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

Light-driven rotary motors (LDRMs) convert light into directed molecular rotation and are used in areas such as nanotechnology, optogenetics, and materials science. Classical overcrowded alkene (OA) motors operate via four steps with two photochemical isomerisations interleaved by two thermally activated helix inversions (THI), which limit efficiency and operating temperature ranges. Recent efforts aim to develop photon-only motors that eliminate THI. The MTDP motor (a 2-stroke, photon-only LDRM) was designed and synthesized, with experiments indicating that both THI steps are bypassed at room temperature. Gas-phase simulations predicted high isomerisation quantum yields (Φiso ~0.9) for both photosteps; however, experiments in methanol measured a substantially lower Φiso for EP → ZP (0.25 ± 0.05), suggesting a strong solvent effect. The central research question is how solvation—particularly hydrogen-bonding interactions—impacts the photoisomerisation dynamics and quantum efficiency of MTDP, and whether chemical modification can mitigate solvent-induced efficiency losses.

Prior work on OA LDRMs established unidirectional rotation controlled by stereogenic centers but often suffered from low quantum yields and the need for THI steps. Alternative photon-only motor designs (e.g., hemithioindigo- and phosphorus-based motors) demonstrated all-photochemical cycles but with very low per-step efficiencies, requiring thousands of photons per full rotation. Design principles involving chiral centers, steric control in the fjord region, and tailoring conical intersection topographies have been proposed. The MTDP motor introduced a simplified 2-stroke photon-only cycle attributed to a strained MCP stator and reduced fjord steric repulsion. Earlier gas-phase NAMD predicted high Φiso and perfect unidirectionality, while limited QM/MM simulations and experiments in methanol indicated reduced Φiso and longer excited-state lifetimes, implicating solvent effects. Methodological advances such as SSR (SI-SA-REKS) for excited states and SHXF surface hopping have been validated in LDRMs and retinal systems, motivating their use here.

The study employed multiscale QM/MM nonadiabatic molecular dynamics (NAMD) to simulate the two photochemical steps (EP → ZP and ZP → EP) of the MTDP motor in methanol and its N-methylated derivative (MMTDP). Electronic structure: SI-SA-REKS (SSR) ensemble DFT with BH&HLYP functional and 6-31G* basis as implemented in a local GAMESS-US (2018.v3) build with analytic derivatives. Dynamics: Surface Hopping from Exact Factorization (SHXF) implemented in pyUNIXMD, incorporating decoherence via nuclear quantum momentum from exact factorization. Environment: QM/MM coupling to TINKER v6.3 with OPLS-AA force field for methanol; simulations used a finite methanol droplet (906 MeOH molecules) around the solute. Initialization: For each photostep, 30 initial conditions were sampled (see Supplementary Note 3), with initial dihedral distributions near gas-phase equilibria. Trajectories were propagated on S1 for up to 1.5 ps or until an S1 → S0 hop occurred, then continued on S0 until reaching a stable product conformation (up to 4–5 ps). Statistical analysis: Averages and uncertainties were refined by bootstrapping with 10^4 replicas to obtain margins of error. Additional analyses included tracking the central torsion angle θ, an auxiliary dihedral η to capture helicity (P/M), S1 population decay fitting to monoexponential forms with latency t0 and decay constant τ, and quantifying shortest hydrogen-bond distances between rotor H-bonding sites (carbonyl O and amino H1) and methanol at hopping geometries. For MMTDP, identical protocols were used to assess changes in lifetimes, directionality, Φiso, and solvent interactions.

