274 FEMTOCHEMISTRY ATOMIC-SCALE DYNAMICS OF THE CHEMICAL BOND USING ULTRAFAST LASERS Nobel Lecture, December 8, 1999 by Aumep H. Zewat California Institute of Technology, Pasadena, USA. Over many millennia, humankind has thought to explore phenomena on an ever shorter time scale. In this race against time, femtosecond resolution (1 fs = 101 second) is the ultimate achievement for studies of the fundamental dy- namics of the chemical bond. Observation of the very act that brings about che- mistry - the making and breaking of bonds on their actual time and length scales — is the wellspring of the field of Femlochemistry, which is the study of molecular motions in the hitherto unobserved ephemeral transition states of physical, chemical and biological changes. For molecular dynamics, achieving this atomic-scale resolution using ultrafast lasers as strobes is a triumph, just as x-ray and electron diffraction, and, more recently, STM and NMR, provided that resolution for static molecular structures. On the femtosecond time scale, matter wave packets (particle-type) can be created and their coherent evolution as a single-molecule trajectory can be observed. The field began with simple systems of a few atoms and has reached the realm of the very complex in isolated, mesoscopic and condensed phases, and in biological systems such as proteins and DNA. It also offers new possibilities for the control of reac tivity and for structural femtochemistry and femtobiology. This anthology gives an overview of the development of the field from a personal perspective, encompassing our research at Caltech and focusing on the evolution of tech- niques, concepts, and new discoveries.Ahmed H. Zewail 275 TABLE OF CONTENTS: Abstract L IL. Mm. Prologue Dynamics and Arrow of Time (1) Origins - From Kinetics to Dynamics ‘The Arrhenius Seminal Contribution The London, Eyring and Polanyi Contributions ‘Transition State and Definition ‘Transition State and Spectroscopy (2) Arrow of Time Femtochemistry: Development of the Field (1) The Early Years of Coherence New Techniques for Molecules Optical Analogue of NMR: Controlling the Phase (2) The Marriage with Molecular Beams The Anthracene Discovery: A Paradigm Shift ‘The Successful 036 Laboratory Changing A Dogma: Development of RCS ransition to the Sub-Picosecond Regime ‘A New Beam Machine: Pump-Probe Mass Spectrometry The First ICN Experiment: Sub-picosecond Resolution (4) The Femtosecond Dream A Piece of Good Fortune ‘The Classic Femtosecond ICN Discovery ‘The Nal Discovery: A Paradigm for the Field The Saddle-Point Transition State ‘The Uncertainty Principle Paradox Bimolecular, Bond Making & Breaking: Bernstein’s Passion Ultrafast Electron Diffraction (UED) Clusters, Dense Fluids & Liquids, and New Femtolands Theoretical Femtochemistry Experimental Femtochemistry (5) Femtocopia ~ Examples from Caltech Elementary Reactions & Transition States Organic Chemistry Electron & Proton Transfer Inorganic & Atmospheric Chemistry The Mesoscopic Phase: Clusters & Nanostructures The Condensed Phase: Dense Fluids, Liquids & Polymers276 Chemistry 1999 (6) Opportunities for the Future Transient Structures from Ultrafast Electron Diffraction Reaction Control Biological Dynamics IV. Impact and Concepts - A Retrospective ‘Time Resolution — Reaching the Transition-State Limit Atomic-Scale Resolution Generality of the Approach Some Concepts (1) Resonance - Non-equilibrium Dynamics (2) Coherence - Single-molecule-type Dynamics (3) Transition Structures - Landscape Dynamics (4) Reduced Space — Directed Dynamics V. Epilogue Acknowledgments References Bibliography: The Caltech Research Some General References I. PROLOGUE In the history of human civilization, the measurement of time and recording the order and duration of events in the natural world are among the earliest endeavors that might be classified as science. The development of calendars, which permitted the tracking of yearly flooding of the Nile valley in ancient Egypt and of the seasons for planting and harvesting in Mesopotamia, can be traced to the dawn of written language. Ever since, time has been an impor- tant concept [1] and is now recognized as one of the two fundamental di- mensions in science. The concept of time encapsulates an awareness of its duration and of the passage from past to present to futureand surely must have existed from the very beginning with humans searching for the meaning of birth, life and death and, in some cultures, rebirth or recurrence. My ancestors contributed to the beginning of the science of time, develop- ing what Neugebauer [1] has described as “the only intelligent calendar which ever existed in human history”. The “Nile Calendar” was an essential part of life as it defined the state of yearly flooding with three seasons, the Inundation or Plooding, Planting, and Harvesting, each four months long. A civil year lasting 365 days was ascertained by about 3000 BC or before, based on the average time between arrivals of the flood at Heliopolis, just north of Cairo; nilome- ters were used in more recent times, and some are still in existence today. By the time of the First Dynasty of United Egypt under Menes in ca. 3100 BG, the scientists of the land introduced the concept of the “Astronomical Calendar” by observing the event of the helical rising of the brilliant star Sothis (orAhmed H. Zewail 277 Sirius). Inscribed on the Ivory Tablet, dating from the First Dynasty and now at the University Museum in Philadelphia, were the words, “Sothis Bringer of the Year and of the Inundation” [1, Winlock]. On the Palermo Stone, the an- nals of the kings and their time-line of each year’s chief events were document- ed from pre-dynastic times to the middle of the Fifth Dynasty [1, Breasted]. Thus, as early as 3100 BG, they recognized a definite natural phenomenon for accurate timing of the coming flood and recounted the observed reap- pearance of the star as the New Year Day — real-time observation of daily and yearly events with the zero-of time being well defined!’ In about 1500 BG, another major contribution to this science was made, the development of Sun-Clocks, or Sundials, using moving shadows (Fig. 1]. Now in Berlin, the sun-clock bearing the name of Thutmose III (or Thothmes after the Egyptian God Thoth of wisdom and enlightenment), who ruled at Thebes from 1447-1501 BC, shows the graduation of hours for daytime measurements. This clock with uneven periods for hours was man-made and transportable. For night time, the water-clock was invented, and the device provided even periods for timing. With these developments, the resolution of time into periods of year, month, day and hour became known and has now been used for more than three millennia. The durations of minutes and se- conds followed, using the Hellenistic sexagesimal system, and this division, ac- cording to Neugebauer, “is the result of a Hellenistic modification of an Egyptian practice combined with Babylonian numerical procedures.” About 1300 AD, the me- chanical clock [1, Whitrow] was advanced in Europe, ushering in a revolution in precision and miniaturization; Galileo began studies of pendulum motions and their clocking, using his heartbeats (seconds), in 1582 AD. Our present time standard is the cesium atomic clock’, which provides precision of about 1:10", ie., the clock loses or gains one second every 1.6 million years. For this work, Norman Ramsey shared the 1989 Nobel Prize in Physics. Until 1800 AD, the ability to record the timing of individual steps in any process was essentially limited to time scales amenable to direct sensory per ception - for example, the eye’s ability to see the movement of a clock or the Recognizing the incommensurability of the lunar month and the solar year, they abandoned the lunar month altogether and used a 30-day month, Thus the year is made of 12 months, 30 days each, with 5 feast days at the end of the year. The “Civil Calendar” was therefore 365 days per year and differs from the astronomical calendar of Sothis by approximately a 1/4 day every year. The two calendars must coincide at intervals of 365 x 4=1460 years, and historians, based on re- corded dates of the reappearance of Sothis in dynastic periods, give the four dates for the coin- cidence of both calendars: 139 AD, 1317 BC, 2773 BC, and 4233 BC (+139-3 x 1460 = 4241 BC). Even though the Egyptians discovered the astronomical calendar, 365-1/4 days, they decided, presumably for “bookkeeping”, to use the civil calendar of 365 without leap years. They also divided the day into two periods of 12 hours each for day and night time. The remarkable calen- dar of years, months, days, and hours was adapted throughout history and formed the basis for the 365.25-day Julian (46 BC, Alexandria school) and 365.2422-day Gregorian (from 1582 AD, Pope Gregory XIII) calendars. In the words of the notable Egyptologist J. H. Breasted [1], “It has thus been in use uninterruptedly over six thousand years,” Many historians regard the date around 4241 BC as the beginning of history itself as it defines the period of written records; anything earlier is prehistory 2 Since 1967 one second has been defined as the time during which the cesium atom makes ex- actly 9, 192,631,770 oscillations.278 Chemistry 1999 Glendar Sundial Mechanical Cock Fast Photography, Laver Femioscopy: teal me/ard, — (Ponptawant (icone bla sa ai Figure 1, Timeline of some events in the history of measurements of time, from yearly calendars to the femtosecond regime (see text). car’s ability to recognize a tone. Anything more fleeting than the blink of an eye (~0.1 second) or the response of the ear (~0.1 millisecond) was simply be- yond the realm of inquiry. In the nineteenth century, the technology