Robot-assisted automated characterization of nanocrystals accelerates intelligent reverse design of materials

Abstract

The present invention discloses an intelligent reverse design for accelerating materials for robot-assisted automatic characterization of nanocrystals. Specifically, the following design methods are disclosed: S1) using a high-throughput preparation platform to conduct small-dose experiments; S2) constructing a small-dose experimental database; S3) designing a decision optimization disk; S4) detecting the longitudinal plasmon resonance absorption peak, absorbance ratio, and half-maximum width of the nanocrystals obtained under different reaction conditions using the decision optimization disk, and selecting the experimental conditions that are closest to the target morphological characteristics as further amplification conditions; S5) conducting amplification experiments and obtaining the morphological characteristics of the nanocrystals in each group of experiments; S6) establishing an equation between the reaction volume in the step-by-step method and the corresponding morphological characteristic parameters. The present invention implements intelligent reverse design by implementing robot-assisted automatic characterization of nanocrystals in nanocrystal amplification experiments.

Description

Technical Field

[0001] The present invention belongs to the fields of novel material preparation technology and digital manufacturing technology, and in particular relates to robot-assisted nanocrystal automatic characterization and accelerated intelligent reverse design of materials. Background Art

[0002] In recent years, both domestic and international research on digitally driven materials innovation has received significant attention. The advantages of integrating materials and artificial intelligence technologies are becoming increasingly apparent: efficiency and accuracy are improving [1-3]; dependence on computing resources is decreasing [4,5]; machine learning algorithms and self-training capabilities are being strengthened [6,7]; learning scales and databases are expanding [8,9]; and reverse engineering of functional materials is gradually maturing [1,10]. However, the lack of reliable experimental data and descriptors of specific properties has limited the development of artificial intelligence in the materials field. Currently, research focuses on combining first-principles computational data with machine learning (supplemented by experiments) [11-15]. Research on digital and intelligent material preparation, combining robotics / process automation with artificial intelligence, is gaining momentum [16-20].

[0003] The process of automated organic material preparation is a hot topic in current research: Li, Ballmer, et al.

[18] studied the fully automated preparation of 14 different organic small molecules, clarifying a feasible roadmap for the automated preparation of more general organic small molecules; Cronin et al.

[19] developed a machine learning-driven organic synthesis robot that can predict possible reactions after a small number of experiments. In the field of biomaterials, He Kai et al.

[21] developed an automated experimental platform for synthetic biology; Ager et al.

[22] combined automation and computer technology to develop a workstation for optimizing the performance of biocatalysts. In the field of inorganic materials research, Cooper et al.

[20] proposed the use of mobile robots to screen photocatalysts for water hydrogen production, focusing on the evaluation experiments of multiple catalyst formulas rather than the preparation of new materials.

[0004] However, the basic ideas of the above research mainly focus on the process automation and mobile robots of organic materials and biomaterials, but there are few reports on the programmable rational design methods of the entire process of mathematical models, databases and AI algorithms for material performance evaluation of nanocrystalline materials.

[0005] 1.Zunger, A., Inverse design in search of materials with targetfunctionalities. Nature Reviews Chemistry, 2018.2(4):p.0121.

[0006] 2.Hautier,G.,A.Jain,and S.P.Ong,From the computer to the laboratory:materials discovery and design using first-principles calculations.Journal ofMaterials Science,2012.47(21):p.7317-7340.

[0007] 3.Liu,Y.,et al.,Materials discovery and design using machinelearning.Journal of Materiomics,2017.3(3):p.159-177.

[0008] 4.Chen,C.,et al.,A Critical Review of Machine Learning of EnergyMaterials.Advanced Energy Materials,2020.10(8):p.1903242.

[0009] 5.Schleder,G.R.,C.M.Acosta,and A.Fazzio,Exploring Two-DimensionalMaterialsThermodynamic Stability via Machine Learning.ACS Applied Materials&Interfaces,2020.12(18):p.20149-20157.

[0010] 6.Jablonka,K.M.,et al.,Big-Data Science in Porous Materials:MaterialsGenomics and Machine Learning.Chemical Reviews,2020.120(16):p.8066-8129.

[0011] 7.Ward,L.,et al.,A general-purpose machine learning framework forpredicting properties of inorganic materials.npj Computational Materials,2016.2(1):p.16028.

[0012] 8.Tabor,D.P.,et al.,Accelerating the discovery of materials for cleanenergy in the era of smart automation.Nature Reviews Materials,2018.3(5):p.5-20.

