Chopper system for time resolved experiments with synchrotron radiation.

Abstract

A chopper system for time resolved pump-probe experiments with x-ray beams from a synchrotron is described. The system has three parts: a water-cooled heatload chopper, a high-speed chopper, and a millisecond shutter. The chopper system, which is installed in beamline ID09B at the European Synchrotron Radiation Facility, provides short x-ray pulses for pump-probe experiments with ultrafast lasers. The chopper system can produce x-ray pulses as short as 200 ns in a continuous beam and repeat at frequencies from 0 to 3 kHz. For bunch filling patterns of the synchrotron with pulse separations greater than 100 ns, the high-speed chopper can isolate single 100 ps x-ray pulses that are used for the highest time resolution. A new rotor in the high-speed chopper is presented with a single pulse (100 ps) and long pulse (10 micros) option. In white beam experiments, the heatload of the (noncooled) high-speed chopper is lowered by a heatload chopper, which absorbs 95% of the incoming power without affecting the pulses selected by the high speed chopper.

Full text

Page 1

Chopper system for time resolved experiments with synchrotron radiation
Marco Cammarata,1Laurent Eybert,1Friederike Ewald,1Wolfgang Reichenbach,1
Michael Wulff,1,a?Philip Anfinrud,2Friedrich Schotte,2Anton Plech,3Qingyu Kong,4
Maciej Lorenc,5Bernd Lindenau,6Jürgen Räbiger,6and Stephan Polachowski6
1European Synchrotron Radiation Facility, BP 220, Grenoble Cedex 38043, France
2Laboratory of Chemical Physics, National Institutes of Health, Bethesda, Maryland 20892-0520, USA
3Institute for Synchrotron Radiation (ISS), FZ Karlsruhe, Postfach 3640, D-76021 Karlsruhe, Germany
4Synchrotron SOLEIL, Saint-Aubin, 91192 Gif-sur-Yvette, France
5Groupe Matière Condensée et Matériaux, Université de Rennes 1, UMR6626 CNRS,
35042 Rennes Cedex, France
6Central Technology Division (ZAT), Forschungszentrum Jülich GmbH, Leo-Brandt-Str., 52425 Jülich,
Germany
?Received 3 July 2008; accepted 4 November 2008; published online 6 January 2009?
A chopper system for time resolved pump-probe experiments with x-ray beams from a synchrotron
is described. The system has three parts: a water-cooled heatload chopper, a high-speed chopper, and
a millisecond shutter. The chopper system, which is installed in beamline ID09B at the European
Synchrotron Radiation Facility, provides short x-ray pulses for pump-probe experiments with
ultrafast lasers. The chopper system can produce x-ray pulses as short as 200 ns in a continuous
beam and repeat at frequencies from 0 to 3 kHz. For bunch filling patterns of the synchrotron with
pulse separations greater than 100 ns, the high-speed chopper can isolate single 100 ps x-ray pulses
that are used for the highest time resolution.Anew rotor in the high-speed chopper is presented with
a single pulse ?100 ps? and long pulse ?10 ?s? option. In white beam experiments, the heatload of
the ?noncooled? high-speed chopper is lowered by a heatload chopper, which absorbs 95% of the
incoming power without affecting the pulses selected by the high speed chopper. © 2009 American
Institute of Physics. ?DOI: 10.1063/1.3036983?
I. INTRODUCTION
It has always been a dream to visualize the structure of
molecules in chemical and biochemical reactions with x rays,
but that has so far been difficult to do due to the short times
involved and the relatively low pulse intensity of state-of-
the-art x-ray sources. Bond formation, isomerization, and
electron transfer are fundamental chemical steps that evolve
on the femtosecond and picosecond time scales and ultrafast
techniques are therefore needed to probe molecules in time.
Three pump-probe techniques are currently used and all use
short optical pulses to initiate and clock the reaction. The
probe pulse is then either an optical pulse, an electron pulse,
or an x-ray pulse. X rays are particularly powerful for struc-
tural work since their wavelengths match atom-atom dis-
tances in molecules and they can penetrate condensed
samples much deeper than electrons. With the advent of third
generation synchrotrons such as the European Synchrotron
Radiation Facility ?ESRF?, pulsed quasimonochromatic
beams of hard x rays are now available with a pulse length of
around 100 ps and with 109to 1010photons/pulse. By vary-
ing the delay between the laser and the x-ray pulse, scatter-
ing patterns can be collected as a function of delay with each
delay being a 100 ps snapshot of the moving molecular
structure. Protein dynamics has been filmed in crystals1–4and
more recently in solution5and that has deepened our under-
standing of how the function of a protein is related to its time
dependent structure. Another new field is dissociation and
recombination dynamics of small molecules in solution,6–9
where transient structures of the molecules and their sur-
rounding medium can be studied in great detail.10,11
The most efficient way to conduct a time resolved ex-
periment is to film it, i.e., initiate a given reaction and then
open an x-ray shutter, and record the change in the scattered
x rays as a function of time. Unfortunately the time reso-
lution of area detectors is milliseconds at best, i.e., 107times
longer than the x-ray pulse. To exploit the single-pulse reso-
lution of a synchrotron, multiple scattering patterns from a
fixed laser/x-ray delay have to be accumulated in the detec-
tor, which slows down the speed of the data acquisition con-
siderably. The frequency of a pump-probe experiment might
vary from single-shot data acquisition in protein crystallog-
raphy ?0 Hz? to 3 kHz in flow cell experiments where the
sample is replaced at high speed. Note that the chopper sys-
tem is designed not only to produce short x-ray pulses but
also to slow down the x-ray frequency to a usable level for a
given sample.
The first high-speed chopper for synchrotrons was de-
signed in 1988 by Wilfried Schildkamp and Claude Prader-
vand, University of Chicago, and used at Chess for the first
single-pulse Laue experiments on lysozyme crystals.12The
beam was chopped by two slits in a rotating disk with me-
chanical bearings. An upgraded version of this design was
later used at ESRF in single pulse experiments in 1994.13In
parallel Kosciesza and Bartunik14selected single pulses with
a rotating mirror at Desy and McPherson et al.15,16designed
a chopper and a rotating crystal for the Advanced Photon
a?Author to whom correspondence should be addressed. Electronic mail:
wulff@esrf.fr.
REVIEW OF SCIENTIFIC INSTRUMENTS 80, 015101 ?2009?
0034-6748/2009/80?1?/015101/10/$25.00 © 2009 American Institute of Physics
80, 015101-1
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 2

