Terahertz-driven, all-optical electron gun

Ultrashort electron beams with narrow energy spread, high charge, and low jitter are essential for resolving phase transitions in metals, semiconductors, and molecular crystals. These semirelativistic beams, produced by phototriggered electron guns, are also injected into accelerators for x-ray light sources. The achievable resolution of these time-resolved electron diffraction or x-ray experiments has been hindered by surface field and timing jitter limitations in conventional RF guns, which thus far are<200 MV/m and>96 fs, respectively. A gun driven by optically-generated single-cycle THz pulses provides a practical solution to enable not only GV/m surface fields but also absolute timing stability, since the pulses are generated by the same laser as the phototrigger. Here, we demonstrate an all-optical THz gun yielding peak electron energies approaching 1 keV, accelerated by 300 MV/m THz fields in a novel micron-scale waveguide structure. We also achieve quasimonoenergetic, sub-keV bunches with 32 fC of charge, which can already be used for time-resolved low-energy electron diffraction. Such ultracompact, easy to implement guns driven by intrinsically synchronized THz pulses that are pumped by an amplified arm of the already present photoinjector laser provide a new tool with potential to transform accelerator based science.


Results and discussion
The electron momentum gain, , can be expressed as ( where is the THz vector potential, is the emission time (i.e. delay), and is the time the electron exits the PPWG. The approximation ( ) ≈ ( ) is valid because ≫ in our setup (shown later). To determine the optimum emission time for acceleration, we record the electron energy gain ( ) spectra and bunch charge versus delay in Figures fig:specgram(a)-(b). The UV emitter can precede (<2 ps), overlap (-2 to 2 ps), or succeed (>2 ps) the THz pulse. In the overlap region (-2 to 2 ps), maps out the phase and amplitude of ( ), similar to THz streaking in gases [Fruhling2009]. One exception is that between -0.25 and 0.4 ps, emission occurs in the positive half-cycle of the THz field, causing a suppression of charge and energy gain. Two delays are selected to be the operating points of the gun. The first delay, 1 = −2 ps, produced the highest peak acceleration while the second delay, 2 = 0.8 ps, produced the most monoenergetic spectra. The total bunch charge was 40 fC at 1 and 32 fC at 2 .
When the emission precedes the THz pulse (<2 ps), a large energy spread centered about ~0.45 keV is observed. The origin of these broadened spectra, enduring for long decay times, is attributed to multiple complex mechanisms encompassing thermal [Herink2014] or time-of-flight effects. Further discussion is provided in Supplementary Information. When the emission succeeds the THz pulse (>2 ps), there is no net acceleration from that pulse. The constituency of electrons slightly elevated to 50 eV is attributed to the aforementioned decay effects probed by a backreflected THz pulse arriving at 18 ps (Supplementary Information). In Figures fig:specgram(c)-(d), we take a closer look at the energy spectra from the two operating points, 1 and 2 , for three different THz energies, . Each spectrum exhibits a unimodal distribution with an average energy gain increasing with . Except for the = 35.7 μJ spectrum at 1 , the spectral shapes are asymmetric with a pedestal toward lower energies and a maximum yield toward higher energies, followed by a sharp cutoff, akin to the shapes observed in RF accelerators [Warren1983]. The 6 high yield near the cutoff indicates that most electrons are emitted at the optimal THz phase and concurrently experience the same acceleration. The pedestal can be attributed to electrons emitted away from the optimal phase, resulting in a lower energy gain.
We continue investigations at 1 and 2 by plotting versus on a spectrogram ( Figures   fig:scaling(a)-(b)) and scatter plot ( Figures fig:scaling [Harris2011] and would be the case in this study for larger field or reduced PPWG spacing.

Simulations
In Figure

Conclusion
In conclusion, we demonstrated high field (>300 MV/m), quasimonoenergetic (few percent spread) THz acceleration of multi-10 fC electron bunches to sub-keV energies in an ultracompact, robust device. No degradation in performance was observed over 1 billion shots. While the operating pressure was 40 µTorr, no change in performance was observable up to 10 mTorr. This first result of a jitter-free, alloptical THz gun, powered by a few-mJ laser, performs in accordance with underlying simulations and is encouraging for future developments. In its current state, it can be used for time-resolved LEED andwith modest improvements in laser stability and -for time-resolved electron energy-loss spectroscopy (EELS) [Piazza2014]. Further improvements on the gun structure and THz field promise relativistic electrons [Fallahi2016].

