Unveiling coupled electronic and vibrational motions of chromophores in condensed phases

Elucidation of the multidimensional reaction coordinate and energy flow in molecular systems provides the typically hidden yet crucial link between the structure, energy, and function.^1–5 On a fundamental level, the intimate correlation between electronic state changes and coordinated nuclear responses guides the directional movement from the reactant to the product: first, electrons take the lead upon photoexcitation, followed by rapid nuclear motions particularly for those light and facile protons, and then the electrons react to the newly established atomic configurations. This action-reaction cycle continues until the system reaches an equilibrium state and acquires its functional properties.⁶ Accordingly, ultrafast electronic and vibrational spectroscopies in tandem can offer an integrated and complementary toolset to track the correlated electronic and structural motions during the photophysical and/or photochemical processes (Fig. 1).

It is well accepted that molecules exhibit chemical reactivity because the constituent atoms move around in their microenvironment and collectively contribute to macroscopic functionality.² As a structural dynamics technique, femtosecond stimulated Raman spectroscopy (FSRS) has demonstrated its unique advantages since the early 2000s because it not only allows the tracking of atomic motions at the chemical bond level under equilibrium or nonequilibrium conditions^4,7,8 but also selectively enhances vibrational signatures of transient species as chemical reactions proceed in real time.^9,10 In other words, the very nature of stimulated Raman and resonance enhancement (see Sec. II) opens exciting new avenues to reveal the highly desired structural dynamics information from competing nonlinear spectral signals, some of which are overwhelming if detected or processed in other ways. Another typically overlooked benefit of FSRS is to reveal the excited state processes of highly fluorescent molecules (e.g., laser dyes, fluorescent proteins),^5,11 which could not be easily studied by conventional Raman spectroscopy (e.g., cw Raman).

Since 2007, several comprehensive reviews^4,7,8,12 on theoretical and experimental development of FSRS have covered small molecules (e.g., β-carotene, rhodamine 6G, charge-transfer complex, spin-crossover metal complex) and biomolecules (e.g., rhodopsin, bacteriorhodopsin, green fluorescent protein or GFP, and phytochrome). In this perspective, we highlight recent advances in the FSRS field including technical information and specific considerations to capture the coupled electronic and vibrational motions of organic chromophores in various solutions and the protein environment. We also address some current issues that could benefit from further experimental and theoretical inquiries. For instance, performing FSRS with a bluer Raman probe than the Raman pump has yielded new results that challenge the conventional wisdom (see Sec. II B).^13–16 Different stimulated Raman lineshapes could be attributed to dynamic resonance Raman enhancement with not only the excited state absorption (ESA)^16,17 but also the stimulated emission (SE) band,^14,15 which has enabled many new studies in tracking vibrational motions of photosensitive systems in the electronic excited state (see an example in Fig. 1).⁵ We recognize that competing theories and opinions still exist, while researchers actively compare results and findings based on various versions stemming from basic FSRS principles. However, an updated account of the field is warranted from the evolving experimental perspectives. Some restatement is necessary in Sec. II to put our discussion in context and present recent results, so general readers can appreciate the research frontier, major findings, remaining challenges, and future opportunities for FSRS studies on functional chromophores in condensed phases.

In particular, we aim to motivate the next phase of theoretical development to better understand FSRS data, gain structural dynamics insights, and enable the spectroscopy-guided molecular engineering (Secs. III C–III D) from targeted design and sample preparation to optical control. The greatly deepened and potentially transformative knowledge of chromophore structural dynamics, which power a plethora of functions beyond fluorescence, will continue to accumulate and be applied as rational de novo design principles of next-generation molecular machines.

With all the energy level diagrams considered for nonlinear pathways during coherent signal generation, ground-state FSRS can be treated as a four-wave mixing (FWM) process, while excited-state FSRS involves six-wave mixing due to the generation of excited-state population by two interactions with an actinic pump on the bra and ket side of the double-sided Feynman diagram.^7,18,19 In particular, coherent nuclear motions induced by the photoexcitation pulse can exert statistically significant effects on the excited state molecular evolution with discernible spectral signatures (see Fig. 1 for illustration). Upon tuning the actinic pump and Raman pump-probe pair from the ultraviolet (UV) to near-infrared (NIR) region, specific electronic states and associated transient species of the target system can be accessed with sufficient spectral and temporal resolution during photophysical or photochemical processes. Essential concepts in this section are useful in enabling the FSRS studies of organic chromophores in condensed phases, with new discussions in several important directions that could power future advances in molecular design and engineering.^4,7,8,20

One of the key advantages of performing FSRS in the mixed time-frequency domain is a direct acquisition of time-dependent Raman spectra of a photoexcited molecular system starting from time zero (Fig. 2). The spectral resolution needs to be sufficient to resolve multiple vibrational motions of the sample under femtosecond and picosecond laser irradiation. For previously reported chromophores in solution or inside a protein matrix (e.g., GFP), typical Raman bandwidth is in the vicinity of 15–35 cm⁻¹ (broader for excited-state peaks than ground-state peaks), reasonable for the ensuing spectral data analysis on a mode-specific basis.^10,21,22 The broadening of ground-state FSRS peaks relative to the natural linewidth in continuous-wave (cw) Raman likely arises from the picosecond duration of the Raman pump,^7,23 while molecular structural inhomogeneity in condensed phases contributes to a Gaussian peak lineshape more than a homogeneously broadened Lorentzian lineshape.^3,24,25

After electronic excitation, time-resolved FSRS peaks become broader (when compared to the ground-state peaks in S₀) because of the shorter lifetime of an excited state (e.g., S₁). However, the center frequency of a Raman band remains a reliable observable to track transient structural evolution as shown for wild-type (wt) GFP using an 800 nm Raman pump (R_pu) and a redder Raman probe (R_pr) after a 400 nm actinic pump (A_pu).²¹ Later examples include mode frequency shifts of the GFP-based Ca²⁺ biosensor chromophores.^26–29 Simulations of such an evolving conformational coordinate were demonstrated using a 2D model system with a slow conformational degree of freedom and a higher-frequency vibration,³⁰ and time-dependent (TD) quantum mechanical wavepacket averaged mode frequencies were calculated along the excited-state trajectory from an initially displaced position.³¹ Time-dependent density functional theory (TD-DFT) of the intrinsic reaction path³² enriches such inquiries by revealing the driving force of the excited-state proton shuttle in GFP, giving strong evidence that the photoinduced structural motions play active roles for the chromophore photoacidity in relation to a dynamic H-bonding network inside the GFP pocket.³³ 

To realize the molecular movie capability of FSRS with a precise time zero and sufficient resolution, it is important to use short (femtosecond) pulses to initiate photochemistry and capture the structural snapshots in action. The experimental strategy to achieve simultaneously high spectral and temporal resolutions lies in the picosecond R_pu and a concomitant use of the femtosecond R_pr and A_pu.^4,7,8 Initial vibrational coherences are generated by the femtosecond actinic pulse, usually the low-frequency skeletal modes within the A_pu bandwidth. Meanwhile, since the probing vibrational coherences are generated by the picosecond-R_pu-femtosecond-R_pr pair, one experimental spectrum can cover a broad range of low to high-frequency modes (e.g., 100—2000 cm⁻¹) within the energy difference between the R_pu and R_pr photons (Fig. 2) at well-controlled time delay points in the electronic excited state. Therefore, the time resolution is solely determined by the cross correlation between the femtosecond A_pu and R_pr pulses. Theoretical work on the effective temporal and spectral resolution of FSRS was performed.^18,34 The sophisticated FSRS technique mitigates typical limiting factors associated with a single-pulse measurement (i.e., the lifetime-bandwidth product of ∼5300 cm⁻¹) by essentially decoupling the temporal and spectral resolutions achievable in FSRS experiments (typically as low as ∼30 fs and 10 cm⁻¹).^8,19,21,23 In such optical setups, a CCD array camera without any shutters (i.e., continuous data collection with the preset number of laser shots per time delay) is then used to collect a series of well-averaged Raman spectra from a dispersive spectrograph during a photoinduced reaction.

Given the wavelength tunability and versatility of the optical setup, it is natural to perform FSRS experiments on both the Stokes and anti-Stokes sides by using a Raman probe on different energy sides of the Raman pump. We consider that the terminology of “Stokes” and “anti-Stokes” should not be reserved for conventional Raman as long as the underlying FSRS pathways in new studies are articulated. Previously, transient anti-Stokes Raman spectroscopy was considered to be a useful technique to measure intramolecular vibrational relaxation (IVR) and other vibrational cooling phenomena because higher vibrational quantum states are involved.^35–37 However, limitations existed due to small spontaneous Raman cross sections, the fluorescence background, and lack of sub-picosecond time resolution. The excited-state FSRS methodology mitigates these problems by the femtosecond-picosecond-femtosecond pulse sequence generating nonstationary wavepackets in the electronic excited state, the noncollinear stimulated Raman scattering (SRS) process, and tunable dynamic resonance conditions realized synergistically by all incident pulses. As a result, specific resonance Raman enhancement can be attained for characteristic vibrational modes, fluorescence is effectively rejected, and femtosecond time resolution is readily achieved to track a variety of photoinduced processes such as intramolecular/intermolecular charge transfer (CT), proton transfer, vibrational cooling, internal conversion, intersystem crossing, isomerization, and radiative emission (see Sec. III for representative advances).

