A comprehensive picture of the ultrafast excited-state dynamics of retinal

Abstract

All-trans retinal is the chromophore of microbial rhodopsins initiating energy conversion and cellular signalling by subpicosecond photoinduced switching. Here, we provide detailed UV-Vis transient absorption experiments to disentangle the complex photochemistry of this polyene, which is governed by its terminal aldehyde group. After photoexcitation to the S₂(¹B_u⁺) state, the system exhibits polarity-dependent branching, populating separate S₁(¹A_g⁻) and intramolecular charge transfer (ICT) species. In all solvents, population of a singlet nπ* state from S₁ is observed which represents the precursor of the T₁ triplet state. While triplet formation dominates in nonpolar solvents (67% quantum yield), it is dramatically reduced in polar solvents (4%). The channel closes completely upon replacing the aldehyde by a carboxyl group, due to an energetic up-shift of ¹nπ*. In that case, internal conversion via the ICT species becomes the main pathway, with preferential formation of the initially excited isomer.

---

1. Introduction

Derivatives of retinal play a key role as chromophores in rhodopsin photoreceptor proteins and fulfil diverse functions, such as energy conversion, cellular signalling, vision and photopigment regeneration.^1–3 In microbial rhodopsins, all-trans retinal (= all-trans β-apo-15-carotenal, compound 1 in Fig. S1, ESI†) is connected as a protonated Schiff base (PSB) to an opsin apoprotein. Upon photoexcitation, the all-trans PSB undergoes excited-state photoisomerisation to the 13-cis isomer (compound 2 in Fig. S1, ESI†). Characterising light-triggered processes of retinyl chromophores and the influence of the environment on their dynamics is therefore of prime importance to understand their function in vivo.

In the simplest model, the photoinduced dynamics of carotenoids involve three singlet electronic states: the S₀(¹A_g⁻) ground state, the one-photon-forbidden S₁(¹A_g⁻) first excited state, and the optically “bright” S₂(¹B_u⁺) second excited state absorbing in the blue-green spectral region. In apocarotenals, the conjugated terminal aldehyde group alters the electronic properties drastically, especially those of the S₁ state. For instance, systems such as β-apo-12′-carotenal, show a strong decrease of the excited-state lifetime with increasing solvent polarity, in contrast to nonpolar carotenoids.^4–7 This is ascribed to the formation of considerable intramolecular charge transfer (ICT) character in the S₁ state.^8–10

Earlier studies suggest that the photochemistry of retinal is more complex. For instance, Yamaguchi and Hamaguchi performed ultrafast transient absorption spectroscopy of all-trans retinal in n-hexane and evaluated their data based on a kinetic model involving branching in the S₁ state:¹¹ after S₂ → S₁ internal conversion (IC), S₁ is depopulated via the “¹nπ*/triplet channel” S₁ → ¹nπ* → T₁ → S₀, forming exclusively the all-trans isomer¹² with 74% quantum yield, and via the “photoisomerisation/IC channel” S₁ → p* → p → S₀ involving perpendicular (90°) excited-state and ground state configurations p* and p, which lead to formation of trans and cis isomers. In a follow-up study, they reported transient absorption spectra in n-butanol,¹³ where the lifetime of the S₁ state was 1.6 ps, and thus slower than in n-hexane (0.56 ps). In n-butanol, stimulated emission (SE) was observed above 620 nm, in agreement with SE kinetics previously detected in other polar solvents.^14,15 Later on, Polívka et al.⁷ ascribed this red emission to an ICT state, based on an analogous assignment of SE features observed for β-apo-12′-carotenal, β-apo-8′-carotenal^5–7 and β-apo-12′-carotenoic acid.¹⁶

Although these experiments have improved the understanding of retinal photochemistry, important pieces of information are still missing, such as (a) the complete spectral characterisation and ultrafast dynamics of the ¹nπ* state in different solvents, (b) the determination of solvent-dependent lifetimes and quantum yields for the processes in the ¹nπ*/triplet and photoisomerisation/IC channels, (c) the possible influence of ICT character in the excited electronic state manifold, and (d) an explanation for the fast excited-state decay of retinal which is a priori unexpected for such a short-chain polyene species. We will address all of these points and arrive at a comprehensive description of the excited-state dynamics of all-trans retinal, with the aid of complementary experiments for the closely related all-trans retinoic acid (compound 3 in Fig. S1, ESI†) and 13-cis retinal.

2. Experimental

2.1 Substances

All-trans retinal and all-trans retinoic acid were kindly provided in double-recrystallized form by BASF SE (Prof. Bernd Schäfer), 13-cis retinal was purchased from Toronto Research Chemicals and used without further purification. Purities were >97% for the all-trans retinoids, 95% for 13-cis retinal, and 99% or better for the organic solvents.