Delay scan
In order to have a fuller understanding of the electron dynamics induced inside the gun by the THz and UV pulses, we acquire a spectrogram over a wide range of delays in Figure fig:longdelay(a). Between -2 and 2 ps, the electron spectra change rapidly with respect to delay due to the temporal overlap with the main THz pulse. Here, the spectra are narrowband and the momentum gain follows the vector potential of the THz field, as described in the main text.
When the UV pulse precedes the THz pulse (<-2 ps), we observe broad, elevated electron spectra enduring over a long delay window to nearly -50 ps. A number of physical processes may contribute to this behavior. Detailed investigations will be the topic of a forthcoming article.
One possibility is thermally-assisted THz field emission, a process investigated in [Herink2014] and more generally in [Fujimoto1984,Hohlfield2000].

THz spatiotemporal properties as a function of energy
In Figure fig:scaling, the THz energy, , was varied by changing the IR pump energy and measured using a pyroelectric detector. Since the acceleration process depends on the spatiotemporal properties of the THz beam, it is important to verify that there are no significant distortions in the THz temporal and spatial profiles as the energy is changed. It is also important to verify that the THz field strength scales proportionally as the square root of the THz energy.
We first measure the temporal profile via EO sampling for various THz energies in Figure   fig:spatiotemporal(a). Aside from scaling in field strength, the shape of the temporal profile has little variation as a function of THz energy, with the carrier-envelope phase and pulse duration remaining roughly constant. In Figure fig:spatiotemporal(b), we sample the THz field as a function of THz energy at two peaks of the waveform as labelled in Figure fig:spatiotemporal(a): Peak 1 (black dots) and Peak 2 (gray dots). The field strengths at these two peaks correspond to the maximum accelerating fields experienced by the electron in the gun. The data fits well to a square root function (dashed lines), thus verifying that the peak accelerating field scales with the square root of the THz energy.
Next we measure the spatial profile of the THz beam at the freespace focus using a THz camera for various energies in Figure

Calculation of coupling efficiency into the gun
Here, we show how the THz coupling efficiency into the gun, , can be calculated from the measured power transmission data, , shown in Figure fig:vna (blue line).

13
First, we make the assumption that the out-coupled freespace mode, denoted by ( , , ) and shown in Figure fig:ccalc(b), varies minimally for different PPWG spacings, . This has been validated by EM simulations for values of within our region of interest: 0 < < 200 μm. This mode is the beam which couples most efficiently from freespace into the TEM mode of the PPWG, with a coupling efficiency denoted by . Our THz beam in-coupled into the PPWG can be approximated as a fundamental Gaussian beam, denoted by ( , , ) and also shown in Figure fig:ccalc(b). The amount power in the ( , , ) component of the ( , , ) mode, as a fraction of the total power, is denoted by . Using EM simulations, we determine by computing the overlap integral over a chosen plane normal to y: Note the integrand is scalar because the modes have only one and the same polarization. We obtained a result of = 0.8. Second, we assume that the large majority of transmission losses come from wall ohmic losses inside the PPWG (region 2↔3 in Figure fig:ccalc(a)) and from reflections at the interfaces between the PPWG and taper sections (regions 1↔2 or 3↔4 in Figure fig:ccalc(a)). We proceed to express the power transmission and reflection at the interfaces as follows.
Reflection 2 → 1: 1 − Also, the propagation along 2↔3 induces ohmic losses. The propagation efficiency of one pass is denoted by Propagation 2 ↔ 3: Using EM simulations with finite conductivity surfaces, we determined = 0.83. We can now calculate the total transmission through the structure: Transmission 1 → 4: After some algebraic simplification, we obtain

