Accurately and effectively predicting ionic diffusion coefficient of cement-based materials is one of the crucial issues in the study of concrete durability. Recently, many novel prediction methods based on General Effective Media (GEM) equation have been proposed and their applicability has been verified. As a significant parameter in diffusivity models based on GEM equation, percolation exponent n characterizes the diffusion properties of micro-pore structures in cement paste. The value of percolation exponent n is affected by the shape and direction of the medium, namely the morphological characteristics of micro-pore structures in cement paste. With the hydration process, the material composition and pore structure of cement change greatly, as well as the value of percolation exponent n. The value-taking of percolation exponent n has a great influence on the accuracy of the diffusion prediction model. However, a systematic study to reveal its physical meaning and determine its value is deficient. In this paper, the strong correlation between percolation exponent n and the morphological characteristic of cement paste was verified, and quantitative expressions of percolation exponent n was given by fitting experimental data.

Reference cement complying GB8076—2008 was used to prepare 25 groups of cylindrical cement paste specimens (Φ=100 mm, h=200 mm) with water-cement ratios (w/c ratio) of 0.30/0.35/0.40/0.45/0.50 and curing periods of 28/60/90/120/180 d, respectively. There were 3 specimens in each group, 2 of which were cut in the center of specimen height (50-150 mm) to obtain 4 standard test specimens (Φ=100 mm, h=50 mm) for NEL test, and 1 of which was drilled at the center of height and section of specimen to obtain a cylinder (Φ=32 mm, h=50 mm) for low-field NMR (Nuclear Magnetic Resonance) test. Saturated saline water was saturated into specimens for NEL test using the NELD-CCM vacuum water saturation instrument at vacuum degree of 87%. The NELD-CCM540 cement chloride ion effective diffusion coefficient tester was used in NEL test, 4 specimens in a group were tested and the test data was averaged as the representative effective diffusion coefficient. To determine the pore characteristics of specimens in each group, NMR test was conducted for corresponding specimen twice. For the first time, deionized water was saturated into specimens for NMR test using NELD-CCM vacuum water saturation instrument at vacuum degree of 98%, and NMR test was performed for saturated specimen to obtain total transverse relaxation time spectra of water in pore and chemically combined water. The specimen was then dried in an oven at 110 ℃ for 12 h and placed for 3 h to reduce the sample temperature to room temperature. NMR test was performed again for dry specimen to get transverse relaxation time spectra of chemically combined water. Combined with total spectra, pore size distribution of tested specimen was available. Moreover, the porosity of specimen was obtained by establishing the quantitative relation between water volume and nuclear magnetic signal.

For all groups of specimens, the effective diffusion coefficient increases firstly and then decreases with hydration process. Considering the change of gel pores and capillary pores quantity and effective diffusion coefficient data simultaneously, it can be known that the auxiliary effect of gel pores on diffusion cannot be ignored. Capillary porosity of all specimens keeps decreasing, while the 2.5-10.0 nm gel porosity increases greatly in the first 60 d of curing and decreases at a relatively smaller rate after 60 d of curing. With the increase of w/c ratio of cement paste, the proportion of diffusion effect of capillary pores to that of gel pores increase. As the hydration process carries on, the pore characteristics of the specimens shows the following rules: the continuous pore size s_c of the pore structure system decreases for specimens with w/c ratio of 0.30/0.35/0.40/0.45/0.50, and the continuous pore size s_c of the pore structure system changes to a relatively small extent for specimens with w/c ratio of 0.50; the tortuosity factor τ of pore structure system decreases at first and then increases, which is influenced by coupling effect of hydration products formation and morphological difference between two kinds of pores.

The sample correlation matrix between percolation exponent n, continuous pore size s_c, and tortuosity factor τ reflects the strong correlation between these three parameters. Percolation exponent n is negatively correlated with continuous pore size s_c, and positively correlated with the tortuosity factor τ. The results prove the morphological meaning at physical level of percolation exponent n. With the increased hydration degree, the percolation exponent n of cement paste with different w/c ratio shows different variation: the percolation exponent n of cement paste with w/c ratio of 0.30 and 0.35 increases, the percolation exponent n of cement paste with w/c ratio of 0.40 and 0.45 decreases at the beginning and then increases, and the percolation exponent n of cement paste with w/c ratio of 0.50 decreases. The different trends are primarily due to different diffusion effect ratios between capillary pores and gel pores.

Based on the theory of micropore structure analysis, the relationship between percolation exponent n and diffusion property of cement paste is explained from the aspect of mechanism. For cement paste with different w/c, the difference in ratio between capillary porosity and gel porosity leads to the difference in change law of pore characteristic parameters and percolation exponent n. With the experimental results, the relational expression of function between percolation exponent n and hydration degree was obtained.