Anal. Chem. 2000, 72, 2613-2617
Trace Chemical Characterization Using Monochromatic X-ray Undulator Radiation Hiromi Eba,† Chiya Numako,‡ Junji Iihara,§ and Kenji Sakurai*,†
National Research Institute for Metals, 1-2-1, Sengen, Tsukuba, Ibaraki 305-0047, Japan, University of Tokushima, 1-1, Minami-Josanjima, Tokushima 770-8502, Japan, and Sumitomo Electric Industries Company Ltd., 1-1-1, Koya-kita, Itami, Hyogo 664-0016, Japan
An efficient Johansson-type X-ray fluorescence spectrometer has been developed for advanced X-ray spectroscopic analysis with third-generation synchrotron radiation. Kr and Kβ X-ray fluorescence spectra for trace metals have been collected by a Ge(220) analyzing crystal with a Rowland radius of 150 mm, under monochromatic X-ray excitation at the undulator beamline at the SPring-8. The energy resolution is ∼10 eV for most of the K lines for 3d transition metals. In light of the greatly improved efficiency, as well as the excellent signal-to-background ratio, the relative and absolute detection limits achieved are 1 ppm and 1.2 ng of copper in a carbon matrix, respectively. The energy resolution of the present spectrometer permits the observation of some chemical effects in Kβ spectra. It has been demonstrated that the changes in Kβ5 and Kβ′′ intensity for iron and cobalt compounds can be used for the analysis of chemical states. Resonant X-ray fluorescent spectra are another important application of monochromatic excitation. In view of trace chemical characterization, the present spectrometer can be a good alternative to a conventional Si(Li) detector system when combined with highly brilliant X-rays. Third-generation synchrotron radiation, especially an X-ray source using undulators as insertion devices, is extremely attractive when it comes to X-ray spectroscopic analysis, because it provides very brilliant X-ray photons and its intensity at the sample position is 103-104 times greater than that of conventional synchrotron radiation from bending magnets.1 In the field of X-ray fluorescence (XRF), chemical characterization of trace elements localized in a small area or on the surface of materials has huge potential in analytical applications. A Si(Li) detector has been employed for most synchrotron XRF experiments so far,2,3 because the wavelength-dispersive system,4 which is usually used with an ordinary laboratory X-ray source, has fairly low detection †
National Research Institute for Metals. University of Tokushima. § Sumitomo Electric Industries Co. Ltd.. (1) Kamitsubo, H. J. Synchrotron Rad. 1998, 5, 162-167. (2) Saisho, H.; Gohshi, Y. Application of Synchrotron Radiation to Materials Analysis; Elsevier: Amsterdam, 1996. (3) Iida, A.; Gohshi, Y. In Handbook on Synchrotron Radiation; Ebashi, S., Koch, M., Rubenstein, E., Eds.; North-Holland: Amsterdam, 1991; Vol. 4, Chapter 9, p 307. (4) Jenkins, R. An introduction to X-ray spectrometry; Heyden & Sons: London, 1974; Chapter 4, p 52. ‡
10.1021/ac991308f CCC: $19.00 Published on Web 04/26/2000
© 2000 American Chemical Society
efficiency, even though its high-energy resolution is of key importance. Some pioneering XRF experiments using a crystal analyzer have been performed with white5,6 or quasi-monochromatic7 excitation. This represents a compromise in order to gain XRF intensity from minor components. However, from the standpoints of variety and quality of analytical information obtained through the experiments, the use of tunably monochromatic X-rays is clearly desirable. The advent of a brilliant X-ray undulator source permits such experiments and opens up new opportunities for further advanced characterization of trace elements. In the previous work,8 the feasibility of XRF experiments with a Johansson-type spectrometer and monochromatic undulator X-rays was examined. The Rowland radius of the spectrometer was set at 350 mm at that time. Although the measurements were quite satisfactory, the results indicated that further enhancement of detection efficiency would be necessary in order to extend the experiments to trace systems in the order of ppm or lower. One promising way would be to make the spectrometer compact and to shorten the Rowland radius. In the present study, a more compact, efficient spectrometer was developed. XRF spectra for trace systems were measured with monochromatic X-ray excitation. Particular emphasis was placed on the chemical-state analysis of trace systems or small samples. The present paper describes the first experimental results obtained by means of the present high-performance spectrometer and monochromatic X-ray undulator source. EXPERIMENTAL SECTION The experiment was carried out at undulator beamline BL39XU at SPring-8, Harima, Japan. Details of the undulator source and the beamline are given elsewhere.9 The ring current during the experiments was 40-60 mA. Incident X-rays were monochromatized by a Si(111) rotated-inclined double-crystal monochromator, and higher-order harmonics were rejected with a single, flat platinum-coated mirror at a glancing angle of 5 mrad. In the present beamline, there were no focusing optics for the primary beam. The beam size was adjusted by a slit of 0.2 mm in width, (5) Rivers, M. Adv. X-ray Anal., in press. (6) Ohashi, K.; Takahashi, M.; Gohshi, Y.; Iida A.; Kishimoto, S. Adv. X-ray Anal. 1992, 35, 1027-1033. (7) Iida, A., Photon Factory, Institute of Materials Structure Science, 1-1 Oho, Tsukuba 305, Japan, 1994; personal communications. (8) Sakurai, K.; Eba, H. Jpn. J. Appl. Phys. 1999, (Suppl. 38-1), 650-653. (9) SPring-8 Beamline Handbook; Japan Synchrotron Radiation Research Institute, 1997; Version 1.1, p 20.