2.2 Pump – supercontinuum probe (PSCP) spectroscopy

Ultrafast broadband transient absorption experiments were carried out using UV-Vis pump – supercontinuum probe spectroscopy at 920 Hz repetition frequency.^17–19 All-trans retinal was excited by 400 nm pulses at every second laser shot (laser pulse fluence ca. 0.23 mJ cm⁻²). Extended spectral coverage into the deeper UV region was achieved by employing a 400 nm-seeded multifilament supercontinuum generated in a 2 mm thick translating CaF₂ plate as probe pulse (260–660 nm, beam diameter ca. 130 μm). The relative pump–probe polarization was set at magic angle (54.7°). The pump–probe intensity cross-correlation time varied in the range 57–84 fs in the different experiments, with a time zero accuracy of ca. 10 fs. To avoid any influence of photoinduced sample degradation, 10–15 mL of a nitrogen-saturated solution of all-trans retinal, 13-cis retinal or all-trans retinoic acid in the organic solvent of interest was passed through a flow cell (path length 1 mm, window thickness 200 μm), so that the sample volume was exchanged after each pump–probe cycle. Solutions with an OD of 0.2 (at 400 nm, 1 mm path length) were employed. Under these conditions, the transient absorption signals were in the linear regime, as confirmed in separate experiments (Fig. S2–S7, ESI†). Steady-state absorption spectra of the sample solutions were recorded on a Varian Cary 5000 dual-beam spectrometer.

2.3 Data processing and global kinetic analysis procedure

The PSCP spectra were chirp-corrected by using the coherent response of the respective pure solvents. The coherent solvent signals were in all cases subtracted from the transient absorption data of retinal to obtain the pure chromophore dynamics (Fig. S8, ESI†).

Each data set, consisting of 512 kinetic traces, was subjected to a global modelling procedure employing the kinetic scheme discussed later on. Species-associated spectra were parametrised by a sufficiently large number of Gaussian functions. During the modelling, the known, separately measured absolute S₀ absorption spectra for the trans/cis isomers of retinal and retinoic acid were kept fixed. In addition, the following boundary conditions based on existing experimental data were imposed: total isomerisation quantum yields of Φ_cis = 12% (n-hexane, n-hexadecane) and Φ_cis = 33% (ethanol), based on the measurements of Waddell et al. and Ganapathy and co-workers,^20,21 were employed with an allowed variation of ±15%. Note that this only slightly influences the amplitude of the T₁ triplet spectrum in the overlap region with the S₀ ground state bleach (GSB), with the T₁ peak being unaffected. Next we assumed that all-trans retinal does not isomerise in the triplet channel, as found in independent transient FTIR experiments of Yuzawa and Hamaguchi.¹² Finally, absolute peak absorption coefficients (all in units of M⁻¹ cm⁻¹) for the T₁ triplet band are available from pulsed laser photolysis experiments of all-trans retinal in a few solvents, albeit partially with either unknown or substantial uncertainties: 70 000 (n-hexane),^22–25 78 600 (THF),²⁶ 68 000 (methanol)^26,27 and 63 000 (acetonitrile).²⁸ During the fitting procedure, the peak value of the fitted T₁ spectrum was forced to stay within ±10% of these values.

The parameters, such as the position, width and height of the Gaussian functions describing the excited-state spectra and the rate constants, were optimised simultaneously to arrive at the best fit, which included convolution with the experimental time response, the latter one being determined from the pure solvent signal.

2.4 DFT/TDDFT calculations

Calculations for all-trans retinal and all-trans retinoic acid were performed based on the procedures described in our previous publications:^29–31 the structure of the ground electronic state was optimized using DFT employing the B3LYP functional and a 6-31G(d) basis set. The energies of the three lowest excited states (vertical excitation) were determined by TDDFT using SVWN, BLYP, B3LYP, MPW1K and CAM-B3LYP functionals and a 6-31+G(d) basis set.

3. Results and discussion

3.1 Ultrafast dynamics of all-trans retinal and all-trans retinoic acid in different solvents

UV-Vis steady-state absorption spectra of all-trans retinal and 13-cis retinal in various solvents are summarised in the ESI† (Fig. S9). The S₀(¹A_g⁻) → S₂(¹B_u⁺) band maximum is located slightly below 400 nm. The spectrum of the all-trans species has a larger absorption coefficient and red-shift than 13-cis. A discussion of the solvent-dependence of the absorption spectra is included in the ESI† (Fig. S9–S13), with the main finding that several factors, such as increasing solvent polarisability, polarity and hydrogen-bonding capability lead to an increased spectral red-shift.