[0013] 9.Chen,C.,et al.,Graph Networks as a Universal Machine LearningFramework for Molecules and Crystals.Chemistry of Materials,2019.31(9):p.3564-3572.

[0014] 10.Sanchez-Lengeling,B.and A.Aspuru-Guzik,Inverse molecular designusing machine learning:Generative models for matter engineering.Science,2018.361(6400):p.360.

[0015] 11.Bruix,A.,et al.,First-principles-based multiscale modelling ofheterogeneous catalysis.Nature Catalysis,2019.2(8):p.659-670.

[0016] 12.Zhong,M.,et al.,Accelerated discovery of CO2 electrocatalystsusing active machine learning.Nature,2020.581(7807):p.178-183.

[0017] 13.Seh,Z.W.,et al.,Combining theory and experiment inelectrocatalysis:Insights into materials design.Science,2017.355(6321):p.eaad4998.

[0018] 14.Ma,X.-Y.,et al.,Large family of two-dimensional ferroelectricmetals discovered via machine learning.Science Bulletin,2021.66(3):p.233-242.

[0019] 15.Lu,S.,et al.,Accelerated discovery of stable lead-free hybridorganic-inorganic perovskites via machine learning.Nature Communications,2018.9(1):p.3405.

[0020] 16.Perera,D.,et al.,A platform for automated nanomole-scale reactionscreening and micromole-scale synthesis in flow.Science,2018.359(6374):p.429-434.

[0021] 17.Mehr,S.H.M.,et al.,A universal system for digitization andautomatic execution of the chemical synthesis literature.Science,2020.370(6512):p.101-108.

[0022] 18.Li,J.,et al.,Synthesis of many different types of organic smallmolecules using one automated process.Science,2015.347(6227):p.1221-1226.

[0023] 19. Granda, JM, et al., Controlling an organic synthesis robot with machine learning to search for new reactivity. Nature, 2018.559(7714): p.377-381.

[0024] 20. Burger, B., et al., A mobile robotic chemist. Nature, 2020.583(7815): p.237-241.

[0025] 21. He Kai, Chen Shu, Zhao Wenliang, Fang Haitao. Automated casting platform system integration technology for synthetic biology: China, 111482212A, 2020-08-04. Under review - open to the public.

[0026] 22.Ager, D., et al., Method and apparatus for optimization of high-throughput screening and enhancement of biocatalyst performance. 2002, GooglePatents. Summary of the Invention

[0027] The purpose of the present invention is to provide a robot-assisted automatic characterization of nanocrystals to accelerate the intelligent reverse design of materials, solve the inherent laws of the existing technology in studying the in-situ automatic characterization of large sample physical property data such as color and spectrum and small sample electron microscope morphology data, reveal the role of experimental parameters related to morphology control agents as "logic gates" in the programming program, and realize the reverse design of nanocrystals of "programmable design-robotic preparation-automatic characterization-redesign".

[0028] To solve the above problems, the present invention establishes a characterization model by utilizing the relationship between reaction color and nanocrystal morphology, and implements redesign in specific nanocrystal research, namely, nanocrystal amplification experiments.

[0029] One aspect of the present invention provides a method for preparing nanocrystals by amplification, which comprises the following steps:

[0030] S1) using a high-throughput preparation platform to perform a small-dose experiment on nanocrystal preparation conditions for the nanocrystal to be prepared, wherein the reaction volume for each experiment is no greater than 1.5 mL;

[0031] S2) detecting the morphological characteristics of the nanocrystals obtained in step S1), and constructing a small-dose experimental database based on the reaction conditions and corresponding morphological characteristic data, further screening experimental schemes that can obtain the target morphological characteristics from the small-dose experimental database, and analyzing the reaction conditions of each experimental scheme, screening the reaction conditions with the highest frequency to form a basic scheme;

[0032] S3) designing a decision optimization disk of reaction conditions based on the basic scheme, wherein the experimental scheme in the decision optimization disk includes one or more reaction conditions that deviate from the basic scheme;

[0033] The reaction conditions are derived from adjustments to the basic scheme, and one or more reaction conditions in the basic scheme reaction conditions are selected for adjustment, with the adjustment range being 80%-120% of the basic scheme value;

[0034] S4) detecting the longitudinal plasmon resonance absorption peak (LSPR), absorbance (OD) ratio, and full width at half maximum (FWHM) value of the nanocrystals obtained under different reaction conditions of the decision optimization disk, and screening the experimental conditions closest to the target morphological characteristics by LSPR, OD ratio, and FWHM values ​​as further amplification conditions;

[0035] S5) further performing amplification experiments using the amplification conditions determined in step S4), performing 4-6 groups of step-by-step amplification experiments, with the reaction volume magnified by 2-6 times, and obtaining the morphological characteristics of the nanocrystals in each group of experiments;

[0036] S6) establishing an equation between the reaction volume in the step-by-step method and the corresponding morphological characteristic parameters; and predicting the corresponding nanocrystal morphological characteristics in different amplification reactions through the equation.

