Carotenoid-Bacteriochlorophyll Energy Transfer in LH2 Complexes Studied with 10-fs Time Resolution

Abstract

In this report, we present a study of carotenoid-bacteriochlorophyll energy transfer processes in two peripheral light-harvesting complexes (known as LH2) from purple bacteria. We use transient absorption spectroscopy with ≈10 fs temporal resolution, which is necessary to observe the very fast energy relaxation processes. By comparing excited-state dynamics of the carotenoids in organic solvents and inside the LH2 complexes, it has been possible to directly evaluate their energy transfer efficiency to the bacteriochlorophylls. In the case of okenone in the LH2 complex from Chromatium purpuratum, we obtained an energy transfer efficiency of η_ET2 = 63 ± 2.5% from the optically active excited state (S₂) and η_ET1 = 61 ± 2% from the optically dark state (S₁); for rhodopin glucoside contained in the LH2 complex from Rhodopseudomonas acidophila these values become η_ET2 = 49.5 ± 3.5% and η_ET1 = 5.1 ± 1%. The measurements also enabled us to observe vibrational energy relaxation in the carotenoids' S₁ state and real-time collective vibrational coherence initiated by the ultrashort pump pulses. Our results are important for understanding the dynamics of early events of photosynthesis and relating it to the structural arrangement of the chromophores.

INTRODUCTION

Carotenoids are among the most investigated organic molecules. They are ubiquitous pigments in photosynthetic systems, performing several vital functions (1). They photoprotect the photosynthetic apparatus by dissipating the excess solar energy and quenching both triplet excited (bacterio)chlorophyll (BChl) and singlet oxygen due to their low-lying triplet state (1–4). This is their essential role and there would be no photosynthesis in the presence of oxygen without them. They also serve as accessory light-harvesting pigments: light absorbed by donor carotenoids in the blue-green region of the spectrum is transferred to acceptor (B)Chls via singlet-singlet energy transfer (ET) with very high efficiency (up to ≈100%), thereby making it available to drive photosynthesis. Carotenoids also have distinct structural functions (as structure stabilization) (5) and are involved in plants in nonphotochemical quenching via the xanthophyll cycle (6).

Carotenoids have a characteristic strong absorption in the visible region, due to the optically allowed electronic transition from their ground state S₀ (of symmetry, according to the C_2h point group, in analogy to all-trans-polyenes), to their second excited singlet state (7,8). The first excited state has the same symmetry as the ground state and is therefore optically forbidden (dark state): it is, however, reached in a few hundreds of femtoseconds after photoexcitation, via an ultrafast internal conversion (IC) from S₂, denoted IC₂₁ in the following (9–12). Nonradiative decay from S₁ to S₀ (indicated as IC₁₀) then occurs on a 10-ps timescale (13,14). Intersystem crossing to the triplet manifold is rather inefficient but does occur in some purple bacterial antenna complexes (15,16). Carotenoid photophysics has been recently excellently reviewed by Polívka and Sundström (17).

Most studies on excited state dynamics of carotenoids have used the three-state model described above to interpret their results. However, previous theoretical work by Tavan and Schulten had indicated the possible existence of additional excited singlet states (e.g., and ) between S₂ and S₁ for carotenoids with more than approximately nine conjugated double bonds (18,19). Recently, experimental results based on resonance Raman profiles (20–22), fluorescence spectroscopy (23), and time-resolved absorption spectroscopy (24–28) have suggested the involvement of these intermediate states in the internal conversion of carotenoids from S₂ to S₁. In particular, transient absorption spectra with 10-fs time resolution showed a photoinduced absorption (PA) band in the near-IR rising instantaneously and decaying with a 10−20-fs time constant (depending on the carotenoid chain length) (25,27,28). This initial step in excited state dynamics was interpreted as a signature of conversion from S₂ to an intermediate state. Experiments suggesting the presence of intermediate states are, however, at odds with fluorescence up-conversion data showing for the emissive state a 150-fs lifetime which corresponds to the time constant for formation of the S₁ state (10,11). Therefore, the existence of intermediate excited states in carotenoids and their involvement in the IC₂₁ process is still a matter of controversy.

Purple photosynthetic bacteria are anaerobic prokaryotes that use carotenoids and BChls as their main photosynthetic pigments (29). Peripheral light-harvesting (LH2) complexes from purple bacteria are excellent model pigment-protein complexes for investigating the basic mechanisms of photosynthetic light harvesting, because their structure is known with high resolution (30,31) and because, unlike in plants and algae, the spectral features of the different pigments are well separated (29). In particular, they enable us to study the ET processes from the carotenoids to the BChls.

