- Email: [email protected]
Modeling the spectral response for the soft X-ray imager onboard the ASTRO-H satellite
Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Modeling the spectral response for the soft X-ray imager onboard the ASTRO-H satellite Shota Inoue a,b,n, Kiyoshi Hayashida a,b, Shuhei Katada a,b, Hiroshi Nakajima a,b, Ryo Nagino a,b, Naohisa Anabuki a,b, Hiroshi Tsunemi a,b, Takeshi Go Tsuru c, Takaaki Tanaka c, Hiroyuki Uchida c, Masayoshi Nobukawa d, Kumiko Kawabata Nobukawa c, Ryosaku Washino c, Koji Mori e, Eri Isoda e, Miho Sakata e, Takayoshi Kohmura f, Koki Tamasawa f, Shoma Tanno g, Yuma Yoshino g, Takahiro Konno g, Shutaro Ueda h a
Department of Earth and Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan Project Research Center for Fundamental Sciences, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan c Department of Physics, Graduate School of Science, Kyoto University, Kitashirakawa, Oiwake-cho, Sakyo-ku, Kyoto 606-8502, Japan d Department of Teacher Training and School Education, Nara University of Education, Takabatake-cho, Nara 630-8528, Japan e Department of Applied Physics, University of Miyazaki, 1-1 Gakuen Kibana-dai Nishi, Miyazaki 889-2192, Japan f Department of Physics, Graduate School of Science and Technology, Tokyo University of Science, 2641 Yamazaki, Noda, Chiba 278-0022, Japan g Department of Physics, Faculty of Science and Technology, Tokyo University of Science, 2641 Yamazaki, Noda, Chiba 278-0022, Japan h Institute of Space and Astronautical Science (ISAS), Japan Aerospace Exploration Agency (JAXA), 3-1-1 Yoshinodai, Chuo-ku, Sagamihara, Kanagawa 252-5210, Japan b
On behalf of ASTRO-H/SXI team art ic l e i nf o
a b s t r a c t
Article history: Received 4 December 2015 Received in revised form 19 March 2016 Accepted 22 March 2016
The ASTRO-H satellite is the 6th Japanese X-ray astronomical observatory to be launched in early 2016. The satellite carries four kinds of detectors, and one of them is an X-ray CCD camera, the soft X-ray imager (SXI), installed on the focal plane of an X-ray telescope. The SXI contains four CCD chips, each with an imaging area of 31 mm × 31 mm , arrayed in mosaic, covering the field-of-view of 38′ × 38′, the widest ever flown in orbit. The CCDs are a P-channel back-illuminated (BI) type with a depletion layer thickness of 200 μm . We operate the CCDs in a photon counting mode in which the position and energy of each photon are measured in the energy band of 0.4–12 keV. To evaluate the X-ray spectra obtained with the SXI, an accurate calibration of its response function is essential. For this purpose, we performed calibration experiments at Kyoto and Photon Factory of KEK, each with different X-ray sources with various X-ray energies. We fit the obtained spectra with 5 components; primary peak, secondary peak, constant tail, Si escape and Si fluorescence, and then model their energy dependence using physicsbased or empirical formulae. Since this is the first adoption of P-channel BI-type CCDs on an X-ray astronomical satellite, we need to take special care on the constant tail component which is originated in partial charge collection. It is found that we need to assume a trapping layer at the incident surface of the CCD and implement it in the response model. In addition, the Si fluorescence component of the SXI response is significantly weak, compared with those of front-illuminated type CCDs. & 2016 Published by Elsevier B.V.
Keywords: X-ray Charge-coupled device P-channel CCD ASTRO-H
1. Introduction Charge-coupled devices (CCDs) are widely used for X-ray
n Corresponding author at: Department of Earth and Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 5600043, Japan. E-mail address: [email protected] (S. Inoue).
astronomy (e.g. [1,2]). The ASTRO-H satellite [3], which is the 6th Japanese X-ray observatory, includes an X-ray CCD camera called the soft X-ray imager (SXI, [4–6]). The SXI contains four CCD chips, each with an imaging area of 31 mm × 31 mm , arrayed in mosaic, covering the field-of-view area of 38′ × 38′, the widest ever flown in orbit. The original pixel size is 24 μm × 24 μm , but we operate it with 2 × 2 pixel binning to reduce the number of readout channels while preserving enough
http://dx.doi.org/10.1016/j.nima.2016.03.071 0168-9002/& 2016 Published by Elsevier B.V.