Source ?APS?. Rotating mirrors and crystals can produce
submicrosecond opening times but the beam position is sen-
sitive to jitter in the rotation speed. A slot-based high-speed
chopper rotating in helium with air bearings and vacuum
windows was built by Gembicky et al.17for the ChemMat
beamline 15 at the APS. Rotating a 140 mm radius disk with
45 slits ?0.35 mm opening? at 502.9 Hz, they produced
2.1 ?s open windows for single-pulse beam at APS at a
record frequency of 22.6 kHz. They later added a heatload
chopper based on the same technology.18Note finally that the
ESRF triangular chopper design described below is also used
on the BioCars beamline 14 at APS, at Spring8, and at
KEK.19
The ESRF chopper system has three components. In
downstream order they are: a heatload chopper in the optics
hutch, a millisecond shutter, and high-speed chopper in the
experimental hutch near the sample. In the original design,
the heatload on the high-speed chopper was reduced by a 5
Hz heatload shutter in the optics hutch, which made white
beam experiments rather inefficient.20–22In the upgraded sys-
tem presented here, the heatload shutter is replaced by a 1
kHz heatload chopper and the previous 1 kHz high-speed
chopper is upgraded with a 1 kHz microsecond mode
?10 ?s? and a 3 kHz mode.
The timing of the chopper system is shown in Fig. 1 for
a single pulse experiment. The 300 ns open window from the
high-speed chopper picks out a single pulse from the
?50 ?s pulse from the heatload chopper. Both pulses repeat
at 1 kHz, so a millisecond shutter is needed to isolate one
single pulse. The heatload chopper is placed 29.5 m from the
x-ray source in front of the monochromator and the focusing
mirror. The heatload is reduced by a factor of ?20, which
means that the full peak power of the undulator can be used
without overheating the high-speed chopper. The high-speed
chopper is placed 53 m from the source in the ?quasi? fo-
cused beam 1.2 m from the sample. By default, both chop-
pers produce pulses at 986.3 Hz synchronized to the radio
frequency ?rf? clock of the synchrotron. The millisecond
shutter is used to either gate the detector during readout or to
produce subharmonics of the high-speed chopper at frequen-
cies up to 80 Hz.
II. BUNCH MODES FOR SINGLE-PULSE
EXPERIMENTS
To exploit the 100 ps time resolution from a single x-ray
pulse, the filling pattern of the storage ring has to provide a
sufficiently wide time gap for the high-speed chopper to se-
lect a single pulse. With present chopper technology and
beam sizes, the single pulse has to be separated by at least
100 ns from side pulses for the chopper to eliminate them. At
the ESRF there are currently four modes for single pulse
experiments: the 4-bunch and 16-bunch mode with equidis-
tant bunch fillings, the hybrid mode, and the 7/8 multibunch
mode. These modes run 80% of the time with 20% in uni-
form mode. We will now shortly describe these bunch modes
in detail.
It takes a 6 GeV electron 2.816 57 ?s to traverse the
844.39 m long synchrotron at ESRF. That corresponds to an
orbit frequency of 355.042 kHz. The energy lost to synchro-
tron radiation is compensated by rf cavities that operate at
352.201 664 MHz, the 992nd harmonic of the orbit fre-
quency. The rf cavities can support up to 992 bunches in a
?quasi? uniform fill with pulse separations of 2.839 ns. That
time gap is too small for mechanical single pulse isolation,
so the shortest achievable pulse is 200 ns given by the me-
chanical opening. The single-pulse modes are shown in Fig.
2. The time gaps in the 4-bunch and 16-bunch modes are 704
and 176 ns, respectively. In the hybrid mode, the time gaps
are slightly asymmetric, 429 and 344 ns. Finally the 7/8
mode has a 352 ns gap with a single bunch in the middle.
The low bunch charge in the 7/8 mode, 2.5–5.0 nC, shortens
the x-ray pulse to 60–80 ps ?full width at half maximum
FIG. 1. ?Color online? Scheme to isolate single x-ray pulses. When operated
in 16-bunch mode, the ESRF produces a train of ?100 ps x-ray pulses
separated by 176 ns. This pulse train is first chopped by the heatload chop-
per into 50 ?s ?FWHM? macrobunches at 1 kHz, thereby reducing by a
factor of 20 the transmitted average power striking the x-ray focusing mirror
and other downstream components. The millisecond shutter opens on de-
mand to isolate one heatload chopper macrobunch from the 1 kHz train of
bunches. The high-speed chopper isolates a single pulse from the center of
the transmitted macrobunch.
FIG. 2. ?Color online? Filling patterns suitable for single-pulse pump-probe
experiments at the ESRF: 4-bunch mode ?40 mA?, 16-bunch mode ?90 mA?,
hybrid mode ?24?8+1, 200 mA?, and 7/8 mode ?200 mA?.
015101-2Cammarata et al.Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 3

