Simulations of solvent effects on excited state dynamics of p-DAPA, a red single benzene-based fluorophore

Abstract

Fluorescence-based techniques are widely used in basic science and medicine for detecting molecules and probing chemical reactions in vitro and in vivo. Red fluorophores are especially of interest due to their potential applicability for deep-tissue and whole-body imaging with low background interference. The smallest red fluorophore is the benzene-based para-diacetylphenylenediamine (p-DAPA), which has a fluorescence quantum yield of 0.06 and high brightness. However, the effect of solvent type on the photophysical properties of p-DAPA has not been extensively explored. We have employed classical force field dynamics and ab initio Born-Oppenheimer molecular dynamics (BOMD) calculations in explicit solvent to explore the ground and excited state potential energy surfaces of p-DAPA. Our results show that p-DAPA exhibits stronger fluorescence in nonpolar solvents due to the appearance of additional nonradiative decay pathways in polar solvents. The type of solvent modulates the type and speed of decay pathways observed, while the overall character of the nonradiative state is maintained. This work provides insight into how p-DAPA might be adapted for biological imaging and new solvent environments.

I. INTRODUCTION

Organic fluorophores have attracted great interest because of their broad applicability in optoelectronic devices, biological sensing, fluorescence imaging, and disease theranostics.^1–5 The main approaches to organic fluorophore design are either construction of an extended π-conjugated system or use of a push-pull system based on an electron-donor-acceptor (D-A) framework.^6–8 Large π-conjugated systems often have limited utility in biological systems due to poor membrane permeability, hydrophobicity, low quantum yield, and the ease of intermolecular energy transfer mechanisms, which lead to quenching.^9,10 Conversely, the push-pull system is constructed using a small π-conjugated system and is mostly used as the fundamental building block for designing larger organic dyes and materials.¹¹

Single benzene fluorophores (SBFs) are a unique class of small fluorophores based on the D-A framework and have tunable photophysical properties with high potential biocompatibility.^12,13 However, SBFs in biological imaging applications are limited by poor water-solubility, low photostability, poor tissue penetration, and low fluorescence quantum yield (FQY) with narrow Stokes shifts.^14,15 Some of these characters may be addressed in a fairly straightforward manner. For example, it has been shown that redder fluorophores are better able to penetrate living tissue, and that increased FQY will generally also result in improved tissue penetration. Some SBFs have been synthesized which exhibit improved emission after crystallization, yet remain poor emitters in solution.¹⁶ This suggests that crystallization results in diminished rotational degrees of freedom in the molecule, constraining the geometry and eliminating non-radiative decay pathways. Improvements on water solubility and photostability are therefore likely to be areas of the greatest opportunity for improvements.

Although the development of small red fluorophores with larger Stokes shifts remains challenging, Kim et. al. recently discovered the smallest red emitter, para-diacetylphenylenediamine (p-DAPA), which consists of a single benzene moiety with hydrogen bond donor (HBD) and acceptor groups and has a FQY of ~ 0.06 in chloroform.¹⁵ The small size, large Stokes shift, and positioning of the HBD contribute to the red emission of p-DAPA. While p-DAPA meets several desired standards, it nonetheless exhibits low FQY in polar solvents.¹⁵

Solvent polarity can affect the photophysical properties of organic dyes, but the type, frequency, and magnitude of quenching mechanisms may vary across solvents. The current literature surrounding p-DAPA is mainly focused on the molecule in a solution of chloroform, with other solvents mentioned only briefly. Indeed, Kim et al. report on 11 solvents ranging from nonpolar (cyclohexane) to polar protic (water) and a corresponding trend in decreasing FQY with increasing polarity.¹⁵ Interestingly, while as expected cyclohexane has a higher FQY of ~ 0.16 and methanol/water have FQYs of ~ 0, DMSO’s FQY is ~ 0.03 as compared to ~ 0.07 for DCM and ~ 0.06 for chloroform. Therefore solvent polarity alone may be insufficient to explain the changes in the photophysics between solvents. Solvent effects are often approximated using the polarizable continuum model (PCM) rather than explicit solvent molecules in the system. The mechanism by which solvent polarity affects photophysics is not made clear, however, and these insights are necessary to inform future design efforts of SBFs to be smaller, redder, or more useful in biological systems.