[0037] Furthermore, step 7) is included, according to S1) small-dose experiment and S2) decision optimization, the experimental conditions under which the LSPR value of the nanocrystal morphology characteristics in the disk is lower than the LSPR value of the amplified experiment, the reaction conditions that reduce the LSPR value are screened, and the amplified experimental conditions are corrected.

[0038] Furthermore, the morphological characteristics of the nanocrystals are LSPR values ​​or RGB values ​​of the nanocrystals.

[0039] Furthermore, the decision optimization disk in S3) includes single-factor, double-factor, and triple-factor adjustment conditions.

[0040] Furthermore, in S3), the number of different reaction conditions contained in a single decision optimization disk is 6 or more, for example, 6-12.

[0041] Furthermore, the screening method for LSPR, OD ratio, and FWHM value in S4) is to sequentially screen 10% of the experimental conditions with LSPR close to the target morphological characteristics in the preferred decision disk, and select the experimental conditions with high OD ratio and low FWHM among them.

[0042] Furthermore, the step-by-step amplification experiment in S5) was respectively set as a reaction volume 2 mL group, a reaction volume 4 mL group, a reaction volume 20 mL group, and a reaction volume 40 mL group.

[0043] Furthermore, the equation described in S6) is a linear equation.

[0044] Furthermore, the nanocrystals are selected from gold nanocrystals; in step S7), the red shift of the LSPR value in the gold nanocrystal amplification experiment is corrected by reducing the HCl concentration in the reaction.

[0045] Another aspect of the present invention provides a modeling method for predicting the LSPR value of nanocrystals based on the color of the nanocrystal reaction solution.

[0046] S01) obtaining the RGB colors and LSPR values ​​of different reaction solutions of corresponding nanocrystals and establishing a database;

[0047] S02) Using a machine learning method to construct a model between RGB colors and LSPR values.

[0048] Furthermore, in step S01), the amount of different reaction solutions collected is more than 80.

[0049] Furthermore, the nanocrystals are selected from gold nanocrystals.

[0050] Another aspect of the present invention provides a method for predicting the LSPR value of nanocrystals by the color of nanocrystal reaction solution. The method is a model constructed using the above method, and the LSPR value of the nanocrystal to be tested is obtained by inputting the RGB value of the color of the nanocrystal reaction solution to be tested.

[0051] Beneficial effects

[0052] 1) The present invention unexpectedly discovered the phenomenon of red shift of LSPR value during nanocrystal amplification experiment, and realized the prediction of LSPR value of nanocrystal after amplification experiment based on the results of step-by-step amplification experiment.

[0053] 2) The present invention confirms that the method of reducing the LSPR value obtained by analyzing the results of small-dose experiments or decision disk experiments, and the correction of the LSPR red shift in the amplified experiment is achieved after applying this condition to the amplified experiment.

[0054] 3) This invention demonstrates for the first time that the RGB color of gold nanocrystal reactants is correlated with their morphological characteristics. Using machine learning, a RGB-LSPR model was derived, demonstrating high accuracy, with a variance exceeding 0.9. As the nanocrystal preparation process becomes electronic, mechanized, integrated, and scaled, the use of digital RGB color can enable faster assessment and screening of nanocrystal morphological characteristics, potentially even becoming a useful tool for monitoring nanocrystal reaction progress.

[0055] 4) The present invention establishes a gold nanocrystal RGB color and LSPR value model based on gold nanocrystals, and through the amplification experiment of gold nanocrystals, obtains a solution to correct the red shift of the gold nanocrystal LSPR during the amplification experiment. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Schematic diagram of the decision disk for scale-up experiments.

[0057] Figure 2 Schematic diagram of the step-by-step amplification experimental process.

[0058] Figure 3 is the LSPR value during the step-by-step amplification experiment.

[0059] Figure 4 is the linear equation for the red shift phenomenon during the step-by-step amplification experiment.

