research papers

572 doi:10.1107/S0021889809006803 J. Appl. Cryst. (2009). 42, 572–579

Journal of

AppliedCrystallography

ISSN 0021-8898

Received 19 March 2009

Accepted 19 May 2009

# 2009 International Union of Crystallography

Printed in Singapore – all rights reserved

An efficient X-ray spectrometer based on thinmosaic crystal films and its application in variousfields of X-ray spectroscopy

Herbert Legall,a* Holger Stiel,a Matthias Schnurer,a Marcel Pagels,a Birgit

Kanngießer,b Matthias Muller,c Burkhard Beckhoff,c Inna Grigorieva,d Alexander

Antonov,d Vladimir Arkadieve and Aniouar Bjeoumikhovf

aMax-Born-Institut fur Nichtlineare Optik und Kurzzeitspektroskopie, Max-Born-Strasse 2A, 12489

Berlin, Germany, bTechnical University Berlin, Institut fur Optik und Atomare Physik, Harden-

bergstrasse 36, 10623 Berlin, Germany, cPhysikalisch-Technische Bundesanstalt, Abbestrasse 2-12,

10587 Berlin, Germany, dOptigraph GmbH, Rudower Chaussee 29/31, 12489 Berlin, Germany,eIAP – Institut fur Angewandte Photonik eV, Rudower Chaussee 29/31, 12489 Berlin, Germany,

and fIfG Institute for Scientific Instruments GmbH, Rudower Chaussee 29/31, 12489 Berlin,

Germany. Correspondence e-mail: legall@mbi-berlin.de

X-ray optics with high energy resolution and collection efficiency are required in

X-ray spectroscopy for investigations of chemistry and coordination. This is

particularly the case if the X-ray source emits a rather weak signal into a large

solid angle. In the present work the performance of a spectrometer based on

thin mosaic crystals was investigated for different spectroscopic methods using

various X-ray sources. It was found that, even with low-power X-ray sources,

advanced high-resolution X-ray spectroscopy can be performed.

1. Introduction

The applicability of X-ray spectroscopic methods for the

investigation of chemical states and coordinations is in some

cases limited by the collection efficiency and the energy

resolution of the X-ray optics. A typical example is chemically

sensitive X-ray emission spectroscopy (Glatzel & Bergmann,

2005), which needs both high energy resolution and high

collection efficiency. In other cases the feasibility of an

experiment can be limited by the low flux supplied by the

X-ray source that is used for the experiment. Prominent

examples include femtosecond-pulse-emitting laser-based

X-ray sources, e.g. laser plasma sources (Thoss et al., 2003;

Zhavoronkov et al., 2005). While in the first case the signal is

emitted in a large solid angle, in the second case the source

used for the experiment radiates isotropically in all directions.

In both cases the solid angle of collection of the X-ray optics

should be as large as possible. Single crystals, which are

commonly used in experiments in which high energy resolu-

tion at photon energies above 1 keV is required, exhibit only

small acceptance angles and the integrated reflectivity is

therefore low. Bending these crystals (Johannson geometry)

increases the collection angle, but single crystals can be bent

only weakly, and large-area crystals are in general not avail-

able. Consequently a multi-crystal analyzer is used to ensure a

large solid angle of detection (Glatzel & Bergmann, 2005).

In the present work another strategy is described, which

makes use of the efficient collecting properties of mosaic

crystals. An ideal mosaic crystal consists of a large number of

small crystallites that are randomly tilted over a small angular

range perpendicular to the surface normal (cf. Fig. 1).

Consequently, these small crystallites can cover a large solid

angle in the scattering plane, even if the surface of the crystal

is flat. The FWHM of the angular distribution of the crystal-

lites is called the mosaic spread. For very thin pyrolytic

graphite crystals, as used in this work, a Lorentzian distribu-

tion was found (Legall, Stiel, Antonov et al., 2006). Because

the mosaic spread of these crystals is much larger than the

single-crystal rocking curve width, the integrated reflectivity

of mosaic crystals can be some orders of magnitude larger than

that for single crystals (Beckhoff et al., 1996; Legall, Stiel,

Arkadiev et al., 2006). If the intrinsic reflection width (Darwin

width for perfect single crystals) of the individual crystallites is

small, high energy resolution can be observed even with

mosaic crystals (Legall, Stiel, Antonov et al., 2006; Legall,

Stiel, Arkadiev et al., 2006). Furthermore, these crystals can be

made in a strongly bent three-dimensional geometry

(Grigorieva & Antonov, 2003), and bending effects on the

reflection properties of the small crystallites are small.

Consequently, nearly no loss in energy resolution is observed

(Legall, Stiel, Antonov et al., 2006). Finally, virtually no area

limitations exist, because the area of mounted foil crystals is

only limited by the size of the substrate.

Highly oriented pyrolytic graphite (HOPG) is a well known

representative of mosaic crystals and can be used in an energy

range from 3 keV up to several tens of kiloelectron volts

(Beckhoff et al., 1996). In addition, it was shown recently that

these crystals can be produced with low intrinsic reflection

broadening, which is a pre-condition for high spectral reso-

lution. This new type of thin pyrolytic graphite, called HAPG

(highly annealed pyrolytic graphite), developed by Optigraph

GmbH for these applications and originally described by

Legall, Stiel, Antonov et al., 2006), was used in this work.