In an early work by Mathies and co-workers, anti-Stokes FSRS was performed on the electronic ground state of Rhodamine 6G (R6G).³⁸ As R_pu was tuned from the red edge of the visible absorption band toward the blue edge, vibrational modes below ∼1000 cm⁻¹ go from negative to dispersive, purely positive, and then oppositely phased dispersion as the pump wavelengths approach the absorption maximum. Even earlier, dispersive lineshapes were observed at the ground-state vibrational frequencies of bacteriorhodopsin after 500 nm excitation, and the time-ordered dipole couplings were formulated to simulate the third-order susceptibility and FSRS lineshapes.³⁹ The interference between the R_pr field and the out-of-sync signal field during heterodyne detection leads to dispersive lineshapes, which are dependent on the vibrational mode frequency, Franck-Condon (FC) factor, and coherence lifetime. Dang et al. later reported a study using FSRS to measure the temperature of condensed matter (CaCO₃ single crystal) at the molecular vibrational level from thermal equilibrium to laser-induced nonequilibrium conditions on the picosecond time scale.⁴⁰ However, the relationship between the Raman loss:gain ratio and temperature and how spectral symmetry between the Stokes and anti-Stokes FSRS signals gets affected by electronic resonances remain actively studied and debated.^13,41–44 In 2017, Fang and co-workers implemented the anti-Stokes and Stokes FSRS to track ultrafast vibrational cooling during excited-state proton transfer (ESPT) from a photoacid pyranine to water based on a fortuitous dynamic resonance condition in S₁.¹⁵ In essence, a fixed 580 nm Raman pump achieves two dominant preresonance conditions in the singlet excited state: first, with an excited-state absorption (ESA) band of the photoreactant (i.e., protonated chromophore) and later, with a stimulated emission (SE) band of the photoproduct (i.e., deprotonated chromophore) as the ESPT reaction proceeds on ultrafast time scales.

Since the double-sided Feynman diagrams for the ground-state FSRS have been discussed in the literature,^{18,34,39,42,44–47} we depict the excited-state Stokes (Fig. 3) and anti-Stokes (Fig. 4) FSRS pathways with an ESA band to lay out central contributions to the sign and lineshape of the observed FSRS signal. Vibronic states are intimately involved in sequential interactions between a molecular system and incident laser pulses, and FWM processes in the electronic excited state (see Fig. 2, for example) can adopt various forms of response functions to generate the Raman gain, Raman loss, or dispersive Raman lineshapes, when the off-resonance [Fig. 2(a)] and on-resonance [Fig. 2(b)] conditions are separately considered for the excited-state FSRS signal generation.

During quantum wavepacket calculations, the top three diagrams in Fig. 3 are commonly grouped into a term called stimulated Raman scattering or SRS(I), especially when R_pu is on resonance and the excited state lifetime is long. Contributions from the resonance Raman scattering RRS(II) and inverse Raman scattering IRS(II) terms are broad and thus affect spectral baselines.^38,47 The hot luminescence HL(I)–(IV) terms become dominant when the Raman pulses approach resonance conditions with the ESA band as shown here. In contrast, due to the longer wavelength of R_pu than R_pr in Fig. 4, all the terms could start from a vibrationally hot state except for IRS(I), which may originate from a vibrationally cold state. This rationale explains the observed excited-state anti-Stokes Raman peaks at lower frequencies than Stokes counterparts,¹⁵ and a FSRS signal sign change from positive to negative [e.g., the dominant RRS(I) converting to IRS(I) pathway] when the shorter-wavelength R_pr is tuned to be on resonance with the ESA band and acts initially in an excitatory role.¹⁶ Further insights from theory groups into photoexcited molecules navigating the multidimensional PES and governing FSRS features are required to systematically understand detailed contributions from all distinct and competing pathways.^19,42,48

There have been two conceptually different theoretical models, perturbative and nonperturbative, which can simulate light-matter interactions in the weak and strong limits, respectively.^45,46,49,50 Early development of FSRS theory adopted the perturbative treatment where the diagrammatic approach represents the third-order nonlinear polarization induced in the molecular system density matrix by incident laser pulses (e.g., using Feynman diagrams and starting from the field-free wavepackets), mostly suitable for cases where R_pu and R_pr are electronically off-resonant. An increasing number of pathways need to be considered when the incident stimulated Raman pulse pair approaches electronic resonance conditions, and the HL terms (see Figs. 3 and 4) were added to describe nascent excited-state vibrational coherences.^45,47 Alternatively, nonperturbative methods have been developed to describe the strong-field effects and directly solve the equation of motion (i.e., the time-dependent Schrödinger equation for the electronic and vibrational wavefunction for a condensed-matter system under irradiation). Such numerical simulations have been recently applied to conical intersections (CIs) located in the FC region (spectroscopically accessible) with an intense actinic pump pulse on a three-electronic-state (⁠$e0$⁠, $e1$⁠, $e2$⁠) two-vibrational-mode (a tuning mode and a coupling mode) model.⁵⁰ The fundamental aspects of light-matter interaction and the resultant third-order polarization spectrum based on the sum of all possible pathways are actively being modeled and calculated to correlate with FSRS experimental results, especially in the electronic excited states.^{19,30,31,42,47,50}

During the early development of FSRS, it was recognized that preresonance enhancement on Stokes side can obtain high-intensity Raman signals while avoiding significant interference from the HL terms.³⁸ Without A_pu, the Feynman diagrams illustrating density matrix evolution during FWM processes were formulated using the classical coupled-wave theory and quantum mechanical density matrix involving nonstationary wavepackets.^7,45–47 Notably, the HL terms may contribute when excited-state FSRS is performed, wherein sequential interactions of a molecular system with the picosecond R_pu and femtosecond R_pr pulses generate a vibrational coherence in electronic excited states (e.g., see Figs. 3 and 4 on resonance with an ESA band). Pertinent energy level diagrams were depicted that involve the probing of multidimensional reaction coordinates during an ultrafast photochemical reaction in S₁.¹⁵ The light-matter interactions now enter the expanded regime because the energetic and reactive PES can have upward (ESA) and/or downward (SE) transitions as the impulsively generated wavepacket slides down the initial PES slope out of the FC region.¹⁸ The resonance Raman enhancement factors achieved throughout molecular structural evolution are intrinsically dynamic after electronic excitation, which corroborates the approach to first perform femtosecond-TA and then select incident pulse wavelengths in time-resolved excited-state FSRS experiments (Sec. II E).

Besides the potentially dominant and informative coherent motions illuminating previously hidden primary molecular events that hold functional importance for photochemical reactions, the R_pu-R_pr pair continues to generate vibrational coherences in the excited state to track the transient population change before the system reaches the CI, fluorescent state or other lower-lying states. In principle, it is possible to follow the entire photochemical reaction trajectory via stimulated Raman as long as the population does not completely return to the original ground state. In reality, the detection time window can extend to nanoseconds.⁸ As for the initial reaction phase (e.g., femtosecond to picosecond time scales, see Fig. 1) where a significant molecular population exists in the electronic excited state with high Raman polarizability and a notably anharmonic PES (Fig. 2), resonance enhancement effects can be exploited to enhance transient vibrational features during the primary events.^5,10

The foundational principle of such a resonance effect is the enhanced third-order nonlinear response and the displacement of vibrational coordinates in resonant electronic states.^9,51,52 When certain resonance conditions are fulfilled in the excited state, previously elusive or unclear reaction coordinates that may involve nuclear motions coupled to those electronic transitions (i.e., vibronic coupling) can be unveiled. This is also intimately related to the coherent Raman approach wherein the nuclear oscillators in various parts of the sample are directly correlated with the incident light fields, establishing a well-defined phase relationship.⁵² In a recent report on the Stokes and anti-Stokes FSRS lineshapes of the model photoacid pyranine (see Sec. III A), we exploited various dynamic resonance conditions by scanning the R_pu wavelength across the dominant ESA band of pyranine in methanol [see Fig. 5(a)].¹⁶ Notably, the mode-dependent Raman lineshapes were observed as a function of the R_pu wavelength, along with the R_pr wavelength range that also contributes to the dispersive lineshapes (more prominent in the anti-Stokes FSRS) [see Fig. 5(b)]. On the Stokes side, Raman peaks are enhanced significantly by a preresonance R_pu, not an on-resonance one [Fig. 5(c)]. Moreover, the ESA dynamics provide an effective means to understand the observed FSRS lineshape change as a function of time after the photoexcitation time zero.¹⁶ 

Using an SE band for resonance enhancement represents a new advance for FSRS (vs the more conventional utilization of absorption bands either from the electronic ground or excited state).^13,16,17 It is understandable that at R_pu off-resonance conditions although A_pu is effective in creating some excited state populations, fluorescence (i.e., spontaneous emission) should be insignificant within the detection window defined by the R_pu-R_pr pair, while R_pu is far away from the electronic transition peak. Also, transient population takes time [typically on the nanosecond (ns) time scale] to undergo radiative emission, while the broad fluorescence band does not pose significant competition or background to the much narrower vibrational peaks evolving on the femtosecond to picosecond time scales. However, when R_pu is tuned toward an electronic transition, preresonance enhancement can significantly enhance the signal-to-noise ratio (SNR) of transient FSRS peaks. R_pr may approach the electronic transition peak (e.g., Stokes when R_pu is on the blue side of the SE band, or anti-Stokes when R_pu is on the red side of the SE band), but its femtosecond time duration and weak intensity^15,23,53 ensures that no significant effect on the excited-state population would occur due to R_pr (e.g., dumping).

