67th ICFA Adv. Beam Dyn.

Workshop Future Light Sources FLS2023, Luzern, Switzerland JACoW Publishing

ISBN: 978-3-95450-224-0 ISSN: 2673-7035 doi:10.18429/JACoW-FLS2023-TU1C4

THE CXFEL PROJECT AT ARIZONA STATE UNIVERSITY

Content from this work may be used under the terms of the CC-BY-4.0 licence (© 2023). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI

W. S. Graves∗ on behalf of the CXFEL collaboration† , Arizona State University, Tempe, AZ, USA

Abstract LABORATORY ENVIRONMENT

The CXFEL Project encompasses the Compact X-ray Environmental factors have a large impact on the stability
Light Source (CXLS) that is now commissioning in the hard and performance of experiments at molecular length and
x-ray energy range 2-20 keV, and the Compact X-ray Free- time scales. In 2018 we constructed a new laboratory build-
Electron Laser (CXFEL) designed to lase in the soft x-ray ing on campus, Biodesign C, with 2 essentiallly identical
range 250 - 2500 eV. CXFEL has recently completed a 3-year groups of labs that are purpose-built for the accelerators, RF
design phase and just received NSF funding for construction systems, lasers, and x-ray endstations used by CXFEL and
over the next 5 years. These instruments are housed in sepa- CXLS. The laser, accelerator, and x-ray labs are in separate
rate purpose-built laboratories and rely on inverse Compton rooms each with their own safety and air-handling systems
scattering of bright electron beams on powerful lasers to pro- but the 3 rooms share a single 2m thick reinforced concrete
duce femtosecond pulses of x-rays from very compact linacs foundation that is physically separated from the adjoining
approximately 1 m in length. Both instruments use recently building spaces to provide a common stable platform for the
developed X-band distributed-coupling, room-temperature, beams. These rooms are verified to meet VC-E vibration
standing-wave linacs and photoinjectors operating at 1 kHz standards similar to an electron microscope lab. The air
repetition rates and 9300 MHz RF frequency. They rely on temperature stability in the laser labs is ±0.25∘ C, and is
recently developed Yb-based lasers operating at high peak ±0.5∘ C in the accelerator lab and x-ray hutch. Air humidity
and average power to produce fs pulses of 1030 nm light at is at 40% ± 5%.
1 kHz repetition rate with pulse energy up to 400 mJ. We
To mitigate electromagnetic interference we installed the
present the current commissioning performance of CXLS,
RF systems including the klystrons in a Faraday cage pre-
and review the design of the fully coherent CXFEL.
venting external pickup and suppressing any effects of the
high power klystron pulses on sensitive x-ray experiments.
INTRODUCTION The majority of equipment is water-cooled for temperature
Future light sources aim to improve performance, cost, stability of high power systems. We designed and built a
and accessibility over today’s instruments. The CXFEL precision processed cooling water system that is tunable and
project is developing femtosecond x-ray light sources at a stable to ±0.05∘ 𝐶 with a future goal to achieve ±0.01∘ 𝐶.
cost and size that makes the novel time-resolved molecular
science they provide accessible to many institutions includ-
ing universities, medical facilities, and industrial labs. The CXLS
first instrument produced and now commissioning is the The CXLS is constructed and now commissioning hav-
CXLS that produces partially coherent synchrotron-like x- ing produced first x-rays in February 2023. The technical
ray pulses at few hundred fs duration in the hard x-ray range. components of CXLS including RF and accelerator sections,
The second instrument, CXFEL, has completed design and magnets, lasers, and diagnostics all serve as prototypes for
is now under construction. It is a further development of the CXFEL equipment. CXLS will continue commissioning
CXLS technology that adds novel nanometer-scale electron through 2023, transitioning to early science phase in 2024.
bunching to produce fully coherent x-rays in the soft x-ray
spectrum. These instruments differ from the major XFEL
and synchrotron facilities as well as current laboratory scale CXLS X-ray Performance
sources, thus require development of new experimental tech-
The CXLS design performance is given in Table 1. So far
niques, sample delivery, detector properties, controls and
for commissioning we are running at lower electron bunch
data analysis methods that are matched to their novel prop-
charge (10 - 20 pC) and laser ICS laser power (80 W) result-
erties. These properties include lower flux than the major
ing in x-ray flux of 3 × 105 photons/shot at 1 kHz repetition
facilities as well as improvements in stability and precision
rate. The charge is limited by damage to UV photoinjector
of beam properties, and the ability to tailor integrated ac-
optics that are currently being replaced with reflective op-
celerator, laser, and x-ray beamline operations to optimize
tics. We are being conservative with the ICS laser power
particular experiments. We first describe the labs that house
while developing automated steering software. In coming
the instruments and then the 2 instruments. For illustrations
months we expect to increase UV and ICS laser powers to
of the various equipment and laboratories discussed here
their full design specs. Simulations indicate that current x-
please see oral presentation TU1C4 in the proceedings of
ray performance (flux, emission angle, energy) is consistent
the 2023 FLS workshop.
with measured values. We anticipate meeting the full x-ray
∗ [email protected] performance specifications as the power from both lasers is
† see Annex A increased.
TU1C4
54 C - Compact Light Sources
67th ICFA Adv. Beam Dyn. Workshop Future Light Sources FLS2023, Luzern, Switzerland JACoW Publishing
ISBN: 978-3-95450-224-0 ISSN: 2673-7035 doi:10.18429/JACoW-FLS2023-TU1C4

