Which detectors can I use to analyze this flare?
From RHESSI Wiki
|1st Author:||Brian Dennis|
|2nd Author:||Kim Tolbert|
|Published:||September 15, 2014|
|Next Nugget:||Poynting Flux|
|Previous Nugget:||RHESSI Resumes Observations|
This is the most common question we are asked by people trying to analyze RHESSI data for spectroscopy and/or imaging. The answer is not always straightforward because the individual detectors respond differently to radiation damage and to the four anneals completed so far (Nov. 2007, Mar. 2010, Jan. - Feb. 2012, and Jun. - Aug. 2014).
The issue is particularly important now because some of the detectors did not separate electronically into a front and rear segment after the two most recent anneals. Following the 2012 anneal, detectors 2 and 4 failed to segment, although both detectors eventually segmented much later. Following the 2014 anneal, detectors 2, 4, 5, and 7 did not segment. Unsegmented detectors operate as single large-volume detectors and have similar effective area to that of segmented detectors. Their background rates are higher and their energy resolutions are worse. Nevertheless they are still perfectly adequate for spectroscopy and imaging above ~20 keV.
We have generated an archive of quicklook front-segment spectral plots for times throughout the mission to help determine which detectors and energy ranges are OK to use for light curves, spectroscopy, and imaging of a specific flare. The plots show the count spectra and the corresponding 'semi-calibrated' photon spectra for the nine detector front segments for a one-minute interval centered on the peak time of each flare in the RHESSI flare catalog, even if the flare is divided into multiple events in the catalog. An example is shown in Figure 1.
The plots are available in Browser via the 'Detector Spectra' check-box under the 'Flare Quicklook' option (click the magnifying glass to expand). Use the flare-seeking button on Browser to step through the data by flare. The PNG files displayed in Browser can be directly accessed here.
Description of Plots
As the example in Figure 1 shows, the plot for each flare has two sections: an upper panel with the count-flux spectrum between 1 and 100 keV for each of the nine detector front segments, and a lower panel with the corresponding 'semi-calibrated' photon-flux spectra.
The count-flux spectra are in units of counts cm-2 s-1 keV-1 corrected for decimation and live time but with no background subtracted. The energy bin widths are 0.3 keV below 10 keV and increase logarithmically to 3 keV at 100 keV to give a total of 138 bins.
The 'semi-calibrated' photon-flux spectra are in units of photons cm-2 s-1 keV-1. They are computed by multiplying the corresponding count-flux spectrum by the diagonal elements of the detector response matrix for the attenuator state at that time. Since no background was subtracted, this is equivalent to assuming that all the counts were from solar photons and that the off-diagonal elements are negligible. Photon spectra determined in this way approximate the actual solar spectrum only in the energy range where the solar count rates are significantly greater than background rates and the off-diagonal elements are small (typically between ~10 and 100 keV for large flares). These photon plots serve primarily to show the relative sensitivity of the front segments and how well the detector response matrices correct for the known sensitivities of the different detectors. More accurate spectral analysis with background subtraction and using all elements of the response matrix for each detector requires the use of OSPEX. This also allows for the inclusion of albedo (X-rays reflected from the Sun) in the analysis, and corrections for instrumental effects such as pulse pile-up (one kind of saturation) and changes in the energy resolution and calibration for individual detectors.
Interpreting these plots depends on the attenuator state at the time the spectrum was obtained. (Attenuators are aluminum disks that can be automatically brought into the field of view of the detectors to reduce the effects of pulse pile-up at high count rates - see Smith et al. (2002) p. 41 & 58.) The hatched area on each plot indicates the attenuator-state dependent lower energy range that should not be used for analysis. The upper energy limit (not shown on these plots) depends on the flare intensity and the energy at which the solar flux dips below the background level.
Figures 1, 2, and 3 show spectra in the three possible attenuator states - A0 (no attenuators), A1 (thin attenuators), and A3 (both thin and thick attenuators). (Note that the thick attenuators are never used by themselves.) The interpretation of these quick-look plots in the three attenuator states is discussed below.
A0 Attenuator State (no attenuators)
An example of a small flare in the A0 attenuator state on 2 May 2014, just prior to the 2014 anneal, is shown in Figure 1. As is usual in this attenuator state, the count-flux spectrum from the flare peaks at ~7 keV. The separate spectra show which detectors have recorded an increase in count rate from the flare. In this case, all detectors except 6F show the solar signal peaking at this energy.
The plotted photon spectra are a useful representation of the solar spectrum which in this case extends above background up to ~15 keV. The lower energy is set by both the electronic lower threshold level for each detector and the absorption of material in front of the detectors including the thermal blankets and the beryllium windows on the cryostat. Thus, counts recorded below ~3 keV are electronic noise and should not be used in the determination of the photon spectrum. The calculated photon spectrum above ~15 keV in this case is an indication of the background spectrum under the false assumption that all the photons came from the solar direction.
Here is a breakdown of what you can determine about each detector from the spectra shown in Figure 1.
- Detectors 1F, 3F, 4F, 5F, and 8F all show similar photon spectra above 3 keV and can be used for spectroscopy and imaging. We'll call these the 'good' detectors for this event. Note that detector 3F has lower count rates than the others but its response matrix takes care of this in computing the photon spectrum. This detector will therefore suffer less from pulse pile-up, but that is not a factor for this weak flare.
- Detector 2F has a similar count-flux spectrum as the 'good' detectors but shows a much broader peak because of its poorer energy resolution. Consequently, you may not want to use it for spectroscopy below ~10 keV.