Experiments covering various spectroscopic methods have

been performed with a variety of X-ray sources to demon-

strate the applicability of a HAPG spectrometer if a high

energy resolution and a high collection efficiency are required.

2. Diffraction properties of the HAPG mosaic crystalfilms

The mosaic spread influences the diffraction properties of a

mosaic crystal in several ways: it determines the integrated

reflectivity (acceptance angle for monochromatic radiation)

and enables broad band reflection for a given angle of inci-

dence. Furthermore, it affects the penetration depth (as a

result of extinction effects), which contributes to energy

resolution (Fig. 1). In addition, in a symmetric geometry,

mosaic focusing (parafocusing) takes place in the scattering

plane. This enhances the intensity in the image plane as well as

the spectral resolution.

The energy resolution is determined by various factors,

which give rise to a smearing in the image plane. Some of these

factors are independent of the chosen distances, such as, for

example, the intrinsic reflection broadening of the crystallites

and the so called ‘flat focusing error’ (Ice & Sparks, 1990). The

latter is a consequence of crystallites not lying on the Rowland

circle in a flat crystal geometry. The contributions of other

factors change with distance. If a constant spatial smearing �s

is induced in the image plane – for example, as a result of

penetration effects into the depth of the mosaic crystal or

simply because of the imaged source size – the contribution of

these factors to the energy resolution can usually be neglected

at larger distances. The reason for this is an increasing spectral

dispersion in the image plane with distance. In Fig. 2 the

influence of the different contributions to an energetic

smearing as a function of distance is displayed. The spectral

resolution of the crystal can be obtained by convolution of

these contributions. The formulae used for these calculations

are given, for example, by Ice & Sparks (1990).

As discussed previously, the mosaic spread affects the beam

penetration. The penetration depth is lower for small mosaic

spread and can limit beam penetration (the screening effect)

more than absorption. In addition, low mosaic spread strongly

reduces the flat focusing error. To summarize, if the source size

and the mosaic spread are small the spectral resolution of a

mosaic crystal can be improved by increasing the distances F

between the source and the crystal and the detector plane,

respectively. The spectral resolution in this case is, similar to

the case for single crystals, limited by the intrinsic reflection

broadening of the small crystallites. The intrinsic reflection

broadening is a consequence of particle size and strain and

gives therefore an upper limit for the spectral resolution. This

upper limit for the energy resolution of the mosaic crystal is

given by the dispersion relation, which can be written as

E=�E ¼ tan �=��intr; ð1Þ

where � is the Bragg angle and ��intr is the averaged intrinsic

width of the Bragg reflection.

The integrated reflectivity is proportional to the number of

crystallites participating in the reflection process. If the

penetration depth is large (e.g. because of large mosaic

spread) or multiple reflections occur in the crystal, losses due

to scattering and absorption must be considered. Both the

atomic scattering and the absorption depend on photon

energy. Consequently, the integrated reflectivity depends on

the photon energy too, as well as the energy resolution

(Beckhoff et al., 1996; Freund et al., 1996; Alianelli et al., 2001).

It is very important to note that, in contrast to the energy

resolution, the effective integrated reflectivity is independent

of the distance between crystal, source and detector (beyond a

certain minimum distance from which the incident angular

emission characteristic can be considered to be constant). If

the crystal size is adapted to the distance, or in other words, if

the same solid angle is collected at all distances, the same

intensity in the image plane is observed. The latter means that

the flux that can be collected, and hence the intensity in the

research papers

J. Appl. Cryst. (2009). 42, 572–579 Herbert Legall et al. � An efficient X-ray spectrometer 573

Figure 2Different contributions to a smearing in the image plane and itsdependence on distance calculated for 8 keV. The formulae forcalculation are given, for example, by Ice & Sparks (1990).

Figure 1Illustration of mosaic focusing which takes place if the distances Fbetween the crystal and the source and the image plane are equal. Alsoshown is the Rowland circle and the focusing errors arising fromreflection from deeper lying crystallites in the film.

image plane, is an intrinsic property of the crystal and does not

change if the energy resolution is maximized by optimizing the

spectrometer geometry. A trade off between energy resolution

and collected flux only arises if the thickness and/or the mosaic

spread of the crystal is varied.

3. Spectrometer setup

The performance of a wavelength dispersive spectrometer

depends on the spectrometer geometry, the crystal and the

detector. Bending the crystal in a cylindrical geometry and

placing it between source and detector in the image plane

results in a so-called ‘von Hamos spectrometer’ setup (von

Hamos, 1932). In this geometry the incident radiation can be

collected in a large solid angle (cf. Fig. 3), while in the

dispersion plane the crystal behaves like a flat crystal (Rm =

1).