?FWHM??. For time resolved experiments over wide time
ranges, the 7/8 mode offers the advantage that for time de-
lays greater than ?1 ?s; the intensity of the pulse can be
boosted by changing the chopper phase by 180°.
III. THEORETICAL OPEN PROFILES FOR
A CHOPPED BEAM
Before discussing the choppers in detail, we will calcu-
late the open profiles for a Gaussian beam for the three main
geometries shown in Fig. 3. In the first case ?Fig. 3?a?? the
beam is chopped by a tunnel in the center of a disk with the
rotation axis perpendicular to the beam. This tunnel geom-
etry offers short open times as the beam is cut from above
and below simultaneously during the close/open/close cycle
by the four end-points. The second case is the tunnel-less
chopper where the tunnel ceiling is removed ?see Fig. 3?b??.
The beam is cut by one edge at a time, which typically
doubles the open time compared to a tunnel. However in the
tunnel-less case, the open time can be varied without chang-
ing the speed of the chopper by varying the chopper-to-beam
distance. In the third type, the slot chopper in Fig. 3?c?, the
beam passes through slots at the periphery of the disk with
the beam parallel to the rotation axis. Having N equidistant
slots around the periphery, the x-ray frequency is N times the
rotation frequency, which is the simplest way to increase the
frequency of the pulse train.
A. Tunnel-based chopping
We will now calculate the open profile I?t? for a rotating
tunnel, i.e., the time dependent spatially integrated intensity
for a close/open/close cycle in a continuous wave ?cw? beam.
Let R be the disk radius, 2h0the height of the tunnel, and f
the rotation frequency. The tunnel is assumed to be small,
2h0?R, and passing through the center of the disk. The ve-
locity of the cutting edges in the tunnel is perpendicular to
the incident beam and v=2?fR. The incoming beam defines
the x-axis, the y-axis is the rotation axis, and the beam is cut
in the z-direction. Let’s assume a Gaussian beam intensity
centered on z=0,
?2??exp?−
f?z? =
1
z2
2?2?,
?1?
where ? is the ?rms? beam size in the chopper ?FWHM
=2?2 ln 2?1/2??2.355??. When the tunnel rotates through a
close/open/close cycle, the four tunnel edges form a slit,
which opens and closes symmetrically around the beam ?see
Fig. 3?d??. When the chopper rotates clockwise away from its
fully open position, the cutting points P0and P0? will move to
P1and P1?, while reducing the aperture from above and be-
low symmetrically. When h?R we have
h?t? ? h0− ???t??R = h0− ?R?t? = h0− 2?fR?t??2?
for times ?t??t0=h0/?2?fR? and h?t?=0 for ?t??t0. Note that
t0=h0/v where v is the speed of the disk. The transmitted
intensity, integrated in the vertical direction, is the slit
integral
I?t? ??
−h?t?
h?t?
f?z?dz = erf?
h?t?
?2??= erf?
h0− 2?fR?t?
?2??,
?3?
where erf is the error function.
The aperture 2h?t? and intensity I?t? functions are shown
in Figs. 4?a? and 4?b?. The intensity is zero outside ?−t0,t0?.
The total open time, i.e. the base line of the open window is
?tbase= 2t0=
2h0
2?fR,
?4?
which is the time it takes for a cutting edge on the periphery
to move the distance 2h0. With typical parameters h0
=0.075 mm, f=1000 Hz, and R=100 mm, we get ?tbase
=239 ns. The tip velocity is v=628.3 m/s, which is greater
than the speed of sound in air ?343.6 m/s?. That implies that
the disk has to rotate in vacuum suspended in magnetic bear-
ings.
With a tunnel-to-beam-ratio h0/? of 1, the open profile
is triangular with a peak intensity of ?0.7 at the center of the
window, whereas for h0/? between 3 and 4, there is a pla-
teau at the center with full transmittance. In single pulse
experiments, this “top line” is a buffer for rotation jitter. The
top line has to be much greater than the timing jitter to avoid
cutting the single pulse when the open window arrives
slightly early or late. If a pulse arrives in the rising or falling-
edge zone, the beam size and the intensity are reduced. In a
cw beam, a ?small? fraction of the transmitted beam is re-
flected in the tunnel ceiling and floor, which degrades the
collimation and spectra purity ?pink beam?. This parasitic
scattering is readily removed by a collimator near the
sample. If the incoming or outgoing beam is reduced in size
by a slit s?2h0, the central part of the open profile is re-
FIG. 3. ?Color online? Rotor geometries employed in three classes of chop-
pers: ?a? the tunnel chopper, ?b? the tunnel-less chopper, and ?c? the slotted
chopper. The geometrical parameters relevant for computing the time-
dependent transmission of a Gaussian beam through a rotating tunnel are
shown in ?d?.
015101-3 Nanosecond chopperRev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 4

duced to a constant value, the integral of the beam through
the slit, erf?s/2/??2s??, whereas the rising and falling edges
are unchanged.
If the tunnel is placed off-center in the disk, the speed
term 2?fR in Eqs. ?2? and ?4? should be replaced with the
speed perpendicular to the tunnel when it is open. For the
triangular rotor used on beamline ID09B the perpendicular
speed is ?3?fR.
B. Tunnel-less based chopping
If the ceiling in the tunnel is removed as shown in Fig.
3?b?, the lower limit in the intensity integral becomes
hmin?t? ? 2?fR?t? − h0,
and the upper
?5?
hmax= ?.
?6?
The open profile is then
I?t? =?
hmin?t?
2?1 − erf?
?
f?z?dz =1
2?1 − erf?
2?fR?t? − h0
hmin?t?
?2???
?1
?2???.
?7?
The tunnel-less and tunnel based chopper are compared
Fig.4?c?
with
?=0.025 mm,
=100 mm, and f=1000 Hz. Note that the tunnel-less inten-
sity goes gradually to zero. The base line is roughly twice of
that for the tunnel. If an aperture s is added in front of the
tunnel-less chopper, the open window becomes
in
h0=0.075 mm,
R
I?t? =?
erf?
s
2?2??
for 0 ? ?t? ?h0− s/2
2?fR
1
2?erf?
s
2?2???− erf?
− h0− 2?fR?t?
?2? ? for
h0− s/2
2?fR
? ?t? ?h0+ s/2
2?fR?
?8?
FIG. 4. ?Color online? Chopper opening profiles: ?a? the angle dependence of the tunnel aperture, ?b? time-dependent intensity profile for beam/tunnel ratios
of h0/?=1, 2, 3, and 4, and ?c? comparison of time-dependent transmission for tunnel chopping vs tunnel-less and slotted wheel chopping. The parameters
used in this comparison are ?=0.025 mm, R=100 mm, h0=0.075 mm, and f=1000 Hz.
015101-4Cammarata et al. Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 5

