updated model parameters θ, the pre-trained model is obtained;

[0065] b) In this embodiment, the target CsPb(Br) is selected. x / I 1-x The operating parameters of three microreaction systems with x values ​​of 1, 0.67, and 0.5 in 3 nm crystals and the conditions under which CsPb(Br) was prepared were described. x / I 1-x The peak wavelength data (520nm, 567nm, and 600nm) of 3nm crystals were used as input and output to form the support set for the experimental task. Relevant parameters were obtained through pre-experiments. Stochastic gradient descent with a learning rate of 0.005 was used to fine-tune the pre-trained model obtained from the meta-learning algorithm, ultimately yielding a model showing the relationship between the operating parameters of the microreaction system and the wavelength positions. Eight wavelength positions (535nm, 546nm, 557nm, 578nm, 591nm, 609nm, 619nm, and 630nm) were selected as the query set for the experimental task to evaluate the model's accuracy. Five of these wavelength positions were within the support set range, and three were outside the support set range.

[0066] When evaluating the model, the target wavelength location data was set ±4nm as the upper and lower limits for the convergence of the evaluation experiment. When the experimentally obtained wavelength was outside this range, the hysteresis algorithm started adjusting the flow rate of the precursor liquid feed pump based on the minimum adjustment flow rate (0.01mL / min). If the experimentally obtained wavelength position data is greater than the target wavelength position data +4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate plus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate minus the minimum adjustment flow rate. If the experimentally obtained wavelength position data is less than the target wavelength position data -4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate plus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate minus the minimum adjustment flow rate. If the experimentally obtained wavelength position data is greater than the target wavelength position data -4nm and less than the target wavelength position data +4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate minus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate plus the minimum adjustment flow rate.

[0067] Figure 2 This is a flowchart illustrating the automated control and wavelength positioning components of this invention. The automated component includes fluid line cleaning, precursor mixing, reaction, and data acquisition. Wavelength positioning is achieved based on automated control, meta-learning, and feedback control. When the target wavelength and fine-tuning flow rate are input, the meta-learning model provides the optimal combination of operating parameters. Online spectroscopy is used to confirm whether the experimental results meet expectations (i.e., within ±4nm of the target wavelength). If not, the system adjusts the experiment according to the set fine-tuning flow rate value and starts a new experiment until the experimental results meet expectations.

[0068] In the above steps, both the Br-containing precursor solution and the I-containing precursor solution contain oleic acid and oleylamine. The concentrations of Cs and Pb in the precursor solution are both 0.02M; the concentration of oleic acid is 0.1-0.2M, and the concentration of oleylamine is 0.02-0.05M; the solvents used include dimethyl sulfoxide and N,N-dimethylformamide, and the antisolvents used include chloroform and toluene.

[0069] In steps 2), a1), and a2) above, the total flow rate of the first and second automatic control feed pumps is 0.8-1.6 mL / min, and the flow rate ratio is controlled between 0.1 and 5. The flow rate of the fourth automatic control feed pump is 10-12 mL / min, and the ratio of the flow rate of the fourth automatic control feed pump to the total flow rate of the first and second automatic control feed pumps is controlled between 7 and 16.

[0070] The above meta-learning algorithm is a model-independent meta-learning algorithm.

[0071] In step 4) above, a suitable flow rate ratio is set between the third and fourth automatic control feed pumps, with the flow rate ratio controlled between 3 and 7. The purpose of introducing the antisolvent by the fourth automatic control feed pump is to clean and displace, and to avoid dead zones and residual solids at the T-junction.

[0072] In steps 1) and 2), the antisolvent is a poor solvent, which causes the crystals to crystallize due to the reduced solubility, thus obtaining the desired product.

[0073] This invention, based on a spiral reactor, combines automatic control with a micro-reaction platform to achieve rapid screening of reaction parameters such as the ratio of antisolvent to precursor solution, acid-base ratio, and residence time, thereby enabling the optimized synthesis of high-quality CsPbBr3 nanocrystals. Furthermore, building upon automatic control, this invention combines meta-learning algorithms with hysteresis algorithms, achieving targeted emission of CsPb(Br3) within the emission wavelength range of 510–635 nm by sampling three data points. x / I 1-x Synthesis of 3 nanocrystals.

