Terahertz radiation from bacteriorhodopsin revealscorrelated primary electron and protontransfer processesG. I. Groma*†, J. Hebling‡§, I. Z. Kozma¶�, G. Varo*, J. Hauer¶, J. Kuhl**, and E. Riedle¶

*Institute of Biophysics, Biological Research Center of the Hungarian Academy of Sciences, H-6726, Szeged, Hungary; ‡Department of Experimental Physics,University of Pecs, H-7624, Pecs, Hungary; ¶Chair for BioMolecular Optics, Ludwig-Maximilians-University, D-80538 Munich, Germany; and **Max PlanckInstitute for Solid State Research, D-70569 Stuttgart, Germany

Edited by Mostafa A. El-Sayed, Georgia Institute of Technology, Atlanta, GA, and approved February 20, 2008 (received for review July 6, 2007)

The kinetics of electrogenic events associated with the different stepsof the light-induced proton pump of bacteriorhodopsin is well studiedin a wide range of time scales by direct electric methods. However, theinvestigation of the fundamental primary charge translocation phe-nomena taking place in the functional energy conversion process ofthis protein, and in other biomolecular assemblies using light energy,has remained experimentally unfeasible because of the lack of properdetection technique operating in the 0.1- to 20-THz region. Here, weshow that extending the concept of the familiar Hertzian dipoleemission into the extreme spatial and temporal range of intramolec-ular polarization processes provides an alternative way to studyultrafast electrogenic events on naturally ordered biological systems.Applying a relatively simple experimental arrangement based on thisidea, we were able to observe light-induced coherent terahertzradiation from bacteriorhodopsin with femtosecond time resolution.The detected terahertz signal was analyzed by numerical simulationin the framework of different models for the elementary polarizationprocesses. It was found that the principal component of the terahertzemission can be well described by excited-state intramolecular elec-tron transfer within the retinal chromophore. An additional slowerprocess is attributed to the earliest phase of the proton pump,probably occurring by the redistribution of a H bond near the retinal.The correlated electron and proton translocation supports the con-cept, assigning a functional role to the light-induced sudden polar-ization in retinal proteins.

coherent terahertz emission � excited-state kinetics � nonlinearspectroscopy � sudden polarization � ultrafast charge separation

The general aim of the majority of terahertz spectroscopic studiesis to characterize the response of a particular sample to an

external radiation source in the previously uncovered 0.1- to20-THz regime (1). In contrast to this, in an emerging new branchof time-resolved terahertz emission experiments, the transientradiation is generated within the sample itself for monitoring theunderlying ultrafast light-induced polarization processes. Thismethod has been successfully applied in the study of carrierdynamics in bulk semiconductors, quantum nanostructures, super-conductors (2), and recently even in ordered molecular systems (3,4). Here, we report the observation of light-induced coherentterahertz radiation from a biological sample. It was found that thecomplex phenomenon of the correlated vectorial electron andproton translocation during the primary energy transduction pro-cess of bacteriorhodopsin (bR) is manifested in an experimentallydetectable electromagnetic radiation in the terahertz region. Weshow that the emitted signal carries important kinetic informationabout these fundamental and previously unexplored processes. Theexperimental techniques and the evaluation methods introducedhere can be applied as a protocol in further studies investigatinglight-driven charge transport phenomena in a broad range of otherbiological systems.

The overall light-induced proton translocation function of bRconstitutes a complex sequence of electrogenic events spanning

many decades of the time scale. The most direct way to study theseprocesses is to measure the electrical signal arising from theintramolecular dipole moment change of oriented samples (5–8).Previous studies of this kind were carried out from the ‘‘physio-logical’’ time domain of microseconds to seconds (5) down to therange of several picoseconds (6–8). They established a generallyaccepted model according to which a strict correlation existsbetween the time constants characteristic for the electrogenicevents corresponding to the step-by-step motion of protons acrossthe membrane and those for the photointermediates of bR (9)explored by spectroscopic methods. In these experiments thelight-induced electrical signal was monitored by conventional elec-tronic (5–7) and recently electro-optic sampling (EOS) methods(8), with an inherent high-frequency limit of �100 GHz. Thistechnical limit excluded the direct observation of the ultrafastcharge translocation events taking place in the retinal chromophoreand its proximity during the excited state and right after itsrelaxation. However, in both early (10) and recent literature (11–16), more and more evidence indicates that these initial polarizationprocesses possess a functional role in triggering the isomerization ofretinal and initiating the proton pump.