A prototypical purple bacterium is Rhodopseudomonas acidophila (R. acidophila); its LH2 complexes have a ring structure, consisting of nine monomeric units, each formed by a pair of αβ-apoproteins which bind, noncovalently, three molecules of BChla and one or two molecules of the carotenoid rhodopin glucoside (30,32). BChls molecules are organized into two groups, commonly denoted BChl 800 and BChl 850, referring to the position of their low-lying absorption bands. There are nine BChl 800 groups forming a ring on the cytoplasmic side of the complex, and 18 BChl 850 groups forming a tightly coupled ring near the periplasmic side of the complex. The well-resolved carotenoids have an extended S-shape, all-trans (33) conformation and span the entire depth of the complex, coming into van der Waals contact with both groups of BChls. The LH2 complex from Chromatium purpuratum (C. purpuratum) has an unusual absorption spectrum in the near-infrared (with a single major absorption peak at ≈830 nm) and contains the carotenoid okenone (34). There is no structural information available for this LH2 complex.

The BChls have two absorption bands relevant to the ET process: the Q_y transition at ∼800–850 nm and the Q_x transition at ∼590 nm (in the following, we will refer to Q_x/Q_y as the states reached by the Q_x/Q_y transitions, respectively). The S₂ state of carotenoids can transfer excitation energy to the Q_x state (denoted ET₂ channel in the following). On the other hand, ET to Q_y (denoted ET₁ in the following) can occur from the carotenoid S₁ state when the energy of this S₁ state is higher than that of Q_y, which happens for carotenoids with shorter conjugation lengths (typically less than 11 conjugated double bonds) (35). Given the 1–2 orders-of-magnitude shorter lifetime of the S₂ carotenoid state as compared to the S₁, one would expect ET₁ to be the dominant channel; however, it has been demonstrated, by fluorescence up-conversion measurements (36–38) and transient absorption measurements (16,38), that ET₂ is active in all photosynthetic bacteria studied so far and even predominant in some of them. The ET₂ process, being a transition between two allowed optical states, can be described to the first order by the Förster-type mechanism, i.e., coupling between transition dipole moments of the molecules; a more accurate theoretical description is obtained using full Coulombic interaction including higher-order multipoles (39). Although the forbidden nature of the S₀–S₁ transition initially suggested that the Dexter mechanism, i.e., transfer through electron exchange, was responsible for the ET₁ process (40), more detailed calculations (39,41) later showed that it is dominated by higher-order Coulombic and polarization interactions.

Given the very fast timescale of both IC₂₁ and ET₂ processes, their full characterization challenges the ≈100-fs temporal resolution of standard pump-probe or fluorescence up-conversion systems; in particular the need of deconvolving the instrumental response from the measured dynamics can lead to ambiguities in the assignment of the underlying processes. The recent availability of pulses with 10−20-fs duration tunable throughout the visible spectrum (42) makes it possible to probe the early events of energy relaxation in carotenoids with unprecedented temporal resolution. In this work we use 10-fs light pulses to study the early steps of excited state dynamics of two carotenoids, okenone and rhodopin glucoside, in organic solvents and in the LH2 complexes from two purple bacteria, C. purpuratum and R. acidophila, respectively. Exploiting the improved temporal resolution, we can follow the details of the IC₂₁ process; by comparing the dynamics of the carotenoids alone and in the LH2 complexes, we are able to measure directly the efficiency of the ET₂ process. By subsequently studying the IC₁₀ process, we can also determine the role of ET₁.

In our experiments on okenone and rhodopin glucoside, we did not detect any short-lived PA signal which could provide evidence of intermediate states. Such short-lived PA is best observed in the near-IR, where the PA bands from S₂ and (possibly) the intermediate states are located. In the LH2 complexes, however, these bands overlap with the photobleaching (PB) and PA bands of BChl. This spectral congestion of signals coming from different species makes it very hard to detect the presence of intermediate states in the antenna complexes. For these reasons, we analyzed our data using the classical three-level picture.

MATERIALS AND METHODS

Sample preparation

Okenone, rhodopin glucoside, and the LH2 complexes (C. purpuratum and R. acidophila) were isolated and purified as previously described (34,38,43). The purity of the carotenoids was determined by thin layer chromatography on silica gel plates while the purity and integrity of the antenna complexes was checked by spectrophotometric measurements. The LH2 complex from C. purpuratum was found to be rather unstable. We therefore used chromatophore membranes rather than the isolated LH2 complex. Previous studies (34) showed that the LH2 complex represents >90% of the pigment-protein complexes in C. purpuratum membranes under the cell growth conditions used for our samples. In this case, no sample degradation was detected.

Ultrafast spectroscopy

The setup used for the experiments starts with a mode-locked Ti:sapphire system with chirped pulse amplification, producing 500-μJ, 150 fs pulses at 790 nm, and 1 kHz repetition rate. These pulses are used to pump two non-collinear optical parametric amplifiers (NOPAs), which have the capability of generating broadband visible pulses (44–46). The first NOPA produces 15-fs transform-limited pulses with tunable center frequency; the second NOPA generates ultrabroadband visible pulses with bandwidth spanning the 500−700 nm wavelength range and sub-10-fs duration. For both NOPAs, the pulses are compressed by multiple reflections onto chirped dielectric mirrors, which are particularly insensitive to alignment and allow obtaining consistently short pulses. The temporal structure of the compressed pulses has been characterized by measuring their spectral amplitude and phase using the SPIDER technique (46). The pulses derived from the NOPAs are synchronized by a delay line and focused on the sample using only reflective optics, in a standard non-collinear pump-probe configuration: an intense pulse from the first NOPA serves as the pump, while a weaker, delayed pulse from the second NOPA probes the differential transmission changes (ΔT/T). To minimize vibronic relaxation effects, our pump pulses were tuned in such a way as to excite predominantly the first vibrational level of the S₂ excited state of the carotenoids. After the sample, the probe beam was spatially selected by means of an iris. By inserting a mirror before the sample, it was possible to perform a cross-correlation between pump and probe pulses, so as to exactly determine the zero time delay and the instrumental response function. Day-by-day verification of the setup performance was made by standard non-collinear cross-correlation, which had a width of 15−20 fs.

