Spectral hole burning: examples from photosynthesis

Abstract

The optical spectra of photosynthetic pigment–protein complexes usually show broad absorption bands, often consisting of a number of overlapping, ‘hidden’ bands belonging to different species. Spectral hole burning is an ideal technique to unravel the optical and dynamic properties of such hidden species. Here, the principles of spectral hole burning (HB) and the experimental set-up used in its continuous wave (CW) and time-resolved versions are described. Examples from photosynthesis studied with hole burning, obtained in our laboratory, are then presented. These examples have been classified into three groups according to the parameters that were measured: (1) hole widths as a function of temperature, (2) hole widths as a function of delay time and (3) hole depths as a function of wavelength. Two examples from light-harvesting (LH) 2 complexes of purple bacteria are given within the first group: (a) the determination of energy-transfer times from the chromophores in the B800 ring to the B850 ring, and (b) optical dephasing in the B850 absorption band. One example from photosystem II (PSII) sub-core complexes of higher plants is given within the second group: it shows that the size of the complex determines the amount of spectral diffusion measured. Within the third group, two examples from (green) plants and purple bacteria have been chosen for: (a) the identification of ‘traps’ for energy transfer in PSII sub-core complexes of green plants, and (b) the uncovering of the lowest k = 0 exciton-state distribution within the B850 band of LH2 complexes of purple bacteria. The results prove the potential of spectral hole burning measurements for getting quantitative insight into dynamic processes in photosynthetic systems at low temperature, in particular, when individual bands are hidden within broad absorption bands. Because of its high-resolution wavelength selectivity, HB is a technique that is complementary to ultrafast pump–probe methods. In this review, we have provided an extensive bibliography for the benefit of scientists who plan to make use of this valuable technique in their future research.

Introduction

Frequency- and time-resolved laser spectroscopic techniques play an important role in the study of relaxation processes of electronically excited states of photosynthetic pigment–protein complexes. Energy transfer between pigments, optical dephasing, spectral diffusion and decay of exciton states are examples of such relaxation processes. To study these processes, lasers are used that have either very short pulses or very narrow spectral widths.

Techniques that make use of narrow-band lasers are called site- or energy-selective spectroscopies (Gooijer et al. 2000), such as fluorescence line-narrowing (FLN; Creemers et al. 1999a; De Caro et al. 1994; Freiberg et al. 2009; Jankowiak 2000; Personov 1983; Personov et al. 1972; Peterman et al. 1997), spectral hole burning (HB; Creemers and Völker 2000; Dang et al. 2008; De Vries and Wiersma 1976; Friedrich et al. 1994; Gorokhovskii et al. 1974; Hayes and Small 1978; Kharlamov et al. 1974; Krausz et al. 2008; Moerner 1988, and articles therein; Reinot et al. 2001; Völker 1989a, b; Völker and Van der Waals 1976) and single-molecule (SM) spectroscopy (Barkai et al. 2004; Berlin et al. 2007; Cogdell et al. 2006; Ketelaars et al. 2001; Moerner 2002; Moerner and Kador 1989; Orrit and Bernard 1990; Rigler et al. 2001; Rutkauskas et al. 2004, 2006; Van Oijen et al. 1999). These experimental methods yield information on dynamic processes in doped crystals and glasses as well as in pigment–protein complexes that cannot be obtained with conventional spectroscopy since their homogeneously broadened bands are buried under largely inhomogeneously broadened spectra.

This educational review is focussed on spectral hole burning (HB); it also provides an extensive bibliography. After an introduction to the processes studied here, we describe the HB principle. This is followed by a discussion of experimental methods. We then demonstrate the potential of this technique to obtain an insight into the dynamics of photosynthetic systems after photo-excitation. A number of examples, obtained in our laboratory, are shown (for references, see below). We prove that information on energy-transfer times and optical dephasing can be obtained for light-harvesting (LH) complexes of purple bacteria by measuring the hole width as a function of temperature. The first example discusses the B800-to-B850 energy-transfer process in LH2 complexes, whereas the second example shows results of optical dephasing in the red wing of the B850 band of LH2. We then follow this discussion on the broadening of the hole as a function of time (spectral diffusion). We show that the amount of spectral diffusion depends on the size of the photosynthetic complex studied. Further, we demonstrate that, in addition to the hole width, the hole depth as a function of wavelength can also yield relevant information that is otherwise hidden under the broad absorption bands. Data reviewed proves the existence of ‘traps’ for energy transfer in photosystem II (PSII) sub-core complexes of higher plants. The final example shows how we uncovered the lowest k = 0 exciton states hidden under the B850 band of LH2 complexes, and how their spectral distributions could be determined. To our knowledge, HB is the only technique that is able to uncover small, hidden spectral distributions characterized by specific dynamics.

Homogeneous linewidths, optical dephasing and spectral diffusion

Absorption and emission bands of pigment–protein complexes and organic molecules dissolved in solvents or polymers are generally very broad (typically a few 100 cm⁻¹, even at liquid-He temperatures), as compared to those found in crystalline systems (of a few cm⁻¹). Such large widths are caused by the slightly different environments of the individual chromophores within the disordered host (the protein or glass at low temperature), leading to a broad statistical distribution of the electronic transition energies and, therefore, to a wide Gaussian profile with an inhomogeneous width Γ_inh (Creemers and Völker 2000; Völker 1989a, b, and references therein).

Information on the dynamics of the excited state of the system is contained in the homogeneous linewidth Γ_hom of the electronic transition of the individual chromophores. Since Γ_hom is usually a factor of 10³–10⁵ times smaller than Γ_inh (Völker 1989a, b), the homogeneous line is buried in the inhomogeneously broadened band. To obtain the value of Γ_hom, laser techniques must be used, either in the time domain, such as photon echoes (Agarwal et al. 2002; Fidder and Wiersma 1993; Fidder et al. 1998; Hesselink and Wiersma 1980, 1983; Jimenez et al. 1997; Lampoura et al. 2000; Narasimhan et al. 1988; Thorn-Leeson and Wiersma 1995; Thorn-Leeson et al. 1997; Wiersma and Duppen 1987; Yang and Fleming 1999), or in the frequency domain, such as FLN, HB and SM (for references, see above).

The lineshape of a homogeneously broadened electronic transition is usually Lorentzian; it is the Fourier-transform of an exponential decay function. Its linewidth Γ_hom is given by the inverse of the optical dephasing time T ₂ and is usually written as the sum of two terms (Völker 1989a, b; Wiersma and Duppen 1987):

$$ \Upgamma_{\hom } = {\frac{1}{{{{\uppi}}\,T_{2} }}} = {\frac{1}{{2{{\uppi}}\,T_{1} }}} + {\frac{1}{{{{\uppi}}\,T_{2}^{*} \left( T \right)}}} $$

(1)

, where T ₁ is the lifetime of the electronically excited state and $T_{2}^{*}$ is the ‘pure’ dephasing time. The first term in Eq. 1 does not depend on temperature T, at low T. It is called the residual linewidth Γ₀ = (2π T ₁)⁻¹ for T → 0. $T_{2}^{*}$ represents the time it takes for the coherence of the electronic transition to be destroyed by chromophore–host (or pigment–protein) interactions. Since such fluctuations of the optical transition are caused by phonon scattering, $T_{2}^{*}$ depends on T.

The functional dependence on temperature of the second term $(\pi T_{2}^{*} (T ) )^{ - 1}$ in Eq. 1 differs for crystalline and amorphous systems. For doped organic crystals, it depends exponentially on temperature as exp (−E  / kT) (Dicker et al. 1981; Molenkamp and Wiersma 1984; Morsink et al. 1977; Völker 1989a, b; Völker et al. 1977, 1978). For doped organic glasses and pigment–protein complexes, it follows a universal T ^1.3±0.1 power law at low temperature (T ≤ 20 K), independent of the host and the chromophore (Breinl and Friedrich 1988; Jankowiak and Small 1993; Jankowiak et al. 1993; Köhler et al. 1988; Meijers and Wiersma 1994; Narasimhan et al. 1988; Thijssen et al. 1982, 1983, 1985; Van den Berg and Völker 1986, 1987; Van den Berg et al. 1988; Völker 1989a, b). Such a T-dependence has been interpreted in terms of two-level systems (TLS), which are low-energy excitations assumed to exist in glasses and in disordered systems in general. The TLSs are double-well potentials representing distinct structural configurations of the glass (Anderson et al. 1972; Phillips 1972, 1981, 1987). The transition or ‘flipping’ from one potential well to another occurs through interaction with phonons that cause a change in the glassy structure. TLSs are assumed to have a broad distribution of tunnelling parameters and energy splittings that lead to a broad distribution of fluctuation rates in the glass (Black and Halperin 1977; Hu and Walker 1977, 1978; Jankowiak et al. 1986; Maynard et al. 1980). If a probe molecule is incorporated in such a disordered host and its optical transition couples to TLSs, the dephasing or frequency fluctuations of the optical transition will be caused by relaxation of the TLSs. In particular, ‘fast’ TLSs that have relaxation rates R much larger than the decay rate (1/T ₁) of the excited state of the probe molecule are assumed to be responsible for ‘pure’ dephasing. The T ^1.3 dependence of Γ_hom has been explained by assuming a dipole–dipole coupling between the probe molecule and TLSs, with a density of states of the TLSs varying as ρ(E) ∝  E ^0.3, where E is the energy splitting of the eigenstates of the TLSs (Huber 1987; Jankowiak and Small 1993; Jankowiak et al. 1993; Putikka and Huber 1987).

