Abstract
Water transport properties in cement are important for the cement industry. At the nanoscale, a nondestructive experimental method, 1 H nuclear magnetic resonance [NMR] relaxometry, can be used to quantify these properties. However, recent results have proven difficult to reconcile with current understanding of cement. The purpose of this work is to use Molecular Dynamics [MD] simulations to try and better understand water in cement and hence better interpret some of the NMR data. In particular, MD simulations are used to investigate water dynamics in two sizes of nanopores in analogues of calcium-silicate-hydrate [C-S-H], which is the active phase of cement paste. These pores are gel pores (3-5 nm) and interlayer spaces (1 nm). First, a bulk water system is studied and the water diffusion coefficient and NMR relax ation times are calculated. The results are compared to literature values and used to validate the methods. Then, different C-S-H analogues based on SiO 2] α -quartz crystal, tobermorite 11 ̊ A and modified tobermorite 14 ̊ A are presented. Two different sets of interatomic poten tials are used for these model simulations: CLAY FF+SPC/E and Freeman+TIP4P. These simulations are then compared. A model called MD4 which is based on modified tobermorite 14 ̊ A and using CLAY FF+SPC/E potentials is selected for further work. The density profile of water oxygen in MD4 is used to identify four water layers with different properties in the gel pore (L1, L2, TL and B) and one water layer in the interlayer pore (IL). Diffusivity and desorption analyses are performed on water populations related to these layers. The importance of the calcium ions close to the surface is highlighted. The NMR dipolar correlation function is generated for water using data from the MD4. This function underpins relaxation analysis. These outputs are compared to Korb’s single water layer model of surface NMR relaxation. Korb’s model is not supported by the new data. However, a new relaxation model of surface relaxation that takes into account water in two layers is supported by the data. Exchange is possible between these layers and is important for diffusivity as well as relaxation. Simulations are carried out as a function of temperature and used to calculate water trans- port activation energies in bulk and in MD4. Finally, the analysis of water exchange between the interlayer and gel pores is performed. It is shown that the exchange time in simulations is ≈69000 times smaller than measured experimentally. Some possible failings in the model that would account for this are discussed.