Nuclear Instruments and Methods in Physics Research A 477 (2002) 155–160 Application of a finite size of the charge cloud shape generated by an X-ray photon inside the CCD H. Tsunemia,b,*, J. Hiragaa,b,1, E. Miyataa,b a Department of Earth and Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 5600043, Japan b CREST, Japan Science and Technology Corporation (JST), Japan Abstract A mesh experiment enables us to specify the X-ray landing position on a charge-coupled device (CCD) with subpixel resolution. By this experiment, we find that the final charge cloud shape generated by Ti–K X-ray photons (4:5 keV) in the CCD is about 1:5 1:1 mm2 (standard deviation). An X-ray photon photoabsorbed in the CCD generates a number of electrons, forming an X-ray event. It becomes up to a 4-pixel-split event since the pixel size of the CCD used (12 mm square pixel) is bigger than the charge cloud size. Using the mesh experiment, we can determine the X-ray landing position on the CCD. In this way, we can compare the estimated X-ray landing position with the actual landing position on the CCD. Employing the charge cloud shape, we can improve the position resolution of the X-ray CCD by referring to the X-ray event pattern. We find that the position accuracy of our method is about 1:0 mm: We discuss our method, comparing it with the charge centroid method. r 2002 Elsevier Science B.V. All rights reserved. PACS: 07.85.m; 29.40.Wk; 95.55.Aq Keywords: Mesh experiment; Charge-coupled device; Charge centroid method; Cloud shape method 1. Introduction A charge-coupled device (CCD) has relatively good spatial resolution and high quantum efficiency for both X-ray and optical regions. It consists of many small pixels about 10 mm in size. Each pixel consists of 2–4 electrodes which *Corresponding author. Department of Earth and Space Science, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 5600043, Japan. Tel.: +81-6-850-5477; fax: +81-6-850-5539. E-mail address: [email protected] (H. Tsunemi). 1 Partially supported by JSPS Research Fellowship for Young Scientists, Japan. are responsible for the nonuniformity of the detection efficiency over the pixel. The output of the CCD comes from each pixel, resulting in a spatial resolution of pixel size. X-ray photons photoabsorbed in the CCD generate a number of electrons of about E=3:65 eV; where E is the incident X-ray energy. These electrons, generated in a very small region [1], become a finite-size charge cloud through diffusion when they are collected into the potential well under the electrodes [2]. They are called ‘an X-ray event’, forming an island, i.e., a series of connecting pixels having signals. The pixel with the largest signal in the X-ray event is called the ‘event pixel’. 0168-9002/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 9 0 0 2 ( 0 1 ) 0 1 8 7 7 - 0 156 H. Tsunemi et al. / Nuclear Instruments and Methods in Physics Research A 477 (2002) 155–160 When an X-ray photon is photoabsorbed in the field-free region, under the depletion layer, the charge cloud expands and forms a multipixel event consisting of more than four pixels. In contrast, when it is photoabsorbed in the depletion layer, the entire charge is collected into the potential well with relatively small spread. When the X-ray landing position is far from the CCD pixel boundary, the entire charge is collected into one pixel. When it is close to the CCD pixel boundary, the charge splits into 2–4 pixels depending on the landing position. Thus, X-rays photoabsorbed in the depletion layer form various types of event patterns. A mesh experiment is introduced to measure the X-ray responsivity with subpixel resolution [3]. Using this method, we can specify the X-ray landing position within the pixel for all X-ray events [10,11]. Then, we developed a method of directly measuring the charge cloud shape generated by X-ray photons [4]. Precise knowledge of the cloud shape enables us to determine the X-ray landing position with subpixel resolution using split pixel events. In this paper, we report experimental results for the improvement of the position resolution of the CCD. mine the mutual alignment between the mesh and the CCD. In this way, we can obtain the CCD pixel map for the X-ray responsivity. The detailed explanation of the mesh experiment is given in the literature [3,5]. The CCD pixel map obtained shows the pixel structure convoluted with the mesh hole shape. Therefore, the accuracy of the pixel structure is limited by the mesh hole size. Similarly, we can easily restrict the X-ray landing position on the CCD with CCD-pixel-size precision if we assume that the X-ray landing position is somewhere inside the event pixel. In this way, we can obtain the mesh hole map of the X-ray transmission. The mesh hole map obtained shows the mesh hole structure convoluted with the CCD pixel shape. The CCD pixel shape is an exact square. Thus, we can estimate the effective mesh hole size [6]. Fig. 1 shows a mesh hole shape convoluted with the CCD pixel shape. Since the CCD pixel shape is much bigger than the mesh hole shape, the result mainly reflects the CCD pixel shape. Once we determine the X-ray landing position with subpixel resolution, we know what type of X-ray events are generated according to the X-ray landing position. This enables us to measure the charge cloud shape generated by an X-ray photon. 