The evolution of the glass (or protein) dynamics may lead to a continuous and irreversible change of the frequency of the optical transition of the chromophore. As a consequence, the observed homogeneous linewidth will depend on the time delay (t _d) between the excitation of the transition and its observation (Bai and Fayer 1988, 1989). This process is called spectral diffusion (Creemers et al. 1997; Den Hartog et al. 1998a, 1999a, b; Friedrich and Haarer 1986; Koedijk et al. 1996; Littau et al. 1992; Lock et al. 1999; Meijers and Wiersma 1994; Silbey et al. 1996; Wannemacher et al. 1993), and the measured width is the ‘effective’ homogeneous linewidth $\Upgamma_{\hom }^{'}$. In a time-dependent hole-burning experiment (see below) $\Upgamma_{\hom }^{'}$ depends on the delay t _d between the burn and probe pulse.

Principles of hole burning

In a spectral hole-burning experiment, the inhomogeneously broadened absorption band is irradiated at a given wavelength with a narrow-band laser. Whenever the molecules resonant with the laser wavelength undergo a photo-transformation (photophysical or photochemical), a hole is created in the original absorption band (see Fig. 1). The width of the hole, under certain conditions (see below), is then proportional to the homogeneous linewidth. The photoproduct will absorb at a different wavelength, either within the absorption band or outside. Since the laser selects molecules absorbing at a given frequency ν ₁, and not molecules in a specific environment, the correlation between transition energy and environmental parameters is, in general, different for the photoproduct and the original molecule. As a consequence, the width of the photoproduct band, or antihole, is larger than that of the hole (Völker and Van der Waals 1976; Völker and Macfarlane 1979). The optical resolution that can be reached with HB is 10³–10⁵ times higher than that with conventional techniques, which makes HB a powerful tool for spectroscopy in the MHz range (Völker 1989a, b).

Fig. 1

Top: Diagram of an inhomogeneously broadened absorption band with a width Γ_inh. The homogeneous bands of width Γ_hom of the individual electronic transitions are hidden under the broad inhomogeneous absorption band. Bottom: Laser-induced spectral hole burned at frequency ν₁. The photoproduct absorbs at a different frequency, here outside the inhomogeneous band (Creemers and Völker 2000)

Hole-burning mechanisms can be divided into two categories: persistent HB and transient HB (THB). Within the first category, there is photochemical HB (PHB; De Vries and Wiersma 1976; Friedrich and Haarer 1986, and references therein; Völker and Van der Waals 1976; Völker et al. 1977) and non-photochemical HB (NPHB; Carter and Small 1985; Hayes and Small 1978; Jankowiak and Small 1987, and references therein; Small 1983). The time scales involved in PHB and NPHB at low temperature are usually seconds to hours, whereas THB often lasts only microseconds (μs) or milliseconds (ms). For more details about these HB mechanisms, the reader is referred to Völker (1989a, b).

Dynamic processes, such as optical dephasing, energy transfer and spectral diffusion, which determine the homogeneous linewidth and therefore the hole width, depend on the interaction of the optical transition of the probe molecule with its surroundings. Thus, the hole width does not depend on the HB mechanism, as long as the latter takes place at a time scale much larger than the dynamic process under study (Creemers et al. 1997; Koedijk et al. 1996).

Experimental methods

A hole-burning (HB) experiment consists of three steps, schematically shown in Fig. 2: (1) the laser is scanned with low light intensity for a time t _p over the wavelength range of interest to generate a baseline in the absorption band; (2) a hole is burnt at a fixed wavelength for a time t _b with a much higher laser intensity (typically a factor of 10–10³); (3) the hole is probed for a time t _p by scanning the laser with low intensity as in step (1). To obtain the hole profile, the difference is taken between the signals in steps (1) and (3). To study spectral holes as a function of time (spectral diffusion), the delay time t _d is varied. Every new hole is then burnt at a slightly different wavelength in a spectral region outside of the previous scan region (Creemers and Völker 2000; Den Hartog et al. 1999b; Völker 1989a, b).

Fig. 2

Pulse sequence used in time-resolved hole-burning (HB) experiments. Top: Timing of the laser pulses with t _p: probe time, t _b: burn time and t _d: delay time. Bottom: Frequency ramp and steps with Δν: change in laser frequency (Den Hartog et al. 1999b)

Experimental set-up for continuous-wave hole burning

The experimental set-up used in our laboratory to perform CW hole-burning experiments is depicted in Fig. 3a. A single-frequency, CW titanium:sapphire ring laser (bandwidth ~0.5 MHz, tunable from ~700 to 1,000 nm) or a dye laser (bandwidth ~1 MHz, tunable between ~550 and 700 nm), both pumped by an Ar⁺ laser (2–15 W), is used. The intensity of the laser light is stabilized with a feedback loop consisting of an electro-optic modulator (EOM), a photodiode (PD) and control circuitry for Light-Intensity Stabilization (LIS). The wavelength of the laser is calibrated with a wavemeter (resolution Δλ/λ ~ 10⁻⁷) and the mode structure of the laser is monitored with a confocal Fabry–Perot (FP) etalon (free spectral range, FSR = 300 MHz, 1.5 GHz or 8 GHz). Burning power densities P/A (P is the power of the laser, and A is the area of the laser beam on the sample) between ~1 μW/cm² and a few 100 μW/cm², with burning times t _b from ~5 to ~100 s, are generally used.

Fig. 3

Top: a Set-up for CW hole burning. Either a CW (continuous wave), single-frequency titanium-sapphire (bandwidth 0.5 MHz) or a dye laser (bandwidth 1–2 MHz) was used. EOM electro-optic modulator, LIS electronic control circuit for light-intensity stabilization, BS beam splitter, PD photodiode, PL polarizer, λ-meter Michelson interferometer, FP Fabry–Perot etalon, M mirror, OD neutral density filters, Sh shutter, L lens, S sample, C liquid-⁴He-cryostat, PM (cooled) photomultiplier, ADC analogue-to-digital converter, DAC digital-to-analogue converter. Bottom: b Time-resolved hole-burning set-up. Either a CW single-frequency temperature- and current-controlled (T- and I-control) diode laser, or a titanium:sapphire laser, or a dye laser (see the above panel, a) was used. OI optical isolator, AOM/D acousto-optic modulator and driver, A diaphragm, Amp amplifier, P&D GEN pulse- and delay generator, WF GEN waveform synthesiser, ⊕ summing amplifier, DIG SCOPE digital oscilloscope, PIA peripheral interface adapter (Adapted from Creemers and Völker 2000)

The holes are either probed in fluorescence excitation at 90° to the direction of excitation or in transmission through the sample, with the same laser but with the power attenuated by a factor of 10–10³. The intensity of the probe pulse is reduced with a neutral density filter. The fluorescence or transmission signal of the hole is detected with a cooled photomultiplier (PM) and subsequently amplified with an electrometer. The signals are digitized and averaged point by point 1,000 times with a computer (PC) and the pulse scheme of Fig. 2 is used only once and not cycled through (see below). The experiments are controlled with a PC (Creemers and Völker 2000; Völker 1989a, b).

Experimental set-up for time-resolved hole burning

To perform time-resolved hole-burning experiments (see Fig. 3b), various types of CW single-frequency lasers are used, in combination with acousto-optic modulators (AOMs), to create the pulse sequence described in Fig. 2. The choice of the laser depends on the absorption wavelength of the sample and the time scale of the experiment (Creemers and Völker 2000; Creemers et al. 1997; Den Hartog et al. 1998a, 1999a, b; Koedijk et al. 1996; Störkel et al. 1998; Wannemacher et al. 1993). For delay times t _d, shorter than a few 100 ms and down to microseconds, we use current- and temperature-controlled single-mode diode lasers. The type of diode laser depends on the wavelength needed. The main advantage of these semiconductor lasers is that their frequency can be scanned very fast, up to ~10 GHz/μs, by sweeping the current through the diode. A disadvantage is their restricted wavelength region (5–10 nm, tunable by changing the temperature of the laser). The bandwidth of these diode lasers is ~3 MHz (Den Hartog et al. 1999b). For delay times t _d longer than ~100 ms, either a CW single-frequency titanium:sapphire (bandwidth ~0.5 MHz) or a dye laser (bandwidth ~1 MHz) is used. The frequency of these lasers can be scanned continuously over 30 GHz with a maximum scan speed limited to ~100 MHz/ms by piezoelectric-driven mirrors. This speed is about 10⁴–10⁵ times slower than that of diode lasers (Creemers and Völker 2000; Den Hartog et al. 1999b).

