Ann. Rev. Phys. Chem.1986.37 :81-104

Copyright © 1986 by Annual Reviews Inc. All rights reserved

SUBPICOSECOND SPECTROSCOPY

Graham R. Fleming

Department of Chemistry and the James Franck Institute, The University of Chicago, Chicago, Illinois 60637

INTRODUCTION

It is now possible to generate visible and near ultraviolet light pulses with durations corresponding to a few optical cycles. At the time of writing the shortest visible pulse yet generated has a duration of about 8 fs, corresponding to about four optical cycles at the center wavelength (1). Many of the techniques developed for picosecond spectroscopy rely only on the laser pulse for their time resolution and can therefore be carried over directly into the femtosecond regime. Many laboratories now possess the capability to generate pulses in the 50-200 fs range, and considerable activity in the field of subpicosecond spectroscopy can be confidently predicted over the next few years. In this review I discuss some of the techniques for generating and using ultrashort pulses, and then describe applications to a range of chemical and biological processes.

What processes of chemical and biological significance take place in the subpicosecond region? Figure I gives a crude graphical representation of the range of time scales appropriate to a number of relaxation processes and other phenomena. The dynamics of chemical reactions in liquids and solutions is an area of much current interest both experimentally (2-5) and theoretically (6-8). The exchange of momentum and vibrational energy between solutes and solvents, the transfer of electrons or protons, and the response of a solvent to a newly created charge or dipole may all be subpicosecond phenomena. A few examples should make this clear. Acti­vated crossing over a barrier represents a simple model for chemical reactions in solution. The overall, solvent-dependent, barrier-crossing rate depends on the barrier height and the balance between two antagonistic solvent effects: a retardation arising from repeated crossing and recrossing

81 0066--426X/86/1101-D081$02.00

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

82 FLEMING

of the barrier as the result of momentum exchange between reactant and solvent, and an acceleration resulting from vibrational energy exchange between reactant and solvent that maintains thc population of activated reactant molecules and deactivates the product molecules. Both processes are related to the "collision rate," which for a liquid may be crudely estimated from the Enskog relation, TeoII = (pd2/6y/) where p is the density, d the diameter, and Y/ the viscosity. Taking parameters appropriate for liquid ethane gives Teall = 0.5 ps for Y/ = 0.04 cp, the latter corresponding to liquid ethane at room temperature. Thus for typical barrier curvatures of say 1013 s-1, it is clear that large time scale separations between the time spent in the barrier region and the interactions' with the solvent cannot always be safely assumed, and it is therefore necessary to make systematic studies of the rates and mechanisms of energy and momentum exchange before a clear picture of activated barrier crossing can be formulated.

If a reaction involves motion of a charge, the time scale of the solvent's dielectric response will be important. For example, if the solvent cannot

CoLUSION TIME IN UQUIoS

� MOLECULAR ROTATION �

SOLVENT L RELAXA,TION

1--E-OL-EEp-CH-

TA-�- I

N- NIG-C--�

I

VIBRATIONAL OEPHASING I � IN LIQUIDS

VIBRATIONAL RELAXATIO�

I VIBRATIONAL M� � ELECTRONI C RELAXATION

�--�I ----�I----�I --·--+I----�I----�I 10-15 10-14 10-13 10-12 10-11 10-10 10-9

INTRAMOLECULAR PROTON TRANSFER

INTERMOLECULAR PROTON TRANSFER

'-=-ENERGY I C PROTE I N I NTERNAL MDT I ON

TRANSFER I N D--IT PHOTOSYNTHESIS ELECTRON TRANSFER

IN PHOTOSYNTHESIS PHOTOIONIZATION �

CffiOTOOISSOCIATION I I PHOTOCHEMICAL ISOMERIZATION

CAGE RECOMBINATION

Figure 1 Approximate time scales of elementary molecular relaxation phenomena and of some chemical and biological manifestations of these phenomena. Where boxes are not closed the process may occur on longer time scales than I ns. In small molecules such as N2 or CO vibrational relaxation in the liquid state is many orders of magnitude slower than indicated in the figure.

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUB PICOSECOND SPECTROSCOPY 83

respond quickly enough, the transition state will not have equilibrium solvation and there will be a "dielectric friction" on the reaction coordinate (9). The time scales of solvent response range from hundreds of pico­seconds in viscous alcohols, to a few hundred femtoseconds in water. The relative importance of rotational and translational motion in the solvent relaxation is not yet clearly understood (9, 10). For example, it has been proposed that dielectric relaxation in methanol occurs largely via trans­lational motion (II), whereas dielectric relaxation in acetonitrile is expected to be dominated by molecular reorientation.

A number of important biological processes occur very rapidly. Pro­cesses involving excited electronic states, such as the primary electron transfer in photosynthesis or the isomerization of rhodopsin or bacterio­rhodopsin in vision and in the proton pump process, respectively, must be rapid in order to compete effectively with other processes leading to deactivation of the excited state. The internal motions of proteins occur over a time scale of 10-13 s to 102 s. There is considerable current interest in the biological significance of the short time motions.

A number of ultrafast processes occur in molecular crystals as a result of the high mobility of delocalized excitations and of the proximity of neighboring molecules. A novel characteristic of ultrashort light pulses is their small size; this characteristic opens up a number of interesting applications in metrology.

Following a description of the techniques for generating ultrashort pulses, I discuss a series of studies of solvent-solute interactions and liquid state dynamics. Attention is then turned to ultrafast biological processes. A molecular description of these processes requires considerable input from the studies described in the first part of this review and also from the concepts developed in studies of molecular crystals. I then present a summary of subpicosecond spectroscopic studies of electronic and vibrational dynamics in molecular crystals, and conclude with a brief discussion of optical ranging studies.

METHODS OF GENERATING SUBPICOSECOND

PULSES

The topic of generating subpicosecond pulses has been discussed at length elsewhere (12, 13) so in this section I only summarize the basic methods and make brief mention of recent developments. Two laser systems are in common use for generating ultrashort pulses: passively mode locked ring dye lasers, and synchronously pumped dye lasers combined with optical fiber pulse compressors.

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

84 FLEMING

Ring Dye Lasers

The development of the colliding pulse ring dye laser generating 90 fs pulses by Fork, Greene & Shank (14) signaled the beginning of reliable subpicosecond spectroscopy. The basic idea of this laser is to use the interaction or interference ("collision") between two counter circulating pulses in the laser cavity to enhance the effectiveness of the saturable absorber. This is most easily achieved in a ring cavity configuration where the absorber and gain media are separated by one quarter of the cavity round trip distance. The system is self-synchronizing (the least loss being when both pulses pass through the absorber at the same time) and the interference between the two pulses creates a standing wave in the saturable absorber and significantly reduces the energy required to saturate the absorber (15).

Widespread use of femtosecond ring lasers was hindered by a certain amount of "black magic" being involved in selecting intracavity mirrors that gave minimum pulse duration. The suggestion of Fork et al (16) of introducing adjustable group-velocity dispersion into the cavity by means of four prisms has greatly simplified this problem. The prisms can intro­duce either positive or negative net intracavity dispersion and thus com­pensate for group-velocity dispersion produced by mirror coatings, dyes, and dye solvents. In addition, the prisms can compensate for self-phase modulation resulting from the nonlinear response of the dye solvents or time-dependent saturation of the laser dyes (17). Martinez et al (18, 19) have discussed how the combination of intracavity self-phase modulation and group velocity dispersion in the elements of the laser cavity produce pulse shortening by a mechanism analogous to soliton formation in optical fibers (20). This soliton-like shaping can be combined with conventional shaping from saturable absorption to produce pulses that are shorter and more stable than those produced from saturable absorption alone. Valdmanis et al (17) used the four-prism method to generate pulses of 27 fs directly from their ring dye laser. They also noted that 33 fs pulses were obtained from a linear resonator geometry, thus casting doubt on the necessity of the colliding-pulse interaction for the generation of such ultra­short pulses.

