- Top of page
A compact microfocus x-ray source based on a high repetition rate, ultrafast laser-triggered field emission cathode is characterized. Bremsstrahlung and Cu Kα x-ray pulses are generated in a transmission geometry from 20-keV electrons incident on a thin Cu target. Timing correlations between optical and x-ray pulses show that the precision timing properties of the optical pulse train are transferred to the x-ray pulse train. Simple electron optics allow a demonstration of point-projection x-ray imaging. Modelling of the electron trajectories shows that sub-µm spot sizes and sub-100-fs x-ray pulse durations are feasible with improved electron optics.
In this Letter, we demonstrate the generation of high repetition rate, pulsed Cu Kα x-rays in an ultrafast x-ray tube based on fs laser-induced electron emission from a field emission tip. Ultrafast x-ray tubes based on photoelectron emission from surfaces have been previously demonstrated [1-4] but require ∼100-µJ pulses from amplifier systems and suffer from space-charge effects that lengthen the x-ray pulse duration. Electron emission from fs laser-triggered tips is actively being studied in several groups [5-10]. These electron sources have nanometric emission areas, leading to a high brightness electron beam that can in principle be focused to a small spot; field emission tips have already been used as sources for DC microfocus x-ray tubes with x-ray emission areas down to 100 nm. Furthermore, field emission tips may be triggered by nJ pulses from a fs laser oscillator without amplification, allowing for high repetition frequencies and mitigating space-charge effects. Applications of such a source include time-gated medical imaging to reduce radiation dose rates, secure satellite communications , and ultrafast time-resolved x-ray studies of materials[14, 15].
The experimental setup is shown in Fig. 1. To generate fs electron pulses, a Ti:sapphire laser's pulse train (150-MHz repetition frequency frep, <10-fs pulse duration, 800-nm central wavelength) is focused ( intensity radius ∼3 µm) onto a field emission tip in ultrahigh vacuum ( torr). The single-crystal tips are either (310)-oriented HfC or W, or (111)-oriented W, each with ∼100-nm radius of curvature. For the results reported here, choice of tip does not affect timing or spectral properties of the x-rays, but W tips yield more current and x-rays than HfC tips. To measure the current (“electron configuration”), the tips are held at −600 V while thin electrodes (E1–E3) are held at ground potential and the x-ray target (XRT) is removed from the electron beam. E1–E3 are spaced 1.5, 3.6, and 5.7 mm from the tip's apex and have apertures of diameter 1.6, 2.4, and 2.4 mm, respectively. A dual-chevron configuration microchannel plate detector (MCP, pore size 10 µm, active area diameter 40 mm) is placed 84 mm from the tip's apex. For 10 mW of average laser power, 0.2 electrons/pulse are detected when HfC is illuminated, and for W, 70 mW of power yields 5 electrons/pulse.
To generate ultrafast x-rays (“x-ray configuration”), the XRT is inserted 28.6 mm from the tip's apex. It is comprised of a 12.7-mm diameter, 100-µm thick piece of Be foil onto which a 500-nm thick layer of Cu has been evaporated. The Cu and Be thicknesses have been chosen to stop 20-keV electrons in the Cu layer while transmitting of Cu Kα x-rays. The XRT is held at +19.3 kV, while the tip is held at +100 V so that stray primary electrons cannot reach the grounded front face of the MCP. In x-ray configuration, E1 and E3 are typically held at +700 V while the potential of E2 can be adjusted away from +700 V to focus the electrons.
To characterize the x-ray pulse timing, we correlate individual x-ray detection events with pulses from the laser detected with a fast photodiode (PD). The electrical impulse from the front face of the MCP is separated from the DC current by a fast bias tee, amplified, and used to trigger the start time of a time-to-amplitude converter (TAC) via a constant-fraction discriminator (CFD). The TAC's stop trigger is activated by the next leading edge of a square wave derived from the PD's pulse train. The TAC output is binned using a multi-channel analyzer (MCA), providing the electron and x-ray timing correlations with respect to the optical pulse train as shown in Fig. 2. Gaussian fits to the data with values greater than of the peak value give full-width half-maxima (FWHM) of 54 ± 1 and 60.3 ± 0.8 ps, respectively.
Figure 2. Timing correlations for electrons and x-rays, each with respect to the optical pulse train. MCP timing jitter is the dominant contribution to the 60-ps widths.
Download figure to PowerPoint
The timing correlation widths are limited by the detection process: 60-ps jitter from the MCP is consistent with . This width is much larger than physical sources of broadening prior to detection: transit time differences due to electrons taking different trajectories (26-degree emission half-angle and 0.5-eV energy spread lead to 8 ps and 100 fs, respectively, under our experimental conditions), the electron emission time scale (10 fs), and multiple scattering in the x-ray target (estimated stopping time for 20-keV electrons is 10 fs). Jitter in the TAC system is negligible; directly-digitized MCP output pulses were fit to template waveforms and timing correlations with widths similar to Fig. 2 were obtained.