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
2
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
resolution. The N-type silicon used for the SXI yields a depletion layer thickness of 200 μm . The CCDs are of back-illuminated (BI) type, which gives the SXI enough tolerance against micro-meteorites and high quantum efficiency both at low and high energy bands. Fig. 1 shows the schematic view of the cross-sectional surface of the SXI. As with the CCD property shown in previous studies (e.g. [7– 10]), the response function of the SXI for monochromatic X-ray photons is reproduced with 5 components: primary Gaussian, secondary Gaussian, constant, Si escape and Si K components (Fig. 2). The origin of each of these components is explained by the following occurrences. When all the charge induced by an X-ray is generated in the depletion layer and collected at the gate, the event is detected as the primary Gaussian component. Sometimes, the charge cloud is split over the neighboring pixels. When spread charge is below the threshold, the event is detected as the secondary Gaussian component. The event appeared as the constant component is generated when a part of charge is lost near the boundary of the depletion layer. A Si K fluorescent X-ray (e.g. Si K α 1.739 keV) following the photon absorption sometimes escapes from the pixel. This event is detected as the Si escape peak and its peak energy is 1.739 keV (Si K α) lower than the primary peak. The Si K component includes the following two cases: (1) an X-ray absorbed at the SiO2 layer is re-radiated as Si K and (2) Si K due to a Si escape event is re-absorbed outside the pixel. Empirically, these 5 components are described as four Gaussians and a constant.
2. Calibration experiments To evaluate the X-ray spectra obtained with the SXI in orbit, an accurate ground calibration of its response function is essential. For this purpose, we performed two calibration experiments: flight model (FM) SXI camera calibration (Fig. 3) and KEK experiments with test CCDs (Fig. 4). The FM camera calibration was performed at Kyoto University from August to September in 2014. By irradiating various targets with α -rays from 241Am, we generated fluorescent X-rays of Ge K α (9.9 keV), Ge L (1.2 keV) and F K α (0.68 keV). We also irradiated CCDs with the Mn K α (5.9 keV) by a 55Fe source. In December 2014, we performed the calibration experiment at the KEK Photon Factory (KEK-PF), where monochromatic X-rays are available. In this experiment, we used mini-sized CCDs that we fabricated in the development stage of the SXI using the same process as the FM. The difference between FM and mini-sized CCDs is in the size, the number of pixels and that of readout nodes. The X-ray energies of 0.50 keV, 0.70 keV, 1.83 keV and 1.85 keV are available in this experiment.
3. Construction of the response function We construct the response function for the SXI using these calibration data. Before modeling the function, we correct the effects of charge transfer inefficiency and charge trailing (see [11]). These are the effects that the collected charge is lost during the charge transfer process, and we then correct the pulse heights of the X-ray line centroids. Using these corrected data, we fit each spectrum with the five components we describe in Section 1. In this fitting we leave most of the parameters free, while the width and center is fixed for some weak lines (e.g., Ni K β , the escape line of Ge K β ). Fig. 5 shows the spectrum taken for the FM camera calibration. In addition to the incident X-rays from the source, there are some other unexpected X-rays (e.g. Cr K α , Fe K α ). We fit each of these components with a single Gaussian template. After estimating the parameters for the SXI response, we examine the energy dependence of these parameters. In Table 1, we summarize the X-ray lines we used. Only in the analysis of the Primary center peak and that of the Primary FWHM, we include not only F K α , Ge L, Mn Kα and Ge K α , but also Mn K α Si escape,
Fig. 1. Schematic cross-section of the SXI CCD. Incident layer of the X-ray is the upper plane of this figure.
Fig. 2. Line profile of the SXI. Each component reproduces the line profile explained in Section 1.
Fig. 3. FM camera calibration at Kyoto University. The FM SXI camera is in the red circle. X-ray sources are attached at the top of the body. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
3
Table 1 X-rays we use to examine energy dependences of parameters of SXI response function. Energy values are weighted-averaged value for several lines (e.g., Kα1, Kα2). We use the KEK data only for the constant intensity ratio to the primary intensity. X-ray
Fig. 4. Calibration experiment at KEK-PF. We use a camera body and refrigerator different from those of the FM, while the CCD is almost the same as the FM. A spectroscope at the upstream side extracts monochromatic X-rays.
Energy (keV)
FM calibration at Kyoto FK α Ge L Mn K α Ge K α Mn K α Si escape Cr K α Fe K α Fe K β Ni K α KEK experiment — — — —
0.6768 1.199 5.895 9.876 4.158 5.412 6.399 7.058 7.472 0.50 0.70 1.83 1.85
Pulse Height [ch]
1500
1000
500
Residual [ch]
0 2 0 −2 0
2
4
6
8
10
8
10
Energy [keV] 250
Primary FWHM [eV]
200
150
100
50
0 Fig. 5. Mn K α (upper panel) and Ge K α (lower panel) spectra of FM camera calibration at Kyoto University. In the fitting of the Mn K α data, the Ti K α line and the escape line of Mn K β shown as the residual between ∼ 700 and ∼ 850 channel are not taken into account.
