Quantification of magnetic resonance relaxation of fluids in tight rocks via molecular dynamics simulations
Nuclear magnetic resonance (NMR) measurements are widely used to quantify storage and fluid-transport properties of rocks and their saturating fluids, including porosity, pore-size distribution, capillary pressure, wettability, and viscosity. Features such as pore geometry and paramagnetic impurities can affect the interpretation of NMR measurements, whereby it is crucial to quantify their effects on proton longitudinal and transverse relaxation times. Laboratory measurements of magnetic resonance represent a spatial average of rock sample properties from which it is challenging to isolate proton relaxation times due to individual effects. Molecular dynamics (MD) simulations provide an efficient and accurate way to quantify the effects of specific rock/fluid properties on NMR measurements. They reproduce rotational and translational motions of molecules after reaching equilibrium, which enables the calculation of proton relaxation times. This dissertation aims to implement MD simulations to investigate and quantify the effects of different pore and fluid factors on NMR relaxation times of fluids specifically contained within tight rocks (pore size under 20 nm). First, MD simulations are performed to investigate the effects of pore size, geometry, and ions in aqueous solution on NMR relaxation properties. Results indicate that NMR relaxation times of water increase with pore size but decrease with ion concentration in nanopores devoid of solid paramagnetic impurities. Presence of ions in aqueous solution, however, has a minor effect on NMR relaxation times. Furthermore, NMR relaxation times of water layers contained in different pores reveal that the effect of solid surfaces on proton relaxation in a 3 nm slit pore is 15.32% with respect to that of a 9 nm slit pore. In addition, water in spherical nanopores has markedly different NMR relaxation times from the cases of slit and cylindrical pores exhibiting the same surface-to-volume ratio. It is found that spherical pores are an adequate representation of pores in natural rocks for the assessment of NMR relaxation times. For the case of solids with paramagnetic impurities, effects due to pore size dominate that of surface properties on self-diffusion coefficients, while the effect of surface properties is significant for NMR relaxation times originating from proton–proton interactions. However, relaxation times stemming from proton–electron interactions are not sensitive to surface density oscillations caused by solid–water interactions. The effect of surface roughness on water properties was further investigated by introducing geometrical features on slit pore surfaces. Results indicate that surface roughness plays a crucial role in altering surface chemistry rather than increasing effective surface areas for relaxation times due to proton–proton interactions. However, the importance of surface roughness in proton–electron interactions lies in exposing additional paramagnetic atoms on the surface. As a result, the proton–electron relaxation time of water reaches approximately 331 ms when a 1.25 µm slit pore contains 1% of paramagnetic minerals on the surface, which is comparable to laboratory measurement of 1% goethite-coated sandstone. Molecular-dynamics-based NMR relaxation times indicate that (a) when surface relaxivity is on the order of nm/s, standard Brownstein-Tarr methods used to interpret NMR measurements can underestimate a 9 nm slit pore as being 3 nm-thick, (b) when surface relaxivity is on the order of μm/s, employing high Larmor frequencies enables more accurate NMR measurements of tight rocks, and (c) when surface relaxivity is associated with the presence of paramagnetic impurities, the Brownstein-Tarr estimation remains accurate. Overall, NMR relaxation times obtained with MD simulations will closely approximate laboratory measurements when pore size, pore geometry, ions in aqueous solution, and appropriate surface properties are included in molecular models.