Open Access
Add to BibSonomyAbstract
We have developed an electron beam excitation ultra-soft X-ray add-on device for a scanning electron microscope with a reflective zone plate mulichannel spectrometer in order to analyse ultra-light elements such as Li and B. This spectrometer has high (λ/Δλ~100) resolving power in the energy range of 45 eV – 1120 eV. Metallic Li samples were examined and fluorescence spectra successfully measured. Energy resolution of 0.49 eV was measured in the ultra-low energy range using the Al L2,3 line at 71 eV. High sensitivity of Boron detection was demonstrated on a B4C sample with layer thicknesses of 1-50 nm, detecting an amount of metallic Boron as small as ~0.57 fg.
© 2015 Optical Society of America
1. Introduction
Even though ultralight elements (Z = 3, 4, 5) are quite rare and only found in compounds, they are used in a number of applications. With the advent of semiconductor industry, Li and B started to be used as dopants. One of the most promising energy storage technologies are Li-Polymer and Li-Ion batteries. All elements are also extensively used for scientific purposes, such as Be X-ray windows. Some of the materials with the highest hardness and melting point, like B4C, B-N-C or c-BN, have such properties because of small additions of B. A scanning electron microscope is a very versatile tool that enables one to obtain different information about the sample. In X-ray emission spectroscopy, the sample is excited using the electron beam and the atoms are relaxed by emitting an X-ray photon with characteristic energy that depends on the atom type, the atom chemical environment and the transition that the electrons in the atom underwent during the process. For this method one needs to use a suitable detector to detect the emitted photons. Since the X-ray fluorescence lines of ultralight elements have relatively low energy (E < 100 eV), they cannot be investigated using standard parallel energy dispersive detectors (EDS), but rather using crystal- or grating based wavelength dispersive spectrometers (WDS). The latter provide much higher resolution and can be designed to detect any energy, but are at the same time very slow. The recent development of JEOL Company enables parallel spectra registration with high energy resolution for a transmission electron microscope [1] and a scanning electron microscope [2].
A novel parallel WDS is being designed by IfG Institute for Scientific Instruments GmbH with optics developed by Helmholtz Zentrum Berlin. It is based on reflection zone plates (RZP), which act as a dispersive, reflective and focusing element at the same time. The low efficiency of the traditional multi element WDS is overcome by only one optical element, where losses are reduced significantly. Instead of using only one RZP that can focus a limited energy range, a multichannel array of them is used. By designing each zone plate in the array to focus a different energy in the range from 45 eV to 1200 eV, the spectrum can be recorded in parallel using a position sensitive detector. Results of our measurements of thin layers of B4C with determined limit of detection in mass units, measurements of resolution on Al L2,3 band spectrum and measurements of Li are presented in this paper.
2. Instrumentation
All the experiments were performed on Zeiss DSM 942 scanning electron microscope (SEM). This SEM has an electron probe current from ~30 nA to ~110 nA at the accelerating voltage of 5 kV. The pressure in the sample chamber was around ~10−6 mbar.
Photons are recorded with a Greateyes GmbH GE 2048 512 BI UV1 CCD camera. It has a full-frame back illuminated sensor with 2048 x 512 pixels. The pixel size is 13.5 μm x 13.5 μm, which corresponds to a total sensor size of 27.6 mm x 6.9 mm. The solid angle of the RZP on the detector in vertical x horizontal dimensions is 27.7 mrad x 110 mrad.
3. Reflection zone plate multichannel spectrometer
The basic principle of a reflection zone plate used in the spectrometer was already covered in detail in [3–5 ].
Throughout the experiments, two different multichannel spectrometers were used. Each has a different energy range and specific parameters in that range. Different spectrometers complement each other by covering ultra-low energy range (14-channel spectrometer) and higher energies up to 1116 eV (17-channel spectrometer).
