This dataset contains example input files, training data sets and potential files related to the publication "First-principles-based Machine Learning Models for Phase Behavior and Transport Properties of CO2." by Mathur et al (2023). In this work, we developed machine learning models for CO2 based on different exchange-correlation DFT functionals. We assessed their performance on liquid densities, vapor-liquid equilibrium and transport properties.
The materials include codes and example input / output files for Monte Carlo simulations of lattice chains in the grand canonical ensemble, for determining phase behavior, critical points, and formation of aggregates.
Fractures in geological formations may enable migration of environmentally relevant fluids, as in leakage of CO2 through caprocks in geologic carbon sequestration. We investigated geochemically induced alterations of fracture geometry in Indiana Limestone specimens. Experiments were the first of their kind, with periodic high-resolution imaging using X-ray computed tomography (xCT) scanning while maintaining high pore pressure (100 bar). We studied two CO2-acidified brines having the same pH (3.3) and comparable thermodynamic
disequilibrium but different equilibrated pressures of CO2 (PCO2 values of 12 and 77 bar). High-PCO2 brine has a faster calcite dissolution kinetic rate because of the accelerating effect of carbonic acid. Contrary to expectations, dissolution extents were comparable in the two experiments. However, progressive xCT
images revealed extensive channelization for high PCO2, explained by strong positive feedback between ongoing flow and reaction. The pronounced channel increasingly directed flow to a small region of the fracture, which explains why the overall dissolution was lower than expected. Despite this, flow simulations revealed large increases in permeability in the high-PCO2 experiment. This study shows that the permeability evolution of dissolving fractures will be larger for faster-reacting fluids. The overall mechanism is not because more rock dissolves, as would be commonly assumed, but because of accelerated fracture channelization.
Geochemical and geomechanical perturbations of the subsurface caused by the injection of fluids present the risk of leakage and seismicity. This study investigated how flow of acidic fluids affects hydraulic and frictional properties of fractures using experiments with 3.8 cm-long specimens of Eagle Ford shale, a laminated shale with carbonate-rich strata. In low-pressure flow cells, one set of samples was exposed to an acidic brine and another set was exposed to a neutral brine. X-ray computed tomography and x-ray fluorescence analysis revealed that samples exposed to the acidic brine were calcite-depleted and had developed a porous altered layer, while the other set showed little evidence of alteration. After reaction, samples were compacted and sheared in a triaxial cell that supplied normal stress and differential pore pressure at prescribed sliding velocities, independently measuring friction and permeability. During the initial compaction, the porous altered layer collapsed into fine particles that filled the fracture aperture. This effectively impeded flow and sealed the fracture, resulting in a decrease in fracture permeability by 1 to 2 orders of magnitude relative to the compressed unaltered fractures. During shear, the collapsed layer of fine-grained particles prevented the formation of interlocking micro-asperities resulting in lower frictional strength. With regard to subsurface risks, this study showcases how coupled geochemical and geomechanical processes could favorably seal fractures to inhibit leakage, but also could increase the likelihood of induced seismicity. These findings have important implications for geological carbon sequestration, pressurized fluid energy storage, geothermal energy, and other subsurface technologies.
This dataset contains input and output files to reproduce the results of the manuscript "Homogeneous ice nucleation in an ab initio machine learning model" by Pablo M. Piaggi, Jack Weis, Athanassios Z. Panagiotopoulos, Pablo G. Debenedetti, and Roberto Car (arXiv preprint https://arxiv.org/abs/2203.01376). In this work, we studied the homogeneous nucleation of ice from supercooled liquid water using a machine learning model trained on ab initio energies and forces. Since nucleation takes place over times much longer than the simulation times that can be afforded using molecular dynamics simulations, we make use of the seeding technique that is based on simulating an ice cluster embedded in liquid water. The key quantity provided by the seeding technique is the size of the critical cluster (i.e., a size such that the cluster has equal probabilities of growing or shrinking at the given supersaturation). Using data from the seeding simulations and the equations of classical nucleation theory we compute nucleation rates that can be compared with experiments.
This dataset is affiliated with the publication https://doi.org/10.1007/s00348-022-03455-0. All of the data provided is necessary to reproduce the results with the aforementioned publication. The data in this repository is for the wake of a wind turbine at high Reynolds numbers. The data is mainly used for reproducing the statistics (deficit and variance profiles) and the phase averaged results.