Temporal resolution in conventional TEMs as described in the previous sections is achieved by using a detector that can be read out fast enough to sample the time evolution of a specimen undergoing reversible or irreversible dynamicsin situ. While conventional full-frame cameras restrict the resolution to the millisecond or sometimes microsecond time scale, some event-driven detectors can time-stamp individual electron events with nanosecond precision (116). Elementary excitations of electrons, lattice and spins, however, evolve on femto- to picosecond time scales.
In order to reach the time resolution required for imaging this kind of dynamics, ultrafast TEMs typically use pulsed electron sources and specimen excitation in a laser pump/electron probe scheme. First, an ultrashort laser pulse (pump) is used to drive the specimen out of its equilibrium state. Subsequently, one or more ultrashort electron pulses (probe) are used to image the transient state of the specimen after a temporal delayΔ𝑡(120,121).
In more detail, two different approaches to time-resolved TEM have emerged. From the 1980s on, Bostanjoglo et al. developed the first time-resolved TEM with 30 ns temporal resolution (122,123). It was able to record three subsequent images of irreversible dynamics using three individual photoemitted electron pulses after a single impulsive laser excitation of the specimen (124). An electrostatic beam deflector was used to record the image series on a single detector frame, thus circumventing the need for a fast detector. Thissingle-shotapproach was further developed at Lawrence Livermore National Laboratory (LLNL) and termed dynamic transmission electron microscopy (DTEM) (125,126). In its latest reported state, the DTEM system records an array of up to16 × 16sub-images on the detector with down to 10 nm spatial resolution using 108to 109electrons in 15 ns pulses (127–129). Combined with a compressive sensing approach, even more sub-images can be encoded in the same detector frame (129).
In contrast, ultrafast transmission electron microscopy (UTEM) uses astroboscopicpump/
probe scheme to record movies of reversible dynamics. The image at every single temporal delayΔ𝑡is integrated over millions of consecutive optical pump and electron probe pulses. In contrast to the DTEM approach, UTEMs are mostly operated in asingle-electron regime, in which Coulomb repulsion within the individual electron pulses can be neglected. This allows
2.7 Ultrafast transmission electron microscopy
for the realization of UTEMs with femtosecond electron pulse durations and a spatial resolution comparable to that of conventional TEMs (120).
The first stroboscopic UTEM was presented by the group of Ahmed Zewail at the California Institute of Technology (Caltech)3 (83, 130, 131). Since then, various implementations of the UTEM methodology have been put into operation in laboratories around the world (66, 132–146). Despite the fundamentally identical working principle, a decisive influence on the electron pulse properties is found in the specific design of the electron source. The most common implementation employs a flat photocathode to emit a single ultrashort electron pulse for each ultrashort laser pulse absorbed by the emitter. A typical choice, due to a low work function of only 2.4 eV, is a lanthanum hexaboride (LaB6) emitter in a truncated-cone shape (120). These emitters are often used in conventional TEMs where they are operated in a thermionic regime. A continuous electron beam is emitted from the LaB6surface when heated by an electrical current flowing through the filament holding the emitter. In ultrafast operation, the work function of the cold emitter is overcome by the photon energy of the photoemission laser. Besides the original implementation at Caltech, LaB6-based UTEMs are operated, for example, at EPFL in Lausanne (133), at the University of Minnesota (136) and at KTH in Stockholm (138).
However, with typical diameters of tens of micrometers, LaB6 emitters emit a relatively incoherent electron beam due to the large effective source size. Applications demanding a high source coherence, such as electron holography, coherent nanodiffraction or quantum optics experiments are therefore challenging in thermionic UTEMs (120,147,148). In conventional TEMs, beam coherence could be increased by spreading the beam in the gun or the illumination system. In the case of UTEM, however, the associated loss in beam current is often not tolerable, which led to the development of high-coherence electron sources based on metallic nanotips (149,150), for example, at the University of Göttingen (66), at the University of Toulouse (140) and at the Chinese Academy of Sciences (143).
The Göttingen UTEM is based on a JEOL JEM-2100F Schottky field emission microscope (66) and was the first to use a tip-shaped photocathode in time-resolved operation (Fig.2.1A to C). The Schottky field emitter shares the advantage of thermionic emitters that it can either be operated in a continuous mode at a temperature of 1800 K and extraction fields on the order of 0.5 V/nm to 1 V/nm at the tip apex, or in photoemission mode. When operating the emitter below 1400 K, continuous electron emission is fully suppressed and photoelectron pulses are generated by the photoemission laser focused onto the tip apex. The emission process is supported by a zirconium oxide overlayer, reducing the work function to about 2.9 eV at the(1 0 0)front facet of
3The Zewail group used the abbreviation UEM (for ultrafast electron microscopy) instead of UTEM.
Figure 2.6: Exemplary experimental results achievable with the current status of the Göttingen UTEM instrument. (A) High-resolution TEM (HRTEM) micrograph of Au/Pd particles on an amorphous carbon film. Visible lattice planes with 2-Å spacing demonstrate the resolution capabilities of the modified instrument (here: using thermal electron emission). Inset: Fourier transform of a four times larger sample region. (BtoI) Measurements acquired with photoelectron beams (typical acquisition times 5–60 s) and at an electron energy of 120 keV. (B) Bright-field image of an ultra-microtomed 50 nm thin sample of 1T-TaS2showing bending contrast of the thin-film membrane. Close-up: drop-casted gold nanorod on the sample surface. (C) Lorentz imaging provides magnetic contrast in UTEM as demonstrated for permalloy islands on a silicon nitride support (see also Ref. (151)). The out-of-focus image reveals the existence of a magnetic vortex in each of the four islands (visible as black and white features, respectively, depending on vortex orientation). The magnetic structure of a single vortex is schematically depicted in the upper panel. (D) Diffraction pattern of the charge-ordered phase of an ion-polished PCMO (Pr0.7Ca0.3MnO3) plan view sample. Weak superstructure spots are visible halfway between the lattice reflections. (E) Diffraction pattern of the nearly commensurate charge density wave (NC-CDW) phase of 1T-TaS2. The first-order NC-CDW diffraction spots are hexagonally arranged around structural reflections. (F) Convergent beam electron diffraction (CBED) pattern of an exfoliated 100 nm thick single-crystalline graphite flake. (G) High dispersion diffraction pattern of a 463 nm spaced grating replica, demonstrating 1.2-µm transverse coherence lengths. (H) Electron hologram obtained using a Möllenstedt biprism at a filament voltage of 9 V, emphasizing the photoelectron coherence properties achievable in the UTEM. (I) Electron energy loss spectra of 1T-TaS2and PCMO. Inset: zero-loss peak (ZLP) with a FWHM of 0.6 eV. Figure and caption reprinted from Ref. (66).