was to change drastically, resolving time intervals into the sub-second domain, The famous motion pictures by Eadweard Muybridge (1878) of a galloping horse, by EtienneJules Marey (1894) of a righting cat, and by Harold Edgerton (mid-1900’s) of a bullet passing through an apple and other objects are ex- amples of these developments, with millisecond to microsecond time resolu- tion, using snapshot photography, chronophotography and stroboscopy, re- spectively [2]. By the 1980's, this resolution became ten orders of magnitude better [see Section 112], reaching the femtosecond scale’, the scale for atoms and molecules in motion. The actual atomic motions involved in chemical reactions had never been observed in real time despite the rich history of chemistry over two millennia [3], as khem became khemia, then alchemy, and eventually chemistry [3,4] — see the Stockholm Papyrus in Fig. 2. Chemical bonds break, form, or geome- trically change with awesome rapidity. Whether in isolation or in any other phase, this ultrafast transformation is a dynamic process involving the me- chanical motion of electrons and atomic nuclei. The speed of atomic motion °"The prefix milli comes from Latin (and French), micro and nano from Greek, and pico from Spanish, Femto is Scandinavian, the root of the word for “fifteen”; nuclear physicists call femto- meter, the unit for the dimensions of atomic nuclei, fermi. Ato is also Scandinavian.Ahmed H. Zewail 279 Figure 2. A page of the Stockholm papyrus (300 AD) describing the recipe for “making” (actual- ly imitating) emerald. Note that change has been at the heart of chemistry from its millennia-old definition KM, to Chémia and Alchemy, and to Chemistry. [Ref. B26, and references therein] Equilibrium, kinetics and dynamics are the foundation for the describtion of chemical changes. is ~ 1 km/second and, hence, to record atomic-scale dynamics over a distan- ce ofan Angstrm, the average time required is ~ 100 fs. The very act of such atomic motions as reactions unfold and pass through their transition states is the focus of the field of femtochemistry. With fs time resolution we can “free- ze” structures far from equilibrium and prior to their vibrational and rotatio- nal motions, or reactivity. The pertinent questions about the dynamics of the chemical bond are the following: How does the energy put into a reactant molecule redistribute among the different degrees of freedom, and how fast does this happen?; What are the speeds of the chemical changes connecting individual quantum states in the reactants and products?; What are the detai-280 Chemistry 1999 semicon | | Plosiew). Cheintesh and Blalegtaal | soso btocas Mode | Figure 3, Time scales. The relevance to physical, chemical, and biological changes. The funda mental limit of the vibrational motion defines the regime for femtochemistry. Examples are given for each change and scale. [Ref. B4, B10] led nuclear motions that chart the reaction through its transition states, and how rapid are these motions? As pointed out by Jim Baggot, “the entire history of chemical reaction dynamics and kinetics has been about providing some approxima- te answers to these three questions [4].” In femtochemistry, studies of physical, chemical or biological changes are the fundamental time le of molecular vi tions [Fig. 3]. In this sense, femtoscience represents the end of the race against time, or, as the report of Ref. [5] puts it, ”. For this same reason, Martens stated, “all chemistry is femtochemistry [5]". The ephemeral transition states, denoted in the past by a bracket [TS]* for th clusiveness, can now be clocked as a molecular species TS!. Moreover, on this time scale, the time-dependent description of a coherent, single-molecule tra- jectory represents the classical nuclear “motion picture” of the reaction as its wave packet proceeds from the . and on to final products ~ the language of the actual dynamics! The fs time scale is unique for the creation of such coherent matter waves on. the atomic scale of length, a basic problem rooted in the development of quantum mechanics and the duality of matter. Figure (4) highlights some steps made in the de- scription of the duality of lightmatter and time scales, and the Appendix cusses the importance of coherence for localization, the actual nuclear tions reaching the end of the road”. nitial state, through transition statesAhmed H. Zewail 281 Matter Waves Particle-type Control & Dynamics de Broglie (1924) Binstein’s light wave/particle E=hv E=cp i=hlp Similarly, matter particlevwave Schrodinger (1926) ‘The Wave Equation ~ Stationary waves HY =EY Schrodinger (1926) Micro- to Macro-mechanies ‘Quantum to Newton Mechanics ¥ to wave group Femtochemistry & Quantum Limit (i): Partiele-type ge Bogie (initial localization) = h/p uncertainty in time measurement AxAp2WGn) ALAE>h/Qn) for force free Ax=pim Ai At At~10fs Ax~O1A At~10 ps Ax~100A Figure 4. Matter waves and particle-type limit of dynamics. The atomicscale de Broglie wave- length in the coherent preparation of a quantum system and in the uncertainty of probe meas rements are both reached on the fe time scale. Sec also the Appendix and its Figure. Figure 5. A femtochemistry apparatus typical of early Femtolands. (A, above) The laser system. (top) the first CPM oscillator used in Femtoland I; (bottom) the continuum generation to the right and the experimensal layout for clocking to the left. (B, next page) The molecular beam apparatus of Femtoland IIT, together with a view of the beam/laser arrangement, (Ref. B1, B16, B28, 49]282 Chemistry 1999 Molecular beam Femtosecond Mice Probing Pulse This powerful concept of coherence lies at the core of femtochemistry and was a key advance in observing the dynamics at atomic resolution. The realiz- ation of its importance and its detection by selectivity in both preparation and probing were essential in all studies, initially of states and orientations, and culminating in atomic motions in reactions. With these concepts in mind, the marriage of ultrafast lasers with molecular beams [Fig. 5A, B] proved to be es- sential for the initial development. Laser-induced fluorescence was the first probe used, but later we invoked mass spectrometry and nonlinear optical techniques. Now numerous methods of probing are known [Fig. 6] and used in laboratories around the world; Coulomb explosion is the most recent po- werful probe developed by Will Castleman [5] for arresting reactive interme- diates, as mentioned in Section IIL Applications of femtochemistry have spanned the different types of chemi- cal bonds ~ covalent, ionic, dative and metallic, and the weaker ones, hydro- gen and van der Waals bonds. The studies have continued to address the vary- ing complexity of molecular systems, from diatomics to proteins and DNA.Ahmed H, Zewail 283 Laser-induced Fluorescence Infrared Pumping Laser Multiphoton Ionization Femtochemistry ‘Mass Spectrometry Stimulated Emission Pumping) | Frequency Ms ‘Time, Velocity & Angular Resolutions ‘Transient Absorption Velocity | Orientation Maltiple Pulses a Photoelectron Spectroscopy I (Came, ZEKE, Coincidence) Coulomb Explosion Diffraction ? Figure 6, Techniques for probing in femtochemistry. Both excited and ground states have been probed by these methods. The correlations of time with frequency, mass, velocity and orientation were essential in the studies of complex systems. Diffraction represents the new effort for prob- ing structures (see text and Fig. 29). Studies have also been made in all phases of matter: gases and molecular be- ams; mesoscopic phases of clusters, nanostructures, particles and droplets; condensed phases of dense fluids, liquids, solids, surfaces and interfaces; and in sibling fields of femtoscience such as femtobiology [Fig. 7]. Twenty-four centuries ago, the Greek philosopher Democritus and his teacher Leucippus gave birth to a new way of thinking about matter’s invisible and elementary entity, the atom. Richard Feynman once asked, if you had on- ly one sentence to describe the most important scientific knowledge we pos- sess, what would that sentence be? He said, “everything is made of atoms.” Democritus’ atomism, which was rejected by Aristotle, was born on a purely philosophical basis, surely without anticipating some of the twentieth centu- ry's most triumphant scientific discoveries. Atoms can now be seen, observed in motion, and manipulated [6]. These discoveries have brought the micro- scopic world and its language into a new age, and they cover domains of length, time and number. The length (spatial) resolution, down to the scale of atomic distance (angstrém), and the time resolution, down to the scale of atomic motion (femtosecond), have been achieved. The trapping and spec- troscopy of a single ion (electron) and the trapping and cooling of neutral atoms have also been achieved. All of these achievements have been recogniz- ed by the awarding of the Nobel Prize to STM (1986), to single-electron and -ion trapping and spectroscopy (1989), to laser trapping and cooling (1997), and to laser femtochemistry (1999).284 Chemistry 1999 Femtochemistry ‘Structures UED & X-rays Femtobiology Elementary Phenomena: Transition States, Energy Redistributions (IVR) & Rates (k(E)) Chemical Bond: Covalent, Ionic, Dative, Metallic, Hydrogen & van der Waals, Molecular Size: ‘Two Atoms to DNA & Proteins Molecular State: Ground & Excited, Neutral & lon (negative or positive) Figure 7. Femtochemistry and the scope of applications. IL DYNAMICS AND ARROW OF TIME (1) Origins ~ From Kinetics to Dynamics The Arrhenius Seminal Contribution At the turn of the 20th century, the study of reactivity was dominated by the question’: How do reactions proceed and what are their