We show H₂O₂ is spontaneously produced from pure water by atomizing bulk water into microdroplets (1 μm to 20 µm in diameter). Production of H₂O₂, as assayed by H₂O₂-sensitve fluorescence dye peroxyfluor-1, increased with decreasing microdroplet size. Cleavage of 4-carboxyphenylboronic acid and conversion of phenylboronic acid to phenols in microdroplets further confirmed the generation of H₂O₂. The generated H₂O₂ concentration was ∼30 µM (∼1 part per million) as determined by titration with potassium titanium oxalate. Changing the spray gas to O₂ or bubbling O₂ decreased the yield of H₂O₂ in microdroplets, indicating that pure water microdroplets directly generate H₂O₂ without help from O₂ either in air surrounding the droplet or dissolved in water. We consider various possible mechanisms for H₂O₂ formation and report a number of different experiments exploring this issue. We suggest that hydroxyl radical (OH) recombination is the most likely source, in which OH is generated by loss of an electron from OH⁻ at or near the surface of the water microdroplet. This catalyst-free and voltage-free H₂O₂ production method provides innovative opportunities for green production of hydrogen peroxide. Significance Water is considered to be a stable and relatively inert molecule in bulk solution. We report an exceptional behavior of water: Water molecules are spontaneously oxidized to form hydrogen peroxide near the water−air interface of micron-sized water droplets. This process does not require any chemical reagent, catalyst, applied electric potential, or radiation. Only pure water in the form of microdroplets in air is necessary for the appearance of hydrogen peroxide. We suggest that this discovery opens various innovative opportunities including green and inexpensive production of hydrogen peroxide, green chemical synthesis, safe cleaning, and food processing.

Experiments show how product pathways can be controlled by irradiation with one or more laser beams during individual bimolecular collisions or during unimolecular decompositions. For bimolecular collisions, control has been achieved by selective excitation of reagent vibrational modes, by control of reagent approach geometry, and by control of orbital alignment. For unimolecular reactions, control has been achieved by quantum interference between different reaction pathways connecting the same initial and final states and by adjusting the temporal shape and spectral content of ultrashort, chirped pulses of radiation. These collision-control experiments deeply enrich the understanding of how chemical reactions occur.

Deep reinforcement learning was employed to optimize chemical reactions. Our model iteratively records the results of a chemical reaction and chooses new experimental conditions to improve the reaction outcome. This model outperformed a state-of-the-art blackbox optimization algorithm by using 71% fewer steps on both simulations and real reactions. Furthermore, we introduced an efficient exploration strategy by drawing the reaction conditions from certain probability distributions, which resulted in an improvement on regret from 0.062 to 0.039 compared with a deterministic policy. Combining the efficient exploration policy with accelerated microdroplet reactions, optimal reaction conditions were determined in 30 min for the four reactions considered, and a better understanding of the factors that control microdroplet reactions was reached. Moreover, our model showed a better performance after training on reactions with similar or even dissimilar underlying mechanisms, which demonstrates its learning ability.