6. Xtend/SXI

6.1 Xtend Basics

Xtend is the soft X-ray imaging telescope of XRISM , with a pair of an X-ray mirror (XMA) and an X-ray CCD imaging spectrometer (Soft X-ray imager: SXI, Hayashida et al., 2018). Xtend has a larger field of view (38$'\times$38$'$) and a low/stable non-X-ray background, ideal for observing diffuse X-ray emission with low surface brightness. The large field of view (FoV) and the finer pixel size (1.$''$74) also provide more detailed spatial information of or around the target, extending the capabilities of Resolve's spectroscopy by evaluating contamination from sources outside Resolve's FoV and measuring the local diffuse background (Figure 6.1). Xtend works in a complementary role to Resolve.

Besides a few improvements, Xtend is almost identical to the Astro-H SXT + SXI telescope system (see Section 6.4.6). Therefore, its publications and archival data should be a good reference for Xtend 's performance. This chapter describes the SXI module and the combined performance of the XMA and SXI. Chapter 4 describes the XMA-specific performance, which has an identical design to Resolve's XMA telescope but with a slightly different performance.

Figure 6.1: X-ray images of the Perseus cluster obtained with the Hitomi observatory (Nakajima et al., 2018). (a) Full X-ray CCD (SXT+SXI) image overlaid with the micro-calorimeter (SXT+SXS) FOV in cyan. (b) Magnified CCD image of the central region overlaid with the SXS pixel boundaries in cyan. (c) Micro calorimeter (SXT+SXS) image of the same region.

The SXI houses four identical X-ray CCD sensors in a 2$\times$2 grid. Each CCD has an identifier, CCD_ID=0$-$3^6.1, used for the science and housekeeping data (Figure 6.2). These CCDs are similar to those on earlier or ongoing X-ray observatory missions, such as Suzaku XIS, Chandra ACIS, and XMM-Newton EPIC-MOS. However, the SXI CCDs utilize the p-channel, backside-illumination technology, which realizes a thick depletion layer of 200 $\mu$m, with high quantum efficiencies in a broad energy range between 0.4$–$13 keV. The following list describes the key instrument performance. Table 6.1 also summarizes Xtend 's characteristics.

Figure 6.2: Left — Schematic layout of the SXI CCDs looking toward the sky from the back of the focal plane. The 4 CCD chips are aligned in a 2$\times$2 format. Each CCD has two electrically separated segments, whose boundary is shown by the dashed line. Each segment has two readout nodes for redundancy (the black dots with the A/B/C/D labels). Each chip has an additional frame store area on the readout node side. The filled gray half circles show approximate areas illuminated by the calibration sources. The red box shows the SXS FOV and the dotted orange circles show the off-axis angles. The distances from the CCD edges are inflight measurement values. There is a tiny area on the far side corner from the aim point shown in yellow, where a camera body structure blocks the view of the sky. Right — SXI CCD sensor photo (Nakajima et al., 2020, Fig. 4). Gold-coated metal frames cover the frame store areas.

Table 6.1: Xtend Characteristics

Field of View	38$'$$\times$38$'$ (Full window)
Sensitive band	0.4 $-$ 13 keV
Effective area	356 cm$^{2}$@1.5 keV, 307 cm$^{2}$@6 keV
On-axis XMA PSF at 6.4 keV $^{\dagger}$	1.47$'$ (HPD), 7.2$''$ (FWHM)
Pixel size	1.74$''$ (48$\mu$m, logical)
Time resolution	4 sec (Full window), 0.5 sec (1/8 window)
Energy resolution (FWHM)	$\sim$180 eV @ 6 keV
Pileup tolerance	2.5 mCrab (Full window)
Total (NXB + Sky) X-ray background $^{\ddagger}$	$\leq$10$^{-6}$ counts keV$^{-1}$ s$^{-1}$ arcmin$^{-2}$ cm$^{-2}$
HPD: Half Power Diameter.
$^{\dagger}$See the XMA chapter (4) for the details.
$^{\ddagger}$See Figure 6.4 right and Figure 18 of Nakajima et al. (2018) for the details.
These values are based on ground or in-flight calibrations.

