The role of mixed vibronic Qy-Qx bands in the light-harvesting dynamics of the major antenna complex, LHCII

The importance of green light for driving natural photosynthesis has long been underappreciated, however, under the presence of strong illumination, green light actually drives photosynthesis more efficiently than red light. This green light is absorbed by mixed vibronic Q y -Q x states, arising from chlorophyll (Chl)-Chl interactions, although almost nothing is known about these states. Here, we employ polarization-dependent two-dimensional electronic-vibrational spectroscopy to study the origin and dynamics of the mixed vibronic Q y -Q x states of light-harvesting complex II. We show the states in this region dominantly arise from Chl b and demonstrate how it is possible to distinguish between the degree of vibronic Q y versus Q x character. We find that the dynamics for states of predominately Chl b Q y versus Chl b Q x character are markedly different, as excitation persists for significantly longer in the Q x states and there is an oscillatory component to the Q x dynamics, which will be discussed. Our findings demonstrate the central role of electronic-nuclear mixing in efficient light-harvesting and the different functionalities of Chl a and Chl b.


Introduction
The fact that leaves are green and the majority of spectroscopic studies on optically active pigment-protein complexes (PPCs) are performed on in vitro systems has led to the misunderstanding that green light has little efficacy on photosynthesis. However, on the contrary, it has been shown by Terashima et al. that for in vivo systems, green light in the presence of strong illumination actually has the ability to drive photosynthesis more efficiently than red light. 1 In this work, we explore the states absorbing green light and their dynamics in light-harvesting complex II (LHCII).
Overall, the success of the photosynthetic apparatus begins with the design and function of the PPCs which harvest solar light, the primary step in photosynthesis. 2 In green plants and algae, LHCII serves as the major antenna complex, which transfers excitation energy towards the photosynthetic reaction center. 2 LHCII, in trimeric form as it is generally found, is composed of 24 chlorophyll (Chl) a, 18 Chl b, and 12 carotenoid (Car) pigments. 3 Spatially, these pigments are held within the protein environment in such a way that electrostatic interactions between nearby pigments promote the formation of delocalized excitonic states, leading to intricately tuned spatial and energetic landscapes. Together, these pigments, preferentially arranged over billions of years of evolution 4,5 , harvest light with a quantum efficiency of near unity. 2 Understanding how the excitation energy transfer (EET) dynamics of LHCII are mapped across the spatial and energetic degrees of freedom (DoF) of the complex has been the focus of significant, sustained attention. [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] This effort has increasingly led to a deeper understanding of Chl-Chl interactions, which predominately manifest energetically in the red edge of the LHCII absorption spectrum (excitonic Qy bands of mainly Chl a and Chl b character give rise to the two peaks centered around 14800 cm -1 and 15500 cm -1 in Figure 1a, respectively). 14 Recent work has 3 also more closely considered the role of Car-Chl interactions, which arise largely at higher energies, i.e. at the blue edge of the LHCII absorption spectrum (peak shoulder rising around 19000 cm -1 in Figure 1a). 17,18 However, over 3000 cm -1 of the spectrum, namely the Chl vibronic Qy-Qx region energetically connecting the Chl Qy states and the Car states 23 , remains little studied. This is not particularly surprising, despite the estimated integrated absorption of the vibronic Qy-Qx region being nearly 75% of that of the Qy region in LHCII, because these states have low oscillator strength and are highly mixed in multiple ways, thus making them challenging to study. Namely, the states of LHCII that span this spectral region have varying degrees of i) mixed inter-pigment electronic character, ii) mixed intra-pigment electronic character (Qx/Qy character), and iii) vibronic character (mixed electronic-vibrational states). The understated importance of the vibronic Qy-Qx states in light-harvesting is also perpetuated by the context in which the complex is typically studied -in vitro -rather than in vivo. In fact, previous work has shown that absorption of the vibronic Qy-Qx states of Chl is enhanced in leaves relative to isolated LHCII, as well as other minor PPCs, to such an extent that the spectrum is essentially uniform over the full photosynthetically active region (PAR). [24][25][26][27] This phenomenon, termed the detour effect, is caused by the highly light scattering environment within the leaf which effectively increases the optical path length of incident green light, therefore, increasing the likelihood of absorption. 