J. Phys. Chem. 1996, 100, 9537-9544
9537
Annihilation Processes in the Isolated D1-D2-cyt-b559 Reaction Center Complex of
Photosystem II. An Intensity-Dependence Study of Femtosecond Transient Absorption†,‡
M. G. Mu1 ller, M. Hucke, M. Reus, and A. R. Holzwarth*
Max-Planck-Institut für Strahlenchemie, Stiftstr. 34-36; D-45470 Mülheim a.d. Ruhr, Germany
ReceiVed: December 11, 1995; In Final Form: February 29, 1996X
The excitation intensity dependence of the kinetics of the primary processes and of the yield of radical pair
formation in the isolated D1-D2-cyt-b559 reaction center of photosystem II has been studied by femtosecond
transient absorption spectroscopy. It is shown that the kinetics is strongly dependent on the excitation intensity.
The radical pair yield as a function of excitation intensity is compared with the theoretical annihilation curve
and a good agreement between theory and experiment is observed, indicating that the intensity effects on the
kinetics and radical pair yield arise primarily from annihilation processes. Sufficiently annihilation-free
measurements require excitation intensities that give rise to e0.06 absorbed photons/RC while maintaining
a high signal/noise ratio of g100:1 at most detection wavelengths in order to resolve the complex kinetics.
It is shown that such low excitation intensities give rise to absorption changes that are at the edge of the
capabilities of present femtosecond absorption equipment. We also compare the excitation conditions that
have been used so far by other research groups for published transient absorption data on the isolated D1D2-cyt-b559 complex. This comparison shows that essentially all published data have been obtained under
conditions where annihilation or quenching effects are expected to significantly distort the kinetics and timeresolved spectra and to reduce the radical pair yield.
Introduction
Photosystem (PS) II is the antenna/reaction center complex
in higher plants and green algae that performs the photoinduced
splitting of water into molecular oxygen and protons and which
is thus of prime importance for the maintenance of the
biosphere.1-3 The so-called D1-D2-cyt-b559 complex was
isolated for the first time in 1987 from PS II particles and has
been shown to be the unit active in primary charge separation,4,5
and it is therefore called the reaction center (RC) complex of
PS II. Since its isolation, there is vivid interest in understanding
the mechanism and the kinetics of the primary processes in this
complex. Quite a variety of spectroscopic techniques have been
employed to elucidate these processes.6 Among those that
potentially can provide direct information on the kinetics and
the mechanisms of the charge separation and other fast
processes, femtosecond transient absorption spectroscopy ranks
high on the list. For this reason a number of femtosecond
studies has been performed on the D1-D2 complex as a function
of temperature, excitation intensity, redox state, etc. (for a review
see e.g. ref 6). Unfortunately the outcome of these experiments
carried out by various groups has led to highly controversial
and fundamentally differing interpretations. Thus the group of
Wasielewski and co-workers assigned a 3 ps component to the
primary charge separation process,7,8 while the group of Klug
and co-workers criticized that assignment and instead attributed
a ∼21 ps component to that process.9-11 Both the groups of
Holzwarth et al.12-16 and van Gorkom et al.17 favored the 3 ps
interpretation for the primary charge separation based on
fluorescence and transient absorption data, respectively. Up to
now no agreement has been reached in this matter. Furthermore
there was an additional disagreement on the lifetimes and
* Author to whom all correspondence should be addressed.
† This work has been presented in part at the ESF-Workshop on
“Structure and Function of the isolated D1-D2 reaction center”, Wye
College, England, April, 1995.
‡ Abbreviations: Chl, chlorophyll; Pheo, pheophytin; RC, reaction center;
D1, D2, polypeptides of the reaction center of photosystem II.
X Abstract published in AdVance ACS Abstracts, May 1, 1996.
S0022-3654(95)03715-4 CCC: $12.00
spectral range of slow energy transfer components present in
the D1-D2 complex between the groups of Klug and coworkers9,18,19 and Holzwarth and co-workers12-16,20 which still
persists in part. A further topic of controversy was the effect
of using parallel vs magic-angle detection conditions.12,21
Despite these controversies a common conclusion prevailing
from all of the studies was that the primary kinetics is highly
complex, and unravelling it represents a challenging task. As
it stands up to now no valid kinetic scheme has been presented
that would be suitable to explain satisfactorily the observed
transient absorption and fluorescence kinetics and would at the
same time provide reasonable species-associated (fluorescence
and/or transient absorption) spectra of the intermediates. This
ongoing controversy is to a large part caused by the fact that
(a) the absorption and absorption difference spectra of all the
involved states are highly congested in the Qy region of the
spectrum and (b) that no detailed kinetic models have been tested
on the data. In addition, intensity-dependent effects, underestimated largely so far, might also be causing at least part of the
discrepancies between different groups. This is the subject of
the present study.
If one sets out to unravel the ultrafast processes in the D1D2 complex it is important to determine exactly the upper limit
of excitation pulse intensity that would allow one to observe
the kinetics in the linear range, sufficiently undistorted from
annihilation and other nonlinear effects that might influence the
kinetics. Surprisingly such a study has not been performed in
detail to our knowledge. Despite some claims in the literature
from most groups that they actually performed their transient
absorption measurements under annihilation free conditions, in
fact rough estimates of the excitation intensities used already
suggest that excitation intensities applied may have been quite
high in many cases. Thus in some published data even the signs
of transient absorption signals differed, let alone the shape of
the kinetics.7,9 Furthermore most groups reported prominent
transient absorption components with lifetimes comparable to
the pulse widths used in their measurements.7,10,19,22 Such
© 1996 American Chemical Society
9538 J. Phys. Chem., Vol. 100, No. 22, 1996
Figure 1. Absorption spectrum of D1-D2 complex at room temperature
(full line) and spectrum of 680 nm femtosecond excitation pulse (dashed
line).
ultrafast components could to a significant part be caused by
annihilation effects since the majority of fast annihilation is
expected to occur in the RC core on a time scale comparable to
energy transfer.19 In view of this situation the aim of our present
study is severalfold: First, to measure the kinetics of D1-D2
particles as a function of the excitation intensity over a large
range. Second, to measure the annihilation curve for radical
pair formation and to compare it with exact theoretical calculations. Third, to present a detailed comparison of the kinetics
with published data from other groups and to compare the
excitation conditions used by various groups. As a main result
of this study we conclude that femtosecond transient absorption
experiments published so far have not been performed under
sufficiently annihilation-free conditions and that in some cases
actually quite high pulse intensities or alternatively high average
powers have been employed, giving rise to severe annihilation
effects or substantial concentration of quenching intermediates,
respectively, and thus to concomitant modifications in the
kinetics. This has profound consequences on the interpretation
and conclusions drawn from these experiments.
