The initial stages of cement hydration at the molecular level

The world remains far from meeting greenhouse gas reduction targets, and the cement industry is a major contributor to industrial emissions. Reducing emissions solely through efficiency improvements is insufficient; material-level strategies are needed. Ordinary Portland cement predominantly contains alite and belite, impure forms of tricalcium silicate (C3S) and dicalcium silicate (C2S), whose hydration governs strength development. Understanding their hydration mechanisms at the molecular scale is critical for designing environmentally friendly cements. Molecular simulations can isolate the intrinsic reactions by excluding impurities. Initial steps involve adsorption (molecular or dissociative) of water on C3S and C2S, followed by proton hopping and detachment of Ca ions from surfaces. Despite prior work, the general dissolution pathway of Ca ions, particularly beyond specific coordination environments in C3S, remains unclear. Therefore, this study simulates the initial hydration to elucidate the general Ca dissolution pathways for C3S and C2S and to clarify the role of interfacial chemistry in enabling Ca desorption.

Prior studies indicate that single water molecules tend to adsorb dissociatively on C3S and molecularly on C2S, while bulk water on C3S prefers molecular adsorption due to strong hydrogen bonding. C3S is generally more reactive than C2S, attributed to more weakly bound ionic oxygen in C3S. After water adsorption, proton hopping into the solid and detachment of Ca ions are key steps. However, existing computational work has examined only limited cases, such as dissolution pathways for two specific coordination Ca ions in C3S, leaving the general mechanisms unresolved. Experimental and computational evidence also suggest faster hydration of C3S than C2S and emphasize the role of surface oxygen reactivity (e.g., Bader charge analyses) and adsorption energetics in governing hydration behavior.

The study employed reactive molecular dynamics (ReaxFF) in LAMMPS to simulate unbiased initial hydration of C3S and C2S surfaces up to 40 ns at 298 K with a 0.1 fs timestep. Surfaces: M3-C3S (010) and β-C2S (100) were selected as the lowest-energy symmetric surfaces. Initial crystal structures were taken from experimental XRD; unit cells and low-index surfaces were optimized by DFT (VASP 5.4.4, PBE-GGA, energy cutoff 600 eV, energy tolerance 1.0×10^-5 eV atom^-1; forces 0.01 eV Å^-1 for bulk and 0.03 eV Å^-1 for slabs; D3 dispersion; appropriate k-point meshes). Slabs were ~43 Å thick with a 15 Å vacuum in DFT; large slab models were constructed for MD with surface thickness ~100 Å and vacuum thickness ~150 Å, replicated in-plane to avoid size effects, giving initial surface areas of 9.08 nm^2 (M3-C3S) and 10.06 nm^2 (β-C2S). The vacuum region was filled with water at 0.99 g cm^-3. ReaxFF parameters merged Ca–O/H and Si–O/H sets; energy minimization used conjugate gradient with tight tolerances; charge equilibration precision 1.0×10^-6. After NPT equilibration for 100 ps, simulations proceeded to 40 ns under temperature control (Nosé–Hoover thermostat, 50 fs coupling). No biasing forces were applied. Umbrella sampling (US) with the Colvars module was used to compute free energy profiles for the transition from Structure 2 to Structure 3 of dissolving Ca in two scenarios: H-jump (proton hopping allowed) and no H-jump (proton hopping constrained), using a reaction coordinate defined by the distance from the dissolving Ca to surrounding Ca. Twelve windows from 5.15 to 5.70 Å (0.05 Å spacing) were sampled for 100 ps each, spring constants 50 (H-jump) and 200 (no H-jump) kcal mol^-1 Å^-2; WHAM with 100 bins recovered PMFs. Post-processing used OVITO for radial distribution functions (time-averaged), Python for densities/structures, and zeo++ (Voronoi decomposition; probe radius 1.1 Å; 100,000 samples in water-removed near-surface regions) for pore size distributions. Bond cutoffs: Ca–O 2.83 Å, O–Si 2.00 Å, O–H 1.20 Å.