Table 1: CXLS Design Performance at 0.1% and 5% band- The low-level RF (LLRF) system is a hybrid analog-digital
width. Brilliance units are ph/(s 0.1% mm2 mr2 . system based on IQ modulation/demodulation developed at

Content from this work may be used under the terms of the CC-BY-4.0 licence (© 2023). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI
ASU. It drives 100 W solid-state power amplifiers that in
Parameter 0.1% BW 5% BW turn drive a pair of L3 L-6145 klystrons to 6 MW saturated
Photon energy (keV) 2 - 20 2 - 20 output power in 700 ns pulses. Output from klystron 1 is
Photons/pulse 5 × 106 1 × 108 split with a waveguide variable phase-shifter power-divider
Pulse rate (Hz) 1000 1000 (VPSPD) between the photoinjector (3 MW) and linac 1 (2
Avg flux (ph/s) 5 × 109 1 × 1011 MW) after losses. Klystron 2 output is fed to a 3 dB hybrid
Avg brilliance 2 × 1012 5 × 1012 that routes 2.5 MW power to each of linacs 2 and 3. We use
Peak brilliance 3 × 1019 9 × 1018 a total of 3 VPSPDs that can shunt klystron power to water
Round RMS src size (𝜇m) 3.0 3.0 loads so that the klystrons always run at max power and
Round RMS src angle (mrad) 4.0 4.0 constant thermal load regardless of what final beam energy
RMS pulse length (fs) < 300 < 300 is required. Scandinova K1 modulators power the klystrons,
RMS timing jitter (fs) < 50 < 50 achieving 100 ppm rms voltage stability. Generally the RF
systems are extremely stable over short and long terms with
phase jitter of < 0.04∘ (12 fs) and amplitude jitter < 5×10−4.
Interaction Point Typical operations during commissioning are to run for 8-10
hours in a day, shutdown, and then repeat the next day. The
The electron and laser beams meet at the interaction point facility reaching thermal stability about 45 minutes after
(IP), crossing at an 8∘ angle from head-on to produce x- startup when running at 1 kHz, or within 10 minutes when
rays. The electron bunch length is 300 fs and the Dira is running at 100 Hz. The day-to-day stability is excellent
1 ps. Timing jitter is 100 fs and is stable over timescales requiring only a few minutes of tuning to rephase all the
of hours. The Dira output is focused by a f=20 cm lens to structures and laser.
a 𝜔0 = 10 µm spot and the electron beam is focused by a
quadrupole triplet located 20 cm upstream of the IP to a 4 µm
Lasers and Timing Systems
rms spot size at the IP. Spatial overlap is stable over periods
of minutes at these spot sizes. We are developing steering CXLS includes a photocathode laser that produces the
feedback to correct slow thermal drift over longer timescales. electron beam and the high power ICS laser that collides
The x-rays emerge with a 1̃0 mrad opening angle. In current with the electron beam. The photocathode laser is a Light-
commissioning no collimating x-ray optic is present. An f = Conversion Pharos Yb:KGW amplifier producing 1.5 mJ
20 cm elliptical Montel x-ray optic at fixed photon energy pulses of 1030 nm light with FWHM 180 fs at 1 kHz repeti-