Burning power densities (Pt/A) between ~50 nW/cm² and 20 mW/cm², with burning times t _b ranging from 1 μs to ~100 s, are generally used. The delay time t _d between burning and probing the holes varies from ~1 μs to ~24 h. For delay times shorter than ~100 s, the burn and probe pulses are produced with two AOMs in series (two instead of one to reduce the laser light leaking through them when switched off, suppression better than 10⁶). For delay times t _d longer than ~100 s, the intensity of the probe pulse is reduced with a neutral density filter.

The holes are probed in fluorescence excitation with a cooled photomultiplier (PM) perpendicular to the direction of excitation. The signals before and after burning are stored in two channels of a digital oscilloscope, amplified and averaged in different ways, depending on delay time. For t _d < 100 ms, a sequence of probe–burn–probe cycles is applied with a repetition rate ≤10 Hz using home-built electronics (see Fig. 3b) and then summed. After each probe–burn–probe cycle, the frequency of the laser is slightly shifted (by a few times the hole width) to obtain a fresh baseline for each hole. Transient holes with a lifetime up to a few milliseconds are averaged 10³–10⁴ times, whereas persistent holes with delay times shorter than ~100 s are averaged 50–100 times with the digital oscilloscope. For delay times t _d > 100 s, the signals are averaged point by point about 1,000 times with the PC, with a total number of 200–1000 points per scan, depending on t _d (see previous section). Experiments are controlled with the PC.

Examples from photosynthesis studied with hole burning

Energy transfer and optical dephasing: hole width as a function of temperature

Examples presented below will show how energy-transfer times and information on optical dephasing can be obtained for light-harvesting (LH) complexes of purple bacteria by measuring the hole width as a function of temperature. LH complexes (antennas) in photosynthetic systems are responsible for the efficient collection of sunlight and the transfer of excitation energy to the reaction center (RC). The primary charge separation, which occurs in the RC, leads to the subsequent conversion of the excitation energy into a chemically useful form. The function of the antenna is to improve the absorption cross-section of the individual RCs. Each RC is surrounded by many LH complexes (Blankenship 2002; Sundström et al. 1999; Van Amerongen et al. 2000; Van Grondelle et al. 1994).

Most purple bacteria contain two types of LH complexes: the LH1 core complex surrounding each RC, and peripheral LH2 complexes that absorb slightly to the blue and transfer energy to LH1 (Cogdell et al. 2006; Fleming and Scholes 2004; Hu et al. 2002; Sundström et al. 1999; Van Amerongen et al. 2000; Van Grondelle and Novoderezhkin 2006). Both the LH1 and the LH2 complexes have concentric ring-like structures. The LH1 complex has only one absorption band at ~875 nm. In contrast, the LH2 complex of Rhodobacter (Rb.) sphaeroides (discussed below) has two absorption bands at 800 and 850 nm, as shown in Fig. 4 (bottom). The pigments in these purple bacteria are bacteriochlorophyll a (BChl a) molecules and carotenoids non-covalently bound to a pair of small trans-membrane α and β polypeptides. The B800 ring in Rhodopseudomonas (Rps.) acidophila consists of nine in-plane BChl a monomers, whereas the B850 ring is formed by a collection of 18 BChls distributed along the ring in 9 dimer subunits (McDermott et al. 1995; Papiz et al. 2003). Their planes are perpendicular to those of the BChls in the B800 ring (see Fig. 4, top). The X-ray structure of Rhodosprillum (Rs) molischianum is similar to that of Rps. acidophila, with 8 BChls in the B800 ring and 16 BChls in B850 (Koepke et al. 1996). Cryoelectron microscopy has shown that the structure of the LH2 complex of Rb. sphaeroides (Walz et al. 1998) is also similar to that of Rps. acidophila.

Fig. 4

Top: Arrangement of the bacteriochlorophyll a (BChl a) molecules in the B800 and B850 rings of the light-harvesting (LH) 2 complex (left: side view, right: top view; Data from www.pdb.bnl.gov.) Bottom: Excitation spectrum of the LH2 complex of Rb. sphaeroides (2.4.1, wt) at liquid-helium temperature (Spectrum obtained in our laboratory)

Energy transfer from B800 to B850 in light-harvesting 2 complexes of purple bacteria

The wavelength selectivity and high-frequency resolution of spectral hole burning is particularly advantageous for the study of pigment–protein complexes that are characterized by broad absorption bands. The first HB experiments on photosynthetic complexes were performed by G. Small and his group in the 1980s on the RC of purple bacteria (Hayes and Small 1986; Lyle et al. 1993, and references therein; Tang et al. 1988), and on photosystem I (Gillie et al. 1989) and the RC of photosystem II (Jankowiak et al. 1989; Tang et al. 1990) of green plants and cyanobacteria. Here, we describe HB experiments performed in our laboratory, in Leiden, The Netherlands, on the red wing of the B800 band of LH2 at liquid-helium temperature (De Caro et al. 1994; Van der Laan et al. 1990, 1993). The results of these experiments proved, for the first time, that the B800 band is inhomogeneously broadened because holes could be burned into this band. As described earlier in this review, the widths of spectral holes are a measure for the homogeneous linewidth Γ_hom of the optical transition, under the condition that the laser bandwidth is negligible compared to Γ_hom. If the ‘pure’ dephasing time $T_{2}^{*}$ in Eq. 1 is much larger than T ₁, i.e. $T_{2}^{*} \gg T_{1} ,$ then Γ_hom will be determined by T ₁ processes. Thus,

$$ \Upgamma_{\hom } \approx {\frac{1}{{2{{\uppi}}T_{1} }}} = {\frac{1}{{2{{\uppi}}\tau_{\text{fl}} }}} + {\frac{1}{{2{{\uppi}}\tau_{\text{ET}} }}} $$

(2)

, where τ _fl is the fluorescence lifetime, and τ _ET is the energy-transfer time. If the latter is much shorter than τ _fl, for example, τ _ET approximately a few picoseconds, Γ_hom will directly yield the energy-transfer rate (2πτ _ET)⁻¹.

In the experiments of De Caro et al. (1994) and Van der Laan et al. (1990), where holes were burnt into the red wing of the B800 band of Rb. sphaeroides 2.4.1 (wild type, wt), hole widths were found of ½Γ_hole = Γ_hom ~ 65 GHz, independent of burning wavelength between 791 and 804 nm. Since this width is much larger than the fluorescence lifetime-limited value, (2πτ _fl)⁻¹ ~100 MHz (corresponding to a τ _fl of a few ns), and the value of Γ_hom proved independent of temperature between ~1.2 and 30 K (no holes could be burnt at T > 30 K), Van der Laan et al. (1990) concluded that Γ_hom is entirely given by the energy-transfer rate from B800 to B850, which corresponds to τ _B800→B850 = 2.3 (±0.4) ps. In Fig. 5, the value of Γ_hom is plotted as a function of temperature. This result was subsequently confirmed by HB experiments from the group of G. Small (Reddy et al. 1991) and by femtosecond time-resolved pump-probe experiments (Scholes and Fleming 2000; Sundström et al. 1999; Van Amerongen et al. 2000, and references therein).

Fig. 5

Temperature dependence of the homogeneous linewidth Γ_hom of the electronic transition in the red wing of the B800 band of the isolated LH2 complex of Rb. sphaeroides (2.4.1, wt), between 1.2 and 30 K. The value of Γ_hom = 69 ± 10 GHz is shown here to be independent of temperature. It represents the energy-transfer rate between B800 and B850 (Van der Laan et al. 1990)

Additional HB experiments from our laboratory on various LH2 mutants of Rb. sphaeroides with blue-shifted B850 bands (Fowler et al. 1992) and on the B800–B820 complex of Rps. acidophila at liquid-helium temperature have shown that the transfer times from B800 to B850 vary at most between 1.7 and 2.5 ps (De Caro et al. 1994; Van der Laan et al. 1993). These results were interpreted with Förster’s mechanism for energy transfer (Förster 1948, 1965), assuming that energy is transferred from the 0–0 transition of B800 to a broad vibronic band of B850 overlapping with B800. From this model, the distance between the B800 donor and the B850 acceptor molecules was estimated to be R _DA = 1.5–1.9 nm for the various LH2 complexes (Van der Laan et al. 1993). These values agreed surprisingly well with the distance of 1.76 nm between the B800 and B850 rings subsequently determined by X-ray crystallography (McDermott et al. 1995). Since, then, more refined methods have been developed to estimate the B800–B850 energy-transfer rates, which are based on a generalized Förster theory for multi-chromophoric systems (Beljonne et al. 2009, and references therein; Cheng and Silbey 2006; Fleming and Scholes 2004; Jang et al. 2004; Scholes and Fleming 2000, 2005) and on a modified Redfield theory (Van Grondelle and Novoderezhkin 2006, and references therein).

