Report

Dynamics of Chemical Bonding Mapped by Energy-Resolved 4D Electron Microscopy

See allHide authors and affiliations

Science  10 Jul 2009:
Vol. 325, Issue 5937, pp. 181-184
DOI: 10.1126/science.1175005

Abstract

Chemical bonding dynamics are fundamental to the understanding of properties and behavior of materials and molecules. Here, we demonstrate the potential of time-resolved, femtosecond electron energy loss spectroscopy (EELS) for mapping electronic structural changes in the course of nuclear motions. For graphite, it is found that changes of milli–electron volts in the energy range of up to 50 electron volts reveal the compression and expansion of layers on the subpicometer scale (for surface and bulk atoms). These nonequilibrium structural features are correlated with the direction of change from sp2 [two-dimensional (2D) graphene] to sp3 (3D-diamond) electronic hybridization, and the results are compared with theoretical charge-density calculations. The reported femtosecond time resolution of four-dimensional (4D) electron microscopy represents an advance of 10 orders of magnitude over that of conventional EELS methods.

Bonding in molecules and materials is determined by the nature of electron density distribution between the atoms. The dynamics involve the evolution of electron density in space and the motion of nuclei that occur on the attosecond and femtosecond time scale, respectively (13). Such changes of the charge distribution with time are responsible for the outcome of chemical reactivity and for phenomena in the condensed phase, including those of phase transitions and nanoscale quantum effects. With convergent-beam electron diffraction (4), the static pattern of charge-density difference maps can be visualized, and using x-ray absorption (5) and photoemission spectroscopy (68) substantial progress has been made in the study of electronic-state dynamics in bulks and on surfaces. Electron energy loss spectroscopy (EELS) is a powerful method in the study of electronic structure on the atomic scale, using aberration-corrected microscopy (9), and in chemical analysis of selective sites (10); the comparison with synchrotron-based near-edge x-ray absorption spectroscopy is impressive (11). The time and energy resolutions of ultrafast electron microscopy (UEM) (1216) provide the means for the study of (combined) structural and bonding dynamics.

Here, time-resolved EELS is demonstrated in the mapping of chemical bonding dynamics, which require nearly 10 orders of magnitude increase in resolution from the detector-limited millisecond response (17). By following the evolution of the energy spectra (up to 50 eV) with femtosecond (fs) resolution, it was possible to resolve in graphite the dynamical changes on a millielectronvolt (subpicometer motion) scale. In this way, we examined the influence of surface and bulk atoms motion and observed correlations with electronic structural changes: contraction, expansion, and recurrences. Because the EEL spectra of a specimen in this energy range contain information about plasmonic properties of bonding carriers, their observed changes reveal the collective dynamics of valence electrons.

Graphite is an ideal test case for investigating the correlation between structural and electronic dynamics. Single-layered graphene, the first two-dimensional (2D) solid to be isolated and the strongest material known (18), has the orbitals on carbon as sp2 hybrids, and in graphite the π-electron is perpendicular to the molecular plane. Strongly compressed graphite transforms into diamond, whose charge density pattern is a 3D network of covalent bonds with sp3 hybrid orbitals. Thus, any structural perturbation on the ultrashort time scale of the motion will lead to changes in the chemical bonding and should be observable in UEM. Moreover, surface atoms have unique binding, and they too should be distinguishable in their influence from bulk atom dynamics.

The experiments were performed on a nm-thick single crystal of natural hexagonal graphite. The sample was cleaved repeatedly until a transparent film was obtained, and then deposited on a transmission electron microscopy (TEM) grid; the thickness was determined from EELS to be 108 nm. The fs-resolved EELS (or FEELS) data were recorded in our UEM, operating in the single-electron per pulse mode (12) to eliminate Boersch’s space charge effect. A train of 220 fs infrared laser pulses (λ = 1038 nm) was split into two paths; one was frequency-doubled and used to excite the specimen with a fluence of 1.5 mJ/cm2, and the other was frequency-tripled into the UV and directed to the photoemissive cathode to generate the electron packets. These pulses were accelerated in the TEM column and dispersed after transmission through the sample in order to provide the energy loss spectrum of the material. Details of the clocking and characterization methodology can be found in (12, 13, 16).