Field of View: The 2$\times$2 CCD array covers a 38$'\times$38$'$ square FOV with small mechanical gaps between the CCD chips. The gaps between the active pixel regions are 44$''-$59$''$ (1.2$-$1.6 mm). The aim point co-aligns with the Resolve aim point and offsets by 5 or 5.4$'$ from the CCD edges of CCD_ID=1 to avoid the chip gap. In the 1/8 window mode, the 2 CCDs (CCD_ID=0, 1) have a narrow rectangular FOV with the aim point at the width center. The long side does not touch the CCD edges, having a relatively big imaging gap with the other 2 CCDs' FOV (Figure 6.5, Section 6.2).

Sensitive Energy Range & Quantum Efficiency: The SXI CCD sensors have a p-type channel and n-type substrate, enabling a thick depletion layer of $\sim$200 $\mu$m and realizing a large X-ray stopping power up to $\sim$13 keV. The CCDs are backside illumination CCDs with high transparency to soft X-rays. The SXI system also has the Optical Blocking Layer (OBL) on the CCD surface to block optical/IR lights and the Contamination Blocking Filter (CBF) at the hood top to avoid molecular contamination build-up and further block optical/IR lights. These components limit the soft X-ray sensitivity to $\sim$0.4 keV. Figure 6.3 left shows the SXI CCD quantum efficiency.

Effective Area: This value primarily depends on the XMA's effective area (see Section 4.2.2) and the CCD's quantum efficiency. Beside, the SXI runs with the charge injection operation through the mission, which produces 1$-$3 insensitive CCD rows every 80 rows and reduces the effective area by 1$-$3%. Figure 6.3 right shows the effective area of the Xtend system for a point source at on-axis. The effective area can decline with radiation damage or contamination condensation in the line of sight. The instrument and science operations teams will calibrate it regularly and update the result in the calibration database.

Angular Resolution: Each CCD has a square imaging area with 1280$\times$1280 physical pixels of 24 $\mu$m length each, which are binned by 2 pixels onboard during the science operation. The output science data have 640 “logical" pixels of 48 $\mu$m, equivalent to 1.74$''$ in the sky. This pixel size is significantly smaller than the on-axis PSF's half-power diameter (1.47$'$) and FWHM (7.2$''$) (see Chapter 4).

Figure 6.3: Left — SXI CCD Quantum efficiency measured with the ground experiments (black). The plot also separately shows the quantum efficiency of a CCD chip combined with the transmission of the Optical Blocking Layer (QE + OBL, red) and the transmission of the Contamination Blocking Filter (CBF, green). Right — Effective area of the Xtend telescope, including the XMA component.

Time resolution: The nominal operation (the Full window mode without the burst option, see Section 6.2) takes a frame exposure for 3.96 seconds and then moves the obtained data to the frame store area in 36.864 milli-second for readout. The imaging area takes another exposure right after this transfer, so, this operation continuously obtains data every 4 seconds. The 1/8 window mode without the burst option runs this cycle every 0.5 sec. The 1/8 window mode with the burst option uses only a fraction of each 0.5 sec exposure. The data-collecting cycle does not increase, but users can use its narrow exposure interval information for pulsar analysis, for example.

Energy resolution: The CCD energy resolution depends on the statistical fluctuation of charges produced by X-ray photons, the number of charges lost during the CCD charge transfer (charge transfer inefficiency: CTI), and the readout electric noise. Normally, the first component, charge statistical fluctuation, is the most significant factor in the CCD energy resolution. Higher energy X-rays have worse absolute energy resolution (${\Delta}E$) with more charges but better relative energy resolution (${\Delta}E$/$E$) with better statistics. The ground calibration verified the energy resolution of $\sim$180 eV at 5.9 keV (FWHM). The CTI normally gets worse through the mission with in-orbit radiation damage. The instrument and science operations teams will calibrate the energy resolution regularly with the $^{55}$Fe calibration sources and/or celestial source observations and updates the result in the calibration database.

