Calculation of XRT CCD contaminant layer thickness.

Link to the Stray Light Correction of the XRT Temperature Response Functions.

(22 August 2016 by Aki Takeda)
(4-Sep-2016, 4-Oct-2016, 6-Nov-2016, 2-Dec-2016, 4-Jan-2017: analysis report by Keiji Yoshimura was added.)
(21-Sep-2017, analysis of 3rd stray light increase by KY added.)

1. Importance of estimating the contaminant layer thickness.

It is officially reported that there exists an accumulation of contamination(unspecified compound) on the surface of XRT's CCD and Focal-plane analysis filters (Narukage, et al. 2011, Solar Physics, vol.269, p.169-236). Since this contamination partially absorb observed X-ray signal (and partially enhance G-band signal), it affects the photometry of the XRT observation. It is also reported that the thickness of the contaminant layer varies with time; generally accumulates with time and cleared by the CCD bakeout operation. So it is important to estimate the thickness of the layer at any time of observation by XRT. For this purpose, The intensities of the G-band full-Sun images are being monitored regularly (see, G-band luminosity monitor page).

2. The current model and handling method.

We handle the absorption/enhancement by contaminant (their spectral dependence) as a factor of XRT's effective area function;
[Effective Area] = [aperture] * [prefilter] * [mirror1] * [mirror2] * [filter1] * [filter2] * [ccdqe] * [contamination]
which is used to derive XRT's temperature response functions.

3. Possible improvements.

(1) More frequent update of DB.

As of this writing(Aug, 2016), the latest DB (xrt_contam_on_ccd.geny, ver_1.1_13-SEP-2015) covers until 13-Sep-2015. Ideally, updates after every CCD bakeout operation (when the level0 data before the latest bakeout become available) is desirable.

(2) Taking into account the gradual increase of G-band baseline.

The G-band luminosity monitor page shows gradual increase of the baseline of G-band intensities since 2011 (even excluding the effect of intensity jump-ups due to the stray light increase of 9-May-2012 and 14-Jun-2015). The contaminant layer is not completely removed by each bakeout operation after 2011, while the current model assumes it is cleared to zero after every bakeout. Taking this residual layer into account would provide more realistic estimation of contaminant layer thickness.

Obtaining the best effective area function possible (with respect to the correction of CCD contamination effect) is crucial to estimate the effect of the increased stray light on the effective area function and the response functions.

4. Recent Analysis Updates.

4.1 Re-calculation of G-band intensity variation (Keiji Yoshimura, 4-Sep-2016).

Using the same data selection and data processing method as the current official version by N. Narukage, the G-band intensity data since 2008 were re-collected. It is noticed from the Fig. 4.3 that the G-band intensity curves between two consecutive bakeouts started to go downward in the late phase since March, 2016. The similar trend is observed in the
"recent plot" at the G-band monitor page at SAO, but seems to have started earlier (Jan-2016). This indicates that the CCD contamination has become so thick (nearly 1000A) that the G-band intensity started to reduce because of the effect of interference. In other words, we are around the first maximum of the model curve (Fig.2). If this interpretation is right, we significantly underestimates the recent contaminant thickness (the current method generally calculates them as less than 300 A), and thus need to modify the current handling method.

[click image to enlarge]

Fig.4.1.1 Fig.4.1.2 Fig.4.1.3
[Update_1: data added after the 3rd stray light increase on 27-May-2017]

Reading from Fig. 4.1.6, the G-band intensity right before 27-May (shown in the lower panel) was 5.84x105 (DN/sec/full-res-pix), while the value interporated back to 27-May in the upper panel is 1.34x106 (DN/sec/full-res-pix). This yields the increase by 2.3 times.

[click image to enlarge]
Fig.4.1.4 : Jan-2008 through Aug-2017.
Plot in a same scale.
Fig.4.1.5 : Jan-2008 through Aug-2017.
Plots in separate scales.
Fig.4.1.6 :
close-up, Mar-2017 through Aug-2017.

4.2 Determination of the baseline of the G-band intensity (Keiji Yoshimura, 4-Oct-2016).

In order to determine the baseline increase of the G-band intensity due to CCD contaminant, you need to eliminate the effect of the stray light, which also increase the G-band intensity. Fig. 4.2.1 shows the G-band intensities after subtracting the stray light contribution.

Fig.4.2.1 (click to enlarge)
Note for Fig. 4.2.1
Even with the scattered data points after the 2nd stray-light event, we can define the baseline of the G-band intensities, which keep increasing since 2008 (not quantified yet). According to this curve, the current value of the baseline is approximately 2.6x10^4 (DN/sec/pix), which is ~30% increase from the value in 2008 (~2.0x10^4 DN/sec/pix). This amount of increase is converted to the contaminant layer thickness of 400 to 500 Å. As the contaminant layer grows after the CCD bakeout operation, the G-band intensities reach up to 2.9 x10^4 (45% increase from 2008). This value is just around the top of the first maximum of the model relation curve (corresponds to contaminant layer thickness of 700 Å).

As of Oct-2016, the thickness of the contaminant layer (around the mid point of the 2 consecutive bakeouts) is estimated as 600 to 700 Å.