[0074] Since the successful implementation of this invention depends on the rational design of reaction system components such as micromixers and microreactors (e.g., dimensions) and the appropriate selection of operating parameters (e.g., flow rate), preliminary experiments using the system of this invention are necessary before implementing the method of this invention to determine suitable design parameters for the reaction system components and suitable ranges for the selection of operating parameters. Examples 1-5 below are examples of preliminary experiments. The data obtained from the preliminary experiments can also be used to train the model of the relationship between the operating parameters of the microreactor system and the wavelength position of this invention.

[0075] Example 1: The Influence of Micromixer (120 cm long) and Microreactor (430 cm long, 14 mm spiral diameter) Designs on the Synthesis of CsPbBr3 Nanocrystals

[0076] For the synthesis of CsPbBr3 nanocrystals, a support structure with a helical height of 3 mm and a helical diameter of 10 mm was fabricated using a 3D printer. A 120 cm long PTFE capillary was wound around this support structure to form a micromixer. A U-shaped support structure with a helical height of 3 mm and a helical diameter of 14 mm was also fabricated using a 3D printer. A 430 cm long PFA capillary was wound around this support structure to form a microreactor. A quartz tube with an inner diameter of 2 mm, an outer diameter of 4 mm, and a length of 10 cm was selected as the detection tube. The spectrometer integration intensity was set to 10 ms, and the laser current was set to 5 mA.

[0077] The ratios of oleic acid and oleylamine in the 0.02M precursor solution of the first and second automatically controlled feed pumps were set to 4 and 32, respectively. The total flow rate of the first and second automatically controlled feed pumps was set to 1.4 mL / min, with flow rate ratios of 0.2, 1, and 4. The flow rate of the fourth automatically controlled feed pump was set to 10.6 mL / min. (The last sentence appears to be incomplete and possibly refers to fluorescence emission.) Figure 3 As shown, the fluorescence emission wavelengths of the obtained CsPbBr3 nanocrystals were 524 nm, 522 nm, and 517 nm, with full width at half maximum (FWHM) values ​​of 19 nm, 19 nm, and 18 nm, and fluorescence intensities of 3100, 28900, and 34000, respectively. The UV absorption peaks of the obtained CsPbBr3 nanocrystals are shown below. Figure 4 As shown, there are no obvious impurity absorption peaks at 315 nm and 325 nm, indicating that high-quality CsPbBr3 nanocrystals have been obtained.

[0078] Example 2: The Influence of Micromixer (150 cm long) and Microreactor (350 cm long, 14 mm spiral diameter) Designs on the Synthesis of CsPbBr3 Nanocrystals

[0079] For the synthesis of CsPbBr3 nanocrystals, a support structure with a helical height of 3 mm and a helical diameter of 10 mm was fabricated using a 3D printer. A 150 cm long PTFE capillary was wound around this support structure to form a micromixer. A U-shaped support structure with a helical height of 3 mm and a helical diameter of 14 mm was also fabricated using a 3D printer. A 350 cm long PFA capillary was wound around this support structure to form a microreactor. A quartz tube with an inner diameter of 2 mm, an outer diameter of 4 mm, and a length of 10 cm was selected as the detection tube. The spectrometer integration intensity was set to 10 ms, and the laser current was set to 5 mA.

[0080] The ratios of oleic acid and oleylamine in the 0.02M precursor solution of the first and second automatic feed pumps were set to 4 and 32, respectively. The total flow rate of the first and second automatic feed pumps was set to 1 mL / min, with flow rate ratios of 0.2, 1, and 4. The flow rate of the fourth automatic feed pump was set to 11 mL / min. (The last sentence appears to be incomplete and possibly refers to fluorescence emission.) Figure 5 As shown, the fluorescence emission wavelengths of the obtained CsPbBr3 nanocrystals were 522 nm, 520 nm, and 518 nm, with full width at half maximum (FWHM) values ​​of 20 nm, 20 nm, and 21 nm, and fluorescence intensities of 11500, 34600, and 33300, respectively. The UV absorption peaks of the obtained CsPbBr3 nanocrystals are shown below. Figure 6 As shown, there are no obvious impurity absorption peaks at 315 nm and 325 nm, indicating that high-quality CsPbBr3 nanocrystals have been obtained.