If the radius (Rs = F sin �) of the crystal cylinder is chosen

properly the radiation collected from the source will be

focused in the image plane geometrically. Owing to the mosaic

spread, the image of the source shows an additional spread

which takes place in the direction out of the dispersion plane

(cf. Fig. 3). For lower mosaic spread this smearing out of the

dispersion plane is lower and better focusing can be obtained

(Sanchez del Rio et al., 1998).

The spectrometer in this work was designed to function in

polychromatic mode and no entrance and/or exit slits were

used. The entrance slit of the spectrometer is the source size

itself. In this mode the whole spectrum can be collected

simultaneously, if an appropriate detector (e.g. a CCD

camera) is used. The advantage of this mode is that it enables

single-shot spectroscopy if a pulsed X-ray source is used. The

disadvantage of a slitless spectrometer is that its energy

resolution is critically influenced by the source size. By placing

an area detector on the cylinder axis (Shevelko et al., 2002) the

energy range for which geometrical focusing in the detector

plane occurs can be, in principle, extended without limitations

(cf. Fig. 4). However, for high energy resolution at large

distances with the latter configuration, a detector with a large

detection area is required. Additionally, if a cooled area CCD

is used the whole setup must be placed in a vacuum, because it

is difficult to realize low incident angles if a window for

vacuum sealing is used. Therefore, the detector plane was

placed perpendicular to the central ray of the spectrum

reflected by the crystal (cf. Fig. 4), as was also done by Ice &

Sparks (1990). In this geometry for a certain energy E0

parafocusing can take place, while other energies are focused

in front of or behind the detector plane. Consequently, in this

geometry the energy resolution drops for energies not equal to

E0. The extent of this loss in energy resolution depends

strongly on the mosaic spread. For the experiments in this

work we found a loss of less than 10% in energy resolution

over the collected energy range of interest. The energy

dispersion in the image plane can be calculated in this

geometry by the following equation:

�xð�Þ ¼ 2F tanð�0 � �Þ: ð2Þ

Here �xð�Þ is the distance of the image of a variable energy E

with respect to that of E0 in the detector plane. The angles �0

and � are the Bragg angles for the two different energies.

Equation (2) can be use to calculated the energy range that

can be collected simultaneously by using an area detector at a

certain distance F. For example, at a distance of 680 mm at

8.3 keV a spectral range of 1 keV can be observed with a

detector length of 25.4 mm.

The dispersive elements in the HAPG spectrometer were

100 mm-thick HAPG crystals. The crystals were bonded by

Optigraph GmbH to cylindrically polished glass substrates

using a glue-free process. This process avoids distortions that

typically arise from non-uniform glue layers and ensures high

surface quality of the crystal optic. The intrinsic width of

reflection of these crystals at 8 keV was estimated to be 27

arcseconds for the 002 reflection from rocking curve

measurements performed with a 15 mm-thick crystal (Legall,

Stiel, Antonov et al., 2006). This very thin crystal was taken

from the same production batch as those used in the present

work. The mosaic spread of this very thin HAPG crystal was

measured to be 0.04�. For thicker HAPG crystals of 100 mm no

rocking curves have been measured before now, but according

to Grigorieva & Antonov (2003) the mosaic spread of this type

of crystal increases with thickness. The thicker HAPG crystal

was chosen for the experiments in this work because the

integrated reflectivity increases with both the thickness and

the mosaic spread (see above). Furthermore, a larger mosaic

spread increases the acceptance angle of the crystal; this is

favorable if the spectrometer is operated in a polychromatic

mode, because a modulation in reflectivity caused by mosaic

research papers

574 Herbert Legall et al. � An efficient X-ray spectrometer J. Appl. Cryst. (2009). 42, 572–579

Figure 3Sketch of the von Hamos spectrometer.

Figure 4Modified von Hamos geometry.

spread restricts the energy range that can be collected simul-

taneously with a given crystal size.

The integrated reflectivity can be estimated from the

rocking curve measurements, which reveal the reflectivity over

the incidence angle of X-ray radiation. For the 15 mm HAPG

crystal the rocking curve gave an integrated reflectivity of

0.7 mrad at 8 keV in the first reflection order, which is nearly

ten times larger than that reported for Ge(111) crystals (Town,

1997). From measured Cu K� emission spectra a decrease in

intensity in the image plane of a factor of five was found if

working with the second-order reflection of the same crystal.

Comparing the measured Cu K� emission spectra of the 15

and 100 mm crystals gives six to seven times more intensity in

the image plane for the thicker crystal.

To collect a broad spectral range, a 16 bit deep depletion

CCD camera (Roper Scientific model PI-LCX 1300) with a

quantum efficiency of about 50% at 8 keV was placed in the

image plane. The CCD was calibrated with an Fe55 radioactive

source. In this setup, the pixel size of 24 mm determines the

accuracy of the absolute calibration of the energy axis in the

detector plane (see below). A thin (250 mm) Be window in

front of the deep depletion CCD was used for vacuum sealing

of the camera, so that a deep cooling (down to 223 K) of the

CCD was possible.