with I?t?=0 for other times. The base line is determined by
the condition that the two erf arguments are equal,
?tbase=2h0+ s
2?fR,
?9?
which is longer than the equivalent tunnel by the ratio ?2h0
+s?/2h0. Pulse lengthening is thus the price to pay for tun-
able pulse length.
C. Slot-based chopping
The integration limits for a small rotating slot with an
aperture 2h0can be approximated by
hmin?t? ? 2?fRt − h0,
?10?
hmax?t? ? 2?fRt + h0,
which gives
I?t? =?
hmin?t?
− erf?
?11?
hmax?t?
f?z?dz ?1
?2???.
2?erf?
2?fRt + h0
?2??
2?fRt − h0
?12?
The rotating slot is essentially identical to the tunnel-less
case for the same h0as shown in Fig. 4?c?. The advantage of
the slot is the simplicity of machining many slots around the
periphery. However as the periphery has to be thin to reduce
the centrifugal pull on the rotation shaft, the absorption effi-
ciency is lower than in the previous choppers where the pre-
and postpulses hit the edges of the tunnel at normal inci-
dence. The above formula is equivalent to the intensity of a
slit scan in position across a Gaussian beam with a fixed gap
2h0. When the slit is much smaller than the beam, the scan
measures the profile of the beam. In time space, that is the
time it takes the radius to sweep across the beam. When the
slit is much greater than the beam the scan measures the slit
aperture.
A slot chopper is being built at Forschungszentrum
Jülich for soft x rays at Bessy, Berlin. It has a radius of 170
mm and 1252 equidistant slots around the periphery. Each
slot is 0.15 mm wide. Rotating at 998 Hz it produces 140 ns
pulses at 1.25 MHz! The rim is only 0.5 mm thick to reduce
the centrifugal force, so the absorption efficiency for x-ray
beams has to be considered. Note that nuclear scattering ex-
periments with microsecond isotopes could be done in 16-
bunch mode with a slot chopper.
IV. THE HIGH-SPEED CHOPPER
The ESRF high-speed chopper has a triangular titanium
rotor which can be inscribed in a circle with radius of 96.8
mm. The vacuum chamber, the rotor, and the drawing of the
beam positions are shown in Fig. 5. The rotation frequency is
986.3 Hz, the 360th subharmonic of the orbit frequency. The
rotor has a tunnel on one of the three end-faces of the tri-
angle. The tunnel is 165 mm long, 3 mm wide, and its height
varies linearly from 0.10 to 0.22 mm across the width. The
tunnel, which is semiopen, is formed by a channel with two
10 mm long roofs on the extremities. When the beam is
chopped by the 0.10 mm part, the open base line is 192 ns.
For the 0.22 mm part, it is 422 ns. When the beam is in the
tunnel, the rotor opens once per turn. The rotor has two 5
mm deep steps on either sides of the tunnel as shown in Figs.
5?b? and 5?c?. This allows making longer pulses up to 20 ?s
by lowering the chopper relative to the beam without having
to change the rotation frequency which simplifies the syn-
chronization of a pump probe experiment. Note that the
beam can also be positioned above the edges of the triangle
where the rotor opens three times per revolution ?3 kHz
mode?. The main parameters for the high-speed chopper are
shown in Table I. We will discuss of the 3 kHz mode in Sec.
IVB.
If the chopper is not needed during mirror alignment for
example, it can be removed horizontally by moving the
vacuum vessel by ?6.5 mm ?H?. In this out position, the
longer pulses from the heatload chopper become available
?50–86 ?s?.
FIG. 5. ?Color online? High-speed chopper. ?a? Side view of triangular rotor
in its vacuum chamber. The 20 mm thick walls of the stainless steel chamber
provides safety and x-ray shielding. ?b? Titanium rotor. Note the tunnel
along one edge and the step used for tunnel-less chopping. ?c? View of the
trapezoidal tunnel, the steps alongside the tunnel, and different working
positions for the x-ray beam ?red?.
TABLE I. Chopper and ms-shutter parameters.
Heat-load
chopper
High-speed
chopper
Millisecond
shutter
Rotor shape
Radius ?mm?
Rotation frequency ?Hz?
Pulse frequency ?Hz?
Tip speed ?m/s?
Tunnel height ?mm?
Tunnel width w ?mm?
Tunnel length ?mm?
?t_base_tunnel ??s?
Step height ?mm?
?t_base_step ??s?
Rotation jitter ??s; rms?
Circle
75
98.63
986.3
46.5
4.0
8.0
150.0
85
0.45–1.65
34–86
2
Triangle
96.8
986.3
986.3
599.8
0.10–0.22
3.0
167.7
0.19–0.42
5.0
0–20
0.003
Bar
30
2.681
0–80
0.51
0.3–4.0
5.0
60.0
600–4000
¯
¯
50
015101-5 Nanosecond chopperRev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 6