In our research group, not only was the inter-band B800 → B850 energy transfer studied but also the intra-band B800 → B800 transfer by means of FLN and HB as a function of excitation wavelength λ_exc. From FLN, i.e. from the shift of narrow emission lines with λ_exc in the red wing of the B800 band (797–805 nm) and from the independence of the hole widths with λ_exc in this spectral region, we concluded that here B800 → B850 transfer dominates with τ _B800→B850 = 2.5 ± 0.2 ps for Rb. sphaeroides and 2.0 ± 0.1 ps for Rps. acidophila at liquid-helium temperature (De Caro et al. 1994). When exciting towards the blue within the B800 band (λ_exc < 798 nm), the fluorescence signals become broad and shift towards the red, while Γ_hom increases from 60–80 to ~250 GHz (between 798 nm and, at least, 788 nm). In this spectral region, inter-band B800 → B850 competes with intra-band B800 → B800 transfer, and intra-band energy-transfer times become τ_B800→B800 ≈ 900 fs between λ_exc ~ 780 and 798 nm. At λ_exc < 780 nm, non-selective excitation in vibronic transitions of the B800 band takes place. The resulting fluorescence is broad with a peak at about 805 nm, independent of λ_exc. In this region, B800 → B800 ‘downhill’ transfer and vibrational relaxation are the dominant processes. We conclude from these examples that FLN in combination with HB are powerful techniques for unravelling energy-transfer rates in photosynthetic complexes at low temperature.

(For discussions on energy transfer in bacterial LH complexes, see also Cheng and Silbey (2006), Novoderezhkin et al. (2003), Scholes and Fleming (2000), Sundström et al. (1999), Van Amerongen et al. (2000), Wu et al. (1996) and Zazubovich et al. (2002a).)

Optical dephasing in the B850 band of purple bacteria

The strong interactions between nearest-neighbour BChl molecules in the B850 band of LH2, with distances of less than 1 nm, lead to delocalization of the excitation to an extent that is limited by static and dynamic disorder (Cogdell et al. 2006; Hu et al. 2002; Krueger et al. 1998; Scholes et al. 1999; Sundström et al. 1999). We will come back to this subject later. Here, we discuss the role of the protein structure in controlling the excited-state dynamics of the BChl a pigments in the B850 band. As shown above, the dynamics of a pigment within a protein is reflected by the homogeneous linewidth Γ_hom. In the case of B800, we saw that $T_{2}^{*} \gg T_{1}$ with Γ_hom determined by inter-band (B800 → B850) and intra-band (B800 → B800) energy-transfer processes. Here, we will show that in the red wing of the B850 band, Γ_hom is dominated by optical dephasing $\left( {T_{2}^{*} } \right)$ processes characterized by a value of Γ_hom that is temperature dependent. Experiments were performed in our laboratory on Rb. sphaeroides (G1C, mutant): holes were burnt at a given temperature and Γ_hole measured as a function of burning-fluence density Pt/A. The hole widths are plotted versus Pt/A in Fig. 6a (J. Gallus and L. van den Aarssen, unpublished results). The value of Γ_hom is obtained from such a plot by extrapolating ½Γ_hole to Pt/A → 0. Similar measurements were done for temperatures between 1.2 and 4.2 K.

Fig. 6

Top: a Hole width, ½ Γ_hole, as a function of burning-fluence density, Pt/A, of a hole burnt in the red wing of the B850 band of the LH2 complex of Rb. sphaeroides (G1C, mutant) at 1.8 K. The extrapolation of ½ Γ_hole to Pt/A → 0 yields the homogeneous linewidth, Γ_hom, at a specific temperature. Inset: Hole burnt at Pt/A ~ 0.2 J/cm². Bottom: b Homogeneous linewidth, Γ_hom, as a function of temperature T between 1.2 and 4 K in the red wing of the B850 band. Γ₀ is the residual homogeneous linewidth for T → 0. Its value is consistent with a fluorescence lifetime of a few nanoseconds (J. Gallus and L. van den Aarssen, unpublished results from our laboratory)

Figure 6b shows a plot of the homogeneous linewidth Γ_hom as a function of temperature (J. Gallus and L. van den Aarssen, unpublished results). We found small values of Γ_hom, between ~0.5 GHz and a few GHz at the red wing of the B850 band, as compared to those in B800. The values in B850 are determined by ‘pure’ dephasing processes $\left( {T_{2}^{*} } \right),$ i.e. by fluctuations of the optical transition arising from coupling of the BChl a pigments to the surrounding protein. The values for B800, in contrast, are limited by T ₁ processes, i.e. by energy transfer from B800 to B850 and from B800 to B800 (De Caro et al. 1994; Van der Laan et al. 1990, 1993). The temperature dependence of Γ_hom, in Fig. 6b, follows a T ^α power law, with α = 1.3 ± 0.1. Similar behaviour was found for chromophores in amorphous hosts (for reviews, see Jankowiak et al. 1993; Moerner 1988, and articles therein; Völker 1989a, 1989b), for BChl a in a triethylamine glass (Van der Laan et al. 1992) and for other photosynthetic systems, such as the B820 and B777 subunits of LH1 (Creemers and Völker 2000; Creemers et al. 1999a; Störkel et al. 1998), and the PSII RC (Den Hartog et al. 1998c, 1999b; Groot et al. 1996) and CP47-RC (Den Hartog et al. 1998b) of green plants between 1.2 and 4.2 K. The dephasing times in photosynthetic systems, however, are about one to two orders of magnitude larger than in glassy systems, indicating that there is rather strong coupling between the pigments and protein. Here, optical dephasing is assumed to arise from coupling of the energy levels of the chromophore or pigment to a distribution of TLSs of the glassy host or protein (Jankowiak and Small 1993; Putikka and Huber 1987; Völker 1989a, b).

In contrast to the systems mentioned above, a crystalline-like T^2±0.2 hole-width dependence was reported for the CP43 and CP47 ‘trap’ pigments in O₂-evolving PSII core complexes between 2.5 and 18 K (Hughes et al. 2005).

The extrapolation value Γ₀ = (2πτ _fl)⁻¹ for T → 0 in Fig. 6b is consistent with a fluorescence lifetime τ _fl of BChl a of a few ns (Sundström et al. 1999). Thus, our dephasing results disprove the existence of residual exciton scattering at T → 0, which was assumed to contribute to the much broader holes reported by Wu et al. (1997c) for the red wing of the B850 band of LH2 of Rps. acidophila. Although a T ^1.3 dependence of Γ_hom was also reported for HB experiments performed between 4.2 and 20 K (Wu et al. 1997b), the value of Γ_hom at 4.2 K was about five times larger than the one obtained in our group at the same temperature. On the other hand, accumulated photon echo (APE) experiments on the same system (Lampoura et al. 2000) yielded values of Γ_hom at 4.2 K that were about five times smaller than those by Wu et al. (1997b). Lampoura et al. (2000) suggested that the discrepancy between the results from the APE experiments and HB experiments was due to spectral diffusion, since the experimental time scales in APE are much smaller than those in HB (picosecond vs. minutes, respectively). However, our HB results at 4.2 K coincide with those of the APE experiments, from which we conclude that the APE–HB discrepancy does not arise from spectral diffusion, but is caused by the much too high burning fluences used in the HB experiments of Wu et al. (1997b). This shows that Γ_hom values extracted from HB experiments are reliable only when obtained from an extrapolation of the hole width to Pt/A → 0, as shown in Fig. 6a and b.

