Subsections

• 7.8.1 Actions
• 7.8.2 Usage and Inputs
• 7.8.3 Data Products

7.8 MERGE_OBSERVATIONS Stage

This stage is run once as shown in Figure 8, after extractions have been performed for each observation. Even if you have only one observation you will run MERGE_OBSERVATIONS.

7.8.1 Actions

• A multi-ObsId PSF is constructed as an average of the 1.49 keV PSFs from all the observations being merged, weighted by the exposure map values in each observation. The multi-ObsId PSF is normalized so that it sums to 1.0.

• The ``neighborhood'' event lists (neighborhood.evt) for each observation are reprojected and merged to form a multi-ObsId neighborhood event list.

• The extracted source event lists (source.evt) for each observation are reprojected and merged to form a multi-ObsId source event list.

• The spectral extraction data products (source and background spectra, ARFs, RMFs) for all the observations of a source are combined into a single set of products ready for spectral fitting.

• Time variability of the source both within and between observations is quantified by comparing a uniform flux model to the distribution of source event time stamps from all observations and computing the 1-sample Kolmogorov-Smirnov statistic. The significance of this multi-ObsId KS statistic is saved in the keyword MERGE_KS in source.stats; values near 0 suggest variability. Currently no background ``subtraction'' is performed in variability analysis--thus background variability could bias the KS statistic downward.

• Two PostScript plots are produced showing light curves,produced by the TIMING stage (§E.4), from multiple observations--one uses a separate set of axes for each observation (Figure 9), the other shows all observations on a single ``broken'' time axis (Figure 10). Examination of the flux and median energy time-series can reveal source spectral changes during a flare.

  BEWARE: none of AE's timing analysis (light curves, median energy time series, variability measures) currently takes the background into account! This can produce misleading results in some cases, e.g. a spurious indication of variability between two ObsIds that have very different backgrounds.

• Wide-band photometry corrected for background is performed, as described in §5.10.

• A 2-sample Kolmogorov-Smirnov¹² statistic is computed to compare the source and background spectra, and the significance of that statistic is reported in the keyword KS_SPECT in source.stats. It is hoped that this statistic will characterize whether a candidate source is similar (significance $\sim$1) or distinct (significance $\sim$0) spectrally from its local background. Note that for some sources the source and background spectra may have very different numbers of points. For sources with enough counts, this may help to distinguish true point sources from clumpy diffuse emission.

• A number of observation-specific keywords from the obs.stats files are combined/propagated to the source.stats file:

  NUM_OBS:

  number of observations extracted

  OVRLP_LM:

  overlap limit used for discarding observations

  OVRLP_LO:

  smallest overlap fraction from all observations

  OVRLP_HI:

  largest overlap fraction from all observations

  BESTOBS:

  least crowded merged observation

  WORSTOBS:

  most crowded merged observation

  MERGNUM:

  number of observations merged

  MERGFRAC:

  fraction of extracted data merged

  OBSNAME:

  list of the OBSNAME values for merged observations

  PRIM_OBS:

  name of the deepest merged observation

  THETA:

  a weighted average of the off-axis angles of the source in merged observations

  THETA_LO:

  the smallest off-axis angle among merged observations

  THETA_HI:

  the smallest off-axis angle among merged observations

  SRC_AREA:

  a weighted average of the SRC_AREA keyword (source area) from merged observations

  SRC_RAD:

  a weighted average of the SRC_RAD keyword (source radius) from merged observations

  PSF_FRAC:

  a weighted average of the PSF_FRAC keyword (PSF fraction) from merged observations

  MSK_RAD:

  largest value of the MSK_RAD keyword (mask radius) from merged observations

  FRACSPEC:

  smallest target PSF fraction for merged observations

  SRC_CNTS:

  total in-band counts extracted from merged observations

  EMAP_TOT:

  sum of EMAP_AVG values for merged observations

  FRACEXPO:

  fraction of time that the source position (not the whole PSF footprint) spent on live portions of the detector, derived from FRACEXPO in the CIAO ARF files

  WARNFRAC:

  fraction of events in the ``warning regions'' of merged observations

  EXPOSURE:

  total exposure in merged observations

  EFFAREA:

  ARF value at nominal energy (1.49 keV)

  SCAL_MAX:

  maximum value of BACKSCAL column in source spectrum

  SCAL_MIN:

  minimum value of BACKSCAL column in source spectrum

  BACKGRND:

  background intensity ( $photons/pixel^2/cm^2/sec$), total-band, merged observations

  PROB_KS:

  smallest single-ObsId KS variability statistic among merged observations

• A ``mean data'' source position estimate (§5.3) is computed as the mean position of the band-limited multi-ObsId event list, and is saved as keywords RA_DATA, DEC_DATA in source.stats. Position uncertainties for each axis are saved as keywords ERX_DATA and ERY_DATA.

• A ``1 deviation root mean square'' (dRMS or 1DRMS) error radius (§5.3) for the position estimate (RA_DATA, DEC_DATA) is computed and saved in the keyword ERR_DATA in units of arcseconds (NOT sky pixels).

• A multi-ObsId DS9 region file is constructed showing the extraction polygons for all observations.