25,27 The result is that such a significant increase in the light-harvesting contribution of the vibronic Qy-Qx region occurs such that these states rival the contribution from the Qy electronic region. 1,24 Through an analogous mechanism, yet on even larger scale, green light is also crucial in stimulating photosynthesis in the lower portions of the canopy where red and blue light have been filtered out by the top layers. 27 It is worth noting here that this comparison is in terms of quantum efficiency rather than overall energy efficiency. 28,29 This is to say that the additional energy over the Qy region 4 contained by the photons absorbed by the vibronic Qy-Qx region is unable to be utilized by photosynthesis because the energy required to drive charge separation in the photosynthetic reaction center is equivalent to that of a red, rather than green, photon. 29 However, in vivo, green light drives photosynthesis more successfully in chloroplasts throughout the leaf and plant (despite diminished energy efficiency). This is because of the improved quantum efficiency of the vibronic Qy-Qx states due to the detour effect and the fact that the photons absorbed by these states ultimately penetrate deeper into the mesophyll of a given leaf and the canopy of a whole plant than do the red photons, which are more readily absorbed by the Qy electronic states arising from PPCs residing nearer to the surface of the leaf. However, the lack of spectral assignments or insight into the EET dynamics of the Chl vibronic Qy-Qx states has long hindered a complete understanding of their in role photosynthetic light-harvesting.
In this work, we utilize recent advances in multidimensional spectroscopy, namely the advent of two-dimensional electronic-vibrational (2DEV) spectroscopy 30 , to study the origin and involvement of these highly mixed states -inaccessible to more conventional spectroscopies.
A major advantage of this technique is the improved spectral resolution, afforded by IR detection, which has successfully allowed for insight into ultrafast energy transfer, charge transfer, and proton transfer dynamics. 6,19,[31][32][33][34] IR detection also inherently makes this technique especially sensitive to the mixing of vibronic states because such mixing significantly alters vibrational transitions moments. Additionally, the further sensitivity provided by polarization-dependent 2DEV spectroscopy has been demonstrated in the spectral assignments of monomeric Chl a and b 35 , as well as in unveiling the role of vibronic coupling in a solar cell dye. 34 Here, building on previous applications of this technique, particularly to LHCII 6,19 and monomeric Chl a and Chl b 35 , we apply polarization-dependent 2DEV spectroscopy to study the 5 origin of the highest-lying mixed vibronic Qy-Qx states (spanning 520-570 nm) arising from Chl-Chl interactions in LHCII, in order to gain mechanistic insight into their function in photosynthetic light-harvesting. In doing this, we present direct evidence that this spectral region is dominated by Chl b character, which together with previous in vivo studies indicates that Chl b enhances the ability of green plants and algae to harvest green light. 1,[24][25][26] Following more definitive assignments, we will discuss the role of these states in the EET dynamics of LHCII at 77 K.
Namely, we show that relaxation from the higher-lying states of mainly Chl b character to the lower-lying Qy states occurs on a timescale of <90 fs (within our instrument response function), demonstrating how mixing between the electronic and nuclear DoF of Chl b drives the ultrafast EET dynamics of LHCII and extends efficient light-harvesting throughout the PAR. Further, we find that relaxation from the Qx→Qy states of Chl b occurs on a timescale of ~200 fs (based on the timescales of an oscillatory component associated directly with these states). Such a timescale for Qx→Qy transfer in Chl b agrees well with recent theoretical work on monomeric Chl b 36 , which suggests that the observed Qx states arise from more highly localized Chl b pigments. The ability of polarization-dependent 2DEV to follow the pathways of energy flow for such highly mixed states of LHCII offers a new direction towards a deeper understanding of photosynthetic lightharvesting across the solar spectrum.