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000 2613
Table 1. Experimental Conditions for XRF Spectra Measurements
sample NIST SRM 610 (460 ppm Fe) NIES No. 8 (67.7 ppm Cu) Fe foil (5 µm in thickness) K3[Fe(CN)6] (15 µg) Co foil (4 µm in thickness) Co complexes (15 µg) Co powder (15 µg) Co3O4 (15 µg) K3[Fe(CN)6] (15 µg)
incident X-ray energy (keV) 8 10 7.13 7.13 10 10 7.695-7.739 (2-eV steps) 7.695-7.755 (2-eV steps) 7.084-7.150 (3-eV steps)
storage ring current spectro(mA) meter
XRF line
angular range (deg) 28.72-29.20 (0.003° step) 22.43-22.84 (0.002, 0.003° steps) 25.75-26.26 (0.003° step) 25.75-26.26 (0.003° step) 23.60-24.11 (0.003° step) 23.60-24.11 (0.003° step) 26.46-26.74 (0.0027° step)
59 56 53 53 51 48 63
B B B B B B A
Fe KR Cu KR Fe Kβ Fe Kβ Co Kβ Co Kβ Co KR
57
A
48
B
total meas time (min)
ref figure
160 175 170 170 170 170 105
20 20 1 10 1 10 4
53 58 3 28 3 28 (for each complex) 7 (for 1 XRF spectrum)
2 3 4 4 5 5 6
Co KR 26.46-26.74 (0.0027° step)
105
4
7 (for 1 XRF spectrum)
6
Fe KR 25.75-26.26 (0.006° step)
85
3
4 (for 1 XRF spectrum)
7
and the irradiation area of the sample was 0.2 (H) × 2 (V) mm2. Although generally the use of an extremely intense X-ray beam could give some damage to a sample, significant radiation effects were not observed in our case. Incident X-ray intensity was monitored by an ionization chamber. Some standard reference materials from NIST and NIES (National Institute for Environmental Science, Tsukuba, Japan) were measured in the form received. Pure metallic cobalt, tricobalt tetroxide (Co3O4), bis(acetylacetonato)cobalt(II) (Co(C5H7O2)2‚ 2H2O), tris(acetylacetonato)cobalt(III) (Co(C5H7O2)3), and potassium hexacyanoferrate(III) (K3[Fe(CN)6]) were pulverized and mixed homogeneously with cellulose powder, and then pellets (13 mm diameter) were prepared by pressing. A wavelength-dispersive XRF spectrometer usually has very low detection efficiency compared with a Si(Li) detector. Therefore, it is important to improve it as much as possible, especially when monochromatic X-rays rather than white or quasi-monochromatic X-rays are used for excitation. The present study employed focusing optics based on Johansson-type curved crystal. As reported in the previous work,8 our spectrometer has a flexible Rowland radius, with the change in the Rowland radius from 350 to 200 mm being achieved via a very slight modification (type A). After performing a series of measurements with the type A spectrometer, we developed a new compact spectrometer for crystals with even smaller Rowland radii (150 mm in the present case, type B). Both spectrometers performed satisfactorily. The energy resolution of the spectrometer depends not only on Rowland radii but also on other geometrical factors, while the efficiency can be improved directly reducing the radii. Therefore, the type B spectrometer could be rather competitive. The present paper mainly reports on data obtained from the type B spectrometer, although some resonant XRF spectra were collected with the type A spectrometer. Both spectrometers employed Johanssontype Ge (220) (Crismatec) analyzing crystals and a NaI:Tl scintillation detector (Rigaku; 22 mm diameter). The receiving slit width for the crystals with Rowland radii of 200 and 150 mm was 0.1 and 0.07 mm, respectively. Table 1 summarizes the other experimental conditions for obtaining the spectra presented in this paper. Figure 1 shows a schematic drawing of the type B spectrometer. XRF generated at the sample (S) is analyzed by a Johanssontype curved crystal (C), which rotates to the angle θ on a small goniometer (I) and moves linearly by a translational stage (II). S 2614 Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
no. integrn of time/ steps step (s)
Figure 1. Schematic drawing of the type B Johansson-type spectrometer. Key: S, sample; C, analyzing crystal; ES, entrance slit; RS, receiving slit; D, detector; O, center of Rowland circle; I, twoaxis goniometer; III, 2θ arm; II, IV, and V, translational stages.