• Solvent significantly reduces quantum efficiency and slightly alters directionality in MTDP compared to gas phase. In methanol: EP → ZP exhibits Φiso = 0.33 ± 0.09 (exp: 0.25 ± 0.05) with S1 lifetime τs = 485 ± 71 fs and unidirectionality 0.92 ± 0.05; ZP → EP shows Φiso = 0.43 ± 0.09, τs = 341 ± 28 fs, unidirectionality 0.97 ± 0.03. Gas-phase values were Φiso 0.87 ± 0.05 (EP → ZP) and 0.91 ± 0.04 (ZP → EP) with shorter S1 lifetimes (413 ± 13 fs and 298 ± 9 fs) and perfect unidirectionality. • Hydrogen bonding with methanol correlates with unproductive trajectories. At S1 → S0 hops for EP → ZP, MeOH···Ocarbonyl distances are shorter in unproductive trajectories (2.066 ± 0.519 Å) versus productive ones (2.620 ± 1.094 Å), with narrower distributions in unproductive cases. N–H (H1)···OMeOH distances show smaller differences (unproductive 2.415 ± 0.511 Å; productive 2.577 ± 0.475 Å), implying lesser impact in this step. For ZP → EP, H1···O distances differ more strongly (unproductive 2.323 ± 0.472 Å; productive 2.634 ± 0.804 Å), suggesting these H-bonds hinder productive motion; MeOH···Ocarbonyl distances also shorten in unproductive cases (1.924 ± 0.273 Å vs 2.184 ± 0.513 Å) but with smaller effect. • Chemical modification (MMTDP, N-methylation at N1) reduces H-bonding capability at the amino site and increases Φiso in ZP → EP from 0.43 ± 0.09 to 0.60 ± 0.09, while EP → ZP remains at 0.33 ± 0.09. The unidirectionality in EP → ZP changes slightly to 0.90 ± 0.05. • MMTDP exhibits longer S1 population decay (especially EP → ZP) and longer times to reach products (up to 5 ps), consistent with increased rotational friction due to the methyl group. • In MMTDP, the influence of MeOH···Ocarbonyl interactions on productivity diminishes: at hops, productive vs unproductive shortest MeOH···O distances become more similar (EP → ZP: 2.543 ± 0.999 Å vs 2.616 ± 0.745 Å; ZP → EP: 2.345 ± 0.837 Å vs 2.315 ± 0.592 Å). A competing intramolecular C–H···O interaction (between Ocarbonyl and methyl hydrogens) in S1 is hypothesized to modestly oppose rotation and contribute to increased latency, requiring further study.

The simulations reconcile experimental observations of reduced quantum yield and longer excited-state lifetimes in methanol with mechanistic insight: persistent hydrogen bonds between methanol and the rotor's H-bonding sites (carbonyl oxygen and amino hydrogen) impede rotor motion near conical intersection regions, biasing trajectories to return to reactants. The correlation between shorter solvent–solute H-bond distances and unproductive outcomes supports a causal role of H-bond pinning in lowering Φiso. Targeted chemical modification that suppresses solvent H-bonding at the amino site (N-methylation) selectively improves the ZP → EP step’s quantum efficiency without degrading overall operation, illustrating a practical design handle to mitigate solvent friction. Slight reductions in unidirectionality and increased latency/friction in the modified motor indicate trade-offs that can be balanced in future designs. Overall, the findings extend design rules for photon-only 2-stroke LDRMs by explicitly accounting for solvent-mediated interactions, suggesting that minimizing protic H-bonding environments or blocking H-bonding sites can enhance performance in solution.

The work identifies solvent hydrogen bonding as a key factor reducing photoisomerisation quantum efficiency of the photon-only MTDP motor in methanol. By methylating the amino nitrogen (MMTDP), the ability to form solvent H-bonds is reduced, leading to a ~40% increase in Φiso for the ZP → EP step (0.43 → 0.60) while maintaining the EP → ZP Φiso. The study proposes extended engineering rules: (1) ensure axial rotation by tuning heterolytic vs homolytic axle bond breaking to access favorable conical intersection topographies; (2) eliminate THI steps by introducing a chiral strained unit and reducing fjord steric strain; (3) minimize solvent H-bonding to the motor (by molecular design and/or solvent choice). The modification may also provide a practical tethering point (“harness”) for device integration without compromising efficiency. Future work should explore aprotic solvents (e.g., acetonitrile), further analyze frictional effects and the origin of the lower EP → ZP Φiso in MMTDP, and expand statistical sampling for more precise estimates.

• Finite sampling: 30 trajectories per photostep limit statistical power; bootstrapping mitigates but does not eliminate sampling uncertainty. • Finite solvent model: a methanol droplet (906 molecules) necessitated restricting simulation times (≤ 4–5 ps) to avoid boundary effects, potentially affecting long-time energy dissipation. • Methodological constraints: results rely on SSR with BH&HLYP/6-31G* within QM/MM and SHXF; while validated for similar systems, functional-dependent limitations are noted in related protonated Schiff base studies. • Solvent scope: only methanol (polar protic) was investigated; results may differ in aprotic or less H-bonding solvents. • Mechanistic ambiguities remain: the precise cause of unchanged EP → ZP Φiso in MMTDP and the impact of intramolecular C–H···O interactions require further simulations; slight losses in unidirectionality and the role of initial solvent-imparted momenta are not fully resolved.