- Detector 6F shows no flare counts peaking at 7 keV so it should not be used at all for this event. The detector electronic threshold level was set high (~11 keV) because of the high rate of noise counts at this time, extending up to tens of keV.
- Detector 7F shows a similar count-flux spectrum to the 'good' detectors but the calculated photon spectrum is much higher than for the other detectors below ~12 keV indicating that there is a problem with the response matrix for this detector. Thus, this detector should not be used at low energies at this time until the response matrix is corrected.
- Detector 9F shows lower count fluxes than the 'good' detectors below ~8 keV, and that discrepancy is not removed in the photon spectrum. Consequently, you might not want to use detector 9F at lower energies.
Weak events like this one require careful background subtraction and considerations of different attenuation at low energies possibly from degrading thermal blankets over the detectors.
A1 Attenuator State (thin attenuators)
An example of an M-class flare in the A1 attenuator state on 03 September 2014, just after the 2014 anneal, is shown in Figure 2. As is usual in this attenuator state, the solar count-flux spectrum peaks at ~11 keV. The solar flux can be seen above background at energies up to ~30 keV. The count-flux spectrum falls at energies below ~6 keV due to the absorption of the overlying material, which now includes the thin attenuator. This strong attenuation at low energies is such that almost all counts below ~6 keV are from K-escape events, i.e. incident photons with energies above the germanium K-edge at 11 keV that have photoelectric interactions that result in the escape of the K-shell photon from the detector - see Smith et al. (2002) p. 56 and Sampaio et al. (2014). As a result, counts below ~6 keV should not be used.
- Detectors 1F, 3F, 6F and 8F, and 9F are the 'good' detectors in this case. The calculated photon spectra closely follow the same curve for these five detectors showing that they can all be used between ~6 and ~30 keV. The Fe-line complex peak at 6.7 keV can be seen in the count-flux spectra from the detectors with the best energy resolution (1F, 3F, and 6F).
- Detectors 2F, 4F, 5F, and 7F were not segmented at this time and their calculated photon spectra show that the current response matrices do not correctly allow for the electronic threshold energies in this state. Their background levels are higher since counts from interactions anywhere in the full detector volume are now recorded in the electronics channel normally reserved for front-segment events. This necessitates using a higher electronic threshold of up to ~20 keV to limit the total count rate. Consequently, these detectors cannot be used at energies below ~20 keV and hence cannot be used at all for this flare. Flares that extend to higher energies should give count rates above the enhanced background rates and it should be possible to use existing front-segment response matrices once the higher electronic energy thresholds are taken into account.
A3 Attenuator State (thin and thick attenuators)
An example of an intense flare in the A3 attenuator state on 20 January 2005 is shown in Figure 3. As is usual in this attenuator state, the solar count-flux spectrum peaks at ~18 keV. The solar flux can be seen above background at energies up to at least 100 keV. As in the A1 attenuator state, the count-flux spectrum falls at energies below ~6 keV due to the absorption of the overlying material, which now includes both the thick and the thin attenuators. Consequently, as in the A1 attenuator state, almost all counts below ~6 keV are from K-escape events, and counts below ~6 keV should not be used.
The close alignment of the photon spectra for all detectors between 6 and 100 keV indicates that they can all be used over this full energy range although there are still small differences between detectors as indicated below.
- Detector 1F has a higher count rate above ~60 keV than the other detectors. This is because the grids above this detector are made of molybdenum instead of tungsten, and they start to become transparent to X-rays above this energy. The photon spectrum for this detector matches the spectrum for the other detectors indicating that this difference is incorporated in the response matrix.
- Detector 2F has poorer energy resolution than the other detectors as indicated by the enhanced count rate below ~4 keV compared to the other detectors. Thus, this detector should not be used for spectroscopy when high energy resolution is important.
- Detectors 3F, 4F, 5F, 6F, 8F, and 9F all have similar count and photon spectra and can be used from 6 to 100 keV for all applications.
- Detector 7F shows reduced count fluxes below ~12 keV and enhanced rates above ~25 keV. The reduced rates at low energies are due to the higher electronic threshold needed because of the high background rate from this detector probably caused by its poorer energy resolution. Thus, even though the photon spectrum matches the other detectors down to 6 keV, it should not be used below ~10 keV since the count rates would be very low. The excess count flux above 25 keV is not taken into account by the response matrix since it also shows up as excess flux in the photon spectrum. It is most probably caused by pulse pile-up because the electronics are less able to reject pile-up events when the electronic threshold is high. It should be possible to take this into account in OSPEX spectral analysis by including the pileup function when analyzing data for this detector alone.
Please check out these quicklook detector spectral plots in Browser or here. Use them to decide which detectors can be used for any particular flare, and over what energy range. You can generate them yourself for any time interval by running the IDL program hsi_spectrum_check_dets.pro. For example, the command to generate the plot in Figure 1 is:
hsi_spectrum_check_dets,time='2-may-2014 00:55:06', /no_gui
Generating the plots yourself will allow you to examine different time intervals (don't include times of attenuator changes!). You can then determine if the detector response matrices have been improved since the archived quicklook plots were created (creation date/time shown in the lower right corner of each plot). Changes to the response matrices and energy calibrations have been infrequent in the past but continue to be necessary as we learn more about variations in detector response with time.
If after all of this, you still don't know which detectors to use for any particular flare, feel free to ask any RHESSI team member.