Thilo Behrends† and Pieter Kleingeld
Department of Earth Sciences, Faculty of Geosciences, Utrecht University, The Netherlands. †behrends@geo.uu.nl
Analyzing heterogeneous samples from earth and environmental sciences with high spatial resolution became an indispensable approach for unraveling geochemical and biogeochemical processes controlling the speciation, coordination environment, and redox state of elements in these samples. A large variety of high-resolution techniques are nowadays available and their potential to provide valuable information about (bio)geochemical processes has been demonstrated in numerous applications. The list of techniques includes electron microscopy, electron microprobe analysis, ion microprobe analysis (secondary ion mass spectroscopy, SIMS), as well as laser ablation - inductively coupled plasma - mass spectroscopy (LA-ICP-MS), and synchrotron radiation based microanalysis as, for example, synchrotron micro X-ray fluorescence (S-μ-XRF ). All these techniques are used for determining elemental compositions in samples from earth and environmental sciences with high spatial resolution. Several of these techniques can provide additional information such as coordination environment, crystal structure, redox state, or isotopic composition depending on configuration. They all have specific advantages and limitations whereas spatial resolution, range of detectable elements, feasibility of quantitative analysis, detection limit, dynamic range, and required sample preparation are relevant criteria regarding the analysis of elemental compositions. But also accessibility and costs often decide whether microanalysis is performed on a set of samples or not.
"Accessibility and costs often decide whether microanalysis is performed on a set of samples" |
Compact μ-XRF instruments were launched into market just a few years ago. These instruments are attractive alternatives to other microanalysis instruments due to the relatively low price and relatively low demands regarding infrastructure, sample preparation, costs, and expertise required for operation. Recently, we decided to order a bench-top μ-XRF and in this article we want to share our experiences in the attempt to evaluate the possibilities and limitations of commercially available μ-XRF instruments, especially in view of other microanalytical techniques.
Micro-XRF has been applied to geochemical samples since years but the usage of μ-XRF analysis was only possible at a relatively small number of synchrotron beamlines providing S-μ-XRF facilities. This is because synchrotron radiation is characterized by orders of magnitude higher photon flux than conventional X-ray tubes which facilitates the generation of sufficient X-ray fluorescence from small irradiated sample spots for elemental analysis. Another reason why μ-XRF instruments have been exclusive to synchrotron facilities might have been the challenge of focusing X-ray beams. Synchrotrons are locations with concentrated expertise of X-ray optics which catalyzes the development and usage of optical elements for focusing hard X-rays. The most common optical device for focusing X-ray beams at S-μ-XRF beamlines are Kirkpatrick-Baez mirrors. KB-mirrors focus the beam by reflecting X-rays at the coated surface of the curved mirrors. The loss of intensity, due to the low reflectance of these surfaces is compensated by the extreme high intensity of the collimated x-ray beam.
Focusing and collimating X-ray beams in commercial μ-XRF instruments
Building μ-XRF instruments principally based on conventional X-ray tubes has been enabled by the development of X-ray capillary optics, which allow designing much more compact and robust optics for collimating and focusing X-ray beams.
The simplest way to reduce the spot size of an X-ray beam is to use apertures to narrow down the size of the outgoing beam. The problem is that only a small spatial segment from the emitted X-rays is then used for the analysis resulting in relatively low yield of X-ray fluorescence. In practice, apertures are used in μ-XRF instruments for spot sizes of 1 mm or larger. Creating smaller spot sizes with apertures is impractical because the final X-ray intensity becomes so low that fluorescence generation in the sample would be insufficient for analysis.
All commercial μ-XRF instruments make use of the effective diffraction of X-rays at the surface of glass capillaries for creating smaller spot sizes. The possibility of using glass capillaries to convert divergent X-rays into a parallel beam or to focus X-rays has been first reported by M. A. Kumakhov in 1984 and extensively investigated since then (Kumakhov and Komarov, 1990). Transformation of divergent X-rays into a parallel beam by using glass capillaries is schematically shown in Fig. 1. For μ-XRF applications two different designs are used: mono-capillary and poly-capillary. Highly collimated X-rays can be produced by mono-capillaries. The advantage of highly collimated X-rays is that μ-XRF analysis can be combined with X-ray transmission imaging. Another advantage is that the vertical positioning (working distance) of the sample is less crucial, and that the probed sample volume is better defined. Furthermore, mono-capillaries are offered for μ-XRF with spot sizes down to 10 μm, which is smaller than spot sizes of presently available poly-capillaries. The downside of these advantages is that the photon flux, emitted by mono capillaries on the sampling spot, is lower compared to poly-capillaries, which are typically arranged to create focused beams.
| Figure 1: Collimating or focusing of divergent X-rays with the help of polycapillaries. Source: X-ray Optical Systems Inc. |
Lateral resolution and fluorescence yield (what is the best capillary option?)