The 400 nm pump pulse used in the time-resolved experiments excites predominantly the S₀ → S₂ transition, with the weakly absorbing S₁(¹A_g⁻) state estimated to lie 2300 cm⁻¹ below S₂.³²Fig. 1 shows transient UV-Vis PSCP broadband absorption spectra of all-trans retinal and all-trans retinoic acid in n-hexane. Fig. 2 compares the solvent dependence of the PSCP spectra of all-trans retinal in n-hexadecane, diisopropyl ether, THF and methanol as contour plots (corresponding transient spectra are included in the ESI,† Fig. S14–S19). The initially populated S₂ state shows pronounced excited-state absorption (ESA) with a peak at 600 nm accompanied by an S₀ → S₂ GSB centred below 400 nm. Four central aspects are immediately clear upon inspection of the subsequent spectral development:

Fig. 1 Transient UV-Vis PSCP absorption spectra of all-trans retinal (A) and all-trans retinoic acid (B) in n-hexane upon excitation at 400 nm. (Top panel) 0.15–0.03 ps with 30 fs steps; (2nd panel) 0.06–0.14 ps with 20 fs steps, and 0.2, 0.3 and 0.4 ps; (3rd panel) 0.5, 1 and 10 ps; (bottom panel) 15, 30, 50, 100 and 700 ps (inset in (B): magnified spectrum at 700 ps). Selected transient spectra are shown as thick coloured lines for guidance. The blue-dashed lines in the bottom panels are the scaled and inverted steady-state absorption spectra. The corresponding contour plots for all-trans retinal (C) and for all-trans retinoic acid (D) are presented with the symbols of the electronically excited species used in the global kinetic analysis. The concentration of retinoic acid was kept low (3 × 10⁻⁵ M) to avoid any influence of dimer formation on the spectra. This leads to a reduced signal-to-noise ratio.

Fig. 2 Contour plots of transient UV-Vis PSCP absorption spectra of all-trans retinal in n-hexadecane (A), diisopropyl ether (B), THF (C) and methanol (D) upon S₀ → S₂ excitation at 400 nm. Note the gradual disappearance of the ¹nπ*/triplet channel with increasing solvent polarity. The absorption feature around 300 nm in (B) at short times (100 fs) is due to solvent contributions which cannot be removed perfectly by subtraction of the pure solvent spectrum.

(a) For retinal in n-hexane, there is the emergence of a sharp absorption band centred at 396 nm (Fig. 1(A), second panel, and (C)). It is assigned to ESA of the ¹nπ* singlet state formed after intermediate population of the S₁ state (see below). The ¹nπ* state represents the precursor of the T₁ triplet state, and the latter one is formed at later times with a characteristic band peaking at 443 nm,²⁶ see Fig. 1(A), bottom panel, and (C).

(b) The ¹nπ*/T₁ channel is closing down completely upon replacement of the terminal aldehyde group (retinal) by a carboxyl function (retinoic acid), see Fig. 1(B), second panel. There are no spectral fingerprints of the ¹nπ* and T₁ states.

(c) Spectral features of the ¹nπ* state of all-trans retinal become weaker with increasing solvent polarity, see Fig. 2(A–D) and Fig. S14–S19 (ESI†). This suggests a strong reduction of the quantum yield for ¹nπ* and T₁ formation. Still, in methanol and acetonitrile, weak ¹nπ* and well-visible T₁ spectral features are present. A comparison of the solvent-dependent band shapes is shown in Fig. 3. In (A), at 0.17 ps the ¹nπ* ESA band shows a sharp peak at 396 nm from nonpolar solvents up to diisopropyl ether. The band becomes much weaker in more polar solvents and also develops at later time (Fig. S16–S19, ESI†), already suggesting delayed formation of ¹nπ* which is confirmed by the global analysis below. In (B), the long-lived T₁ → T_n absorption develops from a narrow band in n-hexadecane/n-hexane into a broader band with a long-wavelength tail for polar solvents, such as methanol or acetonitrile.

Fig. 3 Comparison of broadband transient absorption spectra of all-trans retinal in the organic solvents n-hexadecane, n-hexane, diisopropyl ether, THF, ethanol, methanol and acetonitrile. Spectra are normalised at the maximum GSB amplitude. (A) Spectra at t = 0.17 ps: here, the spectral features of the S₁ and ICT species (above 400 nm) and the rise of the ¹nπ* absorption (around 396 nm) are observed. (B) Spectra averaged over the time range 700–796 ps: the long-lived T₁ → T_n triplet absorption band above 400 nm is clearly visible.

(d) The ESA band of all-trans retinoic acid (Fig. 1(B), second panel) featuring two peaks at 470 and 550 nm arises from an S₁/ICT species, which is populated from S₂, see below. A very similar ESA band is also observed for all-trans retinal, with peaks located at 480 and 540 nm (Fig. 1(A), second panel). The appearance of the S₁ and ICT bands of all-trans retinal changes drastically with increasing polarity: from a double-peak structure in nonpolar solvents to a pronounced single peak band with a blue shoulder and an SE feature to the red (Fig. 3(A)). Most importantly, the kinetics in this ESA band exhibit a biexponential decay (e.g. at 530 nm), strongly suggesting that two distinguishable species contribute in this spectral region which decay independently via separate relaxation channels.