The colliding-pulse lasers discussed above are pumped by a continuous argon laser. For some applications, for example, synchronous ampli­fication (see below), there are advantages to pumping the dye laser with a cw mode-locked laser. Vanherzeele et al (21, 22) and Norris et al (23) have described linear-cavity dye lasers that have one end mirror replaced by an antiresonant ring (24) containing a thin saturable absorber jet. The laser is pumped by a mode-locked argon laser (21, 22) Or by a frequency-doubled

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBCOSECOND SPECTROSCOPY 85

cw mode-locked Nd Y AG laser (23). The latter system generates pulses of 85 fs. Nuss et al (25) have described a synchronously pumped colliding­pulse ring laser that generates 100 fs pulses. Colliding-pulse ring lasers are essentially untunable, although small shifts can be obtained by varying dye concentrations. In addition only one combination of gain and absorber dyes-rhodamine 6G and DODCI-is in common use.

Pulse Compression

The shortest optical pulses generated to date have been obtained by sub­jecting amplified pulses from colliding pulse ring lasers to optical pulse compression. Pulses are focused into short lengths of single mode optical fiber. Self-phase modulation in the fiber introduces both higher and lower frequency components into the spectrum of the pulse while group velocity dispersion chirps the pulse. The light emerging from the fiber is recol­lima ted and passed through a grating pair compressor (26), which undoes the chirp and generates a near transform-limited pulse (27). Pulses of 30 fs (28), 16 fs (29), 12 fs (30), and 8 fs (I) have been reported from the compression of colliding-pulse ring-laser outputs.

Tunable ultrashort pulses have been generated by compression of the output of commercial synchronously pumped dye lasers. A single stage of compression is fairly simple to implement and can produce pulses of about 200 fs duration (31). Palfrey & Grischkowsky used two stages of compres­sion, with an intermediate amplification stage, to generate tunable pulses of 16 fs starting with 5.4 ps pulses (32). The pulse compression technique enables tunable ultrashort pulses to be generated over a wide range of wave­lengths. However production of independently tunable, synchronized, pump and probe pulses remains difficult without white light continuum generation. This generally requires amplification of the oscillator output.

Amplification of Ultrashort Pulses

It is desirable for many applications to have higher peak powers than can be produced by the oscillator laser alone. For example, the lack of tun­ability of colliding pulse ring lasers can be overcome by amplification to power levels sufficient to generate a white light continuum. Continuum generation is easier and more reproducible with subpicosecond pulses; for example, Fork et al ( 15) report that 80 fs pulses focused into a 500 /lm flowing jet of ethylene glycol generated a continuum extending from 190 nm to 1600 nm while maintaining an 80 fs duration. Further ampli­fication of selected regions of the continuum has been described by Migus et al (33), thus opening up the possibility of femtosecond pulse generation throughout the optical spectrum.

Amplification schemes at 10 Hz (15, 34, 35), 1-6 kHz ( 1, 36a,b, 37), and

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

86 FLEMING

532 10 Hz

BS1 BS2

.... 1 5 .... 1 .... 2

AC

CPM L .... 5ER 10<---+----1

DODCI R6G

B53

GP S .... 2 .... 3 S .... 3 M

A ... + LASER

Figure 2 A typical experimental configuration for amplifying a ring laser oscillator at 10 Hz. The Y AG laser beamsplitters (BS) are chosen such that most of thc energy pumps the last amplifier state (A4). The saturable absorbers (SA) consist of flowing ethylene glycol jets with a high concentration of malachite green to minimize amplified spontaneous emission. A grating pair (GP) is used to compensate for group velocity dispersion in the amplifier. A real time autocorrelator (Ae) monitors the colliding pulse mode locked ring laser.

1 MHz (38, 39) have been described. This is an area of rapid development, and reliable sources of ultrashort pulses at multi-kHz repetition rates are likely to become more readily available in the near future. A block diagram of a 10 Hz oscillator-amplifier system is shown in Figure 2.

CHEMICAL REACTION DYNAMICS

Isomerization Reactions

The photoisomerization of stilbene (Figure 3) has been extensively investi­gated for many years. The trans-cis isomerization has been much studied by picosecond spectroscopy as a model system for activated barrier crossing in solution (40, 41). A typical rate for trans-cis isomerization is 60 ps (hexane solution 20°C). The cis-trans isomerization is very much faster, and has had to await the development of subpicosecond laser pulses before direct time-resolved studies have been possible. The isomerization process lead­ing from excited (rans- or cis-stilbene to the ground state isomers has long been supposed to take place via a twisted (perpendicular) state, sometimes

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 87

called a "phantom state," as no direct evidence for its existencc has been obtained. The difficulty in observing this state following excitation of the trans-isomer is apparent. If thc twisted state lives for only a fcw pico­seconds, it will be populated more slowly than it decays, and the population will be undetectably small. However, the rapid decay of the cis-isomer should enable a sizeable population of the twisted state to be built up.

These problems have been taken up in a series of papers by Grcene and co-workers, who utilized amplified pulses from a colliding pulse ring laser. Greene & Farrow (42) and Greene & Scott (43) used excitation at 312.5 nm and multiphoton ionization with the fundamental at 625 nm to monitor the decay of excited cis-stilbene in the vapor phase (42), and in hexane solution (43). The MPI signal decayed with a time constant of 0.32 ps in the vapor (isolated molecule conditions) (42) and 1.4 ps in hexane solution at room temperature (43). Absorption measurements by Doany et al (44), in which a 625 11m probe pulse was used, confirmed the latter value, giving a 1.35 ps lifetime for cis-stilbene in hexane. These very rapid decays imply that there is little or no barrier for excited cis molecules to isomerize, as compared with a 3.5 kcaljmole barrier for excited trans molecule iso­merization. Doany et al (44) also studied transient absorption at 312.5 nm and observed a complex signal. Following a pulsewidth-limited rise in absorption, a rapid partial recovery of transmission followed by a slower further increase in optical density was observed. They interpreted the slower rise as the appearance of ground state cis and trans absorptions at 312.5 nm, while the rapid decay they interpreted as the 1.4 ps decay of

1 �lg

70ps

� 1.4 ps

�

1-�

TRANS CIS

TWISTING ANGLE

Figure 3 Summary of stilbene isomerization dynamics in hexane solntion at room tem­perature.

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

88 FLEMING

excited cis-stilbene into the twisted state. A kinetic analysis in which the lifetime of the twisted state was varied gave a value of 3 ± 2 ps for the phantom state. The lifetime of the twisted state is thus longer than that of the excited cis molecule and constitutes a "bottleneck" on the isomerization pathway. A fairly complete picture of stilbene photophysics is now avail­able. The rates of various processes in hexane solution are summarized in Figure 3.

The radiationless decay of the triphenyl methane dyes is believed to involve viscosity-dependent rotation of the phenyl rings toward an equi­librium geometry displaced from that in the ground state (45). The system malachite green in water has been extensively studied in both the time and frequency domains (33, 46-50). These studies gave a range of lifetimes, so it was not clear whether the time and frequency domain studies were in accord. Engh et al (51) studied the ground state recovery of malachite green in water-methanol mixtures by using subpicosecond pulses from a compressed, synchronously mode-locked dye laser. In order to obtain accurate decay times, deconvolution of the instrument response was necessary. However, in a ground state recovery experiment in which identi­cal pump and probe wavelengths are used, the coherent interaction of the two pulses introduces a distortion around zero time called the coherence spike or coherent coupling artifact (52). The antisymmetrization technique suggested by Palfrey et al was used to remove the contribution from the coherence spike and to enable direct deconvolution by using the measured pulse autocorrelation function as the instrument function. The decay time was obtained by using the usual method of iterative reconvolution.

The ground state recovery time for malachite green in water is 1.29 ± 0.33 ps using the antisymmetrization procedure, whereas a decay time of roughly twice this value would be obtained if deconvolution were not performed. The frequency domain polarization spectroscopy studies of Song et al (48) and Trebino & Siegman (49), when corrected for saturation effects (53), gave decay times of 1.2 ± 0.1 and 1.2 ± 0.3 ps, respectively. Saiken & Sei, also using polarization spectroscopy, found 0.7 ps (50). This value is identical to that reported by Siegman before the saturation correction was applied, so it appears that a decay time of 1.2 ps represents the best single exponential lifetime from the frequency domain studies. Thus time and frequency domain studies are in good agreement, though care is clearly needed in the interpretation of the data in both cases.