Materials and Methods
The D1-D2 cyt-b559 RCs used in this work have been
isolated according to van Leeuwen et al.23 with slight modifications.24 All measurements reported here have been carried out
on fractions of a sample pool which had a Chl/2 Pheo ratio of
6.1 ( 0.2 as determined by HPLC.24 The room temperature
spectrum of the RCs are shown in Figure 1 along with the
spectrum of the femtosecond excitation pulse peaking at 680
nm. For some measurements samples with a slightly higher
Chl/2 Pheo ratio, i.e. 6.3 ( 0.2, have been used but the results
did not differ significantly with respect to the results discussed
here. For measurements, carried out at 4 °C, the sample was
contained in buffer Tris-HCl 50 mM, pH ) 7.2, detergent
dodecyl maltoside 0.1%. The sample was filled into a rotating
cuvette (d ) 1 mm, volume 3 mL) under a nitrogen atmosphere.
The cuvette is air tight so that with the additional use of the
enzymatic oxygen-scrubbing system glucose/glucose oxidase
and catalase25 the sample was kept under anaerobic conditions.
The absorption of the sample at the excitation wavelength (680
nm) was adjusted to about 0.8-0.9/mm (i.e. OD ) 1/mm at
675 nm). A single excitation pulse hits the same sample volume
only every ∼50 s on average.
Femtosecond Absorption and Data Analysis. Excitation
pulses for femtosecond absorption measurements were generated
by chirped pulse amplification of 40 fs pulses from a Tisapphire laser oscillator (Tsunami, Spectra Physics) in a
regenerative Ti-sapphire amplifier and stretcher/compressor
unit (Quantronix model 4810). The output pulses from the
amplifier (FWHM ) 80 fs) were frequency shifted using an
Müller et al.
optical parametric generator (Topas, Light Conversion). The
spectral width of the excitation pulses (∼120 fs FWHM; ca.
180 fs autocorrelation width) was limited to e6 nm spectral
width at a repetition rate of 3 kHz. The pulses were close to
being transform limited. Part of the 800 nm light from the
amplifier was used to generate a white light continuum of
80 fs width (FWHM). The pump and probe pulses were
polarized at magic angle relative to each other in order to
exclude any kinetic depolarization effects. A narrow wavelength
range from the white light continuum, after transmitting the
sample, was selected by a double monochromator (DH10 Vis,
Jobin Yvon, spectral width ∼2 nm) and was detected by a
photomultiplier. Lock-in detection was used for recording the
transient absorption kinetics. Full details of the apparatus will
be published in a separate report. The rms noise of the detection
system was as low as (5 × 10-7 OD units under actual
measurement conditions in the best cases. This allowed data
with a very high signal/noise (S/N) ratio to be obtained at very
low excitation intensities which was actually a prerequisite for
the measurements presented here. The exciting and detecting
light were focused to a spot with a diameter (FWHM) of 0.13
mm.
Determination of the Molar Absorption Coefficient per
RC. The molar absorption coefficient (or absorption cross
section) of a D1-D2 particle at room temperature is a key
parameter for the theoretical calculation of the multiple hit
probability as a function of excitation intensity. Our preliminary
data indicated that this extinction coefficient is much higher
than what was obtained on an estimate based on the absorption
coefficients of Chl a and Pheo a in solution of organic
solvents.26,27 We therefore determined the absorption coefficient
of D1-D2 RCs experimentally in the following way: The
absorption spectrum of a D1-D2 RC solution was measured on
an absolute scale. An aliquot of this sample was taken, and
the pigments were extracted using standard techniques,24 and
the absorption spectrum of the extract was measured. The Chl/
Pheo ratio was determined by HPLC.24 The absorption spectra
of solutions of known concentration of pure Chl a and Pheo a
were measured in the same solvent (Note: It is important to
use exactly the same water content in the solvent as used for
measuring the extract, since in particular the Pheo a spectrum
undergoes shifts and spectral changes in response to changing
water content.) The spectrum of the extract was then composed
as a linear superposition of the pure Chl a and Pheo a spectra
by a fit over most of the Qy band region.24 The extinction
coefficient of the intact D1-D2 particle was then determined
from these data assuming that the integrated absorption coefficient of Chl a and Pheo a in the Qy region remains unchanged
when going from the organic solvent to the protein. In this
way the value of the molar absorption coefficient of the entire
D1-D2 complex (6.1 Chl + 2 Pheo) was determined to be max
) 720.000 ( 10% (675.5 nm). This value is by roughly a factor
of 2 higher than what would be obtained based on adding up
the solution spectra of the chromophores in organic solvents
(even after accounting properly for spectral shifts according to
the spectral decomposition24). A detailed analysis of the
absorption line-shape function is given in ref 24. The reason
for the much higher absorption coefficient is the substantially
smaller (by a factor of about 2) inhomogeneous width of the
spectra of Chl a and Pheo a in the protein as compared to
organic solvent. Thus in the protein the spectra are narrower
and have a higher extinction coefficient.