kinetic rates? Svante Arrhenius [7] gave the seminal description of the change in rates of chemical “The main focus of prior studies was on the thermodynamics and equilibrium characteristics. Interestingly, chemical equilibrium was already established a century before (1798) by Claude Berthollet, during a visit to the Natron Lakes in Napoleonic times, through his studies of the reaction Na,CO, + CaCl, + 2NaCl + CaCO, [S. W. Weller, Bull. Hist, Chem. 24, 61 (1999)].Ahmed H. Zewail 285 reactions with temperature and formulated in 1889 the familiar expression for the rate constant, k=A exp (E,/RT) a) which, as Arrhenius acknowledged, had its roots in van’t Hoff’s (1884) equa- tions [7]. For any two reactants, the rate constant (k) depends on tempera- ture (T) according to an energy of activation (E,), which is different from the thermodynamic net energy change between reactants and products, and the dependence is exponential in form. (R is the universal gas constant). If we think of the reaction as a finite probability of reactant A colliding with B, then the rate is simply the collision frequency times the fraction of successful colli- sions with an energy equal to or more than E,. Besides the value of equation (1) through the well-known plots of “In k vs. 1/T” to obtain the energy of ac- tivation E,, Arrhenius introduced a “hypothetical body’, now known as the “activated complex”, a central concept in the theory of reaction rates: the reaction, because of collisions or other means, proceeds only if the energy is, sufficient to exceed a barrier whose energy is defined by the nature of the complex. Since then, various experimental data for different temperatures T were treated with equation (1), yielding E, and the pre-exponential factor A. A few years after Arrhenius’ contribution, Bodenstein (1894) (7] published a landmark paper on the hydrogen/iodine system, which has played an im- portant role in the development of gas-phase chemical kinetics, with the aim of understanding elementary reaction mechanisms. In the twenties, Linde- mann (1922), Hinshelwood (1926), Tolman (1920), and others [7&8] devel- oped, for unimolecular gas-phase reactions, elementary mechanisms with dif ferent steps describing activation, energy redistribution, and chemical rates. By 1928, the Rice-Ramsperger-Kassel (RRK) theory was formulated, and Marcus, starting in 1952, blended RRK and transition state theory in a direc- tion which brought into focus the nature of the initial and transition-state vi- brations in what is now known as the RRKM theory [8]. The rate constant, k(T), does not provide a detailed molecular picture of the reaction. This is because k(T), which was obtained from an analogy with van't Hoff’s description of the change with T of the equilibrium constant K (thermodynamics), is an average of the microscopic, reagent-state to product- state rate coefficients over all possible encounters. These might include dif- ferent relative velocities, mutual orientations, vibrational and rotational phases, and impact parameters. A new way was needed to describe, by some quantitative measure, the process of the chemical reaction itself: How reagent molecules approach, collide, exchange energy, sometimes break bonds and make new ones, and finally separate into products. Such a description is the goal of molecular reaction dynamics [9]- The London, Eyring and Polanyi Contributions For some time, theory was ahead of experiment in studies of microscopic mo- lecular reaction dynamics. The effort started shortly after the publication of the Heitler-London quantum-mechanical treatment (1927) of the hydrogen286 Chemistry 1999 molecule [10], a breakthrough in thinking not only about the stable structure of the chemical bond, but also about how two atoms can interact at different separations. One year later (1928), for Sommerfeld’s Festschrift (60th birth- day), London [10] presented an approximate expression for the potential energy of triatomic systems, e.g., H, in terms of the coulombic and exchange energies of the “diatomic” pairs, In’ 1931 Henry Eyring and Michael Polanyi [10], using the London equation, provided a semiempirical calculation of a potential energy surface (PES) of the H + Hy reaction describing the journey of nuclei from the reactant state of the system to the product state, passing through the crucial transition state of activated complexes. The birth of “reaction dynamics” resulted from this pioneering effort and, for the first time, one could think of the PES and the trajectories of dynamics on it ~ in those days, often, expressed in atomic units of time! But no one could have dreamed in the 1930's of observing the transient molecular structures of a chemical reaction, since the time scale for those far from equilibrium activated complexes in the transition state was estimated to be less than a picosecond (ps). The time scale was rooted in the theory developed for the description of re- action rates. Building on Arrhenius’ work and the work of Polanyi and Wigner (1928) [11], in 1935, Eyring, and independently Evans and Polanyi, formulated transition-state theory, which gave an explicit expression for Arrhenius’ pre-exponential factor [11]: a k KT gs KD. Q h h 0,05 where k is Boltzmann's constant, h is Planck’s constant and Q is the partition function. The key idea here is to assume the equilibration of the population between reactants and the transition state: A + B «> [TS]? > products. Thus, the rate constant can be related to the equilibrium constant for formation of the transition state, K*, and hence AG?, AH? and ASt; physically, AGt (AH*-TAS*) in the exponent gives the barrier energy (through AH?) along the reaction coordinate, and the pre-exponential entropic term which re- flects the change in vibrational modes perpendicular to the reaction coordi- nate. By comparing (2) with Arrhenius’ equation (1), A can be identified; E,, the activation energy, and E,, the barrier energy, with zero-point energy cor- rections, are related [8, Laidler]. Kramers’ (1940) classic work [11] modified the pre-exponential factor to include friction from the surrounding medium, with transition state theory giving an upper limit, and Casey Hynes, in the 1980's, provided a dynamical theory of friction with emphasis on time scales in the transition-state region and for solvent interaction [11]. In the 1950’s, Marcus [11] obtained a transition-state type expression for reactions of elec- tron transfer in solutions with the Gibbs free energy of activation expressing the dependence on solvent reorganization energies. This work was awarded the 1992 Nobel Prize. According to transition state theory, the fastest reaction is given by KT/h, which is basically the “frequency” for the passage through the transition state exp(—Ey/kT) (2)Ahmed H. Zewail 287 [Equation 2]. At room temperature this value is 6 x 102 second", correspond- ing to ~ 170 fs; the time scale of molecular vibrations is typically 10-100 fs. This estimate is consistent with knowledge of the speeds of nuclei and the distance change involved in the reaction. In 1936 the first classical trajectory from Hirschfelder-Eyring-Topley molecular dynamics simulations of the H+H, reaction showed the fS steps needed to follow the reaction profile, albeit on the wrong PES. Later, Karplus, Bunker and others showed a range for the time scales, ps to fs, depending on the reaction and using more realistic PES’s [see 9 and 26]. Transition State and Definition In general, for an elementary reaction of the type, A+BC> [ABC]* > AB+C (3) the whole journey from reagents to products involves changes in internuclear separation totaling ~10A. If the atoms moved at 10*-10° cm/sec then the en- tire 10A trip would take 1012-10 sec. If the ‘transition state’, [ABCI}, is de- fined to encompass all configurations of ABC significantly perturbed from the po- lentialenergy of the reagents A + BC or the products AB + C, then this period of 1-10 ps is the time available for its observation. To achieve a resolution of ~0.1A, the probe time window must be 10-100 fs. The above definition of the transition state follows the general description given by John Polanyi and the author [12], namely the full family of configur- ations through which the reacting particles evolve en route from reagents to products, This description may seem broad to those accustomed to seeing the TS symbol, t, displayed at the crest of the energy barrier to a reaction. As stated in Ref. 12, even if one restricts one’s interest to the over-all rates of che- mical reactions, one requires a knowledge of the family of intermediates sampled by reagent collisions of different collision energy, angle and impact parameters. The variational theory of reaction rates further extends the range of TS of interest, quantum considerations extend the range yet further, and the concern with rates to yield products in specified quantum states and angles extends the requirements most of all. A definition of the TS that em- braces the entire process of bond-breaking and bond-making is therefore likely to prove the most enduring. This is specially important as we address the energy landscape of the reaction, as discussed in Section IV. The cardinal choice of the transition state at the saddle point, of course, has its origin in chemical kinetics — calculation of rates - but it should be re- membered that this is a mathematical ‘single-point’ with the division made to define the speed of a reaction [8]. Even in thermal reactions, there is enough of an energy distribution to ensure many types of trajectories, Furthermore, transition state theory is not a quantum theory, but a classical one, because of the assertion of a deterministic point. Quantum uncertainty