Spectral diffusion: hole widths as a function of delay time

The dependence of spectral diffusion on the size of photosynthetic complexes

Proteins are materials that display both crystalline and glassy properties. On the one hand, they have rather well-defined tertiary structures reflected in their crystalline properties. On the other hand, and in contrast to crystals, the structures of proteins are not static: they may undergo conformational changes between a large number of somewhat different intermediates called conformational sub-states (CSs; Frauenfelder et al. 1991, 2001; Friedrich et al. 1994; Hofmann et al. 2003; Rutkauskas et al. 2004, 2006). These CSs are separated by a wide distribution of energy barriers with multiple minima on a potential energy landscape, reminiscent of TLSs in glasses. TLSs, however, are randomly distributed, whereas CSs are assumed to be hierarchically organized, possessing a large degree of complexity. Whether conformational changes in proteins have a continuous distribution of relaxation rates as observed in glasses (Koedijk et al. 1996; Littau et al. 1992; Meijers and Wiersma 1994; Silbey et al. 1996; Wannemacher et al. 1993), or are characterized by discrete and sharp rates (Thorn-Leeson and Wiersma 1995; Thorn-Leeson et al. 1997), is still a controversial issue (Baier et al. 2007, 2008; Schlichter and Friedrich 2001; for reviews, see Berlin et al. 2006, 2007).

One way to study the conformational dynamics of proteins is by following their time evolution through spectral diffusion (SD; Berlin et al. 2006; Creemers and Völker 2000; Den Hartog et al. 1999b; Richter et al. 2008; Schlichter and Friedrich 2001; Störkel et al. 1998). Here, we show that the size of the protein influences the amount of SD in photosynthetic pigment–protein complexes. We have investigated three sub-core complexes of photosystem II (PSII) of green plants (spinach) at low temperature by time-resolved spectral hole burning, covering 10 orders of magnitude in time (Den Hartog et al. 1999b; Groot et al. 1996): the isolated PSII RC, the inner core antenna CP47 and the CP47-RC complexes. The samples used in these experiments were prepared by J. Dekker and collaborators (Dekker et al. 1989, 1990; Eijckelhoff and Dekker 1995; Kwa et al. 1992). They were subsequently diluted in buffer and glycerol to work at low temperature (Den Hartog et al. 1998b).

The SD behaviour of the PSII sub-core complexes is compared here with that of B777, the monomer subunit of the LH1 complex of purple bacteria. B777 was obtained from LH1 by adding the detergent n-octyl-β-glucopiranoside (OG) and diluted in buffer and glycerol (Creemers et al. 1999a, and references therein). The B777 complex, in turn, is compared with BChl a embedded in the same OG detergent (diluted in buffer and glycerol) without the protein, which we call here BChl a in OG-glass (Creemers and Völker 2000). The purpose of this experiment was two-fold, to compare the SD behaviour of proteins with that of glasses, and to clear up a long-standing problem: whether the BChl a molecule in B777 is bound or not to the protein (Sturgis and Robert 1994, and references therein). HB results on SD of B820, the dimer subunit of LH1, at various temperatures and delay times, and its comparison to glasses, can be found in Störkel et al. (1998).

Photosystem II (PSII), the ‘engine of life’, is a large complex embedded in the thylakoid membranes of plants, algae and cyanobacteria. Driven by solar energy, PSII catalyzes the splitting of water into oxygen which is essential for the survival of life on Earth (for a review, see Barber 2008). The events that give rise to the primary and secondary electron-transfer processes, which lead to water oxidation start with the absorption of sunlight by a peripheral light-harvesting complex, called LHCII (Kühlbrandt et al. 1994), which transfers the excitation energy to the RC within the PSII core complex. The isolated PSII RC, which is the smallest unit that shows photochemical activity (Nanba and Satoh 1987; Rhee et al. 1997), is composed of the D1 and D2 proteins and bound mainly to the CP43 and CP47 complexes (Boekema et al. 1998; Dekker and Boekema 2005). The D1 and D2 proteins contain the cofactors that bring about charge separation.

The crystal structures of cyanobacterial PSII, determined by X-ray crystallography at 3.5 Å (Ferreira et al. 2004) and 3 Å (Loll et al. 2005) resolution, confirmed the dimeric organization of the isolated complex and the positioning of the major subunits within each monomer, previously obtained by electron crystallography (Eijckelhoff et al. 1997; Rhee et al. 1997). Loll et al. (2005) concluded that there are about 36 Chl a and 11 β-carotene molecules per PSII core, and that the CP43 and CP47 complexes bind 13 and 16 Chls, respectively, while the RC binds 6 Chls, 2 pheophytin (Pheo) molecules, 2 plastoquinone (PQ) molecules, at least one β-carotene and a non-heme Fe. A cluster of four weakly coupled Chl a molecules in the center of the RC forms the delocalized P680* excited state (Durrant et al. 1995), where a short-lived charge-transfer state is created before the subsequent electron-transfer processes take place. This picture is consistent with the so-called multimer models (Durrant et al. 1995; Jankowiak et al. 2002; Prokhorenko and Holzwarth 2000). Other models for energy transfer and charge separation in PSII, based on decoupled pigments with monomeric absorption, have also been reported (Diner and Rappaport 2002).

A discussion on the nature of P680 and the relation to a far red-absorbing (700–730 nm) complex that induces charge separation in intact O₂-evolving PSII RCs, can be found in Hughes et al. (2005, 2006b), Krausz et al. (2008, and references therein) and Peterson-Årsköld et al. (2004).

Time-resolved HB experiments were performed, in our laboratory, in red-absorbing pigments of the isolated PSII sub-core complexes that act as ‘traps’ for energy transfer, i.e. in pigments characterized by a fluorescence decay time of a few nanoseconds and therefore yielding narrow holes. In the presence of SD, the holes broaden with delay time t _d, the time between burning and detecting the hole. From such holes, the ‘effective’ homogeneous linewidth $\Upgamma_{\hom }^{'} (t_{\text{d}} )$ is determined, which reflects the occurrence of time-dependent conformational changes in the protein or glassy host. $\Upgamma_{\hom }^{'} (t_{\text{d}} )$ can be expressed as:

$$ \Upgamma_{\hom }^{'} \;(T,t_{\text{d}} )\; = \;{\frac{1}{{2\,\pi \,T_{1} }}}\; + \;{\frac{1}{{\pi \,T_{2}^{*} \left( {T,t_{\text{d}} } \right)}}} = \Upgamma_{0} \; + \;\left( {a_{\text{PD}} \; + \;a_{\text{SD}} (t_{\text{d}} )} \right)\;T^{1.3\,} , $$

(3)

where in the absence of energy transfer, Γ₀ is determined by the fluorescence lifetime τ _fl, Γ₀ = (2πτ _fl)⁻¹ (see Creemers and Völker 2000; Den Hartog et al. 1999b; Koedijk et al. 1996; Silbey et al. 1996; Wannemacher et al. 1993). The last term in Eq. 3 consists of two contributions: a ‘pure’ dephasing contribution a _PD T ^1.3 (always present) that accounts for fast fluctuations of the optical transition within the lifetime of the excited state of a few ns, and a delay-time-dependent contribution determined by spectral diffusion a _SD (t _d) T ^1.3 that increases with t _d. Hence, following from Eq. 3:

$$ a_{\text{SD}} (t_{\text{d}} )\; = \;{\frac{{\Upgamma_{\hom }^{'} (t_{\text{d}} )\; - \;\Upgamma_{0} }}{{T^{1.3\,} }}}\; - \;a_{\text{PD}} , $$

(4)

where the functional dependence of the coupling constant a _SD on delay time t _d yields the distribution P(R) of relaxation rates R in the protein (see below and Fig. 7).

Fig. 7

Coupling constant a _SD of spectral diffusion (SD) as a function of the logarithm of the delay time between burning and probing, t _d. B777-subunit of the LH1 complex of a purple bacterium (open red triangles), and three ‘trap’ pigments for energy transfer of the isolated PSII sub-core complexes of higher plants: CP47 (open orange ovals), RC (closed green squares) and CP47-RC (closed blue triangles). The data are compared to those of BChl a in the detergent n-octyl-β-glucopiranoside (OG) embedded in a buffer-glycerol glass (BChl a in OG-glass; open grey circles). The mass of the protein is given in parenthesis in kDa. Note the correlation between the amount and onset of SD and the mass of the protein: the larger the mass, the slower the SD (Den Hartog et al. 1999b). The difference between the results obtained for B777 and BChl a in OG-glass proves that the BChl a molecules in B777 are bound to the protein (Creemers and Völker 2000)

The log–log dependence of a _SD on t _d for the three sub-core complexes of PSII is shown in Fig. 7, with t _d varying between 10⁻⁶ s (microseconds) and 10⁴ s (a few hours), and temperatures from 1.2 to 4.2 K. The results are compared in the same figure with those obtained for B777, the monomer subunit of LH1 (red curve), and BChl a in OG-glass (grey curve). The latter shows a typical glass-like behaviour, with a _SD increasing linearly with log (t _d) over at least 15 orders of magnitude in time (10⁻⁹–10⁵ s), indicating that the distribution of relaxation rates P(R) is continuous and proportional to 1/R (Koedijk et al. 1996; Silbey et al. 1996; Wannemacher et al. 1993). In contrast, the B777 subunit of LH1, which consists of a BChl a monomer surrounded by protein and dissolved in OG-glass, qualitatively displays the behaviour of the PSII sub-core complexes: for short delay times, a _SD is constant and the results seem to be determined by ‘pure’ dephasing, i.e. by fast, local fluctuations. Thus, for short times, the protein appears to be rather rigid and to behave as a crystal in the direct vicinity of the excited pigments. The onset of SD at longer delay times and the logarithmic delay-time dependence of $\Upgamma_{\hom }^{'}$ suggest that slow fluctuations are involved in conformational relaxation (at least at low T), implying that protein motions have a broad and continuous 1/R distribution of low-frequency rates R with a cut-off frequency equal to t ⁻¹_d at the onset of SD. These motions probably take place at the interface between the protein and buffer-glycerol glass, where there is more structural flexibility.