One advantage of a triangular rotor geometry is that the
tunnel is off axis. That makes it easier to taper the rotor in
the radial direction to increase the mechanical strength near
the rotation axis where the centrifugal stress accumulates. A
slight drawback of the triangle is the lower vertical speed,
which is reduced by cos?30°?=?3/2=0.866 for the triangle.
The tangential speed is 599.8 m/s ?1.7 times the speed of
sound in air? and the vertical speed 519.4 m/s. As a conse-
quence, the base line in the open window for the triangular
tunnel takes the form
?tbase=
2h0
?3?fR.
?13?
The chopper is installed in the ?quasi? focused beam 1.2
m from the sample where the beam is small which improves
the chopping efficiency. The beam size in the tunnel is
0.37 mm ?H??0.07 mm ?V?,
0.10 mm ?H??0.06 mm ?V? in the focus ?FWHM?.As the
beam is small horizontally compared to the 3 mm wide tun-
nel, the height of the tunnel that cuts the beam can be varied
from 0.10 to 0.22 mm by translating the chopper horizon-
tally.
The phase locking of the high-speed chopper is done
with a pickup signal from a small magnet integrated in the
rotor. This signal is compared with a time to live ?TTL?
reference signal at 986.3 Hz, which is a frequency divided
?360? and phase shifted copy of the orbit clock. The time
delay of the TTL signal can be delayed to a resolution of 5.6
ns, which is sufficient for centering the open window onto
one pulse. The internal clock in the chopper controller runs
at 352 MHz and the pickup signal is measured to 2.8 ns
?rms? resolution. The measured rotation jitter is 2.8 ns ?rms?
depending on the mechanical vibrations around the chopper.
In the direction of the beam, the high-speed chopper is
followed by a slit that adjusts the beam size on the sample. If
the vertical slit is smaller than the tunnel height, the ?cw?
open profile becomes trapezoidal in shape. In the central
part, the open profile is constant for a time and the top line of
the open window is
whichconverges to
?ttop?tunnel? =2h0− s
?3?fR.
?14?
The base line is
?tbase?tunnel? =
2h0
?3?fR.
?15?
The equivalent expressions for the tunnel-less case are
?ttop?tunnel-less? =2h0− s
?3?fR
?16?
and
?tbase?tunnel-less? =2h0+ s
?3?fR.
?17?
These open windows are shown schematically in Fig. 6. Note
that the top line is the same in the two cases, but that the
tunnel-less base line is prolonged by ?2h0+s?/2h0as for a
full Gaussian beam.
We will now compare the open times for tunnel and
tunnel-less chopping in the 16-bunch mode, the most de-
manding chopper mode but also the most frequent mode for
single pulse experiments. In 16 bunch mode the pulses are
separated by 176 ns. With 2h0=0.12 mm, R=96.8 mm, f
=986.3 Hz, and a vertical slit s=0.06 mm, we get ?ttop
=116 ns and ?tbase=231 ns. If the open window is centered
on the pulse, the pulse will pass the tunnel with a clearance
to the walls of only 0.03 mm. Moreover the pre- and post-
pulses are 61 ns from the edges of the open window, which is
much greater than the 2.8 ns ?rms? rotation jitter. The side
pulses are thus perfectly blocked. In addition the 116 ns long
top line, compared with the 2.8 ns ?rms? rotation jitter, en-
sures a high pulse to pulse stability, which is particularly
important for scanning diffraction experiments. In the
equivalent tunnel-less case with h0=0.06 mm, the base line
is 347 ns and the top line 116 ns. The distance to the side
pulses is 2.5 ns, which is insufficient in view of the
2.8 ns ?rms? rotation jitter. Raising the chopper to h0
=0.05 mm, we get ?tbase=308 ns and ?ttop=77 ns, which
increases the side pulse clearance to 22 ns, which is fine.
Finally a long pulse is obtained for h0=2.5 mm, which gives
a nearly rectangular pulse shape with ?tbase=9.74 ?s and
?ttop=9.51 ?s.
A. Tunnel-less chopping at 3 kHz
The high-speed chopper is mainly designed to produce
pulses at 1 kHz, which, compared to the 5.7 MHz frequency
of a 16-bunch beam, makes extremely poor use of the beam.
In liquid experiments where the sample can be exchanged
rapidly in a jet, it would be advantageous to run at higher
frequencies. Rotating the chopper above 1000 Hz is not pos-
sible with the present rotor due to a resonant shaft-bending
mode at 1100 Hz, followed by centrifugal fracture at 1450
Hz. The simplest option is to use the three sides of the rotor
in tunnel-less mode with the beam outside the rotor and in a
lateral position where it never hits the tunnel ?see Fig. 5?c??.
This poses the following challenge: the rotor spins around its
center of mass, which, due to finite tolerances, differs
slightly from the geometrical center. Is the rotor machined
and balanced well enough to isolate single pulses in the 16-
bunch mode at 3 kHz? The challenge is to get the same open
window from the three sides, i.e., that the beam heights h0,
for the three sides are sufficiently identical that the open
windows select one pulse only and that they arrive at the
right time. The rotor turns in the clockwise direction and face
1 is the one with the tunnel and faces 2 and 3 follow in time.
We find faces 2 and 3 to be 16 and 19 ?m closer to the beam
than face 1, respectively. Moreover the angles in the triangle
have to be 60° to avoid systematic shifts in the time position
of the open window. For example, with an angular speed of
0.355 mdeg/ns at 986.3 Hz, it follows that if one rotor angle
is 3.55 mdeg too big, the associated open window will arrive
10 ns too late. By tracing the open profiles in time from the
three faces, the three angles were measured to be 59.998?1?°,
60.003?1?°, and 59.999?1?°, which is just sufficient for 3 kHz
operation ?see below?.
We saw previously that single pulse selection in tunnel-
less typically requires h0=0.05 mm and s=0.06 mm. So the
015101-6Cammarata et al.Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 7