Cr K α , Fe K α , Fe K β and Ni K α , for which we fit with a single Gaussian. Fig. 6 shows the energy dependence of the center peak (upper panel) and that of the FWHM (lower panel) of the primary component. For the center peak of the primary component, we model its energy dependence by a linear function. Typical
0
2
4
6 Energy [keV]
Fig. 6. Energy dependence of the peak channel (upper panel) and primary FWHM (lower panel). All data points are those taken by the FM calibration at Kyoto University.
discrepancies are within ∼2 ch between the data and the model. For the FWHM of the primary component, we employ the following function:
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
F ·W ·EX + N 2 + U ,
(1)
where EX is the energy of the incident X-ray, F is the Fano factor [13], W is the mean ionization energy in silicon, N is the readout noise and U is the unidentified term to originate in the thermal noise or other noise. We adopt F = 0.115 [14], W = 3.65 eV/e− [12] in this paper. While discrepancies between the data and the model are <20 ch , only for high statistic data point (F K α , Mn K α , Ge K α ) discrepancies are within ∼4 ch . In addition to the center peak and the FWHM of the primary component, we model energy dependences of intensity ratio of the secondary, constant, Si escape and Si K components to the primary component, and that of FWHM ratio of the secondary component to the primary component.
100 normalized counts s−1 keV−1
Primary FWHM =
MnKα
MnKβ
10 CrKα
MnKα Si escape
1
FeKβ
0.1 0.01 10−3 10−4 4
ratio
4
2 0
2
3
4
4. Unveiled features of the SXI
5
6
7
Energy (keV)
We find some interesting properties in our model of the SXI response function which we describe here. normalized counts s−1 keV−1
Fig. 7 shows the energy dependence of the integrated intensity ratio of the constant component to the primary component. For the constant component, we integrate the flat intensity from the first PHA channel to the peak of its primary line. Since it is difficult to estimate the constant intensity due to the contamination of other X-ray lines, we use the data of the KEK experiment in the low-energy band. To reproduce the observed data, we perform Monte-Carlo simulations using Geant 4 [15]. We construct a geometry in which the CCD has a layer structure as shown in figure Fig. 1 and input X-ray photons with incident directions perpendicular to the sensor plane. The red line in Fig. 7 represents the result of the simulation for the intensity ratio, which cannot reproduce the data in the low energy band. We then perform the simulation by assuming a charge-trapping layer between the depletion and oxidation layers. To reproduce the data, we assume that X-rays in this layer contribute to the constant component with flat energy distribution. With the charge-trapping layer thickness of 20 nm, we can reproduce the observed intensity ratio. We find that we need to take this layer
MnKα
MnKβ
10 CrKα
MnKα Si escape
1
FeKβ
0.1 0.01 10−3 10−4 4
ratio
4.1. Existence of a trap layer
100
2 0
2
3
4
5
6
7
Energy (keV) Fig. 8. Fitting result with the incident X-ray model convolved by SXI response we modeled as the best estimation. The upper panel shows the fitting result with the assumption of a single delta function model for a X-ray line. The lower panel shows the fitting result with a detailed X-ray model for the incident X-rays. The black and red dot lines represent the model components for Mn K α and those for Mn K β . Green dot-dash lines are additional line components approximated with a single gaussian model (Ti K α : 4.509 keV, Cr K α : 5.412 keV, Fe K α : 6.399 keV, Fe K β : 7.058 keV). The blue solid line represents the unidentified component expressed in Eq. (2). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
0.1
Constant / Primary
w/ trap layer
into account when we model the constant component of the SXI. We have not yet identified whether this layer is originated in this particular CCD model or common for some types of BI-CCDs. 4.2. Detailed incident X-ray model
0.01
10−3
Kyoto Cal KEK
w/o trap layer 0.5
1
2
5
10
20
Energy [keV] Fig. 7. Energy dependence of the intensity ratio of the constant component to the primary component. Red and blue lines show the model without a trap layer and that with a trap layer, respectively. The FM calibration data taken at Kyoto (green) are shown with the uncertainties including the systematic error due to the constant emission of the intrinsic X-ray. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
We fit the SXI spectra by using the response function we model as the best estimation. Fig. 8 shows the result for the Mn K α spectrum. The upper panel shows the result when we assume each X-ray line has a delta function profile. It fails to reproduce the spectrum around ∼ 5.5 keV, ∼ 6.2 keV and ∼ 6.8 keV which correspond to the boundaries between lines. We know that Mn K α and Mn K β emission consists of several lines (e.g. Mn K α1, Mn K α2, Mn K α3 etc.), and each of these lines is broadened with a Lorentzian profile [16]. By referring to the 55Fe spectrum taken with an X-ray micro-calorimeter (ΔE ∼ 5 eV , [17,18]), we model the line profile of Mn K α and Mn K β . We employ seven and five Lorentzian components listed in Tables II and III of [16] and one more Lorentzian component for Mn K α . In this modeling we introduce an extra component between the Mn K α and Mn K β whose intensity is
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
H satellite, planned to be launched in early 2016. To model the SXI response function before launch, we performed the ground calibration for the SXI. Using these data, we construct the SXI response and find the following:
10
Counts s−1 keV−1
1
1. The observed intensity of the constant component for the SXI is not reproduced by the model expected from the design of the SXI CCDs. We find that the data are well reproduced if we assume a charge-trapping layer with a thickness of 20 nm between the depletion layer and the oxidation layer. 2. The Si K component of the SXI is weaker than that of a FI-CCD (e.g. XIS onboard the Suzaku satellite). One reason is that the detection of the re-radiated Si K α from the illumination side occurs less frequent for BI-CCD than that of the FI-CCD. The other reason is the effectively larger pixel size of the SXI ( 48 μm ).