If we take a closer look at Fig. 7, we see that the onset of SD as well as the slope of the curves depend on the complex studied (Den Hartog et al. 1999b). B777 (with a protein mass of ~6 kDa (Sturgis and Robert 1994)) has its onset of SD at the shortest delay time (t _d ~ 10 ms) and shows the largest slope ${\text{d}}\Upgamma_{\hom }^{'} /{\text{d}}\log t_{\text{d}}$, whereas CP47 (~70 kDa; Chang et al. 1994) starts SD at t _d ~ 300 ms, and RC (~110 kDa; Eijckelhoff and Dekker 1995) starts SD at t _d ~ 1 s. Correspondingly, the slope of CP47 is larger than that of RC, indicating a larger amount of SD in CP47. Surprisingly, CP47–RC (~180 kDa; Eijckelhoff et al. 1997) does not show any SD over the time and temperature ranges investigated, from which we conclude that this sample appears to be rather rigid or, at least, it does not show conformational changes at low frequencies, but only undergoes fast, crystalline-like fluctuations (i.e., dephasing) at T ≤ 4.2 K. Thus, motions involving the entire complex (or a part of it) take place in these protein systems, even at liquid-helium temperature. It is further striking that the slopes in Fig. 7 seem to be correlated with the mass or size of the protein, and not with the number of pigments in these proteins (1 in B777, 8 in RC, 16 in CP47 and ~24 in CP47–RC). The results of Fig. 7 indicate that at low temperature and short delay times (t _d < ms), there is no SD, but only ‘pure’ dephasing, i.e. local, fast fluctuations remain. At longer times, very slow motions (with cut-off frequencies of 1–100 Hz) take place, probably at the protein–glass interface (Creemers and Völker 2000; Den Hartog et al. 1999b).

If we assume that the amount of SD is proportional to the pigment–protein interaction ($\propto \left( {r^{n} } \right)^{ - 1}$ for multipolar types with n ≥ 3) and to the number of TLSs present at the surface of the protein $\left( { \propto r^{2} } \right),$ then SD $\approx {\text{d}}\Upgamma_{\hom }^{'} /{\text{d}}t_{\text{d}} \propto \left( {r^{n - 2} } \right)^{ - 1} \propto r^{ - 1}$ (for n = 3; Den Hartog et al. 1999b). SD should thus increase with decreasing r, i.e. with decreasing size of the protein (or with its mass, for constant density). In conclusion, the heavier the protein, the smaller the amount of SD. The nature of the protein motions involved, however, is still unknown and, as mentioned above, it is a matter of controversy whether TLSs are a useful concept for explaining the dynamics of proteins at low temperatures. (For recent reviews, see Berlin et al. (2006, 2007), where an anomalous power law in waiting time was observed for heme proteins at low temperature.) More time-resolved HB experiments on larger complexes, combined with different solvents, and at higher temperatures may shed some light on these unsolved issues.

Hidden spectral bands made visible: hole depth as a function of wavelength

The advantages of HB, as compared to ultrafast time-resolved techniques, are the high spectral resolution (of a few MHz) and the wavelength and burning-fluence selectivity. These properties make HB an attractive tool for disentangling spectral bands ‘hidden’ in strongly heterogeneously broadened and overlapping absorption bands. The disentanglement can be achieved by measuring the hole depth, in addition to the hole width, as a function of excitation wavelength, at constant (and low) burning-fluence density (Pt/A) and at liquid-helium temperature. Such ‘action’ spectra were first reported by the group of G. Small for LH1 and LH2 (Reddy et al. 1992, 1993; Wu et al. 1997a, b, c) and, subsequently, by A. Freiberg and co-workers for the same systems (Freiberg et al. 2003, 2009 and references therein; Timpmann et al. 2004), although the hole widths in these experiments were not determined by Γ_hom.

Holes, which do change their depth but keep their value of Γ_hom constant, are a proof that only those pigments that are involved in a specific dynamic process, with a characteristic decay or dephasing time, have been selected by hole burning. Two examples from our laboratory, in which ‘hidden’ spectra have been made visible in this way, are presented in this review: the first example deals with ‘traps’ for energy transfer in PSII complexes of green plants; the second one discusses the distribution of the lowest k = 0 exciton states in the B850 band of LH2 complexes of purple bacteria. In the first example, we show that, by means of FLN and HB, pigments within the isolated PSII RC, CP47 and CP47-RC complexes that do not participate in energy transfer can be distinguished by their decay times from those that do participate (Den Hartog et al. 1998b). ‘Trap’ pigments display narrow holes because the excited pigments decay in a few nanoseconds to the ground state by fluorescence. They can be separated from the pigments that participate in energy transfer as the latter have fast excited-state decay times and, therefore, show broad and shallow holes. The spectral distribution of the depths of the narrow holes, thus, represents the distribution of ‘traps’ for energy transfer.

The existence of CP43- and CP47-‘trap’ states in O₂-evolving PSII complexes has recently been reported (Hughes et al. 2005), and the assignment of the two quasi-degenerate red ‘trap’ states in CP43 and the origin of the HB mechanism in this system is presently a matter of debate in the literature (Dang et al. 2008; Hughes et al. 2006a; Jankowiak et al. 2000).

Here, we further prove that the spectral distribution of the lowest k = 0 exciton states within the B850 band of LH2 complexes of purple bacteria can be obtained in a manner similar to that described above: by measuring the depths of narrow holes as a function of excitation wavelength in the red wing of B850. In this case, the excited BChl a molecules belonging to the lowest k = 0 states decay directly to the ground state with a lifetime of a few nanoseconds (ns), leading to very narrow holes. Higher-lying k-states, absorbing in the middle to the blue side of the B850 band, have many pathways of de-activation and, as a consequence, their decay times are fast, usually a few tens to hundreds of femtoseconds (fs), even at low temperature (Novoderezhkin et al. 2003; Van Grondelle and Novoderezhkin 2006, and references therein). Such fast decay times correspond to hole widths that are orders of magnitude larger than those burnt in the lowest-lying k = 0 band. Such wide holes are usually not detectable since they are very shallow and disappear in the noise.

‘Traps’ for energy transfer in Photosystem II complexes of green plants

The original motivation for studying the isolated reaction center of photosystem II (PSII RC) of higher plants, the smallest unit in PSII that shows photochemical activity, called the D1–D2–cytochrome b559 complex (Nanba and Satoh 1987), was a lack of consensus in the literature regarding energy-transfer rates and charge-separation rates (Groot et al. 1994; Jankowiak et al. 1989; Klug et al. 1995; Roelofs et al. 1993; Tang et al. 1990). The controversy probably persisted because of the large overlap of strongly inhomogeneously broadened absorption bands in PSII RC between 660 and 690 nm (see Fig. 8a). As a consequence, sub-picosecond time-resolved experiments were difficult to interpret (Groot et al. 1996, and references therein).

Fig. 8

Spectral distributions of ‘trap’ pigments for energy transfer of various isolated sub-core complexes of Photosystem II, PSII (dashed lines) obtained from hole depths measured as a function of excitation wavelength and, subsequently, reconstructed within the fluorescence-excitation spectra. Top: a RC, Middle: b CP47, Bottom: c RC and CP47 ‘trap’ distributions in the RC-, CP47- and CP47-RC-complexes of PSII. The intensities of the ‘trap’ distributions have been normalized to match the red wing of their respective absorption spectra. The RC and CP47 ‘traps’ are also present in the CP47-RC complex (Den Hartog et al. 1998b; Groot et al. 1996)

To verify whether low-lying energy ‘trap’ pigments in PSII RC at low temperature exist, and to solve the contradictions related to energy transfer in PSII RC, spectral hole burning experiments from 1.2 to 4.2 K, between 665 and 690 nm, were performed in our research group (Groot et al. 1996). Since fluorescence-excitation spectroscopy was used to probe the holes, an excited pigment can only be detected if it fluoresces or transfers its excitation energy to another pigment which in turn fluoresces. As the excited primary donor P680* undergoes very fast charge separation, in much less than 30 ps (Greenfield et al. 1996; Klug et al. 1995; Wiederrecht et al. 1994), it practically does not fluoresce. Thus, only accessory ‘trap’ pigments are sensitive to hole burning detected in this way.