During its vertical Franck–Condon transition from the ground tothe excited state, the retinal undergoes a large dipole momentchange (17–19). Molecular dynamics simulations (11) supported byphoton echo and transient absorption measurements (12) revealedan inertial dielectric response of the protein occurring on the 50- to100-fs time scale. bR samples reconstituted with synthetic retinalanalogues showed a direct correlation between the existence of alarge, sudden polarization in the chromophore and the occurrenceof the photocycle (13). The time evolution of the retinal polariza-tion was followed indirectly by observing the transient absorptionof the neighboring tryptophan residues (14, 15), resulting in anunresolved component assigned to the Franck–Condon polariza-tion and a second phase developed in �200 fs and attributed to thecharge translocation along the conjugate chain. A more directobservation of the Franck–Condon polarization was carried out bythe detection of the corresponding resonant optical rectificationsignal from bR in coherent infrared emission experiments (16, 19).The sensitive heterodyne detection technique applied in thosestudies covered the extremely high 20- to 50-THz range, but still did

Author contributions: G.I.G., J. Hebling, I.Z.K., J. Hauer, J.K., and E.R. designed research;G.I.G., J. Hebling, I.Z.K., G.V., and J. Hauer performed research; G.I.G. and J. Heblinganalyzed data; and G.I.G. and E.R. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

†To whom correspondence should be addressed. E-mail: groma@brc.hu.

§Present address: Department of Chemistry, Massachusetts Institute of Technology, Cam-bridge, MA 02139.

�Present address: Menlo Systems GmbH, D-82152 Martinsried, Germany.

This article contains supporting information online at www.pnas.org/cgi/content/full/0706336105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

6888–6893 � PNAS � May 13, 2008 � vol. 105 � no. 19 www.pnas.org�cgi�doi�10.1073�pnas.0706336105

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

1

not provide any information on the detailed kinetics of electronpolarization in the retinal chromophore or on the primary events ofthe proton translocation.

Our approach to explore the electrogenic processes falling in thepreviously unfeasible range of several terahertz is essentially basedon the concept of Hertzian dipole radiation. In this case theelementary dipoles correspond to the individual bR molecules,emitting electromagnetic radiation because of the ultrafast electronand proton translocation taking place on their excitation (Fig. 1A).These elementary sources are coherently summed up and form amacroscopically observable terahertz signal because of (i) thenatural assembling of bR molecules into a two-dimensional crys-talline structure in the membrane (20) and (ii) the preparation ofdried oriented purple membrane samples (see Materials andMethods).

Results and DiscussionThe easiest way to demonstrate the existence of the expectedterahertz radiation is to detect the electromagnetic field at a large

distance from the source with the help of focusing optics. Ourexperimental setup realizing this concept is depicted in Fig. 1B. Thelight-induced coherent terahertz signal emitted by bR and observedin this very sensitive arrangement is presented in Fig. 2. The signaldetected in this way clearly demonstrates that the sampled electricfield indeed corresponds to propagating electromagnetic radiation.This is in contrast to previously published EOS measurements (8)detecting a local, nonradiative field directly at the surface oforiented bR samples. The emitted signal spans the 0.1- to 3-THzrange, as shown in Fig. 2 Inset. This range, the upper limit of whichwas determined by the time resolution of our experimental setup,fills an important part of the gap between the frequency regionsaccessible by electronic measuring techniques (5–7) and coherentinfrared emission methods (16). In a control experiment the samplewas pumped at 775 nm where the absorption spectrum of bRvanishes. No terahertz emission was detected at this wavelength,indicating that the process responsible for the terahertz generationis directly related to the photoexcitation of the molecules.

Although it is ideal for the initial demonstration purposesdiscussed earlier, a terahertz detection arrangement involving fo-cusing optics has several serious weaknesses. The largest problemis the difficulty of the exact simulation of the complex diffractionprocesses taking place during the propagation of the terahertzsignal by the different components of the setup. This diffraction ismanifested in not only the spatial but also the temporal transfor-mation of the traveling signal (2, 21–23). While in the near-field theradiated signal is proportional to the first derivative of the sourcepolarization, traveling to the far-field in the free space it transformsto the second derivative. Applying a collimating and focusing pairof mirrors this transformation can be partially compensated (21,23); hence, the shape of the signal shown in Fig. 2 resembles morethe first than the second derivative. Because of their finite size,however, real mirrors cut the low-frequency components of theterahertz field (23). This phenomenon is very undesirable ininvestigating slow polarization processes like those supposed tooriginate from the early phase of the functional proton transloca-tion in bR. Moreover, the corresponding simulation would dependon many poorly defined parameters and one cannot expect areliable analysis of the signal measured in this kind of setup.Another problem with this arrangement is that the pump beamfocused into a small (�1-mm diameter) spot on the sample couldcause unwanted nonlinearity in the excitation.

