beamlines
J. Synchrotron Rad. (2017). 24, 521–530 https://doi.org/10.1107/S1600577516020579 521
Received 3 November 2016
Accepted 28 December 2016
Edited by J. F. van der Veen
Keywords: inelastic x-ray scattering;
X-ray Raman spectroscopy; beamline;
spectrometer; ESRF.
A large-solid-angle X-ray Raman scattering
spectrometer at ID20 of the European Synchrotron
Radiation Facility
S. Huotari,
a,b
* Ch. J. Sahle,
a
Ch. Henriquet,
a
A. Al-Zein,
a,c
K. Martel,
a
L. Simonelli,
a,d
R. Verbeni,
a
H. Gonzalez,
a
M.-C. Lagier,
a
C. Ponchut,
a
M. Moretti Sala,
a
M. Krisch
a
and G. Monaco
a,e
a
ESRF – The European Synchrotron, CS40220, 38043 Grenoble Cedex 9, France,
b
Department of Physics,
PO Box 64, University of Helsinki, 00014 Helsinki, Finland,
c
Department of Physics, Faculty of Science,
Beirut Arab University, Beirut 11072809, Lebanon,
d
ALBA Synchrotron Light Facility, Carrer de la Llum 2-26,
08290 Cerdanyola del Valle
´
s, Barcelona, Spain, and
e
Dipartimento di Fisica, Universita
`
di Trento,
via Sommarive 14, 38123 Povo (TN), Italy. *Correspondence e-mail: simo.huotari@helsinki.fi
An end-station for X-ray Raman scattering spectroscopy at beamline ID20 of
the European Synchro tron Radiation Facility is described. This end-station is
dedicated to the study of shallow core electronic excitations using non-resonant
inelastic X-ray scattering. The spectrometer has 72 spherically bent analyzer
crystals arranged in six modular groups of 12 analyzer crystals each for a
combined maximum flexibility and large solid angle of detection. Each of the six
analyzer modules houses one pixelated area detector allowing for X-ray Raman
scattering based imaging and efficient separation of the desired signal from
the sample and spuri ous scattering from the often used complicated sample
environments. This new end-station provides an unprecedented instrument for
X-ray Raman scattering, which is a spectroscopic tool of great interest for the
study of low-energy X-ray absorption spectra in materials under in situ
conditions, such as in operando batteries and fuel cells, in situ cata lytic reactions,
and extreme pressure and temperature conditions.
1. Introduction
X-ray Raman scattering (XRS) spectroscopy is a versatile tool
for studying shallow X-ray absorption edges usin g hard
X-rays. It has proven an invaluable technique for the study of
electronic excitations in a variety of sample systems such as
crystals (Mattila et al., 2005; Sternemann et al., 2008; Pylk-
ka
¨
nen et al., 2010; Huotari et al., 2012; Nyrow et al., 2014a,b;
Tse et al., 2014; Pascal et al., 2014; Galambosi et al., 2006;
Conrad et al., 2009; Moretti Sala et al., 2014), liquids (Wernet
et al., 2004; Pylkka
¨
nen et al., 2011; Sahle et al., 2013, 2016 a;
Juurinen et al., 2013, 2014; Niskanen et al., 2015) and gases
(Sakko et al., 2011; Inkinen et al., 2013; Zhu et al., 2011). The
inherent properties of hard X-rays used for this technique
render XRS perfectly suitable for the study of soft X-ray
absorption spectroscopy (XAS) or the corresponding coun-
terpart of electron-energy-loss spectroscopy (EELS), often
named energy-loss near-edge structure (Egerton, 2011)
(ELNES), spectra from complex sample environments that
prohibit other probes such as soft X-rays or electrons. XRS
provides a truly bulk-sensitive probe for samples inside,
for example, in situ catalytic reactors, in operando electro-
chemical cells, and high-pressure diamond anvil cells. Over the
past decades, XRS has been applied to solve geoscientific
questions by studying shallow core edges under extreme
ISSN 1600-5775
pressure and temperature conditions (Sahle et al., 2013; Mao et
al., 2003; Lee et al., 2008; Rueff & Shukla, 2010; Tse et al., 2011;
Shieh et al., 2013; Ding et al., 2014), follow chemical reactions
in situ (Sahle et al., 2016a; Miedema et al., 2012; Inki nen et al.,
2015) and study liquid samples under well defined thermo-
dynamic conditions (Pylkka
¨
nen et al., 2011; Juurinen et al.,
2013, 2014; Niskanen et al., 2015; Sahle et al., 2016 b). It can
even be used as a contrast mechanism for three-dimensional
imaging (Huotari et al., 2011; Sahle et al., 2017a,b).