demands some de- localization [8, Hase] and, as mentioned above, the theory invokes the equi- libration of reactants and transition state populations, using statistical ther- modynamics. In fact, the term transition state is used ambiguously to refer288 Chemistry 1999 both to the quasi-equilibrium state of the reaction and to the molecular struc. ture of the saddle point. As discussed in Ref. 13 [Williams, Doering, Baldwin], the molecular species at the saddle point perhaps could be referred to as the transition structure; the activated complex is even more descriptive. The location at the saddle point provides the highest energy that must be reached, defining the exponential probability factor; the dynamics (forces and time scales) are governed by the nature of the TS region. Provided that the energy landscape of the reaction is controlled by a narrow region, the structure of the transition state becomes important in structure-reactivity correlations; rates vs. Gibbs energy between the TS and the ground state. (It is also useful in designing TS analogs as enzyme inhibitors and TS comple- ments as catalysts [13]). The position of the TS and its energy relative to that of reactants and products along the reaction path becomes relevant. In the mid 1930's, the Bell-Evans-Polanyi principle gave a predictive correlation between changes in barrier heights and the enthalpies of reactions, especially for a series of related reactions; the TS for an exoergic or an en- doergic reaction is very different. In 1955, this led to Hammond’s postulate which characterizes a reactantlike TS (so called “early” TS) for exoergic reac- tions, productlike TS (so called “late” TS) for endoergic reactions, and “cen- tral” TS for energy neutral reactions [8, Shaik]. John Polanyi, the son of Michael, formulated some concepts regarding energy disposal in relation to the position of the TS on the PES [14], using molecular dynamics simulations and experimental studies of chemiluminescence. It should be recognized that selectivities, efficiencies, and stereochemistries are quantified only when the dynamics on the global energy landscape are understood. For example, the picture for simple reactions of a few (strong) bonds being made and broken is changed when the energy surface is nearly flat or there is a significant entropic contribution. Many transition states will exist in the region as in the case of protein folding (13, Fersht]. Finally, in the transition state, chemical bonds are in the process of being grade and broken. In contrast, for intermediates, whose bonds are fully formed, they are in po- tential wells, typically “troughs” in the TS region. However, the time scale is crucial. In many cases, the residence time in intermediates approaches the fs regime characteristic of TS structures and the distinction becomes fuzzy. Moreover, for a real multidimensional PES, the non-reactive nuclear motions can entropically lock the system even though there is no well in the energy landscape. And the presence of shallow wells in the energy landscape does not guarantee that trajectories will visit such wells. Transition State and Spectroscopy Various techniques have been advanced to probe transition states more di- rectly, especially for elementary reactions. Polanyi's analogy [14] of transition- state spectroscopy, from “spectral wing emission”, to (Lorentz) collisional line broadening studies, made earlier by A. Gallagher and others [see 12], set the stage for the use of CW spectroscopic methods as a probe. (In this way, only about one part in a million of the population can be detected.) Emission, ab-Ahmed H. Zewail 289 sorption, scattering and electron photodetachment are some of the novel methods presented for such time-integrated spectroscopies, and the groups of Jim Kinsey, Philip Brooks and Bob Curl, Benoit Soep and Curt Wittig, Dan Neumark, and others, have made important contributions to this area of re- search. The key idea was to obtain, as Kinsey [15] puts it, short-time dynamics _from long-time experiments, (The renaissance of wave packet dynamics in spec- troscopy, pioneered by Rick Heller, will be highlighted in Section 114 and TN6). With these spectroscopies in a CW mode, a distribution of spectral fre- quencies provides the clue to the desired information regarding the distribu- tion of the TS over successive configurations and potential energies. Recently, this subject has been reviewed by Polanyi and the author and details of these contributions are given therein [12], and also in [26] (2) Arrow of Time In over a century of development, time resolution in chemistry and biology has witnessed major strides, which are highlighted in Fig. 8 [16]. As men- tioned before, the Arrhenius equation (1889) for the speed of a chemical reaction gave information about the time scale of rates, and Eyring and Michael Polanyi’s (1931) microscopic theoretical description made chemists think of the atomic motions through the transition state and on the vibratio- nal time scale. But the focus naturally had to be on what could be measured in those days, namely the slow rates of reactions. Systematic studies of reac- tion velocities were hardly undertaken before the middle of the 19th century; in 1850 Ludwig Wilhelmy reported the first quantitative rate measurement, the hydrolysis ofa solution of sucrose to glucose and fructose [8, Laidler]. In 1901, the first Nobel Prize for chemistry was awarded to van't Hoff for, among other contributions, the theoretical expressions (chemical dynamics) which were precursors to the important work of Arrhenius on rates. Arrhenius too received the Prize in 1903 for his work on electrolytic theory of dissociation. ‘A major advance in experiments involving sub-second time resolution was made with flow tubes in 1923 by H. Hartridge and F. J. W. Roughton for solu- tion reactions. Two reactants were mixed in a flow tube, and the reaction products were observed at different distances. Knowing the speed of the flow, one could translate this into time, on a scale of tens of milliseconds. Such measurements of non-radiative processes were a real advance in view of the fact that they were probing the “invisible”, in contrast with radiative glows which were seen by the naked eye and measured using phosphoroscopes. Then came the stopped-flow method (B. Chance, 1940) that reached the milli- second scale. The stopped-flow method is still used today in biological kinetics. Around 1950, a stride forward for time resolution in chemistry came about when Manfred Eigen in Germany and R. G. W. Norrish and George Porter in England developed techniques reaching the microsecond time scale [17]. For this contribution, Eigen and Norrish & Porter shared the 1967 Nobel Prize. The method of flash photolysis was developed by Norrish and Porter a few years after World War Il, using electronics developed at the time. They pro-290 Chemistry 1999 ARROW OF TIME Chemistry & Biology Vee LOE Felacetion Mysthod Flash Photons Stogped Flew Flow Metiod Kingdes Phuiens Figure 8. Arrow of Time in chemistry and biology, some steps over a century of development (see text). (Ref. BI] duced an intense burst of light and created radicals in the sample, and, using other light, they recorded the spectra of these radicals. They achieved kine- on this time scale and observed some relatively stable intermediates. Before the turn of the 20th century, it was known that electrical sparks and Kerr cell shutters could have response times as short as ten nanoseconds. In an ingenious experiment, Abraham and Lemoine (1899) [18] in France de- monstrated that the Kerr response of carbon disulfide was faster than ten na- noseconds; it has now been measured to be about two picoseconds (with femtosecond response) [18]. They used an electrical pulse which produced a spark and simultaneously activated a Kerr shutter. Light from the spark was collimated through a variable-delay path and through the Kerr cell (polarizer, tiAhmed H. Zewail 291 CS, cell and analyzer). The rotation of the analyzer indicated the presence of birefringence in the cell for short optical delays; this birefringence disappear- ed for pathlengths greater than several meters, reflecting the total op- tical/electrical response time of 2.5 ns. In this landmark “pump-probe” expe- riment, they demonstrated in 1899 the importance of synchronization. The setting of time delays was achieved by varying the light path. Bloembergen [18] has recently given a historical perspective of short-pulse generation and Shapiro has reviewed the early developments, including mechanical, streak, spark, stroboscope, and other high-speed photography methods [see Refs. 2 & 18]. As pointed out in these references [18], flash photolysis utilized the above approach but one of the flashes was made very strong to generate high concentrations of free radicals and hence their utility in chemical and spec- troscopic applications. Eigen developed “the relaxation method”, which reached the microsecond and close to the nanosecond scale. By disturbing the equilibrium of a solu- tion by either a heat jump, a pressure jump, or an electric field, the system shifts from equilibrium. This is the point of time zero. Then the system equi- librates, and its kinetics can be followed. (At about the same time, shock-tube methods were used to provide kinetics on similar time scales.) Eigen called these reactions "immeasurably fast” in his Nobel lecture. There was a feeling that this time resolution was the fastest that could be measured or that need- ed to be measured for relevance to chemistry (Section IV). The invention of the laser has changed the picture. Shortly after the realization of the first (ruby) laser by Maiman (1960), the generation