Data Analysis. Kinetic transient absorption data were
analyzed in a global fashion by combining the data sets recorded
at different detection wavelengths and excitation intensities.28,29
The global analysis included a deconvolution of the kinetics
Annihilation Processes in Photosystem II
a
J. Phys. Chem., Vol. 100, No. 22, 1996 9539
a
b
b
Figure 3. Plot of amplitude ratios of lifetime components for a fourexponential fit including all excitation intensities: (a) for 545 nm
detection and (b) for 680 nm detection. Components are as follows:
A1, 46 ps; A2, 1.1 ps; A3, long-lived (nondecaying); A4, 100 fs.
Figure 2. Intensity dependence of the transient absorption signal at a
detection wavelength of 680 nm (a) and at 545 nm (b), λexc ) 680 nm.
The signals were normalized to the maximum at time zero in order to
enable a better comparison.
with the autocorrelation of the excitation pulse. The quality of
the fits was judged in each case on the basis of χ2 values and
plots of the weighted residuals.29
Results and Discussion
Using nearly transform-limited 120 fs (FWHM) pulses with
a spectral maximum at 680 nm (Figure 1) the intensity
dependence of the transient absorption signals was measured
for detection wavelengths of 545 nm (Pheo Qx band) and 680
nm (i.e. near the P680 absorption maximum) over a time range
of up to 300 ps. The excitation energy/pulse was varied over
nearly 3 orders of magnitude. The kinetic traces are compared
in Figure 2a for 680 nm and in Figure 2b for 545 nm detection
wavelength. As can be seen from these figures the kinetics is
dependent on excitation energy down to about 1.7 nJ/pulse.
Below this intensity the kinetics remains essentially constant.
In addition to complex changes in the kinetics at short delay
times (see below) a strong decrease of the intensity-normalized
maximal bleaching at long delay times (300 ps) is observed
with increasing excitation pulse intensity. The long-lived
bleaching is taken as a direct measure of the amount of longlived radical pair formed. Thus these data point to a drastic
decrease in the (relative) yield for radical pair formation at
increasing excitation intensity (cf. Figure 2).
The data from all the kinetic traces for the two detection
wavelengths and all excitation intensities were subjected to
global analysis. The minimal number of exponential components required to achieve a reasonable fit to the kinetics was
four with lifetimes of ∼100 fs, 1.1 ps, 46 ps, and g5 ns
(nondecaying). Figure 3, parts a and b, shows the plots of the
amplitude ratios for the components of the four-exponential fit
as a function of excitation intensity. We chose to plot the
amplitude ratios, rather than the absolute values, because it is
thus easier to recognize any changes in the kinetics, since for
intensity independent kinetics the amplitude ratios should remain
constant. Since the kinetics depends drastically on the excitation
intensity, the lifetimes from the global analysis are dominated
by the (distorted) values from the high excitation intensities if
all kinetic traces (at all excitation intensities) are included. This
procedure thus leads to large distortions in the amplitudes for
the low excitation intensities, since there the lifetimes are in
fact quite different. For this reason we have performed an
additional global analysis leaving out the decays for high
excitation intensities (only for 680 nm detection, cf. Figure 4).
These new lifetimes for the low intensity range are 250 fs, 48
ps, 2.7 ps, and long-lived (nondecaying). From comparison of
Figures 3 and 4 it can be seen that the amplitude ratios change
substantially even at relatively low excitation intensity and for
some of the components (e.g. the ultrafast 100 fs component)
even the sign changes in that intensity region. It is also clear
that not only the amount of nondecaying component (A3) is
reduced, but the entire kinetics is strongly dependent on the
excitation intensity. This demonstrates that kinetics undistorted
by annihilation and other artifacts can be measured only at very
low excitation intensities, i.e. at energies e1.7 nJ/pulse under
our focusing conditions (corresponding to e0.06 absorbed
photons/RC).
Intensity Effects on the Relative Radical Pair Yield. One
of the most easily recognizable effects of high intensity should
9540 J. Phys. Chem., Vol. 100, No. 22, 1996
Figure 4. Plot of amplitude ratios of lifetime components from a global
analysis leaving out the high intensities. Components are as follows:
A1, 48 ps; A2, 2.7 ps; A3, long-lived (nondecaying); A4, 250 fs.
Figure 5. Plot of the relative yield of formation of radical pair
(absorbance change after 300 ps at 680 nm normalized to the excitation
intensity) as a function of excitation density at 680 nm (x axis 1) and
as a function of the average number of absorbed photons/RC/pulse (x
axis 2). Also shown is the annihilation curve (dashed) as calculated
based on the multiple hit probability (See Appendix and text). The
arrows indicate the excitation intensities used for various measurements: (1) lowest intensity measurement at 545 nm detection (cf. Figure
6b); (2) intensity used for recording a complete low intensity data set
(7 nJ at 680 nm with 130 µm spot diameter, see text); (3) estimated
intensity used in the measurements of ref 11; (4) 70 nJ/pulse as used
for measuring the kinetics shown in Figure 6a; (5) estimated intensity
used in the measurements of ref 22. For excitation wavelengths other
than 680 nm the relevant numbers have been recalculated to the
equivalent value for hypothetical 680 nm excitation taking into account
the absorption spectrum.
be a reduction in radical pair yield which can be measured in
transient absorption at long delay time. Thus in order to put
the observed intensity effects on a more quantitative basis and
to be able to compare it to the theory for annihilation and also
to the data from other groups we have plotted in Figure 5 the
intensity-normalized yield of radical pair formation (measured
from the bleaching at 680 nm for a delay of 300 ps normalized
to the excitation intensity). The experimental data are compared
with the theoretical curve using the formalism and the assumptions given in the Appendix. Essentially the underlying theory
is the same as that given by Paillotin30,31 except that we use
here the exact solutions which apply to any optical density of
the sample and also take into account the details of the excitation
and detection conditions relevant for transient absorption
detection. For the sake of simplicity, and since no better
information was available for most studies reported in the
literature, we assumed a square excitation profile (i.e. one
filament only according to the formulae in the Appendix). It
Müller et al.
follows from the plot in Figure 5 that the theory describes the
experimental situation quite well up to the highest excitation
intensities used. Thus it seems that at least the intensity effect
on the relative radical pair yield can be described by a
straightforward annihilation process.
