I. Božičević
Mihalić
*,
S.
Fazinić
*,
T.
Tadić
,
D.
Cosic
and
M.
Jakšić
Ruđer Bošković Institut, Bijenička Cesta 54, HR-10000 Zagreb, Croatia. E-mail: ibozicev@irb.hr; stjepko.fazinic@irb.hr
First published on 16th September 2016
A downsized wavelength dispersive X-ray spectrometer, employing a flat crystal and a CCD detector for use with focused ion beams has been constructed and employed to study ion beam induced chemical effects in Si K X-ray spectra from silicon and its selected compounds. By using ADP, PET and LiF(110) diffraction crystals, the spectrometer can measure X-rays in the energy range between 1.2 and 8.4 keV, with the energy resolution E/ΔE(FWHM) = 1850 on Al Kα and 1580 on Si Kα achieved with 2 MeV protons. This energy resolution enables the study of secondary effects in the K X-ray spectra of light elements, L-shell spectra of medium Z elements and M-shell spectra from heavy elements. The K X-ray spectra of silicon and selected silicon compounds were measured after excitation with 2 MeV protons and 20 MeV carbon ions focused to micrometer size. The results obtained for peak relative intensities were analyzed to study their dependence on the silicon oxidation states and effective charge on Si. The results were compared with the existing data obtained by proton, electron and photon induced ionization mechanisms, and clear differences between the ionization sources were observed. Si Kα multiple ionization satellites were studied with 20 MeV carbon ions on Si, SiO2 and SiC. The variation of apparent average L vacancy fraction pLx with effective charge on Si was studied. It has been shown that, for ionization by 20 MeV carbon ions, the relative KαL2 intensity is more sensitive to the chemical environment compared to pLx values.
A major breakthrough for high resolution PIXE (HR-PIXE) was obtained in the 1960s when flow proportional counters replaced photographic detection enabling direct comparison of energies, intensities and shapes of the lines, especially in the case of low-intensity satellites.3,4 An important limitation of old WD systems was that step by step scanning detectors were used, requiring stable beam parameters during experiments which was not always easy to achieve, having in mind that many experiments required long measuring times. With the development of position sensitive detectors, like position sensitive proportional counters (PSPC), and recent advances in high performance scientific charge couple devices (CCD) with small pixel sizes, a fertile ground for the wider use of high resolution systems was set. Spectrometers range from commercial ones on electron microprobes5 to specially designed systems for dedicated applications,6–8 including chemical effect studies on X-ray spectra and studies of secondary effects in X-ray emission of importance for the definition of fundamental parameters in ion–solid interactions.
During the last decade at the Ruđer Bošković Institute (IRB) Accelerator Facility we studied chemical effects on the Kβ X-ray spectra of 3d transition metals and their compounds.9–12 For these studies we used 2 or 3 MeV broad proton beams collimated to 5 × 1 mm size from the accelerator as an excitation source and a WD X-ray detection system based on a flat LiF(110) crystal and a PSPC X-ray detector. Due to low efficiency, typical experiments required ion beam currents higher than 100 nA on targets. While the targets were in a vacuum, all the other components were in a He balloon to reduce X-ray attenuation on the way to the detector (with the sample to detector distance of about 60 cm).
PIXE ionization cross-sections for the K shell are much larger for low Z elements (Na to Cl) compared to 3d transition metals (Ti to Cu). Therefore, we planned to extend our studies of K X-ray spectra to low Z elements. Such studies are important for two reasons: (i) to exploit the possibilities for the use of chemical effects for chemical speciation, and (ii) to evaluate fundamental parameters like single and multiple ionization probabilities for ion incidence. These issues have been recognized by the International Initiative on X-ray Fundamental Parameters supported by the EXSA (European X-ray Spectrometry Association)13 as very important. For example, the influence of multiple ionization satellites and chemical effects on the PIXE analysis of light elements (Mg, Al, Si) would be important for the exploration of Mars with the APXS (Alpha particle X-Ray Spectrometer) installed on Mars Exploration Rovers.14 Due to the low energy of light element X-rays (≈1–3 keV), their measurement would require that all the spectrometer components are in a vacuum. Instead of redesigning the existing large broad beam spectrometer, we have used an alternative approach to construct a downsized spectrometer that would be used in combination with the focused ion beams to ensure good energy resolution. In such a way we could combine the downsized spectrometer for applications on microscopic samples utilizing micrometer beam size available at our ion microprobe.
