Sicong Ma

Hydroxyl radicals (•OH) play a crucial role in the fate of redox-active substances in the environment. Studies of the •OH production in nature has been constrained to surface environments exposed to light irradiation, but is overlooked in the subsurface under dark. Results of this study demonstrate that abundant •OH is produced when subsurface sediments are oxygenated under fluctuating redox conditions at neutral pH values. The cumulative concentrations of •OH produced within 24 h upon oxygenation of 33 sediments sampled from different redox conditions are 2-670 μmol •OH per kg dry sediment or 6.7-2521 μM •OH in sediment pore water. Fe(II)-containing minerals, particularly phyllosilicates, are the predominant contributor to •OH production. This production could be sustainable when sediment Fe(II) is regenerated by the biological reduction of Fe(III) during redox cycles. Production of •OH is further evident in a field injection-extraction test through injecting oxygenated water into a 23-m depth aquifer. The •OH produced can oxidize pollutants such as arsenic and tetracycline and contribute to CO2 emissions at levels that are comparable with soil respiration. These findings indicate that oxygenation of subsurface sediments is an important source of •OH in nature that has not been previously identified, and •OH-mediated oxidation represents an overlooked process for substance transformations at the oxic/anoxic interface.

It has been documented that contaminants could be degraded by hydroxyl radicals (•OH) produced upon oxygenation of Fe(II)-bearing sediments. However, the dependence of contaminant degradation on sediment characteristics, particularly Fe(II) species, remains elusive. Here we assessed the impact of the abundance of Fe(II) species in sediments on contaminant degradation by •OH during oxygenation. Three natural sediments with different Fe(II) contents and species were oxygenated. During 10 h oxygenation of 200 g/L sediment suspension, 2 mg/L phenol was negligibly degraded for sandbeach sediment (Fe(II): 9.11 mg/g), but was degraded by 41% and 52% for lakeshore (Fe(II): 9.81 mg/g) and farmland (Fe(II): 19.05 mg/g) sediments, respectively. •OH produced from Fe(II) oxygenation was the key reactive oxidant. Sequential extractions, X-ray diffraction, Mössbauer, and X-ray absorption spectroscopy suggest that surface-adsorbed Fe(II) and mineral structural Fe(II) contributed predominantly to •OH production and phenol degradation. Control experiments with specific Fe(II) species and coordination structure analysis collectively suggest the likely rule that Fe(II) oxidation rate and its competition for •OH increase with the increase in electron-donating ability of the atoms (i.e., O) complexed to Fe(II), while the •OH yield decreases accordingly. The Fe(II) species with a moderate oxidation rate and •OH yield is most favorable for contaminant degradation.

Heterogeneous catalysis, for its industrial importance and great complexity in structure, has long been the testing ground of new characterization techniques. Machine learning (ML) as a starring tool in data science brings new opportunities for chemists to interpret, simulate, and predict complex reactions in heterogeneous catalysis. Here we review the current status of ML methods and applications in heterogeneous catalysis by following two main streams: the top-down approach by learning experiment data and the bottom-up approach for making predictions from first-principles, which differ in the data source. We focus more on the latter, where ML interacts intimately with first-principles calculations for predicting the key properties (e.g., molecular adsorption energy) and evaluating potential energy surface (PES) to expedite the atomic simulation. The ML-based PES exploration represents the top gear that can largely replace the traditional roles of first-principles calculations for structure determination and activity evaluation but requires efficient methods for data set generation, sensitive structure descriptors to discriminate structures, and iterative self-learning to refine the ML potential. We illustrate these key ingredients of ML-based atomic simulation using the SSW-NN method developed by our group as the example. Three cases of SSW-NN application are presented to elaborate how ML can expedite the material and reaction simulation and lead to new findings on catalyst structure and reaction channels. The future directions of ML-based applications in heterogeneous catalysis are also discussed.