Molecular Dynamics Simulation Research of the Effect of K+ Ions on the Microstructure of Al2O3–SiO2–K2O Silicate Melts Authors: Mingwei Ma, Fengqi Wang, Lichao Guo, Jun Liu, Zhonglin Zhang, Xiaogang Hao, Muhammad Faizan, Shulin Luo, Guixuan Wu Published: 2026 ABSTRACT Understanding the structural and dynamic properties of Al2O3–SiO2–K2O melts is essential for optimizing their performance in high-temperature industrial applications. In this study, molecular dynamics simulations were employed to study the effect of the K2O content on the microstructure of the Al2O3–SiO2–K2O melt, with a focus on viscosity variations and the distribution of ionic associations. The results reveal that Si–O predominantly exists in a stable tetrahedral configuration. The increase of K2O content increases the coordination number of Al–Si while decreasing that of K–Si, in line with the computed distribution of associations. The K2O content has a minimal effect on the O–Si–O bond angle distribution, suggesting that K+ does not disrupt the [SiO4]4– tetrahedral structure. However, the average Si–O–Si bond angle initially decreases and then increases, indicating a transition from tightly packed to more loosely arranged [SiO4]4– tetrahedra. This structural transition explains the observed viscosity trend, which first increases and then decreases. Furthermore, ion diffusion force analysis reveals that the self-diffusion coefficient of Si particles decreases and then increases with the K2O content, further supporting the viscosity results. Introduction The study of silicate melt is of great significance in industrial processes such as coal gasification, metallurgy, glass manufacturing, and ceramics processing. , In coal gasification furnaces, aluminosilicate melts flow down the gasifier wall as an inorganic molten slag. The viscosity of the slag plays a significant role in determining the optimal operating conditions of the gasifier. If the slag viscosity is too high or too low, it can severely impact the process. For one thing, the viscosity must be low enough to facilitate smooth slag removal; for another, it must be sufficiently high to reduce corrosion and protect refractory materials. Therefore, the study of slag viscosity is of great significance in optimizing the operational conditions of entrained-flow gasifiers. Based on experiments, numerous viscosity models have been developed to describe the effects of composition and temperature on the slag viscosity. However, many of these models are derived through empirical and theoretical fitting, lacking a detailed explanation for the underlying flow behaviors and microstructures [such as radial distribution function (RDF), coordination number (CN), and angular distribution function (ADF)]. In fact, the macroscopic properties of slag, such as viscosity, are fundamentally dependent on its microstructure, making it essential to study the microstructure of silicate melts. Most current methods for studying microstructure rely on characterization techniques. However, these techniques often require high-temperature quenching to obtain structural information, which may not accurately reproduce the true structure of the high-temperature melts. Hence, in situ measurements are needed. However, molecular dynamics (MD) simulations are regarded as a promising method for studying the in situ visual structure measurement of silicate melts. − MD simulations can provide detailed insights into the atomic interactions and motions in the molten state, allowing for a better understanding of how the microstructure influences the fluidity and macroscopic viscosity. Moreover, MD simulations offer an efficient and economical method to obtain microstructural information on melts, and it is helpful for guiding experimental observations for a more thorough understanding. Furthermore, MD allows fine control over composition, pressure, and temperature, which can help fill some gaps that are difficult or impossible to access experimentally. The composition and temperature of the slag have a significant impact on its viscosity. Based on the specific role of slag composition in viscosity, oxides are categorized into three groups: network formers (which increase viscosity), network modifiers (which decrease viscosity by breaking Si–O bonds), and amphoteric oxides (which can either increase or decrease viscosity depending on the composition). One common method for reducing viscosity is the addition of fluxing agents. , Among these, K2O is widely used in metallurgical slags due to its cost-effectiveness compared to other fluxing agents such as Na2O, Li2O, and B2O3. Additionally, K2O possesses unique desulfurization capabilities, which can effectively reduce the sulfur content. However, it is corrosive to refractory materials and affects the stability of molten slag. More research is needed to fully understand