Search for: All records

Abstract Hypergolic reactions have emerged as a new synthetic approach enabling the rapid production of a diverse set of materials at ambient conditions. While hypergolic reactions bear several similarities to the well-established flame spray pyrolysis (FSP), the former has only recently been demonstrated as a viable approach to materials synthesis. Here we demonstrate a new pathway to 2D materials using hypergolic reactions and expand the gallery of nanomaterials synthesized hypergolically. More specifically, we demonstrate that ammonia borane complex, NH₃BH₃, or 4-fluoroaniline can react hypergolically with fuming nitric acid to form hexagonal boron nitride/fluorinated carbon nanosheets, respectively. Structural and chemical features were confirmed with x-ray diffraction, infrared, Raman, XPS spectroscopies and N₂porosimetry measurements. Electron microscopy (SEM and TEM) along with atomic force microscopy (AFM) were used to characterize the morphology of the materials. Finally, we applied Hansen affinity parameters to quantify the surface/interfacial properties using their dispersibility in solvents. Of the solvents tested, ethylene glycol and ethanol exhibited the most stable dispersions of hexagonal boron nitride (h-BN). With respect to fluorinated carbon (FC) nanosheets, the suitable solvents for high stability dispersions were dimethylsulfoxide and 2-propanol. The dispersibility was quantified in terms of Hansen affinity parameters (δ_d,δ_p,δ_h) = (16.6, 8.2, 21.3) and (17.4, 10.1, 14.5) MPa^1/2for h-BN and FC, respectively. 

Neural networks have enabled learning over examples that contain thousands of dimensions. However, most of these models are limited to training and evaluating on a finite collection of points and do not consider the hypervolume in which the data resides. Any analysis of the model’s local or global behavior is therefore limited to very expensive or imprecise estimators. We propose to formulate neural networks as a composition of a bijective (flow) network followed by a learnable, separable network. This construction allows for learning (or assessing) over full hypervolumes with precise estimators at tractable computational cost via integration over the input space. We develop the necessary machinery, propose several practical integrals to use during training, and demonstrate their utility.