Professional Documents
Culture Documents
Each successive generation of X-ray machines has opened up reinforce the synchrotron emission over a length of a few metres.
new frontiers in science, such as the first radiographs and The first demonstrations of wakefield-driven radiation using exter-
the determination of the structure of DNA. State-of-the-art nal wigglers have also been reported, though still being limited to
X-ray sources can now produce coherent high-brightness X- optical or near-optical wavelengths and modest peak brightness15,16 .
rays of greater than kiloelectronvolt energy and promise a However, the particles being accelerated in the plasma acceler-
new revolution in imaging complex systems on nanometre and ator also undergo transverse (betatron) oscillations when subject
femtosecond scales. Despite the demand, only a few dedicated to the focusing fields of the plasma wave.
√ These oscillations occur
synchrotron facilities exist worldwide, in part because of at the betatron frequency ωβ = ωp / 2γ , where ωp is the plasma
the size and cost of conventional (accelerator) technology1 . frequency and γ is the Lorentz factor of the electron beam17 . As
Here we demonstrate the use of a new generation of laser- this betatron wavelength (λβ ≈ 2πc/ωβ ) is much smaller than
driven plasma accelerators2 , which accelerate high-charge the period of comparable external wigglers, again as a result of
electron beams to high energy in short distances3–5 , to produce the extremely large electric fields of the wakefield, the wavelength
directional, spatially coherent, intrinsically ultrafast beams at which these particles radiate (of order λ ≈ λβ /γ 2 ) can be in
of hard X-rays. This reduces the size of the synchrotron the X-ray region for γ > ∼ 200, that is, for an electron beam of
source from the tens of metres to the centimetre scale, energy W > 100 MeV. For small transverse amplitude and thus
simultaneously accelerating and wiggling the electron beam. small strength parameter K = γ rβ ωβ /c 1 (that is, undulator
The resulting X-ray source is 1,000 times brighter than limit), the radiation spectrum will be narrowly peaked about a
previously reported plasma wigglers6,7 and thus has the fundamental energy. As K → 1, radiation also appears at harmonics.
potential to facilitate a myriad of uses across the whole For large K 1 (that is, large-amplitude wiggler limit) a syn-
spectrum of light-source applications. chrotron spectrum with broad emission consisting of closely spaced
There are a number of proposals to use extreme nonlinear harmonics is produced17 .
interactions of the latest generation of high-power ultrashort-pulse 2
d2 I 3e 2
laser systems to produce beams of high-energy photons with high E
' Nβ 3 γ 2 2
· K2/3 (E/Ecrit )
brightness and short pulse duration. For example, high-order dEd 2π h̄c0 Ecrit
θ=0
harmonic generation promises trains of coherent pulselets8 and
Compton scattering could extend energies into the γ -regime9,10 . Here, I is the radiated energy, Nβ is the number of oscillations
An alternative proposal has been the use of compact laser–plasma and Ecrit = 3h̄K γ 2 ωβ is twice the energy above and below which
accelerators to drive sources of undulating/wiggling radiation11 . roughly half of the total energy is radiated. K2/3 is a modified Bessel
These accelerators use the plasma wakefield generated by the function of the second kind. For E > Ecrit , the spectrum of radiated
passage of an intense laser pulse through an underdense plasma12 . energy decays exponentially. The radiation is confined to a cone
Such wakefields can have intrinsic fields of more than 1,000 with opening angle θ ≈ K /γ .