0.1
0.01
10−3
10−4
5
1
2
3
4
5
6
Energy (keV) Fig. 9. Simulated spectrum of the SXI (red) and that of the XIS (FI-CCD, black) for the monochromatic X-rays (5.9 keV). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
expressed by an empirical function as
⎧ a exp ((EX − b)/c ) (EX ≤ b) Intensity = ⎨ (EX > b) . ⎩0
(2)
Such a component has never been reported before and we cannot identify its origin. With this incident X-ray model, we can reproduce the SXI spectrum with our response function as shown in the lower panel of Fig. 8, while we have no idea to reproduce the spectrum at ∼ 5 keV. 4.3. Si K component Fig. 9 shows the simulated spectra of monochromatic X-rays by the SXI and the front-illuminated CCD of the Suzaku X-ray Imaging Spectrometer (XIS, [19,20]). Although the Si escape lines show similar intensities between the two CCDs, the Si K α line of SXI response is significantly weaker than that of the XIS because of two reasons. One is the thickness of the dead layer in front of the depletion layer. The FI-CCD has a relatively thicker dead layer at the illumination side. The other reason is their effective pixel size. The effective pixel size of the SXI is 48 μm by 2 × 2 binning, while that of XIS is 24 μm . Since the attenuation length of Si K α is ∼12 μm , the possibility that Si K α passes through the effective pixel size for the SXI is ∼ 1.8%, while that for the XIS is ∼ 14%.
5. Summary
Acknowledgments We thank all of the members of SXI team. S.I. is supported by Research Fellowships of Japan Society for the Promotion of Science for Young Scientists. This study was also supported by JSPS KAKENHI Grant numbers 15J01845, 23340071, 26109506, 15H03641, 26670560, 23000004, 24684010, 23340047, 25109004, 15H02090, 25870347, 26800102, 24740123, 24740167, 23740199, 15K17610.
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]
S. Sembay, et al., Proc. SPIE 5488 (2004) 264. C.E. Grant, et al., Proc. SPIE 6276 (2006) 48. T. Takahashi, et al., Proc. SPIE 8443 (2012) 84431Z. H. Tsunemi, et al., Proc. SPIE 7732 (2010) 11. K. Hayashida, et al., Proc. SPIE 8443 (2012) 23. K. Hayashida, et al., Proc. SPIE 9144 (2014) 914429. S. Ueda, et al., Proc. SPIE 8145 (2011) 814504. S. Ueda, et al., Nucl. Instrum. Methods Phys. Res. A 704 (2013) 140. H. Yamaguchi, et al., Proc. X-ray Universe (2006) 2005 (ESA SP-604. 949). H. Nakajima, et al., Nucl. Instrum. Methods Phys. Res. A 541 (365) (2005) 371. K.K. Nobukawa, et al., Nucl. Instrum. Methods Phys. Res. A 754 (269) (2014) 274. J.R. Janesick, Scientific Charge-Coupled Devices, SPIE Press, Bellingham, Washington, 2001. U. Fano, Phys. Rev. 72 (1947) 26. R.C. Alig, et al., Phys. Rev. B 22 (1980) 5565. S. Agostinelli, et al., Nucl. Instrum. Methods Phys. Res. A 506 (250) (2003) 303. G. Hölzer, et al., Phys. Rev. A 56 (1997) 4554. K. Mitsuda, et al., Proc. SPIE 7732 (2010) 773211. K. Mitsuda, et al., Proc. SPIE 9144 (2014) 91442A. K. Koyama, et al., Publ. Astron. Soc. Jpn. 59 (2007) 23. K. Mitsuda, et al., Publ. Astron. Soc. Jpn. 59 (2007) 1.
