Introduction Carbon dioxide emissions generated during the production of cement-based materials are one of the important sources of global greenhouse gas emissions. Various technological innovations and improvement measures are introduced to reduce carbon emissions in cement industry. Among these, CO2 can be fixed in concrete to form a stable calcium carbonate by solidifying CO2 or accelerating carbonization, thereby reducing the release of CO2 into the atmosphere. This carbonization process can effectively reduce the carbon dioxide produced by concrete and improve the comprehensive performance of concrete to a certain extent. However, the carbonization process is complicated due to the diverse components of concrete materials. The existing studies indicate that the components such as calcium silicate hydrate, ettringite, and calcium hydroxide in the concrete can participate in the carbonization reaction, and different components have different effects on the formation process of calcium carbonate. Therefore, this paper investigated the nucleation and growth process of calcium carbonate in confined space of different components by a molecular dynamics simulation method to reveal the formation mechanism of calcium carbonate in confined space. In addition, the structural characteristics of each component in concrete and the mechanism of their interaction with carbon dioxide were also analyzed.Methods The adsorption models of calcium carbonate in C-S-H, AFt, CH and SiO2 confined space were established, respectively, to investigate the cluster formation process at the interface between calcium carbonate and different components of concrete. For C-S-H, the model includes C-S-H matrix and aqueous solution containing calcium carbonate. The C-S-H substrate with a Ca/Si ratio of 1.7 was used as a substrate model, and the supercell was cut along the crystal plane parallel to (001) to form the surface of the C-S-H substrate. The overall model size is 43.20 ?×45.04 ?×80.00 ?, and the pore size of the substrate is 4 nm. A total of 60 CO32-, 60 Ca2+ and 2 000 water molecules were randomly placed in the pores, and the ion concentration in the solution was 1.4 mol/L. The other three matrices have the same size as the C-S-H matrix, and the concentration of Ca2+ and CO32- in the pores is the same as that of the above C-S-H system. Also, four matrix models with confined space of 3 nm and 5 nm were established, respectively, to explore the influence of pore size on calcium carbonate. Before the formal simulation, the conjugate gradient method was used to minimize the energy, and then the isothermal isobaric ensemble (NPT) was used in the simulation process. The temperature was set to 300 K, the running time was 10 ns, and the time step was set to 1 fs. The clay force field (ClayFF) was used to simulate the C-S-H, AFt, CH, and SiO2 matrix, respectively, and the carbonate solution used a separate force field parameter. The application of the mixed force field to different components in the model could more accurately simulate the dynamic characteristics and interactions between the components.Results and discussion In the case of bonding and local structure, the C-S-H, AFt and CH matrix form a stable and strong bonding structure between the interface region and the ions in the solution. Especially in the C-S-H system, Cac and CO32- in the pores form more bonds with the ions at the interface of the matrix and have a stronger interaction. In contrast, the adsorption capacity of the ions in the pores is obviously weak, and only a small amount of ions are adsorbed due to the lack of calcium ions on the surface of the SiO2 matrix.The nucleation mechanism of calcium carbonate clusters is more consistent with the pre-nucleation theory. The whole simulation process can be divided into three stages, i.e., rapid agglomeration, cluster growth and cluster densification. The stronger interaction between the ions at the interface of the C-S-H matrix and the free ions in the solution lead to the aggregation of calcium carbonate clusters at the interface, providing more nucleation sites for the further nucleation and growth of calcium carbonate. Therefore, calcium carbonate has the optimum nucleation and agglomeration effect in the C-S-H matrix. Also, the pore size of the confined space has a great influence on the formation process of calcium carbonate. When the pore size is 3 nm, the surface of the C-S-H, CH and AFt matrix has a stronger attraction to the free ions in the solution, which is conducive to promoting the rapid nucleation and growth of calcium carbonate. When the pore size increases to 5 nm, the nucleation and growth process of calcium carbonate clusters slows down due to the weakened effect of the matrix interface on the ions in the solution.Conclusions The ions at the interface of the matrix had a great influence on the formation of calcium carbonate clusters. For different components of concrete, calcium carbonate and matrix surface ions attracted each other to form ionic bonds, which were then adsorbed at the interface for nucleation and growth. It was easier to bond with carbonate and calcium ions in the pores because C-S-H had more free hydroxyl groups and calcium ions at the interface, providing more nucleation sites for the agglomeration and growth of calcium carbonate at the interface. Also, the attraction of the substrate surface to the free ions in the solution was stronger when the pore size was 3 nm promoting the nucleation and growth of calcium carbonate. The nucleation mechanism of calcium carbonate clusters in the matrix pores did not follow the classical nucleation theory, but calcium ions and carbonate ions combined in the form of ionic bonds to form smaller clusters and then aggregated to grow and nucleate. In this process, it was unnecessary to overcome a large nucleation energy barrier, which was consistent with the pre-nucleation mechanism.