Besides true bulk sensitivity, scattering of a hard X-ray
photon by an electron can lead to a significant amount of
exchanged momentum. This can be exploited to study the full
electronic structure of materials beyond the dipole limit
(Mattila et al., 2005; Krisch et al., 1997; Soininen et al., 2005).
Furthermore, the non-resonant nature of the XRS process
renders the signal independent of the de-excitation channel,
which leads to negligible self-absor ption effects and a more
reliable spectral shape in XRS than in the complementary
soft- or tender-XAS techniques. Using XRS, both the XANES
and the EXAFS (Huotari et al., 2012; Bergmann et al., 2007;
Hiraoka et al., 2013, 2016) regions are accessible.
An obvious drawback of XRS is the orders-of-magnitude
weaker scattering cross section in comparison with the prob-
ability for photoelectric absorption. This can be compensated
for by using light sources with a very high brilliance and
efficient signal collection. This has been the design target of
high-efficiency XRS end-stations around the world (see, for
example, Sokaras et al., 2012; Verbeni et al., 2009; Fister et al.,
2006; Cai et al., 2003) and it has also been the guiding motive
for the design of the spectrometer presented here.
In this article, we present the new XRS end-station
currently installed and commissioned at the inelastic X-ray
scattering spectroscopy beamline ID20 of the European
Synchrotron Radiation Facility (ESRF, Grenoble, France).
First, we briefly introduce the theoretical background neces-
sary for XRS in x2, then we will give a detailed description of
this new large-solid-angle spectrometer and provide its key
characteristics in x3. We will further describe in brief the data
analysis (x4), show some representative data taken with the
end-station (x5), and provide conclu sions and an outlook in x6.
2. Theoretical background
The measured signal in XRS, and in non-resonant inelastic
X-ray scatter ing (IXS) in general, is proportional to the
double differential cross section (Schu
¨
lke, 2007),
d
2
d d!
¼
d
d
Th
Sðq;!Þ; ð1Þ
which consists of the Thomson scattering cross section,
d
d
Th
¼ r
2
e
!
2
!
1
^
ee
1
^
ee
2
ðÞ; ð2Þ
multiplied by the dynamic structure factor
Sðq;!Þ¼
X
i;f
p
i
hfj exp iq r
ðÞj
ii
2
E
i
E
f
þ !
ðÞ
: ð3Þ
Here, ! is the energy transferred to the sample, !
1
is the
energy of the incident photons, !
2
is the energy of the scat-
tered photons, r
e
is the classical electron radius, and
^
ee
1
(
^
ee
2
)is
the polarization vector of the incident (scattered) photons.
Further, jii is the electron initial state (typically a shallow or
low-energy core level) weighted by its probability p
i
and the
summation is over all possible final states jfi. The dynamic
structure factor is the physical quantity that contains all
information obtainable from the sample via non-resonant
inelastic X-ray scattering. A few curious facts can be drawn
from the above formalism. First of all, in the limit of small
momentum transfer jqj, the operator can be expanded as
expðiq rÞ1+iq r ðq r Þ
2
=2+.... The second term is a
dipolar operator and dominates in the limit q r <1.Ithas
been shown by Mizuno & Ohmura (1967) that in this limit the
dynamic structure factor is directly related to the photo-
absorption spectrum. However, interesting phenomena can
occur since the transferred momentum can reach large values
in comparison with the radial extent of the wavefunction jii.