Several downsized spectrometers built in the past were used with broad ion beams in combination with sets of collimators to enable achieving acceptable X-ray energy resolution.1,15,16 Their design however required sequential measurements. This approach requires very precise measurements of ion dose on a sample during the acquisition of spectra, which is often not an easy task, and can easily introduce artificial features in the measured spectra due to uncertainties in the dose measurements. Therefore, we decided to use a CCD X-ray detector with good spatial resolution to enable measurements of useful spectra with good energy resolution and an X-ray energy window in a single run. In the past, few groups installed on ion microprobes commercially available WD systems originally designed for electron microprobes but these systems used either flow proportional counters (requiring sequential measurements) or PSPC detectors17–19 and were not easily adaptable for the use with ion beams. A relatively large plane crystal spectrometer with a PSPC on a heavy ion microprobe was built by Mokuno et al. for chemical speciation studies2,20–23 demonstrating the possibility of high energy resolution PIXE mapping.21
In our system, a plane diffraction crystal geometry has been chosen due to its design simplicity and smallest sensitivity to sample tilt misalignments. We assumed that eventually the spectrometer could be employed for the chemical speciation of hot-spots on structured (non-homogeneous) micro-samples with a lateral resolution in the order of magnitude of 10 μm. We have focused our efforts to optimise the system to work with ion beams focused to about 10 μm or below in the X-ray energy range that corresponds to K X-rays of light elements from Al to Cl (although even higher energies can be measured). A Peltier cooled CCD X-ray detector with good spatial resolution, a flat analysing crystal with a simple linear translator and a sample holder have been enclosed in a specifically designed vacuum chamber. In this work we will describe the developed system and its performances. Capabilities for chemical effect studies will be demonstrated on the K X-ray spectra of silicon and selected silicon compounds induced by 2 MeV protons and 20 MeV carbon ions.
Beside the source size and the analysing crystal rocking curve, the spectrometer's energy resolution depends also on the CCD spatial resolution. Although in principle the reduction of the pixel size would increase the CCD spatial resolution, degradation occurs as a result of charge cloud formation and detection of one X-ray photon in more than one CCD pixel. For back illuminated CCD (BI CCD) chips, X-rays enter through the backside, leaving their energy mostly in the field free region. Therefore, induced charge has to diffuse to reach the potential wells defining each pixel. As a consequence, charge can be deposited either in a single pixel or it can be spread to neighbouring pixels causing a split effect. Experiments aimed to measure the charge effect on the 13.5 μm × 13.5 μm pixel size CCD showed that 90% of the intensity is concentrated in a radius of 18 μm, leading to spatial resolution which is double the size of the pixel.28 For the same pixel sized CCD, Ghiringhelli et al. obtained charge spatial distribution with a FWHM of 24 μm, with tails of distribution spread to the area extending to more than 20 μm in radius.7 The ratio of single pixel to split events depends on the energy of detected X-rays, size of the field free region of the CCD and its pixel size. Szlachetko et al. showed that for their BI CCD with a pixel size of 20 × 20 μm2, the ratio of single pixel to split events in the case of the silicon K line is around 40%, falling below 1% above the Si K absorption edge.29 They also noticed that a major part of the charge was shared between four neighbouring pixels. Most of the algorithms for processing X-ray signals detected in CCDs with a small pixel area accept events falling into a 2 × 2 pixel area, discarding those spread out to more pixels. In the case of our CCD, 19% of the events fall to a single pixel for silicon K lines (Fig. 2). For S K lines, less than 1% are single pixel events while 70% spread to more than 2 × 2 pixels and are rejected in standard algorithms. Based on these observations, we developed our image processing procedure which takes into account all events that spread to 3 × 3 pixels and fall into the energy window defined around the average charge of single pixel events.