A special case is the lasing-facilitated FSRS wherein a very weak R_pu close to an emission peak gets enhanced by an SE transition to generate the observed FSRS signal. It has been recently demonstrated by us in a modified version of the excited-state FSRS setup, wherein a chopper (instead of a shutter) is placed in the A_pu beam path while no chopper is used for R_pu. This specific experimental configuration ensures that the Raman signal directly stimulated by the R_pu-R_pr pair is not observed. When R_pu is blocked, the TA signal (I) is recorded. When R_pu is unblocked, the TA signal may mix with the FSRS signal (II) only when the femtosecond A_pu induces notable changes to the picosecond R_pu synchronized with the laser system at half of its repetition rate (i.e., 500 Hz). The difference between signals (II) and (I) achieves the lasing-facilitated FSRS with high sensitivity. This newly developed methodology could provide a versatile and broadly applicable platform to investigate the excited-state structural dynamics of highly fluorescent molecules, even including laser dyes.¹¹ Related work with more control experiments (e.g., the A_pu and R_pu power dependence, sample concentration) are currently ongoing and will be elaborated more fully in a future publication.

The logarithm of a ratio of R_pr with both A_pu and R_pu “on” over R_pr with R_pu “on” has been used to extract the transient FSRS signal,^42,56,57 analogous to femtosecond-TA methodology to calculate transient optical density mostly in the ΔµOD to ΔmOD range. In comparison, the alternative calculation method (i.e., taking the ratio, then minus one) may be beneficial due to the experimental self-heterodyned nature of the FSRS scattering signal collinear with R_pr, the miniature gain due to the typically small Raman scattering cross sections, and the linear regime to collect the FSRS signal with intensity being proportional to the R_pu energy.^23,53,58,59 The spectra need to be calculated as the logarithm of a spectral intensity ratio only at very large gains. In other words, when γ is small (I $≈$ I₀), $log10(II0)≈12.303lnII0=12.303ln1+(II0−1)≈12.303(II0−1)$⁠, while the other calculation method of spectral data directly yields $(II0−1)$⁠.^21,23 The latter approach therefore yields a larger signal numerical value, but the experimental SNR remains comparable in both methods.

A common question naturally arises, what is the benefit of performing ultrafast Raman spectroscopy? Notably, the popular time-resolved IR (TRIR) and 2D IR techniques have recorded transient vibrational features from many functional groups in chemical and biological systems.^3,60–68 A detailed comparison is warranted here to appreciate the differences and complementary information between ultrafast Raman and IR. The specific problems that need to be solved should ultimately dictate which technique or combination thereof may be chosen, drawing inspiration from the list of experimental innovations and applications detailed in this section.

For molecules with a center of symmetry, the mutual exclusion rule dictates that no vibrational normal modes can be both Raman and IR active.⁶⁹ When proteins and most photoacids (i.e., with organic chromophores)^23,62,70,71 in the condensed phase are concerned, the nonsymmetric (point group C₁) molecules typically lead to vibrations that are both Raman and IR active. In this case, low extinction coefficients of vibrational transitions could pose challenges for IR measurements, in combination with some level of difficulty in working with ultrafast IR pulses. Based on years of experience,^63,64,72 we consider that 2D-IR is a powerful technique that excites and detects selective vibrational modes in a 2D Fourier-transform frequency map as a function of the population waiting time, providing insights into spectral diffusion and the time evolution of vibrational lineshapes with an effective separation of homogeneous and inhomogeneous broadening mechanisms.^3,25,51

However, some challenges of TRIR and 2D IR exist as follows: (1) There is a strong overlap between water (solvent) peaks and sample (solute) bands, so water is commonly avoided as a solvent. (2) The optical setup needs to be constantly air-purged. (3) Spectral overlap between vibrational marker bands from similar functional groups leads to limited or no site specificity. (4) Peak intensities are generally small and cannot be enhanced by optical methods. (5) The detection spectral window is relatively narrow, in the range of a few hundred per centimeter. (6) Low-frequency modes below ∼1000 cm⁻¹ are challenging to observe.^3,63,64

Experimental IR strategies include the following: (1′) Use D₂O as a solvent to shift solvent bands out of the spectral region of interest to observe protein modes such as the amide-I or II bands.³ (2′) Build an air-tight compartment to house the IR setup and purge it with dry air without moisture or CO₂. (3′) Incorporate isotopic labels such as ¹³C=¹⁶O or ¹³C=¹⁸O at specific backbone locations in peptide and protein samples^63,72 or use nitrile groups such as C≡N or S—C≡N in sidechains to report on the local structure and dynamics.^64,73,74 In general, more unnatural IR probes have been used to facilitate experimental advances as well as refine computational methods to further interpret experimental results. (4′) Develop nonlinear spectroscopy such as 2D IR to enhance SNR at the single-residue level, e.g., one ¹³C=¹⁸O label in a peptide,⁷² and use high sample concentration typically in the millimolar (mM) range.³ (5′) Probe one vibrational region and then move the grating and monochromator to the next position to probe another vibrational region when separated by over 300 cm⁻¹ or perform two-color 2D IR or broadband IR to observe signals from a wider spectral window.^68,75–77 (6′) Develop advanced optical parametric amplifiers (OPAs) to produce a longer-wavelength femtosecond laser source in the IR range that goes beyond 10 µm and covers a spectral range over 3000 cm^−178,79 or perform a hybrid of IR-Raman measurements.⁸⁰ 

Using FPs as an example, due to technical limitations and spatiotemporal resolution issues in earlier laser spectroscopy, the chromophore atomic motions before radiative emission have remained elusive and development of FPs has been largely trial-and-error.^81,82 The TRIR studies of FPs provided important insights, but overlapping peaks from the chromophore and surrounding protein residues hindered spectral analysis.^83,84 To capture conformational snapshots of the “epicenter” of fluorescence immediately following electronic excitation, specific chromophore interactions with its surrounding and response to the incoming photons need to be tracked with sufficient resolution. The table-top FSRS optical setup provides just that, based on a suitable laser pulse sequence with femtosecond or picosecond pulse durations, enabling the much-needed resonance Raman enhancement for transient molecular species during a photoinduced transformation (see Sec. II C). The niche of FSRS as a time-resolved vibrational spectroscopic technique using visible pulses thus makes FPs almost ideal protein systems to be studied, especially targeting the organic chromophores that govern FP emission properties (see Secs. III B and III E for details).

In particular, because the active site for wtGFP fluorescence is the three-residue Serine65-Tyrosine66-Glycine67 (SYG) chromophore formed by autocatalytic cyclization during protein maturation, the spectroscopic technique of choice needs to single out the chromophore response to light despite being surrounded by a much larger number of protein residues. Neighboring residues of the chromophore are relevant for maintaining the H-bonding network, electrostatics, and sterics to the chromophore, while others are irrelevant or insignificant for fluorescence. The choice of conventional FSRS with a picosecond 800 nm Raman pump utilizes the chromophore ESA band near 900 nm, leading to preresonance Raman enhancement in S₁.²¹ For the first time, structural motions of the chromophore embedded at the center of GFP β-barrel were tracked in real time following photoexcitation. Notably, a ∼120 cm⁻¹ motion was uncovered from coherent intensity and frequency oscillations of high-frequency vibrational marker bands (>800 cm⁻¹) in FSRS to play an active role along the photochemical reaction coordinate.^4,12,21

A general operating procedure has since emerged after recent technical development of tunable FSRS.^26,27,85 First, perform steady-state absorption and fluorescence spectroscopy followed by femtosecond-TA to identify prominent transient electronic bands and then guide the selection of A_pu and R_pu wavelengths. Second, optimize the A_pu and R_pu powers as well as the R_pr spectral window to collect FSRS data in the electronic ground and excited states, ideally with related or contrasting samples to aid spectral assignment and interpretation. Third, tune the A_pu and/or R_pu wavelengths to access and enhance different transient molecular species before, during, or after a photoinduced reaction such as ESPT. These time-resolved Raman studies represent a new way to complement the existing IR results^62,86,87 and further map out the chromophore excited-state structural dynamics prior to radiative and nonradiative relaxation pathways in solution and protein environments.^5,10,21,88

To perform impactful research using the aforementioned FSRS methodology, we highlight two of the most challenging problems that the world faces now: human diseases and the energy crisis. The development of cures for many diseases can be greatly expedited by mechanistic insights into the pertinent cellular processes, which could be tracked using advanced fluorescent bioprobes. A large Stokes shift is desirable to improve the bioimaging quality, which can be effectively achieved by incorporating an ESPT reaction such as that in GFP.^21,89–91 Regarding the energy crisis, converting sunlight into chemical or electrical energy has been a research focus for decades since solar energy is clean and abundant. A few bottlenecks remain to be solved to dramatically increase the energy conversion efficiency. For example, past attempts to fully mimic natural photosynthesis with artificial ones have been unsuccessful, in part due to a lack of fundamental understanding about the multistep “flowchart” that involves various stages of proton-coupled electron transfer (PCET).^92,93 Therefore, tackling these problems can be boiled down to a single proton. A type of small molecules termed photoacid, which decreases its pK_a upon photoexcitation and can donate a proton in a suitable solvent,^10,94,95 affords an excellent model system. Besides the aforementioned FP chromophores, essentially photoacids undergoing ESPT inside the protein matrix,^21,87,90 more photoacids have been implemented in bioimaging applications such as intracellular pH sensors,⁹⁶ photovoltaic devices that convert light into renewable ion transport,⁹⁷ and new biosensors and materials that can be broadly considered as photoactivated molecular machines.⁵ 

