Dark Structures in Molecular Radiationless Transitions Determined by Ultrafast Diffraction

See allHide authors and affiliations

Science  28 Jan 2005:
Vol. 307, Issue 5709, pp. 558-563
DOI: 10.1126/science.1107291


The intermediate structures formed through radiationless transitions are termed “dark” because their existence is inferred indirectly from radiative transitions. We used ultrafast electron diffraction to directly determine these transient structures on both ground-state and excited-state potential energy surfaces of several aromatic molecules. The resolution in space and time (0.01 angstrom and 1 picosecond) enables differentiation between competing nonradiative pathways of bond breaking, vibronic coupling, and spin transition. For the systems reported here, the results reveal unexpected dynamical behavior. The observed ring opening of the structure depends on molecular substituents. This, together with the parallel bifurcation into physical and chemical channels, redefines structural dynamics of the energy landscape in radiationless processes.

Radiationless transitions abound in chemical, physical, and biological systems, yielding such diverse phenomena as the conversion of radiation to heat and the photodamage and photocarcinogenesis of DNA (15). After light absorption, a molecule can undergo radiationless processes of two general types: photochemical, involving bond fragmentation or isomerization; and photophysical, involving transitions between electronic states while either conserving spin (internal conversion) or altering spin (intersystem crossing). For more than eight decades [(6) and references therein], our understanding of such radiationless processes has come from indirect evidence based on yields and decay rates of the radiative population. Theoretical studies have in turn advanced the concepts of a “heat bath” within the molecule (originally thought to violate the rules of quantum mechanics) and of conical intersections in the energy landscape (79).

Experimentally, the presence of nonradiative electronic relaxation processes was first deduced from the decrease in steady-state emission quantum yield of molecules at low pressures (10). With the advent of pico-second time resolution, it became possible to study the time scale of these processes [(11) and references therein]; with femto-second time resolution, the actual nuclear motions were resolved [(12) and references therein]. What remains unknown for these transitions are the accompanying molecular structural changes involving all nuclear motions at once. Such “dark” structures are not amenable to detection by conventional light absorption or emission. With diffraction, however, optically dark structures are as readily observed as bright ones.

Previously, we developed ultrafast electron diffraction (UED) for the study of isolated chemical reactions [(13); (14) and references therein] and for electron crystallography of surfaces and surface molecular adsorbates (1517). Here, the focus is on the nature of transient molecular structures involved in radiationless decays, believed to be nonreactively accelerated by vibrational excitation in the so-called channel-three region (18) and/or by the proximity effect (2) of electronic states. Four prototypical heteroaromatic (pyridine, 2-methylpyridine, and 2,6-dimethylpyridine) and aromatic carbonyl (benzaldehyde) organic molecules have been studied. Pyridine, which we studied before (19), is given here only as a reference. For all molecules, we determined the initial groundstate structure and followed upon excitation the changes in the diffraction pattern with time. The depletion of old bonds and emergence of new bonds elucidated the structural origin of dark transitions in the heteroaromatics and the quinoid-like excited-state structure of triplet benzaldehyde. From these data, we gleaned the influence of parent structure on the dynamical evolution of relaxation pathways and the bifurcation into physical and chemical channels (Fig. 1) on the energy landscape.

Fig. 1.

Schematic of the potential energy landscape of ground and excited states in complex molecular systems. The equilibrium ground-state structure defines the initial wave packet prepared by femtosecond pulse excitation to the excited-state surface. Because of its nonequilibrium nature, the excited-state structure evolves into radiative and radiationless channels. The radiationless transitions can result from bifurcation into reactive chemical processes and nonreactive physical pathways (internal conversion/intersystem crossing).

In our laboratory, the development of the UED methodology over four generations of machines has resulted in current state-of-the-art spatial and temporal resolutions approaching 0.01Å and 1 ps, respectively (14, 15); subpicosecond pulses have also been generated (15). The nearly million-fold increase in scattering cross section for electrons relative to x-rays, and the accompanying sensitivity to population change (∼1%), make UED our technique of choice; for isolated molecules, only UED can provide such resolutions and sensitivity. Briefly, a femtosecond laser pulse initiates the change, and the ensuing structural evolution at specific time delays is probed by ultrashort bursts of electrons. Diffraction patterns are then recorded on a low-noise charge-coupled device (CCD) camera for processing and analysis. Furthermore, through the diffraction-difference method (14), careful choice of the reference image, and establishment of zero-of-time (14), it is possible to determine the particular intermediate structure of the molecular dynamics. Typically we record, at a given time delay, the scattering intensity of the molecular sample, which is converted to the molecular scattering function sM(s). Fourier transform of the sM(s) data generates the radial distribution curves, f(r), which give the relative density of internuclear pairs.