Fig. 2 Histogram of CCD charge distribution for Si K X-rays, single pixel and events with charge spreading to multiple pixels are marked. |
The procedure for event detection is similar to the one described by Szlachetko et al.29 From each signal image the mean background image (average of several images without signals – dark images) is subtracted and is followed by the image processing algorithm. Histograms of charge distribution are created, the average charge of single pixel hits is determined from the centre of the peak in Fig. 2 and the energy window is set around it. For split events, the charge spread into a 3 × 3 pixel area is summed and if the total sum falls into the defined energy window, value 1 is assigned to the pixel with maximal partial charge while neighbouring pixels are given value 0. For the signal processing code, it is crucial to adjust the exposure time of single image frames to avoid pile-up events that can lead to the total charge falling outside of the energy window. Each processed image is summed together to get a final frame (Fig. 3(a)). The origin of the curvature in Fig. 3(a) is in X-ray diffraction geometric effects due to the finite and relatively large height of the crystal. Due to the fact that X-ray lines in the image are curved, curvature correction on the final frame is performed. The image is cut into 40 horizontal slices. In each slice X-ray lines are fitted with one to multiple Gauss functions depending on the number of lines detected. The dependence of centroids of the Gauss functions in each slice (in most cases the Kα1,2 line is the narrowest line) on the y position in the image is determined and fitted with a polynomial function. Shifts are calculated from the difference of the centroid in an individual slice and in the central part of the image. After each slice is shifted for the value of the calculated difference, the final image corrected for curvature is obtained (Fig. 3(b)). X-ray events are projected to the horizontal axis to extract the final spectrum.
All the spectra were fitted with Voigt functions, i.e., convolution of Lorentzian and Gaussian functions describing natural linewidth and instrumental broadening, respectively. A linear background was assumed and subtracted. The final errors in the obtained peak areas were taken from the error estimates of the non-linear least-squares fit parameters and empirical statistical error, which was simply estimated as the square root of the number of counts in the peak. The two errors were added in quadrature. Mg and S diagram lines were fitted with one Voigt and Si and Al with two Voigt functions. The FWHM obtained is 0.8 eV for Kα1,2 of Mg and Kα1 of Al, 1.1 eV for Kα1 of Si and 2.8 eV for Kα1,2 of S. The best energy resolution of E/ΔE = 1850 was achieved in the case of Al, while for the S region it was 820. ΔE is based on the experimental spectra and includes natural linewidth and instrumental broadening. In all the cases, the excellent resolution enables the possibility not only to resolve the KαL1 region from related diagram lines but also to distinguish three components within the KαL1 region: Kα′, Kα3 and Kα4(see Fig. 6). The maximum efficiency was obtained for the Mg target where 220 cps are recorded in the Kα1,2 peak, while in the case of the Al target it was 70 cps. For silicon and sulphur targets measured with the PET crystal, the sensitivity was 151 and 135 cps in diagram peaks, respectively.