The photophysics of malachite green seems rather complex. Trebino & Siegman (49), for example, found in a three-laser, induced grating study that their data could not be well fit by a single exponential but was well fit by a double exponential of the form 0.9 exp[-t/(0.78± 0. 1 ps)]+O. l exp [ - t/(7.4 ± 3.0 ps)] or by a uniform range of exponentials extending

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 89

from 0.34± 0.04 to 7.2± 3.0 ps. Two time constants were also observed for this system by Migus et al (33). However, different time constants were observed at different probe wavelengths. Only a single decay time (0.50± 0.05 ps) was found at 700 nm, but at 600 nm and 500 nm two time constants were observed (0.7±O. l ps and 3.0±0.5 ps at 600 nm, and 1.0± 0.1 ps and 3.0± 0.5 ps at 500 nm). The model of Bagchi et al (54) for large amplitude radiationless decay involving no potential barrier predicts nonexponential decay, since the system is in general quite far from equi­librium. However, in view of the findings of Sundstrom & Gillbro (46) of a solvent-dependent barrier in the triphenyl methane dyes, application of the Bagchi et al (54) model to the data is probably premature. The situation is clearly more complex than a simple viscosity-dependent relaxation, since the studies of Engh et al (51) found essentially identical ground state recovery times for MeOH, H20 and MeOH-H20 mixtures. In other words, the recovery time is constant over a range of viscosity from 0.6 cP to 1.8 cPo The complex photophysics of malachite green will probably provide a testing ground for the techniques of ultrafast spectroscopy for some time to come.

The origin and behavior of the coherent coupling artifact, when the rate of dephasing in the system under study is rapid compared with the pulse duration, is fairly well understood (52). However, with the development of pulses in the femtosecond range, a time scale separation between the coherence time of the excitation and probing pulses and the sample phase memory time can no longer be assumed in general. If the material dephas­ing time, Tz, is comparable to the pulsewidth, the shape and position of the coherence spike contain information about T2• Balk & Fleming have analyzed the coherence spike with a model that depends explicitly on the population decay and dephasing rate constants (55). The T2 time for the system of interest can be simply estimated from a study of the coherence width as a function of pulse duration (55), so that ground state recovery experiments may provide a simple alternative to the three-pulse echo technique of Weiner & Ippen (56), discussed below.

Dissociation Reactions The availability of subpicosecond pulses enables spectroscopy to be carried out on the time scale of the initial separation of products in, for example, a unimolecular decomposition. Scherer et al (57) used 400 fs pulses at 306 nm to study the dissociation reaction, leN -+ eN + 1. The appearance of the eN fragment was monitored by laser-induced fluorescence using a

388 nm probe pulse. A rise time of 600 ± 100 fs was observed, consistent with direct dissociation from a repulsive surface. This time scale implies a rotation of the leN parent of 30-600 prior to dissociation. The separation

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

90 FLEMING

of the fragments on the time scale of the experiment is � 10 A. Clearly developments of this type of study open up fascinating possibilities for the direct study of transition states and their evolution into products.

ELECTRON SOLVATION DYNAMICS

The time scale of the appearance of the visible absorption spectrum of the hydrated electron following photo ionization of a probe molecule was one of the earliest problems tackled by picosecond spectroscopy (58, 59). Although electron solvation was time-resolved in alcohol solutions (60), the appearance of the hydrated electron was always limited by the instru­mental resolutions. Experiments with 300 fs resolution by Weisenfeld & Ippen (61) that involved photoionization of ferricyanide found electron hydration more rapid than the instrumental resolution. Studies of electron hydration in micellar systems by Gauduel et al (62) gave an upper limit of 100 fs to the electron solvation time. Studies were carried out using phenothiazine as the probe in anionic micelles in water and in reversed micelles (containing water) in heptane. In the reversed micelle system the solvated electron absorption at 750 nm appeared within the 100 fs time resolution. In the anionic micelle system a rise time of about 400 fs was observed. Gauduel et al (62) suggested this rise time may result from the transfer of the photoelectron through the hydrocarbon phase before solva­tion. The observed rise time increases with increasing surfactant con­centration [350 fs at 0.025 M sodium laural sulfate (SDS) to 450 fs at 0. 1 M SDS] possibly because the electron mobility is decrcased in the higher density surfactant. The faster electron solvation in the reversed micelles is probably related to a difference in the location of the sensitizer in the hydrophobic region in the two systems. The ultrafast solvation of electrons in bulk water is consistent with localization in pre-existing deep traps, rather than a solvation process requiring substantial solvent relaxation (63).

THE OPTICAL KERR EFFECT

The optical Kerr effect is a transient birefringence resulting from an inten­sity-dependent change in the refractive index induced by an intense light pulse propagating through a medium. Picosecond studies of this phenom­enon reveal a major contribution from molecular reorientation; they have been reviewed by Eisenthal (64). Femtosecond measurements have demon­strated the importance of two further contributions to the Kerr signal: an instantaneous electronic contribution, and a rapid interaction-induced (I­I) contribution resulting from the contribution of interactions between pairs, triplets, etc of molecules to the polarizability. The I-I effects relax

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 91

through intermolecular motions that change the relative position and orientation of molecules. For most systems these motions are more rapid than overall molecular reorientation of a single molecule. In the case of CS2 the static or dc Kerr effect is 10% electronic, 40% I-I, and 50% reorientation (65). However, the contribution of thc rapidly decaying contributions will be enhanced for excitation with pulses much shorter than the reorientation time (66); the molecule simply does not have time to move during the excitation pulse.

Subpicosecond resolution studies of the Kerr effect in CS2 have revealed considerable nonexponential character. In addition to the 1.5 ps reorien­tational component, a short component of 0.24 ps (67) or 0.36 ps (68) was observed. This short component is associated with the interaction-induced effects (69, 70) and corresponds to the wing observed in the Rayleigh spectrum (65). A simulation containing the electronic, I-I, and reorien­tational contributions is able to reproduce the experimental response quite well (70). The reorientational contribution only dominates the decay at quite long times, so that reorientation times should be extracted from these measurements with care.

A rapid response was also observed in benzene (71, 72) and nitrobenzene (71). In the case of benzene the fastest relaxation process occurs on a time scale of 0.15 ps (73). The Kerr effect experiment belongs to a general class of measurements that rely on the third-order nonlinear polarization induced by three light frequencies. The relationship between the third­order polarization of the fields is determined by the elements of the third order susceptibility tensor Xijkl. In an isotropic medium, such as a liquid, there are only three independent coefficients related through Xliii = X1l22 + X!2 12 + X 122 I , where I and 2 stand for x, y, or z. When probe and excitation wavelengths are identical, only two components are independent, since X1212 = X1l22. The Kleinman symmetry rule requires XI2l2 = X1221· This latter equality has been tested by Etchepare et al for benzene (73) by using a circularly polarized pump beam. For this con­figuration the Kerr effect signal is proportional to HXI212 - X 1221]. It was found that the two tensor elements are identical to within 5%. Using this equality the values of the various tensor elements for several substances have been obtained (73).

ULTRAFAST DEPHASING MEASUREMENTS

Optical dephasing studies provide a valuable probe of both static and dynamic properties in condensed matter. In many systems of interest the dephasing times lie below I ps and are often masked by inhomogeneous broadening or spectral congestion. Many techniques in both the time and frequency domain have been developed to obtain the homogeneous

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

92 FLEMING

linewidth in an inhomogeneously broadened sample; photon echoes (74), hole burning spectroscopy (75), and polarization spectroscopy (48, 76) are three popular methods. The issue of spectral congestion has been discussed by Lorincz et al (77). These authors point out that when the excitation spans a large number of zero-order vibronic states, the coherent amplitude so generated decays with a time constant characteristic of the light pulse rather than any intermolecular interactions. As the pulse duration is decreased, more states are excited and the loss of phase of the initial superposition occurs still faster, a result that may lead to the (erroneous) conclusion that the T2 time for the system is still shorter than the current pulsewidth. The reason for this behavior can be readily seen by starting with a two-level system and extrapolating to high-level densities. If the two levels are coherently excited, they will beat against each other and the coherent amplitude will oscillate. The beat frequency depends only on the level spacing, and the oscillations correspond to recurrrences of the initial superposition. If the number of levels in the excitation bandwidth is large, the initial coherence decays very rapidly, the recurrence is pushed to very long times, and the oscillations are washed out. Thus the coherent signal will have a rapid decay that closely follows the light pulse (77), and that may dominate the short-time dynamics for large dye molecules in solution.