We developed the X-ray CCD camera, SXI, onboard the ASTRO-
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Modeling the spectral response for the soft X-ray imager onboard the ASTRO-H satellite Shota Inoue a,b,n, Kiyoshi Hayashida a,b, Shuhei Katada a,b, Hiroshi Nakajima a,b, Ryo Nagino a,b, Naohisa Anabuki a,b, Hiroshi Tsunemi a,b, Takeshi Go Tsuru c, Takaaki Tanaka c, Hiroyuki Uchida c, Masayoshi Nobukawa d, Kumiko Kawabata Nobukawa c, Ryosaku Washino c, Koji Mori e, Eri Isoda e, Miho Sakata e, Takayoshi Kohmura f, Koki Tamasawa f, Shoma Tanno g, Yuma Yoshino g, Takahiro Konno g, Shutaro Ueda h a
Department of Earth and Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan Project Research Center for Fundamental Sciences, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan c Department of Physics, Graduate School of Science, Kyoto University, Kitashirakawa, Oiwake-cho, Sakyo-ku, Kyoto 606-8502, Japan d Department of Teacher Training and School Education, Nara University of Education, Takabatake-cho, Nara 630-8528, Japan e Department of Applied Physics, University of Miyazaki, 1-1 Gakuen Kibana-dai Nishi, Miyazaki 889-2192, Japan f Department of Physics, Graduate School of Science and Technology, Tokyo University of Science, 2641 Yamazaki, Noda, Chiba 278-0022, Japan g Department of Physics, Faculty of Science and Technology, Tokyo University of Science, 2641 Yamazaki, Noda, Chiba 278-0022, Japan h Institute of Space and Astronautical Science (ISAS), Japan Aerospace Exploration Agency (JAXA), 3-1-1 Yoshinodai, Chuo-ku, Sagamihara, Kanagawa 252-5210, Japan b
On behalf of ASTRO-H/SXI team art ic l e i nf o
a b s t r a c t
Article history: Received 4 December 2015 Received in revised form 19 March 2016 Accepted 22 March 2016
The ASTRO-H satellite is the 6th Japanese X-ray astronomical observatory to be launched in early 2016. The satellite carries four kinds of detectors, and one of them is an X-ray CCD camera, the soft X-ray imager (SXI), installed on the focal plane of an X-ray telescope. The SXI contains four CCD chips, each with an imaging area of 31 mm × 31 mm , arrayed in mosaic, covering the field-of-view of 38′ × 38′, the widest ever flown in orbit. The CCDs are a P-channel back-illuminated (BI) type with a depletion layer thickness of 200 μm . We operate the CCDs in a photon counting mode in which the position and energy of each photon are measured in the energy band of 0.4–12 keV. To evaluate the X-ray spectra obtained with the SXI, an accurate calibration of its response function is essential. For this purpose, we performed calibration experiments at Kyoto and Photon Factory of KEK, each with different X-ray sources with various X-ray energies. We fit the obtained spectra with 5 components; primary peak, secondary peak, constant tail, Si escape and Si fluorescence, and then model their energy dependence using physicsbased or empirical formulae. Since this is the first adoption of P-channel BI-type CCDs on an X-ray astronomical satellite, we need to take special care on the constant tail component which is originated in partial charge collection. It is found that we need to assume a trapping layer at the incident surface of the CCD and implement it in the response model. In addition, the Si fluorescence component of the SXI response is significantly weak, compared with those of front-illuminated type CCDs. & 2016 Published by Elsevier B.V.
Keywords: X-ray Charge-coupled device P-channel CCD ASTRO-H
1. Introduction Charge-coupled devices (CCDs) are widely used for X-ray
n Corresponding author at: Department of Earth and Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 5600043, Japan. E-mail address: [email protected] (S. Inoue).
astronomy (e.g. [1,2]). The ASTRO-H satellite [3], which is the 6th Japanese X-ray observatory, includes an X-ray CCD camera called the soft X-ray imager (SXI, [4–6]). The SXI contains four CCD chips, each with an imaging area of 31 mm × 31 mm , arrayed in mosaic, covering the field-of-view area of 38′ × 38′, the widest ever flown in orbit. The original pixel size is 24 μm × 24 μm , but we operate it with 2 × 2 pixel binning to reduce the number of readout channels while preserving enough
http://dx.doi.org/10.1016/j.nima.2016.03.071 0168-9002/& 2016 Published by Elsevier B.V.