The 17-channel spectrometer was already described thoroughly in [5]. It enables the detection of energies from 54 eV (Li Kα) to 1116 eV (Ga Lα). However, the efficiency of the channel at Li Kα energy is only 3%. On the other hand, the B Kα channel has the maximum efficiency of 18%.
A new optical element with 14 channels was designed to extend the usable range of the 17 channel spectrometer into energies below ~100 eV. More specifically, it covers the XUV range from 46 eV to 92 eV. The parameters of 17–channel and 14-channel spectrometers are shown in Tables 1 and 2 .
Table 1. Characteristic fluorescence line energies for the 17-channel RZPs and their efficiency in % (theoretical calculations, B – measured value))
Table 2. Characteristic fluorescence line energies for the 14-channel RZPs and their efficiency in %, measured values.
This spectrometer has two important improvements over earlier reported spectrometers in this range:
- • The channels are carefuly optimized in efficiency. Accurate measurements were done using BESSY II reflectometer for each channel at the optics beamline [6]. The experimental results are in a good agreement with theoretical calculations.
- • Besides the 12 analysing channels, there are 2 alignment channels on each lateral side of the RZP. These two channels are designed for 277 eV (C Kα). The majority of the samples are either coated with a thin layer of carbon for conductivity of the sample or already contain carbon, due to contamination or as a part of the material. Because of that, the alignment process is straightforward, robust and could be done on almost any sample.
Optical design parameters of the new 14 channel spectrometer are shared with the older 17 channel spectrometer. These parameters are the incidence grazing angle α = 2°, diffraction angle of the −1st order β = 0.918°, distance from the sample to the zone plate R’1 = 70 mm, distance from the zone plate to the CCD R’2 = 180 mm and the total optical path ltotal = 250 mm.
4. Alignment and calibration
As with every WD spectrometer, the alignment procedure is a crucial step in connecting the count rate at a certain angle (i.e. pixel) with a correct energy. Many of the WD spectrometers have an inherent disadvantage in the alignment – they are rather tedious to align and therefore limit the user. The RZP spectrometer has a distinct advantage. It reflects both 0th diffraction order, the specular reflex, and the −1st diffraction order onto the CCD. 0th diffraction order always reflects under the angle of 2° and contains an image of the substrate itself. Regardless of the element we want to examine, the position of the specular reflex on the detector stays the same. Only the position of the −1st diffraction order changes according to the incident energy. We can use this feature for the alignment. Parameters used for the alignment are shown in Fig. 1 .
Fig. 1 RAW CCD image with 0th and −1st diffraction orders marked. H0 indicates the calculated distance between the edge of the specular reflex (zone plate) and focus of −1st order.
- • An element with a previously known shape and energy of the emission spectrum is placed into the spectrometer. Using mechanical screws on the enclosure of the spectrometer (they allow for the change of horizontal angle), ideal, correct peak shape and calculated pixel distance H0 between −1st and 0th order on the CCD is obtained. The distance H0 is calculated from the geometrical parameters.
- • In this configuration, the pixel position of the 0th order on the CCD is stored and alignment screws on the enclosure are set. These screws do not have to be moved anymore, as the position of the 0th order is the same for any element.
- • The only parameter that has to be changed with subsequent measurements is the SEM sample table height. It is either lowered or raised until the 0th order position matches the calibrated value as shown in Fig. 1. Energy of the diffracted photons is determined from geometrical parameters (i.e. distance in pixels from 0th order reflection). Regardless of chemical shift or material, pixels are definitely connected to the energy.
Until the RZP optical element is removed, replaced or moved, no further alignment is necessary. This makes the RZP spectrometer very easy and robust to use.
In order to extract the spectra for each RZP from the CCD image the following steps are performed. Firstly, signal for each channel in the diffraction direction is calculated by summing up the signal of three adjacent pixels in the direction perpendicular to it. Then the energy calibration is done as mentioned above. The raw signal which is measured with the CCD camera is normalized to the channel width (in eV) for each pixel in the diffraction direction.