of giant and short pulses became possible: nanoseconds by Q- switching (Hellwarth, 1961) and picoseconds (De Maria, et al 1966) by mode- locking (1964). Sub-picosecond pulses from dye lasers (Schafer & Sorokin, 1966) were obtained in 1974 by Chuck Shank and Eric Ippen at Bell Labs, and in 1987 a six £8 pulse was achieved [19]. In 1991, with the discovery of femtosecond pulse generation from solidstate Ti-sapphire lasers by Sibbett and colleagues [19], dye lasers were rapidly replaced and femtosecond pulse generation became a standard laboratory tool; the state-of-the-art [19], once 8 fs, is currently about 4 fs and made it into the Guinness Book of World Records (Douwe Wiersma’s group [19]). The tunability is mastered using continuum generation [19, Alfano & Shapiro] and optical parametric ampli- fication. In the late sixties and in the seventies, ps resolution made it possible to study non-radiative processes, a major detour from the studies of conventional radiative processes to infer the non-radiative ones. As a beginning student, 1 recall the exciting reports of the photophysical rates of internal conversion and biological studies by Peter Rentzepis [20]; the first ps study of chemical reactions (and orientational relaxations) in solutions by Ken Eisensthal [21]; the direct measurement of the rates of intersystem crossing by Robin Hochstrasser [22]; and the novel approach for measurement of ps vibrational relaxations (in the ground state of molecules) in liquids by Wolfgang Kaiser and colleagues [23]. The groups of Shank and Ippen [19] have made impor-292 Chemistry 1999 tant contributions to the development of dye lasers and their applications in the ps and into the fs regime. Other studies of chemical and biological non- radiative processes followed on the ps time scale, the scale coined by G. N. Lewis as the “jiffy” - the time needed for a photon to travel 1 cm, or 33 pico- seconds [24]. At about the same time in the sixties, molecular-beam studies of reactions were being developed, and, although I was not initially a member of this com- munity, beams later became part of our effort in femtochemistry. Molecular collisions occur on a shorter time scale than a ps and real time studies were not possible at the time. Crossed molecular beams and chemiluminescence techniques provided new approaches for examining the dynamics of single collisions using the postattributes of the event, the reaction products. The contributions by Dudley Herschbach, Yuan Lee and John Polanyi [14, 25] were acknowledged by the 1986 Nobel Prize. From state and angular distri- butions of products, information about the dynamics of the collision was deduced and compared with theoretical calculations of the PES and with mo- lecular dynamics simulations; the goal was to find self-consistency and to deduce an estimate of the lifetime of the collision complex. Crossed molecu- lar beam-laser studies have probed dynamics via careful analyses of product internal energy (vibrational and rotational) distributions and steady-state alignment and orientation of products. The contributions to this important area are highlighted in the article by Dick Zare and Dick Bernstein [25] and in the book by Raphy Levine and Bernstein [9]. An overview of femtoche- mistry (as of 1988) in connection with these other areas is given in a feature article [26] by Zewail and Bernstein. I. FEMTOCHEMISTRY: DEVELOPMENT OF THE FIELD The development of the field is highlighted in this section, from the early years of studying coherence to the birth of femtochemistry and the explosion of research, or, as the report of Ref. 5 puts it, “... the revolution in chemistry and adjacent sciences.” On the way, there were conceptual and experimental prob- lems to overcome and many members of our Caltech group have made the successful evolution possible. The review article published in the Journal of Physical Chemistry [B14] names their contributions in the early stages of development. Here, references are given to the work and explicitly in the Figure Captions. (1) The Early Years of Coherence When I arrived in the US as a graduate student in 1969, nine years after the invention of the first laser, I had no idea of what lasers were about. When ap- pointed to the Caltech faculty as an assistant professor in 1976, I was not thinking or dreaming of femtosecond time resolution. But I had the idea of exploring coherence as a new concept in dynamics, intra-and inter-molecular. This proved to be vital and fruitful.Ahmed H. Zewail 293 New Techniques for Molecules The Caltech offer included start-up funds of $50,000 for capital equipment ($18K for shop services), an empty laboratory of two rooms and an office next to it. A few months before moving to Caltech in May of 1976, I made the decision not to begin with the type of picosecond research I was doing at Berkeley as a postdoctoral fellow. Instead, the initial effort was focused on two directions: (i) studies of coherence in disordered solids, and (ii) the devel- opment of a new laser program for the studies of the phenomena of (optical) coherence. Prior to the final move, I came down from Berkeley for several visits in order to purchase the equipment and to outline the laboratories’ needs for electricity, water, gases and so on; my feeling was that of a man left out in a desert with the challenge to make it fertile. We spent a significant fraction of the $50K on setting up the apparatus for optical detection of mag- netic resonance (ODMR) to study disordered solids. My experience at Penn with Professor Robin Hochstrasser and at Berkeley with Professor Charles Harris was to culminate in these experiments. The key questions I had in mind were: What is the nature of energy migration when a crystal is systema- tically disordered? Is it coherent, incoherent or partially coherent? Is there a relationship between optical and spin coherence? While the ODMR apparatus was being built, I was thinking intensely about new laser techniques to probe the coherence of optical transitions (so-called optical coherence). I had the intuitive feeling that this area was rich and at the time had in mind several issues which were outlined in my research pro- posal to Caltech. Laser experiments were designed with objectives focused on the same issues outlined in the proposal. First, in the work I published with Charles Harris, and alone at Berkeley, the coherence probed had been that of spin (triplet excitons) and I felt that we should directly probe the optical coherence, i.e., the coherence between the excited electronic state and the ground state, not that between two spin states of the same excited state [Fig. 9]. Second, I did not believe that the time scale of spin coherence was the same as that of optical coherence. Later, I wrote a paper on the subject which was published in the Journal of Chemical Physics (1979) with the title: Are the homogeneous linewidths of spin resonance and optical transitions related? I was convinced that essentially all molecular optical transitions in solids are inhomogencous, that is, they do not reflect the true dynamics of a homo- geneous ensemble, but rather the overlapping effects of sub-ensembles. This was the key point outlined in my research proposal to Caltech. If the new set of laser experiments at Caltech proved successful, we should be able to find the answer to these important (to me) questions. However, the funds re- maining were insufficient to realize our dream for the new experiments. Fortunately, Spectra Physics, started in 1961 and now a huge laser company, was proud of a new product and was interested in helping us demonstrate the usefulness of one of the first single-mode dye lasers they produced, At Caltech, we had the pump argon-ion laser, from Wilse Robinson's Laboratory, and added the dye laser we obtained from Spectra Physics with the idea that we would purchase it if the experiments were successful. David Evans of204 Chemistry 1999 ‘Coherent Transients — phase control [EXACT SUPERPOSITION #leate ge Lore pete) + ott) ] LASER Lisa 65 5 7 3 7 Tine (8) Figure 9. Molecular coherence and dephasing. Coherent transients from the superposition of states, and control of pulse phase (x or 3) in multiple pulse experiments, the optical analog of NMR spectroscopy. The photon echo of iodine gas was observed on the spontaneous emission using the described pulse sequence. [Ref. B9, B13, B22, B23, 50] Spectra Physics was instrumental in helping us achieve this goal. What was left were the low-temperature cryostat and electronics. We could not afford a real metal cryostat, so we custom-made a glass Dewar that could go down to 1.8 Kelvin. Most of the electronics were obtained on loan for several months. To generate laser pulses we used switching methods developed at JILA (Joint Institute for Laboratory Astrophysics) and IBM. The work at IBM by Dick Brewer's group triggered our interest in using electro-optic switching methods. We succeeded in making the laser perform according to specifications and began the first experiments with phenazine crystals. This system, phenazine, has unique properties which I learned about in Robin Hochstrasser’s labora- tory. At Berkeley I published a paper outlining the nature of coherence in multidimensional systems and used phenazine as a prototype experimentalAhmed H. Zewail 295 system. Our first laser transient at Caltech was beautiful. Unfortunately, the transient was from the electronics, not from the crystal, but we soon realized this! We decided to abandon this particular system for a while and to try an impurity crystal of pentacene in a host of terphenyl. We also decided to study gases, and some success came our way. With the theoretical knowledge acquired in handling