Laser flash photolyses of 2-, 3-, and 4-diazoacetylpyridines 8 give the corresponding pyridylketenes 7 formed by Wolff rearrangements, as observed by time-resolved infrared spectroscopy, with ketenyl absorptions at 2127, 2125, and 2128 cm(-1), respectively. Photolysis of 2-, 3-, and 4-8 in CH(3)CN containing n-BuNH(2) results in the formation of two transients in each case, as observed by time-resolved IR and UV spectroscopy. The initial transients are assigned as the ketenes 7, and this is confirmed by IR measurements of the decay of the ketenyl absorbance. The ketenes then form the amide enols 12, whose growth and decay are monitored by UV. Similar photolysis of diazoacetophenone leads to phenylketene (5), which forms the amide enol 17. For 3- and 4-pyridylketenes and for phenylketene, the ratios of rate constants for amination of the ketene and for conversion of the amide enol to the amide are 3.1, 7.7, and 22, respectively, while for the 2-isomer the same ratio is 1.8 x 10(7). The stability of the amide enol from 2-7 is attributed to a strong intramolecular hydrogen bond to the pyridyl nitrogen, and this is supported by the DFT calculated structures of the intermediates, which indicate this enol amide is stabilized by 12.8 kcal/mol relative to the corresponding amide enol from phenylketene. Calculations of the transition states indicate a 10.9 kcal/mol higher barrier for conversion of the 2-pyridyl amide enol to the amide as compared to that from phenylketene.

No abstract

1-Bromo-2-iodoethylene (9) was used as a central, pseudosymmetric building block for the fully convergent and modular synthesis of two related natural products, cis (1a) and trans (1b) bupleurynol. In doing so, a 9-step synthesis of 1a (reported previously) has been vastly truncated to one single operation by using queued cross-coupling reactions with Pd catalysis, negating the need for any protecting group chemistry.

The use of machine learning is becoming increasingly common in computational materials science. To build effective models of the chemistry of materials, useful machine-based representations of atoms and their compounds are required. We derive distributed representations of compounds from their chemical formulas only, via pooling operations of distributed representations of atoms. These compound representations are evaluated on ten different tasks, such as the prediction of formation energy and band gap, and are found to be competitive with existing benchmarks that make use of structure, and even superior in cases where only composition is available. Finally, we introduce an approach for learning distributed representations of atoms, named SkipAtom, which makes use of the growing information in materials structure databases.