[0060] Figure 5 In order to reduce the HCl concentration in the gold nanocrystal amplification experiment reaction system, the LSPR value was reduced and the red shift problem during the amplification experiment was corrected.

[0061] Figure 6 High-quality digital photos of 96 single-factor experiments taken with a color ultra-sensitive camera and RGB color blocks automatically digitally converted by the system.

[0062] Figure 7 High-quality digital photos of 96 two-factor experiments taken with a color ultra-sensitive camera and RGB color blocks automatically digitally converted by the system.

[0063] Figure 8 High-quality digital photos of 96 experiments of three factors taken by the ultra-sensitive color camera and RGB color blocks after automatic digital conversion by the system.

[0064] Figure 9The following is a visual expression of the relationship model between color data RGB and LSPR data. DETAILED DESCRIPTION

[0065] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below, but it should not be understood as limiting the scope of implementation of the present invention.

[0066] Example 1 Nanocrystal Amplification Experiment Decision

[0067] The preparation cost of noble metal nanocrystals is high. To save experimental costs, small-dose experiments are typically used to explore experimental conditions, followed by scale-up experiments. During the experimental process, the present inventors unexpectedly discovered that as the experimental dose was gradually increased, the LSPR value shifted. In other words, small-dose experiments could not fully guide scale-up experiments. The present inventors investigated this issue.

[0068] First, using a high-throughput platform, we conducted thousands of experimental explorations to build a nanogold material database, collecting 99 synthesis methods for gold nanorods with a target LSPR of 808±10nm. The specific parameters are shown in Table 1, corresponding to the content in the figure.

[0069] Table 1. Database of experimental parameters for synthesizing LSPR of 808 ± 10 nm gold nanorods (1 mL volume)

[0070]

[0071]

[0072]

[0073] In Table 1, adjusting the CTAB concentration (C(CTAB)), chloroauric acid concentration (C(HAuCl4)), silver nitrate volume (V(AgNO3)), hydrochloric acid concentration (C(HCl)), and ascorbic acid concentration (C(AA)) parameters individually or simultaneously can all yield gold nanorods with an LSPR of 808±10 nm. In the database, chloroauric acid, as a reaction precursor, is reduced to Au(I) and ultimately Au(0) by the reducing agent AA. Lowering the chloroauric acid concentration or the reducing agent concentration can lead to reduced gold nanorod yield or low gold salt reduction recovery. Furthermore, CTAB, as a quaternary ammonium salt of a long-chain hydrocarbon group (molecular weight 364.46 g / mol), has high viscosity at high concentrations and is prone to crystallization at room temperature. Excessively high CTAB concentrations can also result in a slow reaction rate in the gold nanorod growth system. Therefore, from the numerous experimental schemes in the database, the most representative experimental condition combinations were selected as benchmark experiments. The three primary factors in the 1mL benchmark experiment—V(AgNO₃), C(CTAB), and C(HCl) at 10 μL, 0.1 M, and 1.0 M, respectively—served as the basis for subsequent optimization scheme adjustments. Minor factors in the benchmark scheme—C(HAuCl₄) at 0.01 M, V(seed) at 2.4 μL, and V(AA) at 8 μL—were not adjusted in the scale-up experiments because they affect product yield, gold salt reduction recovery, and reduce reaction rate. Scale-up validation and optimization were achieved by adjusting single, double, and triple factors. Representative experimental condition combinations are shown in Table 1. It is worth noting that adjustments to the benchmark experimental conditions were required to obtain the target LSPR gold nanorods.

[0074] Although many reaction conditions can obtain the target LSPR through experimental accumulation and conditions disclosed in existing technologies, further screening from a large number of reaction conditions is still needed to obtain more appropriate amplification experimental results.

[0075] To this end, the inventor designed a decision optimization disk, selecting single-factor, double-factor, and three-factor adjustment methods from the database and adjusting the experimental parameters of the basic solution within ±20% to form an optimized decision disk.

[0076] A comprehensive evaluation was conducted on the experimental results of the optimized disk, examining the degree to which the LSPR approaches 808 nm (the closer to 808 nm, the better), the full width at half maximum (FWHM) (the smaller the FWHM, the more uniform the aspect ratio of the nanorods), and the optical density (OD) ratio (when the LSPR is basically fixed, the higher the OD ratio, the lower the proportion of spherical and other non-rod-shaped gold nanoparticles, that is, the less impurity morphology in the reaction product and the higher the nanorod yield).