In this work, we explore the photodynamics of p-DAPA in a range of nonpolar, polar, and protic solvents using explicit solvation models (see Figure 1) to attempt to identify solvent effects which may be missing in PCM. This is of especial importance for p-DAPA, as its FQY does not correlate directly with the dielectric constant of the solvent experimentally as mentioned above.¹⁵ We use a combination of classical molecular dynamics and hybrid quantum mechanical/molecular mechanical (QM/MM) simulations to explore the electronic structure of the ground and excited states of p-DAPA in our array of solvents. This work highlights a number of structural and dynamic features which likely contribute to quenching of p-DAPA in various solvation environments. Understanding how solvent affects p-DAPA’s electronic structure and potential non-radiative decay pathways is crucial for future rational design of p-DAPA derivatives for use in polar environments. Our findings are consistent with previous theoretical work that N-H···O bond lengths between the amine and carbonyl groups on p-DAPA form an intramolecular hydrogen bond network and that proton transfer is a high energy intersection that is inaccessible on the timescales we examined.^15,17 With the use of ab initio molecular dynamics, we find that increasing solvent polarity allows for greater freedom of geometric distortion and electron transfer events which correspond to decreasing vertical excitation energies and likely quenching. While our work cannot directly address the precise deactivation mechanism or timescale, as we are not computing nonadiabatic dynamics due to the prohibitive computational cost, we instead explore the structural and electronic degrees of freedom of importance for altering the S₀-S₁ energy gap. We find that geometric distortions without corresponding electron transfer events are transient and reversible, while distortions which include electron transfer are generally irreversible on the excited state.

FIG. 1.

(a) Structure of p-DAPA with measured dihedrals. 3D structure of p-DAPA in explicit (b) chloroform, (c) methanol, and (d) water solvent environments with solvent-solvent and solute-solvent hydrogen bonds within 3 Å of p-DAPA shown in dashed lines. (e) Solvents tested for effects on p-DAPA photophysical properties in order of increasing polarity.

II. COMPUTATIONAL METHODS

A. Classical Molecular Dynamics Sampling

In this work, we are primarily interested in Born-Oppenheimer molecular dynamics (BOMD) on the S₁ excited state surface. To achieve a well-sampled set of trajectories, we first use classical force field MD generate a sampling of solvent structure and conformations of p-DAPA. Snapshots selected from the classical trajectories were then set up for QM/MM, minimizing on S₀ with adaptive freezing to achieve a local minimum and adjust from the force field description. We then perform ground state BOMD to relax the QM/MM system and bring it up to temperature (300 K), and finally perform excited state BOMD on S₁.

To follow this protocol, we first optimized geometries of gas-phase p-DAPA which were taken from previous work.¹⁵ These input structures were re-minimized, partial charges were assigned using the R.E.S.P. charge fitting model¹⁵, and other geometric parameters were obtained from Amber GAFF2 and ff14SB force fields¹⁸ using AutoParams.¹⁹ This process was also applied to cyclohexane, dichloromethane, acetonitrile, dimethyl sulfoxide (DMSO), butane, and methanol for use as solvents. All generated force field parameters are available in the SI. For water, the TIP3P water model was used.²⁰ The solvent box parameters available in Amber were used for chloroform and water.

For each solvent tested (cyclohexane, butane, chloroform, dimethyl sulfoxide, dichloromethane, acetonitrile, methanol and water) the molecule was explicitly solvated in a 15Å solvent box for running with periodic boundary conditions. Classical MD simulations were run using the pmemd.cuda module in Amber20.^21,22 The system was minimized, equilibrated to the appropriate density for the solvent (e.g. 1 g/ml for water), and heated to 300K prior to production simulations. Production classical MD simulations were in the isothermal-isobaric (NPT) ensemble with a time step of 1 fs at a temperature of 300 °K and pressure of 1 atm for 100 ns. The SHAKE algorithm was used to constrain the hydrogens during the classical MD.²³ A Langevin thermostat with a friction parameter of δ = 5 ps⁻¹ was used to minimize temperature fluctuations.²⁴ We extracted structures from our MD trajectory at 10 ns intervals for QM/MM simulations. During the classical MD, there were very few geometric distortions of p-DAPA observed as compared to the starting structure, indicating a relatively rigid molecule. The main differences were therefore in solvent molecule organization around the solute.