THz source
We employed the tilted pulse front (TPF) pumping technique in a 5.6% MgO-doped congruent lithium niobate (LN) crystal [Hebling2002] cooled to 100 K. The IR pump beam is diffracted off a 1500 l/mm grating to acquire a TPF, which is then subsequently imaged-in the tangential direction-onto the LN using a 150 mm cylindrical lens. Another cylindrical lens with a sagittal focal length of 100 mm was used to shape the impinging pump beam for highest efficiency. The extracted optical-to-THz energy conversion efficiency was near 1.0% with 35.7 µJ of THz energy. We used a Gentec SDX-1152 calibrated pyroelectric THz joulemeter to measure the THz energy. Concurrently, a thermal power meter (Ophir Optronics) measured 18 mW at 1 kHz. A Spiricon Pyrocam IV camera was used to image the THz beam. Shot-to-shot energy fluctuation was 2%.

EO sampling
We employed an oscillator-based EO sampling setup since the pulses from the amplifier were too long to effectively probe the THz waveform. Oscillator probe pulses were overlapped with THz pulses on a 200 µm, 110-cut ZnTe crystal. The probe pulses sample the THz-induced birefringence as a function of delay and are subsequently interrogated by a quarter-wave plate, polarizer, and photodiode combination, as typical [Wu1995]. Because of the much higher repetition rate of the oscillator pulse train, a boxcar averager (SRS SR250) was used to electronically gate out the pulse that overlapped with the THz. The EO crystal mount was custom fabricated such that the crystal is in the same position as the center of the gun for the PPWG-center measurement in Figure fig:char(c). For the PPWG-thru measurement, the gun was placed in its operating position and the transmitted THz beam was image-relayed via two additional parabolic mirrors onto a second focus, where the EO crystal was then placed.

UV source
16 The UV photoinjection was obtained by frequency-quadrupling the fundamental 1030 nm pump. The first second-harmonic generation (SHG) stage consisted of a 0.5 mm thick type I BBO crystal with φ=23.7°, generating approximately 3 µJ of 515 nm pulses. The second SHG stage consisted of 0.5 mm thick type I BBO crystal with φ=44.6°, generating 600 nJ of 258 nm pulses. The conversion efficiency from fundamental to UV was about 7%. A CaF2 prism was used to spatially separate the various wavelengths.
The prism-induced dispersion over the subsequent 0.5 m propagation was determined through calculations to cause negligible increase of the pulse duration. The UV energy impinging the copper photocathode was 270 nJ. Both nonlinear conversions were in the unsaturated regime and the phasematching bandwidths of the two BBO crystals are broader than the spectral bandwidths of both the 1030 nm and 515 nm pulses. Therefore, we estimate of the UV pulse duration as roughly half that of the fundamental. The focused UV beamwaists on the photocathode were 20 µm (x) and 60 µm (y). Two parallel plates, fashioned with 18° tapers, were fabricated separately and afterwards sandwiched together with high-precision Kapton shims in-between to set the spacing and enforce parallelicity. We ultimately operated with a shim thickness of 75±15 µm after optimization (see next paragraph). EM simulations were performed in HFSS to obtain the optimal taper angle for efficient coupling. Fabrication of the THz waveguide was performed in-house using conventional machining tools. A flatness tolerance of 5 µm over a 1 in 2 area was specified for the parallel plate sections. A 9V reverse bias was applied across the plates to help with electron extraction.

THz gun design and characterization
A THz network analyzer was used to characterize the power transmission, , through the PPWG for various spacings, as shown in Figure

Photocathode
A 25 nm copper film coated on a UV-grade quartz substrate was used as the photocathode. The coating of the photocathode was performed in-house by evaporative physical vapor deposition. A chromium adhesion layer of a few nm was first deposited onto the substrate. In addition to functioning as a photocathode, the copper film functions as one of the PPWG plates. To minimize interface losses, the surface of the photocathode is placed flush with the surface of the parallel-plate structure to a tolerance of a few microns using an optical-flat mirror surface. To minimize THz diffraction losses through the thin film, the film is thickened to 125 nm (equal to the copper skin depth at 0.5 THz) along the THz propagation path from the PPWG input until ~0.25 mm before the 25 nm thick photoemission region. To ensure electrical connectivity between the photocathode and the PPWG, the copper film extended around to the edges of the quartz substrate and the edges were in contact with the PPWG.

