Reactive Intermediate Chemistry

912 REACTIONS ON THE FEMTOSECOND TIME SCALE

Norrish type-I reaction, has been studied over the years in extreme detail, with every imaginable physical and theoretical method at hand. Data gathered through studying such reactions on the femtosecond time scale, together with new theoretical work prompted by the dynamics observed, have provided a detailed picture of

the processes involved and a fresh perspective on nonconcerted a-cleavage events.20–24

Excitation of acetone with two photons at l ¼ 307 nm delivers 186 kcal/mol of energy, more than enough to break both C C bonds and give carbon monoxide and two methyl radicals. Following the process with femtosecond mass spectrometry shows that the excited state of acetone of 58 amu rises and falls very quickly, in a spike-like fashion. It is formed and decays with a time constant of 50 fs. For acetone-d6, a similar time dependence is seen: The rise and fall of the 64 amu excited species take place in 80 and 80 fs (Fig. 20.5).

By using a full panoply of theory—time-dependent density functional theory and CASSCF and CASMP2 ab initio methods—the states and potential-energy surface features associated with the dynamics were uncovered. The high-energy excitation of acetone gives several excited Rydberg states that reach the S2(n, 3s) surface in

Time (fs)

Figure 20.5. A graphical representation of the time evolution of transients for the Norrish type-I a-cleavage 43 and 46 amu fragments from acetone and from acetone-d6. The representative sets of ‘‘data points’’ (& for 43 amu, for 46 amu fragments) are modeled with simple buildup and decay response functions, IðtÞ ¼ A½expð t=t2Þ expð t=t1Þ&; the time constants of buildup and decay are t1 and t2, respectively. A modest isotope effect on the characteristic time for formation of these acyl radicals (60 and 80 fs, respectively) and a more prominent CH3/ CD3 effect on decays through loss of CO (420 and 670 fs, respectively) were recorded.24

50 fs or less, and proceed through cleavage of a C C bond in 50 fs. Larger ketones decay in slightly longer times (e.g., 2-butanone, in 100 fs; 3-pentanone, in 160 fs), but the trend is much less pronounced than would be expected if there were full vibrational energy redistribution prior to the first cleavage reaction. The a- cleavage is highly nonstatistical. The energy redistribution from the C O stretching mode to C C stretching and C C O bending modes leading to bond cleavage is quite efficient.

Acetone in the S2 state gives an excited-state linear acetyl radical along with the

methyl radical; the radical exists in a double-well potential about the CCO angle of 180 .

The fragment radical of 43 amu from acetone, the acetyl intermediate, builds up in 60 fs and decays in 420 fs. The first a-cleavage takes place in a time comparable to the vibrational period of the C C bond, 43 fs, while the second is much slower. The intermediate of 46 amu from acetone-d6 rises in 80 fs and decays in670 fs.

The second cleavage depends on moving vibrational energy into the C C bond stretching coordinate while trajectories along the C C O bending coordinate are nonreactive. This partial vibrational energy redistribution takes in 420 fs, substantially longer than required to break the first C C bond. The CH3/CD3 isotope effect reflects nuclear motions involving the CH3 and CD3 substructures, an effect seen previously in the decomposition of methyl iodide.

When acetone is photoactivated at lower energies, with one photon at l ¼ 307 nm (93 kcal/mol), quite different photochemical events lead to the a- cleavage.25 The initial femtosecond motion leads to very fast dephasing of the wave packet from the Franck–Condon region on the S1(n,p*) surface, with a pulse-limited time constant of <50 fs. The photoexcited acetone on the S1 surface then persists on a nanosecond time scale, eventually giving methyl and acetyl radicals through an a-cleavage. This fragmentation does not occur directly from S1; it takes place indirectly, through intersystem crossing from S1 to T1, a state that has available a much less energetically demanding a-cleavage option.