beam distance to the floor is only 0.02 mm. A quick estimate
shows that h0can vary between 0.03 and 0.05 mm in single
pulse mode. The lower limit is imposed by intensity conser-
vation and the upper by side-pulse discrimination.
We characterized the 3 kHz pulse train with a continuous
cw beam. h0was moved to 0.05 mm on the tunnel side ?side
1? and the slit was s=0.06 mm. This gives a 308 ns base
line, which is fine for the 16-bunch mode. The open window
was measured with a Cyberstar scintillation detector
coupled to a 6 GHz oscilloscope ?LeCroy 6 GHz,
WaveMaster6820A?. The pulse train is shown on the milli-
second scale in Fig. 7?a? and with a nanosecond zoom in Fig.
7?b?. The pulse in the middle ?pulse 3? is from face 1 with
the tunnel and pulses 1 & 4 and 2 & 5 are produced by face
2 and 3, respectively. In Fig. 7?b? the nanosecond open pro-
files are offset in time by 0, 1/3, and 2/3 of the principal
revolution time of the rotor. Note that pulse 3 is trapezoidal
and we use that pulse to define time zero. In contrast, pulse 1
?4? and 2 ?5? are triangular, slightly shorter, and shifted in
time by a few nanoseconds ?see Table II?. That means that
the associated rotor sides are closer to the beam and their
angles differ slightly from 60°.
We conclude that 3 kHz single pulse selection is possible
in 16-bunch mode, but the margins in space and time are
very tight for routine operation. Note that the rising edge of
pulse 1 is 10 ns from the prepulse at −176 ns. This means
that the phase of the rotor has to be stable to 10 ns over the
duration of an experiment, which might run for several days.
A future rotor could have a bigger radius ?150 mm?
while thinning the tips from 6 to 2 mm to minimize the
centrifugal stress on the rotation shaft. With this rotor at
986.3 Hz, the settings for single pulse selection in 16-bunch
mode would be h0=0.09 mm, s=0.06 mm, ?tbase=298 ns,
and ?ttop=231 ns. The distance to the satellite pulses would
be 27 ns, i.e., about ten times the rms jitter and therefore
safe. Also if the tunnel height varies between 0.10 and 0.22
mm, the base lines will vary from 128 to 295 ns. Alterna-
tively the tunnel could be 50% higher and accept larger
beams.
V. THE HEATLOAD CHOPPER
The first white beam experiments were done without a
heatload chopper, which meant that the high-speed chopper
had to withstand 60–120 W of power from the focused white
beam during exposure of the detector. Because a rotor in
magnetic bearings is only cooled by radiation, it was neces-
sary to reduce the beam power by closing down the aperture
of the primary slits and reducing the power of the undulator
by opening the gap to about 9 mm rather than using the full
FIG. 6. ?Color online? Beam positions and corresponding trapezoidal open windows for the high-speed chopper. The tunnel mode is shown on the left and the
tunnel-less mode on the right. The open profiles are for h0=s=0.06 mm. The chopper is capable of isolating a single pulse out of the 16-bunch pulse train
?176 ns separation? with both tunnel and tunnel-less chopping modes.
015101-7 Nanosecond chopper Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 8

power position at 6 mm. In practice the total power was
reduced by a factor 6 to avoid overheating the high-speed
rotor with a similar reduction in peak intensity of the pulses.
The solution was to install a water cooled heatload chopper
upstream the monochromator and mirror in the optics hutch.
The heatload chopper is a Cu disk that rotates in vacuum
about a horizontal axis perpendicular to the beam. The radius
of the disk is 75 mm and the disk has five tunnels ?see Fig.
8?. The tunnels are 4 mm high, 8 mm wide, and 150 mm
long. The size of the beam in the heatload chopper is up to
7 mm ?H??0.7 mm ?V?. Normally the disk rotates at
98.63 Hz at the 3600 subharmonic of the orbit frequency, i.e.
ten times slower than the high-speed chopper, but it can also
rotate at any subharmonic between 3 and 98.63 Hz. As the
five tunnels open ten times per revolution, the heatload chop-
per produces pulses at 986.3 Hz in phase with the high-speed
chopper. The chopper parameters are listed in Table I. The
vertical beam size is smaller than the tunnel height, which
makes the temporal open profile trapezoidal with
?ttop?tunnel? =2h0− s
2?fR
?18?
and
?tbase?tunnel? =
2h0
2?fR.
?19?
The denominator in Eqs. ?18? and ?19? is the vertical speed at
the end of the tunnels. With 2h0=4 mm, s=0.7 mm, R
=75 mm, and f=98.63 Hz, we get v=46.5 m/s, ?ttop
=71.0 ?s, and ?tbase=86.0 ?s. By integrating the trapezoi-
dal in time, we get the duty cycle of the chopped beam
D = N?2h0−s
?R
2?
,
?20?
where N is the number of tunnels. For N=5, we get D
=0.077, which gives a heatload reduction of D−1=12.9. For a
typical white beam in 16-bunch mode at 90 mA, the heatload
chopper reduces the power from 62 to 4.8 W, which ?still?
raises the temperature of the high-speed chopper from 20 to
45 °C.
The tunnel height 2h0in the heatload chopper was cho-
sen to be big ?4 mm? to have margins for rotation jitter,
which was expected to be as high as 10 ?s ?rms? due to the
mechanical bearings and the flow of cooling water in the
rotation shaft. A 70 ?s top line is compatible with a jitter of
10 ?s ?rms? without cutting the central part of the pulse. The
measured jitter at 100 Hz is 2 ?s so the tunnel height could
be reduced in future designs. At 10 Hz the jitter increases to
50 ?s and at 3 Hz, the lowest possible speed, the jitter is
200 ?s.
The low jitter at 100 Hz makes it more efficient to run
the heatload chopper in tunnel-less mode with a small dis-
tance h0to the tunnel floor. By adapting Eqs. ?16? and ?17? to
FIG. 7. ?Color online? Transmission of the high-speed chopper when oper-
ated in 3 kHz tunnel-less chopping mode ?recorded with a cw beam?. ?a?
Pulse train on the millisecond time scale. ?b? Time-dependent transmission
from the three faces of the triangle is superimposed by shifting each by tN0
?theoretical arrival time for pulse N given a perfect equilateral rotor?.
TABLE II. Pulse parameters for the principal pulses from the high-speed
chopper in 3 kHz mode.
Pulse
?tbase
?ns?
?t0
?ns?
h
?mm?
1
2
3
295
270
317
−12.7
5.4
0
0.047
0.04
0.052
FIG. 8. ?Color online? Heatload chopper. ?a? Drawing of chopper rotor in its
vacuum vessel. The position of the chopper relative to the x-ray beam is
optimized by motorized translation in the vertical and horizontal directions.
?b? Photo of the five-tunnel rotor in its vacuum vessel. The rotor position is
sensed by a coil located near the top of the photo. This coil generates a
current pulse when SmCo pickup magnets pass by ?seen as gray dots near
the periphery of the rotor?.
015101-8Cammarata et al.Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 9