[0077] A typical decision optimization disk is shown in Table 2. In the benchmark experimental scheme, V(AgNO3), C(CTAB), and C(HCl) are 20μL, 0.1M, and 1.0M, respectively, which are the basis for adjustment of subsequent optimization schemes. C(HAuCl4) is 0.01M, Vseed is 4.8μL, and V(AA) is 16μL in the benchmark scheme, which are not adjusted here because they affect product yield, gold salt reduction recovery rate, and reduce reaction rate. On the basis of the benchmark experimental scheme, schemes by adjusting single factors (single variable), double factors (double variables), and three factors (three variables) were selected from the database, and extended experiments were added to adjust the experimental parameters within ±20%. Therefore, the database established through high-throughput experiments in 1mL reaction systems provides a reference and amplification guidance in 2mL reaction systems. The decision optimization disk designed in Table 2 can achieve the purpose of optimization by further adjusting the experimental parameters and conducting cyclic studies. The results of Table 2 are shown in Figure 1 The middle (right) image shows a comprehensive examination of LSPR, OD ratio, and FWHM values. Experiments 1, 4, 7, and 12 achieved an LSPR of 808 ± 10 nm. Experiment 7 had the highest OD ratio and the narrowest FWHM. Based on the OD ratio and FWHM, the optimal sample (i.e., Experiment 7) was identified and recommended for the final amplification step.

[0078] Table 2. Experimental decision disks and results for gold nanorods with a target LSPR of 808 ± 10 nm

[0079]

[0080] *Bold data are adjustments made to the benchmark experiment parameters. Due to the doubling of the reaction volume, the V(AgNO3) in the benchmark experiment in a 2 mL reaction system is twice that of the 1 mL database system, that is, 20 μL.

[0081] Three stages of amplification experiments were carried out in sequence, namely various plate amplification experiments on a robotic automation platform, amplification experiments on a magnetic stirrer, and pilot amplification experiments in a stirring container. First, the optimized 633, 780 and 808nm samples were experimentally studied on the robotic platform with 2mL, 4mL (6-well plate), 20mL and 40mL (single-well plate), respectively. During the amplification process, an unexpected discovery was that as the dose increased, the LSPR gradually red-shifted compared with the results on the 96-well plate, thus providing new insights into the exploration of amplification rules. Studying these data, it can be found that there is a linear relationship between the amplification of the dose and the distance of the red shift. The results are shown in Figure 4 From this result, we can see that the results of the scale-up experiment can be predicted by the results of the small-dose experiment, and the results of the scale-up experiment can be guided by adjusting the reaction conditions of the small-dose experiment.

[0082] In further scale-up experiments, in order to obtain nanocrystals with LSPR around 808 nm, the HCl concentration was reduced and experiments were conducted at 200 mL, 1 L, and 15 L volumes, respectively. The measured UV-Vis-NIR spectra are shown in Figure 2. Figure 5 As shown in the figure, their LSPRs are 806nm, 802nm, and 812nm, respectively. These values ​​fall within the target range, achieving good amplification.

[0083] Color conversion and digital processing of experimental results in Example 2.

[0084] For different gold nanocrystal morphology parameters (aspect ratio), in addition to using LSPR for evaluation, the present invention also innovatively uses color value (RGB) to identify the morphology of nanocrystals. Because color features are easy to obtain and identify, they are convenient for machine learning prediction and logical calculation. In this way, when mechanized electronic equipment is used for experiments, trained models can be used to process and digitally identify colors. Color features are expected to be easy to obtain and quickly respond to input logical morphology control, which is crucial for the controllable synthesis of target nanocrystals.

[0085] First, a batch of gold nanorod samples were obtained through different experimental methods, and the obtained samples were photographed using a color ultra-sensitive camera to obtain high-quality photos. Figure 6 , Figure 7 and Figure 8 .

[0086] The colors in the photographs were digitized to obtain color blocks and RGB data for each gold nanorod. The LSPR data corresponding to each sample was measured separately, and the RGB data and LSPR data were mapped one-to-one to form a training database. The LSPR-color correspondences from the experimental results and the LSPR data and color data calculated by the model are shown in Tables 2, 3, and 4.

[0087] Table 2. LSPR of gold nanorod samples obtained from 96 single-factor experiments and the corresponding RGB values ​​of the digitally converted colors.

[0088]

[0089]

[0090] Table 3. LSPR of gold nanorod samples obtained from 96 two-factor experiments and the corresponding RGB values ​​of the digitally converted colors.