The x-ray pulse timing is expected to follow the exceptional stability properties of the Kerr mode-locked laser source. We characterized the timing stability by determining the times associated with the timing correlation distribution peaks (as in Fig. 2) for many consecutive time windows. For example, Fig. 3(a) shows the times of the x-ray arrival time distribution peaks for 300 consecutive 10-s data collection windows. Figure 3(b) shows the Allan deviation of these times as a function of data collection window size. For 1 s s, the stability averages down as ∼5 ps , statistically limited by the detected count rate of 220 x-rays/s. The RF spectral content of the x-ray pulse train follows that of the fs laser pulse train. As an example, Fig. 3(c) shows the second harmonic of frep (chosen for its high frequency that is still within the MCP's rf detection bandwidth) measured with the MCP detecting the x-ray pulse train. For this data ∼630 x-rays/s were detected. Nevertheless, the precision timing of the x-ray pulse train allows detection with 25 dBc/Hz SNR at 300 MHz (far from near-DC noise sources).
Figure 3. Timing stability of the detected x-ray pulses. (a) 3000-s record of x-ray peak arrival times, determined from 10-s data windows. (b) Allan deviation of the x-ray arrival time. A stability of 0.3 ps corresponds to determining the location of the x-ray spot to within 100 µm. (c) Harmonics of frep in the x-ray pulse train are detected by the MCP.
Download figure to PowerPoint
The x-ray energy spectrum consists of a broad Bremsstrahlung background with a narrow Cu Kα peak at 8 keV, as shown in Fig. 4. We measure the spectrum using a room-temperature reach-through avalanche photodiode (APD) (1-keV FWHM energy resolution near 8 keV). Drawing a smooth curve through the Bremsstrahlung background allows us to estimate the APD detects 2500 Cu Kα x-rays/s when the tip emits 1.9 electrons/pulse. The detection efficiency is 0.84 for 8-keV x-rays (from the absorption of 130-µm-thick Si ) and the 3-mm diameter APD is placed 16.1 mm from the XRT point-source. Therefore we estimate the XRT emits 1.4 × 106 Cu Kα x-rays/s isotropically into 4π. Similarly, we estimate 3.3 × 106 Bremsstrahlung x-rays/s with energy between 3 and 19.3 keV are emitted, assuming isotropic emission. This flux approaches the ultrafast x-ray diode of  (107 Cu Kα x-rays/s) when that system is operated at similar target voltages with sufficiently low space-charge to keep the pulse duration below 10 ps.
Figure 4. X-ray energy spectrum: 1-keV resolution. The 2.5-keV measurement threshold is set due to APD noise and the data is not corrected for transmission of 100-µm-thick Be foil on either the XRT or the APD. It is also uncorrected for the detector response, which is essentially the absorption of 130-µm-thick Si.
Download figure to PowerPoint
We characterize the x-ray source's spot size by imaging the APD assembly's edge and coaxial cable, which cast x-ray shadows as shown in Fig. 5. For this image, the exposure time was 200 s, and a background due to scattered laser light has been subtracted. The sharpness of the shadow's edge is used to infer the x-ray spot size, which we determine to be ∼360 µm (FWHM) by fitting to the shadow transition and accounting for a factor of 3.3 image magnification. The spot size is limited by aberrations in the simple electron optics; we expect a 100-nm spot size could be achieved with an improved design such as that in .
Figure 5. Point-projection shadow of the APD assembly. The vertical edge just to the right of the central portion of the image corresponds to the well-defined smooth vertical edge of the APD detector's ceramic support substrate.
Download figure to PowerPoint
In order to estimate the timing jitter, and confirm the spatial width, of the electrons focused onto the XRT, we have simulated electron trajectories in both the electron and x-ray configurations . Space-charge effects are neglected since the experiments are conducted with ∼1 electron/pulse. We find that for a broad range of voltages on E2 and E3 the electron pulse duration at the XRT is ∼10 ps, due largely to the spread in electron emission angles (uniform emission within a 26-degree half-angle at the tip's surface, as observed in the electron configuration). The simulations show a quadratic dependence of the electron arrival time with respect to emission angle. Other tips have been shown to exhibit significantly narrower (5-degree) emission half-angles which would yield a 360-fs, 3-µm source with our proof-of-concept electron optics. Furthermore, these simulations indicate that operation at higher voltages with improved electron optics can shorten this time scale to <100 fs.
It is interesting to consider extending these methods to more-optimized parameters. For example, other fs laser-triggered tips have been reported to emit >100 electrons/pulse , and increasing the XRT voltage from 20 to 60 kV would increase the conversion efficiency from electrons to Cu Kα x-rays by a factor of 10 (as well as increase the bremsstrahlung conversion efficiency). Together with a 1-GHz frep, this would yield a Cu Kα flux of 1010/s. Reducing the spot size and pulse duration will require electron optics optimized for both spatial and temporal focusing of the electron beam onto the target. We expect the optimized source can produce 1010 Cu Kα x-rays/s from a 100-nm spot size with a timing jitter below 100 fs. Such an x-ray tube equipped with a fs laser-triggered field emission cathode would have both average flux and peak brightness (6 × 1014 photons/s/mm2/mrad2/0.1%BW) comparable with plasma-based sources employing mJ- and kHz-scale lasers.