Research towards high-repetition rate laser-driven X-ray sources for imagingapplications

J. Gotzfried,1 A. Dopp,1 M. Gilljohann,1 H. Ding,1 S. Schindler,1 J. Wenz,1 L. Hehn,2 F. Pfeiffer,2, 3 and S. Karsch1, ∗

1Ludwig-Maximilians-Universitat Munchen, Am Coulombwall 1, 85748 Garching, Germany2Lehrstuhl fur Biomedizinische Physik, Physik-Department & Munich School of BioEngineering,

Technische Universitat Munchen, 85748 Garching, Germany3Institut fur Diagnostische und Interventionelle Radiographie,

Klinikum rechts der Isar, Technische Universitat Munchen, 81675 Munchen, Germany

Laser wakefield acceleration of electrons represents a basis for several types of novel X-ray sourcesbased on Thomson scattering or betatron radiation. The latter provides a high photon flux anda small source size, both being prerequisites for high-quality X-ray imaging. Furthermore, proof-of-principle experiments have demonstrated its application for tomographic imaging. So far thisrequired several hours of acquisition time for a complete tomographic dataset. Based on improve-ments to the laser system, detectors and reconstruction algorithms, we were able to reduce this timefor a full tomographic scan to 3 minutes. In this paper, we discuss these results and give a prospectto future imaging systems.

Laser-driven X-ray sources take the middle ground inbrilliance and cost between low-cost microfocus X-raytubes and large-scale synchrotron sources. This appliesespecially to sources based on laser-wakefield accelera-tion [1], which allow for the production of collimated,femtosecond X-ray beams [2]. In particular, Thomsonbackscattering and betatron radiation have proven tobe the most relevant for applications in the hard X-ray regime [3]. While the emission of radiation in bothmechanisms is based on oscillatory motions of relativisticelectrons, their oscillation frequencies differ significantly.Thomson backscattering sources rely on electrons oscil-lating in the electromagnetic field of an intense collidinglaser pulse, whereas betatron radiation is generated byelectrons wiggling transversely in the wake of a highlyintense laser pulse traveling through a plasma while be-ing accelerated [2]. Transverse electric fields of severalGV/m to TV/m force the accelerated electrons with ini-tial transverse momentum onto oscillating trajectories.This wiggler-like movement leads to a broadband X-rayemission, while the duration of the X-ray pulses is on theorder of femtoseconds [4]. Laser-driven sources are there-fore particularly suitable to study ultrafast processes liketransitions in the X-ray absorption near edge structures[5]. However, one of the most important medical appli-cations of X-rays remains radiography. Single-shot X-rayimaging has been shown with both Thomson and beta-tron sources [6, 7] and tomographic imaging has beendemonstrated using betatron sources [8, 9]. Moreover,the intrinsic small source size of a few microns allowsfor phase contrast imaging [10, 11]. While previous re-search focused on the potential of this imaging method,the acquisition time for a tomography in these studies hasbeen on the order of several hours [8, 9]. Here we focuson the duration of such scans and report on a successfulreduction of this time to a more application-relevant fewminute scale. This was achieved by upgrading the experi-mental setup to support 1 Hz repetition rates and making

use of advanced reconstruction algorithms. The latter al-lows to acquire a single image per projection angle andperform a consistent reconstruction despite shot-to-shotX-ray flux fluctuations of the source. As a result, thedata acquisition time for a centimeter-scale human bonesample was reduced from several hours to 180 seconds[12].

The measurements were performed at the Labora-tory for Extreme Photonics in Garching, Germany. The800 nm, 27 fs laser pulse was delivered by the ATLAS(Ti:sapphire) laser system. An off-axis parabolic mirror(f/25) focuses the laser pulse containing an energy at thetarget position of 1.9 ± 0.1 J to a spot size of 30 µm(FWHM intensity), which corresponds to a peak inten-sity of 5.5× 1018W/cm2 and a peak power of 70± 4 TWresulting in a0 ≈ 1.6.

As target a gas cell filled with hydrogen at a densityof ∼ 5× 1018 cm−3 was used in which a movable pistonallowed to adjust its length anywhere between 5 and 15mm. The cell length was optimized with respect to theX-ray yield exploiting the fact that deep in the electrondephasing regime the electrons perform more betatronoscillations. The optimum X-ray yield was found at acell length of 11 mm.