Fig. 5 High resolution Kα and Kβ spectrum of silicon induced with 2 MeV protons. The inserted picture shows an enlarged Si Kβ region. Dots are measured and lines fitted the spectrum. |
Fig. 5 shows the final high resolution Kα and Kβ spectrum of silicon. The inserted picture shows an enlarged Si Kβ region. Linear energy calibration was done using two reference lines: Kα1 1739.98 eV and Kβ 1835.9 eV.49 Due to the closeness of the target and the detector, even small misalignments in target positioning can result in a small line shift in the final spectra that can be misplaced and attributed to chemical shifts. Therefore, we decided to compare only relative distances between X-ray lines of different compounds. Kα1 lines in the spectra related to the measured Si compounds have been all aligned to the position of the Si target spectrum. Fig. 6 shows the Si Kα region with the most intense diagram Kα1,2 line and the inserted picture of the KαL1 region, where three satellite lines are resolved. Fitting of the diagram lines was performed with two Voigt functions and the satellite region with 3: Kα′, Kα3, Kα4. The full width at half maximum (FWHM) of the Si Kα1 profile was 1.1 eV which was good enough to resolve Kα3 and Kα4 satellites which are 2.3 eV apart. The relative intensities of KαL1 components for silicon and its compounds: Ir(Kα′), Ir(Kα3) and Ir(Kα4) are listed in Table 1. Also ratios of Kα3 and Kα4 lines together with ratios of total intensity of KαL1 relative to KαL0 are shown. The obtained values for silicon are compared with the ones from Kavčič who also used excitation with 1, 2 and 3 MeV protons.47 I(Kα3)/I(Kα4) and I(KαL1)/I(KαL0) vary for different silicon compounds and have potential to be used for chemical speciation. From measurements on three different proton energies on pure silicon, it was noticed that the intensity ratio of KαL1 relative to KαL0 decreases from 18.6% to 8.5% while increasing the proton energy from 1 to 3 MeV.47 The same trend is observed in I(Kα3)/I(Kα4) but the change in ratio is less than 10%.
Si | Si47 | SiO2 | SiC | Na2SiO3 | |
---|---|---|---|---|---|
Ir(Kα′) (%) | 4 ± 1 | 3.6 ± 0.3 | 5 ± 1 | 5 ± 1 | 4 ± 2 |
Ir(Kα3) (%) | 71 ± 3 | 67.1 ± 0.9 | 63 ± 3 | 65 ± 2 | 60 ± 6 |
Ir(Kα4) (%) | 25 ± 2 | 29.3 ± 0.8 | 32 ± 2 | 30 ± 2 | 36 ± 5 |
I(Kα3)/I(Kα4) | 2.8 ± 0.2 | 2.29 ± 0.07 | 2.0 ± 0.2 | 2.2 ± 0.1 | 1.6 ± 0.2 |
I(KαL1)/I(KαL0) (%) | 12.8 ± 0.5 | 13.2 ± 0.7 | 14.5 ± 0.7 | 14.7 ± 0.4 | 19 ± 1 |
Fig. 7 shows the Kα satellite region for Si, SiO2, SiC and Na2SiO3. Variation in Kα3 and Kα4 intensities for different Si compounds is clearly visible. Previous studies reported that chemical shifts in Kα lines are directly related to effective charges on the Si atom.33,34 An increase of the effective number of electrons in the valence shell of the central atom in the presence of electronegative neighbouring atoms will cause stronger bonding of the inner shell electrons, shifting the Kα1,2 line to higher energies. To test if the intensity ratio of Kα3 and Kα4 lines and the ratio of KαL1 to KαL0 are sensitive to the chemical environment, variation of these intensity ratios with effective charge on Si is shown in Fig. 8 and 9. Values for effective charges on Si in SiC and SiO2 were obtained from ref. 33 where they were calculated with the DV-Xα molecular orbital method and compared to other calculations.34 For pure silicon, zero effective charge was assumed. The values of 1.83 and 0.59 were taken for SiO2 and SiC. Our experimental values are compared to the reported values obtained using photons,41–43,50 electrons4,41,52 and energetic protons as excitation sources.47 A clear offset between the curves is seen between different excitation methods, where the highest Kα3 to Kα4 intensity ratio is for proton excitation and the lowest one for electrons. Also, the highest KαL1/KαL0 intensity ratios are for proton excitations.