of 9.3 keV will be installed in coming weeks to collimate tion rate. The single-box laser has a 4th harmonic module
the beam. There are several YAG:Ce screens in the vicinity that produces up to 100 µJ of 258 nm light that produces the
of the IP to image the electron/x-ray/laser beams including electron beam at the photocathode. The Pharos oscillator is
one insertable screen directly at the IP. Fitting short focal synchronized to the RF master oscillator via a Menlo Sys-
length optics for all of these beams within a few cm of teh tems RRE-SYNCHRO unit resulting in 120 fs rms timing
IP as well as the diagnostics was an engineering challenge. jitter between RF and UV. This unit runs stably with little
tweaking required. However the fs UV pulses at 1 kHz have
proved to cause nonlinear effects and damage in optical coat-
Accelerator and RF Systems
ings on the MgF lenses used for transport. These effects
The accelerator consists of a 4.5 cell photoinjector and limit the charge available in the accelerator to 20 pC. We are
3 short linac sections of 20 cells each, all powered by 2 RF in the process of replacing the transmissive optics with re-
transmitters. All of the accelerator structures are high effi- flective optics and are testing damage thresholds to increase
ciency standing-wave room-temperature copper structures the fluence to the cathode.
with a repetition rate of 1 kHz and fill time of 170 ns. Peak The ICS laser is a Trumpf Dira 200-1 Yb:YAG thin-disk
cathode field in the photoinjector is 120 MV/m for a 3 MW amplifier producing 200 mJ pulses of 1030 nm light with
input power. The cathode cell is less than a half-cell so that FWHM of 1.1 ps at 1 kHz repetition rate. The Pharos oscil-
the laser arrival timing is close to the peak applied electric lator sends an optical seed signal to the Dira that is amplified
field. Exit energy is 4.0 MeV. producing good synchronization with the cathode laser and
The linacs are innovative distributed-coupling struc- electron beam. Trumpf has also supplied a cross-correlator
tures [1] with a high shunt impedance 165 MOhm/m making that measures and corrects the timing difference between the
them very efficient, gradient up to 30 MV/m and peak sur- Pharos and Dira amplifier outputs resulting in a net timing
face E-field of 120 MV/m. The structures are 20 cells long jitter of 33 fs between amplified pulses at 1 kHz rate. Point-
(32 cm), produce energy gain of 10 MeV each for 2 MW ing stability of the Dira is 4 µrad rms and power stability is
input power and a final maximum linac energy of 34 MeV. 0.2% rms over 24 hours. The Dira has been tested to its full
This energy is adequate to produce up to 20 keV x-rays via specifications but we are typically using it at 80 mJ/shot in
ICS. early x-ray commissioning.
TU1C4
C - Compact Light Sources 55
 67th ICFA Adv. Beam Dyn. Workshop Future Light Sources FLS2023, Luzern, Switzerland JACoW Publishing
ISBN: 978-3-95450-224-0 ISSN: 2673-7035 doi:10.18429/JACoW-FLS2023-TU1C4

