Simulation Study Of The Stability Of Borosilicate Glass

Nuclear energy plays an important role in improving economic efficiency as a safe,reliable and effective clean energy source and has been increasingly used in recent years.However,while nuclear energy benefits mankind,it also generates radioactive waste,of which high-level radioactive waste is characterised by high nuclide toxicity and radioactivity,which will have a serious impact on ecological safety if not disposed of properly.The use of highly radioactive waste to solidify glass and then dispose of it in deep geology is now an internationally accepted and feasible solution.Borosilicate glass has become the preferred glass curing body material for curing HLW in many countries,including China,due to its wide range of inclusion,leaching resistance and excellent irradiation resistance.Currently,China’s existing HLW is characterised by high sulphur and sodium content.However,traditional borosilicate glass has a low solubility for sulphate.During the glass melting process,sulphur exceeding the solubility of the curing body will produce a"yellow phase"in the glass,which will significantly reduce the performance of the curing body.In order to increase the solubility of sulphate in the glass curing body,the incorporation of Ba O and V₂O₅into the glass is considered to be one of the most effective ways.However,there are still gaps in microscopic studies of Ba O and V₂O₅on glass structures.At the same time,the resistance to leaching is one of the important performance indicators of glass curing bodies for highly discharged waste,as they are subject to erosion caused by various complex factors such as groundwater during long-term deep geological disposal.Due to the complexity of the glass components and dissolution mechanisms,it is difficult for existing descriptive models to represent the complexity of the leaching behaviour of glass cured bodies,while machine learning-based data-driven models provide a better fit for the description of the leaching behaviour of glass cured bodies.To address the above issues,this paper firstly uses the gradient descent method to fit to obtain the barium-oxygen pair potential parameters.A model of borosilicate glass containing different barium concentrations was obtained by molecular dynamics methods of calculation.The reliability of the barium-oxygen pair potential parameters was verified by structural analysis of the radial distribution function,bond angle distribution function,boron coordination number and property simulations such as density and elastic modulus,and by comparison with experimentally measured data in the literature.The results show that by increasing the concentration of Ba O in the components with the same proportion of other components in the glass,Ba²⁺disrupts the connections in the glass network,resulting in a decrease in[BO₃]and an increase in[BO₄]in the glass structure;the bridge oxygen in the network structure is transformed into non-bridge and free oxygen;the density of the glass increases with increasing Ba O content,while the bulk modulus,Young’s modulus and shear modulus decrease with increasing Ba O content decreases with increasing Ba O content.The barium-oxygen pair potential reference fitted in this paper provides a good reduction in structure and density of borosilicate glasses.Based on the above studies,this paper investigates the effect of barium-vanadium ratio on the stability of borosilicate glasses by molecular dynamics methods,analysing the radial distribution function,bond angle distribution function,boron coordination number and other structures of the model and simulating the properties of density and elastic modulus.The results show that V⁴⁺and V⁵⁺exist in the glass network as vanadium-oxygen polyhedra;as the barium-vanadium ratio decreases,the links in the glass network break and the bridging oxygen decreases and the non-bridging and free oxygen increases in the glass.However,due to the ability to aggregate oxygen ions,V⁴⁺>V⁵⁺>Ba²⁺,resulting in a decrease in[BO₄]and an increase in[BO₃]in the glass.The density of the system decreases as the barium to vanadium ratio decreases,while the bulk modulus,Young’s modulus and shear modulus increase slightly as the barium to vanadium ratio decreases.To explore the potential of probing deep learning algorithm models for predicting the leaching behaviour of high discharge glass curing bodies.In this paper,artificial neural network（ANN）,convolutional neural network（CNN）and temporal convolutional network（TCN）glass leaching behaviour prediction models were trained and tested using the ALTGLASS database,and the agreement of the three models for the prediction of leaching behaviour of different elements was compared.The results showed that the TCN and ANN prediction models showed better prediction accuracy for some elements in the glass curing body,respectively,while the CNN was less suitable for the prediction of leaching behaviour of elements in the glass curing body.The TCN model showed the highest accuracy in predicting the leaching behaviour of Cr,Si,Li,B,Al,Fe and Mg,while the ANN model showed the best agreement in predicting the leaching behaviour of K,Na and Mo.