Impact of solvation on the photoisomerisation dynamics of a photon-only rotary molecular motor

Light-driven rotary motors (LDRMs) are molecular-level devices that convert light energy into directed atomic motion, offering potential applications in nanotechnology, optogenetics, and other fields. Common LDRM designs utilize the overcrowded alkene (OA) motif, completing a 360° revolution through four steps: two photoisomerization steps (EP → ZM and ZP → EM) and two thermally activated helix inversion (THI) steps. Photoisomerization occurs on the S1 potential energy surface (PES), with nonadiabatic transfer to the S0 state at a conical intersection. THI steps reset the molecular conformation for continued rotation. While chemical modifications are used to optimize quantum efficiency in isolated conditions, the solvent's role is understudied. This research focuses on a 2-stroke photon-only LDRM (MTDP), eliminating THI steps and simplifying the working cycle. Gas-phase simulations predicted high isomerization quantum yield (Φiso) for MTDP, but experimental measurements in methanol showed a much lower Φiso. This discrepancy necessitates investigating the solvent's effect on LDRM efficiency at the atomic level to guide future design improvements. The study uses quantum mechanics/molecular mechanics (QM/MM) modeling and nonadiabatic molecular dynamics (NAMD) simulations to understand how solvent molecules affect MTDP's rotational function.

The literature extensively explores light-driven rotary molecular motors (LDRMs), highlighting their potential applications in various fields. Most designs leverage the overcrowded alkene (OA) motif, characterized by a four-step mechanism including two photoisomerization steps and two thermally activated helix inversion (THI) steps. Several studies have focused on improving LDRM efficiency through chemical modifications, primarily targeting the optimization of the thermally activated steps. However, the influence of the solvent environment on the photoisomerization dynamics and overall efficiency remains less explored. Recent advances have led to the design of photon-only LDRMs, aiming to eliminate the less efficient THI steps and simplify the working cycle to two photochemical steps. While these motors offer potential advantages, their overall efficiency still needs improvement. This work builds on previous research into a 2-stroke photon-only LDRM (MTDP), which demonstrated the feasibility of a simplified working cycle through time-resolved spectroscopy and gas-phase simulations. However, experimental findings revealed a significant discrepancy between the predicted and observed quantum yields in solution, prompting the current investigation to elucidate the role of solvent interactions in determining the overall motor performance.

This study employs multiscale nonadiabatic molecular dynamics (NAMD) simulations to explore the photoisomerization dynamics of the MTDP motor in methanol solution. The simulations combine quantum mechanics (QM) and molecular mechanics (MM) methods (QM/MM). The QM part utilizes the state-interaction state-averaged spin-restricted ensemble-referenced Kohn-Sham (SI-SA-REKS, or SSR) method, an ensemble density functional theory (eDFT) approach to accurately capture multiconfigurational characteristics of electronic states. The SSR method incorporates multi-reference effects, providing an improved description of systems with strong electron correlation. The surface hopping from exact factorization (SHXF) method is used for the NAMD simulations. SHXF combines electronic equations derived from the exact factorization of the electronic-nuclear wavefunction with the conventional trajectory surface hopping (TSH) formalism, seamlessly incorporating nuclear quantum momentum and achieving decoherence of nuclear trajectories. The QM calculations are performed using GAMESS-US, interfaced with the TINKER program for MM calculations using the OPLS-AA force field for solvent molecules. The NAMD simulations are conducted with the pyUNIxMD program, which implements the SHXF method. For each photoisomerization step (EP → ZP and ZP → EP), 30 initial conditions were prepared, and trajectories were propagated on the S1 PES for a maximum of 1.5 ps or until a nonadiabatic transition to S0 occurred. Statistical averages were refined through bootstrapping with 10,000 replicas. The analysis includes tracking dihedral angles (θ and η) to monitor the motor's rotation, analyzing excited-state lifetimes (τs), and calculating isomerization quantum yields (Φiso). Hydrogen bond distances between MTDP and methanol molecules were analyzed to assess the role of solvent interactions in influencing the motor's efficiency. A methylated derivative of MTDP (MMTDP) was computationally investigated to assess the impact of reducing hydrogen bonding.

Gas-phase NAMD simulations of MTDP yielded rapid decay dynamics with short S1 lifetimes (413 ± 13 fs for EP → ZP and 298 ± 9 fs for ZP → EP) and high unidirectionality. However, methanol solution simulations revealed a significant decrease in Φiso (0.33 ± 0.09 for EP → ZP and 0.43 ± 0.09 for ZP → EP) compared to gas-phase values (0.87 ± 0.05 and 0.91 ± 0.04, respectively). The S1 lifetimes were also longer in solution (485 ± 71 fs for EP → ZP and 341 ± 28 fs for ZP → EP). Analysis of hydrogen bonding interactions showed that shorter distances between the motor's hydrogen-bonding sites (carbonyl oxygen and amino hydrogen) and methanol molecules were associated with unproductive trajectories. Specifically, MeOH...O distances were shorter in unproductive trajectories for the EP → ZP step, while H1...O distances were more influential for the ZP → EP step. To mitigate the effect of hydrogen bonding, a methylated derivative (MMTDP) was simulated. The simulations of MMTDP showed no change in Φiso for the EP → ZP step (0.33 ± 0.09) but a significant increase in Φiso for the ZP → EP step (0.60 ± 0.09). This increase is attributed to the elimination of hydrogen bonding at the amino group. While the effect of methanol on the carbonyl oxygen was reduced, other factors, such as increased rotational friction from the methyl group, may influence the dynamics. An increase in intramolecular C-H...O hydrogen bonding might also explain the increased latency times observed in the simulations of MMTDP.

The findings address the research question by demonstrating the significant impact of solvent interactions, specifically hydrogen bonding, on the photoisomerization dynamics and quantum efficiency of the MTDP motor. The substantial reduction in Φiso observed in methanol solution compared to the gas phase highlights the importance of considering solvation effects in designing and optimizing LDRMs. The successful increase in Φiso for the ZP → EP step in the methylated derivative MMTDP validates the hypothesis that reducing hydrogen-bonding capacity can enhance the motor's efficiency. This understanding offers a crucial insight into the engineering rules for improving LDRM design, extending beyond the optimization of the molecular structure itself to include considerations of the immediate chemical environment. The observed difference in the influence of hydrogen bonding on the two isomerization steps suggests a nuanced approach in designing modifications to enhance overall quantum efficiency. The increased latency times and effects of increased rotational friction indicate that further work is needed to completely optimize the function.

This study provides new design rules for efficient 2-stroke LDRMs by showing the critical role of solvent hydrogen bonding on quantum efficiency. Modifying the molecule to reduce hydrogen bonding, demonstrated through methylation of the amino nitrogen, led to a significant (40%) increase in the quantum efficiency of one photoisomerization step. This approach offers a route to improve LDRM performance in a solution phase. Future research should explore other solvents and potential intramolecular interactions to further enhance the motor's efficiency and operation under load.

The study used a limited number of trajectories (30 per isomerization step) due to computational constraints, which may affect the accuracy of the calculated statistical averages. The simulations focused on a specific solvent (methanol) and motor design. Further research should investigate different solvents and molecular architectures to generalize the findings. The timescale of the simulations was relatively short (up to 5 ps), potentially missing long-term solvent effects. The analysis of intramolecular hydrogen bonding requires more investigation to confirm the hypothesis related to latency time and efficiency.