Fig. 8 Variation of Kα3 and Kα4 intensity ratios with effective charge on Si. Colours represent different excitation methods: black squares for protons (filled squares: present work, vertically filled squares47), blue triangles for electrons (filled triangles4, vertically filled triangles51, horizontally filled52) and red circles for photons (filled50, vertically filled41). |
Fig. 9 Variation of KαL1 and KαL0 intensity ratios with effective charge on Si. Colours represent different excitation methods: black squares for protons (filled squares: present work, horizontally filled squares47), blue triangles for electrons (filled triangles4, vertically filled triangles51, horizontally filled52) and red circles for photons (filled41, vertically filled50, horizontally filled42,43). |
(1) |
(2) |
(3) |
Interpretation of pLx is not straightforward since the L vacancy distribution differs in the moment of the X-ray emission from the distribution after heavy ion collision. To extract the true value of L vacancy distribution in the moment of a heavy ion collision, a few corrections are needed: (i) correction for vacancy rearrangement where L shell holes created after ion collision can be filled from outer shells and (ii) fluorescence yield corrections due to the fact that its values differ for different KαLi satellites. However, pLx values represent a good parameter for comparison since they are obtained directly from measurements with no need to use model based corrections.60 Higher pLx values imply a shift of intensity distribution to an increased number of L shell vacancies and vice versa. Detailed studies with different projectile-target combinations can reveal the origin of L shell vacancy production and processes of L shell rearrangements after the collision. Some studies related to 2d and 3d elements show the variation of relative intensities and energy shifts of KαLi satellites between different compounds. Explanation for this different distribution lies in the L vacancy transfer from intra-atomic transitions from higher shells of the same atom and interatomic transitions from neighbouring atoms following heavy ion collisions.16 The decrease of pLx values with the increase of average valence electron density for several aluminium, silicon, sulphur and chlorine compounds,16,61 supported by additional measurements on gaseous samples,46 further revealed the importance of chemical bonding in the transfer of vacancies from valence levels of ligand atoms. In atoms with lower effective charges, valence electrons are more bound to the central atom. In collision with heavy ions there is a greater probability that they are ejected compared to atoms with a higher effective charge, where most of the valence electrons are shifted toward neighbouring atoms. So for higher effective charge compounds more valence electrons are at a disposal to transfer to L shell vacancies created after ion collision. This rearrangement of L shell holes leads to the reduction of L shell holes shifting the relative intensity distribution to lower pLx values.46,62
The goal of our experiment was to test the capability of our spectrometer for chemical speciation with heavy ions on silicon compounds. Measurements were performed with 20 MeV (i.e. 1.67 MeV/amu) carbon ions from a Tandem Van der Graaff accelerator. The spectrometer setup was the same as in the experiment with 2 MeV protons. The ion beam current on the target was around 1 nA. Fig. 11 shows the high resolution Kα X-ray spectra of Si in pure Si, SiO2 and SiC. For the ease of comparison, the compound spectra are normalized to the total intensity of KαL0 lines of pure silicon, with aligned positions of the KαL0 line. Besides the ability to separate each KαLi group of satellites, individual components within a specific KαLi group are also resolved, showing the fine structure of the Kα satellite region. The obtained resolution of 1.7 eV FWHM for KαL0 in pure silicon is much better than 7 eV that was reported in some of the earlier studies on chemical speciation with heavy ions using broad ion beams.16 This resolution allows for the fine structure to be resolved in individual multiple ionization Si Kα X-ray satellite lines. As can be seen from Fig. 11, the fine structures of KαL1, KαL2 and KαL3 components show clear changes in relative intensities between Si, SiO2 and SiC. Table 2 shows relative satellite intensities (Ii/Itot) resulting from fitting of the spectra following the same procedure used to fit the spectra obtained with proton excitation. The intensity ratios are corrected for absorption in foils on the target and on the detector.