a tunnel-less disk with h0=1 mm, s=0.7 mm, R=75 mm,
and
f=98.63 Hz,weget
=58.1 ?s. The duty cycle is
?top=28.0 ?sand
?tbase
D = N2h
?R,
?21?
which gives D=0.042 and a heat reduction of D−1=23.6.
This mode is realized by shifting the chopper 1 mm verti-
cally, which leaves 0.65 mm clearance to the floor when the
tunnel is fully open. That distance is ten times greater than
machining and balancing errors in the tunnel positions and
therefore is a good working point.
The Cu disk is driven by a synchronous direct drive
motor in air. The vacuum passage is a ferrofluidic feed
through with cooling water running in the rotation axis. The
disk is cooled from one side by a stainless steel flange, which
is attached to the ferrofluid feed through. Theoretically, the
heatload chopper can absorb 450 W with a temperature rise
of 104 K ?1.3 K in the cooling water?. The disk temperature
is measured by a Raytech infrared detector. The detector
shows higher temperatures than expected, which indicates a
poor heat transfer between the cold flange and the disk. We
plan to improve this by explosion melting the stainless steel
flange onto the Cu disk, which seems the best way to bond
two different metals together in a water-leak-tight fashion
under UHV conditions.
The disk is dynamically balanced at 100 Hz to minimize
eccentricity and friction in the bearings. The rotation speed is
synchronized using a pick up signal from a SmCo magnet on
the disk. In the first version of the heatload chopper, the
magnet was made of NdFeB, but it demagnetized at tempera-
tures around 150 °C. The new SmCo magnet is weaker ?20
mT?, but it maintains its magnetization up to 250 °C. By
comparing the difference between the pickup and a phase
shifted TTL at 98.63 Hz, the disk is either accelerated or
decelerated until the two signals coincide within the time
resolution of the controller ?11.3 ns, rf/8?.
The vacuum vessel of the heatload chopper is motorized
in the horizontal and vertical positions. The tunnel is 8 mm
wide and receives a 7 mm wide beam. When not used, the
heatload chopper is moved 8 mm horizontally toward the
synchrotron.
We finally note that the heatload chopper is readily
modified for 3 kHz operation by increasing the number of
tunnels from 5 to 15. Also if the tunnel height is reduced
from 4.0 to 1.33 mm, the heatload reduction remains the
same as at present.
VI. THE MILLISECOND SHUTTER
Many pump-probe experiments cannot run at the default
986.3 Hz chopper frequency due to heatload from the laser
or the presence of slower-time-scale kinetics in the sample.
Hence, a millisecond shutter is needed to lower the pulse
frequency on the sample. The millisecond shutter is installed
in vacuum just before the high-speed chopper. The shutter is
water cooled via its ferro-fluid feed-through to the stepper
motor in air. The millisecond shutter can be seen upstream
the high-speed chopper in Fig. 5
The millisecond shutter is a rotating tunnel which is wire
cut from a 60?8.5?4 mm3block of tungsten carbide. The
tunnel is 60 mm long, 5 mm wide, and the tunnel height
increases from 0.3 which gives the tunnel a trapezoidal cross
section. As the beam is small in the millisecond shutter,
0.40 mm ?H??0.07 mm ?V?, the open time can be
changed by moving the shutter horizontally without chang-
ing other parameters ?motor speed and delay?. The bar is
mounted on a rotation axis driven by a high-speed stepper
motor ?PV267-D2.8BA, Oriental Motor?. To optimize the
FIG. 9. ?Color online? Millisecond shutter. ?a? Geometric parameters for the tunnel. ?b? Tunnel transmission as a function of angle and time. ?c? Time-
dependent angular position of the tunnel ?top? when operating the shutter at 10 Hz. The tunnel transmits a 1.4 ms pulse each time it toggles back and forth
over a 6° range. Higher acceleration will be needed to isolate a single pulse when operating the high-speed chopper in the 3 kHz mode.
015101-9 Nanosecond chopper Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