Background: Cosmic Ray (CR) particles or fluorescent X-rays from detector materials excited by those particles interact with the CCD sensors and produce charge clouds in CCD pixels. The grade selection method effectively excludes those events but selected (cleaned) data still include some of them called Non-Xray Background (NXB). The NXB flux changes with cutoff rigidity in orbit, whose distribution varies with the 11-year Solar cycle. (Figure 6.4 left, Nakajima et al., 2018). The amount of NXB depends on the exposed duration of each pixel on a CCD and, therefore, the CCD rows (vertical pixel coordinates). The instrument and science operations teams will study the NXB contribution to cleaned data under various observatory environments and provide a tool that evaluates the NXB spectra during individual observations. The sky background contribution is expected to be similar to the Hitomi SXI at 5.6$\times$10$^{-6}$ counts s$^{-1}$ arcmin$^{-2}$ cm$^{-2}$ in the 5$-$12 keV band (Figure 6.4 right, Nakajima et al., 2018).

Figure 6.4: Left — Hitomi SXI NXB spectrum normalized by the physical pixel area. The plot also shows a Suzaku XIS-BI NXB spectrum for the same physical pixel area. Right — Hitomi SXI total (NXB + sky) background spectrum normalized by the effective area and solid angle of the sky. The plot also shows the CXB spectrum and background spectra of other X-ray CCD instruments (Nakajima et al., 2018).

6.2 Observation Options for General Users

Users can choose the following optional operation modes for CCD_ID=0 and 1 to observe bright sources and/or improve time resolution (Table 6.2). These two sensors run in the same mode as they share the CCD sequences, while the CCD_ID=2&3 sensors always run with the Full window mode.

Figure 6.5: Simulated SXI image of the Perseus cluster with the 1/8 window mode. The green boxes show the approximate chip boundaries.

Table 6.2: Available Window/Burst Mode Options

	$N$	Image Size	Time Res.	Exp. time	LTF	Pile-up
		(H) $\times$ (V)	(sec)	(sec)		(mCrab)
Full Window No Burst	1	640$\times$640	4	3.9631	0.99	2.5
1/8 Window No Burst	8	640$\times$ 80	0.5	0.4631	0.93	21
1/8 Window 0.1s Burst	8	640$\times$ 80	0.5 $^{\dagger}$	0.0620	0.12	160
$N$: Number of exposure in a unit time interval (4 sec).
Image Size: each CCD chip size.
LTF: Live Time Fraction.
Pile-up: On-axis source flux that suffers 10% photon pile-ups, equivalent to an SXI count rate at 6.3 cts s$^{-1}$.
$^{\dagger}$Users can measure each photon's arrival time information with $\sim$0.06 sec accuracy.

Window option

The window option reads only a divided portion of the CCDs. The guest observer's program supports the 1/8 window option, which reads only 80 CCD rows for one exposure. This operation reduces the readout time to 1/8, so the exposure cadence, and therefore the time resolution, increase by 8 (see Figure 6.5).

Burst option

The burst option collects events obtained during a fraction of the nominal exposure time. This mode reduces photon pileups of bright sources with short-frame exposures. However, it only takes a single exposure in an exposure cycle and trashes events outside of the exposure interval. So, it does not increase the time resolution. The product data have time stamps of the beginning of each exposure so that users can refine photon arrival times for timing analysis. The guest observing program supports the 0.06 sec exposures (12% duty cycle) for the 1/8 window mode.

How to Choose the Observing Mode

Many point sources optimal for Resolve observations would exceed the Full window mode pileup limit. Even some diffuse sources like Perseus cluster, Pup A and Cas A have bright knots that go beyond the pileup limit (see e.g., Figure 3 in Tsunemi et al., 2013). Which Xtend mode should users choose?

The answer depends on the goal of their Xtend observation. If users want to collect photons only from the aim point target as many as possible, they should choose the 1/8 window mode and add the burst option if the target is brighter than $\sim$21 mCrab. However, the pileup occurs only near the PSF core, so users may be able to utilize data from the extended PSF outskirts with little photon pileups using the Full window mode. The HEASOFT tool, pileest, would help estimate pileup fraction from obtained event data.

6.3 Things to Be Considered for Bright Source Observations

6.3.1 Photon Pileup

Two or more X-ray photon events are indistinguishable when they fall in 3x3 CCD pixels in a single frame exposure; such an incident is called photon pileup. Many pileup events have two photons at different pixels, recognized as a multiple-pixel event and discarded, while others are just counted as single X-ray events. Either way, the SXI counts two or more pileup X-ray photons as one or zero, underestimating the photon flux. If detected as single X-ray events, pileup events have an energy of two or more X-ray photons, making the spectrum harder. The PSF core has the highest count rates per CCD pixel and prone to pileup. The PSF peak diminishes as the source flux increases and turns into a hole in a severe case.