coherence effects, which requires expertise with density matrix formalism and its manipulation in geometrical frames, I had a novel idea: we should be able to detect coheren- ce on the incoherent emission at optical frequencies. From many discussions with Alex Pines and Charles Harris at Berkeley, I knew the power of “adding” pulses in NMR and ESR. This idea was successful and indeed we were able to observe the photon echo on the spontancous emission [Fig. 9] using three optical pulses. Only months after my arrival in May of 1976, we published our first scientific paper from Caltech. This success gave us confidence in the ap- proach and in our understanding of the principles of coherence and its prob- ing in molecular systems. We applied it to larger molecules with success but also encountered some disappointments. This work was followed by a variety of extensions to studies in gases and solids and also in a home-made (from glass) effusive molecular beam. The small group in our laboratory of optical coherence became productive and we had an exciting time. Our group and that of Douwe Wiersma in Holland were then the two most active in these areas of chemical research. In the meantime, the work on disorder in solids began to yield interesting results and, surprisingly, we observed an unusual change in the degree of energy transfer with concentration, which we published as evidence of Ander- son localization, a hot topic in the 1970's. This was followed by detailed studies of several systems and I wrote my first research proposal to the National Science Foundation (NSF). The research was funded! The then pro- gram director at NSF, Fred Stafford, was supportive of the effort and we have maintained our support from the Division of Materials Research to this day; the NSF’s Chemistry Division continues to support our research. Knowledge of energy transfer was also helpful in another area. Terry Cole, a visiting scholar at Caltech, and I initiated work on the studies of Luminescent Solar Concentrators (LSC) using energy transfer between dyes as a key principle This idea, too, was successful and funding for this research came in from SERI and from ARCO. The work on LSC resulted, ultimately, in a patent (with Sam Batchelder; issued in 1980) and in several publications. My re- search group was rapidly expanding. After a year-and-a-half at Caltech, I was pleased to learn that my colleagues were considering my case for tenure. Tenure was granted a few months later, two years after my arrival at the Institute, I was both appreciative and pleased. We continued research in four areas: optical coherence phenomena and dephasing; disorder in solids; picosecond spectroscopy; and LSC. In my own department, some colleagues were not too excited about ‘this stuff of cohe- rence and dephasing’ ~ thinking that it was not relevant to chemistry! Many che- mists on the outside were also unsure what this was all about. In fact, a notab-296 Chemistry 1999 le chemist once said publicly at a conference I attended that coherence and dephasing had nothing to do with chemistry! On the physics side, I was in- vited to numerous conferences, including one in which the Nobel Laureate Willis Lamb asked me to have dinner to discuss our research. This was a spe- cial experience! I was not convinced by these doubts and my faith helped us to continue along with the development. The concept of coherence turned out to be fun- damental in femtochemistry, and it is now well accepted that coherence is a key process in the probing and controlling of molecular dynamics. With the success we had with observations and studies of coherence in different systems, I wrote an Account of Chemical Research article, published in 1980, with the title: Optical Molecular Dephasing — Principles of and Probings by Coherent Laser Spectroscopy. 1 felt that the nanosecond time scale we had mastered should be extended to the picosecond time scale, but did not wish to repeat the Berkeley experience with glass lasers. Fortunately, the design for the first sub-picosecond dye laser was reported in 1974 and we decided to build one to study the phenomena of coherence ~ but now on a shorter time scale. Optical Analogue of NMR: Controlling the Phase From the studies of optical transients, we learned that coherence can be probed directly in real time in gases (and solids) and that incoherent decay (eg. fluorescence) can be used to monitor such coherences provided that the laser pulse(s) is capable of forming a coherent superposition of states. For two states of a transition (say y,, and y,), the coherent state can be written as Peoherent () =a) ¥, + (DY, (4) where the coefficients, a(t) and b(t) contain in them the familiar quantum- mechanical phase factors, exp(iE,t/h) and exp (iE,t/f), respectively [see Fig. 9]. With pulse sequences, we could directly monitor the behavior of the en- semble-averaged coefficients of Y -'P*, (a(t) b*(t)), which contain informa- tion on the coherence decay time (optical T,); they are the off-diagonal ele- ments of a density matrix, p, The term (a(t) a*(t)) is the population of state y,, and represents the diagonal density-matrix element, p,,; (a(t) a*(t)) de- cays with optical T,. We were thus able to demonstrate the power of the opti- cal analogue of NMR pulse techniques in learning about coherence and the origin of optical dephasing in molecular systems of interest to chemical dy- namics. This advance changed the thinking of many with the recognition that it was impossible to deduce T, and T, from measurements of the line width of inhomogeneous transitions. One feature of this work which later helped us in the study of molecular reaction dynamics was the realization of the importance of the pulse phase (shape) in studies of coherence. With the acousto-optic modulation techni- ques we developed earlier, it became possible to make optical pulse sequen- ces with well-defined phases. This development took us into the domain of se- lective and prescribed pulse sequences which could then be used to enhanceAhmed H. Zewail 297 coherences or suppress them ~ the optical analogue of NMR multiple pulse spec- troscopy. We published several papers on phase control and extended the ap- plications to include photon locking. We were eager to extend these techni- ques to the picosecond time domain in order to study solids, but, for several reasons, our attention was diverted to gas-phase molecular dynamics. By this time, our group's efforts were narrowing on two major areas. (Dick Bernstein, who was on the Visiting Committee for our Division, hinted that T was doing too much in too many areas!) The work on disorder and LSC was gradually brought to completion. Picosecond spectroscopy of rotational dif fusion and energy transfer in liquids were similarly handled. I felt that the latter area of research was too crowded with too many scientists, a characte- ristic I do not enjoy when venturing into a new area. I must add that I was not too thrilled by the exponential (or near exponential) decays we were measur- ing and by the lack of molecular information. Our effort began to emphasize two directions: (i) the studies of coherence and dynamics of isolated molecu- les in supersonic beams and (ii) the development of the optical analogue of NMR spectroscopy. The low-temperature facility was put to use to study the dephasing and polarization of highly vibrationally-excited molecules in the ground state. Coherence in chemical dynamics was occupying my thinking, and I made a detour in the applications that turned out to be significant. (2) The Marriage with Molecular Beams The Bell Labs design for the dye laser (passively mode-locked, CW, and cavi- ty dumped) was too restrictive for our use and, even though we published several papers on studies in the condensed phase with 0.6 ps resolution, we decided to change to a new system. The synchronously pumped mode-locked (CW) dye laser allowed for tunability and also for photon-counting detection techniques. The power of single-photon-counting became apparent, and a new laser system, a synchronously pumped, cavity dumped dye laser was con- structed for studies of gas-phase molecular dynamics, but now with the bene- fit of all the expertise we had gained from building the first system used for studies in the condensed phase and for probing the torsional of DNA. Stimulated by the work on coherence, and now with the availability of pi- cosecond pulses, I thought of an interesting problem relating to the question of intramolecular vs. intermolecular dephasing. In large, isolated molecules (as opposed to diatomics), there are the so-called heat bath modes which can be a sink for the energy. The question arose: Could these bath modes in isolated large molecules dephase the optically-excited initial state in the same way that phonons of a crystal (or collisions in gases) do? This problem has some roots in the question of state preparation, and I was familiar with its re- lationship to the description of radiationless transitions through the work of Joshua Jortner, Wilse Robinson and others. There was much theoretical ac- tivity about dephasing, but I felt that they were standard extensions and they did not allow for surprises. We decided on a new direction for the studies of coherence in a supersonic molecular beam.298 Chemistry 1999 Rick Smalley came to Caltech in May of 1980 and gave a talk entitled “Vibrational Relaxation in Jet-Cooled Polyatomics”, He spoke about his ex- citing work on the naphthalene spectra. From the line width in the excitation spectra, he inferred the “relaxation time”, At the time, the work by Don Levy, Lennard Wharton, and Smalley on GW (or nanosecond) laser excitation of molecules in supersonic jets was