times greater than the best achievable by conventional accelerator The first measurements of betatron radiation from a broad
technology, and thus can accelerate particles to high energies in a energy spread electron beam from a laser wakefield accelerator
fraction of the distance. Recently, it has been demonstrated that at have demonstrated X-ray emission with energies up to '1 keV
high laser power, the wakefield can be driven to sufficient amplitude with a peak brightness up to 1019 photons per second per
to be able to trap large numbers of particles (>100 pC) from the mrad2 per mm2 per 0.1% bandwidth6 . Here we demonstrate
background plasma and accelerate them in a narrow energy spread that by entering the highly nonlinear, so-called bubble regime of
beam3–5 , now producing beams of electrons of gigaelectronvolt- laser wakefield acceleration2 , we always and immediately inject a
scale energy of the order of 1 cm (refs 13,14). much larger amount of charge (>100 pC), which is accelerated
Such electron sources are of interest to replace the accelerators to narrow-energy-spread beams with much higher peak energy
that drive current synchrotron sources, and typically use multiple ('230 MeV over '5 mm) than in earlier experiments. This
periods of alternately poled magnets (undulators or wigglers) to results in a 1–100 keV X-ray source with 1,000 times greater
1 The Blackett Laboratory, Imperial College London, London, SW7 2BZ, UK, 2 Center for Ultrafast Optical Science, University of Michigan, Ann Arbor,
Michigan, 48109, USA, 3 GoLP/Inst. Plasmas and Fusão Nuclear, Laboratório Associado, Instituto Superior Técnico, Lisbon, 1049-001, Portugal, 4 DCTI,
ISCTE, Lisbon University Institute, Lisbon, 1649-026, Portugal, 5 Laboratoire d’Optique Appliquée, ENSTA, Ecole Polytechnique, Palaiseau, 91761, France.
*e-mail: stefan.kneip@imperial.ac.uk; zn1@ic.ac.uk.
b 20 mrad
d 0.5 mm
f 0.25 mm
h 0.5 mm
j
20 µm Ag foil
50 µm
5 µm
3 µm
Figure 1 | The quality of the X-ray beam is assessed by measuring its profile and imaging microscopic objects. a,b, Single-shot X-ray beam profile (a) and
sum of five consecutive shot profiles (b) show the imprint of a wire grid and a 4 × 13 mrad2 beam with 5 mrad pointing stability. c–h, X-ray radiographic
images of wire triplets and a resolution test target. The objects contain features as small as 3 µm, which are resolved on the radiographs, indicating a
betatron source of size ≤3 µm. i,j, Photographic images of the smallest wire triplet and resolution test target, from which the various feature sizes
were obtained.
a b c 0.18
Data 1.1 ± 0.6 Data
1.2 8 keV 1 µm 1.0 1.3 ± 0.7 0.16 Model
8 keV 3 µm 1.3 ± 0.5 (Ecrit = 6–10 keV
2 keV 1 µm 1.7 ± 0.6 0.14
1.0 Gauss-top hat)
8 keV 5 µm 0.5 3.1 ± 0.7
3.3 ± 0.7 0.12
Fringe visibility
Intensity (a.u.)
Intensity (a.u.)
0.8
in µm 0.10
0.6 0
InSb half plane 0.08
0.4
¬0.5 0.06
0.2 0.04
0 ¬1.0 0.02
¬0.2 0
¬0.4 ¬0.2 0 0.2 0.4 0.6 ¬0.5 0 0.5 1.0 0 1 2 3 4 5 6 7
x position on camera (mm) x position on camera (mm) Source size (µm)
Figure 2 | The X-ray source casts a shadow of a half-plane on the detector. a, Close-up of measured intensity distribution (black squares) integrated along
the edge of the half-shadow (inset) and exemplary intensity distributions using Fresnel diffraction modelling, for a source with Gaussian intensity
distribution and synchrotron spectrum Ecrit /wx of 8 keV/1 µm (solid red), 8 keV/3 µm (dashed green), 2 keV/1 µm (dash–dotted blue) and 8 keV/6 µm
(dotted grey). b, Series of measured intensity distributions corresponding to shots with decreasing source size wx (from modelling) and fringe visibility. The
numbers are the 95% confidence interval of the source size determined by a least-squares fit. c, The experimentally obtained fringe visibility is consistent
with the fringe visibility obtained from the modelling, plotted for a 6–10 keV synchrotron spectrum with spatial profile of (super-) Gaussian to top-hat type.