For quantitative analysis the following energy dependent coefficient has been used:
- • The photon energy conversion coefficient
- • The quantum efficiency of the CCD camera
- • The sagittal opening angle of the RZP
- • The relative reflectivity of the RZP
The quantitative analysis is not the subject of this paper and will be published separately.
5. Experimental result
In order to characterize the spectrometer and show that it can indeed detect energies as low as B Kα and Li Kα emission line, the following samples with their corresponding X-ray emission lines were measured: thin layers of B4C (B Kα) and metallic Li foil. Preparation, handling and composition of the samples are described in detail below.
5.1 Boron carbide (B4C)
In order to verify the limit of detection (LOD), a series of samples with varying thin layer thickness of boron carbide (B4C) was measured. The series was obtained from the German national metrology institute (PTB).
Investigated samples have the following composition. Thin layers of B4C are deposited on a Si substrate in the following thicknesses: 1 nm, 3 nm, 10 nm, 20 nm and 50 nm. Both layers are covered with a 2.5 nm thin layer of SiO2 that protects the embedded layers from contamination. Figure 2 shows the cross section of the samples. Because of the quality and smoothness of the samples, as well as the consistency between them, the limit of detection in mass units could be calculated.
The 3rd channel of the 17 channel spectrometer is designed for the energy of B Kα emission at E = 183 eV. With high efficiency of 18%, it is possible to measure small amounts of boron in very reasonable time. B4C layers with a thickness of 50 nm and 20 nm could be measured within t = 600 s and those with a thickness of 10 nm, 3 nm and 1 nm within t = 1200 s. Because of consistency, all of the samples were measured for t = 1200 s. Current on the sample was I 0 = 31 nA. Above that significant sample damage could be observed. Accelerating voltage was U 0 = 5 kV.
The higher order line of silicon Kα was detected at the energy of 184 eV (10th order) which is overlapping with B Kα emission at 183 eV. An additional sample of pure Si was measured at the same conditions. The measured pure Si spectrum was used for subtraction of Si background in the B spectrum.
As can be seen in Fig. 3 , the peak is aligned with a theoretical value of E = 183 eV and the thinnest detectable layer of B4C is 3 nm. Since the layers have a very well defined and small thickness, the limit of detection in mass units can be calculated. The electron beam starts to diverge immediately upon entering the sample, but at such small depth (l = 3 nm), it can be very well approximated with having a round footprint. The spot size was estimated from a spatial resolution measurement and is equal to Δx = 350.2 nm ± 53.8 nm. The value also corresponds well with an empirical formula [7]:
where E0 is the incoming beam energy, Ec is the shell excitation energy, both in keV and ρ is the density of the sample in g/cm3. This yields R ~0.4 µm for the used parameters.Fig. 3 Full B4C spectra with subtracted Si background and normalized data for different B4C layer thicknesses.
Because the layers are so thin, the interaction volume of interest can be approximated by a cylinder of volume V, bound by the values l and Δx:
The number of B4C molecules in the volume V in m3 can be calculated by using the following equation:
where N A = 6.022∙1023 is the Avogadro constant, ρ = 2.52 g/cm3, the density of B4C and M B4C = 55.255 g/mol, the molecular weight of B4C. Each molecule has 4 B atoms, therefore making the total B atoms number equal to: N B = 4N.It is straightforward to calculate the mass of a number of atoms. B has an atomic weight of M B = 10.81 g/mol. The limit of detection in mass units is:
By applying the error of the resolution measurement (~15%) and ~10% uncertainty in density, we finally obtain:5.2 Aluminium
In order to measure the energy resolution of the spectrometer in this energy range, an Al L2,3 band spectrum was measured and analysed. The 10th channel on the 14 channel spectrometer is specifically designed for the energy of the line with a peak energy at E = 72.5 eV. SEM parameters were the following: U 0 = 5 kV, I 0 = 115 nA and t = 600 s.