the effects of K2O on molten slag behavior and its suitability for industry applications. Kim et al. , studied the effects of alkali oxides on the viscosity and structure of blast furnace slag. They found that, at a fixed CaO/SiO2 ratio, the viscosity (at 1773 K) increased with the addition of K2O. This is opposite to the reported effect of Na2O on blast furnace slag viscosity. Zhang and Chou measured the viscosity of CaO–SiO2–Al2O3–K2O slags with varying K2O and Al2O3 contents. They observed that when the molar ratio of K2O/Al2O3 was less than 0.7, viscosity increased with the addition of K2O, but when the molar ratio exceeded 0.9, viscosity decreased with increasing K2O content. The viscosity peak was found in the range of 0.7–0.9. These findings suggest that K+ plays a complex role in the microstructure of the CaO–SiO2–Al2O3 melts. Both Kim , and Zhang and Chou focused on the effect of K2O on slag viscosity using experimental methods. However, structural constraints are still lacking to understand the viscosity changes. Wu et al. developed an association solution model to predict viscosity changes in the Al2O3–SiO2–K2O ternary system by introducing association species such as K2SiO3, K2Si2O5, K2Si4O9, Al6Si2O13, KAlO2, KAlSiO4, and KAlSi2O6. Their predictions align well with experimental data, but the model does not explain the microstructure. Higo et al. then employed nuclear magnetic resonance (NMR) to analyze the CaO–SiO2–Al2O3–K2O slag system and found that increasing K2O content led to the substitution of Ca2+ ions by K+ ions at charge compensator positions. This substitution increased the average bond strength within the silicate-aluminate framework and promoted aggregation of the slag network. Similarly, Kim used FT-IR spectroscopy and found that the addition of K2O increased the complexity of Si–O–Al bending, indicating its selective influence on the microstructure of aluminate slags. These studies relied on the characterization methods to analyze microstructure; however, accurately reproducing the structural information on high-temperature melts remains challenging. Subsequently, Min et al. used MDs simulations to study the effect of replacing an equimolar amount of CaO with K2O in CaO–SiO2–K2O melts. Their results showed that K+ had a more prominent modifying effect on the network structure and enhanced the ion diffusion. Li et al. simulated the effects of K2O and Na2O on the fluidity of SiO2–Al2O3–CaO melt and found that the viscosity trends of the two systems were opposite. In order to completely understand the underlying mechanism, it is necessary to study the effect of K2O on aluminosilicate melts by using MDs simulations. In this study, we investigated the effect of potassium oxide (K2O) addition on the microstructure and fluidity of Al2O3–SiO2 melts. We carried out MD simulations of Al2O3–SiO2–K2O systems with different contents of K2O. We analyzed changes in microstructural parameters (RDF, CN, and ADF) and diffusion behavior (mean square displacement and self-diffusion coefficients). Additionally, we used FactSage , software to measure the influence of viscosity and association distribution with the change of K2O/(Al2O3 + K2O). As one of the world’s largest integrated thermodynamic calculation systems, FactSage’s computational core is based on strictly evaluated experimental data and the CALPHAD (phase diagram calculation) method. Unlike MDs, which relies on atomic movement trajectories, it directly solves the phase composition and properties of the system at equilibrium through thermodynamic databases and model equations. It is completely independent of MD. FactSage and MD have fundamental differences in scale, principle, and application goals. They cannot replace each other but can complement and cross-verify each other. The former is from a macroscopic perspective, while the latter is from a microscopic perspective. Finally, we analyzed the changes in viscosity and association distribution of K2O/(Al2O3 + K2O) with microstructural information. The goal of this work is to elucidate the role of potassium ions in modifying the properties of aluminosilicate melts and to provide a deeper understanding of the underlying mechanisms driving these effects. Methods Molecular Model In the MDs simulation of silicate melts, the key to the simulation is the selection of an appropriate potential function. The empirical potential functions used in silicate simulations are primarily based on empirical data. The most widely used force field function for simulating silicate melts is the Born–Mayer–Huggins (BMH) potential, which is a pairwise potential. This function accounts for long-range Coulombic forces, short-range repulsive forces, and van der Waals forces. The expression for the BMH potential is given by 1 Uij(r)=qiqjrij+Aijexp(−Bijr)−Cijrij6 where U ij (r) is the interatomic potential, q i and q j are the standard charges, r ij is the distance between atoms i and j, A ij is the action parameter describing the long-range Coulomb force between ions, B ij is the parameter for the short-range repulsive interaction, and C ij is a parameter representing the van der Waals interaction between atoms i and j. The atomic parameter information used in this paper is shown in Table . 