2. Experimental setup and image restoration The mesh experiment consists of three parts: a parallel X-ray beam, a CCD operating in the X-ray photon counting mode and a metal mesh placed above the CCD. The CCD used has a 12 12 mm2 pixel size and was manufactured by Hamamatsu Photonics Inc. We employed a gold mesh of about 10 mm thickness with small holes of about 2 mm diameter spaced 48 mm apart. These parameters are each mesh dependent, and were measured for the actual mesh by both using the scanning electron microscopic images and measuring the X-ray transmission. Placing the mesh just above the CCD, we can restrict the X-ray landing position with mesh-hole-size precision, since the raw data show a moire! pattern which comes from the interaction between the periodically spaced CCD pixels and periodically spaced mesh holes. The moire! pattern enables us to precisely deter- Fig. 1. Accuracy of the X-ray position determination based on the event pixel method. Four shapes are shown separated by 48 mm; while the CCD pixel size is 12 12 mm2 : H. Tsunemi et al. / Nuclear Instruments and Methods in Physics Research A 477 (2002) 155–160 The result is a convolution between the actual charge cloud shape and the mesh hole shape [4]. Fig. 2 shows a charge cloud shape obtained in this way using Ti–K X-rays (4:5 keV) [7]. It can be well expressed by an axial symmetric Gaussian function of s ¼ 1:1 mm along the charge transfer direction and s ¼ 1:5 mm perpendicular to it where s is the standard deviation. 3. Determination of the X-ray landing position inside the CCD 3.1. Methods of estimating the X-ray landing position When we detect X-ray events on the CCD, we assume that the X-ray landing position is in the event pixel. This is confirmed since the charge cloud shape is well expressed by a Gaussian function. There are three methods of estimating the X-ray landing position for each X-ray event. The simplest method, the ‘event pixel method’, is to employ the center of the event pixel as the X-ray landing position. It is obvious that the accuracy of the position determination depends on the pixel size which is shown in Fig. 1. Fig. 2. Charge cloud shapes for Ti–K X-ray photons are shown in a 12 12 mm2 square [7]. A linear contour is overlaid. 157 The second method, the ‘charge centroid method’, is to employ the center of gravity of the Xray event. Yoshita et al. [8] studied the relation between the center of gravity of the X-ray events and the actual X-ray landing position on the CCD. They found that the center of gravity is a good indicator of the X-ray landing position, although they pointed out that the actual landing position should be determined taking into account the charge cloud shape. The third method that we employ, the ‘cloud shape method’, is to calculate the X-ray landing position based on the charge cloud shape. We can calculate the X-ray landing position so that the charge splits in such a way as to reproduce the event pattern. 3.2. Comparison between the charge centroid method and the cloud shape method There are several event patterns generated by X-ray photons. The event pixel method is applicable to any event pattern, whereas the position accuracy is limited by the CCD pixel size. The twopixel-split events can yield detailed positional information along the split direction but little information perpendicular to it. If we must determine the X-ray landing position in both X and Y directions, we must used split pixel events in both directions. In practice, we used 3and 4-pixel-split events which are sure to occur in the four corners of the pixel [5,6] and calculated the center of gravity and also estimated the X-ray landing position using the charge cloud shape. In this calculation, we simply assume that the charge cloud shape is well expressed by a point-symmetric Gaussian function of s ¼ 1:3 mm; for simplicity. The actual cloud shapes measured are slightly distorted but this does not seriously affect our results. We calculated the relation between the mesh hole center through which the X-ray photon entered and the estimated X-ray landing position in the pixel. The results using the charge centroid method are shown in Fig. 3 and those using the cloud shape method are shown in Fig. 4. The horizontal axis shows the position of the mesh hole center on the CCD pixel. At 0, the mesh hole center just coincides with the pixel boundary to the 158 H. Tsunemi et al. / Nuclear Instruments and Methods in Physics Research A 477 (2002) 155–160 Fig. 3. Relationship between the mesh hole center and the estimated X-ray landing position based on the charge centroid method. Left: the horizontal