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
2
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
resolution. The N-type silicon used for the SXI yields a depletion layer thickness of 200 μm . The CCDs are of back-illuminated (BI) type, which gives the SXI enough tolerance against micro-meteorites and high quantum efficiency both at low and high energy bands. Fig. 1 shows the schematic view of the cross-sectional surface of the SXI. As with the CCD property shown in previous studies (e.g. [7– 10]), the response function of the SXI for monochromatic X-ray photons is reproduced with 5 components: primary Gaussian, secondary Gaussian, constant, Si escape and Si K components (Fig. 2). The origin of each of these components is explained by the following occurrences. When all the charge induced by an X-ray is generated in the depletion layer and collected at the gate, the event is detected as the primary Gaussian component. Sometimes, the charge cloud is split over the neighboring pixels. When spread charge is below the threshold, the event is detected as the secondary Gaussian component. The event appeared as the constant component is generated when a part of charge is lost near the boundary of the depletion layer. A Si K fluorescent X-ray (e.g. Si K α 1.739 keV) following the photon absorption sometimes escapes from the pixel. This event is detected as the Si escape peak and its peak energy is 1.739 keV (Si K α) lower than the primary peak. The Si K component includes the following two cases: (1) an X-ray absorbed at the SiO2 layer is re-radiated as Si K and (2) Si K due to a Si escape event is re-absorbed outside the pixel. Empirically, these 5 components are described as four Gaussians and a constant.
2. Calibration experiments To evaluate the X-ray spectra obtained with the SXI in orbit, an accurate ground calibration of its response function is essential. For this purpose, we performed two calibration experiments: flight model (FM) SXI camera calibration (Fig. 3) and KEK experiments with test CCDs (Fig. 4). The FM camera calibration was performed at Kyoto University from August to September in 2014. By irradiating various targets with α -rays from 241Am, we generated fluorescent X-rays of Ge K α (9.9 keV), Ge L (1.2 keV) and F K α (0.68 keV). We also irradiated CCDs with the Mn K α (5.9 keV) by a 55Fe source. In December 2014, we performed the calibration experiment at the KEK Photon Factory (KEK-PF), where monochromatic X-rays are available. In this experiment, we used mini-sized CCDs that we fabricated in the development stage of the SXI using the same process as the FM. The difference between FM and mini-sized CCDs is in the size, the number of pixels and that of readout nodes. The X-ray energies of 0.50 keV, 0.70 keV, 1.83 keV and 1.85 keV are available in this experiment.
3. Construction of the response function We construct the response function for the SXI using these calibration data. Before modeling the function, we correct the effects of charge transfer inefficiency and charge trailing (see [11]). These are the effects that the collected charge is lost during the charge transfer process, and we then correct the pulse heights of the X-ray line centroids. Using these corrected data, we fit each spectrum with the five components we describe in Section 1. In this fitting we leave most of the parameters free, while the width and center is fixed for some weak lines (e.g., Ni K β , the escape line of Ge K β ). Fig. 5 shows the spectrum taken for the FM camera calibration. In addition to the incident X-rays from the source, there are some other unexpected X-rays (e.g. Cr K α , Fe K α ). We fit each of these components with a single Gaussian template. After estimating the parameters for the SXI response, we examine the energy dependence of these parameters. In Table 1, we summarize the X-ray lines we used. Only in the analysis of the Primary center peak and that of the Primary FWHM, we include not only F K α , Ge L, Mn Kα and Ge K α , but also Mn K α Si escape,
Fig. 1. Schematic cross-section of the SXI CCD. Incident layer of the X-ray is the upper plane of this figure.
Fig. 2. Line profile of the SXI. Each component reproduces the line profile explained in Section 1.
Fig. 3. FM camera calibration at Kyoto University. The FM SXI camera is in the red circle. X-ray sources are attached at the top of the body. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
3
Table 1 X-rays we use to examine energy dependences of parameters of SXI response function. Energy values are weighted-averaged value for several lines (e.g., Kα1, Kα2). We use the KEK data only for the constant intensity ratio to the primary intensity. X-ray
Fig. 4. Calibration experiment at KEK-PF. We use a camera body and refrigerator different from those of the FM, while the CCD is almost the same as the FM. A spectroscope at the upstream side extracts monochromatic X-rays.
Energy (keV)
FM calibration at Kyoto FK α Ge L Mn K α Ge K α Mn K α Si escape Cr K α Fe K α Fe K β Ni K α KEK experiment — — — —
0.6768 1.199 5.895 9.876 4.158 5.412 6.399 7.058 7.472 0.50 0.70 1.83 1.85
Pulse Height [ch]
1500
1000
500
Residual [ch]
0 2 0 −2 0
2
4
6
8
10
8
10
Energy [keV] 250
Primary FWHM [eV]
200
150
100
50
0 Fig. 5. Mn K α (upper panel) and Ge K α (lower panel) spectra of FM camera calibration at Kyoto University. In the fitting of the Mn K α data, the Ti K α line and the escape line of Mn K β shown as the residual between ∼ 700 and ∼ 850 channel are not taken into account.
