Stability of ionization-injection-based laser-plasma accelerators

Laser-plasma acceleration (LPA) is a compact technique to accelerate electron bunches to highly relativistic energies, making it a promising candidate to power radiation sources for industrial or medical applications. We report on the generation of electron beams from an 80 MeV-level LPA setup based on ionization injection (II) over a duration of 8 hours at a repetition rate of 2.5 Hz, resulting in 72,000 consecutive shots with charge injection and acceleration. Over the full operation time the moving averages of the total beam charge of 14.5 pC and the charge between 70–80 MeV did not drift on a detectable level. The largest source of shot-to-shot jitter was in the beam charge (26% standard deviation), which was most strongly correlated with fluctuations in the plasma density (3.6% standard deviation). Particle-in-cell simulations demonstrate that this was chiefly caused by stronger laser self-focusing in higher density plasmas, which significantly increased the ionized charge along with the emittance of the beam. The nonlinearity of this process imposes tight constraints on the reproducibility of the laser-plasma conditions required for a low jitter II-LPA output if self-focusing plays a role in the laser evolution.

Figure 1

Schematic of the experimental setup. Inset top left: measured plasma electron density (black), its rms deviation (grey shaded area) and the density profile used in the simulations (red line).

Figure 2

Measured charge as a function of shot number and run time. The black line shows the 100 shot moving average of the charge with its standard deviation as grey band. A linear fit of all charge measurements is shown in red. Inset: a histogram of the injected charge.

Figure 3

Pointing and divergence of the electron beams. (a) Average 2D-divergence of 200 shots on the profile screen and the corresponding pointing of the single shots (red dots). (b) and (c) 100 shot moving average of the pointing and divergence (black lines) in the nondispersive axis, with their 100 shot standard deviations shown as grey bands.

Figure 4

Electron energy spectra $(\mathrm{d}N/\mathrm{d}E/E)$. (a) As function of shot number and time (waterfall). (b) Average spectrum with its standard deviation (grey area) and the standard deviation of normalized spectra indicating fluctuations independent of charge fluctuations (dotted line).

Figure 5

Correlations of the data using a subset of 3600 shots (5% of total shots). The subplot titles show the correlation coefficients (Pearson’s $r$) of (a) electron density and charge, (b) laser energy and charge, (c) electron density and electron peak energy, (d) laser energy and electron peak energy, (e) charge and electron peak energy. Red markers represent PIC simulation results.

Figure 6

PIC simulation results. (a) Normalized experimental electron spectrum (black) with its standard deviation (grey), the raw simulated spectrum (blue dashes), and the simulated spectrum propagated to the spectrometer screen (solid blue line). (b) Simulated mean energy (solid blue) and rms energy spread (orange) vs $z$. (c) Simulated charge from II vs $z$ at $n_e\pm (1,2){\sigma }_{n_e}$ (bounded by the blue and pale blue shaded areas). The behavior at $n_e$ is shown in black, as it is in all following plots. The peak area of the laser pulse envelope with sufficient intensity to produce ${\mathrm{N}}^{6+}$ is shown for the lowest (dashed line) and highest (solid line) density simulations in grey, in the region where charge is trapped. (d) Simulated $a_0$ evolution at densities $n_e\pm (1,2){\sigma }_{n_e}$. Normalized $n_e(z)$ is shown in orange. The grey dashed lines are the $a_0$ thresholds required to ionize the core electrons of nitrogen according to the ADK model [61]. (e) Laser waist $w_0$ (blue) and length $c\tau$ (orange) vs $z$ at $n_e\pm (1,2){\sigma }_{n_e}$. (f) Normalized slice emittance in $x$ (blue) and $y$ (orange) of II electrons at $n_e\pm (1,2){\sigma }_{n_e}$ at $z=1\mathrm{mm}$. In dashed, solid, and dotted green are the corresponding current profiles at $n_e-2{\sigma }_{n_e}$, $n_e$, and $n_e+2{\sigma }_{n_e}$. In higher $n_e$ simulations the $w_0$ and $c\tau$ are smaller, while $a_0$, $Q$ and ${\epsilon }_n$ are larger.