3.2 Kinetic model

Several kinetic schemes were explored to model the transient spectra of retinal. The best model reproducing all details of the dynamics is found in Fig. 4(A). As a central feature of the mechanism, separate S₁ and ICT species are populated from S₂ by IC. S₁ and ICT decay independently to the ground electronic state, forming all-trans and cis isomers in S₀. The S₁ state is also the precursor of the ¹nπ* state which forms the long-lived T₁ state (all-trans isomer) by ISC. The resulting time constants and quantum yields are summarized in Table S2, ESI.† Key parameters, such as the polarity dependence of quantum yields and selected time constants are highlighted in Fig. 4(B–E). Species-associated spectra (SAS) featuring absolute absorption coefficients are shown in Fig. 5. Additional SAS, species contributions to the spectra at selected times and kinetic traces including simulations are included in the ESI† (Fig. S20–S24).

Fig. 4 Kinetic model for all-trans retinal photochemistry and key parameters of the global kinetic analysis. (A) Kinetic model: S₂ is initially populated from S_0,trans by photoexcitation. The 13-cis isomer is the predominant cis isomerisation product, and according to the experiments of Yuzawa and Hamaguchi,¹² the T₁ species has an all-trans configuration. The red pathway dominates in nonpolar solvents, whereas the orange one prevails in highly polar solvents. (B) Branching ratios from S₂ into S₁ (red) and ICT (orange). (C) Branching ratios for formation of T₁ (blue), S_0,all-trans (magenta) and S_0,cis (black). (D) Total lifetime τ₁ of the S₁ state and its contributions from the ¹nπ*/T₁ channel (τ_1,nπ*) and internal conversion channel to S_0,trans (τ_1,trans). (E) Total lifetime of ICT (orange) and ¹nπ* (green).

Fig. 5 Species-associated spectra from global kinetic analysis. (A) All-trans retinal in n-hexane, (B) all-trans retinoic acid in n-hexane, (C) all-trans retinal in THF, and (D) all-trans retinal in methanol.

We note that other kinetic schemes failed to reproduce central aspects of the spectral dynamics: for instance, a model employing branching from S₂ forming the three species S₁, ICT and ¹nπ* was abandoned, because it did not reproduce the delayed rise of the ¹nπ* band, suggesting that S₁ must act as an intermediate for ¹nπ* formation, see ESI† (Fig. S25).

Furthermore, a model featuring a single S₁/ICT state was tested for retinal. However, a single state should exhibit monoexponential decay behaviour. For instance, the build-up of population in the ¹nπ* state in nonpolar alkanes and THF (τ_1,nπ* ≤ 0.56 ps in Table S2, ESI†) then would quickly drain population from S₁/ICT. It therefore immediately turned out that the second slower component observed in the spectral decay of the ESA band cannot be modelled, see Fig. S26, ESI.† Therefore separate S₁ and ICT species are clearly identified. The obvious question then arises if the S₁ and ICT species might be in equilibrium. Based on our modelling we conclude that a “fast” equilibrium with forward and backward reactions on the picosecond timescale is not compatible with the experimental data, because the fastest channel dictates the decay of the ESA band, and it becomes impossible to model the slower decay component. Yet, we cannot exclude the presence of a slow equilibrium (with time constants of several picoseconds for forward and backward reaction). This implies the presence of an appreciable barrier on the potential energy surface between S₁ and ICT suppressing interconversion of the two species on the timescale of IC to the ground state. This further corroborates our final model in Fig. 4(A) assuming that the S₁ and ICT species are effectively separated.

Lastly, the inclusion of trans–cis isomerisation in the mechanism of Fig. 4(A) is imperative for two reasons: the “persistent bleach” in the GSB region at 355 nm for all-trans retinoic acid in n-hexane in Fig. 1(B), bottom panel (in the absence of any triplet formation), directly proves the formation of cis-species exhibiting smaller absorption coefficients. We confirmed this by recording UV-Vis steady-state absorption spectra of different retinoic acid isomers (Fig. S27, ESI†). For all-trans retinal, the situation is more complicated, because there are long-lived spectral contributions of the triplet band and its corresponding GSB contribution. Yet, earlier experiments employing HPLC analysis of primary photoproducts of all-trans retinal provided an estimate for the quantum yield of cis-isomer formation and confirmed that trans–cis isomerisation in the excited state indeed takes place, with 13-cis representing the dominant product.^20,21

3.3 Polarity-dependent excited-state dynamics of all-trans retinal

Having established the plausibility of the kinetic scheme in Fig. 4(A), we now turn toward the interpretation of the time constants and species-associated spectra obtained from the analysis. The lifetime of the S₂ state is extremely short in all solvents, e.g. 55 and 60 fs in n-hexane and n-hexadecane, respectively. This supports our previously observed trend that the S₂ lifetime of apocarotenals systematically decreases upon shortening of the polyene chain.³³ In all cases, the S₂ → S_n spectrum is characterised by a sharp peak at ca. 600 nm with a shoulder extending to shorter wavelengths.

