-
MolNet: A Chemically Intuitive Graph Neural Network for Prediction of Molecular Properties
Authors:
Yeji Kim,
Yoonho Jeong,
Jihoo Kim,
Eok Kyun Lee,
Won June Kim,
Insung S. Choi
Abstract:
The graph neural network (GNN) has been a powerful deep-learning tool in chemistry domain, due to its close connection with molecular graphs. Most GNN models collect and update atom and molecule features from the fed atom (and, in some cases, bond) features, which are basically based on the two-dimensional (2D) graph representation of 3D molecules. Correspondingly, the adjacency matrix, containing…
▽ More
The graph neural network (GNN) has been a powerful deep-learning tool in chemistry domain, due to its close connection with molecular graphs. Most GNN models collect and update atom and molecule features from the fed atom (and, in some cases, bond) features, which are basically based on the two-dimensional (2D) graph representation of 3D molecules. Correspondingly, the adjacency matrix, containing the information on covalent bonds, or equivalent data structures (e.g., lists) have been the main core in the feature-updating processes, such as graph convolution. However, the 2D-based models do not faithfully represent 3D molecules and their physicochemical properties, exemplified by the overlooked field effect that is a "through-space" effect, not a "through-bond" effect. The GNN model proposed herein, denoted as MolNet, is chemically intuitive, accommodating the 3D non-bond information in a molecule, with a noncovalent adjacency matrix $\bf{\bar A}$, and also bond-strength information from a weighted bond matrix $\bf{B}$. The noncovalent atoms, not directly bonded to a given atom in a molecule, are identified within 5 $\unicode{x212B}$ of cut-off range for the construction of $\bf{\bar A}$, and $\bf{B}$ has edge weights of 1, 1.5, 2, and 3 for single, aromatic, double, and triple bonds, respectively. Comparative studies show that MolNet outperforms various baseline GNN models and gives a state-of-the-art performance in the classification task of BACE dataset and regression task of ESOL dataset. This work suggests a future direction of deep-learning chemistry in the construction of deep-learning models that are chemically intuitive and comparable with the existing chemistry concepts and tools.
△ Less
Submitted 1 February, 2022;
originally announced March 2022.
-
A simple table-top demonstration of radiation pressure on a macroscopic object
Authors:
G. Jesensky,
D. Dams,
O. Khomenko,
W. J. Kim
Abstract:
We report a simple demonstration of radiation pressure on a table-top experiment. Utilizing dynamic force microscopy in ambient environment, the resonant motion of a cm-sized cantilever driven by an amplitude-modulated diode laser is directly observed. Our versatile setup involves a host of exciting techniques that are relevant in precision force measurements and represents an ideal experiment in…
▽ More
We report a simple demonstration of radiation pressure on a table-top experiment. Utilizing dynamic force microscopy in ambient environment, the resonant motion of a cm-sized cantilever driven by an amplitude-modulated diode laser is directly observed. Our versatile setup involves a host of exciting techniques that are relevant in precision force measurements and represents an ideal experiment in the undergraduate laboratory.
△ Less
Submitted 27 May, 2016;
originally announced June 2016.
-
Laboratory limit on the charge of photons by electric field deflection
Authors:
A. Hankins,
C. Rackson,
W. J. Kim
Abstract:
The deflection of a laser beam traveling through a modulated electric field is measured using phase-sensitive detection to place an upper bound on photon charge. An upper limit of $10^{-14}e$ is obtained. The experiment involves a number of experimental techniques that are commonly encountered in modern precision measurements and is suitable for both advanced undergraduate and beginning graduate s…
▽ More
The deflection of a laser beam traveling through a modulated electric field is measured using phase-sensitive detection to place an upper bound on photon charge. An upper limit of $10^{-14}e$ is obtained. The experiment involves a number of experimental techniques that are commonly encountered in modern precision measurements and is suitable for both advanced undergraduate and beginning graduate students in physical science as a laboratory exercise.
△ Less
Submitted 6 September, 2012;
originally announced September 2012.
-
Observation of the thermal Casimir force
Authors:
A. O. Sushkov,
W. J. Kim,
D. A. R. Dalvit,
S. K. Lamoreaux
Abstract:
Quantum theory predicts the existence of the Casimir force between macroscopic bodies, due to the zero-point energy of electromagnetic field modes around them. This quantum fluctuation-induced force has been experimentally observed for metallic and semiconducting bodies, although the measurements to date have been unable to clearly settle the question of the correct low-frequency form of the diele…
▽ More
Quantum theory predicts the existence of the Casimir force between macroscopic bodies, due to the zero-point energy of electromagnetic field modes around them. This quantum fluctuation-induced force has been experimentally observed for metallic and semiconducting bodies, although the measurements to date have been unable to clearly settle the question of the correct low-frequency form of the dielectric constant dispersion (the Drude model or the plasma model) to be used for calculating the Casimir forces. At finite temperature a thermal Casimir force, due to thermal, rather than quantum, fluctuations of the electromagnetic field, has been theoretically predicted long ago. Here we report the experimental observation of the thermal Casimir force between two gold plates. We measured the attractive force between a flat and a spherical plate for separations between 0.7 $μ$m and 7 $μ$m. An electrostatic force caused by potential patches on the plates' surfaces is included in the analysis. The experimental results are in excellent agreement (reduced $χ^2$ of 1.04) with the Casimir force calculated using the Drude model, including the T=300 K thermal force, which dominates over the quantum fluctuation-induced force at separations greater than 3 $μ$m. The plasma model result is excluded in the measured separation range.
△ Less
Submitted 23 November, 2010;
originally announced November 2010.