Then, the higher-order, i.e. non-dipolar, terms in the Taylor
expansion can become dominant. For instance, transitions
from an s-type symmetry initial state to s-type final states
become allowed. This has been used in the investigations
of dipole-forbidden excitons in the K-edges of elements
(Ha
¨
ma
¨
la
¨
inen et al., 2002) as well as in the studies of pre-
resonances of rare-earth ions (Gordon et al., 2007; Huotari et
al., 2015). This q-dependence can be exploited to measure
natural linear dicroism in cubic systems (Gordon et al., 2009).
Other experimental results demonstrate the high potential
of XRS as a bulk-sensitive technique for the characterization
of the electronic properties of actinide materials by studying
the 5d–5f electric multipole transitions at high momentum
transfer (Caciuffo et al., 2010). A highly interesting use of XRS
is furthermore the study of low-energy core-electron excita-
tion spectra in complex environments (high-pressure diamond
anvil cells, in situ chemical reaction cells, etc.) in which soft-
XAS or EELS studies are not possible due to the highly
absorbing sample containers (Sahle et al., 2013; Lee et al.,
2008).
3. Spectrometer
The large-solid-angle XRS spectrometer has been constructed
and commissioned as one of two spectrometers on port ID20
within the framework of the ESRF Upgrade Phase I. The new
beamline and the spectrometer are an upgrade to one of the
two branches of the inelastic scattering beamline previously
located on port ID16 (Verbeni et al., 2009).
ESRF operates a 6 GeV storage ring with a maximum ring
current of 200 mA. The straight section for ID20 contains one
26 mm-period and three revolver undulators that are switch-
able between 32 mm and 26 mm periods. The optics that
prepare the incident beam are the same as for the second
spectrometer on ID20, which is dedicated to high-resolution
beamlines
522 S. Huotari et al.
X-ray Raman scattering spectrometer at ESRF ID20 J. Synchrotron Rad. (2017). 24, 521–530
inelastic X-ray scattering and which will be described else-
where (More tti Sala et al., 2017). The incident-photon beam is
collimated in the vertical plane by a white-beam mirror, which
also serves as a heat-load filter for a liquid-nitrogen-cooled
double-crystal Si(111) pre-monoch romator. The pre-mono-
chromator can work alone as a fixed-exit monochromator or
be coupled with a post-monochromator to form a fixed-exit
ensemble. Various post-monochromator options are available,
e.g. a Si(311) channel-cut, or four-bounce Si(nnn)(n =3,4,5)
or Si(nn0) (n = 4, 6, 8) post-monochromators. Also, a back-
scattering Si(nnn) channel-cut monochromator is available for
highly specialized ultra-high-energy-resolution applications.
Downstream, a toroidal mirror focuses on a secondary source
at a distance of 53 m from the source and 13.7 m from the
sample position. This secondary source is refocused by a
Kirkpatrick–Baez (KB) mirror ensemble to a 8 mm 16 mm
spot size (V H) at the sample position. The working
distances of the KB mirror pair are 1.0 m and 0.5 m. Typical
photon fluxes at the sample position without post-mono-
chromator are 4 10
12
,7 10
13
and 5 10
13
photons s
1
at
6.5 keV, 9.7 keV and 12.9 keV, respectively. With a Si(311)
channel-cut, these numbers are reduced by a factor of seven.
For details on the energy resolution, see subsequent subsec-
tions.
The spectrometer is optimized for non-resonant IXS
experiments with an energy resolution of 0.3–2.0 eV, and
covers a large solid angle for the collection of the scattered
radiation. The leading goal was to place a large number of
analyzer crystals in positions corresponding to well defined
momentum transfers q. These values of q are tunable to match
the requirements of the sample environment and the scientific
problem at hand.
Since XRS is a non-resonant IXS process, the exact values
of incident and scattered photon energies (!
1
and !
2
) are less
relevant, and the measured quan tity depends only on the
energy transfer ! = !
1
!
2
. Therefore, a certain freedom
exists for the choice of the incident energy. Depending on the
exact setup, most XRS spectrometers use incident photon
energies betw een 5 and 20 keV. The design of non-resonant
IXS spectrometers can be made simpler than those made for
resonant IXS, since !
2
can be kept fixed and only !
1
is
scanned.