B. QM/MM Simulations

For each solvent-solute system, we extracted 10 evenly spaced equilibrated snapshots from the classical molecular dynamics simulations. These were used as starting geometries for the QM/MM simulations to ensure adequate sampling of the solute-solvent configurations. Since we had 8 solvents, this resulted in 80 total frames across all systems.

Each extracted frame from the classical dynamics was reduced to the closest 1000 solvent molecules in a sphere around the p-DAPA molecule to accelerate calculation times without significantly affecting resulting electronic structure calculations. This resulted in solvent spheres with a diameter of ~15Å.

For water, an additional QM region was tested to determine whether the addition of a key water molecule closer to p-DAPA hydrogen-bonding moieties could initiate or alter the proton transfer mechanism observed in this system. This QM region included p-DAPA with one additional water in hydrogen bonding range (3Å). The QM description of the additional water did not strongly impact the resulting ground or excited state dynamics (see Figure S1).

All systems and QM calculations used density functional theory (DFT) for the ground state and TDA-TDDFT for the excited state with the ωPBE functional, the 6–31G** basis set, and a range separation parameter (ω) of 0.18 Bohr⁻¹.^25–27 The ω value was originally tuned to match the vertical excitation energy to experimental absorption spectrum values in the gas phase as a first approximation (Tables I and S1). While a value of 0.10 Bohr⁻¹ was closer to the experimental value in the gas phase, the 0.18 Bohr⁻¹ value was closer to the experimentally observed 2.5 eV in the solvated environments and was used for subsequent calculations. All electronic structure calculations were performed with a development version of TeraChem 1.9 interfaced with OpenMM7.^28–30 The nonbonded interactions between the QM and MM atoms include the field of point charges and the Lennard-Jones potential terms from the MM force field. All force field parameters were consistent with the classical MD.

TABLE I.

Comparison of vertical excitation energies.

Method	Eab (eV)	Eem (eV)	Stokes Shift
Experiment	2.57	2.00	0.57
BH&H	3.25	2.85	0.40
ωPBE	2.76	2.38	0.38

All atomic positions in each extracted frame (10 for each of the 8 solvents, 80 total) were optimized to a local energetic minimum with QM/MM on the ground state (S₀) with adaptive restraints to minimize artifacts from transitioning from the pure MM to the QM/MM description of the system.³¹ The solute and solvent were both allowed to minimize, with the adaptive restraints applying only to molecules with a gradient below the default threshold.³¹ After the system converged, it was minimized with no restraints to achieve the final local minimum. These geometries were then minimized on the first excited state (S₁) to obtain solvated S₁ minima, again according to the same protocol with a final unrestrained optimization of solvent and solute at the end.

We also performed ground and excited state QM/MM Born-Oppenheimer molecular dynamics (BOMD) for each frame. We began from the optimized ground state geometry, and then ran for 2000 fs on S₀ at 300 K. The initial optimization does alter the organization of the solvent slightly, lowering the temperature of the system. The ground state sampling prior to excited state dynamics is therefore crucial to ensure the system is sufficiently relaxed. Previous work on other solvated photoactive molecules showed minimal differences in excited state dynamics following this process as opposed to snapshots taken directly from the classical sampling.³²

The final frame from the S₀ trajectories was run on S₁ for 2000 fs to simulate the excitation event and resulting changes in dynamics. In total, this resulted in 10 ground and excited state trajectories for 8 solvent-solute systems for a total of 160 ps of ground state BOMD and 160 ps of excited state BOMD.

For both S₀ and S₁ BOMD trajectories, we used spherical boundary conditions as a half-harmonic restraining force of 10 kcal · mol⁻¹Å⁻². For all systems the p-DAPA molecule (25 atoms) was treated with QM while solvent molecules were treated with MM. SHAKE was turned off, and a timestep of 1 fs was used. The QM/MM BOMD was run in the NVE ensemble.