Deciding about the most suitable x-ray optics for our expected applications was an important aspect in the selection procedure. Of special interest was the characterization of the sampling spot to obtain an impression of the achievable spatial resolution. For this purpose a gold wire with a diameter of about 30 μm was fixed in calcite containing resin material and a two-dimensional scan was performed with the best possible resolution of the different instruments. Exemplarily, a contour plot of the intensity of Au L α fluorescence is shown on the left side of Fig. 2. The asymmetric shape is due to the fact that the gold wire was not perpendicularly fixed in the resin and that fluorescence from the gold wire below the surface contributes to the signal as well. This illustrates that not only the lateral but also the vertical resolution can be of great importance, which will be further discussed below. The escape depth of the Au M α fluorescence is much smaller compared to the L α lines so that the Au M α contour plot provides a better representation of the lateral dimensions and the intensity profile of the X-ray beam (Fig. 2b). The diameters of the measured peak ranged from 35 μm to 100 μm at half maximum and corresponded to the expected dimension when adding the spot size of the beam, as specified by the producer and varying between 10 μm and 60 μm, to the diameter of the wire. The best lateral resolution was achieved with a mono capillary as anticipated.
It is difficult to directly compare fluorescence yield of different capillary optics when different instruments were used, even when normalizing to the different spot sizes. This is because the counting rates depend on a large number of parameters as geometric arrangement of tube and detector, type of detector, type X-ray tube, applied tube power, application of vacuum or He flushing between sample and detector etc. In our tests we did not aim at determining the fluorescence yield as a function of capillary optics. However, we observed the trend that counting rates were, at comparable spot size and identical sample, significantly lower when using mono-capillaries compared to poly-capillaries. Lower counting rates imply that longer counting times are required to obtain similar counting statistics as with high counting rates. Long counting times can be a drawback in elemental mapping and a counting time per pixel of 5 seconds makes surveying an area of 10,000 pixels (1mm2 in 10 μm steps) already an overnight task.
 | Figure 2: Contour plot of X-ray fluorescence collected from a gold wire (30 μm) imbedded into resin material. The left figure shows the contour plot for the Au L α lines and the right figure the Au M α fluorescence. Due to the greater escape depth of the Au L α lines fluorescence from the inclining gold wire below the surface is detected. The Au M α fluorescence only probes the gold wire at the surface of the sample. |
Lateral vs. vertical resolution (what is the best spot size?)
High lateral resolution can be very useful when analyzing quasi two-dimensional samples but the benefit for analyzing three-dimensional heterogeneous samples also depends on the depth to which the X-ray probes the sample. The vertical extension of the probed volume is controlled by two factors a) the penetration depths of the primary beam and b) the escape depth of the generated X-ray fluorescence. The latter is a function of the absorption length of the matrix, which, in turn, depends on the energy of the X-ray fluorescence and the electron density of the matrix. For illustrating how X-ray energy and matrix characteristics influence the analysis depth, a thin sapphire layer was imbedded between two layers of steel. This sandwich was fixed in a resin with an inclining angle of 30o. The absorption length (at which the logarithmic X-ray intensity ln (I/I0) decreases by one unit) of Fe K α radiation in sapphire is 41 μm while for Al α radiation in iron it is only 0.4 μm. The consequences of this difference can be seen in Fig.3 showing the fluorescence intensities of Fe and Al across the steel/sapphire/steel sequence. On the right side of the figure sapphire is inclining under the steel layer, which efficiently absorbs the Al fluorescence from the sapphire below. The width and shape of the edge is, therefore, predominately controlled by the spot size. In contrary, sapphire is relatively transparent for Fe fluorescence and the Fe signal is detected about 400 μm away from the steel/sapphire interface at the sample surface. This means that Fe is detected about 200 μm below the sapphire layer. The line represents the theoretical intensity change with distance when only considering Fe K α absorption in sapphire. The deviation from the theoretical line is caused by neglecting the change of primary beam intensity with depth. However, in this case, absorption of the primary beam is apparently only of minor importance and the depth of the probed volume for Fe is predominately constrained by absorption of Fe fluorescence in the sapphire layer.
| Figure 3: Transect through a sequence of steelsapphiresteel layers inclining with 30o degrees. The orientation of the different layers is schematically shown at the top of the figure. The fluorescence intensity for Fe and Al is normalized to the maximum intensity measured along the transect and the natural logarithm of this ratio is plotted vs. lateral position. On the right side, steel overlaps the sapphire and effectively screens the Al fluorescence from deeper parts of the sample. This leads to relatively sharp edge shapes on the right side. In contrast, Fe fluorescence is detected down to 200 μm below the sapphire layer, leading to a smooth edge shape for Fe on the left side. The line shows the expected Fe fluorescence calculated from the sapphire thickness and the absorption of Fe fluorescence in sapphire. These measurements demonstrate how the depth of the probed sample volume depends on the energy of X-ray fluorescence and matrix composition. |
This exercise also shows that small beam spots are not necessarily advantageous when dealing with three dimensional, isotropically heterogeneous samples, as we often do in earth sciences. Adsorption lengths of X-ray beams with energies between 4.5 and 8 keV (K α lines of transition elements between Ti and Cu) in silica range from 24 μm to 124 μm, so that beam spots in the range of 10s of μm might be the best compromise for three dimensional heterogeneous samples, in view of spatial resolution and fluorescence yield.