For the studies of the pyridine series, we focused on the so-called channel-three phenomenon: At a given internal energy threshold, the nonradiative decay rate increases abruptly, with a concomitant drop in the emission quantum yield (18). In aromatics such as benzenes and pyridines, upon absorption of light, a substantial portion of the energy is directed into nonradiative pathways. For pyridines, quantum yield measurements of the radiative processes (20), including picosecond studies of the total decay rates (21), have been reported. With methyl group substitution, these nonradiative yields and rates change markedly. In particular, the channel-three onset is much more pronounced for pyridine and picoline (2-methlypyridine) than for lutidine (2,6-dimethlypyridine). Such determinations of quantum yields and decay rates unfortunately contain no molecular structural information. Two fundamental questions remain as yet unanswered: What is the origin of the abrupt change in the radiationless behavior of the pyridines above a certain internal energy threshold? Further, how are such radiationless processes influenced by subtle changes in the molecular structure?

To answer these questions, we determined the transient structures involved in the radiationless transitions of the pyridine series mentioned above. First, however, we established their initial ground-state structures. Fig. 2 shows the two-dimensional (2D) diffraction images along with the experimental and refined radial distribution curves for pyridine, picoline, and lutidine. With the subsequent addition of each methyl group, the 2D images exhibit an increasing density of rings. The starting structures for the fits were obtained from density functional theory (DFT) calculations, followed by least-squares refinement of the internuclear separations and vibrational amplitudes. The peaks in the radial distribution curves reflect the covalent C–C and C–N distances occurring near 1.3Å, the second-nearest-neighbor C··C and C··N distances near 2.3Å, and the third-nearest-neighbor C···C and C···N distances near 2.8Å; all bond distances and angles were obtained for the refined structure and are summarized in figs. S1 and S2. Note that the addition of the methyl group(s) in picoline and lutidine does not affect the mean positions of the first-, second-, and third-nearest-neighbor peaks but increases their density. Simultaneously, the methylation introduces several long indirect distances with concomitant broadening of the peak at distances greater than 3.0Å.

Fig. 2.

Ground-state structures, as shown by 2D diffraction images and experimental radial distribution curves along with least-squares refined structural fits, for (A) pyridine, (B) picoline, and (C) lutidine. Note that the addition of methyl groups in this homologous series of molecules leaves the ring distances nearly unchanged; however, the density of longer, indirect internuclear separations increases with substitution.

To resolve the structural changes during the course of the radiationless transition, we recorded diffraction patterns for a range of time delays between the exciting laser pulse (266 nm) and the probe electron pulse. The 2D diffraction-difference images show the emergence of periodic ring patterns whose intensity becomes more pronounced over time before reaching steady state. The 1D diffraction-difference curves (Fig. 3) consist of negative peaks (blue regions), which correspond to loss of internuclear separations in the parent, and positive peaks (red regions) corresponding to gain of new interatomic distances in the transient intermediate. Note that dispersive contributions arising from vibrationally hot transient species can also lead to positive and negative contributions in the difference curves. Closer inspection of the time evolution reveals that pyridine shows greater loss of the direct covalent bond distances than the indirect ones (as is also the case for picoline). Conversely, in lutidine the noncovalent distances show much larger change.

Fig. 3.

One-dimensional (radial distribution) diffraction-difference curves as a function of time for pyridine (left) and lutidine (right). The negative (blue) regions reflect loss of old bonds; the positive (red) regions correspond to gain of new interatomic distances. Although pyridine shows a more pronounced depletion of the direct covalent distances, lutidine has a greater loss of indirect noncovalent distances. Also, pyridine shows emergence of positive contributions at distances greater than 3.5 Å, a feature conspicuously absent in the f(r) of pyridine ground-state structure (Fig. 2A). Note that at these long distances, the baseline for pyridine at t = 0 only shows weak oscillatory noise; at longer times, positive contributions are present. For lutidine, the positive and negative peaks result from the dispersive contribution of the hot ground state relative to the initial cold structure (Fig. 2C).

Upon excitation, several nonradiative pathways are possible, including valence isomerization, fragmentation, and ring opening (2229). To discriminate among the various possible reaction channels, we compared the UED data with predictions of structural models, as previously done in detail for pyridine (19). Through iterative refinement of the background and the structural parameters, we typically identify multiple best-fit structures based on χ2 criteria. However, fitting with improper choice of the initial structure results in unphysical geometries. Fig. 4 compares experiment and theory for the refined structures, displaying both the molecular scattering and radial distribution curves for the chemical channel in pyridine and picoline and the physical channel in lutidine. We also considered other channels, including the formation of Dewar, Hückel, and azaprefulvene isomers, and fragmentation to HCN; however, the unsatisfactory fits for these structures exclude them from the dominant reaction channel on our time scale (Fig. 4).

