Search for: All records

Nitrogen-containing heterocyclic volatile organic compounds (VOCs) are important components of wildfire emissions that are readily reactive toward nitrate radicals (NO3) during nighttime, but the oxidation mechanism and the potential formation of secondary organic aerosol (SOA) and brown carbon (BrC) are unclear. Here, NO3 oxidation of three nitrogen-containing heterocyclic VOCs, pyrrole, 1-methylyrrole (1-MP), and 2-methylpyrrole (2-MP), was investigated in chamber experiments to determine the effect of precursor structures on SOA and BrC formation. The SOA chemical compositions and the optical properties were analyzed using a suite of online and offline instrumentation. Dinitro- and trinitro-products were found to be the dominant SOA constituents from pyrrole and 2-MP, but not observed from 1-MP. Furthermore, the SOA from 2-MP and pyrrole showed strong light absorption, while that from 1-MP were mostly scattering. From these results, we propose that NO3-initiated hydrogen abstraction from the 1-position in pyrrole and 2-MP followed by radical shift and NO2 addition leads to light-absorbing nitroaromatic products. In the absence of a 1-position hydrogen, NO3 addition likely dominates the 1-MP chemistry. We also estimate that the total SOA mass and light absorption from pyrrole and 2-MP are comparable to those from phenolic VOCs and toluene in biomass burning, underscoring the importancemore »of considering nighttime oxidation of pyrrole and methylpyrroles in air quality and climate models.« less

Recent advances in deep generative models have led to immense progress in 3D shape synthesis. While existing models are able to synthesize shapes represented as voxels, point-clouds, or implicit functions, these methods only indirectly enforce the plausibility of the final 3D shape surface. Here we present a 3D shape synthesis framework (SurfGen) that directly applies adversarial training to the object surface. Our approach uses a differentiable spherical projection layer to capture and represent the explicit zero isosurface of an implicit 3D generator as functions defined on the unit sphere. By processing the spherical representation of 3D object surfaces with a spherical CNN in an adversarial setting, our generator can better learn the statistics of natural shape surfaces. We evaluate our model on large-scale shape datasets, and demonstrate that the end-to-end trained model is capable of generating high fidelity 3D shapes with diverse topology.