Results and Discussion
2DEV spectroscopy, a two-color multidimensional spectroscopic experiment, features visible pump pulses that prepare an ensemble of electronic/vibronic states that evolve during the waiting time, T, and are tracked via an IR probe pulse. 30 The data is presented in the form of excitation frequency-detection frequency correlation plots that map how the electronic/vibronic states evolve with considerable frequency resolution -made possible by IR detection. 6 The 2DEV measurements were performed in two different polarization schemes -one in which the visible pump pair and IR probe were all vertically polarized, S V (ω exc. , T, ω det. ), and one in which the visible pump pair was horizontally polarized while the IR probe was vertically polarized, S H (ω exc. , T, ω det. ). These were combined to generate the perpendicular and parallel polarization-associated (PA) 2DEV spectra, given by: 37,38 and respectively. The perpendicular or parallel distinction indicates the angle between the electronic transition dipole moment (TDM) of states initially populated by the visible pump pair and the vibrational TDM of the probed mode on the states populated during the waiting time, T, that will be amplified in the respective PA spectra. As will be shown below, PA 2DEV spectra are particularly useful for separating the evolution of different states (e.g. states of Qx or Qy character) in highly congested, complex spectra.
Along with the linear absorption spectrum of LHCII ( benefit of this visualization is that if the detection frequency is fixed on a local ground state mode, as will be done here, then this serves as an anchor point to assess the relative orientation between the TDM of this vibrational mode and the TDM of the populated excited state(s). Specifically, to assess the character of this region, the GSB detection frequencies will be set to 1690 cm -1 (Figure   2a and b) and 1680 cm -1 (Figure 2d and e), to track Chl a and Chl b character, respectively. Based on the 2DEV anisotropy data of monomeric Chl b, the angle between the Qy and Chl b-specific GSB (related to the formyl group specific to Chl b) TDMs is approximately 60°. 35 Therefore, as the Qy and Qx TDMs are approximately orthogonal 39 , the angle between the Qx and Chl b GSB should be approximately 30°. When the probed mode is fixed on the Chl b GSB, the S ∥ spectra will selectively amplify the Qx pathway, while suppressing the Qy and vice versa for the S ⊥ spectra.
Within this framework, the significant differences between the perpendicular and parallel PA spectra can be explained in a relatively straightforward way -each of these spectra is mainly   1650 cm -1 . The focus of the ESA analysis will be at a detection frequency of 1670 cm -1 , rather than at 1650 cm -1 , because there is less spectral congestion in this region, although the evolution between at these two detection frequencies is nearly identical (Figure 3a and d versus Figure 3b and e). We note that this is further evidence that these two ESAs have similar origins, i.e. a large degree of character from Chl b pigments only weakly coupled to neighboring chromophores. For the analysis, focus will be on the spectral region around 17800 cm -1 because the energy levels in this region clearly participate more strongly in facilitating transfer to Chl a and therefore play a more significant role in the dynamics. For both of these ESAs, the parallel PA spectra have a significant oscillatory component, as was also seen for the GSB features, and is largely absent in the perpendicular PA spectra for all four detection frequencies. This suggests that the underlying origin of these features is the same, i.e. S ∥ selectively isolates the Qx contribution of these ESA features. This may also indicate that these features in part arise from the formyl group specific to This striking oscillatory behavior is made even more interesting as it is observed in both ESA and GSB features.
Before discussing the origin of these oscillations, it is worth mentioning that the amplitude of the 1670 cm -1 ESA along 17800 cm -1 in S ⊥ undergoes a monoexponential decay on a timescale of 600±200 fs (Figure 3c; shown in red), identical to the rise in the Chl a GSB. Again, this timescale falls within the range of observed Chl b ⟶ Chl a transfer. However, there is clearly a longer timescale component (beyond the duration of the experiment) to the ESA signal because the peak amplitude has yet to fully decay by one picosecond. This is as expected because this feature likely has a significant degree of character from more localized Chl b pigments which undergo Chl b ⟶ Chl a transfer the timescale of a few picoseconds. 15 13 To explain the oscillatory peak dynamics, two main possibilities emerge: 1) rapid vibronic Qy-Qx ⟶ Qy transfer occurs and is followed by energy transfer between the lower-lying Qy states or 2) the oscillatory signal arises from these higher-lying bands themselves. The most straightforward way to verify the origin of this signal is to track the how the optical frequency distribution of this region changes during the waiting time, which is essentially tracking the evolution of the excited state(s) in this region. Dynamical changes in the frequency distribution indicate that the excited state(s) are still populated, which would indicate that the dynamics being observed originate from the higher-lying (Qx) states rather than from the lower-lying Qy states.  14 was determined via fitting with a Gaussian function. It is evident that the frequency distribution changes dynamically along the waiting time. Actually, a fit of this peak evolution reveals oscillatory frequencies of 92 ± 6 cm -1 and 240 ± 20 cm -1 (error indicates 1 interval, see Supplementary Table 2 for complete fit results), which are in agreement within error to those present in the peak amplitude. From the fits, the dominant 92±6 cm -1 frequency component was found to decay on a timescale of 250±50 fs, similar to the timescales of the same component in the ESA and GSB peak amplitudes (190±30 fs and 140±30 fs, respectively). Altogether, this suggests that excitation stays in the Qx bands of Chl b for over 200 fs, in agreement with recent theoretical work on Qx→Qy transfer in Chl b, which was found to occur on a timescale of ~200 fs. 36 Interestingly, the agreement between these timescales suggests that the Qx states of Chl b studied here are not significantly mixed with neighboring pigments, but rather retain significant monomeric-like character (i.e. these observed spectral signatures arise from more localized Chl b pigments). As a control, we show the frequency distribution evolution for the same ESA feature in S ⊥ (Figure 3f). Clearly, no such oscillatory behavior is observed (just as none was observed in the peak amplitudes for this PA component) or in previous work focused on the lower-lying Qy states, 19 suggesting, again, that the oscillatory behavior is unique to the Qx states. Although it remains difficult to assign the exact origin of this signal, especially as the peak evolution analysis of the Chl b GSB feature was hindered due to slightly lower signal, it is likely that these oscillatory signals result from coupling between low frequency chlorin ring distortions and the Qx state. Such a phenomenon has been found to occur in other cases for Chls, where the lifting of the electronic TDM out of the ring results in a coupling and subsequent increase in intensity of the chlorin ring distortion modes. 40,41 The identical, in-phase peak amplitude dynamics between the GSB and ESA of Chl b character is also particularly notable, and may suggest that the observed dynamics arise 15 from non-Condon type coupling. 42 The presence of such a non-Condon effect may help to facilitate rapid internal conversion, therefore further improving quantum efficiency across the PAR.
However, further theoretical efforts are required to support such a conclusion.

