Carbon bonds (C‐bonds) are the highly directional noncovalent interactions between carbonyl‐oxygen acceptors and sp3‐hybridized‐carbon σ‐hole donors through n→σ* electron delocalization. We have shown the ubiquitous existence of C‐bonds in proteins with the help of careful protein structure analysis and quantum calculations, and have precisely determined C‐bond energies. The importance of conventional noncovalent interactions such as hydrogen bond (H‐bonds) and halogen bond (X‐bonds) in the structure and function of biological molecules are well established, while carbon bonds C‐bonds have still to be recognized. We have shown that C‐bonds are present in proteins, contribute enthalpically to the overall hydrophobic interaction and play a significant role in the photodissociation mechanism of myoglobin and the binding of nucleobases to proteins.
Machine-learned representations of potential energy surfaces generated in the output layer of a feedforward neural network are becoming increasingly popular. One difficulty with neural network output is that it is often unreliable in regions where training data is missing or sparse. Human-designed potentials often build in proper extrapolation behavior by choice of functional form. Because machine learning is very efficient, it is desirable to learn how to add human intelligence to machine-learned potentials in a convenient way. One example is the well-understood feature of interaction potentials that they vanish when subsystems are too far separated to interact. In this article, we present a way to add a new kind of activation function to a neural network to enforce lowdimensional constraints. In particular, the activation function depends parametrically on all of the input variables. We illustrate the use of this step by showing how it can force an interaction potential to go to zero at large subsystem separations without either inputting a specific functional form for the potential or adding data to the training set in the asymptotic region of geometries where the subsystems are separated. In the process of illustrating this, we present an improved set of potential energy surfaces for the 14 lowest 3 A′ states of O 3 . The method is more general than this example, and it may be used to add other low-dimensional knowledge or lower-level knowledge to machine-learned potentials. In addition to the O 3 example, we present a greater-generality method called parametrically managed diabatization by deep neural network (PM-DDNN) that is an improvement on our previously presented permutationally restrained diabatization by deep neural network (PR-DDNN).
The photoinduced ring-opening
reaction of 1,3-cyclohexadiene to produce 1,3,5-hexatriene is a classic
electrocyclic reaction and is also a prototype for many reactions
of biological and synthetic importance. Here, we simulate the ultrafast
nonadiabatic dynamics of the reaction in the manifold of the three
lowest valence electronic states by using extended multistate complete-active-space
second-order perturbation theory (XMS-CASPT2) combined with the curvature-driven
coherent switching with decay of mixing (κCSDM) dynamical method.
We obtain an excited-state lifetime of 79 fs, and a product quantum
yield of 40% from the 500 trajectories initiated in the S1 excited state. The obtained lifetime and quantum yield values are
very close to previously reported experimental and computed values,
showing the capability of performing a reasonable nonadiabatic ring-opening
dynamics with the κCSDM method that does not require nonadiabatic
coupling vectors, time derivatives, or diabatization. In addition,
we study the ring-opening reaction by initiating the trajectories
in the dark state S2. We also optimize the S0/S1 and S1/S2 minimum-energy conical
intersections (MECIs) by XMS-CASPT2; for S1/S2, we optimized both an inner and an outer local-minimum-energy conical
intersections (LMECIs). We provide the potential energy profile along
the ring-opening coordinate by joining selected critical points via
linear synchronous transit paths. We find the inner S1/S2 LMECI to be more crucial than the outer one.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.