Detection limit and quantitative analysis
In general, μ-XRF instruments are not suitable to analyze elements lighter than Na. For other elements, detection limits can vary by two orders of magnitudes and more due to differences in the element specific fluorescence yield and the energy of the X-ray fluorescence. The intensity loss of low energy X-rays on their path to the detector can be significant and the importance of X-ray absorption implies that detection limits also depend on the sample matrix. For transition elements detection limits between 10 and 100 ppm seem to be achievable which can be considered as the lowest possible detection limits of current μ-XRF instruments. All instruments we tested offer the possibility to analyze the sample under vacuum or under He flux, which increases the sensitivity of the instrument for lighter elements such as Al and Si.
It should be mentioned that also the characteristics of the X-ray detector contributes to the overall performance of the instruments. At the moment, wavelength dispersive detectors are not available with μ-XRF instruments and only instruments with energy dispersive (ED) detectors are available. Some producers offer the option to select between different detectors, whereas silicon drift detectors are interesting options because they do not require liquid nitrogen cooling and offer good energy resolution and high maximum counting rates.
The challenges of using μ-XRF for quantitative analysis are the same as for conventional XRF instruments and other microanalysis techniques relying on X-ray fluorescence (e.g. electron microprobe) and will not be discussed here. In our tests we used rare earth element (REE) standards in glass matrix. Accounting for spectral interferences of the different REE in complex mixtures are challenging tasks in quantitative XRF analysis. In general, the standard-free (fundamental principle) quantification methods worked well for the different reference samples, in particular when including information from the glass matrix in the quantification. The determined concentrations were in the range of ±20% of the specified concentrations at REE concentrations around 5 mg/kg.
μ-XRF vs other microanalysis techniques
In view of the costs of the instrument and efforts required to perform measurements μ-XRF can be possibly best compared to scanning electron microscopy (SEM). Electron beams can be used to generate X-ray fluorescence and SEM instruments, equipped with ED detectors, can be used for elemental analysis. The great advantage of electron beams is that they can be much easier focused than X-rays and resolutions in the nm range are standard for SEM instruments. In contrast to X-ray fluorescence, the analyzing depth is much shallower because the electron beam does not penetrate into the sample as deep as the primary X-ray beam in XRF. A principle disadvantage of using electron beams for XRF spectroscopy is that they create much higher background radiation than X-rays due to the formation of bremsstrahlung in the sample. Furthermore, SEM instruments are, in the first instance, designed for optimal visualization of the sample and not for elemental imaging. Hence, when elemental analysis is the main goal, μ-XRF is superior to SEM as long as the spatial resolution of the μ-XRF is sufficient. Another aspect is that μ-XRF does not necessarily require vacuum, so that the analysis of wet samples is possible. Finally, the sample chamber of bench-top μ-XRF is much larger compared to SEM instruments so that the analysis of larger samples is no problem.
"When elemental analysis is the main goal, μ-XRF is superior to SEM as long as the spatial resolution of the μ-XRF is sufficient" |
Regarding elemental analysis, transmission electron microscopy (TEM) has the same restrictions as SEM, besides that only tiny samples can be analyzed. The strength of TEM is that visualization and elemental analysis can be combined with complementary information when using electron diffraction (crystal structure) or electron energy loss spectroscopy (redox state, coordination environment).
Bench-top μ-XRF cannot compete with synchrotron μ-XRF (S-μ-XRF) where a beam spot size of 100 nm and lower can be realized (for three dimensional heterogeneous samples the considerations regarding analyzing depth are the same as for bench-top μ-XRF). Another strength of S-μ-XRF is that elemental analysis can be combined with other microanalysis techniques such as micro-X-ray absorption spectroscopy (redox state, coordination environment) or micro-XRD (crystal structure). μ-XRF can certainly also not redundantize other high-performance microanalysis techniques such as electron microprobe, SIMS, or LA-ICP-MS.
In conclusion, bench-top μ-XRF does not open principally new analytical possibilities in Earth sciences, but it has a great potential to be widely used for elemental imaging when the desired information can be obtained within the constraints of resolution (>10 μm), atomic number (>10), and detection limit (>10 ppm ). The plus factors are the user friendliness, relatively low costs, and that practically no or only little sample preparation is required. Besides serving as stand alone instrument for elemental analysis, bench-top μ-XRF can be very useful for selecting samples and sample areas before applying more sophisticated microanalysis techniques.
References
Kumakhov M. A. and Komarov F. F. (1990) Multiple Reflection from Surface X-Ray Optics. Physics Reports-Review Section of Physics Letters 191(5), 289-350.