The apparatus consists in an home-made shear-force microscope (see Figure 1) whose probe is a sharp Aluminium-coated optical fibre (aperture 50 nm) that locally collects the visible light emitted by the sample illuminated by X-Ray radiation (synchrotron environment). The instrument, working in ambient conditions or in liquid environment, allows simultaneous pixel by pixel surface topography measurement and chemical mapping 15. The analysed sample must fit with Scanning Probe Microscopy requirements (solid sample, roughness in the micronscale range). This apparatus is evaluated by characterization of ZnO and ZnWO₄ - ZnO thin layers, exhibiting a high luminescence yield. The luminescence spectra are compared to those obtained in far field

This apparatus also enables the XRF signal local collection of the excited sample, replacing the device tuning fork-optical fibre by a fixed X-ray cylindrical capillary (internal diameter 10 μm, length 50 mm). The sample is excited by a rotating anode (excitation at constant energy, Cu Kα at 8 keV, power 40 kV × 40 mA) while the fluorescence signal is analyzed by EDX (Energy Dispersive X-ray). The excitation beam is focused on the sample by a capillary lens (spot diameter 20 μm) provided by IFG GmbH. The XRF technique is particularly suitable for analysis of heavy elements, typically heavier than sodium.

In Figure 2a we present the XAFS-XEOL spectrum obtained with the apparatus at ESRF ID03 line of a ZnO thin layer (~400 nm), prepared by Zn sputtering on a silicon substrate, followed by a 900°C annealing in air. The threshold, localized at 9664 eV, is characteristic of visible light emitted by Zn atoms after X-Ray absorption. This spectrum is compared with that of the same sample (Figure 2b, bottom) and with that of a commercial stoichiometric ZnO powder sample for reference (Figure 2b, top, shifted), obtained in conventional XAFS-XEOL spectroscopy, in far field, at the same beamline. Spectra shown in Figure 2b are in very good agreement in terms of both peak positions and relative magnitudes measured with respect to the average signal above threshold. This indicates that the ZnO sputtered layer is stoichiometric. The great concordance between spectra Figure 2a and Figure 2b validates the instrument concept.

A ZnWO₄ - ZnO thin layer (~400 nm) was prepared by co-sputtering Zn and W onto a silicon substrate, followed by a 900°C annealing in air. In Figure 3 we show twin images corresponding to the simultaneous record of both topography and luminescence cartography of the ZnWO₄ - ZnO sample at various incident energies. In upper Figures 3-a, b, c, d the topography is presented. Grains of 0.5 to more than 1μm are observed, as was confirmed by conventional Atomic Force Microscopy. In Figures 3-e, f, g, h we present the corresponding luminescence cartography obtained respectively, from the left to the right, before and after the Zn-K edge, as well as before and after the W-L edge. Images 3a to 3 h contain 1024 × 1024 pixels. The remarkable stability of the instrument is noticeable, since it took about 8 h for recording this whole set of images. Image 3 g, obtained at higher X-ray energy than the Zn threshold, also highlights Zn rich regions. The contrast is lower than in Figure 3f since the acquisition is performed far from the maximum emission. Black zones correspond to non emitting or to grains emitting out of the fibre acceptance angle.

Post image processing can be carried out on Figure 3e to 3h to define ZnO and ZnWO₄ rich areas. First, the pixel to pixel difference Figure 3f - Figure 3e (resp. Figure 3h - Figure 3g) gives the distribution of Zn (resp. W) luminescent sites. Then, to enhance the contrast, these two images are further converted in black and white scale. By this way we get two intermediate images, which are then used to obtain a chemical mapping of the layer: the ZnO rich emitting areas can be obtained by difference of these intermediate images (Figure 4a), since Zn is present in both materials while W can be found only in ZnWO₄ grains. Finally, a logic operation 'AND' is applied between the intermediate images to highlight the distribution of emitting ZnWO₄ (Figure 4b) since Zn must be present in both materials. In fact a white pixel in Figure 4b is obtained only if the same pixel appears simultaneously white on both intermediate images. This image processing leads to a two-level (black and white) image which increases significantly the contrast. Since Figure 4a shows only few features, one can conclude the emitting centres are almost pure ZnWO₄, as confirmed by XRD and micro-Raman analysis 16. No obvious correlation with the topography is noticeable, since the emitting zones are not specifically centered in the grains.