When cyclic ketones are pumped with two photons of a l ¼ 307-nm femtosecond pulse, imparting some 186 kcal/mol of excitation energy, they are converted rapidly, within 50 fs, through a Norrish type-I C C cleavage to acylalkyl diradicals of varying size and number of degrees of freedom, Nv. From cyclobutanone to cyclodecanone, Nv ranges from 27 to 81. The decays of the acylalkyl diradicals

5.8. Trimethylene and Tetramethylene Diradicals

The postulation of trimethylene and tetramethylene diradicals as reactive intermediates involved in many thermal isomerization and fragmentation reactions has a long history,31 but not until 1994 had they ever been detected in real time. The validity of the ‘‘diradical hypothesis’’ was tested through femtosecond studies, and the tests provided dramatic evidence confirming that these short-lives species are indeed real, directly experimentally accessible chemical entities.27,32

The pump–probe–detect arrangements for the femtosecond experiments was similar to those described above. When cyclobutanone was pumped with two photons of a l ¼ 307-nm femtosecond pulse, two consecutive C CO bond cleavages led to the formation of the trimethylene diradical, detected as an easily ionized transient at 42 amu, with buildup and decay times of 120 20 fs. The decay presumably involves isomerizations to cyclopropane and to propylene— structures not ionized by the probe pulse and thus undetected during the experiment.

τ ~ 120 fs

Starting with cyclopentanone and a pump pulse delivering two photons at 310 nm, the parent mass at 84 amu grows in intensity and then decays. So does a species at 56 amu, the parent minus CO, with a buildup time of 150 30 fs and a decay time of 700 40 fs.27 It attains a peak intensity in 300 fs.

τ ~ 1150 fs

The decay of the tetramethylene diradical derived from 2,2,5,5-d4-cyclopenta- none is much slower than seen for the C4H8 diradical.24 Both principal decay modes, fragmentation to two ethylenes and ring-closure to cyclobutane, may be dependent dynamically on torsional motions of the terminal methylene groups.

These studies raised many interesting questions and set an ambitious agenda for further work, served admirably to establish the time scales appropriate to tetramethylene and trimethylene diradicals, and confirmed them as distinct molecular species.

Unpublished work by Herek and Zewail on the dynamics of photodecarbonylations of 2,2,4,4-d4-cyclobutanone and 3,3-d2-cyclobutanone uncovered that the

916 REACTIONS ON THE FEMTOSECOND TIME SCALE

CD2CH2CD2 and CH2CD2CH2 diradicals decay at somewhat longer lifetimes, in183 and 129 fs, respectively. Torsional motions of the terminal methylene groups are obviously critical to the reaction coordinate leading to vibrationally excited cyclopropane-di products.33

Calculational estimates of the lifetimes of the trimethylene diradical34,35 based on microcanonical variational unimolecular rate theory and direct dynamics simulations have been reported. The lifetimes derived from theory, 91 and 118 fs, are comparable to the experimental estimate, 120 20 fs. Similar lifetime estimates from theory for tetramethylene are comparable, or slightly below, the experimental value.36

Dynamics calculations have also provided new approaches to the stereochemical modes through which cyclopropanes and trimethylene intermediates may be related.37–39 Full quantum dynamics calculations for the trimethylene diradical based on a reduced dimensionality model that followed wave packet densities and time constants for formation of products led to the conclusion that conrotatory and disrotatory ‘‘double’’ rotations of both terminal methylene groups are favored over a ‘‘single’’ rotation of just one by a 2.2:1 ratio.40

A related study sought the lifetime of the 1,3-cyclopentanediyl diradical, generated photochemically from the 2,3-diazabicyclo[2.2.1]hept-2-ene.41 This diradical may be viewed as a trimethylene diradical constrained geometrically by an ethano tether, or as a tetramethylene diradical bridged by a methylene function. In either view, the 1,3-cyclopentanediyl diradical might possibly decay through cyclization to bicyclo[2.1.0]pentane, or a 1,2-hydrogen shift to form cyclopentene, or with C C bond cleavage to give 1,4-pentadiene. Theory suggests that the ring closure is by far the preferred alternative.

The mass spectrum of diazabicyclo[2.2.1]hept-2-ene shows only a weak molecular ion and a very strong fragment at 68 amu. The femtosecond studies found that the 68 amu easily ionized transient profile could be modeled with a rise time of 30 10 fs and decay time of 190 10 fs, a value comparable to the decay time of trimethylene.

Now the 1,3-cyclopentanediyl diradical is constrained to cyclize in a disrotatory fashion while the trimethylene species might well close in both disrotatory and conrotatory ways. Were all other factors constant one could infer that the geometrical restrictions imposed on the 1,3-cyclopentanediyl diradical entailed no significant deduction in rate of cyclization, and thus that conrotatory cyclization of the trimethylene diradical is not strongly preferred under the given reaction conditions and circumstances.