Weiner & Ippen have developed a three-pulse scattering technique that offers a number of advantages for dephasing studies (56, 78). The method relies on an optically induced grating formed by the interference of pulses I and 2. When the two pulses are separated in time a grating can still be formed provided that the dephasing time, T2, is sufficiently long, since the sample "remembers" the phase information in pulse 1 for time T2• The dephasing time can, in fact, be determined by measuring the grating ampli­tude as a function of the delay between pulses 1 and 2. This is accomplished by using pulse 3 as a delayed probe to scatter off the grating into the background-free directions, k4 = k3 + (k[ - k2) and k5 = k3 - (k[ - k2). The delay of pulse 3 (T) is fixed and is chosen so that T> tp, the pulsewidth, and T < Tt. the population relaxation time.

An alternative technique using incoherent light was proposed by Morita & Yajima (79) and demonstrated by Fujiwara et al (80). The basis of this technique is that temporally incoherent light with a wide spectral width has a very short correlation time te, which corresponds to the inverse of the spectral width. As is well known (12), an autocorrelation function measurement of such light generates a spike of duration te at zero delay, resulting from the substructure. In the technique used by Fujiwara et al (80), a broad band 10 ns dye laser pulse was split into two replicas. One replica traversed a variable delay, and the two pulses with wavevectors k[ and k2 were crossed in the sample. Output beams at k3 = 2k2-k[ and

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 93

k4 = 2k\ - k2 were detected by photomultipliers, and the intensities of the two echo beams along k3 and k4 were monitored as a function of the delay.

In both the techniques discussed above the echoes detected along the two directions are symmetric about zero delay if T2 is zero (78, 79). However, for a finite T2 this is no longer the case, and a shift between the peaks of the two signals as a function of delay should be observed. This shift can be directly related to T2, and thus provides an extremely high time resolution estimate of T2• For example, Fujiwara et al (80) reported a 15 fs uncertainty in their T2 measurements using IO ns laser pulses.

Both the incoherent light and the three-pulse technique have been applied to dye molecules in solution and to dye molecules in polymer hosts at low temperatures (SO, Sl). The experiments in solution were complicated by interference from a thermal grating (80) and by spectral congestion effects. No detectable delays between the two echo signals were observed. Weiner et al suggested that the temporal response of dye molecules in

room temperature solution may be modeled as homogeneous (78). Degenerate four-wave mixing experiments by Diels & McMichael using 100 fs pulses on DODeI solutions gave an upper limit of 10 fs for the dephasing time (S2). Substantial peak shifts were observed for cresyl violet in PMMA at 10 K (81) and cresyl fast violet in cellulose at 10 K (80). In the latter case T2 was 0.7 ps at 10 K, decreasing to 100 fs at 80 K and 20 fs at 300 K for 625 nm excitation (SO). However, for excitation at 594 nm even at 10 K the peak shift is '" 30 fs, and it is hardly detectable at higher temperatures. Fujiwara et al (80) attributed this to the substantially higher density of states for 594 nm excitation, leading to rapid intramolecular relaxation, as opposed to the intermolecular interactions leading to dephasing when excitation is near the origin (625 nm).

An additional feature in the three-pulse technique is the potential to measure spectral cross-relaxation (also called spectral diffusion or T3) by measuring the scattering behavior as a function of the delay of pulse 3 (78). Spectral diffusion refers to the migration of the frequencies of individual absorbers within the inhomogeneous distribution. In the cresyl violet PMMA system the delay of pulse three was varied between 200 fs and 200 ps. No difference in peak shifts were observed either at 15 K or 100 K, indicating the absence of substantial spectral diffusion in this time and temperature range (81).

THE PHOTOCHEMICAL CYCLE OF

BACTERIORHODOPSIN

Bacteriorhodopsin acts as a light-driven proton pump in Holobacterium halobrium, its role being to generate an electrochemical gradient across

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

94 FLEMING

the cell membrane to drive ATP synthesis. The photochemistry of bac­teriorhodopsin is believed to involve an initial isomerization step of the retinal chromophore from the initial (light-adapted) all-trans configuration to the 13-cis configuration (83). The retinal forms a protonated Schiff's base with lysine 216 of the protein matrix. Picosecond studies of deuterated samples led to a suggestion that proton transfer is the primary photo­chemical step (84). Following excitation of the light-adapted pigment, bacteriorhodopsin follows a photochemical cycle of about 10 ms duration. Intermediates, designated J, K, L, M, and 0, are formed during the cycle and are distinguished by their absorption spectra.

Nuss et al (85) used 160 fs, 620 nm pulses to study the primary photo­chemistry of purple membranes (in which the bacteriorhodopsin forms trimers, which are arranged in a two-dimensional hexagonal lattice within the cell membrane) and of detergent isolated bacteriorhodopsin trimers and monomers. In all the samples the first intermediate, J, is formed in 430 ± 50 fs. No deuterium isotope effect was observed on the formation of J, thus suggesting that proton transfer is unlikely to be involved in the primary step. The signals from the purple membrane samples show a peak at zero time corresponding to a decrease in probe intensity. This peak, which has a width of 100 fs, the coherence time of the excitation and probing pulses is absent from the trimeric and monomeric bacterio­rhodopsin signals. Nuss et al suggest that the non-centro symmetric hex­agonal lattice of the purple membrane leads to second harmonic gen­eration and hence depletion of the probe beam during the presence of the strong pump pulse. Aside from this peak the signals from all four samples are very similar, although the monomeric sample contains a longer 1 ps component assigned to the presence of a substantial amount of 13-cis retinal in this preparation. It is thus proposed that excitation of the exci­tonically coupled trimer in purple membrane or isolated trimer leads to localization of the excitation on a single retinal chromophore in less than 50 fs, followed by isomerization to the ground state J intermediate (13-cis, 14-s-cis) in 430 fs. The K intermediate is then formed within 5 ps (85). Evidence that isomerization occurs in the 430 fs step comes from the fact that the absorption spectra of the J and K intermediates are very similar, and that K contains a 13-cis retinal (85). This very rapid isomerization is reminiscent of the cis-stilbene case and again implies the absence of any substantial barrier in the excited state.

Using subpicosecond absorption and emission studies on purple mem­brane solutions, Matveetz et al (86) came to similar conclusions. They studied the transient absorption spectrum from 410-750 nm with 0.3 ps resolution. They found an intermediate absorbing at 460 nm, which decayed to the J state in 0.7 ± 0.3 ps. They assigned this intermediate to

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 95

the excited state of bacteriorhodopsin, rather than a further ground state intermediate before J. This decision was based on the near unit yield of the 460 nm species, compared to the yield of primary photochemistry of 0.3, and on dichroic ratios for 360 nm and 550 nm absorbance. Fluo­rescence measurements using a streak camera with 2 ps resolution showed the bacteriorhodopsin fluorescence lifetime to be < 2 ps, consistent with the above findings (86). Downer et al (87) from their subpicosecond absorption studies of bacteriorhodopsin concluded that the intermediate K (batho­bacteriorhodopsin) is formed directly from the initial excited state in 0.7 ps; a small deuterium effect (r = 1.0 ps in D20) was also observed. Nuss et al (85) suggested that the difference between this analysis and their own results is due to a lack of distinction between the absorbance spectra of J and K in the analysis of Downer et al (87).