At the expense of lower sensitivity, all of the problems above canbe substantially reduced by applying near-field observation with amoderately focused pump beam. In a near-field setup no focusingoptics are required and ideally the observed signal is proportionalto the first derivative of the polarization source, resulting in less

Delayed probe pulse

Pump pulse Sample

ZnT e

THz beam

Pump pulse Sample

Detecting system

ZnT e Delayed probe pulse

Detecting system

L ys216

Asp85

Retina l

Plane of sample

W402

THz wave

∆ P electro n

Pump block

Pump block

B

C

A

Fig. 1. The scheme of creation and detection of the terahertz radiation frombR. (A) The active core of the bR molecules. The elementary source of themajor component of the terahertz radiation is the light-induced suddenpolarization of the retinal chromophore as indicated by the arrow. Theelectronic polarization can induce primary proton transfer processes aroundthe chromophore. In a macroscopically oriented sample the polarizationvectors are distributed on a cone surface and form a resultant perpendicularto the plane of the sample. (B) The experimental arrangement for the tera-hertz signal detection at a large distance from the bR sample with focusingoptics. (C) The experimental arrangement for the terahertz signal detection inthe near-field.

Fig. 2. Time evolution of the light-induced coherent terahertz signal frombR observed at a large distance by electro-optic sampling with focusing optics.(Inset) Power spectrum of the signal.

Groma et al. PNAS � May 13, 2008 � vol. 105 � no. 19 � 6889

PHYS

ICS

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

1

suppression of the slow components. Fig. 3A demonstrates theterahertz signal from bR detected by our near-field arrangement(Fig. 1C). Even with this simplified observation scheme the relationof the elementary intramolecular polarization field to the detectedsignal is still complex and its evaluation requires numerical simu-lation (22, 23). The details of our simulation procedure based on themodels for the elementary polarization described below are pre-sented in the supporting information (SI) Text .

Model I: Nonresonant Optical Rectification. The most plausiblemodel for the terahertz emission from bR is optical rectification.This fundamental second-order phenomenon is widely used forthe explanation of terahertz radiation from solid-state samples(2, 24) and was suggested as the underlying mechanism of thecoherent infrared emission recently observed on bR (16). In thesimplest, nonresonant case of optical rectification the elemen-tary intramolecular polarization obeys the envelope of theexciting laser pulse (24). As explained below, however, ourexperiment was carried out under resonant conditions, wherethis simple relation does not hold. Indeed, our simulationindicates (Fig. 3A, gray line) that the nonresonant opticalrectification model is absolutely inconsistent with the complexwaveform of the observed signal. This result agrees perfectlywith the experimentally observed lack of terahertz emissionwhen the sample was pumped at the nonresonant wavelength of775 nm.

Model II: Intramolecular Electron Transfer (Resonant Optical Rectifi-cation). An alternative mechanism leading to terahertz radiation isexcited-state intramolecular electron transfer, demonstrated re-

cently on macroscopically oriented organic dyes, undergoing a largedipole moment change on excitation (3, 4). In the case of bR, sucha light-induced sudden polarization indeed takes place in the retinalchromophore (17, 18). On the basis of fundamental theory ofnonlinear optics (25), the complete process of the creation and therelaxation of this excited-state polarization can be considered as aresonant extension of optical rectification. Although the timeevolution of the intramolecular polarization generated by a non-resonant pump pulse depends only on the parameters of the pumpitself, under resonance it holds detailed information also about thekinetics of the excited-state electron redistribution processes (16).

Previous coherent infrared experiments (16, 19) carried out witha pump pulse of �10-fs duration demonstrated the phenomenon ofoptical rectification on bR, together with a long-lasting multimodaloscillation process. However, applying a detection technique opti-mal in the midinfrared range but having a sharp low-frequencycutoff at �20 THz, these experiments could follow mainly theinstantaneous rise, but not the decay of the excited state, and didnot provide the above extra information available by resonantpumping. In contrast to that, the experimental approach describedhere allows the observation of the actual time evolution corre-sponding to the excited-state electron polarization (although, atpresent, with lower time resolution). Because the extremely broadabsorption spectrum of bR spans the whole visible range (26),pumping at any wavelength in this region induces resonant pro-cesses. To obtain intense radiation, in our experiment the samplewas pumped at the absorption maximum.