4. Experiments

The spectrometer concept was tested in various experiments

using three different types of X-ray sources. The experimental

conditions were chosen such that the flux emitted by the X-ray

source is at the limit of the detection efficiency of today’s

commonly used spectrometer setups. The X-ray sources were

a low-power micro-focus X-ray tube with an Ag anode

(iMOXS MFR; IfG GmbH) for EXAFS measurements, the

mySpot beamline at BESSY II for the X-ray fluorescence

(XRF) experiment and an ultrashort laser plasma source

emitting femtosecond X-ray pulses at the Max-Born-Institut.

4.1. X-ray absorption measurements

The micro-focus tube used in the EXAFS experiments

delivers a Bremsstrahlung continuum with a flux of about

109 photons s�1 sr�1 in 0.1%BW (bandwidth) at 8.3 keV when

the tube is operated with a voltage of 40 kV and an anode

current of 800 mA (cf. x4.3). The source size of the X-ray tube

was 50 mm. Measurements were performed for the 002 and 004

reflections. The energy resolution of the HAPG crystal used

for these measurements was determined from Cu K� spectra

at 8 keV to be E/�E = 1800 for the 002 reflection at a distance

of 680 mm (which represents an upper limit for the 100 mm

crystal at this reflection order) and E/�E = 2100 at a distance

of 360 mm for the 004 reflection. At these distances the

spectral position of the Ni K-edge on the CCD is close to the

energy that is best focused by a cylindrical HAPG with a

radius R = 150 mm (cf. Fig. 5). The length of the cylindrical

crystal was 3 cm and the height was 5 cm. The EXAFS

measurements were performed in transmission with a 4 mm Ni

foil as sample placed between the source and the crystal.

Images were collected with and without a sample. The images

of the Bremsstrahlung continuum after transmission through

the Ni foil are shown in Fig. 5 for both reflection orders.

From the measured images, the Ni K-edge EXAFS spec-

trum can be calculated if the transmitted signal behind the

sample is divided by the unfiltered Bremsstrahlung continuum.

As a result of this division, a modulation of the spectral

intensities by the mosaic crystal can be ignored in the

evaluation procedure. If the geometrical acceptance angle of

the mosaic crystal is too small to reach a low noise-to-signal

ratio over the whole spectrum the energy range must be

scanned. In the presented EXAFS experiment two scans were

sufficient. From the spectra, the noise-to-signal ratio after

20 min accumulation time was determined to be 1% for the

002 reflection and 5% for the 004 reflection. After subtracting

the atomic background absorption the EXAFS oscillation can

be separated from the measured spectrum. Calibration of the

energy axis was achieved by using an Ni foil reference spec-

trum (Farrel Lytle database; http://ixs.iit.edu/data/Farrel_

Lytle_data/PROCESSED/n/nifoil.k45). The result of this

procedure is shown in Fig. 6. For the sake of comparison the

reference spectrum is displayed as well.

As can be seen from Fig. 6, with the HAPG spectrometer

setup, EXAFS spectroscopy can be performed with laboratory

low-power X-ray sources. A comparison of the HAPG spec-

research papers

J. Appl. Cryst. (2009). 42, 572–579 Herbert Legall et al. � An efficient X-ray spectrometer 575

Figure 5EXAFS images and cross sections. The images show the Bremsstrahlungcontinuum after transmission through an Ni foil of 4 mm thickness. Theacquisition time for the images was 20 min.

trometer with the spectrometer setup presented by Benesch et

al. (2004) shows the advantage of using thin mosaic crystals as

was done in this work. For EXAFS measurements, Benesch et

al. (2004) used a single Si(111) crystal combined with a

capillary lens for focusing X-rays emitted by a micro-focus

X-ray tube (TruFocus TFX-8050, 15 W) on a liquid sample.

Owing to the use of the capillary lens, the flux collected by the

dispersive element after transmission through the sample is

enhanced. On the other hand, the source size given by the

focal spot of about 100 mm on the sample decreases the energy

resolution significantly. The acquisition time for an EXAFS

spectrum reported by Benesch et al. (2004) was 20 h instead of

20 min as reported in this work, and the energy resolution was

E/�E = 1400 at 7–8 keV. Considering the lower power (15 W)

of the X-ray tube used by Benesch et al. (2004) in comparison

with the X-ray tube used in this work (30 W), there is still a

factor of 30 gain in spectrometer efficiency.

4.2. X-ray emission spectroscopy (XES)

Chemically sensitive X-ray 2p–1s (K�) or 3p–1s (K�) K

fluorescence spectroscopy can be performed if the interaction

between the 2p and 3p wavefunctions and the wavefunction of

the binding orbital is large enough to show ligand-dependent

chemical shifts (Glatzel & Bergmann, 2005). This is the case,

for example, in 3d transition metal complexes. Appropriate

targets of these investigations are then, for example, the K�satellite lines, which mainly arise from transitions out of

orbitals higher than the 3p shells. Such investigations are today

usually performed at synchrotron undulator beamlines

(Glatzel & Bergmann, 2005; Reinhardt et al., 2009), because

the signal flux emitted from these satellite transitions is quite

low. In this section, measurements performed at a synchrotron

wavelength-shifter beamline in combination with the HAPG

spectrometer will be presented.

The samples investigated in X-ray emission measurements

were a Ti foil and a Ti oxide compound. The radiation of the

mySpot beamline at BESSY II was focused with a poly-

capillary lens (throughput 40%) on the sample, resulting in a

focal spot size of 30 mm with a flux of about 1010 photons s�1

at 8.5 keV. This is a much lower flux than the 1012–

1013 photons s�1 delivered by undulator beamlines. However,

this higher flux is not useful in an experiment in which the

damage threshold of the sample is exceeded. In this case it can

be helpful to optimize the collecting X-ray optics instead of

increasing the incident flux on the sample to resolve these

weak XRF signals. Taking the flux of the mySpot beamline at

BESSY II into account for the K�5 emission, a signal inten-

sity of about 6 � 103 photons s�1 sr�1 was calculated and

confirmed in the experiment. The XES measurements in this

work were performed with a 100 mm bent HAPG crystal

(radius 100 mm, length 6 cm and height 5 cm) placed at a

distance of F = 250 mm between the sample and the CCD.

In Fig. 7 the measured Ti and TiO2 K� XRF spectra are

displayed. Different contributions to the K� emission spec-

trum could be identified. The spectra were fitted with

Lorentzian line profiles, except for the radiative Auger KMM

contribution, which was estimated to have a Gaussian line

shape because of the multiplicity of lines. For K�1;3 in the case

of the Ti foil two Lorentzian profiles were needed, whereas in

the case of TiO2 three Lorentzian profiles were used to fit the

spectra. The latter is a consequence of the strong asymmetry

of K�1;3, which can also be explained by a modulation of the

intensity due to the reflection curve of the mosaic crystal. A

research papers

576 Herbert Legall et al. � An efficient X-ray spectrometer J. Appl. Cryst. (2009). 42, 572–579

Figure 6Measured EXAFS oscillations after subtraction of the atomic back-ground. Also shown is the Ni foil reference spectrum (Farrel Lytledatabase) used for calibration of the energy axis.

Figure 7Measured K� fluorescence of a Ti foil and a TiO2 pellet using the HAPGspectrometer. The acquisition time was 7 h for the Ti spectrum and 3 h forthe TiO2 spectrum.

comparison of the spectra measured in this work with spectra

of other authors shows that the line intensity in this work is

tailed off by the reflection curve of the crystal, which was

aligned to maximum reflectivity for the energy of the K�5 line.

Therefore the real intensity of the satellites can only be

recalculated from the spectrum once the reflection curve is

known. However, the inaccuracy of this procedure is relatively

large considering the weakness of the fluorescence lines. An

experimental improvement is necessary at this point. The best

way to eliminate such modulation effects is a calibration of the

spectrometer, either by employing radiation scattered elasti-

cally at a thin sample in the XRF beam geometry or by direct

illumination of the HAPG crystal with monochromatic

radiation. Both approaches require the tunability of

synchrotron radiation and knowledge of the radiant power of

the radiation. The incident radiation power can be determined

by photodiodes calibrated absolutely (Beckhoff, 2008;

Beckhoff et al., 2006).

However, from the pixel size and the distances between the

K� and K� lines the accuracy of the measurement can be

calculated to be �0.2 eV with 0.4 eV per pixel. As mentioned

above the accuracy of this experiment can be enhanced

without problems by increasing the distances. No loss in

intensity on the CCD is expected if the crystal size is enhanced

to collect the same solid angle. In the same way the energy

resolution can be significantly improved. In Table 1 the

energies of the second-order contributions in the K�1;3 spectra

are tabulated and compared with the results of other authors.

4.3. Plasma emission spectroscopy

Ultrafast laser-based plasma sources deliver X-ray pulses

with durations of some hundreds of femtoseconds and enable

time-resolved investigations with X-ray techniques such as

diffraction (Bargheer et al., 2004) and absorption spectroscopy

(Lee et al., 2005). Up to now only a few papers have been

published dealing with time-resolved EXAFS experiments in

the picosecond (Bressler et al., 2002; Saes et al., 2003, 2004;

Bressler & Cherugi, 2004; Audebert et al., 2005; Chen et al.,

2002) or sub-picosecond range (Lee et al., 2004; Benesch et al.,

2004). Time-resolved spectroscopy using femtosecond laser

plasma sources is still a challenge because the photon flux

delivered by laser plasma sources and other available ultra-

short X-ray sources is much lower than that obtained using

synchrotron radiation, albeit providing a lower pulse duration

limit of about 20 ps. Typical acquisition times reported for an

EXAFS spectrum with femtosecond laser plasma sources are

currently 10 h (Benesch et al., 2004). However, the ultrashort

pulse duration warrants further effort to reduce the acquisi-

tion time by enhancing the available photon flux in the

experiment. This can be achieved either by laser system

upgrades (repetition rate and in some cases laser pulse

energy) or by optimizing the spectrometer concept. The latter

we prove in the present work by applying the spectrometer for

plasma emission spectroscopy.

The laser plasma source used for the experiments consists of

a tape target system in a vacuum chamber and an off-axis

parabolic mirror as focusing element. Thin metal tapes (Cu, W,

thickness 50 mm) were illuminated using the high-field MBI

Ti:Sa laser facility (815 nm center wavelength, 40 fs pulse

duration, > 1 J pulse energy after compression, 10 Hz repe-

tition rate and contrast ratio down to 10�7 and 10�8 at peak

intensity). In the experiment, a maximum energy of 450 mJ

was applied. It was found that 35% of the energy incident on

the parabolic mirror was focused into a spot of 10 mm

diameter, giving an intensity of 3 � 1018 W cm�2 at the target.

The focused laser beam hits the surface of the target at an

incident angle of 45�. The X-ray emission was measured in

reflection at 90� to the incident laser beam. In Fig. 8 a single-

shot Cu K� emission spectrum of the laser plasma is shown;

this was collected with the HAPG spectrometer after exposing

the Cu target to a 430 mJ laser pulse with a 40 fs pulse dura-

tion. The spectrum corresponds to an averaged X-ray photon

flux of 2 � 1010 photons s�1 sr�1 in Cu K� at a laser intensity

of 3 � 1018 W cm�2. In Fig. 9 the W L� flux and the emitted

Bremsstrahlung in 0.1%BW evaluated from measured spectra

with a W target is depicted as a function of the single-shot

laser pulse energy.

The conversion efficiency for the laser-based generation of

ultrashort X-ray Bremsstrahlung scales with ðI�2Þ1=2, where I is

the laser intensity and � the wavelength (Yu et al., 1999).

Therefore, the conversion efficiency is significantly higher if

using higher laser intensities, as was done in the present work.

Comparing the 6 � 107 photon s�1 sr�1 in 0.1%BW measured

in the present work for a tungsten target at an intensity of

about 3 � 1018 W cm�2 (400 mJ pulse�1) and 10 Hz with the

2.4 � 108 photons s�1 sr�1 measured for a copper target at an

intensity of about 1016 W cm�2 (4.5 mJ pulse�1) and 2 kHz in

about 12%BW (Benesch et al., 2004) gives 30 times higher

photon numbers per second and 40 times higher conversion

research papers

J. Appl. Cryst. (2009). 42, 572–579 Herbert Legall et al. � An efficient X-ray spectrometer 577

Table 1Energies of the second-order contributions in the K� fluorescencespectra relative to the K�1;3 contributions (line energies in eV).

A shift of the K�1;3 lines between mono-elemental Ti and the Ti compoundTiO2 could not be resolved with the present accuracy. A reported value is0.2 eV (Koster & Mendel, 1970). The uncertainties of the tabulated energypositions in this work are calculated from the pixel size and from theuncertainty in data fitting. The deviation of the energy scaling from linearity isaccording to equation (2) less than 0.05 eV for all contributions and wasneglected in the evaluation of the measured spectra.

Lines Author Ti TiO2

K�0 This work �15.7 (13) �15.6 (9)Ursic et al. (2003) �5.62 –Salem et al. (1976) �7.08 –Faessler (1955) – �16

K�00 This work 13.1 (17) 16 (5)Koster & Mendel (1970) – 15.2

K�5 This work 30.5 (4) 31.5 (4)Koster & Mendel (1970) 30.9 31.4Ursic et al. (2003) 29.96 –Ursic et al. (2003) – 30.4

K� L1 This work 52.2 (22) 49 (29)Ursic et al. (2003) 50.4 –Raju et al. (2007) 53.5 –

efficiency for the 10 Hz high-field laser plasma source. Besides

the higher photon flux delivered by the high-field laser system,

the lower repetition rate is desirable in time-resolved pump-

probe spectroscopy because the sample is heated much less by

an exciting UV/VIS pulse after ten photocycles per second

than after the 1000 cycles that occur in the kiloherz mode. On

the other hand, the generation of X-rays using high intensity

increases drastically the amount of background X-ray radia-

tion because of the interaction of the highly energetic elec-

trons with the surrounding medium. This means a careful

shielding of the CCD is a pre-condition for obtaining a high

signal-to-noise ratio.

In Table 2 the photons collected on the CCD using the laser

plasma source are related to the photons collected using the

micro-focus X-ray tube that was used for the EXAFS

experiment in x4.1. The setup was identical in both measure-

ments. For the measurements, the target in the laser plasma

chamber was simply replaced by the tube. A comparison of the

flux collected for both sources allows an estimation of the

acquisition time, which is necessary for measuring an EXAFS

spectrum with the HAPG spectrometer using the laser plasma

source instead of the micro-focus X-ray tube.

As can be seen from Table 2, the laser plasma source can be

expected to require a 10� (800 mA/600 mA) = 13 times longer

acquisition time to obtain an EXAFS spectrum that has the

same accuracy (under low background scattering conditions)

as was obtained with the micro-focus X-ray tube in x4.1. This

acquisition time can be further reduced by a factor of 2–3

using a high-Z target material such as tungsten (cf. Fig. 9),

because the conversion efficiency scales with the atomic

number Z. The latter would result in an acquisition time of

about 2 h for the EXAFS spectrum of an Ni foil measured

with a femtosecond X-ray source. This is a significant reduc-

tion of acquisition time compared with the 10 h reported by

Benesch et al. (2004). Furthermore, comparison of the counts

on the CCD also indicates that a significant enhancement in

accuracy can be expected by using the HAPG spectrometer.

Comparing the 50 photons s�1 in 1 keV (or 0.4 photons s�1 in

0.1%BW) on the CCD reported by Benesch et al. (2004) with

the 2.3 � 104 photons s�1 in 0.1%BW at the maximum of the

reflected Bremsstrahlung continuum for the 002 reflection

using the HAPG spectrometer (x4.1), the number of photons

on the CCD can be expected to increase by a factor of (2.3 �

104 photons s�1)/(0.4 photons s�1)/13 = 4.4 � 103 with the

HAPG spectrometer in combination with the high-field laser

plasma source using Cu as target material.

5. Conclusions

A new spectrometer setup based on HAPG mosaic crystals

was described and applied in various fields of X-ray spectro-

scopy. It was shown that, as a result of the high integrated

reflectivity of the mosaic crystals used, the recording time for a

high-resolution EXAFS spectrum measured with a micro-

focus X-ray tube can be significantly reduced and the noise-to-

signal ratios improved in comparison with a single-crystal

spectrometer reported in the literature. Furthermore, it was

demonstrated that high-resolution X-ray emission spectro-

scopy of weak fluorescence lines can be performed using the

HAPG spectrometer even if the flux provided by the exciting

X-ray beam is in the range of that delivered by table top

sources, such as, for example, X-ray tubes. Towards time-

research papers

578 Herbert Legall et al. � An efficient X-ray spectrometer J. Appl. Cryst. (2009). 42, 572–579

Figure 9X-ray flux emitted by the femtosecond laser plasma source as a functionof the laser pulse energy. (Nph denotes the number of photons.)

Table 2Flux measurements.

Tabulated are the collected photons s�1 in 0.1%BW at 8.3 keV on the CCD tocompare the emitted flux of the micro-focus X-ray tube with a Cu anode andthe laser plasma source with a Cu target. Radiation was collected (bent HOPGcrystal, radius 50 mm, 150 mm thickness, 360 mm distance) over an angle of � =4 � 10�3 sr. The anode current of the tube with Cu as anode material was600 mA and the voltage was 40 kV, resulting in a nearly identical amount ofBremsstrahlung as obtained at the Ag anode at the same voltage and sameanode current. The laser intensity was 3� 1018 W cm�2 and the pulse durationwas 40 fs at 800 nm.

Photons s�1 in 002 reflection 004 reflection

Cu K� (LPP) 3 � 105 6 � 104

Cu K� (tube) 2 � 106 4 � 105

Continuum 0.1%BW (LPP) 1 � 103 2 � 102

Continuum 0.1%BW (tube) 1 � 104 2 � 103

Figure 8Single-shot Cu K� spectrum. The exposure time was 40 fs.

resolved X-ray spectroscopy with femtosecond time resolu-

tion, the spectrometer was applied for the characterization of

an ultrafast laser plasma X-ray source. It was shown that an

application of the HAPG spectrometer in the field of time-

resolved high-resolution femtosecond EXAFS spectroscopy

with ultrafast laser plasma sources is promising. If the crystal

size is adapted to the distance and the same solid angle is

collected by the crystal, higher energy resolution can be

realized in all application experiments by increasing the

distances between the source/sample, the crystal and the

detector without changing the spectrometer efficiency. Addi-

tionally, by increasing the crystal size the spectrometer effi-

ciency can be further enhanced. An improvement of the

HAPG spectrometer concept may be possible by using a

collecting optic to focus the radiation emitted by the source on

the sample. The applicability of different types of optics were

investigated by Bargheer et al. (2005) with respect to focusing

properties, the collecting efficiency and, something that is very

important for time-resolved studies, the conservation of the

femtosecond time structure of the X-ray pulses after reflection

by the optics. In summary, there is still great potential for

improvement of the presented HAPG spectrometer, which

may enable new developments in X-ray spectroscopy in the

near future.

This work was supported by the German national program

of supporting development, innovation and technology

(ProFIT Programm zur Forderung von Forschung, Innova-

tionen und Technologien, Land Berlin) and by EFRE Euro-

paischer Fonds fur regionale Entwicklung (grant No.

10126367).

References

Alianelli, L., Sanchez del Rio, M. & Felici, R. (2001). Proc. SPIE,4509, 135–144.

Audebert, P., Renaudin, P., Bastiani-Ceccotti, S., Geindre, J. P.,Chenais-Popovics, C., Tzortzakis, S., Nagels-Silvert, V., Shepherd,R., Matsushima, I., Gary, S., Girard, F., Peyrusse, O. & Gauthier,J. C. (2005). Phys. Rev. Lett. 94, 1–4.

Bargheer, M., Zhavoronkov, N., Bruch, R., Legall, H., Stiel, H.,Woerner, M. & Elsaesser, T. (2005). Appl. Phys. B, 80, 715–719.

