IACHEC 2019@Shonan International Village The spectral response of X-ray CCDs in the energy band around Si-K edge: a solution to the Si-K edge problem for the XIS onboard Suzaku Kiyoshi HAYASHIDA, Koki OKAZAKI , Riku SHOMURA, Tomokage YONEYAMA, Hironori MATSUMOTO, Hirofumi Noda, Hiroshi TSUNEMI (Osaka Univ.), Hiroshi NAKAJIMA(Kanto-Gakuin Univ.), Koji MORI (Univ. of Miyazaki), Masahiro TSUJIMOTO, Yoshitomo MAEDA, Ken EBISAWA (ISAS/JAXA) Today I’d like to talk about “The spectral response of X-ray CCDs in the energy band around Si-K edge: a solution to the Si-K edge problem for the XIS onboard Suzaku.”　Most of the contents were described in SPIE proceeding by Okazaki et al. but I include recent progress. See also Okazaki et al. , SPIE Proc. 10709, id. 107091F (2018) May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village Outline Introduction Suzaku and XIS Si-K edge problem Solution of the Si-K edge problem Introduce discontinuity in PH_peak-Ex relation Application to several targets Discussion May 20th, 2019 IACHEC 2019@Shonan International Village

The X-ray Astronomy Satellite “Suzaku” X-ray CCD Cameras “X-ray Imaging Spectrometers (XIS)” FI-CCD × 3 (XIS0, 2,3) BI-CCD × 1 (XIS1) Suzaku launched on 2005, and operated 10 years in orbit. 4 X-ray CCD cameras onboard are XIS. 3 of 4, XIS0, 2 and 3 employ an FI-CCD and the other XIS1 does BI-CCD. (JAXA) 2005-2015 (Koyama et al. 2007) May 20th, 2019 IACHEC 2019@Shonan International Village

Two Types of CCDs Soft X-ray Hard X-ray Soft X-ray Hard X-ray electrode Insulation layer depletion layer ~42um depletion layer ~65um Insulation layer You know FI and BI CCDs. electrode FI (Front-Illuminated)-CCD XIS0, XIS2, XIS3 BI (Back-Illuminated)-CCD XIS1 Note: XIS2 was stopped on 2006 XISs were sensitive in 0.3-12 keV May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village Si-K Edge Problem LMC X-3 (Black Hole Binary) model: phabs*(diskbb + powerlaw) black: FI-CCD red: BI-CCD ratio= data model about 10 % residual around Si-K edge (1.839 keV) different direction peak energy We had a response/calibration problems around Si-K edge. About 10% residua;. Different direction between FI and BI. The peak energy is also different. May 20th, 2019 IACHEC 2019@Shonan International Village

Si-K Band was ignored in the analysis In some papers, Si-K band were ignored. That’s not good. Perseus Cluster (Tamura et al. 2009) RX J1347.5 (Ota et al. 2008) May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village Response Matrix 𝑆 𝑃𝐼[0] 𝑆 𝑃𝐼[1] ⋮ ⋮ 𝑆 𝑃𝐼[𝑛−1] = 𝑅 0,0 𝑅 0,1 … 𝑅 0,𝑚−1 𝑅 1,0 𝑅 1,1 ⋮ 𝑅 𝑛−1,0 𝑅 𝑛−1,𝑚−1 𝑀 𝐸[0] 𝑀 𝐸[1] ⋮ ⋮ 𝑀 𝐸[𝑚−1] PI spectrum incident X-ray energy spectrum response matrix You all are familar with reseponse matrix. For XIS, xisrmgen is the software to make the response. The software “xisrmfgen” makes response matrixes for XIS. May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village PI_peak-Ex Relation PI_peak: peak PI when a monochrome X-ray enters CCD Ex: incident X-ray energy assumed 𝑃𝐼_𝑝𝑒𝑎𝑘 channel = 𝐸 x eV 3.65 [eV/channel] response matrix Ex (keV) 4095 PI_peak [channel] PI(channel) So far, in xisrmfgen, PI_peak is assumed to be proportional to the incident X-ray energy Ex. So PI peak gets in a diagonal line in the matrix. PI_peak get in a diagonal line Ex [keV] Si_Kedge=1.839 14.95 May 20th, 2019 IACHEC 2019@Shonan International Village

Introduce New Parameters in PI_peak-Ex relation response matrix Ex (keV) jnct_hiene 4095 PI_peak [channel] jmp_hiene PI(channel) This time, we introduced a jump between PI peak Ex relation. We should have noticed before but. We intoudced three parameters, jmp loene, lump hienem and junction hiene, the last one fixed to be 5.9keV because we don’t want to touch the gain primarily calibrated at 5.9keV. jmp_loene PI_peak is discontinuous jmp: gap from the original linear relation at Si-K edge junct: the energy where the relation returns original 5.9 Ex [keV] Si_Kedge=1.839 14.95 (default) May 20th, 2019 IACHEC 2019@Shonan International Village