In a native bR sample a pure electronic intramolecularpolarization process corresponding to the excited-state chargeredistribution in the retinal chromophore can be adequatelydescribed by a vertical Franck–Condon rise and a single expo-nential decay of �1 � 500 fs (27, 28). This simplified descriptionsupposes a two-state model and neglects the details of theexcited-state dynamics such as wavepacket motion, in-planestretching, and isomerization, observed in many experiments andanalyzed with the help of two-state and three-state models (29,30). All of these events can contribute to the fine details of theexcited-state polarization as observed in the appearance of a�200-fs component in the recent tryptophan transient absorp-tion measurements (14, 15) and the long-lived coherent oscilla-tion detected in IR emission experiments (16). The result of asimulation based on the simplified polarization model above(Fig. 3A, blue line) reproduces the main features of the exper-imental curve, indicating that the major component of theterahertz emission can be clearly attributed to the excited-stateelectron polarization of the retinal.

The experimental terahertz signal contains a small, but welldefined, slow negative component (after 4 ps in Fig. 3A). Theappearance of this component cannot be expected in the frame-work of the model above. As a result of nonexact near-fieldobservation, such a negative phase could take place by a partialtemporal transformation of the signal via a diffraction processduring its travel in space (see SI Text). Indeed, a small, slownegative component also appears in the simulated curve, but itsextent is considerably less than that in the experimental one (seeFig. 3A Inset). Hence, the analysis of the observed signal ischallenged by the problem to attribute this difference to theuncertainties in the simulation of the exact experimental con-ditions or to the existence of a real polarization process of slowkinetics and negative amplitude.

To overcome this difficulty, we repeated the experimentswith the acid blue form of bR, which has a much longerexcited-state lifetime than its native counterpart (31, 32). Thedetected experimental terahertz signal and the correspondingsimulation curves are presented in Fig. 3B. As a result of theprolonged kinetics, the slow negative phase in the terahertzsignal emitted by the acid blue sample is shifted toward longertimes and somewhat reduced in amplitude. In contrast to the

A

B

Fig. 3. Near-field terahertz emission detected from the native (A) and theacid blue form (B) of bR and compared with the results of simulations basedon model I (gray line), model II (blue lines), and model III (red lines). The bumpin the experimental traces at �4 ps is simulated by a reflection originatingfrom the PVC film blocking the pump beam (see Fig. 1 and SI Text). Insets showthe essential part of the signals on an expanded scale.

6890 � www.pnas.org�cgi�doi�10.1073�pnas.0706336105 Groma et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

1

native bR, the acid blue form decays by a biexponential processwith characteristic time constants of �11 � 1.7 ps and �12 � 18ps with an amplitude ratio of a11:a12 � 1:0.3 (31, 32). Calcu-lating with these kinetic parameters (listed also in Table 1)results in an almost completely positive simulated curve (Fig.3B, blue line) with a striking difference from the experimentalone (see Fig. 3B Inset). This clearly indicates that the appear-ance of the slow negative phase observed with both samplescannot be explained by this model. The model therefore has tobe extended by taking into account an additional polarizationmechanism.

Model III: Correlated Electron and Proton Transfer. Whereas themajority of the terahertz radiation observed on different mate-rials has been generated by some kind of electron motion (1, 2),the translocation of any charged particle can be the source ofsuch a radiation, provided that its motion is ordered and theparticle is accelerated on a macroscopic scale. It is reasonable toassume, that the excited-state electron polarization analyzed inmodel II acts as a driving force for the primary phase of thefunctional proton pump, the kinetics of which correlates withthat of the early intermediates of the bR photocycle. Accordingto the generally accepted scheme (29) the majority of the excitedmolecules decays to the J intermediate. The next step of thephotocycle is the formation of the K intermediate, which can bedescribed by exponential kinetics with a time constant (�2) of 3and 6 ps for the native and the acid blue form, respectively (31,32). If these kinetics are associated with a jump of protons fromone metastable position to another, as indicated by previousdirect electric measurements (6–8), on a macroscopic scale thiscorresponds to an accelerated proton current, leading to tera-hertz emission. This component of the emission can be associ-ated with the slow phase observed in our signal. In an energet-ically favored scheme the proton moves in the same direction asthe electron does, explaining the opposite (in our notationnegative) amplitude of this phase.