We observe a systematic decrease of the quantum yield for S₁ formation with solvent polarity. For instance, it is 73% in n-hexane and only 43% in acetonitrile, thus population of ICT is favoured in polar solvents (Fig. 4(B)). The latter finding is in agreement with the characteristic change of the experimental ESA band shape and the appearance of pronounced SE above 600 nm from diisopropyl ether onwards (Fig. 5(C and D), orange lines, and Fig. S20, ESI†).

The species-associated spectra of S₁ (red) and ICT (orange) exhibit systematic changes with polarity: in nonpolar solvents (n-hexane, n-hexadecane), the spectra contain two broad peaks and look similar. In polar solvents, the ICT spectra show a pronounced single peak with a shoulder toward shorter wavelengths and a distinct SE feature above 600 nm. The S₁ spectra are more red-shifted and show no SE band. We therefore conclude that the S₁ species is a higher-energy local minimum of the first electronically excited state (assuming a similar energetic position of the upper state in the ESA transitions of S₁ and ICT). S₁ has access to the ¹nπ* state and is separated from the lower-energy ICT part of the potential by an appreciable barrier. The charge-transfer character of ICT is less pronounced in nonpolar solvents due to the missing dipolar solvent stabilization. Therefore the SE feature is likely hidden beneath the stronger ESA band due to its much smaller Stokes shift in nonpolar solvents.

The formation of ¹nπ* from S₁ significantly slows down with increasing solvent polarity, from 0.21 ps in n-hexane to 3.0 ps in acetonitrile and, at the same time, internal conversion to S₀ accelerates (Fig. 4(D)). These two effects lead to a dramatic reduction of the triplet quantum yield from 64% in n-hexane to 4% in acetonitrile (Fig. 4(C)). It is reasonable to assume that a polarity-induced shift of the ¹nπ*, S₁ and ICT energy levels is responsible for this effect. Most importantly, an increase of polarity tends to increase the energy of ¹nπ* species relative to S₁ and ICT.³⁴ This likely reduces the coupling between S₁ and ¹nπ* and slows down the IC rate for ¹nπ* formation, favouring IC to S₀ instead.

We completely characterise the ¹nπ* state spectra in alkanes and diisopropyl ether (Fig. 5, green lines). The band is centred at 396 nm with a shoulder toward shorter wavelengths. In polar solvents, the shape of the ¹nπ* spectrum should be taken with some caution, because its ESA band is weak (Fig. S17–S19, ESI†), due to the small quantum yield for ¹nπ* formation and its subsequent decay to T₁. Still, it is clearly visible that the ¹nπ* band is located at about the same position as in nonpolar solvents and broadens.

The time constant for T₁ formation from ¹nπ* in alkanes and diisopropyl ether is ca. 31 ps. This ISC process accelerates somewhat with increasing solvent polarity (Fig. 4(E)). The T₁ spectra in alkanes and diisopropyl ether are characterised by a main peak at ca. 443 nm (Fig. 5, blue lines, and Fig. 3), with a typical absorption coefficient of ca. 70 000 M⁻¹ cm⁻¹. In polar solvents, the T₁ spectra broaden significantly.

3.4 Influence of the type of terminal carbonyl group

Fig. 1(B and D) demonstrated that the ¹nπ*/T₁ channel is completely shut down upon replacing the terminal aldehyde by a carboxyl group. Consequently, global modelling for all-trans retinoic acid in n-hexane is possible using the SAS shown in Fig. 5(B). We note that a single spectrum for S₁/ICT suffices, because the whole ESA band decays monoexponentially. It has a shape which is similar to the S₁ and ICT species in all-trans retinal (Fig. 5(A)). The influence of the COOH group is twofold: it shifts the ¹nπ* state to higher energies, so that access from S₁ is no longer possible. Such an energetic upshift of ¹nπ* states of carboxylic acids is known for shorter conjugated systems³⁵ and was also found in our DFT/TDDFT calculations reported in Section 3.6. Secondly, the ICT character is weaker for a COOH group than for a CHO group, as suggested by our previous studies of longer-conjugated β-apocarotenals and β-apocarotenoic acids.^5,6,16,36 Therefore the excited-state ICT character of retinoic acid should be reduced and the barrier between the S₁ and ICT part should be small or even absent, leading to the monoexponential ESA decay. It also appears as if the larger energetic separation of the ¹nπ* state from S₁/ICT increases the IC lifetime for S₁/ICT → S₀ to 19 ps.