1BMH Potential Parameters for Pairs in Al2O3–SiO2–K2Oatom1atom2 A ij (kcal/mol) B ij (1/Å) C ij (kcal/mol·Å6)AlAl95,520.226.250AlK263,8820.126.250AlO1,984,668.5166.060AlSi69,050.233616.250KK72,922,896.676.25225.07KO4,959,4821.546.06300KSi190,7231.36.250OO34,533,8305.88399.87OSi1,448,536.46.060SiSi49,882.266.250 BMH Potential Parameters for Pairs in Al2O3–SiO2–K2O Simulation Method To study the effect of the K2O content on the microstructure and fluidity of the ternary system, the Al2O3–SiO2–K2O simulated system was divided into five groups, with each model containing approximately 6000 atoms. The composition and number of atoms for each system are given in Table . 2Composition and Atomic Count of Different Models mol. fraction (%)atomic numbermodelSiO2 Al2O3 K2OSiAlKOtotal16040094812640379260042602515102985851436026003360221810457746203561600046020201059706706353260035601822107064877835015997 Composition and Atomic Count of Different Models In this study, MD simulations were performed by using the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) software. We used a square model box with periodic boundary conditions. Long-range Coulomb forces were calculated by using the PPPM algorithm with a cutoff radius of 10 Å. The simulation was conducted under both NPT and NVT ensembles with a time step of 1 fs, and the Nose–Hoover method was used for temperature and pressure control. The simulation was divided into three parts: high-temperature relaxation, cooling, and low-temperature relaxation. For the high temperature relaxation, the NPT ensemble was used to run the system for 100 ps at 4000 K and 1 atm to ensure complete random mixing of the atoms within the box. Then the system was cooled to 1773 K at 500 ps under an NPT integration. Subsequently, the system was maintained at 1773 K and underwent low-temperature relaxation under the NPT ensemble for 500 ps. Here, the first 500 ps represents the cooling time, and the second 500 ps represents the relaxation time. Finally, the system was relaxed at the same temperature under the NVT ensemble for 200 ps. It is followed by statistical analysis of structural information such as the RDF, CN, ADF, and fluidity, like mean squared displacement (MSD). The simulation data statistics were conducted after a relaxation step of 200 ps for data analysis. The NPT used in the early stage was to balance the system to obtain the correct density, while the conversion to NVT in the later stage was to measure the microstructure changes of the silicate system under a constant volume. , The MSD and the self-diffusion coefficient (D) are two key physical quantities that describe the diffusion behavior of particles in a medium. They are closely related, and their relationship can be established by analyzing the particle trajectory data from MD simulations. The MSD is a measure of the average squared displacement of atoms over time and is related to the time-dependent behavior of particle motion. The relationship between MSD and time is given by. 2 MSD=⟨r(t)2⟩=1N⟨∑i=1N|ri(t)−ri(0)|2⟩ where r i(t) denotes the position of atom i at time t and the angle brackets denote the statistical average over multiple time intervals. Using the Stokes–Einstein relation, the self-diffusion coefficient D can be derived from the asymptotic limit of the MSD function over a long period of time, as shown in eq . 3 D=limt→∞⁡16d[Δr(t)2]dt here, d­[Δr(t)­2] refers to the mean square displacement of the ion. The diffusivity of ions in the molten slag can provide insights into the variation in its viscosity with K2O contents. The diffusivity and viscosity are, in general, inversely related through Einstein’s Kinetic Theory of Heat. For example, the Eyring relationship , shows 4 η=xkBTDλ where x is a dimensionless prefactor that accounts for the specific atomic jump mechanism and geometric factors, k B is the Boltzmann constant, and λ is the particle jump step. Results Interatomic Bond Lengths and CNs The RDF is a means to describe the bond length between ions and is used to study the distribution of other atoms around the target atom. It can be obtained through the following formula 5 gij(r)=VNiNj⟨Nij(r,Δr)⟩4πr2Δr where N i and N j represent the total number of i atoms and j atoms, respectively, V represents the system volume, and N ij (r,Δr) is the number of atom pairs within a shell volume of 4πr 2Δr. The symbols in parentheses indicate the time average. Figure shows the RDFs and CNs for various