Horizontal error bars are given by the fitting procedure (typically ±30–40%) and vertical error bars are due to the noise in the experimental data
(typically ±0.03–0.04).
peak brightness than previous laser-driven betatron sources6,7 . average and maximum energy of the electron beam follow the
Furthermore, we demonstrate that the radiation has an appreciable typical wakefield electron density scaling law14,18 .
degree of spatial coherence. With the electron beam deflected away from laser axis by the
The experiment was carried out by focusing an intense short- spectrometer magnet, a bright (undeviated) beam of X-rays was
pulse ('30 fs, '2 J) laser onto the front edge of 3, 5 and 10 mm also observed co-propagating along the laser axis. It was imperative
helium gas jets (see Methods). A scintillating screen was placed in to first prove that this X-ray source originates from the plasma itself.
the electron beam to measure its beam profile, and a permanent To do this a grid of silver wires (60 µm diameter, 310 µm separation)
magnet spectrometer to measure its energy spectrum. Electron was placed a few centimetres from the target. X-rays originating
beams with narrow energy spread features were observed from from the interaction region project the outline of the mesh onto
all nozzles, at electron densities of 4–22 × 1018 cm−3 . As there an imaging plate. A strongly directional beam of X-rays is evident
are consecutive phases of injection, the electron beam consists in Fig. 1a. When either the laser power or electron plasma density
of multiple beamlets, which could be seen in both profile and was reduced to inhibit the electron beam, the X-ray beam also
spectral measurements14 . For example, for an electron density of disappeared, showing that the generation of the X-rays is linked to
8 × 1018 cm−3 on the 5 mm nozzle, electron beams of W = (230 ± the electron beam. The profile is elliptical, with a FWHM divergence
70) MeV with 1W /W = (25 ± 10)% energy spread at full-width at of θx × θy = 4 × 13 mrad2 , corresponding to a wiggler parameter
half-maximum (FWHM) were observed with an average of 2.2±0.4 K = θ γ of Kx = 1.5 and Ky = 5 for a simultaneously measured
beamlets per shot, with a root-mean-square (r.m.s.) divergence of electron beam energy W = 200 MeV. The X-ray beam pointing
1.5 × 1.8 mrad2 and r.m.s. pointing fluctuation of 4.8 × 4.7 mrad2 . is extremely stable, as can be deduced from Fig. 1b, which shows
We typically measure 100–300 pC of charge in the beam. The the sum of five consecutive shots. Their combined divergence is
a b c ¬20
Experimental mean ¬10 6
6
Experimental
1.0 Model standard
d2l/dwdS (a.u.)
0.8 4
Energy (a.u.)
4
0.6 0 3 0 3
0 10 20
100 101 102 20 40 60 80 ¬20 ¬10 0 10 20
X-ray energy (keV) X-ray energy (keV) Horizontal divergence (mrad)
Figure 3 | Measurement and modelling of the X-ray spectrum and modelling of the X-ray profile. a, The modelled, normalized betatron spectrum agrees
well with the experimentally measured spectrum obtained within one standard deviation. Inset: The experimental spectrum is inferred from measurements
of the X-ray yield through various filters with known transmission. b, The angularly and spectrally resolved X-ray flux displays a peak on axis at 10 keV and a
tail extending to '100 keV. c, The spectrally integrated X-ray beam profile shows an elliptically elongated beam profile in the direction of laser polarization.
not significantly larger than that of a single-shot measurement, the For the ringing to occur, the radiation needs to have an apprecia-
r.m.s. pointing stability is 5 mrad in the horizontal and vertical ble degree of spatial coherence at the place of the half-plane, which
direction, similar to the pointing stability of the electron beam. is merely u = 50 mm from the source. The van Cittert–Zernike
To give an indication of the X-ray source size, microscopic theorem states that even the radiation of uncorrelated emitters
objects were backlit with the X-ray beam. Figure 1c–h shows X-ray with Gaussian intensity distribution can be spatially coherent19 .
radiographic images of wire triplets of various sizes and of a The transverse coherence length is given by Ltrans = λu/2π wx,y .
resolution test target. Figure 1i,j gives photographic images of the In our case the relevant radiation wavelength is λ ' 6 × 10−10 m,
smallest wire triplet and the test target. Even the smallest features which is the peak of the product of the synchrotron spectrum
of size 3 µm are resolved, indicating that the betatron X-ray source Ecrit ' 8 keV and detector response. For a source size of wx = 2 µm,
is <
∼3 µm, smaller than the size of the plasma wave in which the Ltrans ' 3 µm, which is enough to observe one Fresnel fringe. The
radiating electrons were trapped and oscillate ('20 µm diameter). spectral width of the source also reduces fringe visibility, as the
To quantify the source size more precisely, a half-plane was temporal or longitudinal coherence length Llong = λ2 /21λ ' λ.