4.3 Conversion of G-band intensity to the contaminant layer thickness (Keiji Yoshimura, 4-Nov-2016).

Once Fig. 4.2.1 (i.e., the G-band intensity variation solely due to contamination accumulation without stray light contribution) is obtained, we can convert each data point to the contaminant thickness using the model curve (Fig. 2). Figure 4.3.1 shows an example of the conversion of the G-band intensities between the two CCD bakeouts, performed on 21-Sep-2012 and 9-Oct-2012. The new analysis shows that the initial value of the contaminant layer thickness and the growth rate are 344 Å and 335 Å/month, respectively, while in the old analysis, the initial thickness was always set to zero and the growth rate was calculated as 701 Å/month for the same period (see Fig. 4.3.2, below).

Figure 4.3.1 (click image to enlarge) :
The reference value of the relative intensity increase (the left plot) is chosen to 1.96x10^4 (DN/sec/pix), which is roughly the lowest values of the G-band intensity in 2008. The relative increase is converted to the contaminant layer thickness through the model (the right plot), then linear fitting is applied (red line). The slope of the linear fit and its extrapolation to the end time of the previous bakeout provide the growth rate and the initial thickness of this sample period.
Figure 4.3.2 (click image to enlarge) : The corresponding plots from the current official DB.

The derivation of the initial thickness and the growth rate for the periods after the 2nd stray light increase is more difficult, because of the following reasons.

  1. the data dispersion is large (see Fig. 4.2.1, after 14-Jun-2015).
  2. the range of model curve includes its first maximal (where the conversion function becomes to have two-values).

4.4 Completion of the contaminant layer thickness data base file (Keiji Yoshimura, 2-Dec-2016).

The above two difficulties in converting G-band intensities to contaminant layer thickness were handled with the following measures:
  1. instead of fitting the subtracted values, [G-band(VLS open)]-[G-band(VLS close)], simply use the values of [G-band(VLS close)], i.e., the stray light component. This is reasonable because the behavior of the two quantities are quite similar, and that it is the relative increase (and not an absolute values) that is used for the conversion.
  2. the two-valued nature of the model curve was avoided by applying a separate single-valued model curve depending on which side of (before/after) the time of the intensity reversal the data point falls on.
The contaminant layer thickness (on CCD) was thereby obtained for the period from early 2008 through the recent CCD bakeout ended 9-November-2016. The updated version of the data base file (xrt_contam_on_ccd.geny) was created and ready for submission to SAO. Figure 4.4.1 shows the comparison plot of the old/new values of the contaminant layer thickness.

Figure 4.4.1 (click image to enlarge)

Now MSU is ready to update the database file as soon as the level0 data become available for the G-band VLS-closed synoptic images obtained until the start of the most recent CCD bakeout. The next step is to discuss the method to transfer the updated file to SAO.

4.5 Plots of contaminant layer thickness (Keiji Yoshimura, 4-Jan-2017).

A separate web page was developed to show a series of plots of the variation of G-band intensities and their conversion to the contaminant layer thickness (e.g., Fig. 4.3.1) for all the periods between two consecutive CCD bakeouts performed since 2008. The above web page will be updated after every CCD bakeout (i.e., as soon as the level0 data of the relevant period become available).

Compared with the plots from the previous database (by N. Narukage), the data points of the contaminant layer thickness spread more linearly, and thus the linear fitting works better (as found in the right-hand plot of each pair). This indicates that the conversion of G-band intensity to contaminant layer thickness was performed properly (using the adequate part of the model curve). In the previous cases, the G-band base intensity was always set to zero, and only the initial part of the model curve has been used.

4.6 Summary and the Definitive figure.

A special telecon was held among MSU and the XRT managers on 9-Jan-2017. Yoshimura reported summary of his work and the technical issue (i.e., the way to regularly update the SSW master database at SAO) was discussed.

Summary PDF and Appendix prepared by Keiji Yoshimura for the telecon and future reference.

Figure 4.6.1 (click image to enlarge)

[Update_2: data added after the 3rd stray light increase on 27-May-2017]

Contamination thickness is estimated properly after the 3rd stray light incfrease event.
(The current method will work until G-band VLS=close full-res 1-msec exposure images start to saturate in the future ...)

[Update_3: plot updated using the data until 14-Apr-2018.]

Figure 4.6.2 (click image to enlarge)

[Update_4: plot updated using the data until 19-Aug-2018.]

Figure 4.6.3 (click image to enlarge)

The impact of the improved contaminant database on XRT response function is non-significant (see pages 2-3 of the summary PDF). However, understanding the true status of the contaminant accumulation on CCD is important to support the reliability of our current method of calibration.

4.7 Effect of Long Bakeout in Dec-2017.

The CCD bakeout performed in December 2017 was longer than usual (it was caused by mis-operation). The CCD heater was kept on for a little over 120 hours (12/10 06:27UT --- 12/15 08UT). The plot below (created by K. Yoshimura) shows the variations of G-band intensity and contamination thickness throughout 2017. It turned out that no significant improvement in the level of contamination was observed by this extra-long bakeout.

Figure 4.7.1 (click image to enlarge)