The experimental, static EEL spectra of graphite in our UEM, with graphene (19) for comparison, are displayed in Fig. 1A; Fig. 1B shows the results of theoretical calculations (19, 20). The spectral feature around 7 eV is the π plasmon, the strong peak centered around 26.9 eV is the π + σ bulk plasmon, and the weaker peak on its low energy tail is due to the surface plasmon. All these features have been assigned in the literature (1921). The agreement between the calculated EEL spectra and the experimental ones is satisfactory both for graphite and graphene. Of relevance to our studies of dynamics is the simulation of the spectra for different c-axis separations, ranging from twice as large as naturally occurring (2c/a; a and c are lattice constraints) to 5 times as large (20). This thickness dependence is displayed in Fig. 1B.

Fig. 1

Static and femtosecond-resolved EELS of graphite. (A) The UEM-obtained experimental spectrum of graphite; for comparison with the spectrum of graphene, see (19). (B) Simulated spectrum for natural graphite, together with the calculated spectra obtained for expanded c-axis structures, with the separation being twice, three times, and five times as large as the native one (20). The theoretical spectrum of graphene is also displayed (19). (C and D) Peak intensity changes of surface (C) and bulk (D) plasmons as a function of time. Solid lines are guide to the eyes. The increase in intensity for the bulk corresponds to a decrease in intensity for the surface plasmon (i.e., nearly out of phase). The zero of time was determined as shown in Fig. 3.

As displayed in Fig. 1, the surface and bulk plasmon bands (between 13 and 35 eV) can be analyzed using two Voigt functions, thus defining the central position, intensity, and width. At different delay times, we monitored the changes and found that they occur in the intensity and position; the width and shape of the two spectral components are relatively unchanged. Fig. 1, C and D, shows the temporal changes of the intensity for both the surface and bulk plasmons. As noted, the behavior of bulk dynamics is “out of phase” with that of the surface dynamics, corresponding to an increase in intensity for the former and a decrease for the latter. Each time point represents a 500-fs change. Within the first 1 ps, the bulk plasmon gains spectral weight with the increase in intensity. With time, the intensity is found to return to its original (equilibrium) value. At longer times, a reverse in sign occurs, corresponding to a decrease and then an increase in intensity—an apparent recurrence or echo occurring with dispersion. The intensity change of the surface plasmon in Fig. 1, C and D, shows a π phase-shifted temporal evolution with respect to that of the bulk plasmon.

The time dependence of the energy position of the different spectral bands is displayed in Fig. 2. The least-squares fit converges for a value of the surface plasmon energy at 14.3 eV and of the bulk plasmon at 26.9 eV, in good agreement with literature reports (1921). The temporal evolution of the surface plasmon gives no sign of energy dispersion, whereas the bulk plasmon is found to undergo first a blueshift and then a redshift at longer times (Fig. 2, A and B). The overall energy-time changes in the FEEL spectra are displayed in Fig. 3. To make the changes more apparent, the difference between the spectra after the arrival of the initiating laser pulse (time zero) and a reference spectrum taken at −20 ps before time zero is shown. The most pronounced changes are observed in the region near the energy of the laser itself (2.39 eV), representing the energy-loss enhancement due to the creation of carriers by the laser excitation, and in the region dominated by the surface and bulk plasmons (between 13 and 35 eV). Clearly evident in the 3D plot are the energy dependence as a function of time, the echoes, and the shift in phase.

Fig. 2

EELS peak position and diffraction as a function of time. (A) Energy dispersion of bulk plasmon. (B) Energy dispersion of surface plasmon. (C) Energy dispersion predicted from diffraction experiments at 1.5 mJ/cm2. The energy dispersion associated with structural changes is obtained from (D), where the position of the bulk plasmon is plotted against the c-axis separation (20). The zero of time was determined as shown in Fig. 3.

Fig. 3

3D intensity-energy-time FEELS plot. The plot shows time-energy-amplitude evolution of spectral changes at different time delays. The inset on the right depicts a contour-map version to enhance the visibility of time- and energy-dependent evolution. The zero of change is set to be sky blue in both plots. The inset on the left depicts two transients for the band at the laser energy (3 to 5 eV) and that of the bulk plasmon (26.9 eV). Note the shift from t = 0 (25). The reported time constant and shift were obtained from two independent measurements made on different specimens.