3.5 Starting from the 13-cis isomer

For further testing of our kinetic model for retinal excited-state dynamics, we carried out selected transient absorption experiments starting from 13-cis retinal. We focus on the limiting cases in nonpolar and polar solvents. Transient absorption spectra in n-hexane and acetonitrile are included in Fig. S28, ESI.† Results of the global kinetic modelling are found in Table S2, ESI.† At first glance, the transient spectra look similar to those of the all-trans isomer. Inspection of the global modelling results revealed however significant differences: the ¹nπ*/triplet quantum yield is reduced, and 13-cis formation dominates. The latter becomes immediately clear upon inspection of the transient absorption spectrum of 13-cis retinal in n-hexane: it is shifted toward shorter wavelengths (Fig. S29, ESI†). If a larger amount of all-trans isomer was formed, the triplet band should largely resemble the all-trans T₁ band, which is obviously not the case. A similar argument holds for acetonitrile (Fig. S28(B), ESI†): if an appreciable amount of all-trans isomer was formed, there should be no GSB at long times but a weak absorption feature, because of the higher absorption coefficient of the all-trans isomer (Fig. S9, ESI†). This is also not the case. There is apparently a strong propensity for the recovery of the initially photoexcited isomer. We therefore conclude that the kinetic scheme in Fig. 4(A) is also applicable to the excited-state dynamics of 13-cis retinal, only the trans and cis denominations need to be exchanged.

3.6 Comparison with DFT/TDDFT calculations

We also carried out selected DFT/TDDFT calculations for all-trans retinal and all-trans retinoic acid. Although there are well-known limitations and drawbacks of this method when applied to polyene systems with charge-transfer character,^37,38 valuable semi-quantitative information was extracted regarding the influence of the terminal carbonyl group. Transitions and oscillator strengths are summarised in Table S3 and Fig. S30 (ESI†). Detachment and attachment electron densities³⁹ for the three lowest excited singlet states are plotted in Fig. S31 (ESI†).

We focus on the shifts of the electronic transitions upon replacement of the aldehyde by the carboxyl group. The optically dark S₀ → ¹nπ* transition is clearly identified in the detachment/attachment plots by the transfer of electron density from the oxygen lone pairs into the conjugated π-system (Fig. S31, ESI†). For all functionals, the S₀ → ¹nπ* transition is strongly shifted upward by 7000–10 000 cm⁻¹ (Table S3, Fig. S30, ESI†). In contrast, the S₀ → ¹A_g⁻ and S₀ → ¹B_u⁺ ππ* transitions exhibit only a mild up-shift of 450–800 cm⁻¹. The calculations therefore provide a rationale for the switching-off of the ¹nπ*/triplet channel observed in the time-resolved experiments upon replacing the CHO by the COOH group, because the ¹nπ* state is then no longer accessible from ¹A_g⁻. The calculations also confirm the slight blue-shift observed in the steady-state absorption spectra upon replacement of the aldehyde by the carboxyl group (Fig. 1(A) and (B), bottom panels).

A closer look however reveals substantial discrepancies between experiment and TDDFT calculations. For instance, in the case of all-trans retinal, ¹nπ* is no longer the lowest excited state for the functionals B3LYP, MPW1K and CAM-B3LYP. In addition, the spacing between the ¹B_u⁺ and the ¹A_g⁻ states and their oscillator strengths are not consistent with experiment. For reaching quantitative agreement, more accurate quantum chemical calculations of these systems in different solvents will therefore be needed. This would require the accurate determination of the energetic positions of the S₂, S₁, ICT and ¹nπ* species. However currently, such calculations are very difficult, because this would require the correct treatment of doubly excited configurations in these polyenes,³⁷ a correct inclusion of the charge-transfer character introduced by the terminal carbonyl group,³⁸ as well as an appropriate consideration of local solvent effects beyond a simple polarizable continuum approach, including solvent-dependent relaxation of the excited-state structures. Note also that, so far, even high-level gas-phase CASSCF and CASPT2 calculations for retinal strongly overestimate the energy of the bright S₀ → S₂ transition.⁴⁰ Recent approaches, such as the one used by Frank, Birge and co-workers for the carbonyl carotenoid peridinin⁴¹ might be also valuable for a better understanding of the excited-state manifold of all-trans retinal and all-trans retinoic acid. In any case, we hope that the current experiments will serve as a useful benchmark for future high-level quantum chemical calculations of these systems.