• C3S exhibits significantly faster and more extensive initial hydration than C2S at the molecular level. In M3-C3S (010), Ca–Ow bond count shows a second rapid increase of 3.5 new bonds nm^-2 between 1.5–6 ns, 84.21% more than β-C2S (100) over the same period (1.9 nm^-2). This second increase is linked to Ca dissolution, which is not observed in β-C2S within 40 ns.
• OW–Hab bonds track Ca–Ow formation but indicate partial water dissociation: in M3-C3S, OW–Hab increases at 0.78 new bonds nm^-2 ns^-1 (1.5–6 ns), matching Ca–Ow, while in β-C2S, OW–Hab increases by 36 bonds (0.79 nm^-2 ns^-1), about 1.9× the increase of Ca–Ow (0.41 nm^-2 ns^-1), evidencing greater molecular adsorption in β-C2S.
• Two-step Ca dissolution in C3S: (1) breaking Ca–Os and forming Ca–Ow and Os–H, producing dissolved Ca (no Ca–Os) that remain tethered via Ow to undissolved Ca (Ca_dis–Ow tridentate with “ligand teeth”); (2) a much slower, general bond-breaking process in which six-coordinated dissolved Ca (95.8% of dissolved Ca) transform reversibly from tridentate to bidentate to monodentate by converting Ca_dis–OwH (teeth) to Ca–H2Ow bonds (Structures 1→2→3). The second process begins around 0.0017 ns and proceeds over tens of ns, with stage-wise evolution: 0.0017–0.7 ns (Structure 1 dominates), 0.7–5 ns (Structures 1↔2, onset of 3), 5–20 ns (Structure 2 dominates), 20–40 ns (stable Structure 3 observed).
• Free energy (umbrella sampling): with H-jump allowed, hydroxylation of a surface O coordinating Ca at ~5.25 Å weakens Ca–O and enables a stable desorbed state at >5.55 Å with an energy barrier <1 kcal mol^-1 (~4 kJ mol^-1). Without H-jump, Ca–O remains strong; the free energy rises monotonically up to ~4 kcal mol^-1 (~16 kJ mol^-1) and no stable state is found, demonstrating the critical role of local hydroxylation and the necessity of unbiased simulations.
• In β-C2S (100), Ca remains in the first dissolution process during 40 ns: no stable detachment from Ow is observed. After a common transition involving five Ca–Ow bonds (regardless of initial coordination 5, 6, or 8), six-coordinated Ca dominate subsequent steps, progressing through representative structures (e.g., (4,1,1) at 0.5 ps–0.1 ns; (3,0,3) and (2,1,3) rising 0.1–0.6 ns; (1,1,4) after 2.3 ns). Ligand teeth are prevalent (only 0.05% without teeth in Step 4), though fewer than in C3S, consistent with higher fractions of molecularly adsorbed water and faster OW–Hab growth.
• Interfacial structure at 40 ns reveals five regions (bulk, Guggenheim interface, stern layer, diffuse layer, bulk liquid) in both systems, with a distinct stern layer and small pores (2–6 Å) forming near M3-C3S (010) but not in β-C2S (100) at 40 ns. Pores favor proton transfer in C3S.
• Surface structural evolution: In M3-C3S, the main Ca–Os RDF peak at 2.35 Å weakens and shifts slightly right by 40 ns; many secondary peaks vanish, indicating partial amorphization. β-C2S retains more ordering at 40 ns, consistent with limited dissolution.
• Experimental consistency: Faster hydration/dissolution of C3S vs C2S aligns with measured undersaturated dissolution rates at 20 °C (C3S: −74.00 μmol m^-2 s^-1; β-C2S: −16.78 μmol m^-2 s^-1) and early hydroxylation observations.

The simulations elucidate why C3S hydrates faster than C2S at the molecular level: C3S’s surface oxygen (Os) chemistry and higher reactivity of Ca–Os bonds promote water adsorption, proton transfer, and formation of ligand-teeth intermediates that mediate Ca detachment. The discovery of a two-step dissolution in C3S—initial Ca–Os bond cleavage and subsequent slow release from Ow-mediated teeth—resolves the long-standing ambiguity about Ca’s dissolution pathway and emphasizes that desorption depends on local hydroxylation. Umbrella sampling confirms that hydroxylated environments drastically lower energy barriers and create stable desorbed states, whereas suppressing proton hopping (as in non-reactive simulations) inhibits dissolution. The interfacial stratification and emergence of a stern layer and nanoscale pores in C3S indicate developing pore solution saturation and enhanced proton transport pathways that may further accelerate subsequent hydration. The absence of stable Ca detachment and small pores in β-C2S within 40 ns underscores its lower intrinsic reactivity and stronger retention in first-step dissolution states. Agreement with experimental trends in dissolution rates and early hydroxylation lends credibility to the atomistic mechanisms revealed here.

Unbiased reactive MD up to 40 ns reveals: (1) a previously unreported second dissolution process for Ca in C3S; (2) two general Ca dissolution pathways for C3S and C2S; (3) a key ligand teeth structure that governs Ca detachment; and (4) detailed characterization of the solid/water interfacial arrangement, including aqueous layers and small pores that favor proton transfer in C3S. Hydroxylation of neighboring atoms is identified as a key enabling factor for Ca desorption, necessitating unbiased simulation approaches. Future work should validate these dissolution pathways across other low-index surfaces of C3S and C2S and extend to other Ca-based minerals, and further probe the correlation between Ca dissolution and OH presence as well as the kinetics of Ca detachment from the surface.

The authors note that additional validation is required to determine whether the elucidated dissolution processes generalize to other low-index surfaces of C2S and C3S and to other Ca-based minerals. There are inherent experimental challenges in resolving ultrafast initial dissolution mechanisms at the molecular level, complicating direct validation. Further investigation is needed on the correlation between Ca dissolution and OH presence and on the detailed detachment dynamics of Ca from the surface.