A wealth of information has been obtained on the spectroscopy and structural dynamics of graphite. Of particular relevance here are the results concerning contraction and expansion of layers probed by diffraction on the ultrashort time scale (22). Knowing the amplitude of contraction/expansion, which is 0.6 pm at the fluence of 1.5 mJ/cm2 (22), and from the change of plasmon energy with interlayer distance (Fig. 1), we obtained the results shown in Fig. 2C. The diffraction data, when now translated into energy change, reproduce the pattern in Fig. 2A, with the amplitude being within a factor of two. When the layers are fully separated, that is, reaching graphene, the bulk plasmon, as expected, is completely suppressed (19).

The dynamics of chemical bonding can now be pictured. The fs optical excitation of graphite generates carriers in the nonequilibrium state. They thermalize by electron-electron and electron-phonon interactions on a time scale found to be less than 1 ps (23), less than 500 fs (8), and ~200 fs (24). From our FEELS, we obtained a rise of bulk plasmons in ~180 fs (Fig. 3) (25), consistent with the limits in (8, 23, 24). The carriers generated induce a strong electrostatic force between graphene layers, and ultrafast interlayer contraction occurs as a consequence (22, 26). In Fig. 1D, the increase of the bulk plasmon spectral weight on the fs time scale reflects this structural dynamics of bond-length shortening because it originates from a denser and more 3D charge distribution. After the compression, a sequence of dilatations and successive expansions along the c axis follows, but, at longer times lattice thermalization dephases the coherent atomic motions; at a higher fluence, strong interlayer distance variations occur, and graphene sheets can be detached as a result of these interlayer collisions (2729). Thus, the observations reported here reflect the change in electronic structure: contraction toward diamond and expansion toward graphene. The energy change with time correlates well with the EELS change calculated for different interlayer distances (Fig. 1).

We have calculated the charge density distribution for the three relevant structures. The self-consistent density functional theory calculations were made using the linear muffin-tin orbital approximation (30), and the results are displayed in Fig. 4. To emphasize the nature of the changes observed in FEELS, and their connection to the dynamics of chemical bonding, we pictorially display the evolution of the charge distribution in a natural graphite crystal, a highly compressed one, and the extreme case of diamond. One can see the transition from a 2D to the 3D electronic structure. The compressed and expanded graphite can pictorially be visualized to deduce the change in electron density as interlayer separations change.

Fig. 4

Charge density distributions and crystal structures. (A) Charge density calculated for graphite (c/a = 2.7), compressed graphite (c/a = 1.5), and diamond. All charge densities (ρ) are visualized as constant value surfaces, corresponding to approximately ρmax/3. (B) Crystal structures and geometrical representations of the transition from graphite to diamond (34).

With image, energy, and time resolution in 4D UEM, it is possible to visualize dynamical changes of structure and electronic distribution. Such stroboscopic observations require time and energy resolutions of fs and meV, respectively, as evidenced in the case study (graphite) reported here, and for which the dynamics manifest compression/expansion of atomic planes and electronic sp2/sp3-type hybridization change. The application demonstrates the potential for examining the nature of charge density and chemical bonding in the course of physical/chemical or materials phase change. It would be of interest to extend the scale of energy from ~1 eV, with 100 meV resolution (31), to the hundreds of eV (32) for exploring other dynamical processes (33) of bonding.

  • * Present address: Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne-Dorigny, Switzerland.

References and Notes

  1. Because FEELS records the energy spectra of all bands in the range studied, we were able to monitor the temporal evolution of the peak at the laser energy (3 to 5 eV), t = 0, and compare with that of the bulk plasmon peak (at 26.9 eV). A definite shift (~150 fs) for the latter was observed in two different specimens. With least-squares analysis (rise and decay) we determined the decay of the low-energy band and the rise of the higher-energy band to be ~180 fs. In previous reports (16, 22) the time zero was relative, reflecting the point when the intensity increases (decreases), whereas in this study the time zero was determined from the behavior of the 3 to 5 eV region (at the initial excitation). Thus, all data are analyzed here with this determined value.

  2. This work was supported by the National Science Foundation and the Air Force Office of Scientific Research in the Gordon and Betty Moore Center for Physical Biology at the California Institute of Technology. We thank B. Barwick for helpful and stimulating discussion.
View Abstract

Navigate This Article