The pileup fraction is the ratio of pileup events over all incoming X-ray events. Xtend SXI reaches the 10% pileup limit at a count rate of 6.3 cts s$^{-1}$, corresponding to $\sim$4.5 $\times$10$^{-11}$ ergs cm$^{-2}$ s$^{-1}$ between 2$-$10 keV for the Crab nebula spectrum with a power-law $\Gamma~$ at 2.13 (an IACHEC model^6.2). Pileup tolerance depends on users' science goals. For example, users should minimize the pileup to detect a weak hard spectral tail, while they may allow some degree of pileups to detect emission lines.

6.3.2 Out-of-Time Events

Out-of-time events are X-ray events detected during the CCD frame transfer — parallel transfer of charges from the imaging area to the frame store area. One logical pixel transfer takes only 57.6 $\mu$sec, so only bright sources appear as vertical streaks above the background in images. Out-of-time events rarely suffer photon pileups as detected charges stay in a fraction of time in a pixel, and so can be used for measuring non-pileup spectra of extremely bright sources.

In the normal clocking mode, out-of-time events constitute a uniform brightness streak and contribute only $\sim$0.93% of the total counts from a target. In the burst option, they constitute a significant fraction of the source counts with a relatively long frame transfer interval to the frame exposure, contributing 7.3% of the source count during the exposure.

6.3.3 Sacrificial Charge

Very bright sources create so many X-ray events that their charges also work as fillers of charge traps in the CCD chips (, ). This effect improves the spectral resolution, but it depends on the source brightness and is difficult to estimate. The instrument team does not plan to implement the effect in the standard calibration.

6.4 SXI in Depth

The following sections describe the detail of the SXI instrument. They may not necessarily be for proposal preparation, but they should help understand how the instrument works.

The SXI sensor (SXI-S) is placed at one of the two XMA's focal plane — 5.6 m below the mirror — on the XRISM's optical bench. The SXI instrument system consists of four components: a sensor body (SXI-S), pixel processing electronics (SXI-PE), digital electronics (SXI-DE), and a cooler driver (SXI-CD). The SXI-S (Figure 6.6) includes the CCDs and their supporting systems, driver and video electronics, a single-stage Stirling cooler, and calibration sources. The SXI-PE extracts X-ray event information from initial image data. The SXI-DE packages X-ray event and associated housekeeping (HK) data and delivers them to the Satellite Management Unit (SMU). The SXI-DE controls the Stirling cooler and associated heater, keeping the CCD sensor's temperature at $-$110 $^\circ$C (early operational phase), $-$120 $^\circ$C (late phase), or the desired temperature.

Table 6.3: SXI Characteristics

CCD Type	Frame transfer, p-channel, Backside illuminated
Number of CCDs	4
Number of segments per CCD	2
Imaging area	31 $\times$ 31 mm$^{2}$, 640 $\times$640 logical pixels
	(1280$\times$1280 physical pixels)
Pixel size	24 $\mu$m $\times$ 24 $\mu$m (physical pixels)
CCD gap	1.2$-$1.6 mm (44$-$59$''$)
Depletion depth	200 $\mu$m
Optical blocking layer	Al(100 nm)/Al(100 nm) deposited on the CCD surface
Contamination Blocking Filter (CBF)	Al(80nm)/Polyimide(200nm)/Al(40nm)
CCD operating temperature	$-$110 $^\circ$C/$-$120 $^\circ$C
Calibration sources	two $^{55}$Fe radioisotopes

Figure 6.6: External View of SXI-S

6.4.1 X-ray Detection Mechanics and Operation

Hamamatsu Photonics K.K. fabricated twelve flight-model candidate CCD chips, and the Xtend team selected four CCD flight-model chips among them (Yoneyama et al., 2021). The SXI CCDs have a p-type channel and n-type substrate, in which positively charged holes act as mobile charge carriers instead of electrons for earlier X-ray CCD cameras (See Figure 1 in Bebek et al. (2004)). The high resistivity of the n-type substrate produces a thick depletion layer so that the CCDs are fully depleted to the back surface with a depth of $\sim$200 $\mu$m. For comparison, the Suzaku XIS backside illuminated CCD with a p-type Si substrate was polished to 45 $\mu$m to ensure full depletion. The large SXI CCD depletion depth combined with the backside illumination improves the quantum efficiency at both soft and hard X-ray energies.