5.9. Oxyalkyl Diradical and Formylalkyl Radical Intermediates

Two-photon excitation at l ¼ 307 nm provides 186 kcal/mol of energy to tetrahydrofuran (THF); the excited molecule breaks a C O bond with a characteristic time of 55 15 fs to form an oxytetramethylene diradical. This species, now with 114 kcal/mol of available energy, has a lifetime of 65 15 fs; it decays primarily through a C C b-cleavage reaction, giving the trimethylene diradical (42 amu, t 120 fs). The cleavage to trimethylene is the dominant reaction.42

An alternative b-cleavage process contributes to a lesser extent: Loss of a hydrogen gives a formylalkyl radical of 71 amu, which loses ethylene to provide another formylalkyl radical of 43 amu. The transient species at 71 and 43 amu have decay times of 120 fs, similar to the time seen for the trimethylene diradical.

This femtosecond study confirmed the involvement of the oxytetramethylene diradical as a reactive intermediate, and found that the trimethylene formed from it had the same lifetime as the trimethylene generated through the photodecarbonylation of cyclobutanone. For tetrahydropyran, the oxypentamethylene diradical (86 amu) is formed readily and the 85 amu transient, from the b-cleavage of a C H bond, is the dominant fragmentation product.

5.10. Retro-Diels–Alder Reactions

Retro-Diels–Alder reactions have long been studied and discussed with an emphasis on whether they should be considered ‘‘concerted’’ or ‘‘step-wise’’ processes. Femtosecond real time studies of representative retro-Diels–Alder reactions of

simple hydrocarbons have helped to provide an answer and to sharpen the nature of the question.43,44

Cyclohexene, upon excitation through a two-photon process providing 186 kcal/ mol, gives two species detected through ionization by a probe pulse and mass spectrometry: a species at 82 amu, the parent structure or the diradical species formed through b-cleavage, and at 54 amu, a mass corresponding to butadiene. An ion at M-15, at 67, is also recorded. The femtosecond transients show that the 82 amu

918 REACTIONS ON THE FEMTOSECOND TIME SCALE

species rises in <10 fs and decays in 225 20 fs; the amu 54 signal rises in 15 10 fs and decays in 150 15 fs.

Clearly, the 54 amu species is not formed from the 82 amu entity; rather both derive from the very rapid decay of a common precursor, the photoexcited cyclohexene. The 82 amu species may then be associated with the diradical intermediate on the nonconcerted pathway, and the 54 signal with an excited form of butadiene, one vulnerable to ionization when probed with a l ¼ 615-nm photon (46 kcal/mol).

For norbornene, a similar situation with respect to dynamics obtains. The 94 amu species corresponding to a C7H10 entity rises in <10 fs and decays in 190 10 fs; the fragment species at 66 amu rises in 30 15 fs and decays in 230 30 fs.

For bicyclo[2.2.2]oct-2-ene, the same pattern is observed: a C8H12 species at 108 forms very rapidly, rising in <10 fs and then decays in 190 10 fs; the fragment at 80 amu rises in 15 10 fs and decays in 185 10 fs.

All three of these retro-Diels–Alder reactions give excited diene intermediates that decay in comparable times: the t values range from 150 to 230 fs. The exact structural characteristics of these intermediates is currently unclear. Perhaps this issue could be addressed using femtosecond spectroscopic studies applying laserinduced fluorescence techniques, or through theory-based approaches.

In all cases, the diradical intermediate formed along the nonconcerted reaction pathway persists for characteristic times ranging from 190 10 to 230 20 fs— many times longer than the time associated with a typical C C vibration, 30 fs. The diradical intermediates are real enough, though the lifetimes themselves give no direct information on the relative importance of concerted versus nonconcerted alternatives on ground-state potential energy surfaces. That matter may be contingent on system-dependent couplings between symmetric and antisymmetry C C bond stretching modes leading toward concerted and nonconcerted transition regions in cooperation with other modes contributing to the reaction coordinates.