atomic pairs in model 2 (60% SiO2–25% Al2O3–15% K2O melt). The abscissa corresponding to the first peak of the RDF curve represents the average bond length between two particles, while the sharpness of the RDF curve g ij (r) reflects the stability of the molten particle structure. A narrower half-width of the highest peak in the g ij (r) curve signifies a greater concentration of j particles in the first coordination shell of i particles, indicating a higher stability of the atomic pairs structure. The first minimum value in the g ij (r) curve corresponds to the cutoff radius between atoms i and j. In the CNs curve, the y-axis of the plateau represents the average CN of the particle pair. The wider and flatter plateau of the CN curve indicates a more stable coordination structure between the particles. RDFs (a) and CNs (b) for different particle pairs in model 2 (60%SiO2–25%Al2O3–15%K2O melt). As shown in Figure a, the Si–O peak is high, narrow, and sharp, indicating that the Si–O structure is the most stable, followed by Al–O, Al–Si, and K–O, with K–Si being the least stable. The average bond lengths of Si–O, Al–O, and K–O are 1.61, 1.70, and 2.73 Å, respectively. The simulation results are in good agreement with experimental data, demonstrating the accuracy of the MD calculations. For the Si–O atomic pair, this valley occurs at 2.0 Å, indicating that the O atoms do not exist within a radius of 2.0 Å from the Si atoms, thus defining the cutoff radius for Si–O bonds. In Figure b, one can see that the average CNs for Si–O, Al–O, and Al–Si are 4.00, 3.95, and 2.41, respectively. The CN curve plateaus for Si–O and Al–O are flatter and wider than those of other atomic pairs. The broader first peak and plateau in the Al–O RDF and CN, compared to the sharp Si–O peak, indicate a more variable coordination environment for Al, consistent with its ability to act as both a network former and modifier. This contrasts with the highly stable and consistent 4-fold coordination of Si. Table presents the average bond lengths for each atomic pair across different compositions. For Si–O, Al–O, K–O, and Al–Si, the bond lengths are 1.61, 1.70, 2.72, and 3.21 Å, respectively. These values remain fixed with varying K2O content. This indicates that the presence of K2O has no effect on the local bond structures of Si–O, Al–O, and Al–Si. The formation of Si–O bonds occurs through the transition of electrons of Si from the s orbital to an unoccupied p orbital, forming sp3 hybrid orbitals. Since sp3 consists of four orbitals, and the Si–O structure contains four Si electrons (each occupying one orbital) with the remaining positions occupied by one electron from the O atom, this structure is highly stable. Therefore, the bond length of Si–O does not change with an increase in K2O content. On the other hand, K2O reacts with the bridging oxygens in the slag structure, making K a network modifier. Yet, the bond length of K–O does not change with K2O content (Table ). As the K2O content increases, the average bond length of K–Si first increases and then remains constant. This indicates that the distribution between network modifiers and the network formers becomes sparser with increasing K2O content and then eventually reaches equilibrium. 3Average Bond Length of Each Particle Pair in the ModelspairSi–OAl–OK–OAl–SiK–Si r ij /Å (K2O = 0)1.611.702.723.21 r ij /Å (K2O = 15)1.611.702.723.213.61 r ij /Å (K2O = 18)1.611.702.723.213.63 r ij /Å (K2O = 20)1.611.702.723.213.63 r ij /Å (K2O = 22)1.611.702.723.213.63 Average Bond Length of Each Particle Pair in the Models As shown in Figure , with increasing K2O content, the CN of Al–Si increases, while the CN of K–Si decreases, with the minimum CN observed in model 4 (consisting of 60% SiO2, 20% Al2O3, and 20% K2O melt). This indicates that as the K2O content increases, the number of Al atoms around the Si atoms decreases, while the number of K atoms increases. This trend is further supported by the association species distribution diagram obtained through the quasi-chemical model using FactSage software (Figure ). The “quasi-chemical model” is a type of model that describes the thermodynamic properties of solutions (such as molten salts, slags, etc.). By defining and quantifying the binding energy between these nearest atomic pairs, the model can more accurately predict the nonideal behavior of real solutions. , It can be seen that as K2O content increases, the value of K2O/(Al2O3 + K2O) increases, and the concentration of the Al6Si2O13 association species decreases, while the content of KAlSi2O6 increases. In the Al6Si2O13 association species, Si atoms are surrounded by more Al atoms, whereas in the KAlSi2O6 association species, Si atoms are surrounded by fewer Al atoms and more K atoms. This finding is in agreement with the MDs simulation analysis, where the number of Al atoms around