The measured spectrum can be seen in Fig. 4 . Theoretically, the spectrum should show a perfectly vertically shaped high energy Fermi edge and a broad low energy Auger tail. The spectrometer resolution can be read from the finite broadening of the edge. Measured resolution was estimated as a full width half maximum (FWHM) of the derivative and is equal to ΔE = 0.5 eV ± 0.2 eV. The error of the measurement was estimated to be ± 1 pixel on the CCD, which translates to the energy difference of 0.2 eV.
Fig. 4 Al L spectrum, showing the Fermi edge (left). The resolution is estimated based on the FWHM of the derivative.
5.3 Metallic lithium
Li is much less volatile as the rest of the elements in the first group, however it is still reactive and its exposure to air has to be minimized. It oxidises very quickly and already a very thin layer of LiO on the surface will absorb the low-energy photons associated with Li Kα emission. To avoid this, a Li foil of 600 μm thickness was cut and fitted onto a standard SEM sample holder inside a glove box, filled with Ar. It was then transported to the laboratory with an airtight container. To avoid prolonged exposure to air, the sample was fitted into the SEM chamber and the air evacuated as quickly as possible. As can be seen in Fig. 5 , only a small area of the sample reacted with the air (lighter areas), while the majority of the sample was still intact (darker areas).
Fig. 5 Metallic lithium sample. Lighter color shows oxidized areas and darker color shows pure Li metal.
Because of its specific electron arrangement, lithium is particularly interesting for X-ray spectroscopy. By having one valence electron and two core electrons, it is the first and the simplest arrangement that allows for a K transition.
The 5th channel of the 14 channel RZP spectrometer is designed specifically for the energy of Li Kα emission with E = 54.3 eV. The samples were prepared as described above. Up to the maximum probe current I 0 = 115 nA, no sample damage was observed. Some oxidation was visible before the measurement (Fig. 5). In order to monitor the consistency of the measurements, absorption current was observed. If the position of the sample was kept constant, the absorption current decreased and some C and O contamination was visible. To compensate for this, the position of the beam was carefully changed during the measurement in order to keep the irradiated spot as clean as possible. Measuring time was t = 600 s, accelerating voltage U 0 = 5 kV and pressure in the order of ~10−6 mbar.
Measured spectrum can be seen in Fig. 6 . Our measurement shows a peak position at E = 54.2 eV ± 0.2 eV and the Fermi edge width calculated as FWHM of the derivation of the spectrum to be 0.66 eV. Reported values from previous measurements show a very good match in the peak position, but a small discrepancy in the Fermi edge width, more specifically: 0.55 eV was measured in [8,9 ] and 0.30 eV in [10].
Even though the numbers differ to some extent and cannot be accounted to the systematic error of the measurement, there is a large difference in the parameters of the experimental setups, for example a comparison of the incoming electron energy of U 0 = 5 kV in our case and a U 0 on the order of U 0 ~200 V in case of older measurements, which is connected to the probability for inner shell ionization.
6. Conclusion
We have demonstrated that a previously proposed novel parallel X-ray emission wavelength-dispersive reflection zone plate spectrometer can be successfully used also in the XUV energy range. The whole setup is as robust as conventional ED spectrometers and as accurate as existing WD spectrometers. Demonstrated alignment procedure, described in this article, is short, convenient and consistent. It yields reproducible results with exact matching of pixels to energy. Resolution of the spectrometer is comparable to other XUV WD spectrometers [2] and acquisition time is in the same order of magnitude as with ED spectrometers.
We propose it not as a replacement for conventional ED spectrometers, which still offer very easy operation in combination with mature software, but rather as a complementary tool for the analysis of different applications, where ED spectrometers are inferior, such as chemical shifts, peaks that lie close to one another, samples with trace elements, and detecting energies below ~100 eV.