ELECTRON TRANSFER IN PHOTOSYNTHETIC REACTION CENTERS

The primary photochemical step in photosynthesis involves the transfer of an electron from an excited special pair of molecules to an acceptor. The recent X-ray crystallographic study of the reaction center of thc bacterium Rhodopseudomonas viridis has revealed the spatial arrangement of the bacteriochlorophyll dimer and its attendant acceptor molecules (88, 89). The bacteriochlorophyll dimer (P) has on each side a bacterio­chlorophyll molecule (BChl ) and a bacteriopheophytin (BPh) molecule, the whole structure being organized with C2 symmetry. In other words, two branches of pigments extend from the dimer, but only one of them is directed toward the "primary" acceptor-a quinone molecule QA' The unidirectional nature of the electron transfer probably arises because of different interactions of the protein matrix with the two BChl molecules that constitute P. These interactions may produce a dipole moment in the dimer through hydrogen bond interactions and thus facilitate the electron motion toward QA (90). Previous picosecond studies of the earlier electron transfer steps clearly indicated one of the BPh molecules (BPhA) as an intermediate with the step BPhA + QA -t BPhA + Q- taking about 200 ps (91-93). However, the involvement of the BChl as a possible intermediate was the object of some controversy (94, 95). Martin et al (96) used 150 fs pulses at 850 nm to directly excite the special pair (dimer) in reaction centers of Rhodopseudomonas sphaeroides R-26 and monitored absorbance changes in the 545-1240 nm region with 100 fs resolution. They concluded that the charge-separated state P+ BPh - is formed in 2.8 ± 0.2 ps. No evidence was found for an intermediate state involving BChl -, as had been suggested by Shuvalov & Klevanik (94).

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

96 FLEMING

Meech et al (97) used accumulated photon echoes and hole burning spectroscopy at 1.5 K to study the early time behavior in reaction centers of R. sphaeroides. The accumulated photon echo (74) is similar to the conventional photon echo in that coherence loss is measured by the ability to generate a photon echo. In the accumulated echo a long-lived bottleneck state is required, in the reaction center system this is the charge-separated state, which lives for several ms at 1.5 K. The echo decay was found to be instrument-limited, implying a population relaxation time of less than 100 fs. The hole burning experiment suggests that the absorption spectrum of P at 900 nm is largely homogeneously broadened, since the whole P absorption band is bleached by narrow band laser excitation at 880 nm. The hole width observed, 425 ± 25 cm -I, corresponds to a population relaxation time of about 25 fs. Meech et al (97) suggest that this process corresponds to charge separation within the dimer pair, which then occurs on a time scale about 100 times shorter than the electron transfer to BPh discussed above.

As biochemical and genetic manipulation techniques improve, studies will become possible on green plant reaction centers. It will be fascinating to make comparisons with the bacterial systems and with crystallographic information. Much activity in this area can be confidently predicted.

LIGAND DISSOCIATION KINETICS IN

HEMOGLOBIN

The dynamical behavior of proteins has received a good deal of attention in recent years (98); however, comparatively few direct experimental obser­vations have been made of the internal motions of proteins. Following extensive work on picosecond and longer time scales, Martin and co­workers have been applying femtosecond spectroscopy to hemeproteins in an attempt to provide data on the rates of local motions inside the heme pocket following ligand detachment (99-101). The principle of the experi­ments is that ligands such as CO, O2, or NO can be photodissociated from hemoglobin or myoglobin and the kinetics of dissociation and rebinding followed spectroscopically. One of the key structural rearrangements associated with ligand dissociation is motion of the iron atom in the heme plane, which is thought to trigger substantial changes in the tertiary structure in hemoglobin. An increase in the Fe-N His (F8) interaction gives rise to a red shift in the Soret band. A species labeled Hb, whose main difference from the equilibrium spectrum is a small red shift (3 nm in both HbCO and Hb02), appears in 350 fs. Martin et al (101) assign Hb as a high-spin state in which the porphoryrin ring is already in its domed conformation. The red shift is not observed following dissociation of

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 97

protoheme, which cannot form the Fe-N bond. The assignment is sup­ported by the very small (1 nm) red shift for HbNO that can be related to a weaker Fe-N His (F8) bond resulting from the unpaired electron of NO in the antibonding dz2 orbital. Two species labeled HbI and Hbn absorb in the 450-500 nm region. HbI absorbs in the 470-500 nm region and has a 350 fs lifetime, making it the likely precursor of Hb+. The other species, Hbn, has a lifetime of 2.5 ps and absorbs in the 445--475 nm region. The yield of Hbn is much larger when the ligand is O2 or NO than with HbeO. The Hbn is considered the result from a channel competing with that producing Hb +. This process is, at least in part, responsible for the lower quantum yield of dissociation for O2 and NO. It is not yet clear what the identity of Hbn is. Extension of picosecond Raman studies (102) into the subpicosecond region should be particularly valuable in revealing the rate and extent of the structural changes following ligand dissociation and for comparison with molecular dynamics calculations (98).

The photophysical behavior of iron-porphorin systems such as heme groups has been studied by a variety of techniques, including picosecond spectroscopy (103), polarization spectroscopy (104), and light scattering ( l05). Thc existence of subpicosecond excited state lifetimes has been proposed in several such systems. For example, Andrews & Hochstrasser (104) used polarization spectroscopy on iron (III) tetraphenylporphine chloride (FeTPP) and concluded that the Soret band is homogeneously broadened with a width of 900 ± 200 cm - I. This corresponds to a popu­lation relaxation time of 3.5 fs. Greene (106) studied transient absorption in FeTPP following subpicosecond excitation at either 625 or 312.5 nm. In either case a � 0.75 ps relaxation was observed; this suggested that both excitation wavelengths lead to a common state that acts as a bottleneck in the degradation of energy in this molecule. The bottleneck state is formed in less than 30 fs following 312.5 nm excitation. Greene suggests that this bottleneck state is the lowest energy singlet charge transfer state. The bottleneck state decays in 0.75 ps to a state proposed to be a charge transfer state of higher mUltiplicity.

RELAXATION PROCESSES IN ORGANIC CRYSTALS

Singlet Exciton Fusion

The dynamics of singlet excitons in organic crystals have been studied for many years. At high excitation densities, the process of exciton-exciton annihilation (fusion) becomes an important deactivation mechanism. This process is also important in photosynthetic light harvesting arrays when high intensity excitation is used, and led to distortion in the decay curves recorded in early investigations (107). In principle, time-dependent energy

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

98 FLEMING

transfer rates are expected in measurements of exciton trapping and annihilation. This is simply because adjacent excitons or excitons adjacent to traps in the lattice will interact first. Theories based on diffusion or on Forster dipole-dipole interaction (in the absence of diffusion) lead to a t-I/2 dependence for the bimolecular annihilation rate, 'Y (108, .1 09). Greene & Millard (110) have investigated exciton fusion in polycrystalline thin­films ( � 1000 A) of p-hydrogen phthalocyanine at 5 K. They used 150 fs pulses at 625 nm for excitation and probing was via a white light continuum. The relaxation dynamics were probed between 19000 cm-I and 22000 cm -I, where there is little ground state absorption and the spectrum can be directly related to the exciton number density. Highly nonexponential decays of the induced absorbance were observed. These were assigned to exciton-exciton annihilation and decays for excitation intensities ranging from 5 x 109 W/cm2 to 5 x 1010 W/cm2 fit well to the expected t-I/2 dependence of the exciton population density. The bimol­ecular exciton annihilation rate constant is 1.0± 0.5 x 10-16 cm3 S-I/2. For excitation intensities above 5 x 1010 W/cm2, it is expected that almost all the lattice sites are excited and that the early time portion of the decay corresponds to the lifetime of proximate excitons. By fitting a single exponential time constant to the first two picoseconds of decay, a tem­perature independent lifetime of 5.7 ps was obtained (110). This rate was independent of excitation intensity above 3 x 1010 W/cm2, a finding that supports the assignment to nearest neighbor annihilation. Millard & Greene (110) therefore suggest that their measurement represents the first characterization of an exciton annihilation rate free from complications due to exciton motion.

Greene & Millard (Il l ) found that the transient bleaching observed at a delay of 2 ps in the low energy region became a transient absorption at � 13500 cm -I at longer times. They attributed this absorbance to the formation of vibrationally hot ground state molecules following annihil­ation, and proposed that long-lived (bottleneck) vibrational states are formed as a result of the annihilation process. Greene & Millard suggest that if this is the case, far infrared solid state dye lasers pumped by visible light may be constructed based on exciton annihilation (111).

