Nuclear Instruments and Methods in Physics Research A 436 (1999) 85}90 Soft X-ray response of a CCD with a grating spectrometer M. Shouho , K. Katayama , H. Katayama , T. Kohmura , H. Tsunemi , S. Kitamoto *, K. Hayashida , E. Miyata , K. Hashimotodani , K. Yoshita , K. Koyama, G. Ricker, M.W. Bautz, R. Foster, S. Kissel Department of Earth & Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan CREST, Japan Science and Technology Corporation (JST), 4-1-8 Honmachi, Kawaguchi, Saitama 332, Japan Department of Physics, Graduate School of Science, Kyoto University, Kitashirakawa-Oiwake-cho, Sakyo-ku, Kyoto 606-8502, Japan Center of Space Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA Abstract We calibrate the X-ray imaging spectrometers, which are CCD cameras installed on the ASTRO-E satellite, by using dispersed continuous soft X-rays from a grating spectrometer. We obtained the signal-pulse height and energy-resolution as a function of X-ray energies continuously. However, the wings of the line spread function of the grating distorts the center of the signal-pulse height derived by a simple analysis. An estimation of this distortion is presented. We also describe two methods of extracting the pure signal-pulse-height distribution from the data using the spectrometer. A brief description of the low-energy tail is presented. 1999 Elsevier Science B.V. All rights reserved. PACS: 07.85.F; 85.60.G; 95.55.A Keywords: CCD; X-ray; Grating; Signal-pulse height 1. Introduction The response function of a charge coupled device (CCD) has been investigated by some authors [1}3]. The CCD response for monochromatic X-rays has some structures: e.g., a low-energy tail, an escape peak, and a low-energy peak [4]. Both the amount and the shape of these structures are * Corresponding author. Department of Earth & Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan, Tel.: #81-6-6850-5478; fax.: #81-6-6850-5539. E-mail address: [email protected] (S. Kitamoto) functions of incident X-ray energy. The grating spectrometer is very useful for the measurement of the pulse-height distribution of a CCD, because a CCD can obtain the dispersed spectrum in a certain energy range in one exposure [5]. We are now calibrating the CCD camera, the X-ray imaging spectrometer (XIS) for the ASTROE satellite, which will be launched in early 2000. The CCD chips are of the frame transfer type. Their imaging area has 1024 pixels ; 1026 pixels and the physical size of about 1 in. ; 1 in. The CCD has four readout nodes and 256 pixels of each horizontal line are read out from each node. We call this CCD structure segment, and treat the data from each segment separately. The details of the XIS 0168-9002/99/$ - see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 9 0 0 2 ( 9 9 ) 0 0 6 0 2 - 6 SECTION I. 86 M. Shouho et al. / Nuclear Instruments and Methods in Physics Research A 436 (1999) 85}90 have been previously published by Hayashida et al. [6]. In this work, we report the study of the response function in the soft X-ray region (0.3}2.2 keV) of one of the XIS #ight models (XIS FM), using a grating spectrometer. The e!ect of the line spread function of the grating is discussed, and two methods for the extraction of the signal-pulse-height distribution are described. 2. Experiment and results 2.1. Date reduction We expose the XIS CCD to dispersed soft X-rays from the Si-K edge spectrometer, Model FFS-II (SES) supplied by Hettrick Scienti"c Inc. [7,8]. The details of the spectrometer system have been described by Hashimotodani et al. [5]. The schematic illustration is shown in Fig. 1. We install the CCD chip at the focal plane, 775 mm from the grating center, and set the CCD chip so that the dispersion direction is parallel to the serial register (that is parallel to the CCD X-direction). We move the XIS along the X-ray dispersion direction and obtain data at seven di!erent positions to cover an energy range from 0.25 to 2.5 keV for all segments. We select X-ray events with signal-pulse heights above &180 eV and charge distributions that "t the grade classi"cation, with a split threshold of &70 eV, as described by Gendreau [9]. Fig. 2. Two-dimensional plot of the X-ray events. The horizontal axis is the CCD position, which is the direction of the dispersion, and hence corresponds to the incident X-ray energy. The vertical axis is the signal-pulse height with the unit of the ADC channel (ADU). Each dot shows one X-ray event. The data reduction is as follows (see also Ref. [10]). We produce a two-dimensional