Cr K α , Fe K α , Fe K β and Ni K α , for which we fit with a single Gaussian. Fig. 6 shows the energy dependence of the center peak (upper panel) and that of the FWHM (lower panel) of the primary component. For the center peak of the primary component, we model its energy dependence by a linear function. Typical
0
2
4
6 Energy [keV]
Fig. 6. Energy dependence of the peak channel (upper panel) and primary FWHM (lower panel). All data points are those taken by the FM calibration at Kyoto University.
discrepancies are within ∼2 ch between the data and the model. For the FWHM of the primary component, we employ the following function:
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
F ·W ·EX + N 2 + U ,
(1)
where EX is the energy of the incident X-ray, F is the Fano factor [13], W is the mean ionization energy in silicon, N is the readout noise and U is the unidentified term to originate in the thermal noise or other noise. We adopt F = 0.115 [14], W = 3.65 eV/e− [12] in this paper. While discrepancies between the data and the model are <20 ch , only for high statistic data point (F K α , Mn K α , Ge K α ) discrepancies are within ∼4 ch . In addition to the center peak and the FWHM of the primary component, we model energy dependences of intensity ratio of the secondary, constant, Si escape and Si K components to the primary component, and that of FWHM ratio of the secondary component to the primary component.
100 normalized counts s−1 keV−1
Primary FWHM =
MnKα
MnKβ
10 CrKα
MnKα Si escape
1
FeKβ
0.1 0.01 10−3 10−4 4
ratio
4
2 0
2
3
4
4. Unveiled features of the SXI
5
6
7
Energy (keV)
We find some interesting properties in our model of the SXI response function which we describe here. normalized counts s−1 keV−1
Fig. 7 shows the energy dependence of the integrated intensity ratio of the constant component to the primary component. For the constant component, we integrate the flat intensity from the first PHA channel to the peak of its primary line. Since it is difficult to estimate the constant intensity due to the contamination of other X-ray lines, we use the data of the KEK experiment in the low-energy band. To reproduce the observed data, we perform Monte-Carlo simulations using Geant 4 [15]. We construct a geometry in which the CCD has a layer structure as shown in figure Fig. 1 and input X-ray photons with incident directions perpendicular to the sensor plane. The red line in Fig. 7 represents the result of the simulation for the intensity ratio, which cannot reproduce the data in the low energy band. We then perform the simulation by assuming a charge-trapping layer between the depletion and oxidation layers. To reproduce the data, we assume that X-rays in this layer contribute to the constant component with flat energy distribution. With the charge-trapping layer thickness of 20 nm, we can reproduce the observed intensity ratio. We find that we need to take this layer
MnKα
MnKβ
10 CrKα
MnKα Si escape
1
FeKβ
0.1 0.01 10−3 10−4 4
ratio
4.1. Existence of a trap layer
100
2 0
2
3
4
5
6
7
Energy (keV) Fig. 8. Fitting result with the incident X-ray model convolved by SXI response we modeled as the best estimation. The upper panel shows the fitting result with the assumption of a single delta function model for a X-ray line. The lower panel shows the fitting result with a detailed X-ray model for the incident X-rays. The black and red dot lines represent the model components for Mn K α and those for Mn K β . Green dot-dash lines are additional line components approximated with a single gaussian model (Ti K α : 4.509 keV, Cr K α : 5.412 keV, Fe K α : 6.399 keV, Fe K β : 7.058 keV). The blue solid line represents the unidentified component expressed in Eq. (2). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
0.1
Constant / Primary
w/ trap layer
into account when we model the constant component of the SXI. We have not yet identified whether this layer is originated in this particular CCD model or common for some types of BI-CCDs. 4.2. Detailed incident X-ray model
0.01
10−3
Kyoto Cal KEK
w/o trap layer 0.5
1
2
5
10
20
Energy [keV] Fig. 7. Energy dependence of the intensity ratio of the constant component to the primary component. Red and blue lines show the model without a trap layer and that with a trap layer, respectively. The FM calibration data taken at Kyoto (green) are shown with the uncertainties including the systematic error due to the constant emission of the intrinsic X-ray. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
We fit the SXI spectra by using the response function we model as the best estimation. Fig. 8 shows the result for the Mn K α spectrum. The upper panel shows the result when we assume each X-ray line has a delta function profile. It fails to reproduce the spectrum around ∼ 5.5 keV, ∼ 6.2 keV and ∼ 6.8 keV which correspond to the boundaries between lines. We know that Mn K α and Mn K β emission consists of several lines (e.g. Mn K α1, Mn K α2, Mn K α3 etc.), and each of these lines is broadened with a Lorentzian profile [16]. By referring to the 55Fe spectrum taken with an X-ray micro-calorimeter (ΔE ∼ 5 eV , [17,18]), we model the line profile of Mn K α and Mn K β . We employ seven and five Lorentzian components listed in Tables II and III of [16] and one more Lorentzian component for Mn K α . In this modeling we introduce an extra component between the Mn K α and Mn K β whose intensity is
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i
S. Inoue et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
H satellite, planned to be launched in early 2016. To model the SXI response function before launch, we performed the ground calibration for the SXI. Using these data, we construct the SXI response and find the following:
10
Counts s−1 keV−1
1
1. The observed intensity of the constant component for the SXI is not reproduced by the model expected from the design of the SXI CCDs. We find that the data are well reproduced if we assume a charge-trapping layer with a thickness of 20 nm between the depletion layer and the oxidation layer. 2. The Si K component of the SXI is weaker than that of a FI-CCD (e.g. XIS onboard the Suzaku satellite). One reason is that the detection of the re-radiated Si K α from the illumination side occurs less frequent for BI-CCD than that of the FI-CCD. The other reason is the effectively larger pixel size of the SXI ( 48 μm ).