CXFEL Table 2: CXFEL Design Performance at several photon

energies. Brilliance units are ph/(s 0.1% mm2 mr2 .
CXFEL is a further development of the technologies used
Content from this work may be used under the terms of the CC-BY-4.0 licence (© 2023). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI

in CXLS that is designed to produce fully coherent x-ray Photon energy (eV)
pulses. See the illustrations in the TU1C4 oral presenta- Parameter 250 1000
tion in these proceedings. The accelerator components are
Photons/pulse 8 × 108 1.1 × 108
nearly identical to CXLS with the addition of an electron
Pulse rate (Hz) 1000 1000
diffaction [2] chamber and emittance exchange [3] line. The
Avg flux (ph/s) 8 × 1011 1.1 × 1011
components also apply lessons learned with CXLS to im-
Flux/shot (nJ) 32 18
prove performance. The purpose of the additional equipment
Avg brilliance 1.3 × 1015 1.2 × 1016
is to create bunches of electrons that are short on the x-ray
Peak brilliance 1.2 × 1028 5.6 × 1028
wavelength scale, i.e., < 1 nm in length so that when they in-
Round RMS src size (µm) 0.9 0.5
teract with the ICS laser the output is coherent. ICS sources
Round RMS src angle (µrad) 440 188
are generally excellent at producing hard x-ray photons with
FWHM pulse length (fs) 9.1 4.6
performance decreasing for softer photon energies. However
FWHM bandwidth (%) 0.18 0.09
the technique used by CXFEL to produce nanobunches that
Arrival timing jitter (fs) < 10 < 10
radiate coherently is currently limited by equipment jitter
Electron beam energy (MeV) 14 29
performance to lower energies (longer x-ray wavelengths)
in the soft x-ray range. We believe that with further devel-
opment this technique will reach hard x-rays, but the scope
of CXFEL is currently limited to photon energies less than pulses up to 40 MW. This power is then split among the RF
2.5 keV with most of the development work focusing on the structures. Combining and splitting the power produces a
important energy range from 250 eV to 1.2 keV. system with redundancy that enables operation with a single
Producing x-rays in this lower energy range is not well RF transmitter, as well as providing the opportunity to reduce
suited to the head-on collision geometry of CXLS where ebeam energy and timing jitter by using the correlations
the effective undulator period is just 515 nm. Such a short in phase and amplitude jitters among the linac sections to
period requires an ebeam energy into the few MeV range to partially cancel their effects. The transmitters are upgraded
make soft x-rays. Such low energy electrons are subject to Scandinova K200 units customized with low jitter triggers
strong space-charge forces and difficult to focus to a small and additional charging supplies with the goal to reduce
interaction spot. Instead we adopt an ”overtaking” collision high voltage jitter from 100 ppm to 10 ppm at 1 kHz. The
geometry where the electron beam and laser propagate in compressed high power RF is used to drive the photoinjector
the same direction with a 30∘ angle between them resulting cathode gradient up to 150 MV/m.
in an effective undulator period of 8𝜇m thus raising the Lasers and Timing Systems
ebeam energy to e.g. 29 MeV to produce 1 keV photons.
This geometry does however require a substantially more The photoinjector laser is identical to CXLS except that
powerful and shorter pulse ICS laser as discussed below. we will build our own harmonic module to improve the UV
mode. The ICS laser is a significant upgrade over CXLS with
X-ray Performance the final multipass amplifier producing 500 mJ of 1030 nm
light in 800 fs pulses at 1 kHz. This output is then spectrally
The CXFEL design performance is given in Table 2 across broadened in a large Herriott cell and compressed to 40 fs.
a range of energies. The nanobunching concept that pro- The timing synchronization for CXFEL takes a different
duces coherent emission is very sensitive to space charge approach than CXLS, using BOM-PD components from
forces resulting in the use of very low charge (1 pC) electron Cycle Laser with the goal to achieve approximately 20 fs
bunches vs the 200 pC bunches of CXLS. Thus the flux of locking between RF and photoinjector and ICS lasers. We
CXFEL with its more efficient coherent emission is similar anticipate ebeam and x-ray arrival time jitter below 10 fs
to CXLS. However the coherent photons are emitted into a with proper setup of the accelerator and laser timing.
phase space volume that is orders of magnitude smaller than
the incoherent CXLS and thus the brilliance of the CXFEL s ACKNOWLEDGMENTS
orders of magnitude higher. The x-ray pulses are also much
shorter, ranging from 0.5 - 10 fs. This work is supported by NSF Awards 2153503 and
1935994, and by ASU.
Accelerator and RF Systems
REFERENCES
The CXFEL accelerator reuses the CXLS photoinjector
[1] S. Tantawi, M. Nasr, Z. Li, C. Limborg, and P. Borchard,
and has 3 similar linac sections. However the linacs have
“Design and demonstration of a distributed-coupling linear
subtle design changes that improve the symmetry of the
accelerator structure,” Phys. Rev. Accel. Beams, vol. 23, no. 9,
cells and allow operation at higher gradient up to 75 MV/m. p. 092 001, 2020.
The RF system uses the same 2 klystrons as CXLS, but doi:10.1103/PhysRevAccelBeams.23.092001
combines and compresses the klystron output to make short
TU1C4
56 C - Compact Light Sources
67th ICFA Adv. Beam Dyn. Workshop Future Light Sources FLS2023, Luzern, Switzerland JACoW Publishing
ISBN: 978-3-95450-224-0 ISSN: 2673-7035 doi:10.18429/JACoW-FLS2023-TU1C4