The final frames for four representative trajectories in chloroform were recomputed as single point energies in the gas phase and with a PCM implicit solvation model to deconvolute the effect of the explicit molecules vs. their electronic field (Figure S2). For these calculations, the explicit molecules were removed but the geometry of the solute attained for each change in state (charge reorganization, charge transfer, geometric distortion, and proton transfer) was maintained. The specific PCM model used was COSMO as implemented in TeraChem, with the dielectric constant for chloroform set to 4.81.^33,34

These results were visualized, and natural transition orbitals were visualized using Visual Molecular Dynamics (VMD) 1.9.4.³⁵

C. Benchmarking and Validation

Our method and level of theory were validated by comparing the results to experimentally observed absorption and emission values.¹⁵ p-DAPA undergoes a primarily HOMO→LUMO transition for its lowest energy excited state (S₀→S₁), presumed to be the fluorescent transition. Our benchmarks further confirm that BH&H overestimates the vertical excitation values as shown in the original theoretical work.¹⁵ We note that in the original work, the calculations were performed with the mixed-reference spin-flip formalism to more accurately describe the proton transfer conical intersections, and that this requires functionals with sufficiently large exact exchange. We then benchmarked the chloroform-p-DAPA system with ωPBE(6–31G**), tuning the range separation parameter (ω) for more precise vertical energies to match the experimental absorption as closely as possible (Supplemental Table S1). This resulted in a final vertical energy of 2.76 eV vs. 2.57 eV experimentally (+0.19 eV) and (ω = 0.18) Bohr⁻¹ with maintained correct orbital transitions for optimized structures and dynamics (see Table I). We used ω PBE/6–31G** (ω = 0.18) for both our subsequent optimizations and QM/MM dynamics. We also validated our work against recent nonadiabatic simulations of p-DAPA, where the primary deactivation mechanism observed in chloroform via dynamics was intramolecular proton transfer, similar to the original static mapping.³⁶

III. RESULTS AND DISCUSSION

A. Multiple discrete decay mechanisms are observed

p-DAPA is a symmetrical molecule, with 2 amine groups and 2 acetyl groups placed diagonally across a benzene core (Figure 1). On the ground state, p-DAPA remains essentially rigid both in the benzene core and between the functional groups, with a common but transient hydrogen bond forming between the NH₂ group hydrogens and the acetyl oxygens. The p-DAPA benzene core is highly aromatic as expected, with most of the additional electronic density in the HOMO orbital lying on the amine group. On the excited state, the hole in the hole-electron pair is essentially the same as the ground state HOMO orbital for both the optimized geometries (see Figure S3) and the BOMD dynamics (see Figure 2), and p-DAPA absorbs at 2.57 eV (482 nm). Once p-DAPA has been excited, the electron redistributes across the benzene core to become more antiaromatic, with electronic density moved from the amine groups to the acetyl groups. This essentially π to π* transition, with an extended conjugation from the acetyl groups, results in a strong Stokes shift of 0.57 eV and red fluorescence emission around 2 eV (618 nm) experimentally. We denote this as the charge rearrangement (CR) mechanism. Importantly, the initial excitation at 0 fs is essentially identicalto CR for all of the observed decay pathways, meaning the subsequent changes in orbital character are not determined by initial structure of p-DAPA alone (Figure 2).

FIG. 2.

Representative natural transition orbitals (NTOs) for four different mechanisms associated with charge reorganization (CR) and fluorescence, electron transfer (ET), proton transfer (PT), and geometric distortion (GD). Frames were taken from the excited state dynamics trajectories at 0 fs (left) and 2000 fs (right). Occupations close to 1 indicate that the hole/electron pair is described well by the shown set of orbitals for all systems.

Aside from CR, we observed three behaviors associated with decreasing excited state energy past the point of fluorescence emission (2 eV) across the polar and non-polar solvents within the 2000 fs excited state dynamics. We note that this does not preclude additional events occurring over a longer timescale.

