Three-Dimensional Elemental Imaging Using a Confocal X-Ray Fluorescence Microscope

Micro X-ray fluorescence (MXRF) is a powerful elemental technique that can nondestructively provide both single-point spectra and full spectral elemental maps.1 The X-ray spot size of a commercial instrument is approx. 10-50 μm; thus, an elemental map may contain between 1600 and 40,000 pixels. Full spectral mapping of the sample can therefore generate the same number of discrete spectra. MXRF images are important because they provide several orders of information, including qualitative and quantitative information on elemental species, as well as heterogeneities and spatial distribution of the elemental species present. The mapping capability essentially provides a picture of the elemental distribution within the material, which very easily provides a tremendous amount of information. Collecting full spectra at each pixel generates hyperspectral data sets. When these hyperspectral data sets are processed with chemometric software, unexpected elements may be detected and chemical phases revealed (i.e., elemental correlations are spatially identified).2 The typical MXRF instrument uses an optic to spatially restrict the excitation X-rays to a small spot, with no optic on the detector. Therefore, fluorescent X-ray photons are detected from the full critical depth of the analyte excitation. Elemental discrimination as a function of depth is not possible in this approach.

Confocal X-ray fluorescence uses an optic on both the excitation source and the detector. This arrangement produces the confocal volume, allowing one to spatially discriminate the source of the X-ray fluorescent photon in both the x–y directions as well as the z direction. A schematic layout of a typical confocal MXRF is given in Ref. 3. A more recent demonstration of a laboratory confocal system has been published.4 Confocal MXRF has also been previously discussed using a synchrotron5; however, synchrotron availability is limited. The current confocal instrument fits easily on a 24 × 24 in. breadboard, with the associated electronics on a rack underneath and the computer workstation next to it. The small size of the instrument makes it ideal for on-line, at-line, and off-line measurements.

Confocal MXRF can be used in such applications as forensics, cultural provenance, minerals, materials sciences, thin films, particulate characterization, pharmaceuticals, polymers, fossils, nanotechnology, and many others.

Instrument design

An X-beam Ag X-ray tube (X-ray Optical Systems [XOS] East Greenbush, NY) powered by an XLG high-voltage power supply (Spellman High Voltage Co., Hauppauge, NY), 50 kV, 0.5 mA, 25 W max, was used as the X-ray source. A Si pin diode detector (model XR-100CR, Amptek, Bedford, MA) was used to detect the X-ray fluorescent photons. The source and detector optics consisted of a pair of monolithic polycapillaries (XOS) with a focal spot size of approx. 35 μm. The angle between the optics is approx. 60°, 30° from the surface normal of the sample, producing a working distance of 10 mm. This optical arrangement is a compromise between the ideal 90° geometry for optimum spatial resolution and greater X-ray depth penetration into a specimen. Three 850G actuators (Newport Corp., Irvine, CA) were used to drive the sample stages. Everything was under computer control using XOS software.

Line profiles

Figure 1 - Optical photograph showing tip of capillary optics, sample position, and beam path. Sample is raised up into confocal volume and scanned in all three dimensions.

The confocal volume was determined by profiling a tantalum foil (10 μm thick) in all three dimensions. The confocal volume was found to have dimensions of 40 μm (full width half maximum) in the x and y directions and 60 μm in the z direction. The optics and sample geometry are shown in Figure 1.

Each elemental signal was measured as a function of integrated intensity under the emission band. For each element , the line and energy region of interest (ROI) are as shown in Table 1.

Each element was recorded as a function of position versus intensity. The line profile of each elemental intensity was normalized to one. Dwell time was set at 1 sec, with 30-μm spacing between measurements. The current instrument is limited to elements with energies between 3 and 20 keV due to air absorption.

Figure 2 - Optical geometry of line scans and 3-D imaging of paint chip. Orientation A, sample layers are perpendicular to beam; orientation B, sample layers are parallel to beam; orientation C, sample is laid flat. Orientation and scan directions are offset by 90° between A and B.