Concluding Comments
This application of PA 2DEV spectroscopy to the highly mixed vibronic Qy-Qx region of the LHCII spectrum is one of the first major attempts to understand the origin and function of the

Sample Preparation
The isolation of thylakoid membranes was performed by using sucrose cushion 43 as

Spectroscopic Measurements
The 2DEV experimental setup used in this work, which has been described in detail elsewhere 30 , will be outlined briefly below. A home-built visible NOPA and mid-IR OPA were pumped by a Ti:Sapphire oscillator (Vitara-S, Coherent) and regenerative amplifier (Legend Elite, Coherent). The NOPA was tuned such that the center frequency of the visible pump pulse spectrum was set to ~18415 cm -1 and spanned 17545~19230 cm -1 . A prism pair in combination with a pulse shaper (Dazzler, Fastlite) were used to compress the visible pulses (~35 fs). The energy of the visible pump pulses was ~160 nJ and a f = 25 cm silver coated 90° off-axis parabolic mirror was employed to focus the visible pulses into the sample to a spot size of 250 μm. The mid-IR OPA was tuned to produce an IR probe spectrum with a center frequency of ~1620 cm -1 . The mid-IR pulse was split by a 50:50 ZnSe beam splitter, forming probe and reference beams, where the probe was normalized by the reference, in order to account for shot-to-shot energy fluctuations in the IR source. The IR probe and reference pulses both had an energy of ~100 nJ and duration of ~60 fs.
Both IR pulses were focused into the sample to a spot size of 200 μm with a f = 15 cm gold coated 90° off-axis parabolic mirror. After the sample, the IR probe and reference were dispersed with a spectrometer (Triax 180, Horiba) onto a dual-array 64 element HgCdTe detector (Infrared Systems Development).
As the 2DEV experiments were performed in a partially collinear pump-probe geometry, the pulse shaper was also employed to generate the visible pump pulse pair and to control the relative phase (where the 2DEV signal was isolated with a 3×1 phase cycling scheme) 44,45 and initial time delay, t1, (scanned from 0-100 fs in ~2.4 fs steps) between the pulses. To remove the optical frequency of the pump, the data was collected in the fully rotated frame with respect to t1.
The visible pulse pair was directed towards the sample via a retroreflector on a motorized delay stage used to control the waiting time, T, between the pump pair and probe pulses. In this work, polarization-dependent 2DEV spectra were collected as a function of waiting time in 10 fs increments from 0 fs to 625 fs and in 100 fs increments from 715 fs to 1015 fs. The relative polarization between the pump pulses and the probe pulse was controlled with a λ/2 waveplate in the pump beam line.
Except for the NOPA, the entire setup was purged with dry air, free of CO2 (Perkins Balston FT-IR Purge Gas Generator).

Data Processing
For correct visualization, all data presented in this work have been adjusted for a small residual positive chirp of ~0.1 fs cm -1 in the visible pump pulses, which were unable to be fully compressed to the transform limit by the pulse shaper and prism pair.

Data Availability
The data presented in this study are available from the corresponding author upon reasonable request.