The larger depletion depth causes longer drift length and more extensive diffusion. This effect strongly affects soft X-ray photons absorbed near the CCD surface: because the CCDs are backside illuminated, their charges need to drift from near the surface to the bottom with the electrodes. X-ray event charges rarely fall in a single physical CCD pixel (24 $\mu$m $\times$ 24 $\mu$m) but multiple pixels. To simplify the event processing, the SXI nominally combines charges in every set of 2$\times$2 original pixels before the CCD readout and converts them to a digital value. This 2$\times$2 onboard binning format has $\sim$30% of X-ray events in a single binned pixel, and the rest within 3$\times$3 binned pixels. In the following, the "physical" pixel refers to the original 24 $\mu$m $\times$ 24 $\mu$m CCD pixel, while the "logical" pixel does the 2$\times$2 binned 48 $\mu$m $\times$ 48 $\mu$m, pixel. The SXI science data output in the 640 $\times$ 640 logical pixel format for each CCD.

Each CCD chip has a rectangular shape with a square imaging area (IA) and frame store (FS) area. The IA has 1280 $\times$ 1280 physical pixels, exposed to the sky through the XMA for X-ray photon detection. The FS also has 1280 $\times$ 1280 pixels but is covered with a metal plate and used solely for temporary charge storage. In a normal data acquisition sequence (Figure 6.7), the IA area detects X-ray photons in a frame exposure (3.96 sec in the normal clocking, Full window mode), whose charges quickly drift to the buried channel of the corresponding pixels. After the exposure, the CCD clock driver changes the electrode voltages (so-called clocking) to move the accumulated charges by one vertical pixel at one cycle and repeat this cycle 1280 times to transfer all charges in IA to FS. This “parallel transfer" process transfer all charges in the 1280 CCD columns in parallel. One pixel shift takes 28.8 $\mu$sec, and a whole parallel transfer from IA to FS takes 36.9 msec. During each parallel transfer, the SXI also operates charge injection for the subsequent frame exposure (Section 6.4.2 for the detail) so that the parallel transfer duration depends on the charge injection frequency and the clocking mode

After the parallel transfer, the CCD driver transfers charges in the bottom row horizontally to readouts with two serial registers. A serial register handles a CCD segment, moving charges in the bottom row by one horizontal pixel at one clock cycle toward the active readout node. An analog-to-digital converter measures the number of charges at the readout and stores the value in memory. Once the serial register reads charges in the whole pixels in the horizontal line, the driver vertically transfers all FS charges by one pixel and then read charges moved to the bottom row. The driver continues this sequence until reading all FS charges.

Figure 6.7: Schematic diagram of the charge transfer process on a CCD. (a) At the end of a frame exposure, X-rays or charged particles produce charge clouds around the reacted pixels (red dots). (b) The CCD clock driver transfers the charge clouds vertically to the FS region. It also injects artificial charges to the top serial register every $N$ vertical transfers ($N$ =80 in the prelaunch default setup), which move down to the IA with the subsequent vertical transfers (red lines). (c) Once all IA charges move to the FS area, the CCD clock driver only transfers charges in the bottom FS row horizontally for readout and then moves all FS charges down by one pixel for another horizontal readout. It repeats this process until all FS charges are read out. IA takes another exposure during this readout process. (d) Another charge transfer process begins. The CI spacing and X-ray cloud sizes are not to scale.

6.4.2 Charge Transfer Inefficiency and Charge Injection Operation

Defects in the CCD silicon lattice can hold charges for a short time. If a defect, or so-called trap, is on an electrode, it can capture a portion of charges under transfer and keep them until untrapped charges move to a downward pixel. This process, called charge transfer inefficiency or CTI, occurs stochastically, producing a tail to the original charge distribution and adding noise to the reconstructed event pulse height. The chance of trap increases with the number of pixel transfers, so more charges tend to be ripped off from events detected further from the readout node (i.e., large RAWY coordinates). The SXI chips have relatively large CTIs in the ground testing, and the CTI is expected to progress in orbit as particles continuously bombard CCDs and break the silicon lattice (Kanemaru et al., 2019).