A 100% probability for annihilation upon multiple excitation
within a single RC particle was used in the theoretical
calculations. In order to be able to solve the relevant equations
(see Appendix) the molar absorption coefficient for the D1-D2
particles was required. Using a first estimate for this value on
the basis of solution spectra for Chl a and Pheo a (maxestimated
≈ 310.000 L mol-1 cm-1) we could not get a reasonable
agreement between theory and experiment. Experimental
determination of the absorption coefficient indeed led to an about
two times higher absorption coefficient (maxexp ≈ 720.000 L
mol-1 cm-1; see Materials and Methods). Using this value then
resulted in a very good agreement with the experimental curve
(see Figure 5). From the absorption change at low intensity
excitation and long delay time we can calculate (see Appendix)
a lower limit of 70.000 L mol-1 cm-1 ((10%) for the difference
absorption coefficient ∆ of the radical pair (at 680 nm), i.e.
the difference (P680-Pheo) - (P680+ - Pheo-).42 Even
considering the substantial absorption increase overlapping
with the bleaching, this number is surprisingly low, taking
into account the fact that at least one Chl (from P680) and one
Pheo should contribute to the bleaching. This number will
presumably increase somewhat once a valid kinetic scheme is
available.
Comparison of Excitation Intensities Used by Various
Groups. In order to be able to judge the conclusions drawn
by various groups from their transient absorption experiments
it is important to know the exact excitation conditions and the
resulting annihilation probabilities. We have thus compiled in
Table 1 the excitation and annihilation conditions at which the
transient absorption experiments published in the literature on
D1-D2 have been performed. Where the authors have provided
all the necessary information to calculate the impinging photon
density per pulse and unit area the values given in the respective
references have been used as the basis for our calculations
(performed according to the formalism given in the Appendix).
In those cases where the exact excitation conditions have not
been provided by the authors, we used their published experimental absorption changes at 680 nm and long delay times
(radical pair formation) as the basis for the calculation. In that
case normalization has been performed with our measured
kinetics and absorbance change at low excitation intensity (see
Figure 2).
The group of Wasielewski and co-workers in their first
measurements7 used excitation in the nonselective region around
610 nm. These experiments have been performed under
extremely high excitation conditions leading to annihilation (on
average) in more than 88% of the RCs (relative radical pair
yield only 12%!). It should be noted that the initial suggestion
of a ∼3 ps charge separation kinetics has been made on the
basis of these experiments. By taking the annihilation data in
Table 1 into account, it is clear that in case that this interpretation
(i.e. ∼3 ps charge separation) should finally turn out to be
correct indeed, the initial conclusions would have to be
considered to be highly fortuitous. The Wasielewski group has
recently reported new measurements (mainly at 545 nm detection) at substantially lower excitation intensity.32 Nevertheless
in their new data set the annihilation probability is still ∼22%.
The most extensive (in terms of both spectra and kinetics)
transient absorption data sets on D1-D2 complexes have been
reported by the group of Klug and co-workers.9-11,18,43 Our
estimate of their excitation conditions for measurements at magic
Annihilation Processes in Photosystem II
J. Phys. Chem., Vol. 100, No. 22, 1996 9541
TABLE 1: Comparison of Excitation and Annihilation Conditions Used by Various Groups When Measuring Ultrafast
Absorbance Changes on the D1-D2-cyt-b559 Reaction Center Complex
research group
ref
Φrel
annihilation
(%)
average absorbed
photons/RC/pulse
approximate S/N
ratio (680 nm)
8.7
0.53
0.4
∼5
∼50-60
∼20
0.023
∼20
11
0.23
g200
0.97
3
0.06
∼50
this work
0.994
0.6
0.012
∼40
22
0.61
1.15
- (only spectra)
Wasielewski et al.
Wasielewski et al.
Klug et al.
7
32
11
0.12
0.78
0.82
van Gorkom et al.
17
0.988
Holzwarth et al.
13 and
this work
0.89
Holzwarth et al.
this work
Holzwarth et al.
van Grondelle et al.
88
22
18
1.2
39
comment
very high intensity
data for lowest intensity (60 nJ) used
estimated based on absorbance change at
680 nm for radical pair
valid for 680 nm excitation; ≈0.9 photons/RC
total over 40 pulses due to high repetition
rate (average pump intensity 400 W/cm2);
Qy detection range; from all the RCs
in the irradiated/detected volume under
these conditions 10% are in the radical
pair state and 20% are in the triplet state
measurement series using 7 nJ/pulse
at 680 nm (average pump intensity
0.18 W/cm2)
measurements at 1.7 nJ/pulse (545 nm
detection); average pump intensity
0.04 W/cm2
measurements at 0.34 nJ/pulse, (680 nm
detection); average pump intensity
0.008 W/cm2
77 K measurement for 680 nm excitation;
based on absorption change at 680 nm for
radical pair; the favorable assumption that
∆ at 77 K is about 30% higher than at
room temperature was made
a,b The absorption coefficient for the RC particle at 675 nm (room temperature) has been determined to be 720 000 ( 10% (see text). The lower
limit for the ∆ difference absorption coefficient (P680Pheo-P680+ Pheo-) at 680 nm has been estimated to be 70 000 ( 15% at 680 nm (room
temperature). Φrel is the relative yield of radical pairs calculated as number of radical pairs/number of photons absorbed. b The data presented here
have been calculated based on the data for excitation energy, wavelength, absorbance, and beam diameter given in the indicated references. Where
these parameters were not given, the calculation has been made on the observed absorption change at 680 nm for the radical pair (long delay time).
Good agreement between calculated and measured absorption changes was obtained in all cases where a comparison could be made.
angle11 results in a probability of ∼18% annihilation (relative
radical pair yield 0.82), i.e. similar to the excitation conditions
used for the new data from the Wasielewski group. Clearly
the intensities used by either the Wasielewski or the Klug group
are still too high for a reliable answer to the problem of the
exact charge separation kinetics (see detailed discussion below).