In this perspective, we focus on the structural dynamics of various photoacids: from weak to super-photoacids, from the acid to base form, from the first to higher-lying excited states, and from the naturally occurring to engineered ones. The photoinduced dissociation of a proton from a photoacid molecule in the condensed phase (both solution and protein environments) represents an archetypal system to investigate the coupled electronic and nuclear motions during a chemical reaction, with sufficient spectral and temporal resolutions offered by an integrated spectroscopic platform that consists of femtosecond-TA and FSRS (Sec. II E). Our ultrafast spectroscopic findings have provided novel insights into ESPT mechanisms,^10,15 photoprotection mechanisms (e.g., under high-energy light irradiation or inside photovoltaic devices),⁹⁸ rational design principles to achieve redder and brighter fluorophores,⁹⁹ and effective strategies to increase photoacidity.^14,100

HPTS (8-hydroxypyrene-1,3,6-trisulfonic acid), a conjugated four-ring system with a hydroxyl group at one end (see the inset of Fig. 1), is a well-known photoacid whose pK_a drops from ∼7 at ground state (S₀) to ∼0 in the first singlet excited state (S₁).^70,101,102 In aqueous solution, the hydroxyl proton of HPTS can be donated to a nearby water molecule following photoexcitation. The relevant ESPT reaction steps have been studied by ultrafast spectroscopies and calculations,^{10,62,70,96,101–104} primarily upon exciting HPTS at the main absorption band (∼400 nm). The ESPT process consists of two major stages: formation of an excited-state complex between HPTS and an adjacent water molecule on the 2–3 ps time scale, and a subsequent diffusion-controlled proton dissociation with a ∼90 ps time constant. Certain low-frequency skeletal modes were revealed to play functional roles, including a 180 cm⁻¹ H-bond stretching mode that promotes the formation of an excited state solute-solvent complex [e.g., a transient contact ion-pair (CIP) upon broadly defining the charge or electron transfer in a molecular system]^10,59,70,102 at early times, as well as vibrational and thermal cooling at later times.^10,15

However, little is known about how HPTS behaves in the higher-lying excited states (S_n) above S₁. We systematically investigated its behavior in water using tunable FSRS with a 267 nm [deep UV (DUV)] actinic pump, supplemented by control experiments of HPTS in methanol.⁹⁸ Switching the pump from 400 to 267 nm does not affect the reaction outcome: ESPT still proceeds in water but not in methanol. ESPT pathways for HPTS in water upon 267 nm excitation are as follows: after the initial sub-picosecond S_n → S₁′ transition, ESPT likely proceeds in a lower-lying S₁ state or an S₁′ state (different from the S₁ state accessed by 400 nm). Instead of forming a CIP with water, HPTS in S₁′ undergoes ultrafast solvation (e.g., ∼1 ps in water) before proton dissociation with a conserved time constant (i.e., ∼100 ps vs 90 ps for HPTS in S₁). Notably, several excited-state Raman bands of the chromophore exhibit frequency shift dynamics (Fig. 6) that are closely related to the ultrafast transitions between electronic excited states and the vibrational relaxation therein.

Notably, the 430 cm⁻¹ mode consists of the in-plane ring deformation and is assigned to the excited photoacid species (PA^*). Its frequency redshift in both water and methanol under DUV excitation [Fig. 6(a)] is due to the chromophore structural change (e.g., breaking the four-ring coplanarity), which is less prominent under 400 nm excitation. The bulkier methanol molecules can suppress such structural changes, leading to smaller mode frequency shifts than those in water. In contrast, the 680 cm⁻¹ mode shows an opposite trend in water (blueshift) and methanol (redshift) regardless of the excitation wavelength [Fig. 6(b)]. For this ring hydrogen-out-of-plane (HOOP) with phenolic COH rocking motion, the redshift could also be related to a structural change of the chromophore, similar to the 430 cm⁻¹ mode. However, the blueshift in water is attributed to the ESPT reaction, which overrules the redshift trend. Therefore, the “counteractive” competition between ESPT and conformational change results in a lengthened blueshift time constant (170 ps for 267 nm excitation, and 130 ps for 400 nm excitation) than the ESPT time (∼100 ps).^10,15,98

Also, the previous literature has confirmed that the ∼1530 cm⁻¹ mode (ring C=C stretching and phenolic COH rocking) frequency blueshift tracks the ESPT progress after 400 nm irradiation.^5,10 Interestingly, under DUV excitation, this mode still exhibits a notable blueshift in methanol without ESPT [Fig. 6(c)]. This observation implies that other processes such as vibrational cooling could contribute to the frequency dynamics. The “additive” effect of ESPT and vibrational cooling leads to shorter time constants (e.g., 45 ps with 400 nm excitation vs 25 ps with DUV excitation) than the main ESPT time constant of ∼90 ps in water, while photons with higher energy cause a more dramatic blueshift magnitude of the vibrational marker band of HPTS.

Such mechanistic insights with chemical bond resolution can only be gained through an ultrafast spectroscopic toolset that can track correlated electronic and vibrational motions in a nonequilibrium PES after photoexcitation with a precisely determined time zero (see Fig. 2).^5,8 The specific structural motions of these transient marker bands, from low to high frequencies, are coupled with the photophysical or photochemical events. Therefore, the mode frequency dynamics represent an effective measure to dissect the occurrence and evolution of various processes, which are essential for the molecular stability and integrity under near and deep-UV irradiation.⁹⁸ 

Since a photoacid can donate a proton to surrounding solvent molecules, its deprotonated or basic form has no ESPT capability because the proton is already lost in the ground state. How the basic form of a photoacid behaves in the excited state remains an intriguing question. In this section, we present a systematic comparison between three different photoacids in their basic forms. Each represents a classic category in terms of the molecular twisting coordinate.

The conjugated four-ring system of HPTS eliminates any prominent twisting coordinate. The excited state processes of HPTS were investigated using femtosecond-TA and FSRS in pH = 12 aqueous solution, wherein HPTS is fully converted to the basic form.¹⁰⁵ Upon 400 nm photoexcitation, the SE band maintains its position but shows an unexpected intensity rise time constant that matches the typical solvent reorientation time of water [Figs. 7(a) and 7(b)]. These new results indicate that an efficient transition between the locally excited (LE) state and a charge transfer (CT) state occurs following electronic excitation and is governed by an ultrafast solvation event.

The excited-state Raman bands in FSRS manifest some intriguing mode-dependent dynamics [Fig. 7(c)]. Some modes reach their highest intensity within the cross-correlation time of the A_pu and R_pr pulses, while others exhibit a gradual increase. For instance, the 1155 cm⁻¹ mode consists of the ring-H rocking with small-scale ring in-plane deformation and only shows decay dynamics, indicating that it is more sensitive to the LE state. The 835 cm⁻¹ mode involves the in-plane ring asymmetric deformation with small-scale C—O⁽⁻⁾ stretching and the O—H bending motion of one adjacent water molecule [see the inset of Fig. 7(d)], whose dynamics follow a similar trend as the SE band in femtosecond-TA, inferring that this mode is coupled more to the nascent CT state (i.e., this vibration could facilitate the solvation-controlled LE → CT transition). By contrasting two distinct mode intensity dynamic patterns in Fig. 7(d), we surmise that Raman bands with common in-plane skeletal motions show the delayed intensity maxima, whereas the Raman bands with more out-of-plane motions exhibit an instantaneous rise to their intensity maxima followed by a decay process.

These new structural dynamics data reveal that the in-plane solvation (for the largely planar HPTS) contributes more to the LE → CT transition. Along the solvation coordinate, the 835 cm⁻¹ mode plays a vital role in the ultrafast barrier crossing and charge transfer processes.¹⁰⁵ In essence, the real-time investigation of the deprotonated HPTS chromophore in basic aqueous solution without any ESPT coordinate challenges the common perception of straightforward skeletal motions and solvation. This study also provides clear spectroscopic evidence (Fig. 7) for ultrafast multidimensional solvation events (coupled with characteristic vibrational motions) prior to major energy relaxation pathways (e.g., fluorescence and nonradiative energy dissipation).¹⁰⁵ 

The p-HBDI (4-hydroxybenzylidene-1,2-dimethylimidazolinone) chromophore is the “heart” of GFP. It undergoes ESPT inside the protein matrix on the time scale of a few picoseconds²¹ and thus can be considered as a photoacid. Although two dihedral angles between the phenolic (P-) ring and imidazolinone (I-) ring allow a free HBDI molecule to twist,^88,106,107 the protein environment exerts electrostatics, steric hindrance, and an intricate H-bonding network that hinder any significant twisting motions. Energy relaxation thus occurs via ESPT inside GFP, leading to green emission with high fluorescence quantum yield (FQY).^21,89,90 As an iconic system, GFP chromophore and derivatives have been central to numerous bioimaging applications to date.^21,81,89,91 However, when outside the protein matrix, the synthetic p-HBDI chromophore (see also Sec. III C) undergoes different excited-state relaxation and becomes essentially nonfluorescent [see Fig. 8(a)].^88,106–108

Both femtosecond-TA and FSRS were implemented to interrogate and visualize the twisting process of HBDI in basic aqueous solution (to avoid possible ESPT to water). The results suggest that HBDI undergoes a sub-picosecond transition from the FC region to a nearby charge-separated (CS) state, followed by a small barrier crossing to a twisted intramolecular charge transfer (TICT) state [Fig. 8(a)]. Notably, a few excited-state Raman bands that are coupled to the SE band (hence the preresonance enhancement) exhibit strong intensity oscillations within the first 2 ps upon electronic excitation. Continuous wavelet transform (CWT) of two pronounced modes at 866 and 1572 cm⁻¹ extracted a dominant low-frequency mode at ∼230 cm⁻¹ (Fig. 9), attributed to a global ring OOP deformation and bridge C—C—C OOP bending motion of the chromophore.⁸⁸ Due to its global nature that affects both the P and I-rings, this mode is anharmonically coupled to the 866 cm⁻¹ (phenolate ring HOOP with small-scale bridge HOOP motions) and 1572 cm⁻¹ (global in-plane motions involving C=C, C=N, and C=O stretches) modes on the reactive excited-state PES. The comparison between HBDI in aqueous solution vs the GFP pocket leads to two important discoveries.

