Search for: All records

Rh(I)‐catalyzed C8‐selective C−H alkenylation and arylation of 1,2,3,4‐tetrahydroquinolines with alkenyl and aryl carboxylic acids under microwave assistance have been realized. Using [Rh(CO)₂(acac)] as the catalyst and Piv₂O as the acid activator, 1,2,3,4‐tetrahydroquinolines undergo C8‐selective decarbonylative C−H alkenylation with a wide range of alkenyl and aryl carboxylic acids, affording the C8‐alkenylated or arylated 1,2,3,4‐tetrahydroquinolines. This method enables the synthesis of C8‐alkenylated 1,2,3,4‐tetrahydroquinolines that would otherwise be difficult to access by means of conventional C−H alkenylation protocols. Moreover, this catalytic system also works well in C8‐selective decarbonylative C−H arylation of 1,2,3,4‐tetrahydroquinolines with aryl carboxylic acids. The catalytic activity strongly depends on the choice of the N‐directing group, with the readily installable and removable N‐(2‐pyrimidyl) group being optimal. The catalytic pathway is elucidated by mechanistic experiments. 

A Rh( i )-catalyzed trideuteromethylation of heteroarenes with inexpensive and readily available deuterated acetic acid (CD 3 CO 2 D) with the aid of a N -containing directing groups is developed. The oxidant-free reaction is applicable to a wide range of heteroarene substrates, including 2-pyridones, indoles, aryl rings, pyrroles and carbazoles. It allows installation of CD 3 groups under straightforward reaction conditions. It is expected that the salient and practical features of this trideuteromethylation protocol will be of use to academic and industrial researchers. 

Transformers are expensive to train due to the quadratic time and space complexity in the self-attention mechanism. On the other hand, although kernel machines suffer from the same computation bottleneck in pairwise dot products, several approximation schemes have been successfully incorporated to considerably reduce their computational cost without sacrificing too much accuracy. In this work, we leverage the computation methods for kernel machines to alleviate the high computational cost and introduce Skyformer, which replaces the softmax structure with a Gaussian kernel to stabilize the model training and adapts the Nyström method to a non-positive semidefinite matrix to accelerate the computation. We further conduct theoretical analysis by showing that the matrix approximation error of our proposed method is small in the spectral norm. Experiments on Long Range Arena benchmark show that the proposed method is sufficient in getting comparable or even better performance than the full self-attention while requiring fewer computation resources.