4. Conclusions

The photochemistry of retinal is largely controlled by the ¹nπ* state. According to El-Sayed's rule,^42–44 only an ¹nπ* → T₁ ISC transition is sufficiently fast to efficiently compete with IC processes typically prevailing in carotenoids. We demonstrated that this channel is operative in all solvents and gradually closes with increasing solvent polarity. A complete suppression of the ¹nπ* channel happens upon changing to a terminal substituent, such as carboxyl, which pushes up the energy of ¹nπ*. Exactly this effect is also operative in the all-trans retinal Schiff base of microbial rhodopsins, as its ¹nπ* state is known to be located at higher energy than for the corresponding aldehyde.^45–50 Protonation of the Schiff base even completely removes its ability to form an ¹nπ* state.⁵¹

In all solvents, isomerisation of the initially photoexcited species is inefficient, e.g. <21% for all-trans retinal (Table S2, ESI†). Such an intrinsically low quantum yield is unwanted for retinal's function as an efficient photoswitch. In microbial rhodopsins, nature has optimized it to 64%,^52,53 implying that the protein pocket promotes much more efficient trans–cis isomerisation.¹

Acknowledgements

We acknowledge B. Schäfer from BASF SE for providing high purity all-trans retinal and all-trans retinoic acid samples for the current experiments. We are indebted to N. P. Ernsting (Humboldt University Berlin, Germany) and J. L. Pérez Lustres (University of Santiago de Compostela, Spain) as well as J. Troe, K. Luther, J. Schroeder, D. Schwarzer and A. M. Wodtke (Georg August University Göttingen, Germany) for their continuous support and advice.