The accelerated electrons are deflected onto a scintil-lating screen by a 0.8 m long 0.85 T dipole magnet of anelectron spectrometer. We achieved an electron beamcharge of 736 ± 51 pC with a pointing fluctuation of1.1 ± 0.1 mrad. The X-rays are detected by a scintil-lator based camera located 4.35 m behind the gas cellexit. An array of aluminum filters of different thick-nesses ranging from 5 µm to 610 µm can be insertedinto the X-ray beam. The different transmission coeffi-cients for each filter allow the determination of the X-rayspectrum by iteratively optimizing its calculated filtertransmissions to the ones measured [13]. The light shieldand the Kapton vacuum window (cf. Fig. 1) are opaquefor low energetic X-rays (> 50 % transmission for ener-

arX

iv:1

803.

0541

5v1

[ph

ysic

s.m

ed-p

h] 1

4 M

ar 2

018

2

sBonesSample

SampleHolder

AlsShielding FiltersCake

ElectronBeam

X0RayBeam

AlsShielding

KaptonsWindow

GassTarget

LaserDO3J=sMbfsØs=D5cm

OA

Pfs=

sMOu

m

DipolesMagnets45cm=s5O4uT

ScintillatingsScreen2EnergysSpectrum7

Sci

ntill

ator

sbas

edX

0Ra

ysC

amer

a

(8ucm

)5cms0sD55cm

FIG. 1. Experimental Setup for a tomography of a human bone sample at 1 Hz repetition rate. The laser pulse is focused into agas cell where it ionizes hydrogen gas and drives a plasma wave. The resulting charge separation generates large electromagneticfields. These accelerate electrons longitudinally while wiggling them transversely, which leads to betatron X-ray emission. Thebone sample is protected from the laser light by a 15 µm-thick aluminum foil which is passed by the electrons and X-rays. Theelectrons are deflected and analyzed in a magnetic spectrometer whereas the X-rays hit a scintillator which is imaged by a CCD.The camera itself is shielded by another 15 µm thick aluminum foil. The geometric distances result in an image magnificationof the bone sample of ∼ 4.4 − 7.3 on the camera.

gies above 7 keV) and therefore define a cut-off energybelow which the spectrum cannot be reliably retrieved.Based on the spectral measurements, a time averagedflux up to (1.6± 0.35)× 109 photons/msr/s (behind thebone sample and light shield) at 1 Hz repetition rate iscalculated. Comparing our retrieved spectrum to a syn-chrotron equivalent above the cut-off threshold of 7 keVgives a critical energy of Ecrit = 13.5 ± 0.95 keV (cf.Fig. 2). The size of the X-ray beam at the position of theCCD is larger than the sensor. Two dimensional Gaus-sian fits to the CCD images of the least divergent shotsprovide a lower estimate for the X-ray beam divergenceof 12× 6 mrad2 (s.d.).

To achieve a decent resolution, the sample was placed60 cm to 100 cm behind the target which therefore was lo-cated in front of the electron beam spectrometer (in con-trast to [9]). In order to minimize bremsstrahlung back-ground from electrons colliding with the sample mount,the bone is supported by a thin 3D printed acrylic plasticholder [14]. The signal of bremsstrahlung photons fromthe bone sample, holder and aluminum foil is negligiblecompared to the betatron signal. Furthermore, the sam-ple is protected from laser radiation by an aluminum foil,which is wrapped around the rotation stage (and there-fore rotates with the sample).

In preceding experiments [8], a direct detection X-rayCCD camera was employed for imaging purposes. Thistype of detector offers a high resolution (e.g. 13.5 µm inthe case of a Princeton Instruments PIXIS-XO:2048 ),but its sensitivity drops rapidly for radiation beyond10 keV. Furthermore, due to the high pixel number ofthe CCD chip and low-noise readout architecture, thistype of cameras typically has a readout time of severalseconds. This limits the shot frequency to around 0.1 Hz,which significantly increases the time needed for a tomog-raphy.

EX−ray[keV] 5 10 20 30 50 80χ 1

2[mm] 0.02 0.13 1.8 2.7 8.5 19.7

TABLE I. Half-value thickness χ 12

for human bones at differ-

ent energies [15].