The total polarization response of the sample, including bothelectron and proton transfer, can be described as

r�t� � ��t���i�1

n

ali e�t

�li

� a2 e�t

�2� c�, [1]

where n � 1 for the native and n � 2 for the acid blue form ofbR and �(t) is the Heaviside step function (see SI Text for thedetailed derivation of this formula). Simulating the terahertzradiation of the native bR sample with this polarization model(Fig. 3A, red line) results in a considerable improvement of thefit over that of model II. The parameters of the simulation arelisted in Table 1, a2 and c were determined from the best fit; allof the others were taken from the literature. The improvementis even more pronounced in the case of the acid blue sample (Fig.3B, red line), where this model completely reproduces the slownegative phase and makes the fit considerably better also in thepositive region. Hence, we attribute the slow negative phase tothe primary step of the proton pump.

The amplitudes and the additive constant in Eq. 1 are related tothe polarization changes taking place in the excited state and in theJ and K intermediates. As described in Eqs. 11–13 of the SI Text, for

native bR these relations are simple; hence, theoretically, thepolarization changes could be calculated from the fitting parame-ters. However, in practice, because of the derivative relationshipbetween the molecular polarization and the terahertz signal and themoderate signal-to-noise ratio for the slow negative component, thequality of the fit with our dataset was mainly sensitive to the sumof a2 and c, leaving the individual values of these parametersuncertain within a wide range. According to Eqs. 12 and 13 in SIText, for native bR this sum corresponds to the polarization of theJ intermediate relative to the bR ground state. In the absence of adetailed reaction scheme, for the acid blue form no such perfectassignment can be made, but clearly the above sum is related to theproton movement. As shown in Table 1, the value of a2 � c (relativeto a11) is �0.2 for the native and �0.8 for the acid blue sample,respectively. The negative sign is in accordance with a unidirec-tional electron and proton movement (in the component perpen-dicular to the membrane.) The lower value of the relative amplitudefor native bR is in full agreement with the fact that in this samplethe formation time of the K intermediate of 3 ps is identical to theexcited-state lifetime of a by-product that contains 13-cis retinal andtends to accumulate in dry samples on illumination (32). Hence, thecontribution to the terahertz signal by the proton translocationoriginating from the main bR form and the electron polarizationfrom the by-product could partially cancel each other. In the caseof the acid blue form of bR such an unfortunate coincidence doesnot exist, as manifested in the higher relative weight of the fittedproton contribution.

High-resolution structural studies indicate that a chain of Hbonds formed by amino acid residues and intraprotein watermolecules plays a substantial role in the proton-pumping mecha-nism of bR (33). We suggest that some of these H bonds locatednearby the retinal chromophore can be perturbed by the light-induced sudden polarization inside the chromophore, leading to aninitial proton translocation process. The most likely candidateparticipating in this mechanism is the closest H bond between theLys-216 residue and the W402 water molecule (Fig. 1A). Thismodel is supported by a recent x-ray diffraction experiment carriedout on a K-like intermediate stabilized at low temperature (34). Itwas found that this bond was indeed highly disturbed comparedwith the equilibrium structure of bR.

Because the shape of the terahertz signal and the intramolecularpolarization are not directly related, in Fig. 4 we plotted the timeevolution of the polarization itself simulated in the framework ofthe three models above. These curves represent the molecularresponse to the 100-fs pump pulse. Throughout the simulation it

Table 1. Kinetic parameters used in the simulations as definedin Eq. 1

bR �11, ps a11 �12, ps a12 �2, ps a2 c

Native 0.5 1 – – 3 �0.1 �0.1Acid blue 1.7 1 18 0.3 6 �0.6 �0.2

Fig. 4. Simulated time evolution of intramolecular polarization generatedby a 100-fs pump pulse in the native (continuous lines) and acid blue (dashedlines) form of bR, corresponding to model I (gray line), model II (blue lines),and model III (red lines). See Materials and Methods and SI Text for a detaileddescription of the models and Table 1 for the kinetic parameters used in thesimulation.

Groma et al. PNAS � May 13, 2008 � vol. 105 � no. 19 � 6891

PHYS

ICS

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

1

was assumed that this response is linear. In a recent detailed studyon linearity (30) it was found that the kinetics of the bR excited stateis independent of the pump intensity provided that the excitationdensity (defined as the product of the absorption cross-section ofthe sample and the photon density) is �0.3 and the peak intensityof the pump is �1010 W cm�2. The corresponding values in ournear-field arrangements were 0.03 and 5 � 108 W cm�2, respec-tively, highly justifying the assumption of linearity. To avoid con-fusion, we stress that the term of linearity is used correctly only inthe context above, i.e., to express that the functional molecularpolarization, and consequently the directly measured electric fieldstrength of the terahertz radiation, depends linearly on the pumpintensity. However, this means that the intensity of the terahertzemission is proportional to the square of the pump intensity, whichin the terminology of optics corresponds to a second-order process.This is completely in accordance with the description of the electronpolarization as resonant optical rectification, and formally, theproton translocation could be also treated in a similar manner.