0.1
0.01
10−3
10−4
5
1
2
3
4
5
6
Energy (keV) Fig. 9. Simulated spectrum of the SXI (red) and that of the XIS (FI-CCD, black) for the monochromatic X-rays (5.9 keV). (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
expressed by an empirical function as
⎧ a exp ((EX − b)/c ) (EX ≤ b) Intensity = ⎨ (EX > b) . ⎩0
(2)
Such a component has never been reported before and we cannot identify its origin. With this incident X-ray model, we can reproduce the SXI spectrum with our response function as shown in the lower panel of Fig. 8, while we have no idea to reproduce the spectrum at ∼ 5 keV. 4.3. Si K component Fig. 9 shows the simulated spectra of monochromatic X-rays by the SXI and the front-illuminated CCD of the Suzaku X-ray Imaging Spectrometer (XIS, [19,20]). Although the Si escape lines show similar intensities between the two CCDs, the Si K α line of SXI response is significantly weaker than that of the XIS because of two reasons. One is the thickness of the dead layer in front of the depletion layer. The FI-CCD has a relatively thicker dead layer at the illumination side. The other reason is their effective pixel size. The effective pixel size of the SXI is 48 μm by 2 × 2 binning, while that of XIS is 24 μm . Since the attenuation length of Si K α is ∼12 μm , the possibility that Si K α passes through the effective pixel size for the SXI is ∼ 1.8%, while that for the XIS is ∼ 14%.
5. Summary
Acknowledgments We thank all of the members of SXI team. S.I. is supported by Research Fellowships of Japan Society for the Promotion of Science for Young Scientists. This study was also supported by JSPS KAKENHI Grant numbers 15J01845, 23340071, 26109506, 15H03641, 26670560, 23000004, 24684010, 23340047, 25109004, 15H02090, 25870347, 26800102, 24740123, 24740167, 23740199, 15K17610.
References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]
S. Sembay, et al., Proc. SPIE 5488 (2004) 264. C.E. Grant, et al., Proc. SPIE 6276 (2006) 48. T. Takahashi, et al., Proc. SPIE 8443 (2012) 84431Z. H. Tsunemi, et al., Proc. SPIE 7732 (2010) 11. K. Hayashida, et al., Proc. SPIE 8443 (2012) 23. K. Hayashida, et al., Proc. SPIE 9144 (2014) 914429. S. Ueda, et al., Proc. SPIE 8145 (2011) 814504. S. Ueda, et al., Nucl. Instrum. Methods Phys. Res. A 704 (2013) 140. H. Yamaguchi, et al., Proc. X-ray Universe (2006) 2005 (ESA SP-604. 949). H. Nakajima, et al., Nucl. Instrum. Methods Phys. Res. A 541 (365) (2005) 371. K.K. Nobukawa, et al., Nucl. Instrum. Methods Phys. Res. A 754 (269) (2014) 274. J.R. Janesick, Scientific Charge-Coupled Devices, SPIE Press, Bellingham, Washington, 2001. U. Fano, Phys. Rev. 72 (1947) 26. R.C. Alig, et al., Phys. Rev. B 22 (1980) 5565. S. Agostinelli, et al., Nucl. Instrum. Methods Phys. Res. A 506 (250) (2003) 303. G. Hölzer, et al., Phys. Rev. A 56 (1997) 4554. K. Mitsuda, et al., Proc. SPIE 7732 (2010) 773211. K. Mitsuda, et al., Proc. SPIE 9144 (2014) 91442A. K. Koyama, et al., Publ. Astron. Soc. Jpn. 59 (2007) 23. K. Mitsuda, et al., Publ. Astron. Soc. Jpn. 59 (2007) 1.
We developed the X-ray CCD camera, SXI, onboard the ASTRO-
Please cite this article as: S. Inoue, et al., Nuclear Instruments & Methods in Physics Research A (2016), http://dx.doi.org/10.1016/j. nima.2016.03.071i