[0081] Example 3: The Influence of Micromixer (100 cm long) and Microreactor (430 cm long, 10 mm spiral diameter) Designs on the Synthesis of CsPbBr3 Nanocrystals

[0082] For the synthesis of CsPbBr3 nanocrystals, a support structure with a helical height of 3 mm and a helical diameter of 14 mm was fabricated using a 3D printer. A 100 cm long PTFE capillary was wound around this support structure to form a micromixer. A U-shaped support structure with a helical height of 3 mm and a helical diameter of 10 mm was also fabricated using a 3D printer. A 430 cm long PFA capillary was wound around this support structure to form a microreactor. A quartz tube with an inner diameter of 2 mm, an outer diameter of 4 mm, and a length of 10 cm was selected as the detection tube. The spectrometer integration intensity was set to 10 ms, and the laser current was set to 5 mA.

[0083] The ratios of oleic acid and oleylamine in the 0.02M precursor solution of the first and second automatically controlled feed pumps were set to 4 and 32, respectively. The total flow rate of the first and second automatically controlled feed pumps was set to 0.8 mL / min, with flow rate ratios of 0.2, 1, and 4. The flow rate of the fourth automatically controlled feed pump was set to 11.2 mL / min. (The last sentence appears to be incomplete and possibly refers to fluorescence emission.) Figure 7 As shown, the obtained CsPbBr3 nanocrystals exhibited fluorescence emission wavelengths of 520 nm, 517 nm, and 515 nm, with full width at half maximum (FWHM) values ​​of 22 nm, 24 nm, and 27 nm, and fluorescence intensities of 10200, 26700, and 27000, respectively. The UV absorption peaks of the obtained CsPbBr3 nanocrystals are shown below. Figure 8 As shown, there are no obvious impurity absorption peaks at 315 nm and 325 nm, indicating that high-quality CsPbBr3 nanocrystals have been obtained.

[0084] Example 4: Microreactor length design for the synthesis of CsPb(Br) x / I1-x The effect of 3 nanometer crystals (x = 0.33)

[0085] For the synthesis of CsPb(Br) x / I 1-x A micromixer was constructed using 3-nanometer crystals. A support structure with a 3mm helix height and 14mm diameter was fabricated using a 3D printer. A 150cm long PTFE capillary was wound around this support structure. A U-shaped support structure with a 3mm helix height and 10mm diameter was also fabricated using a 3D printer. PFA capillary tubes with lengths of 100, 200, 300, 400, and 500cm were wound around this support structure to form a microreactor. A quartz tube with an inner diameter of 2mm, an outer diameter of 4mm, and a length of 10cm was used as the detection tube. The spectrometer integration intensity was set to 10ms, and the laser current to 5mA.

[0086] The total flow rate of the first and second automatic feed pumps is set to 0.8 mL / min, with a flow rate ratio of 2. The flow rate of the fourth automatic feed pump is set to 11.2 mL / min. (For example, fluorescence emission...) Figure 9 As shown, the obtained CsPb(Br) x / I 1-x The fluorescence emission wavelengths of the 3-nano crystals were 555 nm, 562 nm, 564 nm, 567 nm, and 566 nm, with a full width at half maximum (FWHM) of 25 nm. Spectral analysis revealed that the wavelength positions and FWHM tended to stabilize under the conditions of a 500 cm long microreactor.

[0087] Example 5: Microreactor length design for the synthesis of CsPb(Br) x / I 1-x The effect of 3 nanometer crystals (x = 0.5)

[0088] For the synthesis of CsPb(Br) x / I 1-x A micromixer was constructed using 3-nanometer crystals. A support structure with a 3mm helical height and 14mm diameter was fabricated using a 3D printer. A 150cm long PTFE capillary was wound around this support structure. A U-shaped support structure with a 3mm helical height and 10mm diameter was also fabricated using a 3D printer. PFA capillary tubes with lengths of 100, 200, 300, 400, and 500cm were wound around this support structure to form a microreactor. A quartz tube with an inner diameter of 2mm, an outer diameter of 4mm, and a length of 10cm was selected as the detection tube. The spectrometer integration intensity was set to 10ms, and the laser current to 5mA.