backlit with the X-ray beam. A typical intensity distribution on We have assessed the amount of coherence by calculating the ex-
the detector looks like a half-shadow (Fig. 2a inset), whose details perimental fringe contrast (Imax − Imin )/(Imax + Imin ) and comparing
convolve information about the X-ray source and half-plane. it with the prediction from the modelling of a (super-) Gaussian to
The half-plane was a 0.25-mm-thick cleaved InSb crystal (<6% top-hat source with Ecrit = 6–10 keV, as shown in Fig. 2c. From this
transmission below 20 keV) and resembles an ideal step function. agreement we infer that the complex coherence factor µ of our beta-
To accurately model the shape of the intensity distribution, tron source must be close to the theoretical upper limit of µ = 0.88,
it is necessary to use Fresnel diffraction (see Methods), where which would be achievable for a Gaussian intensity distribution19 .
the details of the diffraction pattern depend on the spatial and The spectral properties of the betatron radiation were deter-
spectral intensity distribution of the source and the dimensions of mined by measuring the X-ray transmission through a set of filters
the set-up. Figure 2a shows a close-up of a typical experimental (see the inset of Fig. 3a) with an X-ray CCD (charge-coupled
and several modelled intensity distributions based on a Gaussian device) detector7 . Assuming the spectrum is synchrotron-like, the
intensity profile and synchrotron spectrum. For the solid red curve measured Ecrit is (29 ± 13) keV for ne = (1.0 ± 0.4) × 1019 cm−3 , and
in Fig. 2a, Ecrit = 8 keV and a 1/e 2 intensity radius wx = 1 µm the 5 mm nozzle (see Fig. 3a). With twice the density and half the
was assumed, which best reproduced both the sharp rise and the laser power, the measured Ecrit = 6–10 keV is found to be lower,
amplitude and width of the first fringe. Changing wx or Ecrit under consistent with what can be inferred from the fringe measurement
or overestimates the height and/or width of the overshoot and/or (Ecrit = 4–16 keV) that was carried out at these conditions.
the rise, as shown by the other curves in Fig. 2. Numerical modelling was carried out using electron trajectories
To obtain a 95% confidence interval for the source size, a obtained from the fully relativistic particle-in-cell code OSIRIS.
least-squares fit based on the Fresnel model was carried out with the The simulations were run in the boosted frame, which, because
measured intensity distributions (see Methods). Figure 2a shows of relativistic length contraction and time dilation, offers higher
exemplarily that the agreement between model and data depends resolutions for shorter run times. The trajectories were post-
only weakly on the critical energy. This was confirmed by inde- processed to yield the characteristics of the betatron radiation
pendent measurements of the spectrum at identical experimental (see Methods). Figure 3b shows the modelled X-ray spectrum as
conditions, which found Ecrit = 6–10 keV, as discussed later. To a function of energy and angle. The X-ray flux peaks on-axis at
accurately acknowledge the small uncertainty in critical energy and 10 keV and extends to '100 keV. A lineout taken on-axis, as would
also possible deviations from a Gaussian source profile, the parame- be measured by our detector, is plotted in Fig. 3a, comparing well
ter space of the Fresnel model includes spectra with critical energies with the measured spectrum. A total of 108 photons are predicted
6–10 keV and spatial profiles of (super-) Gaussian to top-hat type. between 1 and 84 keV. We measure 106 –108 , depending on electron
Figure 2b shows a series of experimental intensity distributions charge and (albeit small) pointing fluctuation. As a result of the
for which the source size was determined in this way. By changing difficulty in decoupling pointing and yield fluctuations, we take
the plasma density, we have some level of control over the conservatively 5 × 107 photons, a 1/e 2 source radius wx,y = 1 µm,
source size. As the source size increases the visibility of the a divergence of 13 × 4 mrad2 , a bandwidth of 100% and a pulse
first Fresnel fringe decreases. This is in accordance with the duration of 30 fs (refs 20,21). This gives a maximum achieved peak
modelling, from which it is expected that the fringes disappear for brightness of 1 × 1022 photons per second per mrad2 per mm2 per
a source size wx >∼ 5 µm. 0.1% bandwidth.