Collecting the XEOL signal in near field significantly increases the lateral resolution of this technique, which is now only limited by the aperture of the optical fibre. In fact, the resolution of the apparatus is limited by the tip curvature for topography (~100 nm) and by the optical aperture for the light collection (~50 nm).

Replacing the device tuning fork-optical fibre by a fixed X-ray cylindrical capillary the XRF collection concept feasibility is demonstrated on a test sample, composed of bulk Co and Ti juxtaposed sheets. The X-ray beam simultaneously illuminates both Co and Ti samples. Figure 5 shows XRF spectra obtained using 10 μm diameter cylindrical capillary approached at a distance of 5 mm from the sample surface. We obtain the Kα and Kβ characteristic peaks of both Co and Ti, as reported in literature 17. Since the fluorescence yield of Co is twice that of Ti at excitation energy of 8 keV, the incident spot might be slightly shifted on the titanium sheet regarding the Co-Ti separation.

With commonly marketed XRF equipment, without capillary for detection, the lateral resolution is limited by the diameter of the primary probe, in the range of 10 μm. A resolution increase can be achieved by shrinking down the detector aperture. However, increasing the resolution from 10 to 1 μm, would lead to a factor loss of 100 on the signal. To reach the original signal level, the sample-detector distance must be drastically decreased. However the steric hindrance of the EDX detector (surface of about 1 cm²) impedes to approach the detector at distances lower than 5 mm. Consequently, a solution to avoid primary beam shadowing is to use a low diameter capillary to collect the fluorescence signal at the vicinity of the surface. Furthermore, using for example a cylindrical capillary to collect the signal enhances significantly the signal level regarding a pinhole of the same diameter at a given sample-detector distance 18. The gain G is given by:

Where θ_c is the critical angle of the capillary material (in our case fused silica with θ_c of about 5 mrad at the X-ray energy considered in this paper 19), D is the detector-sample distance and d is the capillary diameter. G is about 3.10³ (resp. 3.10⁵) for a 50 mm long and 10 μm (resp. 1 μm) diameter cylindrical capillary approached at 5 mm from the sample surface. Moreover, the use of elliptical instead of cylindrical capillary would further increase the signal level by a factor 20 2021. Our experience shows that we can combine X-ray capillary optics for both excitation and detection to substantially increase the resolution of in-lab XRF technique which can be better than 1μm keeping a significant signal to noise ratio and remaining in satisfactory acquisition times 22.

We have constructed a new Shear-Force Microscopy head that is able to simultaneously record the topography and the light emitted by a sample. We have demonstrated in synchrotron environment the possibility of simultaneous XEOL mapping and surface topography with a resolution of 50 nm. The instrument is thus able to image the surface and to localize a peculiar object that can be further chemically analyzed by XEOL analysis. Thanks to the recent development of new X-Ray capillary lens, we now equip our home-made Shear Force Microscope with a tightly focused laboratory X-ray source for on-table simultaneous Luminescence-Topography measurements. The sensitivity of the technique, limited by the signal to noise ratio, will be evaluated in the future.

We have demonstrated the concept feasibility of XRF analysis at micrometer scale. In fact, replacing the optical fibre of our microscope head by a 10 μm diameter cylindrical capillary, we succeeded in local collection of sample XRF under X-ray illumination using an in-lab source. The signal level obtained in this work enables to estimate that the lateral resolution of the technique can still be improved. Consequently sub-1 μm resolution can be reached in lab, whereas, using brighter excitation sources (synchrotron), sub-100 nm resolution is expected, limited today by capillary technology. The final idea is to use an elliptic capillary as shear force probe to simultaneously obtain topography and the XRF mapping of the sample.