Verification of the flux calibration in the UV clocked grism

Overview

The new uv-grism flux calibration is valid over the whole detector, and makes also a correction for coincidence-loss. The first integrated software version is UVOTPY-0.9.7.0, but we will do the verification to the 0.9.7.1 version which has patches for the sensitivity-loss and flux model interpolation.

With the choice of data, we have an independent set from the spectra used to construct the effective area. Of course, spectra from the same targets are used.

Method

Calibration spectra that were not used in the flux calibration are reprocessed using the new flux calibration and then compared to the known stellar flux. The spectra for verification are located on the detector between the areas with the spectra used to determine the effective areas for the flux calibration.

../_images/calobs_locations_160.png

Figure 1: The anchor positions of the calibration spectra marked by target. The boxes are enclosing sources used to derive an effective area for the location of the box.

Data used for verification

The approximate location of the anchor of the spectra, their obsid, and number of the fits extension is listed. First the cool stars are listed, followed by the hot white dwarfs.

GSPC-P041C spectra (F0 V)

anchor obsid ext % err plot aspect correction
781,1459 00057965002 1,2 24 GSPC-P041C Figure 1 This spectrum with anchor at [781,1459] is affected by the decreasing sensitivity in this part of the detector. yes,yes
641,1275 00057964002 2 18 GSPC-P041C Figure 2 This spectrum with anchor at [641,1275] is affected by the decreasing sensitivity in this part of the detector. yes
940,1553 00057972002 1 1 GSPC-P041C Figure 3 This spectrum has its anchor at [940,1553]. yes
1361,1051 00057960002 1,2,3 1.5 GSPC-P041C Figure 4 This spectrum has its anchor at [1361,1051]. yes,all
1578,1151 00057969001 3 6 GSPC-P041C Figure 5 This spectrum has its anchor at [1578,1151]. yes
1454,1494 00057958001 2 1 GSPC-P041C Figure 6 This spectrum has its anchor at [1454,1494]. yes
1506,1464 00057958001 1 4 GSPC-P041C Figure 7 This spectrum has its anchor at [1506,1464]. yes

GSPC P177D spectra (F0 V)

anchor obsid ext % err plot aspect correction
1586,1756 00056762002 4 17 GSPC-P177D Figure 1 This spectrum has its anchor near [1586,1756]. yes
993,1496 00056763002 6 1 GSPC-P177D Figure 2 This spectrum has it anchor near [993,1496]. yes
821,1491 00056763002 2 3 GSPC-P177D Figure 3 This spectrum has its anchor near [821,1491]. yes
1229, 889 00056760002 1 8 GSPC-P177D Figure 4 This spectrum has its anchor near [1229,889]. yes

The error is calculated as the mean of the flux difference over a good region of the spectrum, and is considered to be a measure of the error in the flux level. The flux was calibrated using uvotpy.uvotio.rate2flux default option with anchor position given.

WD1657+343 spectra

anchor obsid ext % err plot aspect correction
1797,1467 00055903006 1 1 WD1657+343 Figure 1 This spectrum has its anchor near [1797,1467]. A large zeroth order is present contaminating the 4300A region. yes
1338, 941 00055900050 1 4 WD1657+343 Figure 2 This spectrum has its anchor near [1338, 941]. yes
1291, 891 00055900050 3 9 WD1657+343 Figure 3 This spectrum has its anchor near [1291, 891]. yes
1302,1082 00055900050 2 5 WD1657+343 Figure 4 This spectrum has its anchor near [1302,1082]. yes

WD1057+719 spectra

anchor obsid ext % err plot aspect correction
833,1307 00055203007 1,2,3 11.3 WD1057+719 Figure 1 This spectrum has its anchor near [833,1307]. This spectrum is located near the region of the detector with reduced response. Notice the dip in response near 1800A happens in two locations on the detector and which appear to have a slightly reduced sensitivity. These are only seen very occasionally, and do not have a photometric signature of more than 1%. Very few spectra have been taken at those locations. In the grism, with coincidence loss amplifying differences and background subtraction doing it as well, this appears be more significant than for the UVOT photometry, at least for bright sources. These spots fall close to the spots of reduced photometric sensitivity . The issue is still under investigation. yes,all
833,1307 00055203008 1 13 WD1057+719 Figure 1 This spectrum has its anchor near [833,1307]. This spectrum is located near the region of the detector with reduced response. Notice the dip in response near 1800A happens in two locations on the detector and which appear to have a slightly reduced sensitivity. These are only seen very occasionally, and do not have a photometric signature of more than 1%. Very few spectra have been taken at those locations. In the grism, with coincidence loss amplifying differences and background subtraction doing it as well, this appears be more significant than for the UVOT photometry, at least for bright sources. These spots fall close to the spots of reduced photometric sensitivity . The issue is still under investigation. yes
876,1334 00055203007 4 2 WD1057+719 Figure 2 This spectrum has its anchor near [876,1334]. This spectrum is located near the region of the detector with reduced response. yes
1447,1324 00055201011 1,2 4 WD1057+719 Figure 3 This spectrum has its anchor near [1447,1324.] yes,yes
1496,1255 00055201011 3 14 WD1057+719 Figure 4 This spectrum has its anchor near [1496,1255]. yes

Results

Although the number in the table for the error is based on the mean, since the total exposure time was not long enough to beat down the noise enough, it appears that over most of the detector the flux calibration is within 10% or better. As expected, the flux is most uncertain near the sudden drop in response. The current calibration file (version 002) used with the current uvotpy version (0.9.7.1) give a fit to within about 15%. The largest deviations seem related to contamination of the spectrum by the spectra from other sources in the field, and of course from the second order.

I should mention that the limit on the errors is so some degree dominated by the noise in the data. The error in the flux calibration as determined from the calibrated areas (see the boxes in the figure above,) is more reliable since the total exposure time in those areas is larger.

The wavelength scale errors are quite apparent. Errors in the anchor position cause wholesale shifts of the scale, while the dispersion at long wavelengths is not correct (> 4000A) at present. The zemax model which was used for the dispersion turned out to be not correct, and an update of the dispersion is planned. That in turn causes large swings in the comparison of the flux from the extracted spectra (which were not shifted) and the reference spectra. The anchor positions were determined using lenticular files taken before or after the grism exposure in the same Swift “mode”, meaning that no slew took place between grism and corresponding exposure in one of the lenticular filters. However, occasionally drift of the pointing occurs, especially noticable in the weaker spectra with more than 1000 second exposure times.