Time-resolved measurements at specific probe wavelengths were obtained by spectrally filtering, with 10-nm bandwidth interference filters, the probe beam after it passed through the sample, and combining differential detection with lock-in amplification. Transmission difference spectra over the entire bandwidth of the probe pulse at specific delays are then reconstructed from the time traces, after low-pass filtering to suppress the coherent oscillations superimposed on the dynamics. In all measurements, the maximum ΔT/T signal was <10%, thereby ensuring that saturation did not occur. Linearity of the excitation regime was confirmed by varying the intensity of the pump pulses and checking that the measured dynamics were independent of the intensity of the excitation pulse.

The solutions were kept at room temperature in a homemade cuvette, employing 200-μm-thick fused silica windows and with an optical path of ≈200 μm. Since the average power is quite low, in the μW range, no sample degradation is observed during the experiments. Nevertheless, the solution was replaced at the beginning of each experimental run.

Data analysis

The pump-probe time traces have been analyzed using a fitting procedure based on a rate equations model describing population balance. We chose the simplest model able to reproduce the main observed phenomena, namely:

1. Internal conversion processes within the excited states in the carotenoids.

2. Vibrational relaxation within the excited states.

3. Energy transfer from carotenoids to the BChls in the LH2 complexes.

With respect to global fitting analyses, using singular-value decomposition and extracting species-associated difference spectra, this approach has the advantages of simplicity and of providing direct physical insight in the ongoing processes.

A scheme of the electronic and vibrational levels of carotenoids and bacterio-chlorophylls used in the model is shown in Fig. 1. For the carotenoids in solution, we consider three electronic states: the S₂ state, populated by the pump pulse on its lowest vibrational level; this exponentially decays with rate constant k_IC21 to the S₁ state; and this in turn relaxes to the S₀ ground state, with rate constant k_IC10. Vibrational relaxation within the S₁ state is modeled by considering three associated vibrational levels, with k_V2 and k_V1 deactivation rate constants. The rate equations model is

(1a)

(1b)

(1c)

(1d)

(1e)

where N_ij (i,j = 0,1,2) is the population of the S_i electronic state in its j^th vibrational level and g(t) is the pump beam temporal profile, assumed to be an hyperbolic secant square function with 15-fs, full-width half-maximum, duration.

FIGURE 1.

Scheme of the electronic and vibrational levels of carotenoids and bacterio-chlorophylls used in the model. Population dynamics are indicated with dashed arrows, ΔT/T signals with solid arrows. After photoexcitation from the carotenoid S₀ ground state to the S₂ state, internal conversion occurs to the S₁ state (with k_IC21 rate constant), where vibrational relaxation is modeled as energy cascade within three consecutive vibrational levels (with k_V2 and k_V1 rate constants). Relaxation back to the ground state then takes place with k_IC10 rate constant. When the carotenoid is in the LH2 complex, energy transfer can occur from the S₂ state to the bacterio-chlorophyll Q_x with k_ET2 rate constant or from the carotenoid S₁ state to the bacterio-chlorophyll Q_y with k_ET1 rate constant.

For the carotenoids in the LH2 complexes, two additional relaxation pathways are added: ET from S₂ to Q_x (with rate k_ET2) and ET from S₁ to Q_y (with rate k_ET1). We consider the ET₁ process to be active with equal efficiency from all vibrational levels of the S₁ state. The corresponding rate equations are

(2a)

(2b)

(2c)

(2d)

(2e)

Equations 1a–e and 2a–e are solved numerically. The ΔT/T differential transmission signal at a certain probe wavelength λ and pump-probe delay τ is then calculated as

(3)

where d is the sample thickness and σ_pi,qj(λ) is the cross section (in cm²) for the vibronic transition between the i^th vibrational level of S_p and the j^th vibrational level of S_q. The first term of Eq. 3 represents PB of the S₀–S₂ transition (in which i = 0 because only the first vibrational level of the ground state is occupied at room temperature), the second term describes PA from S₁ to a higher-lying S_n state (both represented by three vibrational levels), and the third corresponds to stimulated emission (SE) from S₂ (in which i = 0 because the pump pulse is only resonant with the 0–0 transition) to S₀. Each of these contributions consists of the sum of the vibronic transitions between the different vibrational levels associated to the electronic states. The cross-section is (47,48)

(4)

where A_pq is a coefficient including the square of the dipole moment for the S_p–S_q electronic transition, |〈i,j〉|² are the Franck-Condon factors, and φ(λ) is a Gaussian line-shape. Finally, to take into account the finite probe pulse duration, the differential transmission signal is convoluted with the probe pulse temporal profile, having a 7-fs duration.