direction. Right: the vertical direction. Unit is given in pixel size; see text. Fig. 4. Same as for Fig. 3 but based on the cloud shape method. neighboring pixel. Since the physical situation is symmetrical with respect to the center of the pixel, the horizontal axis starts and ends at half the pixel size. The vertical axis shows the estimated X-ray landing position inside the pixel. Since we only employed 3- and 4-pixel-split events, all the events used appear only near the pixel boundary. It is clear that the charge centroid method does not yield the precise landing position. This is due to the fact that the charge centroid method can produce the correct answer only when the charge cloud size is comparable to or larger than the CCD pixel size. The data spread along the horizontal axis in Fig. 4 is a result of the finite size of the mesh hole. 159 H. Tsunemi et al. / Nuclear Instruments and Methods in Physics Research A 477 (2002) 155–160 Fig. 5. Accuracy of position determination for the charge centroid method (left) and for the cloud shape method (right). The four peaks in both figures are separated by 48 mm: As we explained briefly in Section 2, we can reconstruct the mesh hole shape using the X-ray landing position determined by either the charge centroid method or the cloud shape method. Fig. 5 shows the convolution between the point spread function (PSF) of the two methods and the mesh hole shape. We analyzed the PSF and summarize its results in Table 1. 4. Discussion and conclusion We confirmed that the cloud shape method yields better results than the charge centroid method or the event pixel method. The position accuracy of this method shown in Fig. 5 is s ¼ 1:170:1 mm: If we exclude the effect of the finite size of the mesh hole, we obtain s ¼ 1:070:1 mm for the cloud shape method which is smaller than that of the charge cloud. The charge cloud size depends on the depth at which the photoabsorption occurs. The actual charge cloud size shows scatter up to a few mm; and the charge cloud size employed is the average of various X-ray events. This will produce some ambiguity in determining the X-ray landing position. Table 1 s (mm) of the results of the two methods using Ti–K X-raysa Method Charge centroid Cloud shape Data (Fig. 5) PSFb 2:470:1 2:370:1 1:170:1 1:070:1 a b Note: Quoted errors are 90% confidence level. Excluding the effect of mesh hole. The cloud shape method will work well only when X-rays produce split events. Taking into account the measured charge cloud size, the CCD pixel size of 6 mm square will be small enough to make most X-ray events split pixel events so that we can improve the position resolution over the entire area. The single-pixel events are produced when the landing position is somewhat away from the pixel boundary so that all the charge is collected into one pixel. Therefore, we can improve the landing position for single-pixel events to slightly better than the pixel size. The detailed analysis for all X-ray events will be carried out elsewhere. We measured the final charge cloud shape generated by Ti–K X-ray photons (4:5 keV). Then we applied our results to improve the position determination. The cloud shape method intro- 160 H. Tsunemi et al. / Nuclear Instruments and Methods in Physics Research A 477 (2002) 155–160 duced here shows a PSF of s ¼ 1:0 mm: This method yields more than twofold better results than the charge centroid method. Our method is applied only for split pixel events, whereas there is some improvement in the position determination for single-pixel events. It will be effective for the Chandra ACIS since its image quality is better than the CCD pixel size [9]. Acknowledgements The authors are grateful to all members of the CCD team in Osaka University. They also thank Mr. K. Miyaguchi of Hamamatsu Photonics Inc. for technical support. This research is partially supported by Kurata Research Grant, the Grandin-Aid for Scientific Research by the Ministry of Education, Culture, Sports, Science and Technology of Japan (13440062). References [1] T.E. Everhart, P.H. Hoff, J. Appl. Phys. 42 (1971) 5837. [2] G.R. Hopkinson, Opt. Eng. 26 (1987) 766. [3] H. Tsunemi, K. Yoshita, S. Kitamoto, Jpn. J. Appl. Phys. 36 (1997) 2906. [4] J. Hiraga, H. Tsunemi, K. Yoshita, E. Miyata, M. Ohtani, Jpn. J. Appl. Phys. 37 (1998) 4627. [5] K. Yoshita, H. Tsunemi, K.C. Gendreau, G. Pennington, M.W. Bautz, IEEE Trans. Nucl. Sci. NS-45 (1998) 915. [6] H. Tsunemi, J. Hiraga, K. Yoshita, E. Miyata, Jpn. J. Appl. Phys. 38 (1999) 2953. [7] H. Tsunemi, J. Hiraga, K. Mori, K. Yoshita, E. Miyata, Nucl. Instr. and Meth. A 436 (1999) 32. [8] K. Yoshita, H. Tsunemi, K.C. Gendreau, M.W. Bautz, IEEE Trans. Nucl. Sci. NS-46 (1999) 100. [9] M.C. Weisskopf, S.L. O’Dell, Proc. SPIE 3113 (1997) 2. [10] H. Tsunemi, J. Hiraga, K. Yoshita, S. Kitamoto, Japan. J. Appl. Phys. 37 (1998) 2734. [11] M. Pivovaroff, S. Jones, M. Bautz, S. Kissel, G. Prigozhin, H. Tsunemi, E. Miyata, IEEE Trans. NS-45 (1998) 164.
© Copyright 2025 Paperzz