To ameliorate the CTI problem, the SXI employs the charge injection (CI) technique. The Suzaku observatory's X-ray CCD camera (XIS) employed this technique for the first time in orbit and substantially improved the energy resolution in the middle of the mission to near the initial performance (see, for example, Uchiyama et al., 2009; Bautz et al., 2007). This technique artificially fills a large number of charges every $N$ CCD row, which fill traps before X-ray events pass the traps during the vertical transfer (Figure 6.7). Figure 6.8 clearly shows that the calibration source spectra improves with the CI operation. A downside is that the CI rows stay in the IA during frame exposures, so these lines, and probably their leading and trailing rows, cannot detect X-rays. This limitation reduces the effective area by 1/$N$ or more.

Most trapped charges escape quickly and appear in a few trailing pixels. These charges change the original 5$\times$5 pixel patterns, which is problematic for the grade event extraction classification. This charge migration can be statistically estimated and restored, and so the charge trail correction algorithm runs during the pipeline processing to fix the problem. A small amount of charges reappear outside the 5$\times$5 pixels, causing an artificial pulse height deficit. The CTI correction algorithm in the pipeline processing recovers the expected loss based on the ground and onboard calibration study (Yoneyama et al., 2020).

Figure 6.8: Results of a ground test of charge injection (CI) operation in September 2021 with an SXI-S CCD engineering model. In this test, 5 keV charges were injected to the CCD chip by every 80 binned rows. Left — Peak Pulse Height (PH) values of Mn K$\alpha$ source events with/without (red/black) CI with the ACTY row, equivalent to twice the number of parallel transfers. CI prevents PH (i.e. charge) reduction. It produces a saw-tooth pattern because it works better on pixels following near a CI row. Right — $^{55}$Fe radioactive source spectra accumulated from all events on the CCD chip. CI recovers the line energy and the energy resolution.

6.4.3 Hot Pixels

Defects in the CCD silicon lattice can also produce charge currents without X-ray or particle events, which appear as high pulse height pixels called hot pixels. Some hot pixels are permanent, always showing pulse heights above the defined hot pixel threshold. SXI-PE finds such pixels when taking dark frames onboard and flags them as hot pixels. The other hot pixels, specifically flickering pixels, show high pulse heights randomly or occasionally. They may not appear as hot pixels in dark frames, so ground data analysis finds them through statistical analysis.

Strayed optical lights can produce pseudo-hot pixels under pinholes of the optical blocking layer. If the known pinhole pixels have a significant event rate increase in orbit, the instrument and science operations teams will obtain frame images and investigate the cause. The teams do not expect optical light to leak into the SXI camera with the current observatory and instrument designs.

6.4.4 Area Discrimination

If a CCD constantly outputs enormous numbers of events from a very bright source or hot pixels, the telemetry may be unable to handle all event data and lose some onboard. If it occurs, the science operations team may choose CCD areas to downlink to avoid telemetry saturation. This area discrimination setting is not a selectable option for general observers.

6.4.5 Supporting Components

Optical Blocking Layer and Contamination Blocking Filter

X-ray CCDs are also sensitive to optical and infrared lights, which can work as noise for X-ray observations. To block the lights, the SXI CCDs coat the surface with Optical Blocking Layer (OBL), two Al layers 100 nm thick each. OBL blocks optical lights to infrared, while it transmits X-ray emission above $\sim$0.4 keV (Figure 6.3). The OBL has a small number of pinholes of 0.1$-$0.3%, which might produce high pulse height pixels with stray optical lights though the instrument team does not expect such lights come into the camera body. Pixels under the pinholes are flagged and may not be used for science data analysis if they produce significant noise.