providing new ways to examine the spectros- copy of molecules and van der Waals complexes. Listening to Rick and being biased by the idea of coherence, I became convinced that the way to monitor the homogeneous dynamics was not through the apparent width but by using coherent laser techniques. This was further kindled by the need for direct measurement of energy redistribution rates; we were encouraged by Charlie Parmenter, after he had reported on a chemical timing method using colli- sions as a “clock” to infer the rate of energy redistribution. The first “real” su- personic molecular beam was huge. We did not know much about this kind of technology. However, in a relatively short time, it was designed and built from scratch, thanks to the effort of one graduate student who must have con- sumed kilos of coffee! The molecular beam and picosecond system were in- terfaced with the nontrivial addition of a spectrometer to resolve fluorescen- ce in frequency and time. This was crucial to much of the work to come. The Anthracene Discovery: A Paradigm Shift Our goal in the beginning was to directly measure the rate of IVR (Intramo- lecular Vibrational-cnergy Redistribution), expecting to see a decrease with time (exponential decay) in the population of the initiallyexcited vibrational state and to possibly see a rise in population in the state after the redistribu- tion, thus obtaining T, directly. What we saw in these large systems was con- trary to the popular wisdom and unexpected. The population during IVR was oscillating coherently back and forth [Fig. 10] with well-defined period(s) and phases! We were very excited because the results revealed the significan- ce of coherence at its best in a complex molecular system, now isolated in a molecular beam, with many degrees of freedom. I knew this would receive at- tention and skepticism. We had to be thorough in our experimental tests of the observation and three of us went to the laboratory to see how robust the observation was. We published a Communication in the Journal of Chemical Physics (1981). Earlier there had been attempts by another group to observe such a “quantum coherence effect” in large molecules, but the observation turned out to be due to an artifact. Some scientists in the field were skeptical of our new observation and the theorists argued that the molecule is too big to see such quantum coherence effects among the vibrational states. Further- more, it was argued that rotational effects should wash out such an observa- tion. We followed the initial publication with several others and the effect be- came even more pronounced with shorter time resolution. Physicists appre- ciated the new results, We published a Physics Review Letter on the nature of non-chaotic motion in isolated systems, and Nico Bloembergen and I wrote a review (1984) on the relevance to laser-selective chemistry [B21]. We andAhmed H. Zewail 299 other groups subsequently showed the prevalence of this phenomenon in large molecules. As is often the case in science, after the facts and in retro- spect, the phenomenon was clear and was soon accepted; to some it even be- came obvious! Looking back, this novel and unexpected observation was a paradigm shift of critical importance, for a number of reasons: First, the observation was the first to clearly show the presence of “quantum coherence effect” in isolated complex chemical systems and only among se- lected vibrational states. In other words, out of the expected chaotic motion in the vibrational and rotational phase space we could see ordered and coherent mo- tion despite the presence of numerous vibrational degrees of freedom (from S, and §, states). This point was theoretically appreciated by only a few sci tists. In fact, at one point Stuart Rice and I drafted a paper on the subject, thinking of clarifying the point. At the time, researchers in high-resolution spectroscopy were observing complex spectra and attributing this complexity to chaotic vibrational motion in the molecule. Stuart and I argued that spec- tral complexity does not mean chaos, and the anthracene experiment was a clear demonstration in real time. Second, the observation demonstrated that coherence had not previously been detected in complex systems, not because of its absence but due to the inability to devise a proper probe. Detection of total absorption or total emis- sion (at all wavelengths) from molecules gives a non-selective window on the dynamics and in this way coherence cannot be detected. This was a key point for the success of the anthracene experiment for which both time and wave- length were resolved and correlated. For all subsequent work on wave packet dynamics, nuclear motion in chemical reactions, and femtochemistry, this concept of “window probing” was essential. The concept was further elucidat- ed by resolving the phase character. By probing at two different wavelengths, we found that the quantum oscillations exhibit identical periods, but were phase shifted by exactly 180° (i.e., they are out-ofphase) [Fig. 10]. The two wavelengths resolved were those corresponding to emission of the initial vibrational state and to that of the vibrational state to which the population goes by IVR. Thus, if “total detection” was invoked, the in-phase and outof- phase oscillations will add up to cancel each other and coherence would have remained undiscovered. Third, observation of phase coherent dynamics gave us a new dimension. The phase shift indicates a true transfer of population, in contrast with con- ventional quantum beats, and by analyzing the phases we could understand the nature of IVR: “concerted”, i.e. going at the same time to all states, or “non- concerted” i.e., going in a sequential redistribution of vibrational energy. We could also obtain the time scale and the effect of molecular rotations on coherence of the vibrational motion. Fourth, the observation illustrated the importance of the “preparation of nonstationary states” in molecules. This issue was of fundamental importan- ce in radiationless transition theories involving multiple electronic states, and experiments by Jan Kommandeur and by Doug McDonald have shown this in- terstate coupling. The question of interest to us was: what nuclear states do we300 Chemistry 1999 IVR- Coherent and Dissipative Regimes Coherent MR Clossicol-2 Noo a > Qvontuen-2 > + § a ib 6 & ‘Qvontum-N fe Joy > tani) j & cc 4G Time (nsec) Figure 10. Dynamics of IVR, intramolecular vibrational energy redistribution. The coherent, re- stricted and the dissipative regimes. Note the exact in-phase and out-ofphase oscillatory behavior between the vibrational states of the system (anthracene in a molecular beam). The theory for classical and quantum pictures (to the left) has been discussed in detail in the literature. [Ref. 51]; See also ref B20, B21. prepare in the isolated molecule on a single surface? The anthracene experi- ment taught us that a coherent source spanning the stationary states can pre- pare a non-stationary state which evolves with time. Moreover, we can prepa- re molecules in-phase at time zero to observe the subsequent coherent dyna- mics. This concept indicates that the description, in terms of Schradinger’s molecular stationary states, is not cardinal and that the time-dependent pic- ture is real and directly relevant to dynamics. Most textbooks describe dyna- mics in terms of stationary states and it took some time for this concept of a time-dependent description to be appreciated. I recently found a theoretical article by Roy Gordon published in the 1960's touching on similar issues. In femtochemistry, the concept of time-domain dynamics is what describes ele- mentary motions. Fifth, by directly probing coherence and its extent in isolated, complex mo- lecular systems we advanced some concepts regarding the nature of IVR and its regions. We divided the regions of IVR into three basic ones: no IVR, re- stricted IVR, and dissipative IVR. We also established that the IVR picture of one vibrational state coupled to a continuum of vibrational levels is not adequate. Instead it is a multi-tier coupling among vibrational states.Ahmed H. Zewail 301 This work and its implications were published in two series of papers and reviewed in two book chapters. The Successful 036 Laboratory The laboratory known as 036 was in the sub-basement of Noyes. In thi laboratory, the initial work on IVR was followed by fruitful applications span- ning (i) studies of IVR in other systems, (ii) radiationless transitions, and reaction rates of a variety of systems. One of our first studies of reactions on the ps time scale, isomerization of stilbene, was stimulated by discussion with Robin Hochstrasser about his work on stilbene vapor at room temperature. He felt that if we could resolve the low-frequency modes in the molecular beam, we would derive a great deal of information on the torsional potential. We resolved these torsional modes. Furthermore, we decided to study the rates as a function of energy and in the process found the barrier for twisting around the double bond and observed coherent IVR in reactions, the first such observation. Even now, stilbene remains a member of our molecular fa- mily and continued studies have been pursued by us and others, also on the fs time scale. The following list highlights some of the work [Fig. 11] done in this initial period from 1981 to 1983: (1) IVR in anthracene and stilbene; (2) trans-cis isomerization of stilbene; (3) quantum beats and radiationless transitions in pyrazine; (4) intramolecular hydrogen bonding in methyl salicylate; (5) in- tramolecular electron transfer in donor-bridge-acceptor systems; (6) IVR and dissociation of intermolecular hydrogen-bonded