Interactive Quantum Chemistry Enabled by Machine Learning, Graphical Processing Units, and Cloud Computing.

Modern quantum chemistry algorithms are increasingly able to accurately predict molecular properties that are useful for chemists in research and education. Despite this progress, performing such calculations is currently unattainable to the wider chemistry community, as they often require domain expertise, computer programming skills, and powerful computer hardware. In this review, we outline methods to eliminate these barriers using cutting-edge technologies. We discuss the ingredients needed to create accessible platforms that can compute quantum chemistry properties in real time, including graphical processing units-accelerated quantum chemistry in the cloud, artificial intelligence-driven natural molecule input methods, and extended reality visualization. We end by highlighting a series of exciting applications that assemble these components to create uniquely interactive platforms for computing and visualizing spectra, 3D structures, molecular orbitals, and many other chemical properties.

Bringing chemical structures to life with augmented reality, machine learning, and quantum chemistry.

Visualizing 3D molecular structures is crucial to understanding and predicting their chemical behavior. However, static 2D hand-drawn skeletal structures remain the preferred method of chemical communication. Here, we combine cutting-edge technologies in augmented reality (AR), machine learning, and computational chemistry to develop MolAR, an open-source mobile application for visualizing molecules in AR directly from their hand-drawn chemical structures. Users can also visualize any molecule or protein directly from its name or protein data bank ID and compute chemical properties in real time via quantum chemistry cloud computing. MolAR provides an easily accessible platform for the scientific community to visualize and interact with 3D molecular structures in an immersive and engaging way.

Chiral photochemistry of achiral molecules.

Chirality is a molecular property governed by the topography of the potential energy surface (PES). Thermally achiral molecules interconvert rapidly when the interconversion barrier between the two enantiomers is comparable to or lower than the thermal energy, in contrast to thermally stable chiral configurations. In principle, a change in the PES topography on the excited electronic state may diminish interconversion, leading to electronically prochiral molecules that can be converted from achiral to chiral by electronic excitation. Here we report that this is the case for two prototypical examples - cis-stilbene and cis-stiff stilbene. Both systems exhibit unidirectional photoisomerization for each enantiomer as a result of their electronic prochirality. We simulate an experiment to demonstrate this effect in cis-stilbene based on its interaction with circularly polarized light. Our results highlight the drastic change in chiral behavior upon electronic excitation, opening up the possibility for asymmetric photochemistry from an effectively nonchiral starting point.

ChemPix: automated recognition of hand-drawn hydrocarbon structures using deep learning.

Inputting molecules into chemistry software, such as quantum chemistry packages, currently requires domain expertise, expensive software and/or cumbersome procedures. Leveraging recent breakthroughs in machine learning, we develop ChemPix: an offline, hand-drawn hydrocarbon structure recognition tool designed to remove these barriers. A neural image captioning approach consisting of a convolutional neural network (CNN) encoder and a long short-term memory (LSTM) decoder learned a mapping from photographs of hand-drawn hydrocarbon structures to machine-readable SMILES representations. We generated a large auxiliary training dataset, based on RDKit molecular images, by combining image augmentation, image degradation and background addition. Additionally, a small dataset of ∼600 hand-drawn hydrocarbon chemical structures was crowd-sourced using a phone web application. These datasets were used to train the image-to-SMILES neural network with the goal of maximizing the hand-drawn hydrocarbon recognition accuracy. By forming a committee of the trained neural networks where each network casts one vote for the predicted molecule, we achieved a nearly 10 percentage point improvement of the molecule recognition accuracy and were able to assign a confidence value for the prediction based on the number of agreeing votes. The ensemble model achieved an accuracy of 76% on hand-drawn hydrocarbons, increasing to 86% if the top 3 predictions were considered.

Unmasking the cis-Stilbene Phantom State via Vacuum Ultraviolet Time-Resolved Photoelectron Spectroscopy and Ab Initio Multiple Spawning.

We present the first vacuum ultraviolet time-resolved photoelectron spectroscopy (VUV-TRPES) study of photoisomerization dynamics in the paradigmatic molecule cis-stilbene. A key reaction intermediate in its dynamics, known as the phantom state, has often been invoked but never directly detected in the gas phase. We report direct spectral signatures of the phantom state in isolated cis-stilbene, observed and characterized through a combination of VUV-TRPES and ab initio multiple spawning (AIMS) nonadiabatic dynamics simulations of the channel-resolved observable. The high VUV probe photon energy tracks the complete excited-state dynamics via multiple photoionization channels, from initial excitation to its return to the "hot" ground state. The TRPES was compared with AIMS simulations of the dynamics from initial excitation, to the phantom-state intermediate (an S1 minimum), through to the ultimate electronic decay to the ground state. This combination revealed the unique spectral signatures and time-dependent dynamics of the phantom-state intermediate, permitting us to report here its direct observation.

Voice-controlled quantum chemistry.

Over the past decade, artificial intelligence has been propelled forward by advances in machine learning algorithms and computational hardware, opening up myriads of new avenues for scientific research. Nevertheless, virtual assistants and voice control have yet to be widely used in the natural sciences. Here, we present ChemVox, an interactive Amazon Alexa skill that uses speech recognition to perform quantum chemistry calculations. This new application interfaces Alexa with cloud computing and returns the results through a capable device. ChemVox paves the way to making computational chemistry routinely accessible to the wider community.

Performance of Coupled-Cluster Singles and Doubles on Modern Stream Processing Architectures.