The SXI CCD, with an operating temperature of $-$110/$-$120 $^\circ$C, is significantly colder than the other satellite components. Adhesive outgases from the satellite body tend to stick to such cold surfaces permanently. The Suzaku and Chandra observatories had such buildups on the optical blocking filters of their CCD instruments, which significantly reduced the soft X-ray sensitivities in-orbit. The Xtend SXI includes CBF at the hood top, which is composed of polyimide with a 200 nm thickness sandwiched between aluminum with thicknesses of 80 nm and 40 nm. It blocks outgases from getting inside the SXI body while it is warm enough at about $+$25 $^\circ$C at 50 cm above the CCD chips not to build up contamination on its surface. The CBF also eliminates a vacuum-tight chamber and door-opening mechanism (Mori et al., 2022). Figure 6.3 shows the combined quantum efficiency of the CCDs, OBL, and CBF.

Calibration Sources

The SXI has two $^{55}$Fe radioactive sources installed into the bonnet for in-flight calibration (Figure 6.6). The sources emit Mn K$\alpha$ (5.9 keV) and K$\beta$ (6.5 keV) radioactive decay lines to two circular spots on the CCD detector plane (Figure 6.2). The instrument and science operations teams will collect in-orbit data of these lines and measure each CCD chip's gain and energy resolution from their line centroids and widths. The team regularly updates the result in the calibration database so that users can utilize the latest calibration by reprocessing their datasets.

The holder of each calibration source collimates direct radiation to $\lesssim$4.8 mm ($\sim$3$'$) from an adjoining corner of two CCDs (Figure 6.2). The ground calibration cannot fully reproduce the in-orbit condition, so that the instrument team will measure the irradiated areas in orbit.

Figure 6.9: 55Fe calibration source spectrum, obtained during the thermal vacuum test in August 2022.

Cooling Systems

A mechanical Stirling cooler removes heat from the SXI's focal plane at a constant rate. The SXI-DE monitors the CCD temperature every second and powers the heater on the backside of the cold plate to regulate the focal plane temperature at $-$110/$-$120 $^\circ$C. The CCD temperature was stable on ground EM testing to within 0.1 $^\circ$C. Such a low-temperature fluctuation helps stabilize the CCD performance and ease the calibration effort if realized in orbit.

6.4.6 Change from the Hitomi SXI

The Xtend SXI is almost identical to the Hitomi SXI (Nakajima et al., 2018; Tanaka et al., 2018), but there are design changes to fix the problems found in the Hitomi SXI.

Adding a Notch Structure on the CCD Electrodes: The Hitomi SXI has a relatively large charge transfer inefficiency (CTI), degrading spectral energy resolution (e.g., Kanemaru et al., 2020). The Xtend SXI CCDs add a notch implant to each channel, whose potential well produces a narrow charge transfer path, reducing the chance of encountering charge traps. With the notch, the Xtend SXI has a factor of 3 smaller CTI than the Hitomi SXI. The Xtend SXI also does not show spatial variation in CTI seen in the Hitomi SXI.

Coating Additional Al Layer to the Optical Blocking Layer: The optical blocking layer of the Hitomi SXI had multiple pinholes with high optical/IR light transmission. When scattered optical/IR lights leaked into the Hitomi SXI body, these pinholes produced large pulse height pixels. The Xtend SXI adds a 100 nm thick Al coating to the original 100 nm Al coating on the CCD surface, reducing the number ratio of pinhole pixels from 4.1% to 0.2% (Uchida et al., 2020). This coating, in turn, slightly degrades soft X-ray sensitivity.

Placing an Aluminum Sheet on the Wiring Area: The bonding sheet between the CCD and Si base is transparent to optical lights. When scattered optical/IR lights leaked into the Hitomi SXI body, the lights sneaked in through the bonding sheet, producing pulse height increases of pixels near the CCD edges. The Xtend SXI makes Al coats on the layer of the backside electrode (under the passivation) to block the optical light.

No Openings on the Optical Bench: The scattered lights mentioned above originated from the two openings in the fixed optical bench for the Hitomi hard X-ray instrument (HXI). XRISM has no openings on the optical bench without the HXI and so expects no incoming scattered light inside the satellite and SXI body.

Shifting the Aim Point: The Xtend SXI's aim point is placed at (8 mm, 8 mm) from the two active pixel edges^6.3. These values are larger than the Hitomi SXI values (6.5 mm, 6.5 mm), intended to minimize the portion of the main target's PSF outskirt falling in the chip gap.