The entire spectrometer, shown in Fig. 1, is mounted on a
massive granite support, which also holds the KB mirror
assembly. The spectrometer is based on Johann-type analyzer
crystals. The analyzers are attached to motorized modular
units, three of which operate in the horizontal scattering plane,
and three units in the vertical scattering plane. Each unit
contains 12 spherically bent Si(nn0) crystals of 100 mm
diameter (three rows of four analyzers each) with a bending
radius of 1.0 m, for a total of 72 crystals. Each analyzer goni-
ometer is equipped with three motorized movements (,
and a translation along the incident beam direction, tx). The
analyzers are cut to 80 mm active diameter in one direction in
order to minimize strain due to angular compression (see
following section) and in order to maximize the solid-angle of
collection while keeping a compact form for the analyzer
array. A schematic drawing of a single an alyzer unit is shown
in Fig. 2(a). Fig. 2(b) shows a single analyzer mount with its
two rotational and one translational degrees of freedom. Each
module is equippe d with a lightweight composite carbon fiber
vacuum chamber, which is operated in a rough vacuum of
approximately 1 mbar in order to minim ize absorption and
parasitic scattering from air. At the exit of the vacuum
chambers, a single-chip Maxipix (Ponchut et al., 2011) detector
head imp lementing a Timepix readout chip (Llopart et al.,
2007) is mounted as close as possible to the sample position in
order to allow the spectrometer to operate in a near-back-
scattering geom etry.
beamlines
J. Synchrotron Rad. (2017). 24, 521–530 S. Huotari et al.
X-ray Raman scattering spectrometer at ESRF ID20 523
Figure 1
(a) Schematic view of the entire spectrometer assembly consisting of six
independently movable crystal analyzer chambers, each hosting 12
spherically bent crystal analyzers. Three chambers are movable in the
vertical plane and three in the horizontal plane. All analyzer chambers,
the sample stage and the KB mirror system rest on a common granite
plate. (b) Photograph of the spectrometer as installed in the experimental
hutch.
The naming convention for the six analyzer units are
vertical down (VD), vertical up (VU), horizontal left (HL)
and horizontal right (HR) modules in the forward direction,
and vertical back (VB) and horizontal back (HB) modules for
the backscattering ones. However, for the selection of the
required momentum transfer range, the scattering angles of all
modules can be adjusted individually and a variety of config-
urations is possible; a typical analyzer unit arrangement in a
transmission measurement using, for example, a liquid jet or
liquid flow could have four units (two horizontal, two vertical)
at scattering angles in the forward direction (for example,
centered around 2 =35
) and two units in the reflection
geometry at large scattering angles (one horizontal and one
vertical unit at 2 = 122
). This allows collecting data at low
values of q (for Si660 analyzer reflection, q
min
’ 1.5 A
˚
1
)and
high values of q (for Si660 reflection, q
max
’ 9.5 A
˚
1
) simul-
taneously while maximizing the collected solid angle at both
q values.
3.1. Energy resolution
The main contributions to the overall energy resolution
arise from the incident beam bandwidth, the analyzer crystal
bandpass and the so-called off-Rowland contribution. In the
following, we will briefly describe the origin for the contri-
butions related to the spectrometer.
Most geometrical factors to the energy resolution scale as
cot
B
, where
B
is the Bragg angle of the analyzer reflection,
and it is therefore beneficial to design the instrument such that
B
is as close to 90
as possible. However, the fact that all
analyzer crystals are to be illuminated by the inelastically
scattered radiation from the sample and because of space
constraints for extended sample environments,
B
is neces-
sarily smaller than 90
. In particular, the Bragg angles for the
three rows of analyzers within one analyzer module will differ
slightly. For example, for the most commonly used config-
uration with a free sample-to-detector distance of 140 mm, the
Bragg angle values for the three rows of analyzer crystals
are 87.8
, 88.3
and 88.9
(assuming an arrangement of the
analyzer foci on the detector active area that results in as
similar Bragg angles for all analyzers as possible). For a
distance of 75 mm between the detector active surface and
the sample, the Bragg angles are increased to 88.5
, 88.7
and
88.9
, respectively, resulting in a slightly improved overall
energy resolution.