is in a compact vacuum chamber to reduce scattering X-rays. The analyzing crystal receives XRF in a direction perpendicular to the incident X-ray beam, to reduce scattering background by making full use of the linear polarization of the synchrotron radiation. The receiving slit (RS) and the detector (D) are on the 2θ arm (III), rotate around the center of C, and move by translational stages (IV and V). S, C, and RS are always on the same Rowland circle to satisfy the focusing condition. When XRF is analyzed, the scan is performed so that both S-C and C-RS are equidistant. The spectrometer covers a limited angular range of 17-32°, which corresponds to the S-C distance of 88-159 mm. The spectrometer is very compact. RESULTS AND DISCUSSION Figure 2 shows the XRF spectrum of iron KR from NIST SRM 610, containing many kinds of trace metals in a glass matrix. As shown in the inset of the figure, it is not usually easy for a conventional Si(Li) detector to separate KR lines from the Kβ lines of the neighboring element, simply because the energy resolution is insufficient. The use of a crystal analyzer not only solves the
Figure 2. XRF spectrum of iron KR from NIST SRM 610 (450 ppm Fe). The inset is measured by a Si(Li) detector for comparison (10keV excitation, at the bending magnet beamline at the Photon Factory).
Figure 3. XRF spectrum of copper KR from NIES reference material No. 8, vehicle exhaust (67.7 ppm Cu). The inset is measured by a Si(Li) detector for comparison (10-keV excitation, at the bending magnet beamline at the Photon Factory).
overlap of iron KR and manganese Kβ peaks, but also enables the separation of iron KR1 and KR2 lines. The fwhm of iron KR1 was ∼8 eV, whereas the overlapped peak of iron KR and manganese Kβ observed by a Si(Li) detector has a width of 290 eV (although the intrinsic energy resolution would be ∼170 eV). The intensity obtained here was ∼40 times higher compared with our previous report.8 The improvement was achieved mainly by the enhancement of the detection efficiency of the spectrometer, but partially by the number of incident photons, which were ∼3 times greater than before, because of the higher ring current. Another advantage of enhancing the energy resolution is an improved signal-to-background ratio. Figure 3 shows the XRF spectrum of NIES reference material No. 8 (vehicle exhaust particulates), which contains 67.7 ppm copper and other 3d transition metals as well as light elements of orders ranging from ppm to 0.5%. One can clearly see the reduction in the background from the low-energy-side tail of strong neighboring lines, which poses a serious problem for a Si(Li) detector (see the inset of Figure 3). The detection limit is estimated as 1 ppm for copper in a carbon matrix. In this case, the analyzing area is 0.4 mm2, and therefore, the corresponding absolute amount of copper is 1.2 ng. So far,
Figure 4. Kβ spectra of metallic iron foil and 15 µg of potassium hexacyanoferrate(III) (K3[Fe(CN)6]). The data are normalized by the Kβ1,3 peak intensity (13 522 and 23 882 counts, for metallic iron foil and K3[Fe(CN)6], respectively). The inset is an enlarged spectrum to show the intensity change in Kβ′′.