Excimer Formation

The formation of excited state dimers or excimers in organic molecules is well known. Excimer formation in crystals provides an excellent model system for chemical reaction studies since the positions and orientations of the reactants are known, thus enabling detailed potential function calculations to be carried out. Williams et al (112) performed sub-

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 99

picosecond resolution transient grating experiments on pyrene crystals, in which excimer formation is well known. The two pump pulses and the probe pulse were 75 fs 620 nm pulses. The decay of the diffracted probe signal showed a prompt component (arising from four-wave mixing) and a decay of 220 ± 50 fs. The pyrene molecules were excited via two photon absorption. Williams et al discussed two plausible explanations for the 220 ps decay: (a) the localized excimer is formed from the initial deloca­lized excitonic state; (b) the decay corresponds to relaxation of the initially excited state, which lies well above the SI origin, down to the SI "pre­excimer" state, which is still delocalized and whose energy lies within the exciton band. The authors tentatively concluded that the excimer form­tion process is responsible for their signal. If this is confirmed, there are interesting implications for the mechanism of excimer formation. This process can be described as an exciton self-trapping process followed by vibrational relaxation of the self-trapped state into the distorted excimer

configuration (113, 114). In the very similar crystal ct-perylene, the motion corresponding to a vibration of a pair of molecules against each other has a frequency of 104 em -I. A minimum excimer formation time should be roughly half the time for a single vibration of this mode, i.e. about 150 fs. Thus the excimer formation time may be simply controlled by the time required for the two molecules to move into the appropriate configuration (112). More extensive studies are clearly needed before this conclusion can be confirmed.

Phonon Dynamics

Lattice energy redistribution in crystals is mediated by optic phonons. The behavior of optic phonons is also important in connection with changes in lattice structure that occur during structural phase transitions, chemical reactions, or melting. De Silvestri et al (115) have made subpicosecond resolution studies of optic phonon oscillations and dephasing in ct-perylene crystals over the range 20 K to 300 K. A pair of ultrashort pulses are crossed with wavevector difference k in the sample. The two pulses excite counter-propagating phonons at ± k, which form a standing acoustic wave. The pulses are shorter than the time of a single vibrational period so that the vibrational mode is driven impulsively. The phonon frequency is derived from the spectral width of the excitation pulses; with pulse durations of less than 100 fs, optic phonon frequencies up to 200 em -I can be investigated. The coherent scattering of a probe pulse incident at the phase-matching angle is monitored as a function of time delay. All three pulses are of the same wavelength. The technique is not highly mode­selective, and any Raman-active mode whose frequency lies within the

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

100 FLEMING

laser linewidth may be excited. If several modes are simultaneously excited, oscillations (beats), corresponding to sums and differences of the fun­damental frequencies, will be observed in the coherently scattered signal. This was the case with ct-perylene. When k was aligned along the b cry­stallographic axis, the 33 and 80 cm - I librational phonons were primarily observed, whereas with k parallel to the a axis the 80 cm - I libron and the 104 cm - I translational mode contributed to the signal. The decay of the 80 cm - I and 104 cm -I modes was very similar (the beat pattern was preserved throughout the entire decay), but the 33 cm- I mode decayed substantially more slowly than the 80 cm - I mode. De Silvestri et al sug­gested that the decay process involves energy- and wavevector-conserving three-phonon processes in which the initial phonon either splits into two lower energy phonons or combines with a thermally excited phonon to produce a higher energy phonon. The long lifetime of the 33 cm -I mode is expected on the basis of density of states arguments; however, the same argument would predict a more rapid decay for the 104 cm - I translational phonon than for the 80 cm-I libron. The 104 cm-I mode, however, is essentially an "internal" vibration of the two molecules of a pair against each other, and may be less well coupled to the other lattice phonons than the 80 cm-1 libron (115). As noted by the authors, study of the 104 cm-1 mode in the excited state will be of some interest since this motion is likely to be closely related to the initial part of the excimer formation reaction coordinate. Thus, the intermolecular interactions relevant to excimer for­mation could be studied. A second area where interesting data are expected is in the behavior of optic phonons near phase transitions.

The initial relaxation of optically excited carriers in polar semicon­ductors is dominated by the emission of small-wavevector LO phonons. Kash et al (116) studied the Raman spectrum of optic�lly pumped GaAs with subpicosecond resolution. The Raman signal was detected by a microchannel plate photomultiplier with a position sensitive anode. This detector enabled the complete Raman spectrum to be recorded and allowed detection of the Raman spectrum with probe powers of less than 1 m W and excitation linewidths in excess of 30 cm -I. For carrier densities of less than 1017 cm-3 the growth of the optically induced non­equilibrium LO phonon population was resolved, and an average electron­phonon scattering time of about 165 fs was deduced. For higher injected carrier densities, substantial time-dependent changes were observed in the spectra that result from screening of the LO phonons by the free carriers and the relaxation of the free carriers to the band edges (116). The new detector used by Kash et al (116) should enable subpicosecond-resolution Raman spectra to be recorded for a range of systems of chemical interest.

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

OPTICAL RANGING

SUBPICOSECOND SPECTROSCOPY 101

The small spatial extent of ultrashort light pulses enables optical ranging to be carried out with very high spatial resolution. For example, Fujimoto et al (117), using 65 fs pulses, demonstrated a spatial resolution of less than 1 5 Jim, with a detection sensitivity to the remitted signal (i.e. that returned from the sample) of less than 10-7 of the incident pulse energy. Their technique required only a slight modification of the standard zero background second harmonic generation method for measurement of pulse autocorrelation functions. Two replicas of pulses from a colliding pulse ring laser were generated by a beam splitter. One replica was focused into the sample by a 5 x microscope objective, which also collimated the remitted or reflected signal Is in the retroreflected direction. The second replica, In traversed a variable delay line. The cross-correlation of IT and I, was measured by combining both pulses in a KDP crystal and =easuring the ultraviolet signal as the variable delay was scanned. The FWHM of the cross-correlation traces resulting from reflection from a pair of microscope slides spaced by 40 Jim was 15 Jim (corresponding to 65 fs sech2 shaped pulses). Thus, boundaries separated by 15 Jim could be resolved, and distances determined with an accuracy of a few micrometers.

Two systems were investigated by Fujimoto et al (117), the cornea of rabbit eyes in vivo and the human skin in vitro. In the corneal system two peaks were observed in the cross-correlation, one corresponding to reflection from the anterior surface of the cornea and the second to reflec­tion from the boundary between the cornea and the aqueous humor. The separation between the two peaks corresponded to 320 Jim-a result in line with previous studies using ultrasonic and optical pachymetry methods. The technique is also applicable to partially transparent bio­logical systems, as evidenced by the studies on human skin (117). Thus, optical ranging with ultrashort light pulses could become a valuable, non-invasive technique for investigating the optical properties (such as refractive index, scattering) and microstructure of biological systems. Fujimoto et al suggest that the technique may enable non-invasive moni­toring of pathological processes and theraputic intervention in non-trans­parent tissues (117).

CONCLUDING REMARKS

I hope that this review has given some indication of the breadth and potential of subpicosecond spectroscopy. As the experimental techniques become more widely disseminated and easier to implement, a rich harvest

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

1 02 FLEMING

of information on fundamental processes In biology, chemistry, and physics can be confidently predicted.

ACKNOWLEDGMENTS

I acknowledge support from the National Science Foundation and the Camille and Henry Dreyfus Foundation (Teacher-Scholar Grant). I thank my students and postdoctorals, particularly Scott Courtney and Mark Maroncelli, for their comments on this review.