3.1.1. Analyzer resolution function. The spe ctral broad-
ening due to the elastic deformation of a bent analyzer crystal
is discussed in detail elsewhere (Verbeni et al., 2005;
Honkanen et al., 2014a). The currently used analyzer crystals
are based on 300 mm-thick anodically bonded Si wafers that
are not stress relief cut (so-called bent-dicing). Such stress
relief cuts have been shown to decrease the contribution of so-
called angular compression (Honkanen et al., 2014a,b), which
is currently the main contribution to the crystal analyzer
energy resolution. Complete elimination of this angular
compression would lead to an energ y resolution of the
analyzer crystals as described by the theory of Takagi and
Taupin (Takagi, 1962; Taupin, 1964). Relief cuts to minimize
the contribution to the energy resolution by this angula r
compression can be implemented in the future. Strain-induced
contributions to the energy resol ution can furthermore be
reduced by using apertures in front of the analyzers (i.e.
masks) and thus reducing the illuminated crystal area, i.e.
reducing the solid angle of detection. At present, available
choices for mask apertures are circular masks with diameters
of 40 mm, 60 mm and 80 mm. Calculated contributions to
the energy resolution induced by the strain in the analyzer
crystal are summarized for different analyzer apertures in
Table 1.
3.1.2. Off-Rowland contribution. Since the spectrometer
should operate at analyzer Bragg angles as close to 90
as
possible and, at the same time, it is necessary to reserve a
certain amount of space for the sample environment, the
beamlines
524 S. Huotari et al.
X-ray Raman scattering spectrometer at ESRF ID20 J. Synchrotron Rad. (2017). 24, 521–530
Table 1
Calculated strain-induced analyzer crystal contributions to the resolution
function in eV with masks with varying diameter and at different
reflection orders. These values do not depend strongly on the exact Bragg
angle.
Reflection
Energy
(keV) 40 mm 60 mm 80 mm
No
mask
Si(4, 4, 0) 6.46 0.1 0.3 0.5 0.7
Si(6, 6, 0) 9.69 0.3 0.5 0.7 1.0
Si(8, 8, 0) 12.92 0.8 1.0 1.2 1.5
Si(10, 10, 0) 16.15 1.5 1.6 1.8 2.0
Figure 2
(a) Sketch of an analyzer module hosting 12 analyzer crystals on a 1 m
Rowland circle. (b) Technical drawing of an individual analyzer crystal
mount with its two rotational and one translational degrees of freedom.
detector is placed inside the Rowland circle by a distance of
2z. In order to maintain the foci of the analyzer crystals on the
detector, the analyzers are mo ved away from the sample
correspondingly by a distance z. This causes the Bragg angles
to vary across the analyzer and thus generates a dispersion
given approximately by
E
E
¼
zD
R
2
cot
B
: ð4Þ
Here, D is the size of the analyzer crystal in the dispersive
direction (80 mm in the current case). This dispersion is called
the off-Rowland contribution to the resolution function and it
may become an important limiting factor in the total resolving
power. To minimize the off-Rowland contribution, the relative
offset z=R and the individual analyzer opening D=R should be
minimized and
B
kept as close to 90
as possible. However,
the simultaneous requirements of a sufficient space for sample
environments and a high energy resolution may become
mutually excluding and certain compromises are necessary.
At present, there are two choices of sample-to-detecto r
distances to adopt for different sample environments and
requirements for energy resolution. The different choices are
realised by two different detector housings (see next section)
leading to sample-to-detector distances of 2z = 75 mm and
2z = 140 mm. The off-Rowland contributions of these
different choices are summarized in Table 2.
3.1.3. Energy resolution of the incident X-rays.The
monochromator configurations avai lable are a cryogenically
cooled Si(111) pre-monochromator only or a combination
of this pre-monochromator with a varie ty of post-mono-
chromators that were described earlier in this article. With
the utilized spherically bent analyzers crystals, the Si(311)
channel-cut monochromator matches the spectrometer reso-
lution best so it is the most commonly used option for a post-
monochromator. Contributions to the overall energy resolu-
tion due to the incident bandwidth are summarized for the
two most useful monochromator configurations [Si(111) pre-
monochromator, and Si(111) pre-monochromator augmented
by a Si(311) channel-cut monochromator] in Table 3.