The aim of the numerical simulations carried out at the initialstate of our study was to show that the observed terahertz emissionis consistent with the kinetics of the fundamental ultrafast processestaking place in bR. For this reason, although our simulationincluded many kinetic and optical parameters (see SI Text), themajority of them were fixed to reasonable published values, andbeyond the scale and location, only the weights of the protonpolarization were free-fitting parameters. Under these circum-stances the applied simulation excellently reproduced the observedterahertz emission. We emphasize, however, that with improvedexperimental techniques leading to better quality of data, one cantest more sophisticated polarization models with the protocoloutlined here. For example, the small, but significant, differencebetween the observed and the simulated signal in the time differ-ence between the main negative and positive peak could be anindication of a more complex electron translocation process ac-companying the population of the excited state. This is in accor-dance with the observation of Schenkl et al. (14, 15), who reportedthe appearance of a second phase in the polarization evolving in�200 fs after the Franck–Condon process in their tryptophantransient absorption measurements. A more detailed study of thetwo phases above, as well as other neglected events of the excited-state dynamics mentioned in the description of model II, requiresincreasing the time resolution of our measuring apparatus byapplying sub-100-fs pump and probe pulses.

In conclusion, we have shown that the time-resolved measure-ment of light-induced terahertz radiation is a very effective tech-nique to gain previously unavailable direct information about themechanism of primary charge translocation in bR. We believe thatthe method is applicable to a broad range of other orderedbiomolecular assemblies using light energy, such as the increasingnumber of retinal proteins discovered and the numerous variants ofphotosynthetic systems. In the case of bR, in particular, it wasdemonstrated that the excited-state electron transfer inside theretinal chromophore and the primary events of proton transfertogether form a correlated sequence of intramolecular polarizationprocesses leading to the generation of the coherent terahertzemission. Our result confirms the emerging idea that such anelectron polarization has a functional role in the initiation of thesubsequent proton pump. From the point of view of fundamentalenergy conversion mechanisms, this corresponds to the extension of

Mitchell’s chemiosmotic theory (35) from the territory of quasi-electrostatics to that of ultrafast electrodynamics.

Materials and MethodsPurple membranes of Halobacterium salinarum prepared by standard meth-ods were macroscopically oriented and deposited by electric field on a glasssubstrate covered by an indium-tin-dioxide thin layer as described in ref. 5.After drying, this procedure resulted in a sample thickness of �5 �m, corre-sponding to �1,000 membrane layers. To obtain the acid blue form of bR thedried samples were rinsed for 2 min first in distilled water, then in H2SO4 ofpH � 0.5, washed by deionized water, and dried again.

The samples were excited by a noncollinearly phase-matched optical para-metric amplifier (NOPA) pumped by the pulses of a 1-kHz Ti:sapphire amplifier(CPA 2001; Clark-MXR) tuned to 570 nm and compressed to a 100-fs pulseduration. Because of the symmetry of the sample only the dipole momentchange perpendicular to the membranes could be observed (Fig. 1A). Becausethe corresponding radiation is zero in this direction, the exciting beamreached the sample surface at 45° (from the side of the glass substrate to avoiddegradation of the terahertz signal by passing through it).

The terahertz radiation was detected by electro-optic sampling (EOS) in a0.4-mm-thick 110-oriented ZnTe crystal (36). To avoid side effects on thecrystal, the pump beam passed through the sample was blocked by a Teflonwindow, transparent at terahertz frequencies. The ZnTe crystal was, in addi-tion, illuminated by a probe pulse provided by a small fraction of the Ti:sap-phire amplifier output (775 nm, pulse length 225 fs). The field-induced bire-fringence acting on the probe beam was detected by a combination of a �/4plate, a Rochon-prism, and a pair of balanced photodiodes. The temporalevolution of the terahertz electric field was sampled by the variation of thetime delay between the pump and probe pulses.