First, Fourier transform of spectral oscillations for the HBDI chromophore in wtGFP retrieved a 120 cm⁻¹ phenol ring-OOP wagging motion within ∼1 ps of photoexcitation.²¹ Since the protein matrix greatly suppresses the isomerization of HBDI in favor of ESPT as an effective way for energy dissipation, the specific configuration of the proton acceptor (water) on top of the proton donor (phenolic hydroxyl end of HBDI) “selects” the 120 cm⁻¹ mode to gate the ESPT reaction.^12,21 In stark contrast, an aqueous environment favors a different ∼230 cm⁻¹ mode to facilitate isomerization in the absence of ESPT.⁸⁸ These results imply the importance of local environment in governing the anharmonic vibrational coupling between specific vibrational motions during photochemical and/or photophysical events. Further control by adding glycerol to water leads to a notable change of the modulation modes: the ∼230 cm⁻¹ mode weakens and redshifts to 212 cm⁻¹, while the ∼130 cm⁻¹ OOP mode and 276 cm⁻¹ phenolate ring in-plane deformation mode become prominent.

Second, the 120 cm⁻¹ mode of protonated HBDI in wtGFP is impulsively excited by the actinic pump, so it displays the strongest oscillation at time zero.²¹ However, the appearance of the spectral modulation caused by the 230 cm⁻¹ mode of anionic HBDI shows a clear sub-picosecond delay as reflected by two “probing” modes (Fig. 9).⁸⁸ Due to the unique PES of HBDI in aqueous solution [Fig. 8(a)], the CS state formation on the sub-picosecond time scale could correlate with ultrafast energy flow from FC-active modes into the global OOP motion at 230 cm⁻¹ although such an evolution needs to be fast enough (typically during the initial tens of femtosecond after photoexcitation) to retain the quantum coherent effects and create a delayed coherent vibrational motion.^109,110 An alternative interpretation is that the 230 cm⁻¹ mode is impulsively excited but because the nascent CS state represents a different Hamiltonian (vs the FC state) with a larger off-diagonal element (i.e., anharmonic coupling) between the modulation/tuning mode and probing/coupling modes,^5,12,111 and the apparent vibrational coupling peaks in Fig. 9 exhibit a discernible ∼500 fs delayed onset from the time zero of photoexcitation. Regardless of its exact origin, the 230 cm⁻¹ mode likely plays a functional role in breaking the two-ring coplanarity and enabling the CS → TICT barrier crossing [Fig. 8(a)] to aid the chromophore isomerization process. The HBDI chromophores with various protonation states in the condensed phase thus substantiate the importance of anharmonic couplings between vibrational modes for energy transfer and relaxation, as well as the selection of one dominant skeletal motion to facilitate the photophysical and/or photochemical reactions.^4,12

Another type of photoacid involving a twisting coordinate is 3-nitrotyrosine (3NY). A nitro group at the ortho position to the phenolic hydroxyl of tyrosine could occur in living systems due to nitro-oxidative stress and protein nitration.¹¹² Recent femtosecond-TA study and quantum calculations of anionic 3NY in water reveal that upon electronic excitation, the nitro group twists on the ∼100 fs time scale to a TICT state and then relaxes back to a hot ground state (HGS) through a slightly sloped CI between S₁ and S₀ states [Fig. 8(b)]. The PESs of anionic HBDI and 3NY in water both involve a TICT state; however, the route to this state and relevant time scales are dramatically different (Fig. 8). The nitro group in 3NY is a much smaller twisting unit than the I- or P-ring of HBDI, so the nitro twisting is more than an order of magnitude faster. Furthermore, the 3NY twisting process is barrierless, whereas a small barrier-crossing event occurs for HBDI. When both structures are optimized in the excited state using the TD-DFT method in Gaussian,³² calculations yield a twisted structure with a nitroaromatic dihedral angle of ∼100° for 3NY in the TICT state¹¹² but a planar geometry for HBDI in the FC or CS region.^88,108 Due to the large electron withdrawing capability of the nitro group in 3NY, it is directly excited into an intramolecular CT state [Fig. 8(b)], while the CT character of the FC region is not obvious for HBDI. Instead, a CS state is later formed en route to the TICT state for HBDI [Fig. 8(a)], supported by the bulkier twisting units and longer time required to achieve notable electron redistribution (e.g., CS and CT) as well as the “delayed” functional low-frequency motion in modulating characteristic high-frequency modes (Fig. 9).

After reaching the TICT state, 3NY and HBDI return to a HGS through a sloped and peaked CI, respectively. This important result is mainly due to a competition between the chromophore twisting and solvent (water) response time: a swift twisting motion is correlated with a higher tendency to encounter a sloped CI due to necessary involvement of other nuclear motions to turn an avoided crossing into a CI.^112,113 The HGS relaxation is also more than two orders of magnitudes slower in HBDI than 3NY, in accord with the size of twisting units along the main photophysical coordinate (see Fig. 8).

The capability and versatility of wavelength-tunable FSRS provide a unique tool to understand molecular properties in both the electronic ground and excited manifolds along reaction coordinates of a photosensitive chromophore (see above).^5,8 Therefore, the combined electronic and nuclear information enables rational design of various molecular machines with targeted photochemical functionalities. First of all, the FQY of HBDI needs to be greatly improved to become a viable bioprobe, and one needs to understand how the chromophore structure affects excited state dynamics and the competition between energy dissipation pathways. In particular, without structural constraints, HBDI efficiently returns to the ground state via the methine bridge twisting (see Sec. III B 2) and isomerization-induced nonradiative deactivation. Time-resolved FSRS study showed that the photoexcited anionic HBDI rapidly slides into a TICT state and reaches the ground state through a CI [see Fig. 8(a)].⁸⁸ This process occurs with a ∼2 ps time constant, so by considering a typical nanosecond fluorescence lifetime, the FQY is on the order of 10⁻⁴ [i.e., Φ = k_r/(k_r + k_nr), where k_r and k_nr denote the radiative and nonradiative decay rate constants, respectively].

Accordingly, internal locking of the methine bridge by a BF₂ group (using synthetic borylation methods) suppresses both single- and double-bond isomerization pathways and increases FQY.^14,99,114 The phenolic ring of HBDI was then fluorinated to enhance ESPT by increasing the chromophore photoacidity due to the strong electron-withdrawing ability of fluorine.¹⁴ The newly synthesized chromophores [MnF, n = 0, 1, 2, 3; see Fig. 10(a)] undergo ESPT and are highly fluorescent (FQY > 0.3).⁹⁹ In contrast, the exo-locked p-HBDI with single-bond rotation inhibited by cyclization shows a low FQY of ∼5 × 10⁻⁴ in water,¹¹⁵ suggesting that cis-trans isomerization is the major nonradiative pathway for HBDI. Nuclear motions of these molecular motors are usually accompanied by electron redistribution such as TICT formation (Fig. 8)^88,112,116 with vibrational signatures. Deeper insights into such structural dynamics will benefit the rational design of brighter fluorophores by selectively restraining the flexibility of moieties contributing to the nonradiative decay. However, the MnF series remain as green emitters, which may be a shortcoming for in vivo bioimaging due to the limited tissue penetration depth of light with short wavelengths.

Toward this goal, we have discovered a unique “double-donor-one-acceptor” molecular framework [Fig. 11(d)] to effectively achieve redder emissions by separately tuning the electronic energetics of the ground and excited states.⁹⁹ With respect to the parent molecule HBDI (see Sec. III B 2), the addition of an electron-donating group (EDG) to an adjacent site of the present donor and an electron-withdrawing group (EWG) to the acceptor moiety leads to energy increase and decrease in the electronic ground and excited states, respectively, resulting in redder emission in a highly beneficial “additive” manner. The photoinduced intramolecular charge transfer [ICT; see Fig. 8(a) for the related TICT] plays a pivotal role in this additive effect, wherein the EWG as an electron acceptor would not significantly change the electron density distribution in the ground state. In this context, FSRS offers an excellent experimental platform to characterize the electron density distribution or ICT by inspecting the vibrational (Raman) frequency change [e.g., see Fig. 11(c)], reflective of the photoinduced electronic perturbations on intrinsic molecular time scales.^5,21,34

In particular, a phenyl group was added to the C2 site at the imidazolinone ring [PnF, n = 0, 1, 2, 3; see Fig. 10(a)] to achieve larger conjugation and redder emission of the chromophore. This site-specific modification red-shifts the emission compared to MnF, but theoretical and spectroscopic studies suggest a main effect on the excited state due to ICT from the phenolic ring to the phenyl moiety.⁹⁹ TD-DFT calculations showed that the first singlet excited state results from the HOMO → LUMO excitation and exhibits CT character [Fig. 10(b)]. The LUMO of M0F also shows increased electron density on the methyl group, confirming ICT toward the C2 site.

Notably, M0F and P0F have a similar electron density distribution in HOMO where most of the electron cloud is located in the phenol-methine-imidazolinone plane, leading to a nearly identical HOMO energy of both M0F and P0F but a lower LUMO energy of P0F than M0F [top panel, Fig. 11(a)]. This result is expected if one considers π-electron delocalization based on the “particle-in-a-box” theory: the LUMO of P0F represents a larger quantum “box” than M0F, while their HOMO “boxes” are similar in size [Fig. 10(b)]. Also, we found that fluorination of the electron donor (i.e., phenolate ring) does not induce a notable change in the emission wavelength. This is consistent with the calculation results that HOMO and LUMO are stabilized to a similar extent by electrophilic substitution at those electron-rich sites [top panel, Fig. 11(a)].^99,117 The HOMOs of MnF and PnF remain isoenergetic upon fluorination regardless of the phenyl substitution at the C2 position.