white or quasi-monochromatic excitation has typically been used for wavelength-dispersive XRF,5-7 but the present result demonstrates that trace element analysis can be performed even by monochromatic excitation. Although it is possible to gain further intensity by slightly degrading the energy resolution, the most feasible option is to maintain a resolution of ∼10 eV at K-lines for 3d transition metals, as this can provide information on the chemical state in addition to the usual identification of elements. Figure 4 shows Kβ spectra of metallic Fe and K3[Fe(CN)6]. While Kβ1,3 and Kβ5 were independent peaks, satellite lines Kβ′ and Kβ′′ were observed as the lower and higher energy side shoulders of Kβ1,3, respectively. In this case, in addition to the regular transition of Fe 3d to Fe 1s, C 2p to Fe 1s can contribute to Kβ5. Kβ′′ corresponds to C 2s to Fe 1s.10,11 Such additional crossover transitions from electrons in the neighboring atoms could increase both Kβ5 and Kβ′′ intensity for K3[Fe(CN)6], in comparison with the metallic Fe, where the Fe 4s to Fe 1s transition is not forbidden. One can see that the Kβ5 and Kβ′′ intensity definitely changes according to the chemical environment. Slight intensity changes are visible as well in Kβ′, which are due to an exchange interaction and/or plasmon oscillation, although quantitative discussion is not easy in this case. Accordingly, it is possible to study some chemical effects by the observation of the profiles of, and intensity changes in, Kβ lines. The intensity ratio observed is summarized in Table 2. Figure 5 shows the Kβ spectra of metallic Co, Co(C5H7O2)2‚ 2H2O, and Co(C5H7O2)3. Again, these is a significant difference in Kβ5 and Kβ′′ intensity, which reflects the contribution of the crossover transition from electrons in the oxygen atoms (see Table 2). The intensity of Kβ′′ for complexes is higher than that for metallic Co. Furthermore, the intensity depends on the oxidation numbers for the complex. The intensity for Co(C5H7O2)3 is apparently stronger than that for Co(C5H7O2)2‚2H2O. This indicates the feasibility of its potential application to chemical-state (10) Koster, A. S.; Rieck, G. D. J. Phys. Chem. Solids 1970, 31, 2505-2510, 25112522. (11) Deutsch, M., Ho ¨lzer, H., Ha¨rtwig, J., Wolf, J., Fritsch, M., Fo ¨rster, E. Phys. Rev. A 1995, 51, 283-296.
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
2615
Figure 5. Kβ spectra of metallic cobalt foil and 15-µg cobalt complexes. The data are normalized by the Kβ1,3 peak intensity (5205, 61 468, and 49 618 counts, for metallic cobalt foil, bis(acetylacetonato)cobalt(II) (Co(C5H7O2)2‚2H2O), and tris(acetylacetonato)cobalt(III) (Co(C5H7O2)3), respectively). Table 2. Relative Peak Intensities of Kβ5 and Kβ′′
compound
oxidn no.
Kβ5 intens ratio IKβ5/IKβ1,3
Kβ′′ intens ratio IKβ′′/IKβ1,3
ref figure
Fe metal K3[Fe(CN)6] Co metal Co(C5H7O2)2‚2H2O Co(C5H7O2)3
0 III 0 II III
0.021 0.035 0.025 0.020 0.025
0.007 0.011 0.009 0.015 0.022
4 4 5 5 5
analyses. The ease of quantitative interpretation of satellite lines would be one important advantage of monochromatic excitation. This is because intensity can be affected by several different factors in the case of white X-ray excitation.12 The amount of cobalt contributing to the spectra is 15 µg. It is possible to discuss chemical states for trace systems, if the energy resolution of the present spectrometer is maintained at ∼10 eV at K-lines for 3d transition metals. One could study chemical shifts of the peaks by a highresolution spectrometer; however, in this case, the efficiency loss becomes very severe when the concentration of target elements becomes low. Observing changes in the intensity ratio13,14 would be a promising way for trace chemical characterization. Furthermore, a monochromatic X-ray undulator source throws up another interesting application, i.e., resonant XRF spectra.15 Figure 6 shows the spectra for metallic Co and Co3O4 measured as a function of incident X-ray energy around the K absorption edge. The energy dependence of XRF intensity can provide the same information as the X-ray absorption spectrum around the edge. The advantage of using XRF is its high sensitivity, which permits its application to trace systems. Detecting chemical shifts of absorption edges using a Si(Li) detector has been particularly widely used for chemical characterization.16,17 The absorption edge energy for the oxide is higher than that for the metallic Co as (12) Deslattes, R. D.; LaVilla, R. E.; Cowan, P. L.; Henins, A. Phys. Rev. A 1983, 27, 923-933. (13) Iihara, J., Izawa, G., Omori, T., Yoshihara, K. Nucl. Instrum., Methods 1990, A299, 394-398. (14) Iihara, J., Kawai, J., Sekine, T., Yoshihara, K. J. PIXE 1993, 3, 177-183. (15) Kotani, A. J. Phys. IV Fr. 1997, 7, C2, 1-8. (16) Sakurai, K.; Iida, A.; Gohshi, Y. Anal. Sci. 1988, 4, 37-42. (17) Nakai, I.; Iida, A. Adv. X-ray Anal. 1992, 35B, 1307-1315.