The fitting procedure was iterated until the best set of parameters was found to minimize the root-mean-square error between the measured and calculated ΔT/T(λ,τ). A nonlinear minimization procedure based on the simplex method was used. To evaluate the uncertainties of the extracted parameters, we have deliberately changed each parameter from the best fit until we obtained an increase in the root-mean-square error of 10%, which we chose as a reasonable estimation of the goodness of the fit.

RESULTS

Ultrafast dynamics of okenone in solution

The main aim of this study was to determine the efficiency of the carotenoid-to-BChl ET by directly comparing the IC dynamics of carotenoids in organic solvents and in the LH2 complexes. To obtain a meaningful comparison, it is essential to choose the correct solvent for the carotenoid that best mimics the protein environment in the LH2 complex. We therefore performed an extensive study of the dynamics of okenone's excited states in various organic solvents. It is well known that the energy of the S₀−S₂ transition in carotenoids depends on the refractive index n of the solvent and it shifts almost linearly with its polarizability R = (n²−1)/(n²+2) (11,49), both for polar and nonpolar solvents. On the other hand, the position of the dark S₁ state does not strongly depend on the solvent. Fig. 2 a shows steady-state absorption of LH2 complex from C. purpuratum and its carotenoid (okenone) in various solvents: acetone (R = 0.219), cyclohexane (R = 0.256), benzyl alcohol (R = 0.314), and carbon disulfide (CS₂, R = 0.354). Carotenoid absorption extends in the visible region from ≈400 nm to ≈600 nm; BChl Q_y absorption has two peaks in the near-infrared region at 800 nm and 830 nm, and BChl Q_x band is partially visible at ∼590 nm. As expected, a clear trend in the position of the S₂ state energy of the carotenoid was seen upon changing solvent: the absorption peak moved from ≈486 nm to ≈520 nm going from low to high polarizability. The closest superposition with okenone's absorption in the LH2 complex was found for CS₂.

FIGURE 2.

(a) Steady-state absorption spectra of okenone in cyclohexane, acetone, benzyl alcohol, CS₂ solution, and of the LH2 complex of C. purpuratum. (b) Steady-state absorption spectra of rhodopin glucoside in acetone and benzyl alcohol solution and of LH2 complex of R. acidophila. Pump central wavelengths are indicated by an arrow.

Fig. 3 a shows −ΔT/T spectra of okenone in various solvents for 500 fs pump-probe delay. Pump central wavelengths (indicated with an arrow in Fig. 2 a) correspond to 530 nm (cyclohexane), 535 nm (acetone), 570 nm (benzyl alcohol), and 580 nm (CS₂). A broad PA band in the visible region is observed (PA₁ band), peaking at 590−650 nm, depending on the solvent. The PA₁ band is a well known feature of carotenoids corresponding to the S₁→S_n absorption and is the signature of the dark S₁ state. The transmission increase on the short wavelength side of these spectra is due to PB of the ground state. In general, PA₁ is formed with a time constant of ≈100 fs and decays with time constant of ≈4.5 ps. This signal reveals the sequential steps of the IC₂₁ and IC₁₀ processes. The measured dynamics are observed to be solvent-dependent. Table 1 reports the time constants obtained fitting with exponential rise and decay the ΔT/T signals at the peak of the PA₁ band. It can be observed that the time constant for S₁ formation decreases from 130 fs in cyclohexane to 95 fs in CS₂, in agreement with the reduction of the S₂−S₁ energy difference and the energy gap law (50,51). Due to the insensitivity of the S₁ state energy to the solvent, the time constant for the S₁ decay only varies from 4.2 ps to 4.65 ps.

FIGURE 3.

(a) Chemical structure and transient absorption spectra of okenone in different solvents at ≈500 fs probe delay. (b) Chemical structure and transient absorption spectra of rhodopin glucoside in acetone and benzyl alcohol solution at ≈600 fs probe delay.

TABLE 1.

PA₁ band formation and decay time constants for okenone and rhodopin glucoside in various solvents, measured at the indicated wavelength corresponding to the peak of the S₁–S_n photoinduced absorption band

Sample	Solvent	PA1 band peak wavelength	PA1 formation time constant	PA1 decay time constant
Okenone	Cyclohexane	600 nm	130 fs	4.2 ps
	Acetone	600 nm	110 fs	4.3 ps
	Benzyl alcohol	640 nm	100 fs	4.65 ps
	CS2	650 nm	95 fs	4.2 ps
Rhodopin	Acetone	560 nm	145 fs	4.02 ps
glucoside	Benzyl alcohol	590 nm	180 fs	4.45 ps

Carbon disulfide has been chosen as the best solvent to reproduce the in vivo environment for okenone since this solvent gives the closest match to the absorption of okenone in the LH2 complex. This solvent was also used in a previous study on the carotenoid-to-BChl ET in the LH2 complex from C. purpuratum (52).

