Considerations for future UTEM instruments

The progress in conventional TEM over the last decades procured electron sources aiming for the highest possible brightness by maximizing the transverse beam coherence and minimizing the beam’s energy spread. Both are achieved by employing nanoscale tip-shaped field emitters. In order to transfer the wealth of well-established TEM techniques to the ultrafast timescale, laser-triggered field emitters display most promising properties (cf. Fig.4.7). In contrast, large-area photocathodes are particularly useful for applications that require only minor transverse beam coherence but benefit from high pulse charges.

While Ch.4constitutes a leap forward for coherent beam applications in UTEM, future development will warrant further improvement of the beam quality in terms of transverse coherence, pulse durations and beam current, which is discussed in the following. The underlying central question is what the emitter should be designed for: either optimized single-electron pulse operation or the flexibility for upscaling the beam current.

Optimized nanoscale electron emitters If single-electron pulses are the central design objective, an emitter with lowest possible transverse emittance is desirable. That would require a reduction of the initial transverse momentum spread of the electrons or further reduction of the emitter size. The minimum intrinsic kinetic energy distribution of the photoelectron beam in Ch.4is0.6 eV, which includes the overall short-term stability of the instrument. Calculations indicate the possibility to decrease the energy spread of the laser-driven Schottky field emitter [158], which is ultimately limited by the width of the

Fermi-Dirac distribution∆Ethermal≈2k_BT≈260 meVatT =1500 Kand employed laser spectral bandwidth of about60 meVfor 30-fs laser pulses at 400-nm central wavelength (cf.

Fig.3.7). Furthermore, sharper metal tips (r0<10 nm) feature a smaller effective electron source size and improved beam coherence, but were so far only capable of producing low electron yields in single-photon photoemission, e.g. Ne=0.03electron per pulse [157].

Further development of tunable tip-based ultrafast electron sources is needed, paving the way towards point-source-like monochromatic emission characteristics. Despite the enormous advantages of single-electron pulses without Coulomb repulsion, larger tip emitters as used in this thesis withr₀∼100 nmhave the advantage of a high stability and allow for easy upscaling of the electron current.

For classical photoemitters, the shot-to-shot distribution of the electron number is given by Poisson statistics. A mean of hNei=1 results in a probability P(N>1)=26%of finding at least 2 electrons in an individual pulse, and only for hNei=0.15, P(N >1) drops to 1%. To avoid these effects, a photon-triggered single-electron source would be needed, as already demonstrated in tunneling currents from single quantum dots [350].

Photoemission from nanoscale tips aims for the ultimate control of the electron’s phase space, but so far does not treat the electron spin degree of freedom. New materials, as demonstrated for planar photocathodes [334], might yield spin-polarized electron beams.

Gun geometry The Göttingen UTEM instrument is based on a custom modified JEOL JEM-2100F Schottky field emission TEM. In continuous operation of the Schottky emitter, the beam brightness is optimized by radially cutting the transverse electron phase space distribution multiple times, thus reducing the beam current from∼100µA to thepA−nA range. The electron transmission from the tip emitter to the sample is drastically increased by individually adjusting the voltages applied to the electrostatic gun electrodes (cf. Ch.4).

Additionally, the original relatively blunt tip (r0∼430 nm) was replaced to improve the beam emittance at constant currents. These tailored operation conditions illustrate the opportunities of a customized ultrafast electron gun design. In order to reduce the impact of space charge and dispersion of the electron pulse, high acceleration fields should be applied.

Simultaneously, the overall gun transmission must be optimized to achieve high beam currents at space charge-free operation. Finally, the correction of temporal aberrations should be considered (e.g. off-axis electron trajectories accumulate a small relative time delay).

Laser

generation THz generation THz generation

Detector 200 µm

a b

Figure 7.2:Electron pulse compression using THz-fields. (a) Schematic experimental setup with synchronized photoelectron pulse generation (930-fs initital duration) and two THz butterfly resonator structures for pulse compression and streaking. (b) Electron pulse duration (FWHM) at the streaking stage for varying incident THz fields used for compression (inset: shortest measured pulse profile of75 fs). From [229]. Reprinted with permission from AAAS.

Shaping & compression of electron pulses Ultrashort pulse generation by linear photoemission adds a large flexibility to freely structure the emitted electron density on a sub-pstostimescale. Such a capability is especially useful ifns-processes, like magnetic vortex gyration [351] and skyrmion dynamics [352] are of interest, allowing for three orders of magnitude higher beam currents at fixed repetition rate without beam degradation.

In ultrafast electron diffraction setups, the use of RF-gun and RF-compressor technology allows for a tailoring of the longitudinal electron phase space on thefs- tonstimescales.

The same concept was employed in early UEM implementations [124] and is also proposed for future UTEM designs using high-charge electron pulses from a planar photocathode [119,120]. In a similar manner, RF-compression could significantly improve the peak brightness of linearly chirped electron pulses from field emitter photocathodes (cf. Fig.4.5).

The high required stability of RF-cavities in a TEM environment has recently been demon-strated by transverse deflection and chopping of a continuous low-emittance electron beam [122,126] and will enable novel applications when used in a time-of-flight type energy spectrometer [230,336]. Nevertheless, RF-technology is challenging to be implemented and is currently still limited to a 30-fs laser/RF-phase synchronization for high-power cavities. A promising alternative is the phase space manipulation by phase-locked optical fields.

Optical (THz-) control of phase space density in the point-particle limit Optical bunching of electron pulses requires intense oscillatory fields with a period longer than the electron pulse duration (200 fs−1 psfor the Göttingen UTEM). Here, ultrashort terahertz

a b

Figure 7.3:(a) Ultrashort terahertz (red) and800 nmnear-infrared (blue) pulses are focused onto a metallic nanotip with variable time-delayt₀[231]. (b) Time evolution of the phase space density distribution without THz field (left) and for relative time delay oft₀=0.2 ps(right) (snapshots taken with time differences of50 fs). Note the negative pre-chirping and self-compression of the phase space density [355]. (a) Adapted by permission from Springer Customer Service Centre GmbH: Springer Nature, Nature Physics, [231], © Macmillan Publishers Limited (2014). (b) Reprinted figure with permission from [355] Copyright 2017 by the American Physical Society.

pulses deliver the highest available field amplitudes, e.g. up to13 MV/cmin free space at a190 kHzrepetition rate [353]. Recent experiments could demonstrate the compression of 930 fs(FWHM) single-electron pulses by a factor of 12, down to 75 fs (FWHM) in a 200µm-gap metal resonator (cf. Fig. 7.2) [229]. The compressed electron pulses can be shorter in time than the initial photoemission laser pulse by a broadening of the kinetic energy distribution, conserving the longitudinal phase space volume according to Liouville’s theorem [354].

A novel concept by Wimmeret al.discusses the control of photoemitted electrons with an additional terahertz pulse directly at the field emission tip (cf. Fig.7.3) [231]. By adjusting the relative timing of the optical photoemission pulse and the terahertz field, the longitudinal phase space distribution can be modified. Notably, in such a geometry Liouville’s theorem does not apply due to non-conservative forces acting on the electrons in the spatially and temporally inhomogeneous terahertz near-field. Hereby, elaborate phase space manipulations are feasible, like negative pre-chirping or monochromatization of the electron pulses [355].