From holes burnt in the red wing of the absorption band of PSII (between ~665 and 690 nm) as a function of burning-fluence density (Pt/A) and temperature, and by extrapolation of the hole widths to Pt/A → 0 to obtain Γ_hom and, subsequently, by extrapolation of Γ_hom to T → 0, hole widths were found that are limited by a fluorescence lifetime of ~4 ns. This proved that accessory pigments acting as ‘4 ns traps’ for energy transfer are, indeed, present in PSII RC, at least at temperatures up to 4.2 K, with dynamics controlled by ‘pure’ dephasing processes (Groot et al. 1996). Such ‘traps’ at T < 50 K had been previously predicted from a kinetic model (Groot et al. 1994; Roelofs et al. 1993). They were later proven to exist by FLN experiments, in addition to HB experiments (Den Hartog et al. 1998b). In contrast, Tang et al. (1990) concluded from broad holes burnt at ~682 nm at 1.6 K, corresponding to decay times of 50 ps, that all accessory pigments transfer their energy to P680, implying the absence of accessory ‘trap’ pigments in PSII RC. The accessory pigments burnt at ~682 nm were attributed to pheophytin a (Pheo a). The hole widths in these experiments had not been extrapolated to Pt/A → 0.

In addition to hole widths, the spectral distribution of these ‘traps’ has also been determined in our laboratory by measuring the hole depth as a function of excitation wavelength at a constant, low burning-fluence density Pt/A (Groot et al. 1996). In the far red wing of the absorption band, the holes change their depth but not their width, indicating that this method indeed selects pigments involved in a specific dynamic process; here, it selects pigments decaying in 4 ns that do not transfer energy ‘downhill’. The distribution of ‘traps’ in PSII RC at 1.2 K is illustrated in Fig. 8a. Its shape is approximately Gaussian, with a width of ~143 cm⁻¹ and a maximum at ~682 nm (Groot et al. 1996). The linear electron–phonon coupling strength S of these ‘4 ns trap’ pigments was also determined by HB to be S ~ 0.73 (Groot et al. 1996), a value that agrees well with that reported for the Pheo a Q_y-state by Tang et al. (1990).

The contradictions in the literature about the existence of ‘traps’ for energy transfer are not only valid for PSII RC but also for the CP47 and CP47-RC complexes of PSII (Den Hartog et al. 1998b, and references therein). The CP47 protein, contained within the central core of PSII and proximate to the RC, is the last complex to be separated from the RC during isolation. It binds 16 Chl a molecules (Barber 2008; Ferreira et al. 2004; Loll et al. 2005) and two β-carotenes (Chang et al. 1994). To clear up the contradictions, it was important to determine the spectral distributions of pigments hidden under the broad absorption bands of these complexes. Two types of experiments were performed for this purpose in our research group: FLN at 1.2 K and HB between 1.2 and 4.2 K, both as a function of excitation wavelength. We will not discuss here how the results were obtained. A detailed account on the subject can be found in Den Hartog et al. (1998b), where it was shown that CP47 and CP47-RC at low temperature have distributions of pigments absorbing in their red wings (at ~690 nm) acting as ‘traps’ for the excitation energy and, therefore, do not transfer energy ‘downhill’. The CP47 ‘trap’ distribution, which has a width of ~200 cm⁻¹ and a maximum at ~690 nm, is depicted in Fig. 8b.

Results on CP47-RC, furthermore, suggested that the fluorescence in this complex originates from two types of ‘trap’ pigments, the CP47 component at ~690 nm and the RC component at ~682 nm, both fluorescing independently from each other. This is shown in Fig. 8c, where the CP47-RC absorption band has been decomposed into its components, CP47 and RC, each displaying its own ‘trap’. Since they fluoresce independently but have partly overlapping spectra, these two ‘traps’ have to be either at a distance from each other that is significantly larger than the Förster radius (R ₀ ≫ 3–8 nm), or they have to be in unfavourable relative orientations such that no energy transfer takes place from the RC to the CP47 complex. From the study of Den Hartog et al. (1998b), we conclude that the combination of HB and FLN experiments prove to be very powerful in unravelling spectral distributions of ‘traps’ for energy transfer in large photosynthetic complexes at liquid-helium temperatures, such as in CP47–RC, CP47 and the RC of PSII of green plants.

Lowest k = 0 exciton states in the B850 band of light-harvesting 2 complexes of purple bacteria

We know, from X-ray crystallography, that the B850 ring of the LH2 complex of Rps. acidophila consists of 18 close-lying BChl a molecules that are at distances of less than 1 nm from each other (McDermott et al. 1995; Papiz et al. 2003). Similar distances have been found within the B850 ring of Rs. molischianum (Koepke et al. 1996) and have been implied for Rb. sphaeroides from cryoelectron microscopy (Walz et al. 1998). Such short distances lead to strong electronic interactions of a few 100 cm⁻¹ and thus to delocalization of the excitation energy and the formation of coherent exciton states (Alden et al. 1997; Dahlbom et al. 2001; Freiberg et al. 1999; Hu et al. 1997, 2002; Krueger et al. 1998; Linnanto et al. 1999; Novoderezhkin et al. 1999, 2003; Sauer et al. 1996; Scholes and Fleming 2000; Scholes et al. 1999; Sundström et al. 1999; Wu et al. 1997b; Zazubovich et al. 2002b). The intensity of the B850 absorption band originates principally from two degenerate components of the excitation manifold, the k = ±1 (‘allowed’) states, labelled according to the assumed change in (pseudo) angular momentum. For a perfectly circular B850 ring, the excitation energy is delocalized over all 18 BChl a molecules and the lowest k = 0 exciton state is forbidden. Any deviation from this ideal situation, such as disorder, will localize the excitation energy over fewer BChl a molecules, allowing k = 0 to become (somewhat) radiative (Cheng and Silbey 2006; Freiberg et al. 1999, 2003; Hofmann et al. 2004; Jang and Silbey 2003; Jang et al. 2001; Novoderezhkin et al. 1999, 2003; Van Oijen et al. 1999; Wu et al. 1997a, b, c). The relative intensity of k = 0 with respect to that of k = ±1 is thus a measure of the extent of disorder in the B850 ring.

The degree of excitation-energy delocalization, which is limited by static and dynamic disorder, however, remains a subject of debate. Although the majority of the calculations are based on disordered Frenkel-exciton models (for reviews, see Cogdell et al. 2006; Hu et al. 2002; Jang et al. 2001; Scholes and Fleming 2000; Van Grondelle and Novoderezhkin 2006), an alternative polaron description leading to self-trapped excitons has been put forward by Freiberg and co-workers (Freiberg and Trinkunas 2009; Freiberg et al. 2009).

Static energy disorder, which reduces the coherence length of excitons in the B850 ring, may be caused by perturbations, such as local variations in hydrogen bonding between the BChls and the polypeptides, dielectric fluctuations and structural variations. These perturbations break the symmetry of the B850 ring that, in turn, affects the degree of delocalization. It is not clear yet whether the controversial measurements reported in the literature (Freiberg et al. 2003; Ketelaars et al. 2001; Rätsep et al. 2005; Reddy et al. 1992, 1993; Timpmann et al. 2004; Wu et al. 1997a, b, c; Zazubovich et al. 2002b) are related to the different experimental procedures used and/or to the differences in the bacteria studied.

We wanted to get a better understanding of the controversies and of the interplay between the coherence of the excitation that originates from the strong electronic coupling and the energy disorder in the B850 ring that tends to destroy the coherence. To this end, we have performed experiments in our laboratory on four types of LH2 complexes of purple bacteria at low temperature with one technique, spectral HB, for comparison (L. van den Aarssen, V. Koning and N. Verhart, unpublished results). In addition, we have done simulations of the total absorption band of the B850 ring, of the lowest k = 0 band and of their relative spectral positions and intensities (R Vlijm, L. van den Aarssen, V. Koning and N. Verhart, unpublished results) to test whether the assumptions made in a theoretical model developed by Silbey and collaborators (Jang et al. 2001; R. J. Silbey, personal communication) agree with the experiments. In the simulations, we have taken into account various types of static disorder, in addition to different coupling strengths and fast relaxation rates from higher-lying exciton states. Here, we focus on one system only, Rb. sphaeroides (2.4.1, wt), as an example, to show how we have made visible the spectral distribution of the lowest k = 0 exciton states, hidden under the broad B850 absorption band, by measuring the hole depth as a function of excitation wavelength.