The effect of sample orientation is critical in confocal MXRF. This is illustrated using a multilayer paint chip. The paint chip, 1 mm thick, is composed of calcium-, titanium-, iron-, zinc-, and lead-based pigments. Figure 2a–c shows the sample orientations of the cross- section. The paint chip appears to have at least 18 distinct layers, with lead more prevalent in the layers on one side. This side will be defined as the bottom of the sample, i.e., early painting.

In orientation A, shown in Figure 2a, the sample cross-section strata are perpendicular to the excitation–detection alignment. The z position was determined by the height of maximum signal; therefore, the confocal volume was imbedded in the surface of the sample. The sample was then profiled from top to bottom, which shows the lighter elements first, followed by the bottom lead layers. As the beam enters the sample, it will be attenuated slightly by the upper and lower layers of paint.

In the second orientation (Figure 2b), the sample cross-section is parallel to the excitation–detection alignment. The line profile was completed by moving the sample left to right such that the beam entering the sample remains approximately in the layer being probed.

Finally, in orientation C (Figure 2c), the sample is laid flat and the profile is measured by moving the sample up through the confocal volume. In this instance, the upper layers will attenuate the sampling beam and the detection beam. In addition, the spatial resolution is decreased by almost 50%.

3-D imaging

Figure 3 - Series of x–y maps at different z depths within the sample for orientation B. All dimensions are in millimeters.

Figure 4 - Three-dimensional elemental image of titanium from the x–y maps shown in Figure 3. All dimensions are in millimeters.

Three-dimensional imaging is possible because the detector optic is collecting photons given off at the focus; thus, a map as a function of the x, y, and z positions versus elemental X-ray fluorescence intensity can be collected. As shown in Figure 3 for titanium, x–y maps at each z position are collected successively. Importing each of the x–y maps into MATLAB 7.04 (MathWorks, Inc., Boston, MA), it is possible to then generate 3-D images of each element individually, as seen in Figure 4. In MATLAB, it is possible to generate 2-D images, 3-D images, and slices, as well as rotate 3-D constructs that can be created and converted into movies.

Figure 5 - Three-dimensional elemental images of paint chip in three different orientations. Images are rotated to have the same orientation. Layers for A and B are up–down with respect to the page, and are not seen to deeper depths because of attenuation. The layers are oriented left–right in image C. The elements are color coded as follows: Ca = green, Ti = red, Fe = yellow, Zn = blue, and Pb = gray. All dimensions are in millimeters.

Full 3-D scans of the paint chip were completed with the same orientations as the line scans with 41 × 41 (x–y) pixel 2-D maps at 14 different z depths. The step size per voxel is 30 × 30 × 50 μm, giving a total of 23,534 voxels, with a 1-sec dwell at each. This produced elemental images 1.23 × 1.23 × 0.65 mm in size. In orientation C, only 11 z positions were mapped due to attenuation of the signal with depth. Each x–y 2-D map took approx. 40 min to collect. The isosurface of each element in the 3-D construct is determined by the selected elemental intensity value. Isosurfaces for each element are 12% Ca, 16% Ti, 34% Fe, 4% Zn, and 28% Pb of their maximum intensity. The images are displayed as 50% transparent. These multiple-element 3-D images are shown in Figure 5a–c for each orientation, respectively.

Results and discussion

Figure 6 - Line scans of elements of normalized elemental intensity in each of the three sample orientations shown in Figure 2.

Comparison of the line profiles for orientation A versus orientation B (Figure 6a and b), shows that the sample layers that are perpendicular to the beam exhibit a loss in intensity at the edges. The titanium and iron layers are almost completely lost on the upper surface, and the bottom lead layer is greatly attenuated in orientation A (note arrows in Figure 6b). While the two line scans comparing the different orientations are not of the exact same area of the sample, they are close. Multiple line scans were obtained in each orientation to verify these observations.

Comparison of orientations A and B with C (Figure 6c) produces a stark contrast. Since the beam size in the z direction is almost 50% larger, a concomitant loss in spatial resolution is seen. This loss leads to an overlap of the layers in the individual line scans, which are nondestructive depth profiles. Orientations A and B were obtained at the cross-sectional surface while orientation C was done in the center of the sample. This means that both the excitation and detection beam will be increasingly attenuated with depth, whereas in A and B a much smaller portion of the beam path was in the sample and thus experienced less attenuation.