2616 Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
Figure 6. Resonant XRF spectra (KR1, KR2) at the cobalt K absorption edge for (a) 15 µg of metallic cobalt and (b) tricobalt tetroxide (Co3O4). (b′) is an enlarged spectrum for (b) to show a weak fine structure.
shown in Figure 6. In the present case, not only the integral intensity of XRF is available but also the spectral profile, at quite good resolution too. Faint peaks of X-ray inelastic scattering can be seen on the low-energy side of the absorption edge (see Figure 6b′). The energy of scattering shifts to the low-energy side as the incident X-ray energy is lowered. This is because the energy difference between incident and scattering X-rays is constant (the energy of LIII absorption edge). Figure 7 shows the resonant Kβ spectra of K3[Fe(CN)6]. Since the energy of the Kβ line is rather close to the incident X-rays, it is almost impossible for a Si(Li) detector to measure even integrated Kβ intensity as a function of incident energy around the absorption edge. A preedge resonant peak, which is specific to the iron complex, was observed for both the Kβ1,3 and Kβ5 lines. As discussed earlier, Kβ spectra can be used for chemical-state analysis, and therefore, the spectrum collected around the resonant condition holds out the promise of further detailed information. Although resonant XRF spectra have been studied from the perspective of gaining a fundamental under-
Figure 7. Resonant XRF spectra (Kβ1,3, Kβ5) for 15 µg potassium hexacyanoferrate(III) (K3[Fe(CN)6]) at the iron K-absorption edge.
standing of X-ray emission and inelastic scattering in terms of charge transfer,18,19 most of these measurements have been performed only for pure or highly condensed samples. Therefore, it should be noted that, in the present research, resonant XRF spectra were obtained for diluted samples, 15 µg of cobalt or iron in a carbon matrix. CONCLUSIONS A versatile compact Johansson-type XRF spectrometer has been developed for trace chemical characterization using the monochromatic X-ray undulator source at SPring-8. The spectra (18) Udagawa, Y.; Tohji, K. Chem. Phys. Lett. 1988, 148, 101-106. (19) Iwazumi, T.; Kobayashi, K.; Kishimoto, S.; Nakamura, T.; Nanao, S.; Ohsawa, D.; Katano, R.; Isozumi Y. Phys. Rev. B 1997, 56, 14267-14270.
obtained by a NaI:Tl detector and Ge(220) analyzing crystals with a Rowland radius of 150 mm clearly demonstrate its feasibility. Efficiency has been considerably improved in comparison with typical wavelength-dispersive spectrometers. The energy resolution is ∼10 eV for most of the K-lines for 3d transition metals. Since the signal-to-background ratio has been significantly improved, the technique is suitable for trace systems. Another important application is chemical-state analysis using some chemical effects observed in Kβ spectra. It has been found that changes in intensities of main (Kβ5) and satellite lines (Kβ′, Kβ′′) can be used to ascertain differences in chemical states. Obviously it would not be so easy to apply the technique to entirely unknown samples. However, the capability of detecting the change in chemical environment such as the oxidation numbers is very feasible for practical analytical application. In view of trace chemical characterization, the present spectrometer can be a good alternative to a conventional Si(Li) detector, by combining with a highly brilliant undulator X-ray source. ACKNOWLEDGMENT The authors gratefully acknowledge the kind assistance of Drs. S. Goto and M. Suzuki (SPring-8) during the experiments at the beamline. Their thanks also go to Mr. M. Goto (CI Industry Co. Ltd.) for his technical cooperation in designing and assembling the spectrometer. C.N. expresses special thanks to Prof. Y. Koto (Tokushima University) for his continuous encouragement. This work was performed with the approval of the SPring-8 Program Advisory Committee (Proposal No.1999A0016-NM-np). Received for review November 15, 1999. Accepted February 24, 2000. AC991308F
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
2617
