Unlocking the time resolved nature of electron microscopy.

D uring the past few years, Zewail (1) and others (2, 3) have reported a remarkable series of advances that unlock the inherent time resolution in EM by introducing a laser-excited photocathode to produce timed pulses of electrons in the microscope in synchronism with a pump laser that interacts with the specimen. These experiments use variable pumpprobe delays to measure the temporal evolution of the specimen after the pump excitation. If the electron-probe pulses are short and limited in intensity, then they will each contain essentially one electron, producing a well-defined single-electron wave packet scattering in the specimen. This single-electron picture provides the foundation for interpretation of highresolution spatial information in lateral x–y directions averaged over the specimen thickness (4). In this experiment, temporal resolution depends on the length of the wave packet in the z direction and its velocity. Since its invention in the late 1930s, EM has served as a workhorse tool for structural characterization in materials physics. Therefore, it would seem that the technology that makes it possible, and the associated physical science that it reveals, should be well-explored by now. Despite this, it has only been within the last 10 y that control of aberrations has allowed the routine preparation of sub–Angstromsized probes for use in EM (5). As a result, we are now experiencing a swift development of new microscopy techniques based on our improved capability to control the lateral properties of kilovolt electron beams, even to the Angstrom level, allowing us to locate, identify, and discover the function of atoms, singly and in small groups, imbeddedwithin bulkmaterials (6). A few moments of thought about the longitudinal size of the electronwave packet produces an expectation for the temporal accuracy of scattering in the electron microscope. Electron emission in the EM typically produces electrons having an energy spread of about 1 eV, corresponding to a longitudinal size of about 1 μm. During acceleration to half the speed of light, and propagation to the sample, the electron wavepacket spreads to as much as 100 μm. Therefore, the transit time of the wavepacket through the thin sample is about 500 fs, defining a measurement uncertainty for the timing of a specimen interaction. Thus, EM, equipped as described above with a method for timed creation of single electron wave packets, provides an ideal platform for measuring the temporal evolution of excitations in materials on the picosecond time scales. In an experiment reported in PNAS, Yurtsever and Zewail (7) use Kikuchi bands as sensitive measures of the tilt of crystallographic planes in single crystal silicon. As illustrated in Fig. 1 (adapted from their discussion), these bands arise when a primary electron scatters in the specimen, producing a broad scattering distribution around its initial direction. Some of these electron paths satisfy Bragg scattering conditions for particular sets of planes, producing bands of intensity in the far field (8). Kikuchi bands are rigidly tied to the orientation of the planes themselves, whereas diffraction spots have positions defined relative to the unscattered electron beam. Thus, diffraction spot positions are only weakly dependent on crystal tilt, while Fig. 1. Scattering diagram for Kikuchi bands. Subsequent to a diffuse scattering event, some scattered electrons satisfy the Bragg condition (θB) for a local crystal plane. Multiple Bragg scattering then produces a band of intensity, and its position accurately tracks the local tilt of the crystal plane. Intensity falling within a narrow section of the Kikuchi band can be assembled in a map of intensity as a function of scattering angle vs. time, as depicted here, below an imaginary slit that crosses the Kikuchi band intensity (modified from ref. 7).