plot, shown in Fig. 2, by assigning one axis as the position along the dispersion direction and the other as the signal-pulse height of the X-ray events measured with the CCD. Since X-ray energy is speci"ed by the position along the dispersion direction, this Fig. 1. The side view of the silicon edge spectrometer. M. Shouho et al. / Nuclear Instruments and Methods in Physics Research A 436 (1999) 85}90 87 two-dimensional plot directly shows the relation between the incident X-ray energy and the signalpulse height. However, there are higher-order dispersed X-rays. Furthermore, we can see several horizontal lines in Fig. 2, which means that there are X-rays which are not dispersed by the grating. 2.2. Energy-PH and energy-FWHM relation In this subsection, we derive the mean signalpulse height and the energy resolution as functions of the incident X-ray energy, by simple Gaussian "tting. We extract each order light from the twodimensional plot, shown in Fig. 2, by applying the upper and lower boundaries on the signal-pulse height. The boundaries are determined so as to delineate an appropriate area which includes the position of the dispersed light. The resultant X-ray spectrum of the "rst-order light as a function of the incident X-ray energy is shown in Fig. 3. A histogram of the signal-pulse height obtained using the CCD (signal-pulse-height distribution) in a narrow range of incident X-ray energy exhibits a peak whose shape can be approximately simulated by a Gaussian function. We "t the signal-pulse-height distributions with a Gaussian model. The center value of the Gaussian (PH) is plotted as a function of the incident X-ray energy (energy-PH relation) in Fig. 4. The width of the Gaussian in terms of FWHM is plotted as a function of the energy (energy-FWHM relation) in Fig. 5. Fig. 4. Energy-PH relation of the XIS #ight model (FM) (Segment B) as a function of X-ray energy. The data are for events of Grade 0#2#3#4#6, in "rst-, second-, third-, and fourthorder light. Fig. 5. Energy resolution (FWHM) of the XIS FM (Segment C) as a function of X-ray energy. The data are for events of Grade 0#2#3#4#6, in "rst-, second-, third-, and fourth-order light. Fig. 3. Dispersed X-ray spectrum of "rst-order light. The lower panel of Fig. 4 shows the residuals of the data from the best-"t linear model. We obtain the relation between the incident X-ray energy and PH with an error of less than &7 eV (&2 ADU) for the energy range of 0.3}1.85 keV. However there is a jump of &12 eV (&4 ADU) around 1.9 keV. The interpretation of this jump will be discussed in Section 3. SECTION I. 88 M. Shouho et al. / Nuclear Instruments and Methods in Physics Research A 436 (1999) 85}90 We "t the energy-FWHM relation with a formula written as E FWHM"2.355;u p#f; u (eV) (1) where E is the incident X-ray energy, u is the mean ionization energy per one electron}hole pair, f is the Fano factor, and p corresponds to system noise including readout noise. The best-"t values are p"0.62$0.02 and f"0.123$0.001, where we assume that u is constant at 3.65 eV. We found an energy resolution of &40 eV at O-K (0.52 keV), and &60 eV at Mg-K (1.25 keV). 2.3. Signal-pulse-height distributions In the above subsection, the mean and the "rst moment (energy resolution) of the signal-pulseheight distribution were described. However, in order to construct the response function of the CCD, more detail on the shape of the signal-pulseheight distribution is required. We developed the following two methods for the extraction of the pure signal-pulse-height distribution from the data, using the spectrometer. In the "rst method, some emission lines, which can be seen in Fig. 3, are used to extract the pure shape of the signal-pulse-height distribution. In the following, we describe this method for the O-K line. The events around the O-K line (0.52 keV) consist of three components, the O-K line, the dispersed continuum, and nondispersed X-rays. First, we produced an energy spectrum from a limited region which includes the O-K line. Next, we produced energy spectra from the two neighboring regions which do not include the peak of the O-K line, where we used only the data below the O-K edge. These consist of the dispersed continuum and nondispersed X-rays, although a small amount of the O-K lines are included. Finally, we subtract the average of the two spectra from the spectrum including the O-K line, after correction of the area used. If the nondispersed X-rays are smoothly distributed around the O-K line, they are not included in the