The first and most common mechanism we observe is electron transfer (ET). In this scenario, rather than an even distribution of electrons across the benzene core and both acetyl groups, ET results in the localization of the entire electron on a single acetyl group (see Figure 2). We denote a mechanism as ET if > 0.5 electrons have transferred to one of the two acetyl groups in p-DAPA. The speed of the electron transferring varies depending on solvent and geometry, with some systems showing a near instantaneous reorganization of charge and others showing a slower redistribution of electronic density over the course of the trajectory. As a rule of thumb, the redistribution is faster in polar as opposed to nonpolar solvents.

For electron transfer to occur in the dynamics, there needs to be a symmetry-breaking geometric change in the p-DAPA molecule, which we denote geometric distortion (GD). GD involves the twisting of specific dihedral bonds on the acetyl group, which we denote φ_A and φ_B (see Figure 1a). It is possible to observe GD without ET, though the S₁ energy decreases are not as large.

Proton transfer (PT), in this case, is an intramolecular process where a proton transfers from the amine group to the acetyl group. The conical intersections associated with this pathway were obtained from prior work, though they were essentially all uphill from the S₁ minimum.¹⁵ This is consistent with our work, where PT is only observed as a reversible process in the shorter timescale dynamics.

Generally, all of the observed potentially non-radiative pathways have a consistent movement of charge towards one of the two acetyl groups and thus orbital character. Interestingly, we observed that PT and GD without ET are often transient, with decreased vertical excitations indicating a temporarily destabilized ground state at the corresponding excited state geometries rather than a substantial decrease in excited state energy. Each of the four pathways appear to varying degrees in the different solvents, with their ratios and accessibility further explained for each below.

B. Excitation leads to charge rearrangement

Consistent with previously published work, we observe that p-DAPA exhibits CR in all tested solvents upon excitation (see Figure 2). The ground- and excited state Mulliken charges were calculated from their respective optimized structures, as well as for each frame of the dynamics. To ensure reproducibility, as Mulliken charges can be unreliable, RESP charges for each decay mechanism were also computed and shown to be similar in overall trend, and quantitatively similar for most atoms (see Table S5). The partial charges for the atoms in each functional group were summed and the difference between adjacent amine/acetyl pairs was obtained, which we refer to here as “electron flow”. The electron flow in non-polar solvents was generally lower than in polar/protic solvents, and we observed the total flow across the central benzene ring to follow a similar trend. This corresponds well to previously reported experimental results, where the fluorescence quantum yield had a maximum of 16% in cyclohexane, 7% in dichloromethane, 3% in DMSO, and negligible fluorescence in water or ethanol.³⁶ We attribute the decrease in observed FQY to the corresponding increase of electron flow and ET events. The magnitude of observed electron flow provides insight into the decay mechanisms including fluorescence and quenching, as the excited state is affected by inter- and intramolecular interactions.

C. Electron transfer character increases with solvent polarity and asymmetry

In the nonpolar solvents cyclohexane and butane, we observed no ET character and only one instance of GD, which was itself transient and rapidly returned to a more planar orientation. In chloroform, we observed 4 trajectories with GD character, one of which also had ET character and another which exhibited the intermittent PT events previously discussed. By comparison, DMSO and DCM showed 4–5 trajectories with ET/GD character. In water, 9 of the 10 trajectories exhibited ET/GD character. Interestingly, acetonitrile showed some deviation from this trend with only one trajectory exhibiting ET character. We ultimately find that as solvent polarity increases, so too does the likelihood of electron transfer between functional groups (see Figures 3 and 4). As solvent is inherently disordered and heterogenous, the resulting electronic effects on p-DAPA are asymmetric as shown in the unequal charge flow between function group pairs, despite the molecule itself being symmetrical. In non-polar solvents and more homogenous solvents, we see less asymmetric charge flow between functional group pairs. However, the asymmetry is still present in nonpolar solvents and does not exactly follow the trend of solvent polarity alone. p-DAPA’s asymmetric character is also to be affected by the organization of the solvent molecules as well. The presence of the explicit solvent thus alters the accessibility of the asymmetric geometries of p-DAPA. While the MM solvent cannot repolarize in response to the excited state as a PCM description can, we do note that recomputing the excited state geometries in implicit solvent gives a very similar description of the excited state (Figure S2). This further suggests that the driving factor for the discrepancies observed experimentally for FQY for chloroform, dichloromethane and DMSO vs. their polarity is primarily geometric.