Page 10

speed of the stepper motor, the angular step size is rather
coarse with 0.9° per step. When working with a 0.6 mm
tunnel, the tunnel is already closed when the shutter is ro-
tated by ?0.6°, i.e., the shutter is only open for one single
position of the motor. At full speed, the angular speed is
965°/s ?f=2.681 Hz?. The open time can be calculated with
the heatload chopper formulas ?18? and ?19?, with R
=30 mm. The base line open window is 0.6 ms for 2h0
=0.3 mm and 4.0 ms for 2h0=2.0 mm ?see Table I?.
In the pulsed mode the shutter is operated in the follow-
ing way. In its close position, the tunnel angle is rotated
+2.7° away from the beam direction, which is three steps
from the open position ?see Fig. 9?a??. The shutter opens “on
the fly” by moving the tunnel to −2.7°. The motion of the
shutter is monitored by a 2000 step/turn linear encoder as
shown in Fig. 9?b?. Note that the shutter does not stop per-
fectly at −2.7°; it overshoots slightly but settles quickly at
the correct value after 2–3 small oscillations without cross-
ing the open region ?0.6 deg while settling down. The am-
plitude of the “overshoots” was reduced by mounting a
damper on the stepper motor. The millisecond shutter is
phased in time with the chopper pulse using a TTL signal
derived from the timing module to start the execution of a
program in the motor driver card. By carefully optimizing
the current in the motor, the close/open/close cycle can be
repeated at 80 Hz without loosing steps.
VII. CONCLUSION AND OUTLOOK
The chopper system for pump-probe experiments on
ID09B has been presented together with the bunch structures
for single pulse experiments at the ESRF. The use of high
brightness white beams is now possible, thanks to a heatload
chopper that reduces the thermal load on downstream ele-
ments by a factor of around 20. In default mode the two
choppers produce pulses at 1 kHz, but they can also be con-
figured at other subharmonics of the orbit frequency between
0 and 3 kHz. With the high-speed chopper in tunnel mode,
open times as short as 200 ns can be obtained with a jitter of
2.8 ns ?rms?. That is short enough to isolate single pulses of
x rays from all timing modes at the ESRF. The shortest x-ray
pulse is 60 ps long and produced by a low bunch charge in
the 7/8 mode. Longer pulses train up to 20 ?s can be made
by the step on the high-speed chopper or by the heatload
chopper alone ?86 ?s?. Finally a pulsed millisecond shutter
is used to lower the frequency of the experiments from 80 Hz
to single shot.
The prospect of running experiments at 3 kHz is demon-
strated with the existing triangular rotor in the high-speed
chopper. In general the high frequency option might be use-
ful for lower electron energy synchrotrons between 2 and 3
GeV to compensate for their lower brightness above 10 keV.
Likewise at the ESRF with 6 GeV electrons, a 3 kHz option
might compensate for the loss of intensity from undulators
above 20 keV and compensate for losses associated with the
use of multilayer optics to control the bandwidth of the white
beam.
ACKNOWLEDGMENTS
We would like to thank Wilfried Schildkamp, Claude
Pradervand, Keith Moffat, Dominique Bourgeois, Thomas
Ursby, Loys Goirand, Jean-Luc Revol, Roland Taffut, Herve
Gonzales, and Ernesto Paiser for their contributions to this
project.
1V. Šrajer, T. Teng, T. Ursby, C. Pradervand, Z. Ren, S. Adachi, W.
Schildkamp, D. Bourgeois, M. Wulff, and K. Moffat, Science 274, 1726
?1996?.
2B. Perman, V. Šrajer, Z. Ren, T.-Y. Teng, C. Pradervand, T. Ursby, F.
Schotte, M. Wulff, R. Kort, K. Hellingwerf, and K. Moffat, Science 279,
1946 ?1998?.
3D. Bourgeois, B. Vallone, F. Schotte, P. Anfinrud, M. Wulff, and M.
Brunori, Proc. Natl. Acad. Sci. U.S.A. 100, 8704 ?2003?.
4H. Ihee, S. Rajagopal, V. Šrajer, R. Pahl, S. Anderson, M. Schmidt, F.
Schotte, P. A. Anfinrud, M. Wulff, and K. Moffat, Proc. Natl. Acad. Sci.
U.S.A. 102, 7145 ?2005?.
5M. Cammarata, M. Levantino, F. Schotte, P. A. Anfinrud, F. Ewald, J.
Choi, A. Cupane, M. Wulff, and H. Ihee, Nat. Methods 5, 10 ?2008?.
6A. Plech, M. Wulff, S. Bratos, F. Mirloup, R. Vuilleumier, F. Schotte, and
P. A. Anfinrud, Phys. Rev. Lett. 92, 125505 ?2004?.
7H. Ihee, M. Lorenc, T. K. Kim, Q. Y. Kong, M. Cammarata, J. H. Lee, S.
Bratos, and M. Wulff, Science 309, 1223 ?2005?.
8T. K. Kim, M. Lorenc, J. H. Lee, M. Lo Russo, J. Kim, M. Cammarata, Q.
Kong, S. Noel, A. Plech, M. Wulff, and H. Ihee, Proc. Natl. Acad. Sci.
U.S.A. 103, 9410 ?2006?.
9J. Davidsson, J. Poulsen, M. Cammarata, P. Georgiou, R. Wouts, G.
Katona, F. Jacobson, A. Plech, M. Wulff, G. Nyman, and R. Neutze, Phys.
Rev. Lett. 94, 245503 ?2005?.
10M. Wulff, S. Bratos, A. Plech, R. Vuilleumier, F. Mirloup, M. Lorenc, Q.
Kong, and H. Ihee, J. Chem. Phys. 124, 034501 ?2006?.
11M. Cammarata, M. Lorenc, T. K. Kim, J. H. Lee, Q. Y. Kong, E.
Pontecorvo, M. Lo Russo, G. Schiró, A. Cupane, M. Wulff, and H. Ihee, J.
Chem. Phys. 124, 124504 ?2006?.
12A. D. LeGrand, W. Schildkamp, and B. Blank, Nucl. Instrum. Methods
Phys. Res. A 275, 442 ?1989?.
13D. Bourgeois, T. Ursby, M. Wulff, C. Pradervand, A. Legrad, W.
Schildkamp, S. Labouré, V. Srajer, T. Y. Teng, M. Roth, and K. Moffat, J.
Synchrotron Radiat. 3, 65 ?1996?.
14D. Kosciesza and H. Bartunik, J. Synchrotron Radiat. 6, 947 ?1999?.
15A. McPherson, J. Wang, P. L. Lee, and D. M. Mills, J. Synchrotron Radiat.
7, 1 ?2000?.
16A. McPherson, W.-K. Lee, and D. Mills, Rev. Sci. Instrum. 73, 2852
?2002?.
17M. Gembicky, D. Oss, R. Fuchs, and P. Coppens, J. Synchrotron Radiat.
12, 665 ?2005?.
18M. Gembicky, S. Adachi, and P. Coppens, J. Synchrotron Radiat. 14, 295
?2007?.
19S. Nozawa, S. Adachi, J. Takahashi, R. Tazaki, L. Guerin, M. Daimon, A.
Tomita, T. Sato, M. Chollet, E. Collet, H. Cailleau, S. Yamamoto, K.
Tsuchiya, T. Shioya, H. Sasaki, T. Mori, K. Ichiyanagi, H. Sawa, H.
Kawata, and S. Koshihara, J. Synchrotron Radiat. 14, 313 ?2007?.
20M. Wulff, A. Plech, L. Eybert, R. Randler, F. Schotte, and P. Anfinrud,
Faraday Discuss. 122, 13 ?2003?.
21F. Schotte, S. Techert, P. A. Anfinrud, V. Srajer, K. Moffat, and M. Wulff,
in Third-Generation Hard X-ray Synchrotron Radiation Sources, edited by
D. Mills ?Wiley, New York, 2002?, pp. 345–401.
22B. Lindenau, J. Raebiger, S. Polachowski, and J. K. Fremerey, AIP Conf.
Proc. 705, 1019 ?2004?.
015101-10Cammarata et al.Rev. Sci. Instrum. 80, 015101 ?2009?
Downloaded 03 Aug 2009 to 128.231.5.75. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp