X-ray Spectrometers using Polycapillary optics

Polycapillary, X-ray Spectroscopy, X-ray spectrometers, XAS


Polycapillary optics provide a promising approach for coupling highly-divergent x-ray emission or inelastic scattering to high-resolution crystal analyzers. We present recent results looking at the application of polycapillary collimators to emission spectrometers. The first application uses a collimating optic and a flat crystal to provide a tunable x-ray fluorescence detector. At high-flux synchrotron radiation sources there is sufficient flux (~1013 ph/sec) to allow application of X-ray Absorption Spectroscopy (XAS) to ppb concentrations if the fluorescence signal can be isolated from an intense background. The polycapillary based analyzer easily achieves the >106 background reduction needed for such measurements. It has the additional advantage of being confocal, only collecting the signal from a small volume at the optic focus, effectively eliminating background from sample substrates, windows, or air scattering. Second, the same type of analyzer can be used for higher-resolution emission spectroscopy if operated close to 90° Bragg angle, and we report results of the commissioning of a user-available instrument suitable for few-eV resolution emission spectroscopy, including the demonstration of roughly order-of-magnitude improved measurement times compared to use of a traditional, single spherically-bent crystal analyzer. As part of this effort, we have developed a process for enhancing the integral reflectivity of Si analyzer crystals through plastic deformation at high temperatures.


Polycapillary x-ray optics are based on bundles of narrow capillaries generally made of glass. X-rays can be channeled through the capillaries by multiple reflections at extreme glancing angles. The capillaries bundles can be tapered or gently bent to guide and focus the x-rays. In this way a full x-ray lens can be produced that accepts x-rays from a point source with a large acceptance angle (typically 10-20°) and focuses them to a focal spot. A half lens version can be used to collimate the beam from a point source or focus a collimated beam. The main limitation of such optics is the critical angle of the glass. This is the angle that if exceeded the x-ray reflectivity drops sharply. For glass it is typically a few milliradians and varies inversely with the x-ray energy. The output angle spread of the x-rays from each individual capillary is approximately twice the critical angle of the glass. This limits the collimating or focusing potential of the bundle. However, since the same limitation applies to incoming x-rays emanating from the focal spot, polycapillaries are excellent for confocal applications. As shown in Figure 1,x-rays from outside the focal spot are strongly attenuated since they exceed the critical angle of reflection.

Fig. 1 Basic principle of polycapillary optics in confocal applications. For each individual channel, x-rays can only be accepted and guided along the capillary if their reflection angle is less than the critical angle of the capillary. The individual channels are arranged as shown with their acceptance angles intercepting a common focus. X-rays incident on the capillaries from parts of the x-ray beam outside of the focus will exceed the critical angle θe and be rejected

Polycapillary optics have been widely used with x-ray tube sources where their large acceptance angle is a big advantage. They have seen less use at synchrotron radiation sources with only a few applications as primary focusing optics. The applications at synchrotron sources have mainly concentrated in two areas: confocal imaging applications, and collection optics for x-ray spectrometers. In this paper we will be concentrating on the application of polycapillaries to x-ray spectrometers, but will discuss one aspect of confocal imaging that is important for x-ray spectroscopy. This is the reduction of background signals in x-ray microprobe applications.


In x-ray microprobe applications samples are often thin sections mounted on low-Z substrates. In this case the substrate as well as the surrounding atmosphere and sample cell windows can generate a strong background signal. It can overwhelm a weak signal of interest. For thicker samples such as biological tissues it is often desired to collect the signal from some structure buried within the sample. In both of these cases confocal detection as illustrated in Figure I can be a big advantage.

To illustrate this, Figure 2 shows some data obtained from a ‘thick’ section of a pigeon inner ear obtained at beamline 20-ID-B at the Advanced Photon Source.

Fig. 2 Comparison of data taken at the Fe edge with and without a polycapillary optic providing background suppression using confocal imaging. The polycapillary case used a single vortex detector while a 4 element vortex detector was used without the polycapillary optic

In this case the experimenters were interested in a specific structure inside the inner ear that is difficult to capture in a thin section. The tack ground for the standard 4-element Si drift detector was excessive, severely degrading the data quality. This background comes from the overlying tissue and the substrate. Since this is data for the Fe edge, and much of the beamline equipment contains Fe, there is also a small contamination signal from metallic Fe. This signal is difficult to completely eliminate from the wide open 4 element detector. For the single element case the entrance optic was a polycapillary half-lens collimator that had a focal spot size of about 30 μm. The background was almost completely eliminated greatly improving the data quality. Note that in both cases the intrinsic energy resolution of the vortex detector (~180 eV) was used to separate the Fe signal from the scattering background. Without the polycapillary the background was so strong that some of the background leaked into the Fe channel. The background also made it necessary to increase the detector-to-sample distance to avoid exceeding the maximum detector count rate. This also reduced the Fe signal that could be detected. Background scans indicated that with the polycapillary the interfering Fe metal signal was also completely eliminated.

Based on these results the 4-element detector is being equipped with a conical polycapillary collimator. This provides background suppression and has a less than 100 um focal spot for coarser confocal applications.


3.1 Application to dilute samples

Modern synchrotron radiation beamlines can provide extremely high fluxes, enabling detection of very dilute components. One of our current goals is the ability to detect parts per billion levels of trace elements and measure the x-ray absorption signals. This is much more stringent than simply quantifying the presence an element. A typical X-ray Absorption Near Edge spectrum (XANES) will have 50-100data points each requiring statistical accuracy of better than 1% (e.g. more than 10 signal counts per point). Such capabilities are of particular importance in environmental or biological systems, where environmental contaminants or biologically important trace elements are present at very high dilution.

A specific example is useful to estimate the required charateristics of a detector. A high flux beamline should provide focused monochromatic photons with flux close to 10 Hz. A sample containing 100 ppb of a heavy element in a lighter matrix (i.e., Fe in a silicate matrix) will have approximately 107 photons/sec absorbed by the element of interest. Assuming 30% fluorescence efficiency and a 1% detection efficiency, 3×104 fluorescence photons/sec can be detected. This is more than enough for rapid XANES studies. However, these photons will be accompanied by a large background of scattering and possibly fluorescence from other components in the sample. Such backgrounds can easily be greater than 1% of the incident flux or 109 photons/s into the collection solid angle of the 1% efficient detector. This can overwhelm standard solid-state detectors that are limited to about 106 Hz total count rate. Even for detectors with high-multiplicity capable of high-total count rates, the finite peak-to-background ratio(typically less than 104) of solid-state detectors makes the detection of very weak peaks impractical. A better approach is to reject the background using a crystal spectrometer.

It has been suggested previously that collimating polycalpillary optics can be combined with a simple flat crystal spectrometer to provide an efficient and simple crystal spectrometer. Figure 3 shows the basic idea that has been commonly applied to laboratory spectrometers, but only sparingly at synchrotron sources.

Fig. 3 Basic geometry of a collimating polycapillary combined with a flat crystal spectrometer

A typical polycapillary collimator operating at 10 keV can collect a cone up to about 20° wide with 20-30% transmission efficiency. This somewhat short of the 1% collection efficiency in the example above, but it should be possible to employ several analyzers simultaneously. Only the polycapillary needs to be close to the sample. Once the beam is collimated there can be a long flight path to the more massive spectrometer and detector.

Tests of these ideas were carried out using a polycapillary collimating optic with parameters as listed in Table 1.

Table 1. Nominal properties of the polycapillary collimating optic purchased from X-ray Optical Systems.

wdt_ID Attributes Parameter
1 Input focal distance 9.5mm
2 Input capture angle 13.7°
3 Output beam diameter 6mm
4 Input focal spot size (8 keV) 50um
5 Transmission efficiency (8 keV) 30%

As a first test a Ni foil was used as the source on the 20-ID-B microprobe station that has a focused beam of about 5 μm. This provided an incoming flux of about 1.5 x 1012 ph/sec. The output beam was visualized with a Pilatus 100K pixel array detector. The many individual pixels of this detector allow it to handle the high count rates encountered. The results are summarized in Table 2.

Table 2. Summary of count rates from flat crystal spectrometer tests.

wdt_ID Attributes Parameter
1 I0 1.5x10^12
2 Ni foil through Polycapillary 4.6x10^8
3 Ni foil through Polycapillary with Ge (111) analyzer 2.1x10^7
4 Ni foil through Polycapillary with graphite (002) analyzer 1.4x10^8
5 Mineral ref BHVO-2G (110 ppm Ni) with graphite (002) analyzer 1.3x10^4

The Ni fluorescence energy is about 7.5 keV so the properties in Table 1 should apply. The fluorescence signal through the polycapillary is close to that expected from these properties. Thus, the optic was well aligned. Two analyzer crystals were tested. Ge (111) provides high energy resolution, but has reduced intensity since it is not well matched to the output angle spread of the polycapillary. The HOPG graphite analyzer has a mosaic spread (nominally 0.4°) sufficient to capture all of the angles from the polycapillary, providing higher intensity but reduced energy resolution. A comparison of the two cases is shown in Figure 4.

From the energy resolution of the Ge data, the output angle spread of the optic can be estimated. The resolution is given by △E/E=△θcot(θ). In this case △E/E is ~0.007, giving △θ~2 mrad. For HOPG the resolution is dominated by the mosaic spread since it is much larger than the output spread of the optic. Using 0.4° as the mosaic spread gives △E/E of

Fig. 4 Comparison of the energy resolution of Ge(111) and HOPG graphite (002) analyzer crystals. The Ge signal is multiplied by 5 for clarity

2.8%. The measured value is about 4% indicating the actual mosaic spread of the HOPG is somewhat worse than the specified value.

The expected efficiency is given by a combination of the angular acceptance of the crystal compared to the output angle of the polycapillary, and the diffraction efficiency of the crystal. For Ge (111) the acceptance angle at 7.5 keV is 95 urad (or △E/E=3.5×10-4), and its reflectivity is nearly 100%. Since the acceptance is about a factor of 20 smaller than the measured value for the polycapillary divergence, we expect that the efficiency will be reduced by a similar factor. The measured efficiency of 4.6% is in good agreement with this expectation. Similarly the large mosaic spread of the graphite should allow it to collect the entire output angle of the optic. However, HOPG typically has a reflectivity of about 30%. Our measured efficiency of 30% is the expected value. From these considerations it is obvious that the ideal analyzer crystal for this application would be intermediate between the two cases measured here. Ideally it would have a mosaic spread of about 2 mrad combined with a high diffraction efficiency.

As a final test a calibrated mineral sample(Basalt glass BHV/O-2G from the USGS) was placed at the focus. This sample contains 110 ppm of Ni and 8.4 wt% Fe. Since the Fe fluorescence is at 6.4 and 7keV, this sample provides a good test of the background rejection of the spectrometer. After alignment using the Pilatus detector, it was replaced with an energy resolving Vortex Si drift detector. The allowed us to examine the energy spectrum of the beam transmitted by the spectrometer. Figure 5 shows the energy spectrum.

Fig. 5 Energy spectrum of the x-rays transmitted through the HOPG spectrometer for the standard sample BHVO-2G. The incident energy was 8350 eV.

It is seen that the Fe Kα and Kβ fluorescence and scattering background are virtually eliminated. The Ni Ka was about 14 kHz while the Fe background was about 10 Hz. Thus we have a background suppression of about 106, even without accounting for the additional background removal that can be provided by the energy resolving Vortex detector. This demonstrates the necessary background suppression for working on ultra-dilute samples.

3.2 Application to intermediate energy resolution spectrometers

For any x-ray spectrometer using diffractive analyzers, best performance ensues when collection solid-angle is maximized and when the physically-needed energy resolution is matched by each of the net spectrometer energy resolution and the integral reflectivity of the analyzer crystal itself. For high-resolution studies (i.e., ~1 eV or finer energy resolution) in the x-ray regime from a few keV to perhaps 10 keV, this latter condition can at least approximately be met by spherically-bent crystal analyzers (SBCAs) or by flat analyzers of semiconductor-grade single crystals, while the former criterion is a matter of ongoing development for SBCA arrays or other approaches. For experiments where the underlying physics requires much poorer energy resolution, i.e, ~100 eV, solid-state detectors or HOPG analyzers generally suffice.

Experiments requiring only intermediate energy resolutions, i.e., 5-10 eV, are more problematic. The energy resolution of solid-state detectors or HOPG-based spectrometers is too coarse, while state-of-the-art SBCA systems have small collection solid angles and unduly small integral reflectivities, each resulting in experimental inefficiencies. Here, we report two activities aimed at improving XES studies at such intermediate energy resolutions. First, we have developed a polycapillary coupled Bragg spectrometer, of the form described in the previous section (see Fig. 3), that is capable of working very near to a backscattering geometry. This gives the needed energy resolution despite the nontrivial divergence of the quasi-collimated beam emerging from the polycapillary optic. Second, we are investigating the use of Bragg analyzers that have been intentionally damaged by plastic deformation to introduce a mosaic spread and consequently increased integral reflectivity. Below, we describe the spectrometer itself, our progress toward enhancing the integral reflectivity, and then move to a representative application: the problem of measuring the Lγ1 XES from lanthanide species, an important diagnostic for the underlying physics of the volume collapse of lanthanides at high pressures. We show a picture of our spectrometer in Figure 6.

Fig. 6 Polycapillary coupled spectrometer as setup at the AFS beamline 16-ID-D (HP-CAT). The incident beam is nearly into the page.

To minimize the contribution from elastic scatter, the spectrometer works at 90 degrees from the incident beam (polarized in the horizontal direction). The entrance focal length of the collimating optic is ~1 cm. A long flight path (~50 cm) to the crystal analyzer is needed to keep the detector from intersecting the transmitted beam when working close to backscatter. This provides the additional benefit of cutting down on background scatter. For convenience, a scintillation detector was used to measure the fluorescence since it provides both a high count rate before saturation and also a modest energy resolution that is useful for further limiting higher energy background scatter. The flight paths and crystal enclosure are He filled to reduce absorption.

For optimal spectrometer performance the integral reflectivity, i.e. the integral of the reflectivity as a function of energy in a symmetric Bragg configuration, should be enhanced as much as possible. In the present case, the angular divergence of the collimating optic(△θ~2 mrad) provides an additional pragmatic threshold: any treatment of the analyzer crystal that increases its integral reflectivity while maintaining a mosaic spread somewhat below △θ will improve count rates at no cost in energy resolution.

Consequently we are investigating the use of mosaic Si crystals, generated from semiconductor-grade wafers by plastic deformation. This general approach has a long history, and has seen a recent reinvigoration with the development of hot-pressing techniques at the University of Sendai. Here, we use a less sophisticated approach that is well-suited to our purposes. At room temperature, the crystals were clamped in a cylindrical geometry using machined HOPG molds having various radii of curvature, Rc, as small as 30 cm. The HOPG clamp was then rapidly heated to~1300 C in an RF vacuum induction furnace, annealed for ~5 minutes, then allowed to rapidly cool back to room temperature. While this method of elastically deforming and plastically relaxing has far less flexibility for creating sophisticated engineering surfaces compared to hot-press techniques, it still introduces a high density of dislocations and corresponding mosaic spread in our simple Bragg analyzers. The cylindrically-deformed crystals can be elastically clamped onto a flat plate or heat treated again (on flat HOPG forms) to give the final, flat Bragg analyzer.

Several deformation treatments were tested at beamline 20-ID-B at the Advanced Photon Source. The Cu Kα emission lines were used as a reference standard for a comparison of Si 440 reflection efficiencies for different preparations. The same polycapillary collimator was used as is described above (see Table 1). Representative spectra are shown in Figure 7.

Fig. 7 Comparison of Cu Kα XES spectra recorded using untreated and deformed Si 110. The abscissa is kept in degrees to emphasize the angular resolution. The ’30 cm’ crystal has been bent to Rc=30 cm and then flattened. The ‘50cm x2’ crystal was heat treated twice, bent to Rc=50 cm both ways and then flattened. The peak at δθ=0 deg is Kα1 (8.05 keV) and the smaller peak at δθ=0.2 deg is Kα2 (8.03 keV).