3.1.4. Overall energy resolution. Tables 4 and 5 report the
overall measured energy resolution based on the full width at
half-maximum (FWHM) of quasi-elastic lines measured off a
thin polymer foil for the most commonly used monochromator
and spectrometer settings.
To further increase the overall resolving power of the
spectrometer while conserving the large solid angle, one can
take advantage of the fact that the dispersion across a
spherically bent crystal analyzer can be compensated for using
an off-focus geometry as was shown recently by Honkanen et
al. (2014b). In comparison with Honkanen et al., however, the
off-focus condition with the XRS spectrometer described here
is achieved by displacing the analyzer crystals along the scat-
tered beam direction (as opposed to displacing the area
detector). Obviously, the off-focus geometry comes along with
a loss of spatial resolution of the spectrometer and is thus not
suitable for samples contained in highly complex sample
beamlines
J. Synchrotron Rad. (2017). 24, 521–530 S. Huotari et al.
X-ray Raman scattering spectrometer at ESRF ID20 525
Table 2
Calculated off-Rowland contribution to the resolution function in eV
with masks with varying diameter and at different reflection orders. These
are values for the middle row of analyz ers (there is in addition one row of
analyzers with slightly higher values and one row with slightly lower
ones).
Sample
distance Reflection 40 mm 60 mm 80 mm
No
mask
Si(4, 4, 0) 0.11 0.17 0.23 0.28
75 mm Si(6, 6, 0) 0.18 0.27 0.37 0.46
Si(8, 8, 0) 0.23 0.34 0.45 0.56
Si(10, 10, 0) 0.28 0.42 0.56 0.70
Sample
distance Reflection 40 mm 60 mm 80 mm
No
mask
Si(4, 4, 0) 0.22 0.33 0.44 0.55
140 mm Si(6, 6, 0) 0.36 0.54 0.76 0.90
Si(8, 8, 0) 0.44 0.66 0.88 1.10
Si(10,10, 0) 0.55 0.83 1.10 1.38
Table 4
Measured FWHM of the spectrometer resolution functions at different
reflection orders, monochromator settings and sample–detector distances.
All results were obtained using 60 mm masks. The reported values are in
eV. The calculated (Calc.) values are the root mean squares of the values
reported in Tables 1, 2 and 3.
Reflection
Energy
(keV)
Si(111) +
Si(311) Calc. Si(111) Calc.
2z =75mm
Si(4, 4, 0) 6.46 0.32 0.01 0.42 0.89 0.01 0.79
Si(6, 6, 0) 9.69 0.65 0.02 0.69 1.39 0.02 1.21
Si(8, 8, 0) 12.92 1.15 0.02 1.18 2.03 0.01 1.77
Si(10, 10, 0) 16.15 1.72 0.08 1.78 2.62 0.02 2.43
2z = 140 mm
Si(4, 4, 0) 6.46 0.42 0.02 0.51 0.90 0.02 0.84
Si(6, 6, 0) 9.69 0.73 0.03 0.84 1.49 0.05 1.30
Si(8, 8, 0) 12.92 1.24 0.03 1.31 2.08 0.01 1.86
Si(10, 10, 0) 16.15 1.77 0.10 1.92 2.70 0.04 2.53
Table 3
Calculated incident bandwidth contribution to the overall resolution
function with different monochromator ensembles and different reflec-
tion orders (Shvydko, 2004). Values are in eV.
Energy
Si(111)
monochromator
Si(311)
channel-cut
Si(4, 4, 0) 0.71 0.25
Si(6, 6, 0) 1.07 0.40
Si(8, 8, 0) 1.42 0.54
Si(10, 10, 0) 1.78 0.67
Table 5
Measured FWHM of the spectrometer resolution function in eV at the
Si(660) analyzer reflection order for different mask sizes, using a Si(111)
pre-monochromator and a Si(311) channel-cut.
Mask size 2z =75mm 2z = 140 mm
40 mm 0.50 0.01 0.53 0.01
60 mm 0.65 0.02 0.73 0.03
80 mm 0.86 0.03 1.03 0.06