In order to minimize the sample illumination - andtherefore the exposure to ionizing radiation - the energyof the X-rays should be chosen such that the correspond-ing half-value thickness is on the order of the sample’sphysical thickness (cf. Table I). This poses a major dif-ficulty for medical imaging with direct-detection CCDssince they are insensitive to radiation transmitted bye.g. human bones. However, the detection efficiency forhigh energetic photons can be drastically increased byconverting the incoming X-rays into photons of visiblewavelengths via scintillators.

For the quick tomography experiments we thereforehave used a scintillator-based CCD. It features an imageintensifier with variable gain. This camera uses a SonyICX285 Progressive Scan 2/3 rectangular CCD chip andsupports frame rates up to 30 fps. In practice, the max-imum shot frequency was limited to 1 Hz due to ourlaboratory data acquisition system. Due to the built infiber-optics taper (50:11), the camera’s effective pixel sizeis 29 µm. The P43 phosphor scintillator offers sensitivityalso for high X-ray energies (up to 100 keV) and thereforemakes this indirect detection method suitable for medicalimaging of thicker samples (cf. Fig. 2).

Figure 3 shows a comparison for the two types of detec-tors at the same angle of the sample and similar X-rayspectrum. As shown in the insets, the direct-detectionCCD has a better spatial resolution. In contrast, thephoton energy range detected by the camera is not welltransmitted by the bone and only limited informationabout its inner structure can be extracted.

3

FIG. 2. Comparison of the two different detection meth-ods (blue) and reconstructed X-ray spectrum at the detector(red). The shaded red area indicates the corresponding rmserror. At X-ray energies relevant for imaging bone samples,i.e. above 20 keV, indirect detection cameras have to be useddue to their much higher sensitivity for high energetic X-rays.

Another approach to reduce the acquisition time formedical applications is to minimize the necessary dataunderlying the tomographic reconstruction. This re-quires advanced reconstruction algorithms capable ofhandling reconstruction artifacts which are prone to ap-pear for small data sets [12]. The research of such al-gorithms has proven a prosperous field over the last twodecades and lead to the development of statistical itera-tive algorithms (SIRs) which are now becoming state ofthe art for computed tomographic technology [16] (seeFig. 4). These algorithms iteratively improve the recon-structed sample and apply weighting factors to emphasizetomograms with better signal to noise ratio. As a proofof principle we performed a quick tomography encom-passing 180 consecutive individual tomograms spanning180◦ at 1◦ step size. The entire data set was taken withinthree minutes at 1 Hz repetition rate [12].

The acquisition rate in these last experiments was lim-ited by the data acquisition system, while both the vac-uum and laser system would have supported shot fre-quencies up to 5 Hz. Beyond this, nowadays 100-TW-class laser systems with 10 Hz repetition rate are com-mercially available. If this could be fully exploited, atomography as presented in this paper would take 18 sec-onds to acquire and a high-resolution tomography with720 projection angles would only need slightly more thana minute - even with current laser technology.

Nevertheless, demands on the photon energy continueto grow as full body CTs or non-destructive testingof thick samples require much higher photon energies,larger X-ray beam diameters and a higher time-averagedflux. Without any new mechanisms in the generationof betatron X-rays, the laser repetition rate must be in-creased in order to gain higher mean brilliances. How-

(b)

(d)

(c)

(a)X-ray CCD

Scintillator

FIG. 3. Comparison of imaging using an X-ray CCD cam-era and a scintillator-based detector. While the X-ray CCDproduces sharper images, showing for instance signs of edgeenhancement, the lower part of the bone sample remains al-most opaque. In contrast, the scintillator camera is more sen-sitive in the > 10 keV regime, clearly showing the trabecularstructure.

ever, this poses a challenging task which current com-mercially available laser systems cannot satisfy: con-ventional Ti:sapphire laser systems have typical aver-age powers of ∼ 50 Watt which is limited by currentcrystal cooling concepts, such that an increased repe-tition rate usually comes at the cost of lower single-pulse energy. With new cooling concepts and/or laserarchitectures, this limit may be overcome in the future.Even though laser-wakefield acceleration at kHz repeti-tion rates has recently been demonstrated using mJ-classlaser systems [17–19], the reduced pulse energy results inelectron beams of lower energy (a few MeV) and negligi-ble betatron emission. For the generation of high ener-getic X-rays with kHz repetition rates, sources based onThomson-backscattering might be a viable alternative,since in this case the X-ray energy follows a favorableE ∝ 4γ2 scaling (e.g. 50 MeV electrons scattered with

4

FIG. 4. Quick tomography of a bone sample and compar-ison of two different reconstruction algorithms [12]. In theleft case filtered back projection (FBP) is used whereas thereconstruction in the right case was done via statistical itera-tive reconstruction (SIR). In contrast to FPB, SIR is capableof handling under-sampling artifacts which are due to the lim-ited number of acquired tomograms.