In this connection one may in particular also question the
interpretation of the ultrafast (100 fs) component which has been
assigned by the Klug group to equilibration among the strongly
coupled RC core excited states. Undoubtedly a substantial
contribution to this ultrafast component will arise due to
annihilation under the excitation conditions used. Furthermore
the exact kinetics for the equilibration process might be
significantly slower if undistorted by annihilation. This is shown
by our data at lower intensity. While higher excitation
intensities in our intensity-dependent measurements led to an
ultrafast component of about 100 fs (i.e. approximately the
excitation pulse width), measurements at lower intensity gave
a longer-lived ultrafast component of about 250-350 fs. Using
picosecond excitation pulses of very high repetition rate, the
group of van Gorkom and co-workers has reported a series of
transient absorption data at very low excitation intensity per
pulse.17 Our calculations indeed show that their measurements
were essentially annihilation-free (∼1.2% annihilation) if judged
on a per pulse basis. Thus these measurements are the only
ones performed so far under sufficiently low excitation intensity
to exclude singlet-singlet annihilation. Nevertheless the measurements from the van Gorkom group suffer from a severe
problem: Due to the high repetition rate of 80 MHz every RC
particle absorbed on average 0.9 photons since it “sees” about
40 excitation pulses on average before the sample could be
exchanged. Consequently the multiple excitation probability
is very high (according to multiple hit theory 12% double hits,
∼4% triple hits, i.e. altogether g36% multiple hits). These
multiple excitations, due to the fact that they occur from
consecutive pulses with a pulse-pulse distance of multiples
of ∼12 ns, will occur mostly in RCs that have already undergone charge separation and are thus either in the radical pair
state or in the triplet state. The very efficient quenching effect
of the radical pair on the newly created excited state is
unpredictable in detail but is expected to lead to severe distortion
in the observed kinetics.33-36 Similarly, singlet-triplet annihilation could be quite substantial as well (cf. Table 1). Thus
despite the low excitation intensity per pulse any conclusions
drawn from these measurements should be considered with
caution.
We have recently measured an extensive multi-wavelength
femtosecond data set.37,44 The aim was to avoid as much as
possible annihilation but nevertheless keep a high S/N ratio that
would also allow to measure the 545 nm range under the same
excitation conditions with a sufficiently high S/N ratio. The
series was performed using 7 nJ/pulse in the excitation, focused
to a spot of 130 µm diameter. Under these low intensity
conditions (this corresponds to about two times lower effective
photon density than used by the Klug group, corrected for the
different excitation wavelengths) nevertheless the probability
for annihilation events is still ∼11% (see Table 1). For some
selected detection wavelengths we have also measured the
kinetics at excitation intensities that were 20 times lower than
that (leading to negligible annihilation).
Low-temperature transient absorption measurements have
been performed recently by the group of van Grondelle and
co-workers.22 On the basis of the observed transient absorption
changes of the radical pair at long delays and by making the
favorable assumption of assuming a some 30% higher difference
absorption coefficient than at room temperature, our calculation
yields a relative radical pair yield of 0.61, i.e. about 40% annihilation probability, at their excitation conditions. This must be
9542 J. Phys. Chem., Vol. 100, No. 22, 1996
considered very high annihilation and is expected to lead to
severe interference of annihilation and primary charge separation
kinetics.
In Figure 5 we have indicated by arrows on the annihilation
curve the intensities at which transient absorption experiments
on the D1-D2 complex have been performed so far by different
groups. Arrow 3 indicates the excitation intensity at which most
of the published measurements of the Klug group have been
performed. This intensity is by a factor of ∼2 higher than the
one used in the majority of our own measurements37 (arrow 2)
and is in good agreement with the photon density and excitation
probability that can be obtained from the parameters given in
ref 11. However a direct comparison of our intensity-dependent
transient absorption kinetics with the results published in refs
9 and 10 suggests, that for measuring the 545 nm range (Pheo
Qx band) probably a g5 times higher excitation intensity has
been used by these authors (indicated by arrow 4 in Figure 5)
as compared for the measurements in the Qy range.45 This
intensity is quite high in the annihilation region. Given the fact
that a decisive part of the kinetic assignment of the Klug group
is in fact based on this kinetics around 545 nm, one may question
the validity of their assignment. In order to directly compare
the signals we have plotted in Figure 6 our own measured
kinetics at 545 nm together with that of the Klug group10 at a
comparable excitation intensity (Figure 6a) and at a ∼40 times
lower intensity (Figure 6b). Note in particular the substantial
change in the zero-crossing time with excitation intensity and
the differences in the ratio of absorption (at zero time) to
bleaching amplitudes (at long time). Without any detailed
analysis it is clear from this comparison that the published
kinetics9,10 is distorted by nonlinear effects and that the radical
pair yield (as judged by the signal at long times) is substantially
decreased.
Reasons for Excitation Intensity Effects on the Kinetics.
The results presented above indicate that both the kinetics and
the yield of radical pair are strongly dependent on the excitation
intensity. The D1-D2 complex contains at least eight chromophores. Thus we expect at first glance the usual annihilation
behavior of multiple chromophore assemblies, as has been
studied intensively on various photosynthetic antenna systems
(see e.g. refs 38 and 39). However, the situation is more
complex in this case. First it is now well known that energy
transfer among chromophores in the D1-D2 complex occurs on
largely different time scales.12,13,19 Thus equilibration within
the RC core pigments occurs most likely on a time scale of a
few hundred femtoseconds.12,19 This should then also correspond approximately to the time scale of the major annihilation
component since most of the absorption occurs initially by the
chromophores in the core. Our data show that at high excitation
intensity the lifetime of the fastest component drops from about
250 fs (at low excitation intensity) to about 100 fs at moderate
and high excitation intensity. That is not the only effect,
however, since the energy transfer of the external Chls with
the core occurs on a 10-100 times slower time scale (∼6-30
ps).12,13 Thus also the annihilation kinetics should have secondary component(s) on a comparable picosecond time scale. This
can explain the changes in the values and amplitudes of the
intermediate lifetimes upon changing the excitation intensity.
There is another effect however. This will come into play when
a particle is hit by two photons and one of them gives rise to
charge separation (on a ∼3 ps time scale as deduced from our
time-resolved fluorescence experiments12,13). Thus after a few
picseconds a Chl cation and most probably a Pheo anion are
formed. Both of these species are known to be efficient quenchers of excited states.33-36 Thus we expect for particles with
multiple excitation another component in the excited-state decay
Müller et al.
Figure 6. Comparison of the transient absorption traces at 545 nm
for measurement carried out at (a) 70 nJ/pulse (dashed line) and (b)
1.7 nJ/pulse. In each case the absorption trace at 545 nm (full line)
from ref 11 is overlaid, normalized to the absorption change of traces
a and b at time zero. Note the difference in absorbance scales and
also the ratio of absorbance maximum at time zero and at 55 ps. The
measurement for b is averaged for a longer time in order to get a
comparable S/N ratio as in a. The numbers in b represent the relative
amount of radical pair formation, as deduced from the bleaching signal
at about 55 ps delay.
due to the quenching by radical species. No exact data are
available on the corresponding quenching rates, which will depend anyway strongly on the unknown distance of the different
chromophores in the complex. Nevertheless the quenching is
expected to be very efficient. Possible quenching mechanisms
in that case are either Förster energy transfer and/or electron
transfer. Finally the radical pair evolves into the triplet with
high yield. Under such measuring conditions thus a relatively
high steady state concentration of triplet may arise that gives
rise to singlet-triplet quenching which may also be very
efficient. These two latter effects may have caused substantial
distortion of the kinetics reported by the van Gorkom group.17
Which Level of Annihilation Can Be Accepted? In conclusion, multiple excitation of an RC is expected to lead to very
complex and not easily predictable changes in the overall
kinetics and to a reduction in radical pair yield. It is clear from
our data set presented here that even at low intensities the
kinetics of primary processes in the D1-D2 RC is extremely
complex (more components than have been deduced so far from
transient absorption experiments are required for the full kinetic
description37). If any meaningful analysis and interpretation of
this complex kinetics is to be performed, we must be absolutely
sure that we can exclude any nonlinear effects of the type(s)
discussed above. Thus transient absorption kinetics have to be
measured under conditions where the contribution of nonlinear
effects is substantially less than the smallest amplitude component. It is important to note that most of the three or four
relevant kinetic components indicative of energy transfer and/
or charge separation have small amplitude, generally e10% of
the long-lived radical pair signal (for example at 680 nm and
similarly at most other wavelengths).9,10,13 In this case our
estimates indicate that a S/N ratio of better than 100:1 will be
minimally required to unravel the five or even six exponential
kinetics. In essence this amounts to the requirement that the
Annihilation Processes in Photosystem II
probability of multiple excitation per particle and pulse must
be reduced to a maximum of about 2%, which is still by about
a factor of 4 lower than what has been used in our own extended
low intensity measurement series (i.e. 1.8 × 1014 photons cm-2
pulse-1 corresponding to 7 nJ/pulse at 680 nm with a spot
diameter of 130 µm). This figure is a tremendous challenge to
the experiment since it is necessary to measure transient
absorption changes in the range of 10-5 to a maximum of 10-3
OD units with a S/N ratio of g100, which for the lowest signals
would require a noise level equivalent to e10-7 ∆OD units.
Such low noise levels in a femtosecond transient absorption
experiment are outside the reach of present transient absorption
setups but may become possible with new developments on the
experimental side. The apparatus used in our measurements
allows under optimal conditions a noise level of e5 × 10-7
∆OD units, as is demonstrated in Figure 6. This requires
however rather long integration times on the order of 10 s per
data point. Such long integration times are in partial conflict
with the further necessity to measure the transient absorption
changes at a large number of different detection wavelengths
and the further requirement to keep the sample intact, i.e. in a
photochemically undamaged state.40,41 Thus some compromise
between sample stability, total sample volume required, S/N
ratio, and acceptable annihilation effect has to be made.
Conclusions
In summary we conclude that recording sufficiently annihilation-free transient absorption signals on the D1-D2 complex is
at present at the very edge of the technical feasibility even for
the most sophisticated instruments. No such measurements have
been performed so far. Even the best published measurements
were obtained under excitation conditions that yield about 18%
annihilation, which is clearly too high in the light of the above
discussion, taking into account the high complexity of the
kinetics. Any detailed conclusions drawn from the published
transient absorption data must therefore at present be considered
with great care. In contrast fluorescence kinetic measurements,
in particular by single-photon-counting methods, are normally
performed under substantially lower excitation conditions (typically e1 × 1013 photons cm-2 pulse-1) where annihilation can
be excluded. If the repetition rate of the excitation is suitably
reduced and if the sample is pumped through a flow cuvette all
other possible kinetic artifacts can be excluded as well. In
fluorescence measurements by single photon counting the main
problem is to achieve a sufficiently high time resolution. It
has been shown, however, that the time resolution that can be
obtained in our single-photon-counting instrument is just
sufficient to perform these measurements and resolve the ∼3
ps component.13 We conclude that the interpretation based on
fluorescence kinetic experiments must at present be considered
to be more reliable than those from most published ultrafast
transient absorption measurements.
J. Phys. Chem., Vol. 100, No. 22, 1996 9543
ability. Near the front surface of the cuvette the annihilation
probability is very high and it decreases to low values at the
backside of the cuvette. Furthermore the excitation and detection may be inhomogeneous across the laser beam depending
on the beam profiles. This means that the region with high
annihilation (e.g. in the center part of the excitation spot) will
contribute most to the ∆A signal due to the fact that most of
the light is absorbed in that region. Taking average values for
the various parameters in this situation leads to substantial errors
in the estimation of the annihilation and detection probability
and in the average number of absorbed photons. We have thus
numerically integrated the equations for excitation and multiple
hit probability in the excitation beam across the cuvette and
the beam profile using a finite element approach. For simplicity
we assume a constant beam waist across the cuvette. Thus the
beam profile has been decomposed into nF concentric rings
(filaments) and the cuvette has been splitted into nS layers where
nF has been chosen to be in the order of 20 and nS about 10.