Si atoms decreases, while the number of K atoms increases as the K2O content rises. (a) The CNs of model 1, 2, 3, 4, 5: Al–Si, (b) the CNs of model 2, 3, 4, 5: K–Si, (c) the variation of the mean CN of Al–Si with the content of K2O, and (d) the variation of the mean CN of K–Si with the content of K2O. Normalized distribution of association species in the ternary system Al2O3–SiO2–K2O at 1773 K. A more detailed analysis of Figure reveals that initially, when no K2O is added, the system primarily contains SiO2, Al2O3, and the association species Al6Si2O13 (mullite) formed by SiO2 and Al2O3. The formation of Al6Si2O13 can be understood as Al atoms disrupting the clusters between silicon–oxygen tetrahedra, leading to the formation of a new association species. Here, Al2O3 acts as a network modifier, lowering the viscosity of the system. Upon addition of K2O, the content of Al6Si2O13 significantly decreases, while the content of KAlSi2O6 increases substantially. The contents of other association species, such as K2SiO3, K2Si2O5, and K2Si4O9, remain largely unchanged. This suggests that the addition of K2O leads to an increased K+ content, which primarily compensates for the charge of Al2O3, generating [KAlO4]4– tetrahedra. These tetrahedra then cluster with [SiO4]4– tetrahedra, forming the association species KAlSi2O6. With a further increase in K2O content, Al2O3 reaches complete charge compensation, and the excess K+ acts as a network modifier, disrupting the network formers and forming association species such as K2SiO3, K2Si2O5, and K2Si4O9. Distribution of Bond Angles Figure a shows the O–Si–O bond angle distribution curve, which is smooth, indicating that there is no distortion of the Si–O polyhedra by K2O addition. The dotted lines in Figure a and Figure b refer to the bond angle of the ideal tetrahedron. The O–Si–O bond angle distribution is unimodal and symmetric, with a peak of around 109°, closely matching the ideal value of 109.5° observed in amorphous silicon. The basic overlap of the five curves in Figure a indicates that the K/Al molar ratio has a minimal impact on the distribution of the O–Si–O bond angles, further suggesting that K+ ions do not disrupt the stable [SiO4]4– tetrahedral structure. Bond angle distribution of five groups of molten slag (a) O–Si–O, (b) O–Al–O, and (c) Si–O–Si. The O–Al–O bond angle distribution (Figure b) exhibits a broad peak at approximately 110°, indicating significant fluctuations. This indicates that the O–Si–O bond angle is less likely to be distorted than the O–Al–O bond angle, implying that the Si–O network structure is more stable than the Al–O network structure. The first O–Al–O curve is notably different from the other curves, indicating the distribution for the O–Al–O bond angle in slag with 0% K2O is distinct from the other compositions. This suggests that K2O has a strong effect on the O–Al–O bond angle distribution. As seen in Figure , the introduction of K2O leads to the formation of association species such as KAlO2 and KAlSi2O6, resulting from charge compensation of Al2O3 by K+ ions. Figure c depicts that the Si–O–Si bond angles are primarily distributed between 120° and 180°. The calculated average Si–O–Si bond angles for the five groups are 169.2°, 167.8°, 166.2°, 163.0°, and 166.6°, respectively. The data reveal a trend where the average Si–O–Si bond angle gradually decreases as K2O increases from 0 to 20 wt % and then begins to increase at higher concentrations. The variation in the Si–O–Si bond angle indicates that as the K+ content increases, the distance between adjacent [SiO4]4– tetrahedra first decreases and then increases, suggesting that the clustering of [SiO4]4– tetrahedra first tightens and then loosens. This behavior corresponds to the observed trend in viscosity, which first increases and then decreases. Mean Square Displacement, Self-Diffusion Coefficient, and Viscosity The calculated MSD curves for elements in the 60% SiO2–25% Al2O3–15% K2O system are shown in Figure . The greater the slope of the MSD curve, the higher the mobility of the ion. It can be observed from Figure a that K ions exhibit the highest mobility, while Si particles show the lowest. This difference is primarily attributed to the fact that Si ions form a highly stable [SiO4]4– tetrahedral network structure with an oxidized O, which severely restricts their mobility. In contrast, K ions do not form stable structures and can move freely without being constrained by the network, resulting in higher mobility. In Figure b, the curve tends to stabilize in the later stage, indicating that the simulation is in the diffusion mode. MSD of different ions in the 60% SiO2–25% Al2O3–15% K2O system. The vertical axes in panels (a,b) use linear and logarithmic scales, respectively. Figure illustrates the effect of the K2O content on the diffusivity of Si particles