To confirm the calculation results, FSRS was employed to provide vibrational signatures as indicators for local electron density in ground and excited states.^10,21 Ground-state FSRS showed no significant CT in S₀ from phenolate to imidazolinone moiety when a phenyl ring is incorporated into PnF: this is corroborated by the largely unchanged frequency of the ∼1340 cm⁻¹ mode, which is a localized imidazolinone C—N/C=N stretching and bridge-H rocking motion [Fig. 11(b)]. In comparison, this mode exhibits a notable frequency blueshift (23 cm⁻¹ for M3F, 13 cm⁻¹ for P3F) from S₀ to S₁, indicating that large-scale charge migration or ICT from the chromophore phenolate to imidazolinone moiety occurs in S₁ [Fig. 11(c)]. These new results reveal that EWG at an electron acceptor moiety primarily stabilizes the excited state and barely affects the ground state.

Further calculations indicate that the addition of EWGs (other than halogens) and EDGs to electron-rich sites at the donor ring could blue and red-shift the emission wavelength, respectively [middle and bottom panels, Fig. 11(a)]. In particular, strong EDGs destabilize the HOMO to a much greater extent than the LUMO, leading to a significantly decreased HOMO-LUMO energy gap. Essentially, one could construct a “double donor” structure (i.e., an EDG next to —O⁻ in this case) to achieve redder emission by destabilizing the HOMO. Several series of HBDI derivatives with various functional groups from —CN to —NH₂ adjacent to the —O⁻ group were synthesized and validated by this tuning strategy.⁹⁹ We thus proposed an effective red-shifting strategy: (1) construct “double donor” or two EDGs (next to each other in the push-pull π-conjugated plane) at the donor moiety to mainly increase the ground state energy; (2) incorporate an EWG at the acceptor moiety to stabilize the excited state [see green arrows in Fig. 11(a), bottom panel].

Notably, in contrast to the existing “push-pull” practices or the ICT-assisted (LUMO-centric) redshifting strategy,^117–119 our main innovation is the elucidation of new mechanistic insights that could enable the rational design for redder emission. For the first time, we proposed a new “double-donor-one-acceptor” strategy [see Fig. 11(d)] that allows the separate and synergistic tuning of electronic ground and excited states, destabilizing HOMO while stabilizing LUMO [Fig. 11(a)], thus achieving much redder emission with an efficient use of the molecular framework.⁹⁹ In other words, our targeted site-specific substitution is based on a deepened physical chemistry understanding of the electron density distribution and structural dynamics of the molecular system (aided by ultrafast spectroscopy and quantum calculations), instead of a semiempirical or synthesis-driven design process typically with steady-state spectroscopic characterization.

Second, ESPT dynamics are strongly dependent on photoacidity and H-bonding interactions between the solute and solvent.⁹⁵ Weak photoacids usually undergo ESPT only in aqueous solution. The dynamics have been discussed using the Eigen-Weller model, where a CIP is initially formed after photoexcitation followed by a diffusion-controlled proton separation into free ions.^10,62,70 Stronger photoacids or “superphotoacids” with a negative pK_a^* can transfer a proton not only to water but also to nonaqueous solvents including certain protic and basic solvents. Different from weak photoacids such as HPTS (Fig. 6 and Sec. III A), the appearance of conjugate bases is temporally delayed by a distinct solvent reorientation process for strong photoacids, which highlights the importance of H-bonding in ESPT on a wide range of molecular time scales.^5,10,95

Recently, we have engineered the GFP chromophore into superphotoacids and studied their ESPT reaction in solvents of weaker proton affinity such as methanol.^14,100 Since the locked p-HBDI derivatives are highly fluorescent and fluorination of the phenol ring enables ESPT in methanol (Sec. III C),¹⁴ the photoacidity increases with more —F groups incorporated, confirmed by the lowering of pK_a^* values from 0F to 3F [Fig. 12(a)]. The additive fluorination (1F → 3F) lowers the pK_a but has a nearly unchanged ΔpK_a, i.e., pK_a − pK_a^*. The PnF (n = 1, 2, 3) compounds exhibit the same trend except with a larger ΔpK_a than the MnF counterparts.¹⁰⁰ In comparison, other reported superphotoacids such as 5,8-dicyano-naphthol (DCN2)¹²¹ and N-methyl-6-hydroxyquinolinium (NM6HQ)¹²² have large pK_a and even larger ΔpK_a values [Fig. 12(b)], and hence a small pK_a^*. Notably, the large ΔpK_a of these superphotoacids are associated with their pronounced ICT character, especially for the anionic form (i.e., with a stronger electron donor —O⁻ than —OH group). The larger ΔpK_a values of PnF compounds than MnF are due to an enhanced ICT by the phenyl group at the acceptor moiety (see Sec. III C and Fig. 10). However, the EWGs/EDGs that are proximal to the phenol —OH or phenolate —O⁻ group mainly affect pK_a due to interactions at short distances. We thus proposed a tuning strategy [Fig. 12(b)] to effectively achieve superphotoacidity: (1) incorporate EWGs to the electron donor moiety (proximal ring bound to the —OH or —O⁻ group) to increase ground state acidity (i.e., lower pK_a); (2) incorporate EWGs to the electron acceptor moiety (distal ring) to increase ΔpK_a due to enhanced ICT.¹⁰⁰ 

Among these enhanced photoacids, M3F and P3F are the two strongest superphotoacids (−5.0 and −5.5 pK_a^* values, respectively) and can transfer a proton efficiently to methanol. Time-resolved femtosecond-TA and FSRS results reveal the ultrafast formation of a CIP-like S₁ complex on the ∼300 fs time scale followed by further PT dynamics.¹⁰⁰ As an example, Fig. 12(c) displays the spectral evolution of femtosecond-TA and FSRS for P3F in methanol. They clearly show the disappearance of the photoexcited acid at early times and appearance of the deprotonated product, the conjugate base, at later times. Global analysis of the femtosecond-TA spectra reveals two processes after initial CIP formation and before fluorescence on the nanosecond time scale: a fast process (5–6 ps) and a slow process (∼300 ps). Further insights have been provided by the FSRS mode-dependent dynamics [Fig. 12(c), right panel with prominent transient Raman modes above 1000 cm⁻¹], which show that more collective nuclear motions are required beyond the first solvation shell to control the ESPT rate in alcohols. All the reactant modes (e.g., 1355, 1553 cm⁻¹) decay with a dominant sub-picosecond time constant, supporting the formation of an initial CIP-like S₁ complex before crossing the solvation-imposed PT barrier. The product modes of the conjugate base (1184, 1414 cm⁻¹) exhibit rise dynamics with two components: the fast process (∼8 ps time constant) largely agrees with femtosecond-TA and involves solvent reorientation in response to the photoinduced dipole moment change,^10,95 and the slow process time constants span from ∼90 to 300 ps for further proton separation.^70,100

Such inhomogeneous dynamics is reminiscent of the anisotropic solute-solvent interactions of an asymmetric solute molecule and likely involves the slow solute (i.e., photoacid) rotational diffusion-controlled PT along with the fast-responding solvent reorientation. Moreover, several Raman mode frequencies blueshift on similar time scales, implying that the PES is stabilized and reshaped to minimize dipolar interactions between the photoacid and solvent. This mechanism is in accord with the continuous SE band redshift in femtosecond-TA spectra on ultrafast time scales [Fig. 12(c), left panel]. In contrast, M3F and P3F in water do not show a rotational diffusion step: ESPT is completed during the solvation process.¹⁰⁰ These recent ultrafast spectroscopic results highlight the importance of intermolecular H-bonding configurations in affecting the ESPT of photoacids in different solvents, which will guide further development of bioprobes and biosensors.

Besides the solution phase, the interior of a protein represents another important condensed-phase environment to track correlated electronic and structural motions of organic chromophores. Different from GFP with a primary function to absorb blue light and glow green,^21,81,91 photoactive yellow protein (PYP) from Halorhodospira halophila is a small, 125-residue protein that functions as a blue light receptor and triggers a negative phototactic response in bacteria.^123,124 PYP is a model system to study photochemistry, chromophore dynamics, protein folding, and signal state formation. Therefore, extensive research efforts have been devoted to PYP,^125–144 reaching a level comparable to the structurally different but mechanistically similar photosensory protein rhodopsins. A major research aim remains to track the position of all atoms upon light activation (i.e., electronic excitation) of the protein chromophore on intrinsic time scales.^{123,140–142,145}

FSRS is a promising tool to study PYP, since the structural motions associated with ultrafast isomerization can be resolved while different resonance conditions of various intermediates can be exploited (see Sec. II C). This strategy allows the selective probing of free chromophores in solution¹³⁸ (also see Secs. III A–III D) and chromophores within the protein binding site,^43,139,143 hence exposing the chromophore-solvent or chromophore-protein interactions for investigation. Both FSRS^43,139 and impulsive Raman (TR-ISRS with all femtosecond pulses) techniques¹⁴³ were used to study PYP and its Glu46Gln (E46Q) mutant. The photocycle governing PYP is driven by the weakly fluorescent (FQY = 1.4 × 10⁻³) trans-p-coumaric acid (pCA) chromophore bound to the protein via a thioester linkage to a nearby cysteine residue (Cys69).¹²⁵ 