[0089] The total flow rate of the first and second automatic feed pumps is set to 0.8 mL / min, with a flow rate ratio of 1. The flow rate of the fourth automatic feed pump is set to 11.2 mL / min. (For example, fluorescence emission...) Figure 10 As shown, the obtained CsPb(Br) x / I 1-x The fluorescence emission wavelengths of the 3-nano crystals were 580 nm, 588 nm, 593 nm, 601 nm, and 600 nm, with a full width at half maximum (FWHM) of 40 nm. Spectral analysis revealed that the wavelength positions and FWHM tended to stabilize under the conditions of a 500 cm long microreactor.

[0090] Example 6

[0091] The meta-learning model was trained based on the wavelength position data of CsPbBr3 nanocrystals at 520 nm in Example 3, and the wavelength position data of the microreactor with a length of 500 cm in Examples 4 and 5. The resulting model is as follows: Figure 11 As shown.

[0092] For the synthetic target CsPb(Br) x / I 1-x A micromixer was constructed using 3-nanometer crystals. A support structure with a 3mm helix height and 14mm diameter was fabricated using a 3D printer. A 150cm long PTFE capillary was wound around this support structure. A U-shaped support structure with a 3mm helix height and 10mm diameter was also fabricated using a 3D printer. A 500cm long PFA capillary was wound around this support structure to form a microreactor. A quartz tube with an inner diameter of 2mm, an outer diameter of 4mm, and a length of 10cm was used as the detection tube. The spectrometer integration intensity was set to 10ms, and the laser current to 5mA.

[0093] The total flow rate of the first and second automatic feed pumps was controlled at 0.8 mL / min, and the flow rate of the fourth automatic feed pump was set to 11.2 mL / min. The target wavelength position was set within ±4 nm as the upper and lower limits for experimental convergence. The wavelengths 535 nm, 546 nm, 557 nm, 578 nm, 591 nm, 609 nm, 619 nm, and 630 nm (the wavelength data corresponding to the highest point of each curve) were input into the computer as data points for the test model. The obtained fluorescence emission is shown below. Figure 12 As shown, the actual deviations of each wavelength from the target wavelength are as follows: Figure 13 As shown, the absolute mean deviation is 3.08 nm, proving that the model obtained in this invention has good generalization performance.

[0094] For the deviation at 630nm, a hysteresis algorithm was used to automatically adjust the flow rates of the Br-containing and I-containing precursor solutions. The original Br-containing flow rate was adjusted from 0.22mL / min to 0.24mL / min, and the original I-containing flow rate was adjusted from 0.58mL / min to 0.56mL / min. The final wavelength position obtained was 628nm.

[0095] Figure 14 This is a TEM image of the experimentally prepared CsPbBr3 nanocrystals. It can be seen that the nanocrystal material prepared by this system has a uniform cubic shape. Figure 15 It is an experimentally prepared CsPb(Br) x I 1-x TEM image of 3 nanometer crystals, where x = 0.66. Figure 16 This is the control interface of the meta-learning microreaction system developed in this invention.