We have taken the maximal number of absorbed photons/RC
to be m ) 20. Likewise we have integrated the probability for
detection of the annihilation (caused by the pump beam) by
the probe beam across the beam profile. It is thus possible to
also take into account different spot sizes of the pump and probe
beams, as often used in experiments. Necessary input parameters are the total pulse energy Ep, the excitation wavelength
λexc and the OD of the sample at the excitation wavelength,
and the intensity profile or halfwidth HW of the excitation beam.
Furthermore the particle concentration c is required which can
be obtained from the absorption spectrum and the molar absorption coefficient per RC particle (see Materials and Methods).
The relevant parameters are given by the following relationships:
Pj )
G(j,HW)*Aj
nF
ptot with ptot )
∑G(j,HW)*Aj
Calculation of Multiple Excitation Probability. The general
theory for annihilation has been developed by Paillotin and
others for arrays of chromophores undergoing efficient energy
transfer.30,31 For a proper description of the annihilation
conditions we have to take into account the details of the specific
experiment which are not considered in Paillotin’s equations,
however. In a transient absorption experiment two beams, the
excitation and the detection beam, are used. In order to increase
the signal, the experiment is usually carried out at high optical
density (usually near OD ) 1) at the excitation wavelength.
This has the effect that the excitation probability across the
cuvette is extremely unequal, and so is the annihilation prob-
hcL
j)1
(in)
pi,j
) Pj10-(OD/nS)(i-1)
(out)
pi,j
) Pj10-(OD/nS)(i)
tj ) cNA(d/nS)10-3Aj
ui,j )
Φrel )
(in)
(out)
pi,j
- pi,j
tj
total number of radical pairs
)
total number of absorbed photons
[
]
(ui,j)k
tj
e-ui,j
∑
∑
∑
k!
i)1 j)1 k)1
nL nF
nL nF
Appendix
Epλexc
m
m
[
(ui,j)k
tj
e-u
∑
∑
∑
i)1 j)1 k)1 (k - 1)!
i,j
]
(1)
G(j,HW) is the intensity in filament j of the beam profile with
halfwidth HW (full-width-half-maximum). This allows one to
take into account nonuniform excitation profiles. The parameter
Aj is the front surface of filament j, ptot is the total number of
photons/pulse impinging on the surface of the cuvette as
calculated from the total pulse energy and the excitation
wavelength λexc, cL the speed of light, OD the optical density
(out)
are the number of
at the excitation wavelength, p(in)
i,j and pi,j
photons entering and leaving segments i, j where i stands for
the layer and j for the filament and NA the Avogadro number.
The function tj gives the total number of RC particles in a section
9544 J. Phys. Chem., Vol. 100, No. 22, 1996
Müller et al.
of filament j. The function ui,j then gives the probability of
photon absorption per RC particle in section i, j. Using the
relationships for multiple hit probability the number of radical
pairs created and the total number of absorbed photons can be
calculated which then allows one to calculate the normalized
yield of radical pair formation Q.
The absorbance change ∆OD observed at a particular
wavelength is given by the relationship:
[
∆OD ) -log
nF
]
G(j,HW)10∆(ca )dAj
∑
j)1
j
nF
G(j,HW)Aj
∑
j)1
(2)
where
caj )
[
]
(ui,j)k
∑
∑tj k! e-ui,j
i)1 k)1
nL
m
NAd10-3Aj
where caj is the average concentration of radical pairs in the
filament j and ∆ is the difference absorption coefficient and d
the length of the cuvette. The average number of absorbed
photons/RC nA is given by
nS
nA
h)
nF
∑
∑ui,j[pi,j(in) - pi,j(out)]
i)1 j)1
nS
nF
(in)
(out)
[pi,j
- pi,j
]
∑
∑
i)1 j)1
We have used a 100% probability of annihilation in the case of
multiple hits. Comparison with the experimental data shows
that this is reasonable for the D1-D2 complex. Lower probabilities can be taken into account in a straightforward manner
if desired, however.
Acknowledgment. We thank Mrs. Iris Martin for developing
the data analysis software used for the transient absorption
data and Mrs, R. Kesslau and Mr. U. Pieper for able technical assistance. Partial financial support by the Deutsche
Forschungsgemeinschaft (Sonderforschungsbereich 189, Heinrich-Heine-Universität Düsseldorf and Max-Planck-Institut
für Strahlenchemie, Mülheim a.d. Ruhr) is gratefully acknowledged. We thank Prof. K. Schaffner for supporting these
investigations.
References and Notes
(1) Renger, G. Angew. Chem. 1987, 99, 660.
(2) Sauer, K. Annu. ReV. Phys. Chem. 1979, 30, 155.
(3) Renger, G. Photosynth. Res. 1993, 38, 229.
(4) Danielius, R. V.; Satoh, K.; van Kan, P. J. M.; Plijter, J. J.; Nuijs,
A. M.; van Gorkom, H. J. FEBS Lett. 1987, 213, 241.
(5) Nanba, O.; Satoh, K. Proc. Natl. Acad. Sci. U.S.A. 1987, 84,
109.
(6) Seibert, M. In The Photosynthetic Reaction Center; Anonymous,
Ed.; Academic Press: New York, 1993; p 319.
(7) Wasielewski, M. R.; Johnson, D. G.; Seibert, M.; Govindjee. Proc.
Natl. Acad. Sci. U.S.A. 1989, 86, 524.
(8) Seibert, M.; Toon, S.; Govindjee; O’Neil, M. P.; Wasielewski, M.
R. In Research in Photosynthesis. II; Murata, N., Ed.; Kluwer Academic
Publishers: Dordrecht, 1992; p 41.
(9) Hastings, G.; Durrant, J. R.; Barber, J.; Porter, G.; Klug, D. R.
Biochemistry 1992, 31, 7638.