Search for Best Jump Values Calibration target : LMC X-3 (black hole binary) Use model: phabs*(diskbb + powerlaw) Fit in 0.7-10.0 keV Fix 𝑁 𝐻 and Γ Refit in 1.4-2.5 keV search the jump values which make reduced-chi square ( 𝜒 𝑟 2 ) minimum Search Range (step 0.1 channel) XIS0 XIS3 XIS1 jmp_loene [channel] −2.0 - +2.0 −3.0 - +1.0 −1.0 - +5.0 +2.0 - +5.5 −5.0 - +1.0 We then searched the jump values that minimize the chi square value with LMC X-3 data. May 20th, 2019 IACHEC 2019@Shonan International Village

Best Jump Values for XIS0 (FI) 2.5 black: response without jump 𝜒 𝑟 2 =2.3159 for 297 d.o.f red: response with jump jmp_loene=+0.7, jmp_hiene=+4.9 𝜒 𝑟 2 =1.1468 for 297 d.o.f This is one example. Black data points is fitted either with reseponse without jump and wth jump. You can clearly see the improvement. May 20th, 2019 IACHEC 2019@Shonan International Village

Best Jump Values for XIS3 (FI) black: response without jump 𝜒 𝑟 2 =2.2799 for 297 d.o.f red: response with jump jmp_loene=+0.2, jmp_hiene=+4.2 𝜒 𝑟 2 =1.1046 for 297 d.o.f This is for XIS3, another FI May 20th, 2019 IACHEC 2019@Shonan International Village

Best Jump Values for XIS1 (BI) black: response without jump 𝜒 𝑟 2 =2.6515 for 297 d.o.f red: response with jump jmp_loene=−1.0, jmp_hiene=−4.1 𝜒 𝑟 2 =1.7205 for 297 d.o.f This is XIS1, BI case. We found the improvement, but not as good as FI case. May 20th, 2019 IACHEC 2019@Shonan International Village

Summary of Best Jump Values jmp_loene [channel] jmp_hiene 𝝌 𝒓 𝟐 [for 297 d.o.f] without jump XIS0 (FI) +0.7 +4.9 1.1468 2.3159 XIS3 +0.2 +4.2 1.1046 2.2799 XIS1 (BI) −1.0 −4.1 1.7205 2.6515 This table summarize the best jump values. For all sensor, reduced-chi2s were improved by about 1. Jump direction is positive and negative for FI abd BI respectively. May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village Perseus Cluster (XIS0) w/o jump -> “before” w/ jump -> “after” model : phabs*(vapec1 + vapec2 + powerlaw) before after 𝜒 𝑟 2 =1.15107 for 2060 d.o.f 𝜒 𝑟 2 =1.08207 for 2060 d.o.f We also applied the reseponse with jump to other sources, this case Perseus cluster. You see the improement. May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village PKS2155-304 (XIS0) NOT set jump -> “before” set jump -> “after” model : phabs*logpar !!! We fitted in 0.7-10.0 keV and zoomed 1.4-2.5 keV 2005 2014 before/after before/after reduced-chi square 1.3908 for 296 d.o.f 1.0998 for 296 d.o.f reduced-chi square 1.1159 for 296 d.o.f 1.0116 for 296 d.o.f This is another example, PKS2155 observed in different epoch. May 20th, 2019 IACHEC 2019@Shonan International Village

X-ray Attenuation Length in Silicon The attenuation length is discontinuous at Si-K edge(1838.9 eV) 14.3 𝜇m at 1838.5 eV (= 𝐸 below ) 1.3 𝜇m at 1839.1 eV (= 𝐸 above ) Let us consider possible origin of this jump. We think the key should be in the jump in the attenuation length in siilion. 14.3micrometer below the edge and 1.3 micrometer above the edge. The path length of the electron cloud changes discontinuously at Si-K edge. (http://henke.lbl.gov/optical_constants/atten2.html) May 20th, 2019 IACHEC 2019@Shonan International Village

jump = jmp_hiene –jmp_loene Summary of Jump jmp_loene [channel] jmp_hiene jump = jmp_hiene –jmp_loene [eV] XIS0 (FI) +0.7 +4.9 4.2 15 XIS3 +0.2 +4.2 4.0 XIS1 (BI) −1.0 −4.1 −3.1 −11 This is the summary of jump again. May 20th, 2019 IACHEC 2019@Shonan International Village

Possible Origin of Jumps ※Si-K edge : 1838.9 eV 𝐸 below : 1.838.5 eV 𝐸 above : 1.839.1 eV Possible Origin of Jumps The longer the path length is, the larger the charge loss is. Some amount of charges is lost in the depletion layer? PI_peak PI_peak Ex Ex 𝐸 below 𝐸 above BI-CCD FI-CCD 𝐸 below 𝐸 above X-ray X-ray electrode 1.3 𝜇m 1.3 𝜇m depletion layer 1.3 𝜇m Schematically PI_Peak Ex relation is illustrated like this for FI and BI. X-ray absorption depth is different like this. The charge have to travel with a distance illustrated with green arrows. f we assume some charges are lost in their travel, and the longer the path is the lager the loss is. We can explain the jumps qualitatively. 14.4 𝜇m 14.4 𝜇m 42 𝜇m 14.4 𝜇m 65 𝜇m 40.7 𝜇m 27.6 𝜇m depletion layer electrode May 20th, 2019 IACHEC 2019@Shonan International Village