Compound | I i /Itot | p L x 20 MeV C (this work) | p L x 22 MeV C16 | p L x 32.4 MeV O16 | |||||
---|---|---|---|---|---|---|---|---|---|
i = 0 | i = 1 | i = 2 | i = 3 | i = 4 | i = 5 | ||||
Si | 0.047 | 0.194 | 0.339 | 0.258 | 0.130 | 0.032 | 0.291±0.005 | 0.267 | 0.348 |
SiO2 | 0.046 | 0.193 | 0.378 | 0.261 | 0.100 | 0.023 | 0.280±0.006 | 0.258 | 0.317 |
SiC | 0.046 | 0.201 | 0.361 | 0.254 | 0.103 | 0.035 | 0.284±0.005 | 0.331 | |
Si3N4 | 0.322 |
p L x values are calculated according to formula (1) and compared to the values obtained by Watson et al.16 They used 22 MeV (i.e. 1.83 MeV/amu) C ions, 5.4 MeV (i.e. 1.36 MeV/amu) He ions and 32.4 MeV (i.e. 2 MeV/amu) O ions. pLx values increase, i.e. L-shell vacancy distribution shifts to higher multiple ionization states, going from lower to higher Z ions (Fig. 12). This is in agreement with previously obtained results, where pLx values increase with the projectile atomic number until the saturation point is reached when the atomic number of the projectile is close to the one of the target.15
Fig. 12 Relative intensity distribution of KαLi satellites for pure silicon with 5.42 MeV He,16 20 MeV C (this work), 22 MeV C,16 and 32.4 MeV O ions.16 |
Fig. 13 shows the variation of pLx values with the effective charge on the Si atom. A decrease of pLx with increasing effective charge on silicon can be observed, although the effect is not so pronounced and variation between the values is small for C ions with an energy of about 2 MeV/amu. For He ions only KαL1 and KαL2 satellites are observed and the variation of pLx between different Si compounds is within experimental error, making pLx values insensitive to chemical speciation. Contrary to He and C ions, the variation of pLx values for 32.4 MeV O ions among different silicon compounds is much higher, attributed to their higher pLx. We can conclude that for silicon chemical speciation studies based on pLx values in the range of energies around 2 MeV/amu, ions heavier than carbon, i.e. those with the atomic number close to silicon should be selected.
Fig. 13 Dependence of pLx values with effective charge on silicon excited with 20 MeV C (this work), 22 MeV C,16 and 32.4 MeV O.16 |
Fig. 11 suggests that relative KαL1 and KαL2 relative intensities have larger differences between different compounds. In their study of sulphur compounds with Ar and Kr ions, Vane et al.63 found a correlation between the relative KαL2 intensity and effective charge on the S atom and the relative intensity of KαL5 with the average valence-electron density of S compounds. Fig. 14 shows the variation of relative intensity I2/Itot with effective charge on the Si atom obtained from our spectra from silicon and its selected compounds. The figure also shows the values obtained by Watson et al. for 20 MeV C and 32.4 MeV O ions.16 It seems that for excitation with 20 MeV carbon ions, relative KαL2 intensities show larger variations with the effective charge on the Si atom than the variation of pLx values do.
Fig. 14 Variation of relative I2/Itot intensity with effective charge on the Si atom for 20 MeV C (this work), 22 MeV C and 32.4 MeV O.16 |
In this work we employed the spectrometer to study ion beam induced chemical effects in Si K X-ray spectra from silicon and some selected compounds induced by focused 2 MeV protons and 20 MeV carbon ions. We particularly analysed the intensity ratios of silicon Kα3 to Kα4 satellite lines and of KαL1 satellites to KαL0 diagram lines induced by 2 MeV protons. Both ratios showed variations with effective charge on silicon. A clear distinction between different excitation methods was observed for both intensity ratios with the highest I(KαL1)/(KαL0) for proton excitation as a direct signature of higher multiple ionization probability for protons compared to electron and photon excitation.
It is known that interaction with heavy MeV energy ions considerably increases the probability for multiple ionizations. The obtained resolution of 1.7 eV FWHM at Si Kα1,2 lines induced with 20 MeV carbon ions on Si, SiO2 and SiC enabled the fine structure to be resolved into multiple ionization Si Kα X-ray satellite components, which showed clear differences for KαL1, KαL2 and KαL3 bands. Apparent average L vacancy fractions calculated from the measured relative intensities of multiple ionization satellites showed small variations with effective charge on silicon, which is in agreement with the trend observed by other authors. A much stronger variation with effective charge and greater chemical sensitivity was seen in relative KαL2 intensities.
This journal is © The Royal Society of Chemistry 2016 |