Fig. 4.

Transient structural determination. Experimental and theoretical difference curves are shown for (A) pyridine, (B) picoline, and (C) lutidine. Left panels: diffraction-difference molecular scattering curves; right panels: diffraction-difference radial distribution curves. The theoretical ΔsM(s) for ring-opened structures of pyridine and picoline and the hot ground-state structure of lutidine are from the least-squares refined structural fits according to χ2 criteria (see text). Details of bond distances and angles are given in figs. S1 and S2.

On the basis of the final best-fit transient structures (Fig. 4), we conclude that upon 266-nm excitation, pyridine and picoline undergo C–N bond scission to open the aromatic ring and form a diradical structure. Lutidine does not undergo ring opening, but instead gives vibrationally hot ground-state species. The refined structures of ring-opened pyridine and picoline show alternating single (near 1.4Å) and double-bond (near 1.3Å) character for the skeletal distances, indicating disruption of the aromaticity of the parent ring structure. Moreover, the farthest C–N distances (reflected as positive contributions in Fig. 3A) are >4Å, and these are absent in the f (r) of the parent ground-state structure (Fig. 2A). The refined hot lutidine structure(s) are insensitive (within our resolution) to the position of the two methyl groups around the ring; however, the distinct retention of aromatic distances in the product rules out ring opening in this molecule. A least-squares fit of the populations of these transient structures gave the temporal growth with time constants of 17 ± 1 ps, 28 ± 7 ps, and 16 ± 2 ps for pyridine, picoline, and lutidine, respectively.

For aromatic pyridine and picoline, then, the nonradiative channel-three behavior is due to direct ring opening (30) to form the diradical structure obtained above. This observation is in stark contrast to the prevailing view that an ultrafast internal conversion pathway, mediated by the proximity of the first and second excited-state surfaces, opens up at the channel-three threshold and leads to vibrationally hot ground-state molecules. Lutidine does not undergo ring opening, consistent with its lack of channel-three behavior. The ring opening thus explains the quantum yield behavior observed in this series of molecules (3133). In the channel-three region, pyridine and picoline show a change in quantum yield with vibronic energy (35,000 to 40,000 cm–1) whereas lutidine does not, even though for all three molecules the initial (0,0) decay is similar and the emission yield is low because of the proximity of states. The ring opening can also explain observations in gas-phase photolysis (34, 35) and in the condensed phase (27, 28).

Why are these profound differences in photophysical and photochemical behavior induced by subtle changes in the parent molecular structure? In pyridine, optical excitation at 266 nm involves transfer of a nonbonding electron on the nitrogen atom to the antibonding π* orbital of the ring, thus lowering the strength of the C–N bond and resulting in its facile scission. If the time scale of this bond rupture is shorter than the lifetime of the state at that particular total energy, we expect this highly efficient, nonradiative photochemical process to actively deplete the radiative population, causing a decrease in emission quantum yield. The abrupt onset of the channel-three phenomenon is therefore a manifestation of the energy threshold for the ring-cleavage chemistry.

Upon methyl substitution, the electron-donating nature of these groups increases the electron density on the ring, leading to greater stability. Despite this increase in ring stability, structural refinement shows that 2-methylpyridine does indeed relax through ring opening, preferentially on the side opposite the methyl group. This favored scission of the bond between the nitrogen atom and the methine group can be understood in terms of its weaker character relative to other skeletal bonds in the S1 state (36). On the other hand, the presence of two methyl groups in the 2- and 6-positions of lutidine further stabilizes the ring through increased charge density (37). Thus, it is not surprising that the lutidine nonradiative process is photophysical internal conversion rather than photochemical ring opening. Addition of electron-donating methyl substituents also tends to shrink the gap between the close-lying first (nπ*) and second (ππ*) excited states, which expedites the overall decay even with no excess vibrational energy.