800 nm light produce 60 keV radiation) [20].

To conclude, we have demonstrated 1 Hz operation ofa laser-driven betatron source for imaging applications.In the near-term, these betatron sources might be fur-ther improved by using controlled-injection schemes [21]and operation at 10 Hz should be possible. At higherrepetition rates it will be easier to reach the requiredX-ray energies for medical tomographies using Thomsonbackscattering sources.

Based on the premise of laser-driven sources, the Mu-nich universities LMU and TUM have established thenew Centre for Advanced Laser Applications (CALA),which hosts several landmark laser installations aimingat both delivering higher laser pulse energies as well asproviding a joule-scale kHz laser system. The formeris met by the ATLAS-3000 laser, one of the few 1 Hzmulti-petawatt laser systems in the world. The PFS-prolaser system in contrast will provide hundreds of mJ oflaser pulse energy at kHz repetition rates. Both lasersystems will drastically increase the average X-ray fluxand therefore constitute another step towards real lifeimaging applications of laser-driven X-ray sources.

ACKNOWLEDGEMENTS

The authors thank F. Schaff and T. Baum (TUM) forproviding bone samples. This work was supported byDFG through the Cluster of Excellence Munich-Centrefor Advanced Photonics (MAP EXC 158), the DFG Got-tfried Wilhelm Leibniz program, TR-18 funding schemes,by EURATOM-IPP and the Max-Planck-Society.

∗ E-mail: [email protected]

[1] E. Esarey, C. B. Schroeder, W. P. Leemans, Physics oflaser-driven plasma-based electron accelerators, Reviewsof Modern Physics 81 (3) (2009) 1229–1285.

[2] S. Corde, K. Ta Phuoc, G. Lambert, R. Fitour,V. Malka, A. Rousse, A. Beck, E. Lefebvre, Femtosec-ond x rays from laser-plasma accelerators, Reviews ofModern Physics 85 (1) (2013) 1–48.

[3] F. Albert, A. G. R. Thomas, Applications of laser wake-field accelerator-based light sources, Plasma Physics andControlled Fusion 58 (10) (2016) 103001.

[4] K. T. Phuoc, R. Fitour, A. Tafzi, T. Garl, N. Artemiev,R. Shah, F. Albert, D. Boschetto, A. Rousse, D. E. Kim,A. Pukhov, V. Seredov, I. Kostyukov, Demonstration ofthe ultrafast nature of laser produced betatron radiation,Physics of Plasmas 14 (8) (2007) 080701.

[5] M. Saes, Observing Photochemical Transients by Ultra-fast X-Ray Absorption Spectroscopy, Physical ReviewLetters 90 (2003) 047403.

[6] S. Fourmaux, S. Corde, K. T. Phuoc, P. Lassonde,G. Lebrun, S. Payeur, F. Martin, S. Sebban, V. Malka,A. Rousse, J. C. Kieffer, Single shot phase contrast imag-ing using laser-produced Betatron x-ray beams, OpticsLetters 36 (13) (2011) 2426–2428.

[7] A. Dopp, E. Guillaume, C. Thaury, J. Gautier, I. An-driyash, A. Lifschitz, V. Malka, A. Rousse, K. T. Phuoc,An all-optical Compton source for single-exposure x-ray imaging, Plasma Physics and Controlled Fusion58 (034005) (2016) 1–5.

[8] J. Wenz, S. Schleede, K. Khrennikov, M. Bech,P. Thibault, M. Heigoldt, F. Pfeiffer, S. Karsch, Quan-titative X-ray phase-contrast microtomography from acompact laser-driven betatron source, Nature Communi-cations 6 (2015) 1–6.

[9] J. M. Cole, J. C. Wood, N. C. Lopes, K. Poder, R. L.Abel, S. Alatabi, J. S. J. Bryant, A. Jin, S. Kneip,K. Mecseki, D. R. Symes, S. P. D. Mangles, Z. Naj-mudin, Laser-wakefield accelerators as hard x-ray sourcesfor 3D medical imaging of human bone, Scientific Reports5:13244 (2015) 1–7.