The ΔT/T spectra of okenone in CS₂ at different pump-probe delays are shown in Fig. 4 a. The rapid formation and subsequent decay of the PA₁ band peaking at 650 nm can be observed in detail. The increased transmission present on the short wavelength side of the spectra is due to a superposition of PB of the ground state (530−580 nm) and, for early delays, stimulated emission (SE) from the S₂ state (580−630 nm). Note the strong blue-shift of the isosbestic point for increasing delays. The ΔT/T dynamics at selected probe wavelengths for short and long timescales are shown in Fig. 5, a and b, as circles, respectively. There is a significant wavelength dependence of the dynamics of the buildup and decay of PA₁: faster at the red edge and slower at the blue edge. The same behavior is seen in the PA₁ spectra (Fig. 4 a), which change in position during the IC processes. The oscillatory patterns superimposed on some of the population dynamics, observed in this and other figures, are due to coherent vibrational motions initiated by the short pump pulse and will be addressed in Discussion, below.

FIGURE 4.

(a) Experimental ΔT/T spectra of okenone in CS₂ solution at selected pump-probe delays; (b) simulated ΔT/T spectra of okenone according to the model described in the text; (c) experimental ΔT/T spectra of LH2 complex from C. purpuratum; and (d) simulated ΔT/T spectra of the LH2 complex.

FIGURE 5.

(Circles) ΔT/T dynamics at different probe wavelengths of okenone in CS₂ (a and b) and the LH2 complex from C. purpuratum (c and d). Thick solid lines are fits obtained using the model described in the text. Pump-probe cross-correlation is shown as shaded area.

The ΔT/T spectra and dynamics were fitted globally with the model described above and the results are shown in Fig. 4 b and in Fig. 5, a and b, as solid lines. The obtained parameters for the best fit are listed in Table 2. The rate constants for the IC₂₁ and IC₁₀ processes are k_IC21 = (95 fs)⁻¹ and k_IC10 = (4.46 ps)⁻¹, respectively. Note that our value of k_IC10 differs by a factor of 2 from the result reported by Andersson et al. (52).

TABLE 2.

List of the parameters used in the global fitting procedure for the four samples studied

	Okenone/C. purpuratum	Rhodopin glucoside/R. acidophila
S1–Sn energy gap	1.91 eV (Okenone)	2.12 eV (Rhodopin glucoside)
	2 eV (C. purpuratum)	2.145 eV (R. acidophila)
S1–Sn vibrational energy	110 ± 10 meV	125 ± 10 meV
Gaussian linewidth	σ = 200 ± 12 meV	σ = 63 ± 7 meV
S1–Sn Huang-Rhys factor γ	0.105 ± 0.005	0.124 ± 0.005
(kIC21)−1	95 (−5 +8) fs	137 (−11 +10) fs
(kIC10)−1	4460 (−160 +180) fs	4280 (−190 +160) fs
(kET2)−1	55.7 (−6 +2) fs	140 (−15 + 10) fs
(kET1)−1	2780 (−110 +270) fs	80,000 ± 10,000 fs

Ultrafast dynamics of okenone inside the LH2 complex of C. purpuratum

Fig. 4 c shows the ΔT/T spectra of the LH2 complex in membranes of the bacterium C. purpuratum, upon resonant photoexcitation of the okenone chromophore. The visible pump pulses are tuned to ≈550 nm central wavelength (as indicated by the arrow in Fig. 2 a) to minimize the excitation of the Q_x transition of the BChls. This leads, however, to partial excitation of the second vibrational level of S₂. Again, the formation of the PA₁ band and its subsequent decay are clearly observed. The increased transmission seen at the blue side (580−500 nm) is due to a combination of PB and SE. Within the protein environment, the PA₁ band of okenone peaks at 620 nm, i.e., ≈20 nm red-shifted with respect to its position in CS₂. The rise time of the PA₁ band is faster in LH2 and is essentially completed within 200 fs. Relaxation of S₁ back to the ground state is also considerably faster in LH2 as compared to okenone in CS₂. The faster buildup and decay of PA₁ are seen more clearly by following the detailed kinetics at selected probe wavelengths (see Fig. 5, c and d).

The ΔT/T spectra and dynamics were fitted using the model described previously, keeping the same parameters used for okenone in solution and adding the energy transfer channels. The results are shown in Fig. 4 d and in Fig. 5, c and d, as solid lines; the obtained ET rates are k_ET2 = (55.7 fs)⁻¹ and k_ET1 = (2.78 ps)⁻¹. A blue-shift of the PA₁ band peak and isosbestic point on the picosecond timescale is also observed in okenone within the LH2 complex. Note that our measurements differ from the previously reported ones (13) because the fast decay component of S₁ (400–500 fs) is not present.