To obtain high-quality detection at a large distance from the source (Fig. 1B),the terahertz radiation generated in the sample was projected onto the EOScrystal in a 4:1 ratio by a pair of off-axis parabolic gold mirrors (focal length, 100and 25 mm). The total distance between the creation and the observation of theradiation was �150 mm. In this arrangement, the probe beam was combined totravel collinearly with the terahertz beam through a thin pellicle. In the othersetupbuilt fornear-fieldobservation, theEOScrystalwaspositionedimmediatelybehindthesample(Fig.1C).Becauseoftheconstraintofthe45°arrangement, theactual average distance of detection was �5 mm. In this case the probe pulseilluminated the EOS crystal from its back side and the reflected beam wasdetected. To avoid damaging and nonlinear response of the bR film by the 5-�Jpulse energy of the pump and to fulfill the criteria of near-field detection, thediameterof the illuminatedspotonthesamplesurfacewaskept large(�3.5mm).In this arrangement the EOS crystal was protected from the pump beam by a thinblack PVC film.

In a control experiment carried out in the sensitive double-parabolic-mirrorsetup the sample was pumped directly by part of the 775-nm pulse of theTi:sapphire laser. Because at this wavelength bR has practically no absorption,in this measurement the pulse energy of the pump could be increased to 20 �Jwithout causing any photodamage. This further increased the effective de-tection sensitivity for any possible terahertz signal generated in the sample.

Because the signals detected both in the near-field and at a large distancedo not carry any valid kinetic information �4 THz, all of the presented traceswere frequency-filtered above this value.

To exclude possible artifacts we repeated the terahertz generation proce-dure by exposing the indium-tin-dioxide electrode to the exciting beam bothwithout any and with unoriented bR sample on it. These control experimentsresulted in no detectable signal beyond the noise level.

The details of the simulation procedure applied for the evaluation of theobserved terahertz signals are presented in SI Text.

ACKNOWLEDGMENTS. This work was supported in part by Hungarian Scien-tific Research Fund Grants T048725 (to G.I.G. and J.Hebling) and T048706 (toG.V.), National Office for Research and Technology, Hungary Grant NKFP1/0007/2005 (to G.I.G and J.Hebling), and the Ultrafast Structural Dynamics inPhysics, Chemistry, Biology, and Material Science Program of European Sci-ence Foundation (G.I.G and J.Hebling).

1. Ferguson B, Zhang XC (2002) Materials for terahertz science and technology. Nat Mater1:26–33.

2. Schmuttenmaer CA (2004) Exploring dynamics in the far-infrared with terahertzspectroscopy. Chem Rev 104:1759–1779.

3. Carey JJ, et al. (2002) Terahertz pulse generation in an organic crystal by opticalrectification and resonant excitation of molecular charge transfer. Appl PhysLett 81:4335– 4337.

4. Beard MC, Turner GM, Schmuttenmaer CA (2002) Measuring intramolecular chargetransfer via coherent generation of THz radiation. J Phys Chem A 106:878–883.

5. Varo G, Keszthelyi L (1983) Photoelectric signals from dried oriented purple mem-branes of Halobacterium halobium. Biophys J 43:47–51.

6. Simmeth R, Rayfield GW (1990) Evidence that the photoelectric response of bacterio-rhodopsin occurs in less than 5 picoseconds. Biophys J 57:1099–1101.

7. Groma GI, Hebling J, Ludwig C, Kuhl J (1995) Charge displacement in bacteriorho-dopsin during the forward and reverse bR-K phototransition. Biophys J 69:2060 –2065.

8. Xu J, et al. (2003) Direct measurement of the photoelectric response time of bacterio-rhodopsin via electro-optic sampling. Biophys J 85:1128–1134.

6892 � www.pnas.org�cgi�doi�10.1073�pnas.0706336105 Groma et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

1

9. Lanyi JK (1992) Proton-transfer and energy coupling in the bacteriorhodopsin photo-cycle. J Bionenerg Biomembr 24:169–179.

10. Lewis A (1978) Molecular mechanism of excitation in visual transduction and bacte-riorhodopsin. Proc Natl Acad Sci USA 75:549–553.

11. Xu D, Martin C, Schulten K (1996) Molecular dynamics study of early picosecond eventsin the bacteriorhodopsin photocycle: Dielectric response, vibrational cooling, and theJ, K intermediates. Biophys J 70:453–460.

12. Kennis JTM, et al. (2002) Ultrafast protein dynamics of bacteriorhodopsin probed byphoton echo and transient absorption spectroscopy. J Phys Chem B 106:6067–6080.

13. Zadok U, Khatchatouriants A, Lewis A, Ottolenghi M, Sheves M (2002) Light-inducedcharge redistribution in the retinal chromophore is required for initiating the bacte-riorhodopsin photocycle. J Am Chem Soc 124:11844–11845.

14. Schenkl S, van Mourik F, van der Zwan G, Haacke S, Chergui M (2005) Probing theultrafast charge translocation of photoexcited retinal in bacteriorhodopsin. Science309:917–920.