complexes; and (7) isome- rization of diphenylbutadiene and styrene. Over the years, in the same laboratory (036 Noyes), members of our group have made new extensions covering the following topics: isomerization in iso- lated molecules ys. in bulk solutions; nonchaotic multilevel vibrational ener- gy flow; mode-specific IVR in large molecules; [VR dynamics in alky-anthra- cenes; isotope effects on isomerization of stilbene; charge transfer and exciton dynamics in isolated bianthracene; isotope effects on the intramole- cular dephasing and molecular states of pyrazine; IVR dynamics in alkylanili- nes (the "ring + tail” system); mode-specific (non-RRKM) dynamics of stil- bene-are gas vdW complexes; solvation effects on intramolecular charge transfer; IVR dynamics in p-difluorobenzene and p-luorotoluene (real time vs. chemical timing); IVR dynamics in deuterated anthracenes; dynamics of interstate coupling in chromyl chloride; dynamics of IVR and vibrational pre- dissociation in anthracene-Ar, (n = 1,2,3); structural effects on the dynamics of IVR and isomerization in stilbenes; and dynamics of IVR and vibrational predissociation in n-hexane solvated transstilbene. The research resulted in a series of publications. Changing A Dogma: Development of RCS The success with the anthracene experiment made us ask a similar question, but now regarding the coherent rotational motion of isolated, complex mole- cules. There were theories around which discarded its possibility because of= Chemistry 1999 Figure 11. Some examples of studies in the 036 Laboratory. The isomerization of stilbene, in- tramolecular electron transfer, and solvation in clusters are examples of the studies made in the early 1980's (see text); IQ = isoquinoline. [Ref. B3, 52] the general belief that Coriolis interactions, anharmonicity and other inter- actions would destroy the coherence. We worked out the theoretical implica- tions and the results suggested possibly another surprise: if we could align the molecules with a polarized picosecond pulse and probe (polarization-selective) the rotating molecules, we should be able to observe rotational recurrences which would give the full period of rotations of the isolated (large) molecule. Classically, it is as if the molecule rotated back to its initial configuration. This rotation period gives the moment of inertia and, since the masses of the atoms are known, we can deduce distances, and hence obtain information on molecular structures of very large molecules. Indeed the recurrences in stilbene were observed with high precision, and its molecular structure deduced. Coherence in rotational motion was clearlyAhmed H. Zewail 303 evident and could be probed in a manner similar to what we had done with vibrational coherence. The approach was again met with some skepticism re- garding its generality as a molecular structure technique. However, it is now accepted by many as a powerful Dopplerfree technique; more than 120 struc- tures have been studied this way. The method [Fig. 12] is termed “Rotational Coherence Spectroscopy (RCS)” and is successfully used in many laboratories. Some book chapters and review articles have been published on the subject [see Bibliography]. Out of this first marriage between ultrafast lasers and molecular beams came the developments and concepts discussed above. We were now poised to study molecules and reactions with even shorter time resolution. We could study their vibrational and rotational dynamics and align (“orient”) them by controlling time. Figure 12. Rotational Coherence Spectroscopy (RCS), the concept and first experimental obser~ vation made in a beam of transstilbene. [Ref. B7, B13, 53]304 Chemistry 1999 (3) The Transition to the Sub-Picosecond Regime By the early 1980's, our laser time resolution for studying molecules in super- sonic jets was 15 picoseconds and detection was made using a microchannel plate (~ 40 ps). With this resolution, we had already studied reactions such as the isomerization (twisting) of stilbene, charge-transfer, and intra- and inter- molecular proton and hydrogen atom transfer. How could we improve the time resolution and study, in a general way, the elementary steps of reactions? The only approach I knew of was to use two pulses, one to “pump” and the other to “probe”. Unlike liquid-state studies, where the approach was proven successful, in this case, the density of a molecular beam is very low. Further- more, it was not clear how to establish the zero of time in sitwin the molecu- lar beam and how to avoid temporal broadening due to propagation effects. A New Beam Machine: Pump-Probe Mass Spectrometry I thought we should build a second generation beam apparatus to house a time-of-flight mass spectrometer, From the physics literature, it was clear that single-atom detection could be observed using ionization techniques with lasers, and such detection had already been successful with nanosecond lasers. Unlike many nanosecond studies, we should propagate the two pico- second (and later femtosecond) pulses in the same direction, otherwise we would lose the ultrashort time resolution! The same beam machine was equipped with optics for laser-induced fluorescence detection. We began a new direction of research in a separate laboratory of our group. The new beam was built and integrated with two independently tunable dye lasers. This proved to be a precursor to the femtochemistry work as this taught us to master pump-probe picosecond and sub-picosecond experiments on chemi- cal reactions. In this same laboratory, we studied with a resolution of a few ps: (1) dissociation reactions; (2) ground-state, overtone-initiated reactions; (3) van der Waals reactions and others [Fig. 13]. We wrote a series of papers on state-to-state microcanonical rates, k(E), and addressed theoretical consequences and deviations from the statistical regime. It was in two of these systems (reactions of NCNO and ketene) that we found that the statistical phase-space theory, although successful in describ- ing productstate distributions, failed in describing the microcanonical rates k(E) as a function of energy. Moreover, we made careful studies of the effect of rotational population on k(E), and the effect was dramatic near the thres- hold. Rudy Marcus, stimulated by these studies of k(E), applied variational RRKM theory and we published some papers in a collaborative effort. The key point here is that the TS “moves” to different (shorter) distances along the reaction coordinate at different energies; the cardinal definition is relax- ed [sce above, Section II]. In another system (H,O,), we studied the ground- state (“thermal”) reaction for the first time in real time by initiating the reac- tion with direct excitation of the overtones of the OH stretch vibration [Fig 13]. The coupling between theory and experiment stimulated my interest inAhmed H. Zewail 305 Ground State, Local-Mode Reactions Reaction Rates of NCNO as a Function of Energy vo! sxid" cio em S00 FS Eanes é ° 16808 WWouo 17800 1740017800 T7800 Vibrational Energy, E (wavenumbers) ye(E) (/seconds) Figure 13. Microcanonical Rates of Reactions: (top) Ground-state reaction of HO, initiated by loca-mode excitation; (bottom) the dissociation of NCNO as a function of energy, showing the breakdown of conventional phase-space theory at energies above threshold. (Ref. Bl, 54] the nature of transition states which generally live for less than a picosecond. The thirst for even shorter time resolution became real! ‘The First ICN Experiment: Sub-picosecond Resolution In the early 1980's, the technology of pulse compression became available and we ordered, from Spectra Physics, a pulse compressor ~ a fiber optic ar- rangement to reduce the laser pulse width to sub-picosecond. The company indicated that it would take them several months to build one, and that the only one available was at Purdue University in the laboratory of Professor Duane Smith, one of the first two graduate students I had at Caltech. I men- tioned to Duane my excitement about the experiment, which was intended to directly monitor the elementary bond breakage in a molecule, and asked if it was possible to borrow his compressor. The triatomic molecule ICN was chosen because the CN radical could be conveniently monitored by laser-in-Ahmed H. Zewail 305 Ground State, Local-Mode Reactions Reaction Rates of NCNO as a Function of Energy vo! sxid" cio em S00 FS Eanes é ° 16808 WWouo 17800 1740017800 T7800 Vibrational Energy, E (wavenumbers) ye(E) (/seconds) Figure 13. Microcanonical Rates of Reactions: (top) Ground-state reaction of HO, initiated by loca-mode excitation; (bottom) the dissociation of NCNO as a function of energy, showing the breakdown of conventional phase-space theory at energies above threshold. (Ref. Bl, 54] the nature of transition states which generally live for less than a picosecond. The thirst for even shorter time resolution became real! ‘The First ICN Experiment: Sub-picosecond Resolution In the early 1980's, the technology of pulse compression became available and we ordered, from Spectra Physics, a pulse compressor ~ a fiber optic ar- rangement to reduce the laser pulse width to sub-picosecond. The company indicated that it would take them several months to build one, and that the only one available was at Purdue University in the laboratory of Professor Duane Smith, one of the first two graduate students I had at Caltech. I men- tioned to Duane my excitement about the experiment, which was intended to directly monitor the elementary bond breakage in a molecule, and asked if it was possible to borrow his compressor. The triatomic molecule ICN was chosen because the CN radical could be conveniently monitored by laser-in-