Literature Cited

1 . Knox W. H., Fork, R. L., Downer, M. c., Stolen, R. H., Shank, C. V., Valdmanis, J. A. 1985. Appl. Phys. Lett. 46 : 1 120-2 1

2. Fleming, G. R., Balk, M. W., Court­ney, S. H. 1 986. J. Stat. Phys. 42 : 83-104

3 . Lee, M., Holton, G. R., Hochstrasser, R. M. 1985. Chern. Phys. Lett. 1 18 : 359--63

4. Berg, M., Harris, A. L., Harris, C. B. 1985. Phys. Rev. Lett. 54 : 951-54

5. Abul-Haj, N. A., Kelley, D. F. 1986. J. Chern. Phys. 83 : 1 335-44

6. Hynes, J. T. 1985. Ann. Rev. Phys. Chem. 36 : 573-97

7. Adelman, S. A. 1 983. Adv. Chem. Phys. 53 : 61-223

8. Hanggi, P. 1 986. J. Stat. Phys. 42 : 105-48

9. Van der Zwan, G., Hynes, J . T. 1 984. Chem. Phys. 90 : 2 1-35

10. Stiles, P. J., Hubbard, J. B. 1 984. Chern. Phys. 84 : 43 1-39

1 1 . Van der Zwan, G., Hynes, J. T. 1 983. Physica A 12 1 : 227-52

12 . Fleming, G. R. 1986. Chemical Appli­cations of Ultrafast Spectroscopy. New York : Oxford

1 3. J. Opt. Soc. Am. B. 1 985. (Special issue, femtosecond optical interactions.) 2 : 582--680

14. Fork, R. L., Greene, B. I., Shank, C. V. 1 98 1 . Appl. Phys. Leu. 38 : 67 1-72

1 5. Fork, R. L., Shank, C. V., Yen, R., Hirlimann, C. A. 1 983. IEEE J. Quant. Electron. QE I9 : 500--6

16. Fork, R. L., Martinez, O. E., Gordon, J. P. 1 984. Opt. Leu. 9 : 1 50--52

1 7 . Valdmanis, J. A., Fork, R. L., Gordon, J. P. 1 985. Opt. Lett. 10 : 1 3 1-33

1 8 . Martinez, O. E., Fork, R. L., Gordon, J. P. 1984. Opt. Lett. 9 : 1 56--58

19. Martinez, O. E., Fork, R. L., Gordon, J. P. 1985. J. Opl. Soc. Am. B 2 : 753-60

20. Mollenauer, L. F., Stolen, R. H., Gordon, J. P. 1980. Phys. Rev. Lett. 45 : 1095-98

2 1 . Vanherzeele, H., Torti, R., Diels, J.-c. 1984. Appl. Opt. 23 : 4182-1;4

22. Vanherzeele, H., Diels, J.-c., Torti, R. 1 984. Opt. Lett. 9 : 549-5 1

23. Norris, T., Sizer, T. II, Mourou, G. 1 985. J. Opt. Soc. Am. B 2 : 6 1 3- 1 5

24. Siegman, A. E. 1 973. IEEE J. Quant. Electron. QE9 : 247-50 ; Siegman, A. E. 1 98 1 . Opt. Lett. 6 : 334--35

25. Nuss, M. C., Leonhardt, R., Zinth, W. 1985. Opt. Lett. 1 0 : 1 6--1 8

26. Treacy, E. B . 1969. IEEE J. Quant. Electron. QE5 : 454-58

27. Nakatsuka, H., Grischkowsky, D., Balant, A. C. 1 98 1 . Phys. Rev. Lett. 47 : 9 10--1 3

28. Shank, C . V., Fork, R . L., Yen, R., Stolen, R. H. 1 982. Appl. Phys. Lett. 40 : 761-63

29. Fujimoto, J. G., Weiner, A. M., Ippen, E. P. 1 984. Appl. Phys. Lett. 44: 832-34

30. Halbout, J.-M., Grischkowsky, D. 1 984. Appl. Phys. Lett. 45 : 1 28 1-83

3 1 . Nikolaus, B., Grischkowsky, D. 1 983. Appl. Phys. Lett. 43 : 228-30

32. Palfrey, S. L., Grischkowsky, D. 1 985. Opt. Lett. 1 0 : 562-64

33. Migus, A., Antonetti, A., Etchepare, J., Hulin, D., Orszag, A. 1985. J. Opt. Soc. Am. B 2 : 584--94

34. Fork, R. L., Shank, C. V., Yen, R. 1982. Appl. Phys. Lett. 41 : 223-25

35. Migus, A., Martin, J.-L., Astier, R., Antonetti, A., Orszag, A. 1982. Pico­second Phenomena 3 : 6--9. New York : Springer

36a. Sizer, T. II, Kafka, J. D., Duling, I. N. III, Gabel, C. W., Mourou, G. A. 1 983. IEEE J. Quant. Electron. QE I 9 : 506--1 1

36b. Duling, I. N., Norris, T., Sizer, T., Bado, P., Mourou, G. 1 985. J. Opt. Soc. Am. B 2 : 6 1 6--18

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

SUBPICOSECOND SPECTROSCOPY 1 03

37. McDonald, D. B., Jonah, C. D. 1 984.

Rev. Sci. Instrum. 55 : 1 1 66-68

38. Waldeck, D. H., Lotshaw, W. T., McDonald, D. B., Fleming, G. R. 1982. Chem. Phys. Lett. 88 : 297-300

39. Gustafson, T. L., Roberts, D. M. 1 982.

Opt. Commun. 43 : 1 4 1--44 40. Rothenberger, G., Negus, D. K.,

Hochstrasser, R. M . 1983. 1. Chem . Phys. 79 : 5360-67

4 1 . Courtney, S. H., Fleming, G. R. 1 985. J. Chem. Phys. 83 : 2 1 5-21

42. Greene, B. I., Farrow, R. C. 1983. J. Chem. Phys. 78 ; 3336-37

43. Greene, B. I., Scott, T. W. 1 984. Chem. Phys. Lett. 106 ; 399--402

44. Doany, F. E., Hochstrasser, R. M., Greene, B. I. , Millard, R. R. 1 985. Chem. Phys. Lett. 1 1 8 ; 1-5

45. Forster, Th., Hoffmann, G. 1 97 1 . Z. Phys. Chem. 75 ; 63-76

46. Sundstrom, J., Gillbro, T. 1984. J.

Chem. Phys. 8 1 ; 3463-74 47. Erskine, D. J., Taylor, A. J., Tang, C.

L. 1 984. J. Chem. Phys. 80 : 5338-39

48. Song, J. J., Lee, 1. H., Levenson, M. D. 1 978 . Phys. Rev. A 1 7 ; 1439--47

49. Trebino, R., Siegman, A. E. 1983. J. Chem. Phys. 79 : 362 1-26

50. Salken, S., Sei, 1. 1983. J. Chem. Phys. 79 : 41 4fi-:.53

5 1 . Engh, R. A., Petrich, J. W., Fleming, G. R. 1 985. J. Phys. Chem. 89 : 6 1 8-21

52. Palfrey, S. L., Heinz, T. F. 1 985. J. Opt. Soc. Am. B 2 : 674-79

53. Siegman, A. E. 1984. Applications of Picosecond Spectroscopy to Chemistry, ed. K. B. Eisenthal, pp. 363. NATO ASI Ser. C, Vol. 2 1 7 ; 325-5 1 . Dordrecht : Reidel

54. Bagchi, B., Fleming, G. R., Oxtoby, D. W. 1983. J. Chem. Phys. 78 ; 7375-85

55. Balk, M. W., Fleming, G. R. 1985. J. Chem. Phys. 89 : 4300-7

56. Weiner, A. M., Ippen, E. P. 1984. Opt. Lett. 9 : 53-55

57. Scherer, N., Knee, J. L., Smith, D. D., Zewail, A. H . 1985. J. Phys. Chem. 89 : 5 14 1-43

58. Rentzepis, P. M., Jones, R. P., Jortner, J. 1 973. J. Chem. Phys. 59 ; 766-73

59. Huppert, D., Rentzepis, P. M. 1 976. J. Chem. Phys. 64 ; 191-96

60. Kenney-Wallace, G. A., Jonah, C. D. 1 982 . J. Phys. Chem. 86 : 2572-86

6 1 . Weisenfeld, J. M., Ippen, E. P. 1 980.