Conclusion (until Okazalo et al. SPIE proc 2018 and HEASOFT v6.25) Introduction of “jump” in response PI_peak-Ex relation, the residual around Si-K edge was reduced. Jump parameters were determined by fitting LMC X-3. The reduction of the residual was confirmed also for other targets; Perseus Cluster with emission lines, PKS2155 observed a few times. Jumps are ∼4 channel (15 eV) for XIS0 and 3, and ∼−3 channel (11 eV). We speculate some amount of charges is lost in the depletion layer; the longer the path length is, the larger the charge loss is. “xisrmfgen” in which jumps are implemented with their optimized values was released on Oct. 2018 as a part of HEASOFT v6.25. Here is the summary. The revised xisrmfgen and its default jump were released in HEAsoft v6.25 last year. May 20th, 2019 IACHEC 2019@Shonan International Village

Further Optimization after HEASOFT v6.25 Jump values are tuned also with M87 and Perseus Cluster in which line emissions are present. However, significant residual remains for the BI. LMC X-3 (XIS1) red: response w/ jump jump=−2.8 channel 𝜒 𝜈 2 =1.72 for 297 d.o.f in V6.25 after loene (ch) hiene (ch) XIS0 +0.2 +4.5 -1.9 2.4 XIS1 -2.0 -4.7 -0.1 -2.9 We have tried further optimization after that. Jump values are tuned with other sources. Altough amount of jump are the same, absolute energy scale is different. Even after this, we found the residual is relatively larger for BI. May 20th, 2019 IACHEC 2019@Shonan International Village

Further Optimization after HEASOFT v6.25 LMC X-3 (BH Binary) QE of BI-CCD is reconsidered. red: w/o edge model 𝜒 𝜈 2 =1.72 for 297 d.o.f purple: w/ edge model 𝜒 𝜈 2 =1. 16 for 297 d.o.f Au-M BI surface design Si-K Hf-M Hf O 2 (5 nm) Si O 2 (3 nm) Si Depletion Layer Ag (1 nm) HfO2: 4nm SiO2+Si deadlayer: 160nm We then reconsidered the QE, in particular the surface dead layer of the BI CCD. Tuning some values from its design values, we can improve the fit. We are considering update of the CALDB describing the QE. 1.4 1.6 1.8 2 2.2 2.4 Energy (keV) May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village For XRISM Xtend Response generator is being updated from that for Hitomi SXI so as to implement possible jumps in the Ex – PI_peak relation. Note: PH->PI conversion should not have jumps. We need matrices to implement a jump. Experiments around Si-K edge performed with a test CCD before the Hitomi launch have been investigated in detail. Irradiation of continuum X-rays are considered. This is the final slide. The XRISM Xtend CCD has different design from XIS and we have not yet concluded similar jump is needed or not. But, we introduce a function to implement possible jumps in the software, at least May 20th, 2019 IACHEC 2019@Shonan International Village

IACHEC 2019@Shonan International Village Backup Slide May 20th, 2019 IACHEC 2019@Shonan International Village

XIS Experiments on Ground with a Grating Gaussian Fit (Response Profile is NOT taken into account) W-M line contamination makes analysis around Si-K edge difficult FM Spare FI0 FM Spare BI1 May 20th, 2019 IACHEC 2019@Shonan International Village

Application to Cluster and SNR Spectra Tycho (SNR) Perseus Cluster 𝜒 𝜈 2 =5.06 for 284 d.o.f 𝜒 𝜈 2 =1. 72 for 284 d.o.f 𝜒 𝜈 2 =22.91 for 272 d.o.f 𝜒 𝜈 2 =27.53 for 272 d.o.f model: phabs*(vapec*2 +PL) (Tamaru+2009) model: phabs*(vnei*3 + PL +Gaussian) (Katsuda+2015) 左がペルセウス座銀河団、右が超新星残骸Tychoでのフィッティング結果です。SNRではgaussianの輝線の速度分散を考慮しています。ここまでと同様同一のデータに対し、ジャンプ無しの応答行列用いた結果が青、ジャンプ有りの応答行列を用いた結果が赤になります。銀河団では、ジャンプ導入により最大で15%程度あった残差が5%以下にまで改善されました。一方、SNRでは改善したキャリブレーション精度を大きく超えた残差が残りました。 1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4 Energy (keV) Energy (keV) We need to Update Emission model for this case. May 20th, 2019 IACHEC 2019@Shonan International Village

Study of Si-K edge response in 1994 Theory (Fraser et al., NIMA, 1994) W-value jump by +0.2% Experiment (Owens et al., NIMA, 1994) W-value jump by +0.5%