[0096] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. Those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for the intelligent and precise preparation of target perovskite nanocrystals based on a microreaction system, characterized in that, The microreactor system includes an automatically controlled feed pump, an automatically controlled multi-channel switching valve, a micromixer, a microreactor, a monitoring device, and a capillary. The automatic control feed pump includes a first automatic control feed pump, a second automatic control feed pump, a third automatic control feed pump, and a fourth automatic control feed pump. A Y-type connector is provided between the first automatic control feed pump, the second automatic control feed pump, and the automatic control multi-channel switching valve. The outlet of the first automatic control feed pump is connected to the first inlet of the Y-type connector via a capillary tube. The outlet of the second automatic control feed pump is connected to the second inlet of the Y-type connector via a capillary tube. The outlet of the Y-type connector and the third automatic control feed pump are respectively connected to the inlets of the automatic control multi-channel switching valve via capillary tubes. The outlet of the automatic control multi-channel switching valve is connected to a micro-mixer via a capillary tube. A T-type connector is provided between the microreactor, the micro-mixer, and the fourth automatic control feed pump. The outlet of the micro-mixer is connected to the first inlet of the T-type connector via a capillary tube. The fourth automatic control feed pump is connected to the second inlet of the T-type connector via a capillary tube. The outlet of the T-type connector is connected to the microreactor via a capillary tube. The monitoring device includes an optical detection device and a computer; the optical detection device detects the optical properties of the products discharged from the microreactor and transmits the data to the computer for processing. Then, the computer sends instructions to the automatic control feed pump and the automatic control multi-channel switching valve based on the data processing results to control the reaction process. The method for intelligent and precise preparation of target perovskite nanocrystals includes the following steps: 1) The first automatic control feed pump introduces a Br-containing precursor solution, the second automatic control feed pump introduces an I-containing precursor solution, wherein the Br-containing precursor solution is a precursor solution containing CsBr and PbBr2, and the I-containing precursor solution is a precursor solution containing CsI and PbI2; the third automatic control feed pump introduces a solvent, and the fourth automatic control feed pump introduces an anti-solvent. 2) Prepare the target CsPb(Br) x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is used as input and input into the relationship model between the operating parameters of the micro-reaction system and the wavelength position obtained after training by the meta-learning algorithm; After the computer reads the fluorescence emission wavelength position data, it determines the flow rates of the first, second, and fourth automatically controlled feed pumps based on the relationship model between the microreaction system operating parameters and the wavelength position, and then performs target CsPb(Br) processing. x / I 1-x Preparation of 3 nanocrystals; wherein, the Br-containing precursor liquid and the I-containing precursor liquid are mixed in a micro mixer, and then contacted with the antisolvent at the T-joint and enter the microreactor for reaction. The product after reaction is obtained by optical detection device to obtain wavelength position data. 3) If the wavelength position data obtained from the experiment does not match the target CsPb(Br) x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is within ±4 nm. The computer uses a hysteresis algorithm to adjust the flow rate ratio of the first and second automatic control feed pumps until the wavelength position data of the reaction product meets the target CsPb(Br) emission wavelength. x / I 1-x The fluorescence emission wavelength position data of 3 nanometer crystals is within the range of ±4 nm. 4) When the target CsPb(Br) x / I 1-x After the 3-nano crystals are prepared, the multi-channel switching valve is automatically controlled by computer commands to open the channels of the third and fourth automatic feed pumps, so that the solvent and antisolvent are introduced into the micro-reaction system to flush the fluid channels of the micro-mixer, microreactor and optical detection device.

2. The method as described in claim 1, characterized in that, In step 2), the training process of the micro-reaction system operating parameters and wavelength position relationship model obtained after training by the meta-learning algorithm includes the following steps: a) The initialized meta-learning algorithm neural network model Pre-training is performed on a meta-learning algorithm neural network model that is a single-input, single-output multilayer perceptron with two hidden layers. The support and query sets for pre-training are sampled using a sine function, in the following form: ； Among them, amplitude The phase varies within the range of [0.1, 7.0]. Variation within the range [0, 2π]; Input and output The dimension is one-dimensional; for each sinusoidal task T i Multiple samples were uniformly sampled from [-7,7]. , The support set is composed of ( ) elements; the mean squared error is used as the loss function, and takes the following form: ； Based on the i-th sine task T i Support concentration ( , The mean squared error loss is calculated using a one-step gradient descent method to convert the initialized neural network model parameters θ into θmin. i Here, the original model parameters θ are not updated; only the new model parameters θ are calculated. i The calculation of parameters follows the following form: ； in, For learning rate, For loss function The gradient; Another number of new data points were extracted ( , To form the sinusoidal task T i The query set, the new mean squared error loss is based on θ i The model's original parameters θ are calculated using the new data; the Adam optimization algorithm is used to update the model's original parameters θ, and the update follows this form: ； in, θ is the learning rate; based on the updated model parameters θ, the pre-trained model is obtained; b) CsPb(Br) prepared under various microreaction system operating parameters and conditions respectively. x / I 1-x The peak wavelength data of 3 nanometer crystals are used as input and output to form the support set for the experimental task. The pre-trained model obtained in step a) is used to obtain the final model of the relationship between the operating parameters of the micro-reaction system and the wavelength position.