Chem. Phys. Lett. 73 : 47-50 62. Gauduel, Y, Migus, A., Martin, 1. L.,

Antonetti, A. 1 984. Chem. Phys. Lett. 1 08 : 3 1 9-22

63. Kenney-Wallace, G. A. 1980. Philos. Trans. R. Soc. A 298 : 309-1 9

64. Eisenthal, K. B . 1 977. Ann. Rev. Phys. Chem. 28 : 207-32

65. Battaglia, M. R., Cox, T. I., Madden, P. A. 1 979. Mol. Phys. 37 : 1 4 1 3-27

66. Sala, K., Richardson, M. C. 1 975.

Phys. Rev. A 1 2 : 1036--47

67. Greene, B. I., Farrow, R. C. 1982. 1. Chem. Phys. 77 : 4779-80

68. Halbout, 1. M., Tang, C. L. 1 982. Appl. Phys. Lett. 40 : 765-67

69. Greene, B. I., Farrow, R. C. 1 983. Chem. Phys. Lett. 98 : 273-76

70. Madden, P. A. 1 984. In Ultrafast Phenomena, ed. D. Auston, K. B. Eisenthal, 4 : 244-5 1 . Berlin : Springer. 509 pp.

7 1 . Etchepare, J., Grillon, G., Astier, R . , Marlin, J. L., Bruneau, C . , Antonetti, A. 1 982. In Picosecond Phenomena, ed. K. B. Eisenthal, et ai, 3 : 2 1 7-20. Berlin : Springer. 401 pp.

72. Tang, C. L., Halbout, J. M. 1982. See Ref. 72, pp. 2 1 2-16

73. Etchepare, J., Grillon, G., Thomazeau, I., Migus, A., Antonetti, A. 1 985 . J. Opt. Soc. B 2 : 649-53

74. Hesselink, W. II., Wiersma, D. A. 1983. In Spectroscopy and Excitation

Dynamics of Condensed Molecular Sys­tems, ed. V. Agranovitch, R. M . Hoch­strasser, pp. 249-99. Amsterdam : North-Holland

75. Thijssen, H. P. H., Von der Berg, R., Volter, S. 1 983. Chem. Phys. Lett. 97 : 295-302

76. Levenson, M. D., MacFarland, R. M., Shelby, R. M. 1980. Phys. Rev. B 22 : 49 1 5-29

77. Lorincz, A, Novak, F. A., Rice, S. A. 1984. See Ref. 70, pp. 387-89

78. Weiner, A. M., De Silvestri, S., Ippen, E. P. 1985. J. Opt. Soc. Am. B 2 : 654-62

79. Morita, N., Yajima, T. 1 984. Phys. Rev. A 30 : 2525-36

80. Fujiwara, M . , Kuroda, R., Nakatsuka, H. 1 985. J. Opt. Soc. Am. B 2 ; 1 634-

39

8 1 . De Silvestri, S., Weiner, A. M., Ippen, E. P. 1 984. Chem. Phys. Lett. 1 1 2 : 1 95-

99

82. Diels, J. C., McMichael, I. C. 1 986. J. Opt. Soc. Am. B. In press

83. Braiman, M., Mathies, R. 1 982. Proe. Natl. Acad. Sci. USA 79 ; 403-7

84. App1ebury, M . L., Peters, K. S., Rent­zepis, P. M. 1978. Biophys. 1. 23 : 375-82

85. Nuss, M. C., Zinth, W., Kaiser, W., Kolling, E., Oesterheld, D. 1 985.

Chem. Phys. Lett. 1 1 7 : 1-7 86. Matveetz, Yu. A., Chekalin, S. V. ,

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.

104 FLEMING

Sharkov, A. V. 1985. J. Opt. Soc. Arn. B 2 : 634-39

87. Downer, M. C., Islam, M., Shank, C. V., Harootunian, A., Lewis, A. 1 984. See Ref. 70, pp. 500--2

88. Deisenhofer, J., Epp, 0., Miki, K. , Huber, R., Michel, H. 1984. J. Mol. BioI. 1 80 : 385-98

89. Deisenhofer, J., Epp, 0., Miki, K., Huber, R., Michel, H. 1985. Nature 3 1 8 : 6 18-24

90. Michel, H. Private communication 9 1 . Kaufmann, K . J., Dulton, P. L., Netzel,

T. L., Leigh, J. S., Rentzepis, P. M . 1975. Science 1 88 : 1 301-4

92. Rockley, M. G., Windsor, M . W., Cog­dell, R. J., Parson, W. W. 1975. Proc. Nat!. Acad. Sci. USA 72 : 2251-55

93. Holten, D., Hoganson, C., Windsor, M . W., Schenck, C. C., Parson, W. W., Migus, A., Fork, R. L., Shank, C. V. 1 980. Biochirn. Biophys. Acta 592 : 461-77

94. Shuvalov, V. A., Klevanik, V. A. 1983. FEBS Lett. 1 60 : 5 1-55

95. Kirmaier, C., Holten, D., Parson, W. W. 1985. FEBS Lett. 185 : 76-82

96. Martin, J.-L., Breton, J., Hoff, A. J., Migus, A., Antonetti, A. 1 986. Proc. Nat!. Acad. Sci. USA 83 : 957-61

97. Meech, S. R ., Hoff, A. J. , Wiersma, D. A. 1 985. Chern. Phys. Lett. 1 2 1 : 287-92

98. McCammon, J. A., Karplus, M. 1983. Acc. Chern. Res. 1 6 : 1 87-93

99. Martin, J.-L., Migus, A., Poyart, C., Lecarpentier, Y., Astier, R., Antonetti, A. 1983. EMBO J. 2 : 1 8 1 5-19

100. Martin, J.-L., Migus, A., Poyart, C., Lecarpentier, Y., Astier, R., Antonetti, A. 1983. Proc. Natl. Acad. Sci. USA 80 : 1 73-77

1 0 1 . Martin, J.-L., Migus, A., Poyart, C.,

. Lecarpentier, Y., Astier, R., Antonetti, A. 1984. See Ref. 70, pp. 447-5 1

102. Findsen, E. W., Friedman, J. M., Ondrias, M . R., Simon, S. R. 1985. Sci­ence 229 : 661-65

103. Cornelius, P. A., Steele, A. W., Cher­noff, D. A., Hoehstrasser, R. M. 198 1 . Chern. Phys. Lett. 82 : 9-1 3

104. Andrews, J. R., Hochstrasser, R. M. 1 980. Proc. Nat!. Acad. Sci. USA 77 : 3 1 10--14

105. Friedman, J. M., Rosseau, D, L. 1 978. Chern. Phys. Lett. 55 : 488-92

106. Greene, B. 1. 1985. Chern. Phys. Lett. 1 17 : 191-95

107. Breton, J. , Geacintov, N. 1980. Biochirn. Biophys. Acta 594 : 1-32

108. Powell, R. c., Soos, Z. G. 1973. J. Lurnin. 6 : 35 1-{i1

109. Suna, A. 1970. Phys. Rev. B 1 : 1 7 1 6-39

1 10. Greene, B. I . , Mil1ard, R. R. 1 985. Phys. Rev. Lett. 55 : 1 33 1-34

1 1 1 . Greene, B. I., Mil1ard, R. R. 1986. J. Phys. In press

1 12 . Williams, L. R., Gamble, E. B., Nelson, K. A., De Silvestri, S., Weiner, A. M., Ippen, E. P. 1 986 . Chern. Phys. Lett. In press

1 1 3. Dissado, L. A., Walmsley, S. H. 1984. Chern. Phys. 86 : 375-87

1 14. Collins, M. A., Craig, D. P. 198 1 . Chern. Phys. 54 : 305-21

1 1 5. De Silvestri, S., Fujimoto, J. G., Ippen, E. P., Gamble, E. 8., Williams, L. R., Nelson, K. A. 1985. Chern. Phys. Lett. 1 16 : 146-52

1 16. Kash, J. A., Tsang, J. C., Hvam, J. M. 1985. Phys. Rev. Lett. 54 : 2 1 5 1-54

1 17. Fujimoto, J. G., De Silvestri, S., Ippen, E. P., Puliatito, C. A., Margolis, R., Oseroff, A. 1986. Opt. Lett. In press

Ann

u. R

ev. P

hys.

Che

m. 1

986.

37:8

1-10

4. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by U

nive

rsity

of

Cal

ifor

nia

- B

erke

ley

on 0

5/23

/14.

For

per

sona

l use

onl

y.