The retro-Diels–Alder reactions studied may well involve many transient configurations appropriate to the potential energy surface and vibrational possibilities approaching and traversing transition regions.45,46

Comprehensive theoretical investigations of such retro-Diels–Alder reactions, with particular attention devoted to the very rapid transitions between photoexcited and ground-state potential energy surfaces at conical intersections, have appeared.47

920 REACTIONS ON THE FEMTOSECOND TIME SCALE

Figure 20.6. Comparisons of experimental and theoretical electron diffraction radial distribution curves based on ab initio geometries for the bridged C2F4I radical structure (on left) and a sum of classical anti and gauche structures (on right).48 The intermediate present 5 ps after the pump pulse was defined structurally through 2D diffraction difference images. [Reproduced with permission from H. Ihee, V. A. Lobastov, U. M. Gomez, B. M. Goodson, R. Srinivasan, C.-Y. Ruan, and A. H. Zewail, Science 2001, 291, 458. Copyright # 2001 American Association for the Advancement of Science.]

undertaken. The time-dependent diffraction data recorded by the CCD detector provided structural data for the cyclic diene in fine agreement with the literature. Difference curves at various delay times were consistent with depletion of covalent and next-nearest-neighbor C C pairs in the starting diene; a number of new positive

new C C interactions at separations longer than those found in the cyclic diene. Given the conformational flexibility of the product triene, and the lack of heavy atoms in the system, the agreement between experimentally recorded and theorybased diffraction-difference-defined radial distribution functions may be considered remarkable. The electrocyclic ring opening was observed directly following femtosecond excitation: Still greater sensitivity and more narrowly defined electron pulses may bring the day of full structural visualizations of reactions in progress closer.

and conformers

Another approach would use X-ray crystallography; promising preliminary examples have been reported, but much more needs to be done before the prospect is realized in any general sense.55

7. MORE COMPLEX REACTIONS

This introduction to experiments following chemical reactions in the femtosecond time domain could be extended to considerations of other gas-phase reactions

involving molecular structures of similar complexity. Or one could extend this chapter extensively and attempt to survey femtosecond chemical investigations of far more complex systems and dynamic issues, including studies involving other phases of matter—gases, liquids, solids, clusters, surfaces, and biological systems. There have been reports on the femtosecond dynamics of DNA assemblies, and bacteriorhodopsin photochemistry, and electron-transfer processes in proteins, and the photodissociation of carbon monoxide from its complex with myoglobin: one might say, ‘‘The sky’s the limit!’’

There are technical challenges still to be overcome before every thought experiment in this brave new realm of molecular science can be realized in practice, but there is good reason for optimism. While ‘‘simple’’ gas-phase reactions of comparatively small molecules will continue to attract serious attention, the forefronts of femtochemistry today encompass far wider perspectives.

8. CONCLUSION AND OUTLOOK

The experimental and theoretical strategies of femtochemistry have provided telling insights on chemical dynamics over the past 15 years. The breakthrough examples and many of the prototypical organic reactions that have been reported already permit some important generalizations.

Experimentally, it has become possible to follow chemical reactions on a femtosecond time scale with excellent time resolution—with precise definitions of t ¼ 0, time separations between pulses, and with very narrow pulse widths.

Each pump–probe sequence may involve on the order of a million or a billion independent molecules; the excitations give a spatially well-defined coherent matter wave packet. An initially randomly oriented ensemble of molecules is launched through the narrow pump laser pulse to give molecules with nuclear–nuclear dis-

˚

tances defined to 0.1 A. The sequence typically is repeated many, many times, over the full range of relevant times, so that the relatively weak signals detected and the intrinsic scatter of the experimental data may be signal averaged. Hence, the absolute imperative of precisely defined timing for the pulses, to initiate a reaction with a pump pulse and to monitor progress along the reaction coordinate using a probe pulse and some detection strategy such as mass spectrometry. Without excellent synchronization, the computer implemented signal averaging would give blurred composites of the individual runs, and little useful information.

Excitations of molecules with femtosecond laser pulses lead to excited-state matter wave packets coherently, launching them with such well-defined spatial resolution and coherence in nuclear motions that they evolve like single-molecule trajectories. Both electronically excited and vibrationally excited ground-state species may be studied. The structural change versus time profile of a reaction turns out to be compatible with classical modes of thinking.

Reaction dynamics may be controlled by the cooperative conjunction of two or more vibrational variables: A rate may be dependent on the correct phasing of such modes, rather than simply upon achieving or exceeding an energy barrier.