resultant spectrum. The dispersed continuum X-rays are also almost completely subtracted, if their intensity is a smooth function of the X-ray Fig. 6. The pure signal-pulse-height distribution of the grade 0#2#3#4#6 events of the O-K line and the Mg-K line. The nondispersed component has been subtracted (see text). energy. A slight di!erence in the center energy of the dispersed X-rays does not a!ect the resultant spectra, because such a di!erence is much smaller than the energy resolution of the CCD. Therefore, the obtained spectrum should show the pure signalpulse-height distribution for the O-K X-rays. The obtained spectrum is shown in Fig. 6. The same method is applied for the Mg-K line and the spectrum obtained for Mg-K line is shown in Fig. 6. The tail component of the Mg-K line is substantially smaller than that of the O-K line. In the above method, we can obtain a clear pure signal-pulse-height distribution, but we can apply this method only for some emission lines. Thus, we develop a second method for obtaining the signal-pulse-height distribution at other energies, where the estimated nondispersed X-rays are subtracted. The energy spectrum of the nondispersed X-rays is estimated from the spectrum obtained near the zeroth-order light, where no re#ectioncondition of the grating is satis"ed for the X-rays below the applied high voltage (5 kV) of the X-ray generator. The intensity of the nondispersed light as a function of the distance from the zeroth-order position is also derived from the data of the same region. The exponential function with a constant component "ts the data well. Extrapolating this function, the spectrum and the intensity of the nondispersed X-ray are estimated. M. Shouho et al. / Nuclear Instruments and Methods in Physics Research A 436 (1999) 85}90 Fig. 7. The signal-pulse-height distribution at "ve energy positions, O-K line (0.525 keV), 0.8, 1.0 keV, Mg-K line (1.25 keV), 1.49 keV. The grade 0#2#3#4#6 events are selected. All were derived by subtracting the estimated nondispersed X-rays. The derived signal-pulse-height distributions for "ve positions (O-K line (0.525 keV), 0.8, 1.0 keV, Mg-K line (1.25 keV), and 1.49 keV) are shown in Fig. 7. Comparing the signal-pulse-height distributions of the O-K line and Mg-K line derived by the two independent methods, the systematic di!erence is found to be of the order of 0.1% of the peak count rate. 3. Discussion We found a jump of the center value of the signal-pulse-height distribution at the energy around 1.9 keV. Figs. 8(a) and (b) shows the expanded dispersed X-ray spectrum, and the residuals of the energy-PH relation from the best-"t linear model. Although the discontinuity is near the Si-K edge (1.840 keV), the energy is signi"cantly di!erent from it. Thus, the discontinuity is not due to the characteristics of the CCD. The obtained result can be interpreted as follows. We derived the signal-pulse height and the energy resolution from a simple Gaussian "t. Although the energy resolving power of the grating is much higher than that of the CCD, the line spread function of the grating has a "nite width. This e!ect becomes large at the high-energy part. 89 Fig. 8. The dispersed spectrum (a) and the residuals of the energy-PH relation from the linear model (b). The estimated deviation of the center from the real X-ray energy is plotted in (c), where the gaussian pro"le with a standard deviation of 72 lm is assumed for the line spread function of the grating. The contribution onto the energy resolution from this line spread function is plotted in (d) by the standard deviation. If the intensity of the dispersed X-rays are smooth, the "nite width of the line spread function reduces the energy resolution, but the e!ect on the center energy is small. However, if the intensity of the dispersed X-rays changes steeply as a function of X-ray energy, the e!ect of the line spread function distorts the center energy toward the intense X-ray direction. In Fig. 8(b), it can be seen that the intensity-decreasing part shows negative residuals and intensity-increasing part shows positive residuals. It should be noted that the residuals shown in Fig. 8(b) are those from the best-"t linear model, but not from the real center. A simple estimation was performed assuming the line spread function of the grating: a Gaussian pro"le with a standard deviation of 72 lm, which is the length of three pixels of the CCD. This resolution corresponds to &5.5 eV at 1.25 keV. If the real center energy of the CCD response function is linear to the incident X-ray energy, the deviation *E of the derived center energy from the real center energy can be estimated as Ef (E) exp(!