RESEARCH

Structure and function of ion channels

Energy and charge transfer in exciton-polariton systems

Exciton-polaritons are quasiparticles formed when molecular transitions resonantly exchange energy with light trapped in an optical cavity. This gives rise to unique properties that are distinct from purely molecular and purely photonic systems. For example, it is known that light-harvesting properties of natural photosynthetic systems and energy transfer in organic semiconductors and dyes can be enhanced by coupling to the cavity field. We study the effect of strong light-matter coupling on the population transfer in natural and artificial light-harvesting systems such as the Fenna-Matthews-Olson complex and porphyrin nanorings with butadiyne linkers incorporated into the porphyrin rings to direct the energy flow. This multporphyrin system is of huge interest because it behaves like a natural light-harvesting system LH2 with the same timescales corresponding to intra- and inter-ring energy transfer. We also explore the fundamental properties of strong light-matter coupling such as the breaking of the parity-time symmetry and their signatures in multidimensional spectra.

Representative publications:

1. Cavity-mediated enhancement of the energy transfer in the reduced Fenna–Matthews–Olson complex, L. E. H. Rodriguez, A. Sindhu, K. J. R. Espinosa, and A. A. Kananenka, Journal of Chemical Theory and Computation 20, 7393-7403 (2024)
2. Photoexcited energy relaxation in porphyrin nanorings, K. J. R. Espinosa and A. A. Kananenka, The Journal of Physical Chemistry C 128, 14347-14356 (2024)

Machine learning for nonadiabatic quantum dynamics

Exact numerical simulations of quantum dissipative systems are out of reach for present and likely future classical computing architectures. To overcome this limitation, we pioneered the use of machine learning (ML) for propagating quantum states for arbitrarily long times without solving the full many-body problem and at a fraction of the cost of numerically exact physics-based methods. Our approach exploits the fundamental property that the information about the underlying dynamical correlations in open quantum systems is encoded at the initial stages of their evolution. This property allows us to predict long-time quantum dynamics using only short-time information. We have shown, for the first time, that a ML model trained on a set of short-time reduced density matrices of a two-level quantum system linearly coupled to a dissipative environment (the spin-boson model) can very accurately predict the long-time evolution of such system in various regimes from weak to strong system-bath coupling, high to low temperature, and in the presence of non-Markovian effects. We performed an exhaustive benchmark of over 20 ML models including neural network models (fully-connected, convolutional, bidirectional, recurrent, etc.) and kernel ridge regression. Recently, inspired by large language models such as ChatGPT and BERT, we developed a transformer-based neural network model, and showed that it outperforms all previously published neural network models for predicting the long-time dynamics of the spin-boson system.

Representative publications:

1. A comparative study of different machine learning methods for dissipative quantum dynamics, L. E. H. Rodriguez, A. Ullah, K. J. R. Espinosa, P. O. Dral, and A. A. Kananenka, Machine Learning: Science and Technology 3, 045016 (2022)
2.  Convolutional neural networks for long time dissipative quantum dynamics, L. E. H. Rodriguez and A. A. Kananenka, The Journal of Physical Chemistry Letters 12, 2476–2483 (2021)
3. A short trajectory is all you need: A transformer-based model for long-time dissipative quantum dynamics, L. E. H. Rodriguez and A. A. Kananenka, arXiv:2409.11320 (2024)

Water: structure and vibrational spectroscopy

Water is the most essential substance to the existence of life. Vibrational spectroscopy (IR, 2DIR, SFG, Raman) provides important insights into the hydrogen bonding structure and dynamics of condensed-phase water and water clusters. Interpreting experimental spectra of condensed-phase water is difficult due to the multitude of hydrogen-bonding environments. We use electronic-structure theory, MD simulations, and line shape simulations to interpret IR, Raman, and SFG spectra of liquid and interfacial water. For example, in a recent collaboration with experimentalists we used simulations to interpret weak combination bands in femtosecond stimulated Raman spectra (FSRS) of water. Some of the combination bands we identified and interpreted have not been studied before. We are also interested in water clusters as they provide a convenient platform to study hydrogen bonding in water. For example, we designed a particular (H₂O)₂₀ cluster and showed that it has the strongest hydrogen bond ever reported for neural water.

Representative publications:

1. Combinational vibration modes in H₂O/HDO/D₂O mixtures detected thanks to the superior sensitivity of femtosecond stimulated Raman scattering, M. Pastorczak, K. Duk, S. Shahab, and A. A. Kananenka, The Journal of Physical Chemistry B 127, 4843–4857 (2023)
2. Unusually strong hydrogen bond cooperativity in particular (H₂O)₂₀ clusters, A. A. Kananenka and J. L. Skinner, Physical Chemistry Chemical Physics 22, 18124–18131 (2020)

Electric fields, molecular structure and dynamics in solutions and at electrochemical interfaces

Molecular vibrations are used to characterize the local electric fields in solutions, enzymes etc. through vibrational Stark effect (VSE) spectroscopy. VSE is, however, not very sensitive in non-polar environments and gives non-monotonic frequency shifts in the presence of hydrogen bonds. In a recent collaboration with experimental ultrafast spectroscopists we have shown that electric-field induced excited-state symmetry breaking (ESSB) in octopolar molecules is a much more sensitive technique for measuring electric fields in solutions. Unlike the ground state VSE spectroscopy, ESSB-based sensing of the local electric fields is free from nonsystematic frequency shifts due to specific interactions with the probe. Additionally, we collaborate with experimentalists to study the structure and dynamics of electrolytes and water at the electrochemical interfaces. Using MD simulations and computational spectroscopy we interpret 2DIR experiments which measure the nitrile vibrational frequencies of a molecule attached to the gold electrode as a reporter of the local electrostatic environment. 2DIR line shape dynamics (spectral diffusion) report on the picosecond ion and water dynamics near the electrode which is very difficult to study using other techniques. Understanding how to control the properties of molecules near electrified interfaces and reliable characterization of interfacial electric fields are important for heterogeneous catalysis.

Representative publications:

1. Excited-state symmetry breaking is an ultrasensitive tool for probing microscopic electric fields, B. Dereka, N. Maroli, Y. M. Poronik, D. T. Gryko, and A. A. Kananenka, Chemical Science 15, 15565-15576 (2024)