In this study, a double bending procedure at 50 cm results in an increase in the integrated intensity by a factor of ~ 2.5 while maintaining the energy resolution achieved using untreated Si. This analyzer was prepared by plastically deforming the crystal to Rc = 50 cm, then plastically reversing the cylindrical bend (i.e., Rc = -50 cm), and finally plastically deforming back to a flat geometry. We have not studied the mosaic spread in detail, but initial measurements suggest that it may be somewhat reduced in comparison to that from the hot-press technique. While further work is needed to determine an optimal heat treatment, the results of Figure 7 already demonstrate an important technical improvement for polycapillary coupled spectrometers.

The measurement of Lγ nonresonant x-ray emission spectra (XES) of rare-earth elements has a direct sensitivity to the underlying physical mechanisms responsible for the volume collapse of lanthanides at high pressures. In such studies, an energy resolution of only ~5-10 eV is needed: the L3 core-hole lifetime is itself ~5 eV and the splitting of the main fluorescence line from the exchange satellite is typically ~20eV. The details of this spectrum determine the bare atomic magnetic moment as a function of pressure, thereby offering a test of, e.g., the role of the Kondo effect. While SBCA-based spectrometers can (and have) been used, significant inefficiencies result. The intermediate resolution provided by our polycapillary spectrometer presents a useful alternative. A comparison of the polycapillary spectrometer to a single SBCA (Si 333) is shown in Figure 8.

Fig. 8 Comparison of Nd Ly nonresonant XES spectra measured using a single SBCA (Si 333) and a polycapillary spectrometer coupled with a plastically deformed Si. The polycapillary instrument showed an increase of a factor of ~20 in the count rate at peak intensity.

In both cases, spectra were collected with (details of experiment at 16-ID). In this case, the polycapillary spectrometer used the same plastically-deformed Si 440 as provided the enhanced count rates in Figure 7, but a slightly different polycapillary collimator with modestly reduced collection solid angle. The polycapillary coupled spectrometer achieved a ~20 times improved count rate with no relevant loss in energy resolution or scientific potential.

Before concluding, we note that for all lanthanide Lγ x-ray emission lines there are simple Si or Ge orientations, shown in Table 3, that can be operated near to backscatter for energy resolutions of better than 10 eV, and often better than 5 eV. The estimated energy resolution comes from error propagation of the polycapillary exit divergence △θ through Bragg’s law. Hence, the general approach described here will be broadly applicable to studies of the lanthanide Lγ nonresonant XES.

Table 3. Simple Si and Ge wafer cuts for Lγ1 fluorescence energies. The general format is {reflection, energy resolution (eV), Bragg angle}

wdt_ID Element Lγ1 Energy (eV) Simple Si cut Simple Ge cut
1 Ce 6052 {333,2.2 eV,78.5 deg} {333,3.8 eV,70.2 deg}
2 Pr 6332 {333,4.1,69.5} {440,2.3,78.2}
3 Nd 6602 {440,2.5,78.0} {440,4.2,69.9}
4 Pm 6892 {440,4.5,69.5} {440,5.9,64.1}
5 Sm 7178 {440,6.1,64.0} {440,7.3,59.7}
6 Eu 7480 {440,7.6,59.7} {440,7.3,59.7}
7 Gd 7786 {444,3.1,77.2}
8 Tb 8102 {444,3.1,77.4} {444,5.3,62.5}
9 Dy 8419 {444,5.4,69.9} {444,7.1,64.4}
10 Ho 8747 {444,7.2,64.7} {444,8.7,60.2}


The paper has demonstrated the versatile nature of polycapillary optics in improving the detection of fluorescent or scattered photons. The confocal nature of the optics is important in rejecting the background, and the large collection angles possible can be used to improve the efficiency of a simple flat crystal spectrometer. It was shown that such a spectrometer can also provide the needed resolution and background rejection to allow the detection of very dilute components when coupled with an intense synchrotron source. It was also shown that when this type of spectrometer is operated at large Bragg angles, it can provide the useful combination of good energy resolution and high efficiency. For both of these applications matching the intrinsic resolution of the analyzer crystal to the output of the polycapillary is essential for achieving the best throughput. Intentionally damaging ideal Si crystals by plastic deformation is one possibility for improving the throughput at high resolution, and some promising initial results were presented.