There are still a number of improvements to be tackled, such as the traditional problem associated with WD spectrometers, high-order diffraction. With our simple model elements, there is hardly any high order diffraction and if it exists, we can subtract the signal of the pure element. With complex compounds, it would be rather hard to extract different signals. There is a possible and promising solution to this problem. For example, a correct choice of coating can reject high-order diffraction of higher energies.
The new device could enable scientists with expertise in spectroscopy to address an energy range which was previously harder to reach. Because of the versatility of the instrument, this could also open up new opportunities to investigate Li as the first and therefore a model element which emits X-rays upon excitation. Different Li compounds could also be investigated. They are of interest because they exhibit a proportionally large chemical shift (~10%) in Li Kα emission.
Acknowledgments
The authors acknowledge support by ZIM-SOLO financed by BMWi. This work was also financially supported by the BMBF project No. 05K12CB4 and Marie Curie FP7-Reintegration-Grants within the 7th European Community Framework Program (project no. PCIG10-GA-2011-297905), fellow Dr. Maria Brzhezinskaya. We are grateful to Robert Kaußmann and Andreas Günther from IfG Scientific Instruments GmbH for their technical support in the experiment and Ms Analía Fernandez-Herrero for the efficiency calculations. We are also grateful to Neutron Optic Berlin GmbH for providing the SEM Zeiss DSM 942 for the measurements and Dr. Burkhard Beckhoff and Dr. Rainer Unterumsberger for providing B4C test samples.
References and links
1. M. Terauchi, H. Takahashi, N. Handa, T. Murano, M. Koike, T. Kawachi, T. Imazono, M. Koeda, T. Nagano, H. Sasai, Y. Oue, Z. Yonezawa, and S. Kuramoto, “A new WDS spectrometer for valence electron spectroscopy based on electron microscopy,” JEOL News 47(1), 23–28 (2012).
2. H. Takahashi, P. McSwiggen, and C. Nielsen, “A unique wavelength-dispersive soft X-ray emission spectrometer for electron probe X-ray microanalyzers,” Microsc. Anal. (Am. Ed.) 28(7), 4–7 (2014).
3. T. Wilhein, D. Hambach, B. Niemann, M. Berglund, L. Rymell, and H. M. Hertz, “Off-axis reflection zone plate for quantitative soft x-ray source characterization,” Appl. Phys. Lett. 71(2), 190–192 (1997). [CrossRef]
4. A. Erko, A. Firsov, and F. Senf, “Novel parallel vacuum ultra-violet/X-ray fluorescence spectrometer,” Spectrochim. Acta B At. Spectrosc. 67, 57–63 (2012). [CrossRef]
5. A. Erko, A. Firsov, R. Gubzhokov, A. Bjeoumikhov, A. Günther, N. Langhoff, M. Bretschneider, Y. Höhn, and R. Wedell, “New parallel wavelength-dispersive spectrometer based on scanning electron microscope,” Opt. Express 22(14), 16897–16902 (2014). [CrossRef] [PubMed]
6. A. A. Sokolov, F. Eggenstein, A. Erko, R. Follath, S. Künstner, M. Mast, J.-S. Schmidt, F. Senf, F. Siewert, T. Zeschke, and F. Schäfers, “An XUV optics beamline at BESSY II,” Proc. SPIE 9206, 92060J (2014). [CrossRef]
7. J. John, Friel, X-Ray and Image Analysis in Electron Microscopy 2nd ed. (Princeton Gamma-Tech, 2005), p. 26.
8. J. A. Catterall and J. Trotter, “Soft x-ray emission spectra from lithium and lithium-magnesium alloys,” Philos. Mag. 4(46), 1164–1170 (1959). [CrossRef]
9. D. E. Bedo and D. H. Tomboulian, “K-emission spectrum of metallic lithium,” Phys. Rev. 109(1), 35–40 (1958). [CrossRef]
10. R. S. Crisp and S. E. Williams, “The K emission spectrum of metallic lithium,” Philos. Mag. 5(53), 525–527 (1960). [CrossRef]