(E!E )/2p) dE *E" !E f (E) exp(!(E!E )/2p) dE (2) SECTION I. 90 M. Shouho et al. / Nuclear Instruments and Methods in Physics Research A 436 (1999) 85}90 where f (E) is the intensity of the dispersed spectrum, p is the standard deviation of the Gaussian pro"le, and E is the X-ray energy. The estimated deviation (*E) is plotted in Fig. 8(c). The rough shape of the residuals can simulate not only the jump around 1.9 keV but also the small change at 1.25 keV by the Mg-K line, and 1.55 keV by the Al-K edge. Although the energy of the jump around 1.9 keV in the data seems to be slightly smaller than that of the simulation, this is partly due to the large scatter of the data around 1.9 keV. In addition, the assumed line spread function is probably too simple to reproduce the exact deviation. The standard deviation due to the line spread function of the grating is plotted in Fig. 8(d). The e!ect on the energy resolution plotted in Fig. 5, due to this line spread function, is roughly 1.2 eV at 1.25 keV, which is much smaller than the error of the data. The two methods for the extraction of the signal-pulse-height distribution are described in Section 2.3. The "rst method is completely new, and can be widely applied for the dispersed spectrometer. As long as the continuum is smooth, the resultant signal-pulse-height distribution has no ambiguity. The e!ects of the line spread function of the grating are less than &1 ADU on the center energy and &1.2 eV (FWHM) on the energy resolution for the Mg-K line. For the O-K line, these e!ects are much smaller. In the second method, we estimate only the e!ect from the zeroth-order light. This is insu$cient for a detailed analysis. However, in the analysis shown in this work, a systematic error of the order of 0.1% of the peak count is con"rmed, indicating that it is still useful for the study of the shape of the signal-pulseheight distribution. Fig. 7 shows the complex behavior of the tail component of the signal-pulse-height distribution. The tail component consists of at least three components. One is the deviation from the Gaussian pro"le at the low-energy part, which always appears. The second is the large hump at the lowenergy part, which is prominent for the O-K line and disappears for X-rays above 1 keV. The third is the long #at tail on the low-energy side. This component seems to decrease according to the energy of the incident X-rays. The quantitative analysis will be published elsewhere. 4. Conclusion The simple analysis and results of the calibration using the dispersed continuous soft X-rays (0.25}2.2 keV) from our grating spectrometer are presented. In the simple analysis, the line spread function of the grating causes a discontinuity of the pulse height. We described two methods of extracting a pure signal-pulse-height distribution from the data obtained using the spectrometer. We demonstrated that the tail of the signal-pulseheight distribution consists of at least three components. References [1] M. Pivovaro!, S. Johns, M. Bautz, S. Kissel, G. Prigozhin, G. Ricker, H. Tsunemi, E. Miyata, IEEE, Trans. Nucl. Sci. NS-45 (1998) 164. [2] K.J. McCarth, A. Owens, A.D. Holland, A.A. Wells, Nucl. Instr. and Meth. A 362 (1995) 538. [3] G.W. Fraser, A.F. Abbey, A. Holland, K. McCarthy, A. Owens, A. Wells, Nucl. Instr. and Meth. A 350 (1994) 368. [4] G.A. Prigozhin et al., Nucl. Instr. and Meth., 1999, in press. [5] K. Hashimotodani, T. Toneri, S. Kitamoto, H. Tsunemi, K. Hayashida, E. Miyata, K. Katayama, T. Kohmura, R. Asakura, K. Koyama, K. Yamamoto, K. Miyaguchi, H. Suzuki, Rev. Sci. Instr. 69 (1998) 392. [6] K. Hayashida, S. Kitamoto, E. Miyata, H. Tsunemi, K. Katayama, T. Kohmura, R. Asakura, K. Yoshita, H. Katayama, M. Shouho, K. Koyama, T.G. Tsuru, H. Awaki, T. Dotani, M. Ozaki, G.R. Ricker, J.P. Doty, M.W. Bautz, S.E. Kissel, R.F. Foster, Proc. SPIE 3445 (1998) 278. [7] M. Hettrick, S. Bowyer, Appl. Opt. 22 (1983) 3921. [8] M. Hettrick, Appl. Opt. 23 (1984) 3221. [9] K.C. Gendreau, X-ray CCDs for Space Applications: Calibration, Radiation Hardness, and Use for Measuring the Spectrum of the Cosmic X-ray Background, Ph.D. Thesis, MIT, 1995. [10] K. Katayama, T. Kohmura, H. Katayama, M. Shouho, H. Tsunemi, S. Kitamoto, K. Hayashida, E. Miyata, K. Hashimotodani, K. Koyama, G. Ricker, M. Bautz, R. Foster, ASTRO-E XIS Team, Advances in Space Research, (1999) in press.
© Copyright 2026 Paperzz