[0091]

[0092]

[0093] Table 4. LSPR of gold nanorod samples obtained from 96 three-factor experiments and the corresponding RGB values ​​of the digitally converted colors.

[0094]

[0095] The machine learning method is used to study the data in the training database to obtain the relationship between the LSPR value and the RGB value of the sample color. The LSPR data and color data (RGB) model is established and gradually optimized.

[0096] Table 5. Relationship model between LSPR data and color data (RGB)

[0097]

[0098] Through data collection from large-scale experiments, the present invention has discovered for the first time that the morphological characteristics of gold nanocrystals can be obtained by the RGB values ​​of the sample color of gold nanocrystals, which will help to quickly obtain the sample characteristics of the nanocrystals. The easy availability of the sample color RGB color makes the experimental verification method simpler and easier to obtain. At the same time, it also helps mechanized, electronic, and batch synthesis of nanocrystals, and provides a new option for the digitization of nanocrystals. The visualization of color data and LSPR data model, the results are shown in Figure 9 .

Claims

1. A method for preparing nanocrystals by amplification, characterized in that: It includes the following steps: S1) using a high-throughput preparation platform to perform a small-dose experiment on the nanocrystal preparation conditions, wherein the small-dose experiment comprises a reaction volume of no more than 1.5 mL per experiment; S2) detecting the morphological characteristics of the nanocrystals obtained in step S1), and constructing a small-dose experimental database based on the reaction conditions and corresponding morphological characteristic data, further screening experimental schemes that can obtain the target morphological characteristics from the small-dose experimental database, and analyzing the reaction conditions of each experimental scheme, screening the reaction conditions with the highest frequency to form a basic scheme; S3) designing a decision optimization disk of reaction conditions based on the basic plan, wherein the experimental plan in the decision optimization disk includes one or more reaction conditions that deviate from the basic plan; The reaction conditions are derived from the adjustment of the basic scheme. One or more reaction conditions in the basic scheme reaction conditions are selected for adjustment. The adjustment range is 80%-120% of the basic scheme value. S4) detecting the longitudinal plasmon resonance absorption peak (LSPR), absorbance (OD) ratio, and full width at half maximum (FWHM) value of the nanocrystals obtained under different reaction conditions of the decision optimization disk, and screening the experimental conditions closest to the target morphological characteristics by LSPR, OD ratio, and FWHM values ​​as further amplification conditions; S5) further performing amplification experiments using the amplification conditions determined in step S4), performing 4-6 groups of step-by-step amplification experiments, with the reaction volume magnified by 2-6 times, and obtaining the morphological characteristics of the nanocrystals in each group of experiments; S6) establishing an equation between the reaction volume in the step-by-step method and the corresponding morphological characteristic parameters; and predicting the corresponding nanocrystal morphological characteristics in different amplification reactions through the equation.

2. The method for preparing nanocrystals by amplification according to claim 1, wherein: The method further includes step 7), wherein the experimental conditions under which the LSPR value of the nanocrystal morphology characteristics in the disk is lower than the LSPR value of the amplified experiment are optimized according to the small-dose experiment in S1) or the decision in S4), reaction conditions that reduce the LSPR value are screened, and the amplified experimental conditions are corrected.

3. The method for preparing nanocrystals by amplification according to claim 1, characterized in that: The morphological characteristics of the nanocrystals are the LPRS values ​​or RGB values ​​of the nanocrystals.

4. The method for preparing nanocrystals by amplification according to claim 1, characterized in that: The screening method for LSPR, OD ratio, and FWHM values ​​in S4) is to sequentially screen 10% of the experimental conditions in the decision-making optimization disk that have LSPRs close to the target morphological features, and select the experimental conditions with high OD ratio and low FWHM among them.

5. The method for preparing nanocrystals by amplification according to claim 1, characterized in that: The decision optimization disk in S3) includes single-factor, double-factor, and triple-factor adjustment conditions.

6. The method for preparing nanocrystals by amplification according to claim 1, characterized in that: The equation described in S6) is a linear equation.

7. The method for preparing nanocrystals by amplification according to claim 1, characterized in that: In S3), the number of different reaction conditions included in the decision optimization disk is more than 6.

8. The method for preparing nanocrystals by amplification according to claim 2, characterized in that: The nanocrystals are selected from gold nanocrystals.

9. The method for preparing nanocrystals by amplification according to claim 8, characterized in that: In step S7), the red shift of the LSPR value in the gold nanocrystal amplification experiment is corrected by reducing the HCl concentration in the reaction.

Images