Bargheer, M., Zhavoronkov, N., Gritsai, Y., Woo, J. C., Kim, D. S.,Woerner, M. & Elsaesser, T. (2004). Science, 306, 1771–1773.

Beckhoff, B. (2008). J. Anal. At. Spectrom. 23, 845–853.Beckhoff, B., Kanngießer, B., Langhoff, N., Wedell, R. & Wolf, H.

(2006). Editors. Springer Handbook of Practical X-ray Fluores-cence Analysis, 1st ed. Berlin: Springer.

Beckhoff, B., Kanngießer, B. & Malzer, W. (1996). Proc. SPIE, 2859,190–199.

Benesch, F., Lee, T. W., Jiang, Y. & Rose-Petruck, C. G. (2004). Opt.Lett. 29, 1028–1030.

Bressler, C., Saes, M., Cherugi, M., Grolimund, D., Abela, R. &Pattison, P. (2002). J. Chem. Phys. 116, 2955–2966.

Bressler, C. M. & Cherugi, M. (2004). Chem. Rev. 104, 1781–1812.Chen, L. X., Liu, T. M., Thurnauer, C., Csencsits, R. & Rajh, T. (2002).

J. Phys. Chem. B, 106, 8539–8546.Faessler, A. (1955). Roentgenspektrum und Bindungszustand, in

Landolt–Boerstein: Zahlenwerte und Funktionen, 6th ed., Vol. 1,Part 4. Berlin: Springer.

Freund, A. K., Munkholm, A. & Brennan, S. (1996). Proc. SPIE,2856, 68–79.

Glatzel, P. & Bergmann, U. (2005). Coord. Chem. Rev. 249, 65–95.Grigorieva, I. G. & Antonov, A. A. (2003). X-ray Spectrom. 32, 64–68.Hamos, L. von (1932). Naturwissenschaften, 20, 705–706.Ice, G. E. & Sparks, C. J. (1990). Nucl. Instrum. Methods Phys. Res.

Sect. A, 291, 110–116.Koster, A. S. & Mendel, H. (1970). J. Phys. Chem. Solids, 31, 2511–

2522.Lee, T., Benesch, F., Jiang, Y. & Rose-Petruck, C. G. (2004). Chem.

Phys. 299, 233–245.Lee, T., Jiang, Y., Rose-Petruck, C. G. & Benesch, F. (2005). J. Chem.

Phys. 122, 1–8.Legall, H., Stiel, H., Antonov, A., Grigorieva, I., Arkadiev, V.,

Bjeoumikhov, A. A. & Erko, A. (2006). Proc. FEL, BessyFRAAU04, 798–801.

Legall, H., Stiel, H., Arkadiev, V., Bjeoumikhov, A. A. & Erko, A.(2006). Opt. Express, 14, 4570–4576.

Raju, S. S., Seetharami Reddy, B., Murti, M. V. R. & Mombasawala, L.(2007). X-ray Spectrom. 36, 35–41.

Reinhardt, F., Beckhoff, B., Eba, H., Kanngiesser, B., Kolbe, M.,Mizusawa, M., Muller, M., Pollakowski, B., Sakurai, K. & Ulm, G.(2009). Anal. Chem. 81, 1770–1776.

Saes, M., Bressler, C., Abela, R., Grolimund, D., Johnson, S. L.,Heimann, P. A. & Chergui, M. (2003). Phys. Rev. Lett. 90, 1–3.

Saes, M., van Mourik, F., Gawelda, W., Kaiser, M., Chergui, M.,Bressler, C., Grolimund, D., Abela, R., Glover, T. E., Heimann,P. A., Schoenlein, R. W., Johnson, S. L., Lindenberg, A. M. &Falcone, R. W. (2004). Rev. Sci. Instrum. 75, 24–30.

Salem, S. I., Hockney, G. M. & Lee, P. L. (1976). Phys. Rev. A, 13, 330–334.

Sanchez del Rio, M., Gambaccini, M., Pareschi, G., Taibi, A.,Tuffanelli, A. & Freund, A. (1998). Proc. SPIE, 3448, 246–255.

Shevelko, A. P., Kasyanov, Yu. S. & Yakushev, O. F. (2002). Rev. Sci.Instrum. 73, 3458–3463.

Thoss, A., Richardson, M., Korn, G., Faubel, M., Stiel, H., Vogt, U. &Elsaesser, T. (2003). J. Opt. Soc. Am. B, 20, 224–228.

Town, R. (1997). LLE Rev. 70, 82.Ursic, M., Kavcic, M. & Budnar, M. (2003). Nucl. Instrum. Methods

Phys. Res. Sect. B, 211, 7–14.Yu, J., Jiang, Z. & Kieffer, J. C. (1999). Phys. Plasmas, 6, 1318–1322.Zhavoronkov, N., Gritsai, Y., Bargheer, M., Woerner, M., Elsaesser,

T., Zamponi, F., Uschmann, I. & Forster, E. (2005). Opt. Lett. 30,1737–1739.

research papers

J. Appl. Cryst. (2009). 42, 572–579 Herbert Legall et al. � An efficient X-ray spectrometer 579