Computational Atomistic Modeling of Cement Based Materials

1. Tobermorite unit cell

Different views (crystal cordinates direction a, b, and c) of a 3x3x3 topography supercell structure of 11 Å tobermorite (green – calcium Ca; dark yellow – silica Si).

Different views (crystal coordinates direction a, b, and c) of a 3x3x3 topography supercell structure of jennite (black – water Wa; green – calcium Ca; dark yellow – silica Si).

2. Jennite unit cell

3. Portlandite unit cell

4. Ettringite unit cell

5. Hydrocalumite unit cell

6. Kuzelite unit cell

7. Hydrogarnet unit cell

Quantification of chloride diffusion in hydrated phases: supercells for chloride diffusion analysis

Periodic boundary conditions were then applied in all three dimensions to produce models of the interfaces formed by the solid layers interspersed with 25 - 45 -Å-thick layers of aqueous solutions, thus representing a simple model of a slit-like cement nanopore. On the other hand, the thickness of the solution layer was sufficiently large to effectively exclude direct interactions between two different solution/solid interfaces created due to the periodicity of the system. The number of H2O molecules in this layer was chosen to reproduce the density of bulk aqueous solution under ambient conditions (1 g/cm3).

The MD simulation model for portlandite contains 8 × 4 × 4 crystallographic unit cells in the x-, y-, and z-directions of the Cartesian coordinate system, respectively. The following figure shown is a layer of 495 water molecules above the four layers of unit cells of portlandite. The model simulates the water-portlandite system at a room temperature of 25˚C (298 K) with a water density of 0.997 g/cm3 and portlandite density of 1.419 g/cm3. Notably, with the periodic boundary condition applied, the simulation model can be considered as consisting another four layers of unit cells of portlandite above the water molecule layer; therefore this model actually simulates a 40-Å-diameter portlandite channel filled with water.

Portlandite_8x4x4supercell (0 0 1) with a 40-Å-diameter portlandite channel filled with water

Tobermorite_2x2x1supercell (0 0 1) with a 25-Å-diameter tobermorite channel filled with 269 water molecules

Jennite_1x4x4supercell (0 0 1) with a 45-Å-diameter jennite channel filled with 541 water molecules

Modeling Study to Find the Atomic Structure of C-S-H

1) C-S-H gel formation

The C-S-H gel is the main component of the cement paste, up to the 70% in volume. It is the main responsible of the cement properties, including the strength, chemical durability and stability. If we want to improve the performance of cement materials, the C-S-H gel is the component in which we should act, modifying its structure in the desired direction. However, the C-S-H gel formation is still an open question in cement research. The amorphous nature of the gel makes vary difficult a proper characterization of its chemistry, which is variable in a wide range and dependent on hydration parameters such as the water/cement ratio, temperature, size of the cement particles, presence of additives, etc. Therefore, a better understanding of the formation from atomistic simulation could give us the clues to modify and improve cement properties.

2) Properties of the hardened paste

The cement paste is an "alive" material that evolves with time even after the main hydration process. The cement clinker phases continue dissolving at low rates, the C-S-H nanoparticles rearrange causing creep, water enters or leaves the material depending on the environment humidity and temperature causing shrinkage, chemical reactions take place due to alkaly attack, sulphate attack, etc. Within such a complex situation, the properties of the hardened paste are as important or even more than the formation of the material itself.The changes of the cement components' (C-S-H gel, Portlandite, Ettringite, Hydrocalumite, and Kuzelite) chemical-physical-mechanical properties and structure under pressure or under chemical attack are therefore of crucial importance for a durable concrete materials.

The following images provide a snapshot of the cement based materials modeling that I did:

Nanoscale Characterization of Tobermorite-1-1x4x4 Crystal Lattice with Crack Defects

Here is one phase of the cement hydrate (shown in Polyhedron style)

In real cement paste, three phases are of the main components in the hydrates.

This image shows the AFm-AFt crystal domain constructed in a layered manner.

From top to bottom, they are Ettringite, Hydrocalumite, and Kuzelite crystals.

The lattice box size is (44.667,44.667,55.149)Angstrom, (90.00ᵒ, 95.207ᵒ,120.00ᵒ) angles

The thickness of each phase is 21.35,16.29, and 26.80 Angstrom, respectvely.

Large-scale (domain size over 100nm) atomistic study of aqueous-cement hydrate paste system

Here is a system contains cement hydrate immersed in a water. The system size is (50nmX50nmX50nm) cube, where the hydrate (solid domain) size is (100.200A, 115.010A, 250.321A, 96.41ᵒ, 92.73ᵒ, 90.00ᵒ). There are 1000000 water molecules surrounding the hydrates domain to form a hydrated cement paste. Since the advanced nano-CT scanning technology can be used to connect the large-scale atomistic study scale, diffusion corrosion caused by transport of ions, e.g. Cl-, K+, Na+, SO4, and S2-, can be simulated employing this cross-scale system.

Mechanics of quasibrittle failure of cementitious materials

1) Mechanics and nano-mechanics of quasibrittle failure of fiber composites, concrete and geomaterials, including probabilistic strength and lifetime.

2) Creep and chemo-mechanical effects in concrete structures and their nano-mechanical origin (development of the chemo-mechanical algorithms)

Development of simulation methodologies

Quantuum Chemistry Based ReaxFF uses a general relationship between bond distance, bond order and bond energy that leads to a proper formation and dissociation of chemical bonds. Other bonding terms and torsional or angular potentials are adapted to the potential so they go smoothly to zero as the bond breaks. Coulombic and Van der Waals interactions are taken into account between all the atoms. The parameters are derived from quantum mechanical calculations on small systems of reference. ReaxFF can reproduce chemical reactions with a good accuracy in systems with tens of thousands atoms, far from the typical size of less than 100 atoms affordable by quantum mechanical methods. ReaxFF has a strong a very well proved performance, and it was developed by Prof. Adrian van Duin from Pennsylvania State University. The new Reaxff will be used to simulate cement hydration process, e.g. the formation of C-S-H.

DFTB is an approximate DFT scheme with a Local Combination of Atomic Orbitals (LCAO) representation of the orbitals. The Hamiltonian elements and calculated and store from previous high level quantum simulations in a system of reference, avoiding any empirical parametrization. The Hamiltonian integrals are also reduced to a two centered interactions, eliminating long range interactions as in typical tight-binding schemes. The result is a method which can give us electronic information about molecules and solids, and considerably faster than typical DFT simulations.