The S₁ decay does not follow a single exponential law: most of the PA₁ signal decreases on a few-picoseconds timescale, whereas a small residual signal lasts a few tens of picoseconds. This slow component can be assigned to photoinduced absorption either from the Q_y BChl electronic level (which is long-lived) or from an additional carotenoid excited state, named S*, which is a precursor on an ultrafast fission reaction pathway to the carotenoid triplet state. A similar energy level has been already observed in LH1 complexes from Rhodospirillum rubrum (15) and in LH2 complexes from Rhodobacter sphaeroides (16). Both interpretations give rise to a long-lived plateau in the PA signals, which was introduced in the fittings.

Ultrafast dynamics of rhodopin glucoside in solution

Ultrafast dynamics of the LH2 system of R. acidophila, together with its carotenoid rhodopin glucoside, has already been studied, with lower temporal resolution, by several groups (38,53,54). Our results partially confirm those findings, but also provide some new insights into the early events during carotenoid-to-BChl energy transfer.

Fig. 2 b shows steady-state absorption spectra of rhodopin glucoside and the LH2 complex from R. acidophila. Pump central wavelengths are indicated with an arrow and correspond to 540 nm (R. acidophila and rhodopin glucoside in benzyl alcohol) and 515 nm (rhodopin glucoside in acetone). BChl has the Q_x absorption band at ≈590 nm and the two Q_y peaks at 800 nm and 850 nm. The carotenoid absorbs in the green-blue region and has a marked vibronic structure, with the 0–0 transition peaking at ≈524 nm in the LH2 complex. Absorption of rhodopin glucoside in benzyl alcohol and acetone solution is also presented in Fig. 2 b, showing the 0–0 line at 510 and 530 nm, respectively. It can be seen that benzyl alcohol mimics well the protein environment both for the absorption peak and its vibronic structure.

The −ΔT/T spectra of rhodopin glucoside in two different solvents at the probe delay corresponding to the maximum signal from S₁ (≈600 fs) are shown in Fig. 3 b. PB of the ground state matches steady-state absorption of the carotenoid: it extends up to 520 nm when dissolved in acetone and 540 nm in benzyl alcohol. The broad PA₁ band in the visible peaks at 560 nm in acetone and 590 nm in benzyl alcohol. Also in rhodopin glucoside the dynamics are solvent-dependent, as shown in Table 1, reporting PA₁ formation and decay time constants at the peak of the band. The S₁ buildup and decay are faster in acetone than in benzyl alcohol. The S₁ decay in acetone is even faster than for LH2 (see later).

The −ΔT/T spectra of rhodopin glucoside in benzyl alcohol, for different pump-probe delays, are shown in Fig. 6 a. At early times, SE from S₂ (640–540 nm) and PB bands (500–540 nm) appear throughout the visible. The corresponding ΔT/T dynamics are shown in Fig. 7, a and b, as circles. Blue-shift in time of the PA₁ band and of the isosbestic point, previously observed in okenone, is also present in rhodopin glucoside, and is clearly seen in both the time evolution of the spectra (Fig. 6 a) and the wavelength dependence of the dynamics, which are faster at the red edge than at the blue edge (Fig. 7, a and b).

FIGURE 6.

(a) Experimental ΔT/T spectra of rhodopin glucoside in benzyl alcohol solution at selected pump-probe delays; (b) simulated ΔT/T spectra of rhodopin glucoside according to the model described in the text; (c) experimental ΔT/T spectra of LH2 complex from R. acidophila; and (d) simulated ΔT/T spectra of the LH2 complex.

FIGURE 7.

(Circles) ΔT/T dynamics at different probe wavelengths of rhodopin glucoside in benzyl alcohol (a and b) and the LH2 complex from R. acidophila (c and d). Thick solid lines are fits obtained using the model described in the text. Pump-probe cross-correlation is shown as shaded area.

Global fittings of the ΔT/T spectra and dynamics (shown in Figs. 6 b and 7, a and b, as solid lines) yields, for the IC₂₁ and IC₁₀ processes, the following rate constants: k_IC21 = (137 fs)⁻¹ and k_IC10 = (4.28 ps)⁻¹. These results are in good agreement with those reported by Macpherson et al. (38), who found a lifetime of the S₂ state of rhodopin glucoside in benzyl alcohol, determined by fluorescence up-conversion, of 124 ± 8 fs, and a lifetime of the S₁ state of 4.8 ± 0.2 ps.

Ultrafast dynamics of rhodopin glucoside inside the LH2 complex of R. acidophila

Fig. 6 c shows a sequence of −ΔT/T spectra of the LH2 complex of bacterium R. acidophila, upon resonant photoexcitation of the rhodopin glucoside chromophore. At early pump-probe delays a broad SE band from S₂ is seen, which rapidly gives way to the buildup of the PA₁ band, while PB is left over below ≈540 nm. Within the protein environment, the peak position of the PA₁ band is blue-shifted by <10 nm with respect to that of rhodopin glucoside in benzyl alcohol, indicating that this solvent is indeed a good approximation of the protein matrix. PB of the BChl Q_x transition burns a weak hole in the PA₁ band that is observable at ∼590 nm. In the LH2 complex, the rise time of the PA₁ band is faster than in benzyl alcohol and is essentially complete within 250 fs. The subsequent relaxation of S₁ back to the ground state is only very slightly faster than for rhodopin glucoside in benzyl alcohol. The faster PA₁ buildup and decay were investigated in greater detail by measuring the kinetics at selected probe wavelengths (Fig. 7, c and d). Analogously to C. purpuratum, a small residual PA signal was still present after 40 ps, in addition to the Q_x PB. This PA probably reflects the formation of the S* state of the carotenoid within the LH2 complex (53,54). The usual wavelength dependence of the S₁ rise and decay (slower in the blue than in the red) is also seen in this LH2 complex.