Figure 3 demonstrates the output of the x–y maps from the paint chip mapped in orientation B. Each 2-D slice is 50 μm successively deeper into the sample. Heterogeneities in the titanium layer can be seen, as well as the number of paint layers containing titanium. As the confocal volume probes deeper and deeper into the sample, attenuation effects are seen, and titanium becomes increasingly difficult to detect, eventually disappearing below 300 μm. Figure 4 shows the titanium layers in 3-D after importing the data set into MATLAB for visualization.

Comparison of the 3-D images in Figure 5a–c obtained in the different orientations is also instructive. Figure 5a shows the 3-D composite elemental images of all the elements mapped. While the edges appear sharper in orientation B (Figure 5b), as would be expected, orientation C (Figure 5c) shows a Ca inclusion that is not seen when the sample is imaged on edge. This demonstrates that the elemental distribution information gained will vary depending on sample orientation.

When a z depth profile is taken of a solid single-element sample, as the beam penetrates into the sample, an increase will be seen in the element intensity until the signal reaches a maximum. The signal will then decrease as the upper layers of the sample attenuate the beam at greater and greater depths, creating a null point for the isovalue and the bottom surface of the construct. This can be seen in Figure 5c, where there are two apparent surfaces for each element. Therefore, 3-D imaging is useful for the upper 0–400 μm of this particular sample. The depth of penetration will vary for each element and its associated matrix.

Conclusion

Confocal MXRF changes the elemental analysis paradigm for materials characterization. This laboratory-based approach offers nondestructive single-point line scans, depth profiles, 2-D maps, and 3-D elemental images.

The current software implementation is limited to ROI intensities when collecting 3- D images. Future work is focusing on development of enhanced software that will collect full spectral data at each point to utilize the power of chemometrics for image processing. This enhanced software will provide several advantages, including reducing the scan time, allowing for cluster analysis in the images, as well as postcollection processing for elements that may not have been noted at the beginning of the data acquisition.

Voxel volume can also be reduced through the use of different capillary optics, down to a 10-μm spot size. Although the spatial resolution would be increased, scan times would also be increased. The current 60° angle between the optics is not optimal. A 90° geometry would reduce the confocal volume in the z direction to a value comparable to the x and y. This would improve spatial resolution in the z direction, but would also improve sensitivity since the same flux would be focused into a smaller volume, at a loss in X-ray penetration depth.

The current data are based only on raw signal intensity for the 3-D imaging. Modeling of the beam attenuation on both the excitation and detection sides is needed to provide true elemental intensities at depth. This would enable nondestructive, quantitative 3-D elemental imaging.

Although there is much work remaining to make confocal MXRF a routine method, the current capabilities show the potential power of this method for materials characterization.

References

  1. Havrilla, G.J.; Miller, T. Micro X-ray fluorescence in materials characterization. Powder Diffr. 2004, 19(2), 119–26.
  2. Miller, T; Havrilla, G.J. Elemental imaging for pharmaceutical tablet formulation analysis by micro X-ray fluorescence. Powder Diffr. 2005, 20(2), 153–7.
  3. Ding, X.; Gao, N.; Havrilla, G. Monolithic polycapillary X-ray optics engineered to meet a wide range of applications. Proceedings of SPIE—The International Society for Optical Engineering, 2000, 4144, 174–82.
  4. Kanngiesser, B; Malzer, W; Reiche, I. A new 3D micro X-ray fluorescence analysis set-up. First archaeometric applications. Nucl. Instrum. Meth. Phys. Res., Sect. B (Beam Interactions with Materials and Atoms) 2003, 211(2), 259–64.
  5. Vekemans, B.; Vincze, L.; Brenker, F.E.; Adams, F. Processing of three-dimensional microscopic X-ray fluorescence data. J. Anal. At. Spectrom. 2004, 19(10), 1302–8.

Dr. Patterson is a Post-Doctoral Research Associate, and Dr. Havrilla is a Technical Staff Member, Los Alamos National Laboratory, P.O. Box 1663, MS K484, Los Alamos, NM 87545, U.S.A.; tel.: 505-667-9627; fax: 505-665-5982; e-mail: [email protected]. This work was funded by the Department of Energy’s (DOE) Office of Science.