[10] S. Kneip, C. McGuffey, F. Dollar, M. S. Bloom,V. Chvykov, G. Kalintchenko, K. Krushelnick, A. Mak-simchuk, S. P. D. Mangles, T. Matsuoka, Z. Najmudin,C. A. J. Palmer, J. Schreiber, W. Schumaker, A. G. R.Thomas, V. Yanovsky, X-ray phase contrast imaging ofbiological specimens with femtosecond pulses of betatronradiation from a compact laser plasma wakefield accel-erator, Applied Physics Letters 99 (9) (2011) 093701–093701–3.

[11] Z. Najmudin, S. Kneip, M. S. Bloom, S. P. D. Mangles,O. Chekhlov, A. E. Dangor, A. Dopp, K. Ertel, S. J.Hawkes, J. Holloway, C. J. Hooker, J. Jiang, N. C. Lopes,H. Nakamura, P. A. Norreys, P. P. Rajeev, C. Russo,M. J. V. Streeter, D. R. Symes, M. Wing, Compact laseraccelerators for X-ray phase-contrast imaging, Philosoph-ical Transactions of the Royal Society A: Mathematical,Physical and Engineering Sciences 372 (2010).

[12] A. Dopp, L. Hehn, J. Gotzfried, J. Wenz, M. Gilljohann,H. Ding, S. Schindler, F. Pfeiffer, S. Karsch, Quick x-raymicrotomography using a laser-driven betatron source,Optica 5 (2) (2018) 199–203.

[13] E. Y. Sidky, L. Yu, X. Pan, Y. Zou, M. Vannier, A ro-bust method of x-ray source spectrum estimation fromtransmission measurements: Demonstrated on computersimulated, scatter-free transmission data, Journal of Ap-

5

plied Physics 97 (12) (2005) 124701.[14] A. Dopp, E. Guillaume, C. Thaury, J. Gautier,

K. Ta Phuoc, V. Malka, 3D printing of gas jet nozzlesfor laser-plasma accelerators, Review of Scientific Instru-ments 87 (7) (2016) 073505.

[15] A. Akar, H. Baltas, U. Cevik, F. Korkmaz,N. Okumusoglu, Measurement of attenuation coef-ficients for bone, muscle, fat and water at 140, 364and 662 kev γ-ray energies, Journal of QuantitativeSpectroscopy and Radiative Transfer 102 (2) (2006)203–211.

[16] L. L. Geyer, State of the art: iterative CT reconstructiontechniques, Radiology 276 (2) (2015) 339–357.

[17] D. Guenot, D. Gustas, A. Vernier, B. Beaurepaire,F. Bohle, M. Bocoum, M. Lozano, A. Jullien, R. Lopez-Martens, A. Lifschitz, J. Faure, Relativistic electronbeams driven by kHz single-cycle light pulses, NaturePhotonics 11 (5) (2017) 293 –296.

[18] F. Salehi, A. J. Goers, G. A. Hine, L. Feder, D. Kuk,B. Miao, D. Woodbury, K. Y. Kim, H. M. Milchberg,MeV electron acceleration at 1 kHz with <10 mJ laserpulses, Optics Letters 42 (2) (2017) 215.

[19] D. Gustas, D. Guenot, A. Vernier, S. Dutt, F. Bohle,R. Lopez-Martens, A. Lifschitz, J. Faure, High-chargerelativistic electron bunches from a khz laser-plasma ac-celerator, Physical Review Accelerators and Beams 21 (1)(2018) 013401.

[20] K. Khrennikov, J. Wenz, A. Buck, J. Xu, M. Heigoldt,L. Veisz, S. Karsch, Tunable All-Optical Quasimonochro-matic Thomson X-Ray Source in the Nonlinear Regime,Physical Review Letters 114 (19) (2015) 195003.

[21] A. Dopp, B. Mahieu, A. Lifschitz, C. Thaury, A. Doche,E. Guillaume, G. Grittani, O. Lundh, M. Hansson,J. Gautier, M. Kozlova, J. P. Goddet, P. Rousseau,A. Tafzi, V. Malka, A. Rousse, S. Corde, K. T. Phuoc,Stable femtosecond X-rays with tunable polarizationfrom a laser-driven accelerator, Light: Science & Appli-cations 6 (11) (2017) e17086–7.