in Al2O3–SiO2–K2O molten slag. As K2O content increases, the self-diffusion coefficient of Si ions initially decreases, reaching its minimum value when the K2O content is 20% (i.e., when K2O/Al2O3 equals 1), and then increases again with further K2O addition. The trend in the diffusion coefficient of Si particles can reflect, to some extent, changes in system viscosity (the change in viscosity in the Al2O3–SiO2–K2O system is fundamentally reflected in the diffusion performance of Si particles). As shown in eq , the diffusion coefficient and viscosity are inversely related. Therefore, Figure indicates that the effect of K2O addition on the system’s viscosity is not straightforward but follows a trend of first increasing and then decreasing, with the maximum viscosity occurring at the K2O/Al2O3 ratio of 1. Calculated self-diffusion coefficients of Si ions in 60% SiO2–(40 – x)% Al2O3–x% K2O (x = 0, 15, 18, 20, 22) melts. Figure shows the variation in viscosity with the Al2O3 composition in the Al2O3–SiO2–K2O system, calculated by using the quasi-chemical models implemented in FactSage software. Quasi-chemical models are usually used to describe the thermodynamic properties of solutions, including molten salts, slags, etc. By defining and quantifying the binding energy between these nearest atomic pairs, quasi-chemical models can more accurately predict the nonideal behavior of real solutions. As illustrated, with the addition of K2O, the viscosity of the system initially increases and then decreases, reaching a maximum when the K2O/(Al2O3 + K2O) ratio is approximately 0.5 (i.e., when K2O/Al2O3 is 1). This trend is consistent with the results in Figure , which further confirms the accuracy of the MD simulations. Wu et al. predicted a similar peak-shaped variation in viscosity with Al2O3/(Al2O3 + K2O) in the Al2O3–SiO2–K2O ternary system using the association solution model. They used association species such as K2SiO3, K2Si2O5, K2Si4O9, Al6Si2O13, KAlO2, KAlSiO4, and KAlSi2O6 to describe the structural changes induced by K+ and Al3+ ions. Among these species, K2SiO3, K2Si2O5, K2Si4O9, and Al6Si2O13 act as network modifiers, while KAlO2, KAlSiO4, and KAlSi2O6 serve as network formers. Variation graph of the viscosity of the SiO2–Al2O3–K2O system with K2O /(Al2O3 + K2O) calculated by the quasi-chemical model (at 1773 K). The predicted viscosity variation can be comprehensively analyzed by examining the changes in the microstructure (RDF, CN, and ADF) in Figure . Without the addition of K2O, Al2O3 acts as a network modifier, breaking the clusters between [SiO4]4– tetrahedra and forming the Al6Si2O13 association species. The formation of Al6Si2O13 reduces the viscosity of the system, making it lower than that of the pure SiO2 component. When K2O is added, K+ preferentially compensates for the charge of Al2O3 (at this stage, K+ does not act as a network modifier), forming the [KAlO4]4– tetrahedral structure. This structure then clusters with [SiO4]4– tetrahedra to form the new association species KAlSi2O6, which increases the viscosity of the system. When the K2O content reaches full charge compensation, the concentration of the KAlSi2O6 association species reaches its maximum, and the system’s viscosity also reaches its peak value. Beyond this, with further addition of K2O, the remaining K+ acts as a network modifier, breaking the clusters between [SiO4]4– tetrahedra and forming association species such as K2SiO3, K2Si2O5 and K2Si4O9. The formation of these compounds decreases the viscosity of the system. Therefore, after achieving complete charge compensation, further additions of K+ ions lead to a decrease in viscosity. Discussion We investigate the effect of K2O content on the microstructure (polymer network stability and bonding environments) and the subsequent control on the transport properties (diffusivity, viscosity) of Al2O3–SiO2–K2O melt by MDs simulations. We analyze the trends of the association species and viscosity through the CN and self-diffusion coefficients, respectively. Nevertheless, the data obtained from MD simulations can guide experimental work and lead to a more comprehensive understanding of the factors governing viscosity. Conclusions In this study, MD simulations were employed to elucidate the effects of K2O on the structure and dynamic properties of the Al2O3–SiO2–K2O melts. The viscosity and association compound distribution data were calculated using FactSage software. Based on the analysis, the following conclusions were drawn:(1)The Si–O bonds primarily exists in a stable tetrahedral form within the network structure, while the stability of Al–O, K–O, K–Si, and Al–Si is lower than that of Si–O.