|
ИСТИНА |
Войти в систему Регистрация |
ИСТИНА ПсковГУ |
||
Finding optimal carbon dots synthesis parameters for quantitative analysis of components in multi-component aqueous solutions using machine learningтезисы доклада
- Авторы: Guskov A.A., Isaev I.V., Laptinskiy K.A., Dolenko T.A., Dolenko S.A.
- Сборник: DLCP Abstracts
- Тезисы
- Год издания: 2025
- Аннотация: Optical nanosensors based on carbon dots (CD) introduced into the object of the study are widely used for analyzing the content of multicomponent liquid media. Their applicability stems from the high sensitivity of CD photoluminescence to changes in medium parameters, such as pH and solution temperature, type and concentration of dissolved substances. In addition, such sensors offer stable optical properties, biocompatibility, low production costs, and enable both rapid and remote diagnostics of the investigated objects. At the same time, the physicochemical properties of CD strongly depend on the synthesis method and the precursors used. Therefore, CD synthesized under different conditions can show significantly different photoluminescence responses even when placed in identical solutions. This study focuses on finding the optimal CD synthesis parameters that ensure high-precision determination of the concentration of specific heavy metal salts in aqueous solutions. Creation of nanoparticles with such features represents an urgent scientific and practical challenge. In the physical experiment, 74 types of CD were prepared by hydrothermal synthesis from citric acid and ethylenediamine under a wide range of precursor ratios, temperatures and reaction times. Then each type of CD was placed in aqueous solutions of Co(NO₃)₂ and Cu(NO₃)₂ with salt concentrations ranging from 0 to 6 mM in increments of 0.67 mM. As the result, 100 samples were prepared for each type of CD. The excitation-emission spectra of carbon dots fluorescence were registered for all obtained aqueous solutions of CD and salts. To solve the stated task, it is reasonable to apply machine learning methods that are capable of revealing hidden relationships in multiparametric systems. In this study, an artificial neural network based on a multilayer perceptron architecture was used to approximate the dependence of the target variable – the error in determining the concentrations of the studied heavy metal salts – on the CD synthesis parameters. The target variable values for training the approximating model were obtained as a result of solving the inverse problem of determining the concentrations of heavy metal salts from the excitation-emission spectra of fluorescence of CD by machine learning algorithms. This two-step approach may allow one to find carbon dot synthesis parameters that yield the desired accuracy in determining the concentrations of various components in multicomponent solutions.
- Добавил в систему: Гуськов Артём Алексеевич