15. Schenkl S, et al. (2006) Insights into excited-state and isomerization dynamics ofbacteriorhodopsin from ultrafast transient UV absorption. Proc Natl Acad Sci USA103:4101–4106.

16. Groma GI, et al. (2004) Resonant optical rectification in bacteriorhodopsin. Proc NatlAcad Sci USA 101:7971–7975.

17. Mathies R, Stryer L (1976) Retinal has a highly dipolar vertically excited singlet-state -implications for vision. Proc Natl Acad Sci USA 73:2169–2173.

18. Huang JY, Chen ZP, Lewis A (1989) 2nd-harmonic generation in purple membrane-poly(vinyl alcohol) films—Probing the dipolar characteristics of the bacteriorhodopsinchromphore in bR570 and M412. J Phys Chem 93:3314–3320.

19. Colonna A, Groma GI, Martin J, Joffre M, Vos MH (2007) Quantification of suddenlight-induced polarization in bacteriorhodopsin by optical rectification. J Phys Chem B111:2707–2710.

20. Amos LA, Henderson R, Unwin PNT (1982) 3-dimensional structure determination byelectron-microscopy of two-dimensional crystals. Prog Biophys Mol Biol 39:183–231.

21. Kaplan AE (1998) Diffraction-induced transformation of near-cycle and subcyclepulses. J Opt Soc Am B 15:951–956.

22. Wynne K, Carey JJ (2005) An integrated description of terahertz generationthrough optical rectification, charge transfer, and current surge. Opt Commun256:400 – 413.

23. Cote D, Sipe JE, van Driel HM (2003) Simple method for calculating the propagation ofterahertz radiation in experimental geometries. J Opt Soc Am B 20:1374–1385.

24. Bonvalet A, Joffre M (2005) Terahertz Femtosecond Pulses. Femtosecond Laser Pulses,ed. Rulliere C (Springer, New York), pp 309–331.

25. Mukamel S (1995) Principles of Nonlinear Optical Spectroscopy (Oxford Univ Press,Oxford, UK).

26. Varo G, Lanyi JK (1989) Photoreactions of bacteriorhodopsin at acid pH. Biophys J56:1143–1151.

27. Dobler J, Zinth W, Kaiser W, Oesterhelt D (1988) Excited-state reaction dynamics ofbacteriorhodopsin studied by femtosecond spectroscopy. Chem Phys Lett 144:215–220.

28. Mathies RA, Cruz CHB, Pollard WT, Shank CV (1988) Direct observation of the femto-second excited-state cis-trans isomerization in bacteriorhodopsin. Science 240:777–779.

29. Abramczyk H (2004) Femtosecond primary events in bacteriorhodopsin and its retinalmodified analogs: Revision of commonly accepted interpretation of electronic spectraof transient intermediates in the bacteriorhodopsin photocycle. J Chem Phys120:11120–11132.

30. Schmidt B, et al. (2005) Excited-state dynamics of bacteriorhodopsin probed bybroadband femtosecond fluorescence spectroscopy. Biochim Biophys Acta1706:165–173.

31. Logunov SL, El-Sayed MA, Song L, Lanyi JK (1996) Photoisomerization quantum yieldand apparent energy content of the K intermediate in the photocycles of bacterio-rhodopsin, its mutants D85N, R82Q, and D212N, and deionized blue bacteriorhodop-sin. J Phys Chem 100:2391–2398.

32. Colonna A, Groma GI, Vos MH (2005) Retinal isomerization dynamics in dry bacterio-rhodopsin films. Chem Phys Lett 415:69–73.

33. Luecke H (2000) Atomic resolution structures of bacteriorhodopsin photocycle inter-mediates: The role of discrete water molecules in the function of this light-driven ionpump. Biochim Biophys Acta Bioenerg 1460:133–156.

34. Matsui Y, et al. (2002) Specific damage induced by x-ray radiation and structuralchanges in the primary photoreaction of bacteriorhodopsin. J Mol Biol 324:469 –481.

35. Mitchell P (1961) Coupling of phosphorylation to electron and hydrogen transfer by achemi-osmotic type of mechanism. Nature 191:144–148.

36. Wu Q, Litz M, Zhang XC (1996) Broadband detection capability of ZnTe electro-opticfield detectors. Appl Phys Lett 68:2924–2926.

Groma et al. PNAS � May 13, 2008 � vol. 105 � no. 19 � 6893

PHYS

ICS

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Aug

ust 2

, 202

1