References

1. O. P. Ernst, D. T. Lodowski, M. Elstner, P. Hegemann, L. S. Brown and H. Kandori, Chem. Rev., 2014, 114, 126.
2. A. Wand, I. Gdor, J. Zhu, M. Sheves and S. Ruhman, Annu. Rev. Phys. Chem., 2014, 64, 437.
3. P. D. Kiser, M. Golczak and K. Palczewski, Chem. Rev., 2014, 114, 194.
4. D. A. Wild, K. Winkler, S. Stalke, K. Oum and T. Lenzer, Phys. Chem. Chem. Phys., 2006, 8, 2499.
5. M. Kopczynski, F. Ehlers, T. Lenzer and K. Oum, J. Phys. Chem. A, 2007, 111, 5370.
6. K. Oum, P. W. Lohse, F. Ehlers, M. Scholz, M. Kopczynski and T. Lenzer, Angew. Chem., Int. Ed., 2010, 49, 2230.
7. T. Polívka, S. Kaligotla, P. Chábera and H. A. Frank, Phys. Chem. Chem. Phys., 2011, 13, 10787.
8. J. A. Bautista, R. E. Connors, B. B. Raju, R. G. Hiller, F. P. Sharples, D. Gosztola, M. R. Wasielewski and H. A. Frank, J. Phys. Chem. A, 1999, 103, 8751.
9. D. Zigmantas, T. Polívka, R. G. Hiller, A. Yartsev and V. Sundström, J. Phys. Chem. A, 2001, 105, 10296.
10. D. Zigmantas, R. G. Hiller, A. Yartsev, V. Sundström and T. Polívka, J. Phys. Chem. B, 2003, 107, 5339.
11. S. Yamaguchi and H. Hamaguchi, J. Chem. Phys., 1998, 109, 1397.
12. T. Yuzawa and H. Hamaguchi, J. Mol. Struct., 1995, 352/353, 489.
13. S. Yamaguchi and H. Hamaguchi, J. Phys. Chem. A, 2000, 104, 4272.
14. E. J. Larson, L. A. Friesen and C. K. Johnson, Chem. Phys. Lett., 1997, 265, 161.
15. E. J. Larson, S. J. Pyszczynski and C. K. Johnson, J. Phys. Chem. A, 2001, 105, 8136.
16. S. Stalke, D. A. Wild, T. Lenzer, M. Kopczynski, P. W. Lohse and K. Oum, Phys. Chem. Chem. Phys., 2008, 10, 2180.
17. A. L. Dobryakov, S. A. Kovalenko, A. Weigel, J. L. Pérez Lustres, J. Lange, A. Müller and N. P. Ernsting, Rev. Sci. Instrum., 2010, 81, 113106.
18. K. Oum, T. Lenzer, M. Scholz, D. Y. Jung, O. Sul, B. J. Cho, J. Lange and A. Müller, J. Phys. Chem. C, 2014, 118, 6454.
19. L. F. Tietze, B. Waldecker, D. Ganapathy, C. Eichhorst, T. Lenzer, K. Oum, S. O. Reichmann and D. Stalke, Angew. Chem., Int. Ed., 2015, 54, 10317.
20. W. H. Waddell and K. Chihara, J. Am. Chem. Soc., 1981, 103, 7389.
21. S. Ganapathy and R. S. H. Liu, J. Am. Chem. Soc., 1992, 114, 3459.
22. B. Veyret, S. G. Davis, M. Yoshida and K. Weiss, J. Am. Chem. Soc., 1978, 100, 3283.
23. R. Bensasson, E. J. Land and T. G. Truscott, Photochem. Photobiol., 1973, 17, 53.
24. T. Rosenfeld, A. Alchalel and M. Ottolenghi, J. Phys. Chem., 1974, 78, 336.
25. R. Azerad, R. Bensasson, M. B. Cooper, E. A. Dawe and E. J. Land, presented on the Int. Conf. at the Calouste Gulbenkian Foundation Centre, Lisbon, Portugal, 1974, in: Excited States of Biological Molecules, ed. J. B. Birks, Wiley (London), 1976, p. 531.
26. W. Dawson and E. W. Abrahamson, J. Phys. Chem., 1962, 66, 2542.
27. M. M. Fisher and K. Weiss, Photochem. Photobiol., 1974, 20, 423.
28. P. K. Das and R. S. Becker, J. Am. Chem. Soc., 1979, 101, 6348.
29. K. Golibrzuch, F. Ehlers, M. Scholz, R. Oswald, T. Lenzer, K. Oum, H. Kim and S. Koo, Phys. Chem. Chem. Phys., 2011, 13, 6340.
30. K. Oum, P. W. Lohse, J. R. Klein, O. Flender, M. Scholz, A. Hagfeldt, G. Boschloo and T. Lenzer, Phys. Chem. Chem. Phys., 2013, 15, 3906.
31. K. Oum, O. Flender, P. W. Lohse, M. Scholz, A. Hagfeldt, G. Boschloo and T. Lenzer, Phys. Chem. Chem. Phys., 2014, 16, 8019.
32. R. R. Birge, J. A. Bennett, L. M. Hubbard, H. L. Fang, B. M. Pierce, D. S. Kliger and G. E. Leroi, J. Am. Chem. Soc., 1982, 104, 2919.
33. F. Ehlers, M. Scholz, J. Schimpfhauser, J. Bienert, K. Oum and T. Lenzer, Phys. Chem. Chem. Phys., 2015, 17, 10478.
34. T.-i. Lai, B. T. Lim and E. C. Lim, J. Am. Chem. Soc., 1982, 104, 7631.
35. J. D. Coyle, Chem. Rev., 1978, 78, 97.
36. P. W. Lohse, R. Bürsing, T. Lenzer and K. Oum, J. Phys. Chem. B, 2008, 112, 3048.
37. J. H. Starcke, M. Wormit, J. Schirmer and A. Dreuw, Chem. Phys., 2006, 329, 39.
38. A. Dreuw and M. Head-Gordon, Chem. Rev., 2005, 105, 4009.
39. M. Head-Gordon, A. M. Grana, D. Maurice and C. A. White, J. Phys. Chem., 1995, 99, 14261.
40. M. Merchán and R. González-Luque, J. Chem. Phys., 1997, 106, 1112.
41. N. L. Wagner, J. A. Greco, M. M. Enriquez, H. A. Frank and R. R. Birge, Biophys. J., 2013, 104, 1314.
42. M. A. El-Sayed, J. Chem. Phys., 1962, 36, 573.
43. M. A. El-Sayed, J. Chem. Phys., 1963, 38, 2834.
44. M. A. El-Sayed, J. Chem. Phys., 1964, 41, 2462.
45. K.-H. Grellmann, R. Memming and R. Livingston, J. Am. Chem. Soc., 1962, 84, 546.
46. W. H. Waddell, A. M. Schaffer and R. S. Becker, J. Am. Chem. Soc., 1973, 95, 8223.
47. R. S. Becker, G. Hug, P. K. Das, A. M. Schaffer, T. Takemura, N. Yamamoto and W. Waddell, J. Phys. Chem., 1976, 80, 2265.
48. R. S. Becker, P. K. Das and G. Kogan, Chem. Phys. Lett., 1979, 67, 463.
49. P. K. Das, G. Kogan and R. S. Becker, Photochem. Photobiol., 1979, 30, 689.
50. R. S. Becker, Photochem. Photobiol., 1988, 48, 369.
51. E. W. Abrahamson, R. G. Adams and V. J. Wulff, J. Phys. Chem., 1959, 63, 441.
52. R. Govindjee, S. P. Balashov and T. G. Ebrey, Biophys. J., 1990, 58, 597.
53. J. Tittor and D. Oesterhelt, FEBS Lett., 1990, 263, 269.

---

Footnote

† Electronic supplementary information (ESI) available: Chemical structures; pump fluence dependence for transient absorption signals of all-trans retinal in n-hexane and acetonitrile; subtraction of the solvent contribution in the broadband transient absorption spectra; steady-state absorption spectra of all-trans and 13-cis retinal, their characteristic parameters and solvatochromic analysis; UV-Vis transient broadband absorption spectra of all-trans retinal; overview of time constants and quantum yields from the global analysis; species-associated spectra for all-trans retinal in different solvents; spectral fits and kinetics for all-trans retinal in different solvents; comparison with fit results from other kinetic models; steady-state absorption spectra for isomers of retinoic acid; UV-Vis broadband transient absorption spectra of 13-cis retinal; results of DFT/TDDFT calculations for all-trans retinal and all-trans retinoic acid. See DOI: 10.1039/c6cp01335j

---