FIG. 3.

Representative trajectories for DCM (left) and DMSO (right) simulations. (Top) The dihedral angle over time for each acetyl group relative to the plane of the benzene ring was calculated using multiple atom sets (dashed lines) and averaged (solid line) to reduce noise and identify trends in dihedral rotation. (Bottom), The vertical excitation energy (S₀→S₁) calculated on S₁ over time. The dashed black line was calculated as the “maximum future energy” (see SI for full details), indicating whether reductions in bandgap were reversible over time. Vertical lines for Electron Transfer (ET) events (green) and reversals (red) are included at their respective time indices to provide further context for these time-dependent metrics. Full details in SI.

FIG. 4.

Representative trajectories for acetonitrile simulations. (Top) The dihedral angle over time for each acetyl group relative to the plane of the benzene ring was calculated using multiple atom sets (dashed lines) and averaged (solid line) to reduce noise and identify trends in dihedral rotation. (Bottom), The vertical excitation energy (S₀→S₁) calculated on S₁ over time. The dashed black line was calculated as the “maximum future energy” (see SI for full details), indicating whether reductions in bandgap were reversible over time. Vertical lines for Electron Transfer (ET) events (green) and reversals (red) are included at their respective time indices to provide further context for these time-dependent metrics. Full details in SI.

D. Electron transfer events are reversible in nonpolar solvents

We recalculated the electron flow for each timestep of the excited state simulations to identify ET events. At each frame, if the total electron flow for either functional group pairing exceeded 0.5, an ET event was declared and the Mulliken charge sums at that time were set as the new baseline for subsequent frames. After an ET event, if the total flow for a group was 0.5 in the opposite direction, an ET reversal event was declared. This cycle repeated through to the end of the simulation for each replicate trajectory of all solvents, and these time points were identified on their corresponding time-dependent metric plots (see Supplemental Information for complete data). We identified a number of ET events with rapid corresponding ET reversals in five (5) cyclohexane trajectories, each of which corresponded to temporary dihedral rotations and corresponding decreases in vertical excitation energies (see Figure 5). Interestingly, no events were observed in the butane simulations, suggesting that the increased packing of butane may be contributing to further restriction of dihedral rotation in p-DAPA and subsequently preventing access to other decay mechanisms. As solvent polarity increases, we observed multiple ET/reversal event pairs with longer times between, culminating in 9 out of 10 water simulations with ET events that have no corresponding reversals (see Figure 6).

FIG. 5.

Representative trajectories for cyclohexane (left) and butane (right) simulations. (Top) The dihedral angle over time for each acetyl group relative to the plane of the benzene ring was calculated using multiple atom sets (dashed lines) and averaged (solid line) to reduce noise and identify trends in dihedral rotation. (Bottom), The vertical excitation energy (S₀→S₁) calculated on S₁ over time. The dashed black line was calculated as the “maximum future energy” (see SI for full details), indicating whether reductions in bandgap were reversible over time. Vertical lines for Electron Transfer (ET) events (green) and reversals (red) are included at their respective time indices to provide further context for these time-dependent metrics. Full details in SI.

FIG. 6.

Representative trajectories for methanol (left) and water (right) simulations. (Top) The dihedral angle over time for each acetyl group relative to the plane of the benzene ring was calculated using multiple atom sets (dashed lines) and averaged (solid line) to reduce noise and identify trends in dihedral rotation. (Bottom), The vertical excitation energy (S₀→S₁) calculated on S₁ over time. The dashed black line was calculated as the “maximum future energy” (see SI for full details), indicating whether reductions in bandgap were reversible over time. Vertical lines for Electron Transfer (ET) events (green) and reversals (red) are included at their respective time indices to provide further context for these time-dependent metrics. Full details in SI.