(2)With increasing K2O content, the CN of Al–Si increases, while that of K–Si decreases. The association species distribution, calculated by FactSage software, shows that with increasing K2O content, the number of Al atoms surrounding Si decreases, while the number of K atoms around Si increases. The trend observed in both approaches is consistent.(3)The K2O content has little effect on the distribution of O–Si–O bond angles, indicating that K+ ions do not significantly disrupt the [SiO4]4– tetrahedral structure. Additionally, the average Si–O–Si bond angle initially decreases and then increases, reaching a minimum value at model 4 (consisting of 60% SiO2, 20% Al2O3, and 20% K2O melt) composition. This suggests that with the increase in K+ content, the distance between adjacent [SiO4]4– tetrahedra first decreases and then increases, indirectly implying that the clustering of [SiO4]4– tetrahedra initially becomes denser and then loosely packed. This behavior further explains the viscosity trend of an initial increase followed by a decrease.(4)The self-diffusion coefficient of Si particles first decreases and then increases, with increasing K2O content, reaching a minimum value at model 4. At this point, the viscosity of the Al2O3–SiO2–K2O melt also reaches its maximum when K2O/(Al2O3 + K2O) ratio is 0.5. This result is consistent with the viscosity change predicted by the FactSage software, which also shows a peak in viscosity at the same composition ratio. (1)The Si–O bonds primarily exists in a stable tetrahedral form within the network structure, while the stability of Al–O, K–O, K–Si, and Al–Si is lower than that of Si–O. The Si–O bonds primarily exists in a stable tetrahedral form within the network structure, while the stability of Al–O, K–O, K–Si, and Al–Si is lower than that of Si–O. (2)With increasing K2O content, the CN of Al–Si increases, while that of K–Si decreases. The association species distribution, calculated by FactSage software, shows that with increasing K2O content, the number of Al atoms surrounding Si decreases, while the number of K atoms around Si increases. The trend observed in both approaches is consistent. With increasing K2O content, the CN of Al–Si increases, while that of K–Si decreases. The association species distribution, calculated by FactSage software, shows that with increasing K2O content, the number of Al atoms surrounding Si decreases, while the number of K atoms around Si increases. The trend observed in both approaches is consistent. (3)The K2O content has little effect on the distribution of O–Si–O bond angles, indicating that K+ ions do not significantly disrupt the [SiO4]4– tetrahedral structure. Additionally, the average Si–O–Si bond angle initially decreases and then increases, reaching a minimum value at model 4 (consisting of 60% SiO2, 20% Al2O3, and 20% K2O melt) composition. This suggests that with the increase in K+ content, the distance between adjacent [SiO4]4– tetrahedra first decreases and then increases, indirectly implying that the clustering of [SiO4]4– tetrahedra initially becomes denser and then loosely packed. This behavior further explains the viscosity trend of an initial increase followed by a decrease. The K2O content has little effect on the distribution of O–Si–O bond angles, indicating that K+ ions do not significantly disrupt the [SiO4]4– tetrahedral structure. Additionally, the average Si–O–Si bond angle initially decreases and then increases, reaching a minimum value at model 4 (consisting of 60% SiO2, 20% Al2O3, and 20% K2O melt) composition. This suggests that with the increase in K+ content, the distance between adjacent [SiO4]4– tetrahedra first decreases and then increases, indirectly implying that the clustering of [SiO4]4– tetrahedra initially becomes denser and then loosely packed. This behavior further explains the viscosity trend of an initial increase followed by a decrease. (4)The self-diffusion coefficient of Si particles first decreases and then increases, with increasing K2O content, reaching a minimum value at model 4. At this point, the viscosity of the Al2O3–SiO2–K2O melt also reaches its maximum when K2O/(Al2O3 + K2O) ratio is 0.5. This result is consistent with the viscosity change predicted by the FactSage software, which also shows a peak in viscosity at the same composition ratio. The self-diffusion coefficient of Si particles first decreases and then increases, with increasing K2O content, reaching a minimum value at model 4. At this point, the viscosity of the Al2O3–SiO2–K2O melt also reaches its maximum when K2O/(Al2O3 + K2O) ratio is 0.5. This result is consistent with the viscosity change predicted by the FactSage software, which also shows a peak in viscosity at the same composition ratio. This study not only unravels the complex relationships among K2O content, microstructure, and fluidity in Al2O3–SiO2–K2O melts but also provides valuable insights that pave the way for optimizing slag viscosity and enhancing material properties in industrial applications.