(10) Durrant, J. R.; Hastings, G.; Joseph, D. M.; Barber, J.; Porter, G.;
Klug, D. R. Biochemistry 1993, 32, 8259.
(11) Klug, D. R.; Rech, T.; Joseph, D. M.; Barber, J.; Durrant, J. R.;
Porter, G. Chem. Phys. 1995, 194, 433.
(12) Holzwarth, A. R.; Müller, M. G.; Gatzen, G.; Hucke, M.; Griebenow, K. J. Lumin. 1994, 60/61, 497.
(13) Gatzen, G.; Müller, M. G.; Griebenow, K.; Holzwarth, A. R. J.
Phys. Chem. 1996, 100, 7269.
(14) Gatzen, G.; Griebenow, K.; Müller, M. G.; Holzwarth, A. R. In
Research in Photosynthesis. II; Murata, N., Ed.; Kluwer Academic
Publishers: Dordrecht, 1992; p 69.
(15) Roelofs, T. A.; Kwa, S. L. S.; van Grondelle, R.; Dekker, J. P.;
Holzwarth, A. R. Biochim. Biophys. Acta 1993, 1143, 147.
(16) Roelofs, T. A.; Gilbert, M.; Shuvalov, V. A.; Holzwarth, A. R.
Biochim. Biophys. Acta 1991, 1060, 237.
(17) Schelvis, J. P. M.; van Noort, P. I.; Aartsma, T. J.; van Gorkom,
H. J. Biochim. Biophys. Acta 1994, 1184, 242.
(18) Rech, T.; Durrant, J. R.; Joseph, D. M.; Barber, J.; Porter, G.; Klug,
D. R. Biochemistry 1994, 33, 14768.
(19) Durrant, J. R.; Hastings, G.; Hong, Q.; Barber, J.; Porter, G.; Klug,
D. R. Chem. Phys. Lett. 1992, 188, 54.
(20) Holzwarth, A. R.; Roelofs, T. A. J. Photochem. Photobiol. B 1992,
15, 45.
(21) Wiederrecht, G. P.; Seibert, M.; Govindjee; Wasielewski, M. R.
Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 8999.
(22) Visser, H. M.; Groot, M.-L.; van Mourik, F.; van Stokkum, I. H.
M.; Dekker, J. P.; van Grondelle, R. J. Phys. Chem. 1995, 99, 15304.
(23) van Leeuwen, P. J.; Nieveen, M. C.; van de Meent, E. J.; Dekker,
J. P.; van Gorkom, H. J. Photosynth. Res. 1991, 28, 149.
(24) Konermann, L.; Holzwarth, A. R. Biochemistry 1996, 35, 829.
(25) McTavish, H.; Picorel, R.; Seibert, M. Plant Physiol. 1989, 89,
452.
(26) Porra, R. J.; Thompson, W. A.; Kriedemann, P. E. Biochim. Biophys.
Acta 1989, 975, 384.
(27) Lichtenthaler, H. K. Methods Enzymol. 1987, 148, 350.
(28) Holzwarth, A. R.; Schatz, G. H.; Brock, H.; Bittersmann, E.
Biophys. J. 1993, 64, 1813.
(29) Holzwarth, A. R. In Biophysical Techniques. AdVances in Photosynthesis Research; Amesz, J., Hoff, A., Eds.; in press.
(30) Paillotin, G.; Swenberg, C. E. Dynamics of excitons created by a
single picosecond pulse; Ciba Foundation Symposium 61; Excerpta
Medica: Amsterdam, 1979.
(31) Paillotin, G.; Swenberg, C. E.; Breton, J.; Geacintov, N. E. Biophys.
J. 1979, 25, 513.
(32) Greenfield, S. R.; Wasielewski, M. R.; Govindjee; Seibert, M. In
Photosynthesis: from Light to Biosphere; Mathis, P., Ed.; Kluwer Academic
Publishers: Dordrecht, 1995; p 663.
(33) Dobek, A.; Deprez, J.; Geacintov, N. E.; Paillotin, G.; Breton, J.
Biochim. Biophys. Acta 1985, 806, 81.
(34) Geacintov, N. E.; Breton, J. Energy transfer and fluorescence
mechanisms in photosynthetic membranes. 5; Chemical Rubber Company: Boca Raton, 1987.
(35) France, L.; Geacintov, N. E.; Lin, S.; Wittmershaus, B. P.; Knox,
R. S.; Breton, J. Photochem. Photobiol. 1988, 48, 333.
(36) France, L. L.; Geacintov, N. E.; Breton, J.; Valkunas, L. Biochim.
Biophys. Acta 1992, 1101, 105.
(37) Müller, M. G.; Hucke, M.; Reus, M.; Holzwarth, A. R. J. Phys.
Chem. 1996, 100, 9527.
(38) van Grondelle, R.; Amesz, J. In Light Emission by Plants and
Bacteria; Govindjee, Amesz, J., Fork, D. C., Eds.; Academic Press: New
York, 1986; p 191.
(39) van Grondelle, R. Biochim. Biophys. Acta 1985, 811, 147.
(40) Chapman, D. J.; Gounaris, K.; Barber, J. Photosynthetica 1989,
23, 411.
(41) Seibert, M.; Picorel, R.; Rubin, A. B.; Connolly, J. S. Plant Physiol.
1988, 87, 303.
(42) A lower limit only is obtained for the difference absorption
coefficient ∆ because the radical pair even at 300 ps is not populated to
100% due to efficient back reaction to the excited state. A more exact
value can be obtained after proper kinetic modeling.
(43) In the calculations and discussion of the data from the group of
Klug et al. we refer to their most recent data set measured under magic
angle conditions.11
(44) A preliminary account of these data has been given at the ESF
Workshop on “Structure and Function of the D1-D2 reaction center
complex”, Wye College, England, April 1995.
(45) We should like to note that an increase in excitation intensity is
not mentioned in the work of Hastings et al.9 and that the authors claim
that the intensity has not been increased as compared to the measurements
in the Qy range (J. Durrant, private communication).
JP953715A