E. Electron transfer initiates smaller energy gap between ground and excited state

We classified each simulation such that any system with electron flow greater than 0.5 total electrons across either functional group pair was tagged as “ET”, any system which had either acetyl group dihedral rotate past 50° was tagged as “GD”, and any system in which the amine proton was closer to the acetyl oxygen than to the nitrogen was tagged as “PT”. Nearly all decay mechanisms observed include ET character with most progressing to GD pathways. In only one instance of chloroform simulation, we observed PT character concurrent with GD, however this was observed to be temporary as the p-DAPA molecule returned to its more planar configuration and corresponding excitation energy values (see Figure S16). A small set of trajectories exhibited GD character without ET, however the majority of decay mechanisms was observed to be ET→GD. This shift from CR to ET→GD decay followed with increasing polarity, with 9 of the 10 water simulations exhibiting this decay mechanism.

By analyzing the Mulliken charges for each atom on the ground and excited state, we determined that the acetyl groups act as electron acceptors and the amines act as donors during ET (Figure 7). Atoms O1/2, C4/9, N1/2 and C1/6 are the most electronegative centers on p-DAPA in both ground and excited states for the electron transfer pathway in all tested solvents (see Tables S3 and S4). We also observe that the difference in total charge on the benzene ring between ground and excited state correlates with the average electron flow of both functional group pairs, regardless of solvent. We further examined the excitation energies of each simulation at the start and end, then compared decrease in S₁ energy values across classified decay mechanisms. We find that in all systems in which ET occurs, the decrease in S₁ energy is at least 1.5eV over the course of the simulation, and that no system which lacks ET character shows more than 0.9eV of decrease in S₁ energy.

FIG. 7.

(a) Atom names for p-DAPA. (b) Acetyl bridge showing the dihedral angles that lead to the different decay mechanisms. (c) Comparison of Mulliken atomic charges of p-DAPA in the ground and excited states in the electron transfer and (d) proton transfer pathway in chloroform.

We also examined the distribution of vertical excitations for the first and last 500fs of each simulation and note that those simulations which we identified as having ET character also exhibited the largest peak separations between distributions (see Figures S7, S12, S17, S22, S27, S32, S37, S42).

F. Solvent polarity affects acetyl group rotation

With increasing solvent polarity, the average dihedral angle across both acetyl groups in all simulations shifts further from coplanarity with the benzene ring and also exhibits increasing fluctuation about the average. Nonpolar and weakly polar solvents like cyclohexane, butane, and chloroform maintain average dihedrals within 5° of planarity with standard deviations of < 20°. By comparison, strongly polar solvents like methanol and water show average dihedrals considerably further from planarity with significantly larger standard deviations, indicating an increased mobility in these solvents which allows access to the electron transfer decay mechanism (see Tables III F and S5–S12). The enhanced mobility in polar solvents can be attributed partially to the stabilization of charge-separated configurations: in planar geometries, conjugation is maximized, leading to more delocalized electrons, whereas deviations from planarity localize charge. Polar solvents preferentially stabilize these charge-separated states, lowering the energetic cost of non-planar distortions. At the same time, polar solvents tend to be more organized around the solute and have a stronger influence on its structure sterically. For example, in snapshots where a water molecule is competing for a hydrogen bond between the amine and acetyl groups of p-DAPA, the water molecule hydrogen bond will take precedence and immediately disrupt p-DAPA’s planarity.

G. Dihedral angles correspond to HOMO-LUMO gap

In excited state simulations, we observed an interesting trend in trajectories which lacked ET character. When the dihedral remains nearly planar with respect to the plane of the central benzene ring, the excitation energies are consistently around 2.8eV. During the simulations, when the acetyl group rotates beyond approximately 30° from the plane, we see that the excitation energy reported drops sharply, sometimes as low as 0.5eV. This energy drop correlates roughly with the magnitude of acetyl rotation. In a number of simulations, however, this angle then returns to a planar orientation with a corresponding return to the higher excitation energies. These simulations specifically lack any ET character, and we find that the drop in excitation energy is actually an increase in the corresponding ground state energy of the geometric conformation at that point in time. This increase in ground state energy shrinks the HOMO-LUMO gap temporarily. However, as the simulation continues back towards a planar geometry, the corresponding ground state energy is reduced once more and the HOMO-LUMO gap is restored (see Figures S14–S16). Conversely, in simulations in which the same dihedral rotations are observed concurrently with electron transfer, the energy loss is not observed to be reversible (see Figures S33–S42).