We develop a new implementation of coupled-cluster singles and doubles (CCSD) optimized for the most recent graphical processing unit (GPU) hardware. We find that a single node with 8 NVIDIA V100 GPUs is capable of performing CCSD computations on roughly 100 atoms and 1300 basis functions in less than 1 day. Comparisons against massively parallel implementations of CCSD suggest that more than 64 CPU-based nodes (each with 16 cores) are required to match this performance.

Strong, Nonresonant Radiation Enhances Cis-Trans Photoisomerization of Stilbene in Solution.

Previously, it has been demonstrated that external electric fields may be used to exert control over chemical reactivity. In this study, the impact of a strong, nonresonant IR field (1064 nm) on the photoisomerization of cis-stilbene is investigated in cyclohexane solution. The design of a suitable reaction vessel for characterization of this effect is presented. The electric field supplied by the pulsed, near-IR radiation (εl = 4.5 × 107 V/cm) enhances the cis → trans photoisomerization yield at the red edge of the absorption spectrum (wavelengths between 337 and 340 nm). Within the microliter focal volume, up to 75% of all cis-stilbene molecules undergo isomerization to trans-stilbene in the strong electric-field environment, indicating a significant increase relative to the 35% yield of trans-stilbene under field-free conditions. This result correlates with a 1-3% enhancement in the trans-stilbene concentration throughout the bulk solution. Theoretical analysis suggests that the observed change is the result of dynamic Stark shifting of the ground and first excited states, leading to a significant redshift in cis-stilbene's absorption spectrum. The predicted increase in the absorption cross section in this range of excitation wavelengths is qualitatively consistent with the experimental increase in trans-stilbene production.

Nonadiabatic Dynamics of Photoexcited cis-Stilbene Using Ab Initio Multiple Spawning.

The photochemistry of cis-stilbene proceeds through two pathways: cis-trans isomerization and ring closure to 4a,4b-dihydrophenanthrene (DHP). Despite serving for many decades as a model system for photoisomerization, the photodynamics of cis-stilbene is still not fully understood. We use ab initio multiple spawning on a SA-2-CASSCF(2,2) potential energy surface to simulate the nonadiabatic dynamics of isolated cis-stilbene. We find the cyclization (to DHP and cis-stilbene) and isomerization (to trans- and cis-stilbene) reaction coordinates to be orthogonal; branching between the two pathways is determined on the S1 excited state within 150 fs of photoexcitation. Trajectory basis functions (TBFs) undergoing cyclization decay rapidly to the ground state in 250 fs, while TBFs moving along the isomerization coordinate remain on the excited state longer, with the majority decaying between 300 and 500 fs. We observe three avoided crossing regions in the dynamics: two along the isomerization coordinate (displaying pyramidalization and migration of an ethylenic hydrogen or phenyl group), and one DHP-like conical intersection along the cyclization coordinate. The isomeric form of the vibrationally hot photoproducts (as determined by measurement 2 ps after photoexcitation) is determined within less than 50 fs of decay to the ground state mediated by passage through a conical intersection. Excess vibrational energy of ground state cis- and trans-stilbene is channelled into phenyl torsions (with mostly opposing directionality). Our simulations are validated by direct comparison to experiment for the absorption spectrum, branching ratio of the three photoproducts (44:52:4 cis-stilbene:trans-stilbene:DHP), and excited state lifetime (520 ± 40 fs).

Optical response of electro-tuneable 3D superstructures of plasmonic nanoparticles self-assembling on transparent columnar electrodes.

Electrically tuneable, guided self-assembly of plasmonic nanoparticles (NPs) at polarized, patterned solid-liquid interfaces could enable numerous platforms for designing nanoplasmonic optical devices with new tuneable functionalities. Here, we propose a unique design of voltage-controlled guided 3D self-assembly of plasmonic NPs on transparent electrodes, patterned as columnar structures-arrays of vertical nanorods. NP assembly on the electrified surfaces of those columnar structures allows formation of a 3D superstructure of NPs, comprising stacking up of NPs in the voids between the columns, forming multiple NP-layers. A comprehensive theoretical model, based on quasi-static effective medium theory and multilayer Fresnel reflection scheme, is developed and verified against full-wave simulations for obtaining optical responses-reflectance, transmittance, and absorbance-from such systems of 3D self-assembled NPs. With a specific example of small gold nanospheres self-assembling on polarized zinc oxide columns, we show that the reflectance spectrum can be controlled by the number of stacked NP-layers. Numerical simulations show that peak reflectance can be enhanced up to ∼1.7 times, along with spectral broadening by a factor of ∼2-allowing wide-range tuning of optical reflectivity. Smaller NPs with superior mobility would be preferable over large NPs for realizing such devices for novel photonic and sensing applications.

Towards Electrotuneable Nanoplasmonic Fabry-Perot Interferometer.

Directed voltage-controlled assembly and disassembly of plasmonic nanoparticles (NPs) at electrified solid-electrolyte interfaces (SEI) offer novel opportunities for the creation of tuneable optical devices. We apply this concept to propose a fast electrotuneable, NP-based Fabry-Perot (FP) interferometer, comprising two parallel transparent electrodes in aqueous electrolyte, which form the polarizable SEI for directed assembly-disassembly of negatively charged NPs. An FP cavity between two reflective NP-monolayers assembled at such interfaces can be formed or deconstructed under positive or negative polarization of the electrodes, respectively. The inter-NP spacing may be tuned via applied potential. Since the intensity, wavelength, and linewidth of the reflectivity peak depend on the NP packing density, the transmission spectrum of the system can thus be varied. A detailed theoretical model of the system's optical response is presented, which shows excellent agreement with full-wave simulations. The tuning of the peak transmission wavelength and linewidth is investigated in detail. Design guidelines for such NP-based FP systems are established, where transmission characteristics can be electrotuned in-situ, without mechanically altering the cavity length.