[2] L. Malin et al., “Quantitative agreement between dynamical [3] E. A. Nanni, W. S. Graves, and D. E. Moncton, “Nanomodu-
rocking curves in ultrafast electron diffraction for x-ray lasers,” lated electron beams via electron diffraction and emittance

Content from this work may be used under the terms of the CC-BY-4.0 licence (© 2023). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI
Ultramicroscopy, vol. 223, p. 113 211, 2021. exchange for coherent x-ray generation,” Phys. Rev. Accel.
doi:10.1016/j.ultramic.2021.113211 Beams, vol. 21, no. 1, p. 014 401, 2018.
doi:10.1103/PhysRevAccelBeams.21.014401

ANNEX A: CXFEL COLLABORATION

A. Botana, S. Botha, B. Cook, O. Erten, P. Fromme, A. Gardeck, W. Graves, M. Holl, J. Houkal, S. Jachim, R. Kaindl,
S. Karkare, R. Kirian, E. Lttman, B. Liebich, S. Rednour, A. Ros, K. Schmidt, D. Smith, S. Teitelbaum, T. Thornton,
S. Tilton, S. Tongay, J. Vela, K. Warble, U. Weierstall, D. Winkel, Arizona State University, Tempe, AZ
H. Loos, Yorktown Heights, NY
M. Mitrano, Harvard University, Cambridge, MA
D. Rolles, A. Rudenko, Kansas State University
Y.-D. Chuang, O. Gessner, J. Qiang, S. Roy, T. Weber, Lawrence Berkeley National Laboratory, Berkeley, CA
R. Comin, K. Nelson, Massachusetts Institute of Technology, Cambridge, MA
L. DiMauro, Ohio State University
N. Shivaram, Purdue University
G. Phillips, Rice University
J. Cryan, V. Dolgashev, Z. Li, E. Nanni, S. Tantawi, M. Trigo, SLAC National Accelerator Laboratory, Stanford, CA
R. Fung, A. Hosseinizadeh, A. Ourmazd, M. Schmidt, P. Schwander, University of Wisconsin - Milwaukee,
Milwaukee, WI T. Grant, University of Buffalo, Buffalo, NY
M. Brown, A. Sandhu, University of Arizona, Tucson, AZ
M. Frank, T. Kuhl, University of California Davis, Davis, CA
F. Mahmood, University of Illinois Urbana-Champagne, Urbana-Champagne, IL
M. Centurion, University of Nebraska, Nebraska
Undergraduate students: G. Babic, C. Bell, E. Boyd, T. Brown, T. Dela Rosa, R. DeMott, A. Dupre, K. Eckrosh,
E. Everitt, A. Eyler, J. Falconer, R. Jaswal, R. Larson, X. Ma, A. Martinez, S. Ramkumar, E. Ros, A. Semaan, A.
Staletovic, J. Stanton, S. Tripathi, D. Valentin, Arizona State University, Tempe, AZ

TU1C4
C - Compact Light Sources 57