The uniqueness of structural determination as a direct probe of molecular behavior emerges in comparison with previous spectroscopic investigations. Using femtosecond time-resolved mass spectrometry in the gas phase, our group had monitored pyridine after 277-nm excitation (22). We observed a fast decay component of 400 fs, assigned to the initial displacement of the wave packet and energy redistribution in the reactive channel. Two slower decays, with time constants of ∼3.5 ps and ∼15 ps, were assigned to isomerizations to Dewar and Hückel pyridines, respectively, purely on the basis of the calculated energetics. With the direct structural information elucidated here by UED, we can reconsider these mass spectrometric results (22). The ∼3.5-ps time scale can still be assigned to the Dewar isomer, which is a minor channel in the time-resolved mass spectra as well as in our UED data (19). However, the time scale of the major channel (∼15 ps) is remarkably similar to the time scale of the UED ring-opening process (17 ± 1 ps). Furthermore, the reported insensitivity of the time scales to deuteration (∼15 ps for pyridine-h5 versus ∼16 ps for pyridine-d5) would be surprising for the Hückel pathway, given that valence isomerization involves the motions of two deuterium atoms. On the basis of these considerations, therefore, we reassign the ∼15-ps decay rate from the mass spectrometric experiment to the ring cleavage process. This decay time is close to the reported lifetime (20 ps) of pyridine fluorescence at 1636 cm–1 above the S1 origin (21).

As noted above, we have also applied UED to an aromatic carbonyl. Benzaldehyde, upon light absorption, undergoes efficient nonradiative intersystem crossing to the triplet state, as evidenced by its high phosphorescence yield (38). Previous investigations have revealed photochemical dissociation into benzene and carbon monoxide (39) above an internal energy threshold (∼35,000 cm–1). We sought to determine at 266 nm whether the photophysical and photochemical processes occur consecutively or instead proceed competitively as a result of bifurcation.

Perusal of the experimental Δf(r) curves for excited benzaldehyde in Fig. 5B supports the rupture of covalent C–C bonds (near 1.4Å), loss of next-nearest-neighbor distances (near 2.5Å), and depletion of longer distances (>3.5Å). These results suggest a fragmentation and repositioning of the atomic nuclei. As in the case of the azines, several possible candidate structures were considered, and from the molecular scattering sM(s) and radial distribution f(r) curves, the two best structures were found to be a quinoid-like triplet benzaldehyde and a dissociated product of benzene and carbon monoxide (Fig. 5A). The data were then fit to a sum of these two structures, with the partitioning parameter floated at each delay time. Shown in Fig. 5B is the structural fit at t = 50 ps, with 40% partitioned to the dissociation. In Fig. 5C, we display the temporal rise for both channels.

Fig. 5.

Excited-state structure and bifurcation pathways in benzaldehyde. (A) Refined structural parameters for the transient species; the distances are in angstroms and the angles are in degrees. (B) Radial distribution curve of difference data and refined theoretical model for excited triplet benzaldehyde, benzene, and carbon monoxide at t = 50 ps. (C) Population increase and fit of a first-order exponential to the rise of excited triplet benzaldehyde and benzene with time constants of 25 ± 4 ps and 38 ± 5 ps, respectively. The parameters marked with an asterisk were not independently refined but were derived from geometric constraints.

The quinoid structure is the optically dark, excited-state (ππ*) triplet benzaldehyde formed as a result of intersystem crossing (40, 41). As seen from the refined geometric parameters (Fig. 5A), this excited-state structure exhibits well-defined single and double bonds, indicating disruption of aromaticity in the ring. On the other hand, benzene formed in the dissociation pathway is in its ground electronic state, as indicated by the refined C–C bond distance in Fig. 5B. The simultaneous emergence of the photophysical and photochemical products in the diffraction data indicates a bifurcation on the excited singlet surface. The time-dependent contribution of the two pathways in the data is described by apparent rise time constants of 25 ± 4 ps and 38 ± 5 ps for excited triplet benzaldehyde and benzene, respectively (Fig. 5C). The observed difference in the rise times suggests the presence of an intermediate for the formation of benzene.

This observed bifurcation into physical and chemical pathways resolves key long-standing issues. Our quantum chemical calculations on the excited singlet surface suggest a low barrier for hydrogen shift and subsequent dissociation. Additionally, intersystem crossing from the excited singlet to the excited triplet is highly efficient. These features of the excited-state potential energy surface, combined with the similar rates of the physical and chemical channels, make these pathways competitive. When viewed in the context of previous mass spectrometry experiments, the new results account for the presence of ground-state benzene as well as triplet benzene on the nanosecond time scale, the latter produced by dissociation of excited triplet benzaldehyde (42, 43).

The UED-determined dark transient structures, including those of excited states, and the time scales for bifurcation into physical and chemical channels underscore the critical importance of structural dynamics in determining the true nature of complex molecular behaviors and the energy landscapes of radiationless transitions.

Supporting Online Material

Figs. S1 and S2

References and Notes

View Abstract

Navigate This Article