Exit anode
The exit anode was cut out of a slab of 100 µm polished stainless steel shim stock. The precise slit width of 20 µm over a 2 mm length was achieved by picosecond laser micromachining. An optical microscope was used to verify the dimensions to within a tolerance of 2 µm. EM simulations in Figure fig:char(a) confirm that, with a width of 20 µm ( 33 ⁄ ), the slit causes minimal distortion to the THz field distribution.

Electron detection
Following its exit from the gun, the electron bunch drifts into a retarding field analyzer (RFA) [Brunner2013], consisting of a channel electron multiplier (CEM) (Photonis, Inc.) and two static, uniform field regions formed by two biased mesh electrodes. The first region (between the gun and first electrode) boosts the electron energy by 300 eV to enhance the detection efficiency of the CEM. The second region (between first and second electrodes) acts as a highpass filter for the electron energy by retarding the electron trajectory using a variable bias − . Electrons having energy less than are repelled by the electrode while those with more energy pass through, being subsequently detected by the CEM. Each spectrum was collected by taking the derivative of the measured current with respect to . The intrinsic energy resolution of the analyzer is about 2 eV. With post-process smoothing, the effective resolution is about 16 eV.
The meshes were TEM grids (Ted Pella, Inc.) with a thickness of 13 µm and a pitch of 12.5 µm. Each mesh had a transmission of 36%, so the total bunch charge was determined from dividing the detected charge by 0.36 2 = 0.13. Each mesh was placed on 100 µm thick stainless steel soldering plates and sandwiched by 500 µm PEEK shims which enforce their spacing and parallelicity as well as providing electrical isolation. The soldering plates contained "fingers" on which high voltage biasing wires were soldered. The RFA was placed 1.5 mm from the exit anode of the electron gun, measured by the distance between their nearest planes.
Absolute charge measurements were obtained by rewiring the grounded input terminal of the CEM to a Keithley 6514 picoammeter and turning off the CEM bias. In this configuration the CEM essentially acts as a Faraday cup. Secondary electron emission on the CEM is not taken into account, but it would only increase the total charge count if it were. The picoammeter had a RMS noise level of 300 fA.

Network analyzer measurements
Our vector network analyzer (VNA) setup for characterizing the PPWG power transmission consisted of an Agilent E8363B and millimeter wave extender V03VNA2-T/R with 70 dB of dynamic range at 0.220-0.325 THz. The transmitter and receiver were connected to corrugated horns designed for coupling the VNA waveguide mode to a free-space Gaussian mode with a waist of 6 mm. We placed two THz polyethylene lenses with focal lengths of 25 mm a distance of 2f apart between the transmitter and receiver and set the background level. Given the waist of 6 mm (diameter of 12 mm) and focal length of 25 mm, the f-number is about 2, which is well-matched to the optimal f-number of our PPWG's taper section. For the VNA measurements we replaced the photocathode with a polished aluminium blocks to eliminate diffraction losses in the PPWG. The PPWG was placed in the center between the two THz lenses and the PPWG transmission was characterized. The PPWG spacing was varied by changing the thickness of the Kapton shims between the two plates (see Methods -THz gun design and characterization).

Particle tracking simulations
3D particle tracking simulations incorporating space charge were used to model the electron bunch evolution in the presence of the THz field. The emitted electron bunch had a Gaussian spatial profile with beamwaists of 20 µm (x) and 60 µm (y) and a Gaussian temporal profile with FWHM of 275 fs. The initial kinetic energy was 0.18 eV (equal to the excess energy) with a uniform momentum distribution over a half-sphere [Dowell2009]. 5000 macroparticles were used to represent a total bunch charge of 32 fC, corresponding to a charge of -40e per macroparticle. The trajectories were modeled by integrating the kinematic equations for every particle using a 4th order Runge-Kutta solver: Here, is the relativistic mass, is the electric force due to the THz pulse, is the electric force due to the 9V reverse DC bias, , is the force on the th particle due to the th image particle, and , is the particle-particle Coulomb force. The THz beam was modeled as a plane wave with a Gaussian distribution in amplitude in the x direction. The THz waveform in was directly imported from the EO sampling trace.