Absorption at ∼445 nm initiates the PYP photocycle that can be described in three broad steps (Fig. 13): initial trans-to-cis isomerization of the photoexcited pCA^–, protonation and reisomerization to form the signaling state, and ground state recovery.^123,124,136 Local structural changes, driven by a series of intermediates, lead to global protein rearrangement during signaling. These later states and their lifetimes have been inferred from pH,¹²⁹ isotope-dependent studies,¹³² electronic spectroscopy,^{125,126,128,146} vibrational spectroscopy including IR and resonance Raman (see Sec. II E),^{131,137,147,148} (¹H-)NMR spectroscopy,¹²⁷ and crystallography.^{130,140–142,144,145} An early resonance Raman study helped establish that the anionic chromophore is stabilized in the protein.¹⁴⁹ By using time-resolved resonance Raman (TR³), Pan and Unno separately characterized the vibrational signatures of later intermediates such as the deprotonated cis form (I₁, pR, λ_max = 465 nm) and the reprotonated state (I₂, pB, λ_max = 355 nm) at room temperature with precisely controlled sample states in a flow or spinning cell.^135,150

The photoinduced processes of pCA^– chromophore in solution on a short time scale (<10 ps) were studied using resonance UV actinic and Raman pump pulses.¹³⁸ The evolution of vibrational modes from 600 to 1800 cm⁻¹ on the sub-picosecond time scale tracks FC relaxation, and the S₁ lifetime was found to be 2.4 ps. The spectral similarity between the ground and a relaxed S₁ (ππ^*) state, especially in the ethylenic stretch region (∼1600 cm⁻¹), suggests that the chromophore maintains a largely planar trans configuration. The ∼2 ps lifetime was observed in small-molecule studies¹⁵¹ and in proteins.^131,136 This process was attributed to the formation of a mostly cis chromophore intermediate in S₀ (named I₀, Fig. 13) following charge redistribution from the phenolate ring. Much like the HBDI chromophore in GFP (see Secs. III B–III D),^21,87,88,107 H-bonding and other electrostatic interactions within the protein matrix play a significant role in photochemistry. However, in sharp contrast to anionic HBDI,^21,88,89,106 free pCA^– chromophore in solution displays a comparable isomerization yield to the protein chromophore.^{125,134,138,151} To this regard, FSRS provides a unique means to dissect which structural motions aid efficient entry into the photocycle as an ultrafast in-plane deformation could bring the PYP chromophore back to S₀ in the trans configuration. Meanwhile, the ethylenic torsional motions (favored in protein environment) could facilitate isomerization and lead to the eventual phototactic response.^{134,137,138,148}

On longer time scales (e.g., during the first 300 ps of the photocycle), Mathies and co-workers performed FSRS experiments¹³⁹ to track the excited-state intensity and frequency dynamics of the pCA⁻ chromophore carbonyl stretch, carbonyl out-of-plane motion, and the phenol HOOP mode. They concluded that the chromophore adopts a distorted cis conformation (assigned to I₀, Fig. 13) on a 2.6 ps time scale. The observed kinetic intermediates (Fig. 14) reveal an initial charge redistribution in the FC region, followed by a relaxed excited state¹³⁸ and the appearance of the distorted I₀ state that may drive the remaining PYP photocycle. Notably, frequency redshift of the carbonyl out-of-plane mode from ∼665 to 640 cm⁻¹ (spectral evolution shown in Fig. 14) implied that successful entry into the photocycle requires breaking an H-bond to the Cys69 backbone, as observed in previous studies.^131,137 Such a mechanism that the isomerization originates near the carbonyl is not very surprising as the phenolate moiety on the other end of the chromophore forms strong H-bonds with nearby Glu46 and Tyr42 residues in the PYP pocket [see Fig. 15(a)].^136,144

FSRS can also be used to study the impact of local H-bonding network on PYP chromophore. The ultrafast charge redistribution was correlated with the weakened H-bonds between pCA⁻ and nearby residues.¹³⁹ An earlier UV resonance Raman study tracked the Tyr42 contributions to the H-bonding network, showing that PYP in the excited state transitions from a low-barrier hydrogen bond (LBHB) between Glu46 and the chromophore to a weaker bond that later recovers.¹⁴⁷ Later, both Stokes and anti-Stokes FSRS (see Sec. II B) were used to track the ∼1555 cm⁻¹ mode intensity of the chromophore (involving vinyl bond C=C stretching and phenolate ring vibrations; see Figs. 13 and 14) in wtPYP and the E46Q mutant that has weaker ground state H-bonding.⁴³ Rearrangement of the H-bonding network in wtPYP was found to occur in the excited state within ∼150 fs, consistent with the assignment to a LBHB between Glu46 and the pCA^– chromophore.

A recent TR-ISRS report showed a ∼100 fs decay of a coherently excited 135 cm⁻¹ intermolecular mode in PYP [Fig. 15(b)] associated with the chromophore and nearby residues [Tyr42 and Glu46 in Fig. 15(a)].¹⁴³ Since the same low-frequency H-bonding marker band exhibits dramatically decreased intensity in the weakly H-bonded E46Q mutant, the sub-picosecond dynamics was attributed to ultrafast charge migration resulting from perturbation of the LBHB [Fig. 15(c)]. This low-frequency mode was also observed in a femtosecond pump-probe experiment with a slight redshift of ∼1 cm⁻¹ upon H/D exchange.¹⁴³ Interestingly, a similar ∼135 cm⁻¹ mode was observed in wtPYP using femtosecond fluorescence spectroscopy,^133,152 wherein the 2004 report assigned the mode to an intramolecular vibration and proposed that the motion effectively triggers isomerization as the initiating step of PYP’s photocycle.¹³³ A recent paper by Oltrogge and Boxer studied the putative LBHB between the S65T/H148D-GFP chromophore and Asp148 to explore the consequences of proton affinity matching across this bond by halogenation at the phenolic ring. They proposed a short ionic H-bond without significant proton delocalization that may be relevant for PYP.¹⁵³ 

As an interesting comparison, the intermolecular H-bond stretching mode between a photoacid HPTS (Sec. III A) and a water molecule in aqueous solution was observed at ∼180 cm⁻¹ by tunable FSRS and vibrational coupling analysis.¹⁰ Another example concerns the heme protein myoglobin (Mb). Besides focusing on vibrational modes coupled to the resonant absorption bands of Mb with the TR³ technique,^154,155 the low-frequency heme Fe–proximal histidine stretching (∼220 cm⁻¹) and heme doming (∼75 cm⁻¹) motions¹⁵⁶ were studied using femtosecond-TA spectroscopy that identified key “reactive” underdamped skeletal motions in complex biomolecules (see Fig. 9 and Sec. III B 2 for related discussions).¹⁵⁷ Such nuclear motions could play functional roles along multidimensional photophysical and/or photochemical reaction coordinates in condensed phases, particularly for the highly diverse yet functional biological systems as discussed above.^5,10,12,21

Using the TR-ISRS approach^143,158 that resembles a purely time-domain version of FSRS,^4,8,10,12 Kuramochi et al. observed the 135 cm⁻¹ mode intensity drop but not its anharmonic coupling to higher frequency motions (they used a 290-fs A_pu),¹⁴³ nor does the mode appear to survive beyond 2.5 ps [the approximate time of I₀ formation from P^*; see Figs. 13 and 15(b)]. While Creelman et al. did not find evidence of vibrational modes between ∼400—1650 cm⁻¹ (Fig. 14) coupled to a low-frequency motion, the observed FSRS peaks were broad and overlapped in part due to the Raman pump pulse duration (590 fs).^53,139,159 Therefore, PYP represents an interesting case where TR-ISRS and FSRS results agree well in that (1) the I₀ intermediate forms on a ∼2.5 ps time scale, (2) I₀ adopts a distorted cis form although the disruption of the carbonyl H-bond to the backbone is still debated, and (3) no vibrational modulating mode was observed (until 2019) despite time-resolved electronic studies tracking coherent oscillations;^133,152 however, both ultrafast Raman techniques found that the H-bonding network of PYP and charge distribution undergo substantial changes on the femtosecond to picosecond time scales with vivid structural dynamics insights.^{43,138,139,143}

Modern energy and biological sciences, including engineering efforts, demand an improved understanding of molecular machines from the bottom up. Among them, photosensitive chromophores in condensed phases including solution and biomolecular systems have gained increasing interest and momentum due to their myriad applications in bioimaging, biomimetics, bioengineering, optogenetics, and optoelectronics. This perspective highlights several important groups of intercorrelated organic chromophores mainly with C, N, O, F, and H atoms, which could add a welcoming sustainable chemistry dimension to this line of inquiry. With technical advances in ultrafast lasers and spectroscopic toolsets, more powerful structural dynamics methods have been implemented to elucidate the working mechanisms of photoinduced “nanomachines” in condensed phases. From weak to strong photoacids, the higher-lying electronic states and basic forms have been studied to yield new information complementary to the typically accessed S₁ state and neutral forms. Equipped with deepened structural dynamics insights, two effective strategies in tuning the fluorescence emission colors, brightness, and ESPT capabilities (i.e., photoacidity) by the targeted functional group substitutions have been demonstrated for GFP chromophore derivatives. Since the excited-state FSRS with dynamic resonance enhancement has been breaking new grounds for advanced spectroscopic characterization of various chromophores, challenges and opportunities co-exist for the broad community to develop FSRS into a readily available and broadly tunable structural dynamics toolset with more exciting applications toward optogenetics, molecular motors, photocatalysts, and other light-controlled nanomachines.