Also in this case, the ΔT/T spectra and dynamics were fitted using the model described previously, keeping the same parameters used for rhodopin glucoside in solution and adding the energy transfer channels. The resulting ET rates are k_ET2 = (140 fs)⁻¹ and k_ET1 = (80 ps)⁻¹. The present data are in good agreement with the results of Macpherson et al. (38), who reported an S₂ lifetime of the carotenoid in the LH2 complex, measured by fluorescence up-conversion, of 57 ± 2 fs. On the other hand, Rondonuwu et al. (54), using pump-probe spectroscopy with 160 fs instrumental response width, proposed a quite different relaxation pathway for the LH2 complex of R. acidophila. This involves initial energy relaxation to the state with a time constant of ≈80 fs and subsequent delayed formation of the S₁ state, with a time constant of ≈400 fs. Our measurements do not find any evidence of a delayed formation of the S₁ state.

DISCUSSION

Vibrational relaxation

In all the ΔT/T spectra and dynamics presented in this article, both in solution and within the LH2 complexes, we observe common features in the evolution of the PA₁ band:

1. Faster buildup in the red than in the blue side.

2. Slower decay in the blue than in the red side.

3. Blue-shift of the isosbestic point and of the PA₁ peak.

4. Spectral narrowing of the band within the first ≈500 fs (best observed in rhodopin glucoside because of its narrower spectral features).

These effects are attributed to intraband vibrational relaxation associated to dissipation of the excess energy (≈5000 cm⁻¹) initially deposited in S₁. Previous studies (53,55–57) have observed a wavelength-dependent dynamics in S₁, typically occurring on the 500-fs timescale, and attributed it to vibrational relaxation. Our enhanced temporal resolution shows that an additional faster relaxation process is taking place on the 50-fs timescale, giving rise to the wavelength-dependent buildup time of the PA₁ band and its spectral narrowing. Therefore, in our model we introduced two excited vibrational levels in S₁ and assumed that relaxation from the v = 2 to the v = 1 level proceeds with a k_v2 = 50 fs time constant, while relaxation from the v = 1 to the v = 0 level occurs with a slower k_v1 = 500 fs time constant. Using these parameters, the model gives satisfactory fits for both carotenoids in the different environments (solution and LH2 complex), indicating that vibrational relaxation in S₁ is a general feature in carotenoids. Although the values of the k_v2 and k_v1 relaxation constants cannot be determined to a great accuracy, nevertheless our simple model appears to give an accurate description of the ongoing photophysics.

We note that an alternative explanation of the wavelength-dependent PA₁ dynamics has been given (58). It has been proposed that conformational relaxation of the carotenoid from a twisted configuration—created by the photoexcitation—to the all-trans form takes place with a time constant of 300−400 fs during the IC₂₁, resulting in a blue-shift of the PA₁ band.

Carotenoid-to-BChl energy transfer

The shortening of the S₂ and S₁ lifetimes of okenone and rhodopin glucoside in their corresponding LH2 complexes is due to the opening of new relaxation channels. In the LH2 complexes, in addition to internal conversion of S₂ to S₁ and S₁ to S₀, energy transfer from both S₂ and S₁ to the BChls is possible. To determine the ET rates and efficiencies in the LH2 complexes, the lifetimes of both S₂ and S₁ must be determined for both carotenoids in organic solvents and in the LH2 complexes. Ideally, this comparison requires the environment of the carotenoids to be the same in both situations, so that the k_IC21 and k_IC10 rates remain virtually unchanged; in practice this is not perfectly achievable. By comparing the ground state absorption and excited-state dynamics of both carotenoids in a range of organic solvents, it has been possible to select a solvent in which the carotenoid's environment is the best approximation to that in the protein. Calculation of the efficiency of ET from the carotenoids to the BChls from analysis of the rate constants for the decay of the S₂ and S₁ states depends critically on the values for these decays in the absence of ET. As we showed above, both of these decays depend on the solvent in which the carotenoids are dissolved. All of the subsequent calculations and discussions assume that our choice of the optimal solvent has been reasonable. It is important, however, to realize this limitation.

For okenone, the results of our fitting procedure are as follows:

1. Internal conversion in CS₂ solution from S₂ to S₁ states occurs with rate k_IC21 = (95 fs)⁻¹, and from S₁ back to S₀ with rate k_IC10 = (4.46 ps)⁻¹.

2. Energy transfer from okenone S₂ state to Q_x state in LH2 complex of C. purpuratum membranes has a rate constant k_ET2 = (55.7 fs)⁻¹.