3. The method as described in claim 2, characterized in that, Step b) specifically involves: b1) Selecting three microreaction system operating parameters and conditions to prepare CsPb(Br) x / I 1-x The peak wavelength data of 3 nanometer crystals is used as input and output, and in one of the operating parameters of the micro-reaction system, the flow rate of the second automatic control feed pump is 0. b2) The operating parameters of the microreactor system with the second automatically controlled feed pump flow rate at 0 and the CsPb(Br) prepared under these conditions x / I 1-x The peak wavelength data of 3 nanometer crystals was obtained by the following method: the Br-containing precursor liquid was loaded into the first automatic control feed pump, and the solvent and antisolvent were loaded into the third and fourth automatic control feed pumps; at room temperature, the Br-containing precursor liquid was mixed by a micro mixer and then contacted with the antisolvent at the T-joint, and then entered the microreactor for reaction. After the reaction, the product was detected by an optical detection device to obtain a wavelength position data. b3) The other two sets of microreaction system operating parameters and conditions under which CsPb(Br) was prepared x / I 1-x The peak wavelength data of the 3-nanometer crystals were obtained as follows: Br-containing and I-containing precursor solutions were loaded into the first and second automatically controlled feed pumps, respectively, while the solvent and antisolvent were loaded into the third and fourth automatically controlled feed pumps. At room temperature, the Br-containing and I-containing precursor solutions were mixed in a micro-mixer, then contacted with the antisolvent at the T-joint and entered the microreactor for reaction. The wavelength position data of the reaction product was obtained by an optical detection device. By changing the flow rate ratio of the first and second automatically controlled feed pumps, two sets of microreactor system operating parameters and their corresponding two wavelength position data were obtained. b4) Use the data obtained in the above steps to form the support set for the experimental task, and train the pre-trained model obtained in step a); finally, obtain the relationship model between the operating parameters of the micro-reaction system and the wavelength position.

4. The method as described in claim 1, characterized in that, The hysteresis algorithm is as follows: If the experimentally obtained wavelength position data is greater than the target wavelength position data + 4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate plus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate minus the minimum adjustment flow rate. If the experimentally obtained wavelength position data is less than the target wavelength position data - 4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate plus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate minus the minimum adjustment flow rate. If the experimentally obtained wavelength position data is greater than the target wavelength position data - 4nm and less than the target wavelength position data + 4nm, then the flow rate of the new bromine-containing precursor solution feed pump is equal to the current flow rate minus the minimum adjustment flow rate, and the flow rate of the new iodine-containing precursor solution feed pump is equal to the current iodine precursor solution feed pump flow rate plus the minimum adjustment flow rate.

5. The method as described in claim 1, characterized in that, Both the Br-containing and I-containing precursor solutions contain oleic acid and oleylamine. The concentrations of Cs and Pb in the precursor solutions are both 0.02 M; the concentration of oleic acid is 0.1-0.2 M, and the concentration of oleylamine is 0.02-0.05 M; the solvent includes dimethyl sulfoxide or N,N-dimethylformamide; and the antisolvent includes chloroform or toluene.

6. The method as described in claim 1, characterized in that, In step 2), the total flow rate of the first and second automatic control feed pumps is 0.8-1.6 mL / min, and the flow rate ratio is 0.1-5; the flow rate of the fourth automatic control feed pump is 10-12 mL / min, and the ratio of the flow rate of the fourth automatic control feed pump to the total flow rate of the first and second automatic control feed pumps is 7-16.

7. The method as described in claim 1, characterized in that, The micro mixer includes a helical tube and its 3D support structure. The helical tube has a helical height of 2-5 mm, a helical diameter of 7-16 mm, and a helical length of 100-150 cm. The micro mixer is used to mix the precursor liquid introduced by the first automatic control feed pump and the second automatic control feed pump. The first automatic control feed pump introduces the precursor liquid containing Br, and the second automatic control feed pump introduces the precursor liquid containing I.

8. The method as described in claim 1, characterized in that, The microreactor includes a fluid reversing helical tube and its 3D support structure. The helix height of the fluid reversing helical tube is 3-5 mm, the helix diameter is 8-14 mm, and the helix length is 300-500 cm. The microreactor achieves fluid reversal by winding around the 3D support structure with helical grooves.

9. The method as described in claim 1, characterized in that, The optical detection device consists of a detection bracket, optical fiber, spectrometer and laser; the detection bracket includes a detection tube and its 3D support structure, the detection tube is made of transparent polytetrafluoroethylene or quartz, the inner diameter of the detection tube is 1-5mm, the outer diameter is 2-7mm, and the length is 10cm.

Images