It is noteworthy that FSRS enables the elucidation of multidimensional PESs after electronic excitation via two unique attributes: (1) dynamic resonance enhancement of transient molecular species in situ and in operando during a photoinduced process, while wavelength tuning of Raman pump can be guided by the femtosecond-TA experiments, and (2) direct observation of vibrational coupling in the reactive excited states and identification of functional modes, beyond the commonly observed vibronic coupling and the associated degrees of freedom. Earlier studies have shown that molecular systems generally adopt one dominant vibrational mode as the “gating” motion to facilitate photophysical and/or photochemical reactions,^{5,10,12,21,22,143} which is responsible for the directionality and efficiency of an intricate molecular Hamiltonian and anharmonic coupling matrix to generate the product state (i.e., without extensive detours or long-time search of reactive phase space).^5,12 The relevant 2D Raman spectroscopy and vibrational coupling map (i.e., frequency-frequency correlation map)⁵ have been achieved using tunable FSRS at preresonance conditions for wtGFP²¹ and the HBDI chromophore in aqueous solution,⁸⁸ and more recently, using TR-ISRS with sub-7-fs pulses (on resonance with the SE band) for wtPYP via specific cross peaks.¹⁶⁰ Comparisons between these two approaches in the frequency or time domain mainly concern resonance conditions, data acquisition time and processing methods, detection window, sample stability, etc.^5,160,161 Such femtosecond Raman studies providing new knowledge about molecular reactivity and photochemistry in condensed phases are highly complementary to other time-resolved techniques such as TRIR, ultrafast fluorescence, and third-order nonlinear spectroscopies.

Future development in this area could entail the following: (1) generation of stable phase-locked sub-10-fs laser pulses to allow the impulsive excitation of coherent nuclear motions with higher vibrational frequencies,^8,12 (2) generation of a more broadly tunable picosecond pulse as the Raman pump in the mixed time-frequency domain FSRS, aiming to cover the currently challenging spectral window gap (e.g., 410–470 nm) and also extending into the deep-UV region,^162,163 (3) optimization of resonance Raman conditions and data collection schemes to track the excited state reaction species free from the ground state species, in tandem with an effective suppression or removal of the spectral baseline,^164,165 and (4) extension into more functional or naturally occurring systems with intricate energy transfer pathways such as those in light-harvesting complexes,^166,167 as well as other solid-state or thin-film devices. The structural dynamics information on the femtosecond-to-picosecond time scales and the associated anharmonic coupling matrix of the molecular Hamiltonian, provided by time-resolved FSRS and corroborated by transient absorption and advanced quantum calculations, will enable a more complete and unified understanding of the nonequilibrium reaction mechanisms and pertinent energy states in physical, chemical, and biological systems.

One exciting direction of protein structural dynamics concerns the phototransformable molecular systems including FPs, specifically the tracking of chromophore species throughout the transformation. For instance, the bright, monomeric, reversibly photoswitchable GFP variant Dreiklang that operates by a reversible hydration/dehydration reaction was first reported in 2011,¹⁶⁸ followed by a recent femtosecond-TA study showing ultrafast deprotonation at the chromophore phenolic end ∼100 fs after 405 nm photoexcitation that involves coherent oscillations.¹⁶⁹ Our recent findings using a combination of femtosecond-TA, tunable FSRS, and quantum calculations on a photoswitchable monomeric teal FP revealed a multistep photoactivation; a crucial intermediate after chromophore deprotonation and half-way isomerization¹⁷⁰ was found before passage through an S₁/S₀ CI. More structural dynamics studies, in conjunction with transient absorption experiments to establish a self-consistent excited-state potential energy landscape, are envisaged to deepen the current understanding of molecular motors¹⁷¹ and optical switches in various environments.

We hereby propose the following experimental strategies. For systems that transform more slowly on time scales longer than a nanosecond but shorter than a millisecond (e.g., 1 kHz laser repetition rate corresponds to 1 ms detection window for each pulse),^5,10 excited-state FSRS could still be used to reveal the primary structural events that contribute to or govern the subsequent transformation, as well as potential photoproduct species that differs from the original S₀ species. For even longer processes (e.g., pH-induced aqueous aluminum nanocluster formation¹⁷² and UV-light-induced organometallic complex reaction in solution¹⁷³) on the minutes time scale or beyond, ground-state FSRS could be readily used to monitor time-resolved vibrational features in the thermally equilibrated state. In this case, the much higher peak density of the picosecond Raman pump than a cw pump (in spontaneous Raman) leads to much higher sensitivity to molecular vibrations being recorded as the stimulated Raman scattering signal (see Fig. 2).^5,172

Another research front concerns the proton-coupled electron transfer (PCET) that involves motions of two low-mass particles in a correlated manner.^93,174,175 With the knowledge gained from prior excited-state FSRS works on ESPT reactions (see Sec. III D)^{5,10,22,23,100,176} or H-bonding dynamics,^{10,105,143,177} tunable FSRS could be further employed to investigate the photoinduced PCET, wherein the electron and proton motions are strongly coupled.^178,179 New knowledge about the underlying mechanisms, if obtained with sufficient atomic details and time resolution in the condensed phase, will have far-reaching impacts on the design of renewable and sustainable energy sources such as artificial photosynthesis systems. For instance, photosynthesis II represents an excellent example of photoinduced PCET;^92,93 however, the artificial design efforts have been hindered due to the lack of detailed mechanistic insights into these highly efficient light-driven processes.^179,180 To date, the experimental work to study photoinduced PCET, particularly the ultrafast concerted pathway, remains sparse^178,181,182 with respect to theoretical efforts that have made notable advances.^179,180,183

A major experimental challenge is associated with a rigorous characterization of photoinduced PCET on ultrafast time scales down to the femtosecond regime, which is intrinsically different from the thermally driven PCET due to the nonequilibrium nature of the excited state(s). Nevertheless, key features such as time scales of electron transfer and proton transfer, the existence of distinct intermediates, and kinetic isotopic effect are still important attributes of a PCET reaction. With simultaneously high spectral and temporal resolutions (Sec. II A), FSRS is poised to become a powerful toolset to collect the nonequilibrium frequency and intensity data with robust analysis (like those representative examples in Sec. III) and provide previously elusive knowledge about the reactant, product, and intermediate (for sequential PCET) states. A recent work using ultrafast Raman loss spectroscopy (Sec. II B) has shed some light on an organic photoacid-external base adduct undergoing a photoinitiated PCET-like process on the ∼500 fs time scale.¹⁸⁴ Looking forward, the dynamic resonance enhancement (Sec. II C) should be exploited to uncover characteristic nuclear motions that are intrinsically coupled to electronic transitions, e.g., the ESA or SE bands of the transient ICT or electron-proton transfer (EPT) state, for various sample systems. Pertinent focus will be on the elucidation of different electronic and nuclear structures as well as the concerted or sequential nature of intramolecular electron transfer and intramolecular or intermolecular proton transfer dynamics (likely on the ∼100 fs time scale),^{5,62,95,175,179} aided by the gradually increased theoretical understanding and modeling of the nuclear quantum effects of protons [e.g., within the nuclear-electronic orbital (NEO) framework].^183,185

Last but not least, the synergistic implementation of complementary techniques is paramount in establishing a solid and more complete understanding of the complex structural dynamics problems at hand. Although a typical research lab would not have all the X-ray, NMR, electronic and vibrational spectroscopic tools in one setting, the literature and collaborations can always pave a path moving forward. One area that could greatly substantiate the experimental observations and advance the detailed elucidation of coupled electronic and vibrational motions of photosensitive chromophores is theoretical development and computational chemistry in the following directions (not an exhaustive list) to (1) simulate the FSRS signal based on the response functions of molecular systems under incident laser irradiation, leading to accurate spectral lineshapes at various resonance conditions across electronic states, (2) predict the excited state Raman spectrum starting from the FC region and involving relevant higher-lying electronic states,¹⁸⁶ (3) model the chromophore-surrounding interactions in both equilibrium and nonequilibrium states as well as in solution and biomolecular environments with reduced computation costs, (4) correlate spectral results through a CI with the highly coupled electronic and nuclear motions along internal conversion coordinates, identifying the prominent gating, cooling, or probing modes, and (5) use current results and machine learning to perform de novo design of chromophores with improved or new properties, drawing strength from the nature-inspired bioorganic synthesis,^99,114 metal-organic complexes with targeted substitutions,^187–190 systematic mutagenesis, and the noncanonical amino acid methodology.^191,192

In essence, the knowledge, skills, discussions, and perspectives in this contribution are envisioned to offer an extra “kick” in time-resolved vibrational spectroscopy because the research areas are vast, pertinent knowledge is expanding, and research questions keep pressing. With physical chemists, chemical physicists, laser spectroscopists, and biophysicists alike joining forces to make great strides in the biological and materials sciences, the future will be brighter with more powerful bioprobes, biosensors, molecular motors, and nanomachines, all improved by design and controlled by light.

The support by NSF CAREER Grant (No. CHE-1455353), NSF Grant (No. MCB-1817949), and the Oregon State University Faculty Research Start-Up Grant to C.F. is gratefully acknowledged. The authors thank all the Fang Group members since 2010, particularly Dr. Weimin Liu, Dr. Liangdong Zhu, Dr. Fangyuan Han, Dr. Breland Oscar, Dr. Yanli Wang, Sean Tachibana, and Taylor Krueger. The authors also appreciate the support from all our collaborators, particularly the Campbell Group at the University of Alberta (Canada) and Baranov Group at the Institute of Bioorganic Chemistry (Russian Academy of Sciences) for some of our recent work discussed herein.