3. Similarly, the energy transfer from okenone S₁ state to Q_y state in LH2 has a rate constant k_ET1 = (2.78 ps)⁻¹.

It then follows that the energy transfer efficiencies are from S₂ and from S₁. The overall ET efficiency is then calculated by adding to the η_ET2 the η_ET1 properly weighted by the fraction of photoexcited molecules that enter the S₁ state: η^okenone = η_ET2 + (1−η_ET2) × η_ET1 = 85.6 ± 1.5%. In a previous study, Andersson et al. (52) determined a value of 95 ± 5% for the ET efficiency of the LH2 complex from C. purpuratum from the fluorescence excitation spectrum. However, that measurement appears to be not very well corrected, since in some spectral regions (400−500 nm) the efficiency exceeds 100%. This overestimate can explain the discrepancy with our result.

Using rhodopin glucoside (and using the same notation and methodology described above for okenone) the following rate constants were determined: k_IC21 = (137 fs)⁻¹, k_ET2 = (140 fs)⁻¹, k_IC10 = (4.28 ps)⁻¹, and k_ET1 = (80 ps)⁻¹. This translates into the following ET efficiencies: η_ET2 = 49.5 ± 3.5%, and η_ET1 = 5.1 ± 1%, which produces an overall η^RG = 52.1 ± 3%. These results are in excellent agreement with those obtained using fluorescence up-conversion where the time resolution was much longer (140–200 fs) (η_ET2 = 51–56%, η_ET1 = 5 ± 2%, 38). Other values for this overall ET efficiency in LH2 complexes from R. acidophila also obtained from transient absorption studies have yielded values of 53% (53) and 54% (54).

Coherent oscillations analysis

The dynamics in Figs. 5, a and c, and 7, a and c, have a superimposed oscillatory pattern, which is due to collective vibrational coherence initiated in the ensemble of photoexcited molecules by the ultrashort pump pulse (59,60). The oscillations are slowly damped and persist on a timescale longer than 1 ps. Fourier analysis of the signals, after subtraction of the slowly varying component, is shown in Fig. 8 a for okenone in CS₂ and in the LH2 complex from C. purpuratum, and in Fig. 8 b for rhodopin glucoside in benzyl alcohol and in LH2 complex from R. acidophila. To test the accuracy of this mathematical analysis, the inset of Fig. 8 a shows a plot (solid line) of the fit obtained by simply adding two sinusoidal functions at the extracted frequencies (with appropriate amplitudes and phases) corresponding to those obtained in the Fourier analysis. An excellent agreement was found with the experimental data (dashed line).

FIGURE 8.

Fourier-transform of oscillatory component of ΔT/T signal in (a) okenone, with LH2 of C. purpuratum, and (b) rhodopin glucoside, with LH2 of R. acidophila. Inset shows the oscillatory component superimposed on the temporal trace of okenone at 580 nm, together with a fit based on the frequencies extracted from the Fourier analysis.

Several coupled modes are present, the strongest being at ≈1155 cm⁻¹ and ≈1510 cm⁻¹ for both carotenoids in the organic solvents. By comparing these to the vibrational frequencies obtained for a variety of carotenoids with resonance-Raman spectra (61,62), it is possible to assign them to a_g-type C−C and C=C symmetric stretching modes of the conjugated backbone, respectively. At ∼1000 cm⁻¹ another smaller feature is visible in some of the pump-probe traces, corresponding to the overlap of methyl in-plane rocking and C–H out-of-plane wagging Raman lines (62). The vibrational frequencies for rhodopin glucoside in benzyl alcohol and in the LH2 complex are very similar. In the case of the LH2 complex from C. purpuratum the main vibrational frequencies are shifted to somewhat lower values.

It now remains for us to discuss upon which electronic state(s) the observed vibrational wavepacket(s) is/are moving. Excitation by an ultrashort pulse can create a vibrational wavepacket in the excited electronic state (S₂ in our case) as well as in the ground state (S₀), due to a mechanism known as resonant impulsive stimulated Raman scattering (63). In addition, in principle, the internal conversion process to S₁ could take place coherently, i.e., preserving the phase of the collective nuclear motion in the final state (64). In our case, due to the short lifetime of the S₂ state, the long-lasting oscillations must be assigned to either S₀ or S₁, or both. Since our measured frequencies match very well those obtained from ground state resonant Raman scattering (62) and the frequencies coupled to S₁ are known to be different from those coupled to S₀ (65), we can assign at least the major observed vibrational wavepackets to the ground state.

CONCLUSIONS

In conclusion, we have presented pump-probe measurements, with 10-fs temporal resolution, on carotenoids both in solution and inside the LH2 complex of two different light-harvesting purple bacteria. Our data provide a real-time picture of the IC, ET, and vibrational relaxation processes and enable us to calculate the energy transfer efficiencies. This study has been restricted to the dynamics that can be measured in the visible region of the spectrum. It will be interesting to see, when these measurements are extended with equal or better time resolution into the near-IR, whether involvement of any excited singlet states, which are located between S₂ and S₁, can be seen to contribute to ET to the BChls.