H. Proton transfer is a transient event

While we did not expect to observe proton transfer events at these timescales, we nonetheless examined all simulations to identify any in which an amine proton transferred to the acetyl group oxygen. Each simulation was analyzed to measure the N-H···O distances for both functional group pairs over time, and these values were normalized to the N-O distance at each frame such that a value of 0.5 indicates the proton is equidistant from both heavy atoms. Values above 0.5 correspond to simulation frames in which the proton has moved closer to the oxygen than to the nitrogen. This valuation was selected to ensure that the O-H distance measurement was not taken as a metric without the context of the N-H distance at that same time. We identified only one simulation in chloroform in which the proton was closer to the oxygen than the nitrogen, however these events were transient and corresponded to a combination of dihedral rotation and reduced vertical excitation energies (see Figure 8). This suggests that, as previously reported, the proton transfer mechanism in p-DAPA is high-energy and largely inaccessible at this timescale.

FIG. 8.

Representative trajectory for chloroform simulations. (Top) The dihedral angle over time for each acetyl group relative to the plane of the benzene ring was calculated using multiple atom sets (dashed lines) and averaged (solid line) to reduce noise and identify trends in dihedral rotation. (Middle), The vertical excitation energy (S₀→S₁) calculated on S₁ over time. The dashed black line was calculated as the “maximum future energy” (see SI for full details), indicating whether reductions in bandgap were reversible over time. (Bottom) Proton sharing ratio over time was calculated such that points on the dashed black line in the center are time indices at which the proton is exactly equidistant from both the amine nitrogen and the acetyl oxygen (adjacent functional group pairs separated into red and blue lines). Points above this center line indicate the proton is closer to the oxygen, while points below this line indicate the proton is closer to the nitrogen. Values were normalized such that the N-H distance and O-H distance at any timestep summed to 1.0, establishing 0.5 as the equivalence point. Vertical lines for Electron Transfer (ET) events (green) and reversals (red) are included at their respective time indices to provide further context for these time-dependent metrics. For full details, see SI.

IV. CONCLUSION

The photophysics of p-DAPA have been examined in a range of diverse solvents based on differences in polarity and protic nature. For all tested solvents, decreasing polarity improves the photostability of p-DAPA by slowing electron transfer to the decay-associated state character. By analyzing QM/MM simulations with explicit solvent we identified key factors that lead to narrowing ground/excited state energy gaps. These included that dihedral twisting of the acetyl group can lead to electron transfer or geometric distortions instead of the previously discussed proton transfer deactivation pathway. This demonstrates that solvent character and geometric structure plays a key role in the excited state dynamics of p-DAPA.

To observe fluorescence instead of non-radiative relaxation, rigidifying the p-DAPA by eliminating the flexible acetyl group from twisting will help maintain planarity and reduce or eliminate the geometric distortion pathway towards non-radiative decay. Another design rule for creating red emissive compounds is to increase the conjugation by using cyclic compounds or sterically bulky electron withdrawing or donating substituent. Forcing planarity, such as by the inclusion of nonpolar functional groups, in addition to careful choice of functionalization position, is likely a consistent design point for creating new SBFs which remain emissive in polar or protic solvent environments.

Supplementary Material

V. SUPPLEMENTARY MATERIAL

Zenodo link to external dataset including xyz files, omega tuning gasphase results, natural transition orbitals for optimized, gas phase and implicit solvated systems, optimized geometries and natural orbitals for polar and nonpolar solvents, absorption and emission energies, computed partial charges, dihedral angles by solvent, and detailed simulation data for each trajectory and solvent.

TABLE II.

Average dihedral angles observed for both acetyl groups across all replicates and solvents.

Solvent	Dihedral (avg) ± σ
Cyclohexane	−1.090 ± 11.294
Butane	−1.074 ± 9.552
Chloroform	−2.577 ± 11.478
DMSO	1.989 ± 11.452
DCM	−2.468 ± 11.565
Acetonitrile	1.182 ± 10.108
Methanol	−10.846 ± 10.642
Water	8.860 ± 18.095