Similar type of hole depth experiments on B850 have been reported by Freiberg et al. (2003, 2009, and references therein), and by Wu et al. (1997a, b, c) and Zabubovich et al. (2002b, and references therein). The burning-fluence densities used in the latter HB experiments, however, were more than 1,000 times larger than those used in our laboratory. Also, the detection of individual k = 0 states by single-molecule experiments on B850 of LH2 has been reported, but not their spectral distribution (Ketelaars et al. 2001).

The B850 band of LH2 consists of a number of exciton states with their homogeneous and inhomogeneous bandwidths. The inhomogeneous bandwidth of B850 is determined by intra- and inter-complex disorder, i.e. by disorder arising from within the B850 ring and between the rings. The individual exciton bands are thus hidden in the total B850 band. To determine the position and width of the lowest k = 0 exciton band, we first measured narrow holes of a few GHz with equal widths on the red wing of the B850 band at low temperature and plotted the hole width Γ_hole as a function of burning-fluence density, Pt/A, (not shown). The hole widths were then extrapolated to Pt/A → 0 (as in Fig. 6a) at each burning wavelength λ_burn to obtain the homogeneous linewidth Γ_hom. The depths of the narrow, homogeneously broadened holes (of equal width) at a given wavelength is proportional to the number of BChl a molecules contributing to the k = 0 band at this wavelength. The dependence of the hole depth on λ_burn, thus, represents the distribution of the lowest k = 0 exciton state.

The reason for the appearance of narrow holes in the red wing of the B850 band is that their width is limited by the fluorescence lifetime of a few nanoseconds of the lowest k = 0 exciton state. In contrast, higher-lying k-states decay to lower-lying k-states in tens to hundreds of femtoseconds (Alden et al. 1997; Novoderezhkin et al. 1999, 2003; Sundström et al. 1999, and references therein), which correspond to homogeneous linewidths that are 4–5 orders of magnitude larger. They contribute to extremely broad and very shallow holes that disappear within the noise, as mentioned above. The hole depths of the narrow holes burnt in the red wing of the B850 band of LH2 of Rb. sphaeroides (2.4.1, wt) are plotted as a function of burning wavelength in Fig. 9. They are well-fitted by a Gaussian curve with a width of ~190 cm⁻¹ and a maximum of ~866.0 nm. We have interpreted these data as representing the spectral distribution of the lowest k = 0 exciton states.

Fig. 9

Hole depth as a function of burning wavelength, for holes burnt in the red wing of the B850 band of Rb. sphaeroides (2.4.1, wt) at 1.2 K. The data were fitted with a Gaussian curve (hole-depth distribution) with a maximum at ~866.0 nm and a width of ~190 cm⁻¹ (V. Koning and N Verhart, unpublished results from our laboratory)

In Fig. 10, the hole-depth (k = 0) distribution of Fig. 9 has been inserted into the B850 band. This was done by matching the red wing of the k = 0 distribution to that of the B850 excitation spectrum. The intensity of the hole-depth distribution was scaled in such a fashion that the two red wings overlap. The result yielded a relative area of k = 0 / B850 ~ 9.5% and an energy difference between the two bands, Δ(B850 – k = 0) ~ 176 cm⁻¹ for Rb. sphaeroides (2.4.1, wt) (V. Koning and N. Verhart, unpublished results). Although the latter value is of the same order as that reported in the literature (~200 cm⁻¹), no values for the relative area for Rb. sphaeroides have been published.

Fig. 10

Excitation spectrum of the B850 band of Rb. sphaeroides (2.4.1, wt) at liquid-helium temperature with the hole-depth distribution from Fig. 9 (see also inset) built into it. The energy difference between the maxima of the B850 band and the hole-depth distribution is Δ(B850 − k = 0) ~ 176 cm⁻¹. The relative area of the hole-depth distribution with respect to that of the B850 band is k = 0/B850 ~ 9.5% (V. Koning and N. Verhart, unpublished results from our laboratory)

The four experimental parameters determined here, i.e. the widths of the B850 and k = 0 bands, the energy difference, Δ(B850 – k = 0) and the relative area, k = 0 / B850, were then used to find simulations that would fit the experiments. In the simulations, we have used nearest-neighbour interactions of ~300 to 400 cm⁻¹ (Cogdell et al. 2006; Jang et al. 2001; Sundström et al. 1999; Van Grondelle and Novoderezhkin 2006) and varied the amount of diagonal and off-diagonal disorder (Jang et al. 2001; R. J. Silbey, personal communication) until the calculated shapes, widths, positions and relative areas of the B850 and k = 0 bands would coincide with the experimental ones. Figure 11 shows both simulations and the experimental results for Rb. sphaeroides (2.4.1, wt). We note that the data are well-reproduced for this complex and for a mutant, Rb. sphaeroides (G1C) (results not shown), but are not so well-reproduced for other LH2 complexes examined in our laboratory. A detailed analysis of the data and the simulations for all the LH2 complexes of purple bacteria investigated in our research group and their comparison to data reported in the literature will be published elsewhere. With the examples presented here, we have demonstrated how hole depths measured as a function of burning wavelength can yield the spectral distribution of the lowest k = 0 exciton states hidden inside the broad B850 absorption band containing many higher-lying k-states. To our knowledge, HB is the only technique that is able to make such weak, hidden exciton distributions visible.

Fig. 11

Comparison of simulations, taking into account static correlated disorder (see text), with the experimental results obtained for the B850 band of Rb. sphaeroides (2.4.1, wt) at liquid-helium temperature, and the hole-depth distribution of Fig. 10. The simulation of B850 is shown in orange, while the experimental B850 is shown in grey. The simulation of the lowest k = 0 exciton band is shown in blue, while the hole-depth distribution is shown in red. A good match between simulations and experiments was found for Rb. sphaeroides (2.4.1, wt) as shown here, and for Rb. sphaeroides (G1C, mutant) (not shown; V. Koning and N. Verhart, unpublished results from our laboratory)

Concluding remarks

In this review, we show that spectral hole burning in its CW and time-resolved versions, in combination with site-selection spectroscopy (fluorescence line-narrowing), yields quantitative information on a number of dynamic processes taking place in the electronically excited states of photosynthetic pigment–protein complexes. Using very narrow-band (MHz), tunable, CW (dye, Ti:sapphire and semiconductor) lasers, it is possible to determine the homogeneous linewidth Γ_hom of an optical transition that is hidden in an inhomogeneously broadened absorption band. To obtain a reliable value of Γ_hom, a careful measurement is required of the hole width as a function of burning-fluence density Pt/A, and an extrapolation of Γ_hole to Pt/A → 0 must be carried out.

From HB experiments performed in this way, we were able to obtain excitation energy-transfer times from BChl a molecules in the B800 ring to those in the B850 ring at low temperature. In addition, experiments on the red wing of the B850 band yielded a T ^1.3±0.1 temperature dependence of Γ_hom (optical dephasing), similar to organic disordered systems, and an extrapolation value of Γ_hom for T → 0 that is consistent with a fluorescence lifetime of the excited state of a few nanoseconds. These results proved that no scattering processes, but only decay from the excited state takes place in the red wing of B850 at T → 0.

By measuring hole widths as a function of delay time between burning and probing, we are able to obtain an insight into spectral diffusion processes in photosynthetic complexes, i.e. into irreversible low-frequency fluctuations of the protein. We found that a decrease of the amount of spectral diffusion is correlated with an increase of the size of the complex for the systems studied: the B777 monomer subunit of bacterial LH1, and the CP47, the RC and the CP47–RC complexes of PSII of higher plants.

Furthermore, we have demonstrated that not only the hole widths but also the hole depths reveal quantitative information that is otherwise hidden within a broad absorption band. On the one hand, ‘traps’ for energy transfer in the isolated PSII RC, CP47 and CP47-RC complexes of higher plants could be disentangled. On the other hand, the lowest k = 0 exciton distributions buried within the B850 band of purple bacteria were made visible.

Finally, it is worth mentioning that spectral hole burning is not only a powerful technique for studying photosynthetic complexes but its value has been demonstrated for other biological systems, such as green, yellow and red fluorescent proteins (GFPs and DsRed), also studied in our group (Bonsma et al. 2005; Creemers et al. 1999b, 2000). In these autofluorescent proteins, HB spectroscopy was used to obtain a ‘fingerprint’ of the species under study. For example, photo-convertible forms and their 0–0 transitions were identified and pathways of photo-conversion and energy transfer were determined.

Owing to the wavelength selectivity of HB, when using very narrow-band lasers, questions on the intricate electronic structure of proteins can be answered that cannot be solved with ultrafast (femtosecond) techniques, because of the inherently large optical bandwidths of short laser pulses. These two techniques are thus complementary for the study of complex biological systems.