• Plots summarizing the results of MERGE_OBSERVATIONS are presented. These plots are diagnostic tools that help you spot problems in your data, in your catalog, in your CIAO configuration, in the CIAO software, or in the AE software. Some plots may also convey scientifically interesting information about your sources.

Figure 9: Flux and median energy time series for multiple observations of a source, shown on separate sets of axes. Three curves are shown. A binned light curve is shown in black, and the median energy of the events in the light curve bins is shown in red; the unequally sized time bins are represented by horizontal bars shaped like the letter ``H''. Example Poisson errors within a single time bin are represented by black and red error bars along the Y-axis. An adaptively smoothed light curve is shown in blue.

Figure 10: The same flux and median energy time series in Figure 9, shown on a single non-physical time axis. Time intervals between observations are indicated by black dotted vertical lines.

7.8.2 Usage and Inputs

acis_extract, srclist_filename, /MERGE_OBSERVATIONS, MERGE_NAME=name, EBAND_LO=energy, EBAND_HI=energy, OVERLAP_LIMIT=value, /MERGE_FOR_PB, /MERGE_FOR_POSITION, /MERGE_FOR_PHOTOMETRY, MIN_QUALITY=value

acis_extract, srclist_filename, /MERGE_OBSERVATIONS, /PLOT, CARTOON_TEMPLATE=cartoon_template

EXAMPLE:
idl $\vert\&$ tee MERGE_OBSERVATIONS.log

eband_lo = [0.5, 0.5, 1.0, 2.0, 4.0, 6.0]
eband_hi = [8.0, 1.0, 2.0, 4.0, 6.0, 8.0]
acis_extract, 'all.srclist', /MERGE_OBSERVATIONS, EBAND_LO=eband_lo, EBAND_HI=eband_hi

acis_extract, 'all.srclist', /MERGE_OBSERVATIONS, /PLOT, CARTOON_TEMPLATE='1875.emap'

• In situations where you wish to construct photometry, spectra, responses, etc. on a subset of the observations you've processed, you can supply an optional obsname parameter (§7.3) after the srclist_filename parameter. The parameter should be a vector of observation names, e.g. ['6420','6421','6403','8460','8461']. This unusual mode of execution is not shown in the example above to avoid confusion.

• The optional parameter MERGE_NAME allows the observer to name the merge session; the stage's output files will be structured with an extra directory level using the supplied name, separating those files from the results of other merge sessions. This parameter is analogous to the optional EXTRACTION_NAME parameter (§7.3) which can be used to retain multiple named extractions.

• EBAND_LO, EBAND_HI (optional) should be vectors defining the desired energy bands (in keV) over which photometry will be performed. If omitted these bands are used:
  $\begin{array}{ccccc} 0.5:8.0 \\ 0.5:2.0 & 2.0:8.0 \\ 0.5:1.7 & 1.7:2.8 & 2.8:... ...5:1.0 & 1.0:2.0 & 2.0:4.0 & 4.0:6.0 & 6.0:8.0 \\ 0.5:7.0 & 2.0:7.0 \end{array}$

• The optional parameter OVERLAP_LIMIT (range [0,1], default 0.1) specifies the threshold applied against each extraction's OVERLAP property (assigned by the tool ae_make_catalog; §7.5) to decide if that extraction should be discarded, i.e. not included in the merge (§5.13). If all extractions have excessive OVERLAP, then the overlap limit for that source is relaxed enough to allow at lease one extraction to be merged, and a warning is printed. The source properties OVRLP_LO and OVRLP_HI in source.stats record the OVERLAP range of the extractions merged.

• If /MERGE_FOR_PB is specified, then the set of extractions merged will be chosen to optimize source validity (§5.13). The MIN_NUM_CTS parameter can be used to impose a minimum requirement on SRC_CNTS in the merge, for example to avoid asserting the validity of a source with only two counts (see our AE recipes (§7.1) for example usage).

• If /MERGE_FOR_POSITION is specified, then the set of extractions merged will be chosen to optimize the accuracy of the source position estimate (§5.13).

• If /MERGE_FOR_PHOTOMETRY is specified, then the set of extractions merged will be chosen to optimize source signal-to-noise ratio. The MIN_QUALITY parameter must be supplied to specify the minimum acceptable ratio between the signal-to-noise ratio achieved by the merge and the optimal signal-to-noise ratio achievable by discarding more extractions (§5.13). For example, specifying MIN_QUALITY=0.50 means that you want AE to discard observations, risking photometric bias, only when necessary to prevent the signal-to-noise ratio of the merge from dropping below one-half of its optimal value.

• The optional keyword parameter CARTOON_TEMPLATE should be the name of a FITS image that will be used to define the ``scene'' (image dimensions, pixelization, and placement on the sky) used to build two cartoon images (cartoon_acis.img, cartoon_sky.img) of the point sources. The first cartoon uses the ACIS PSF scaled by the flux in the first photometry band; the second uses a Gaussian PSF of fixed size (currently hard-coded to FWHM=3 image pixels).

7.8.3 Data Products

For each source the following files are produced:

{sourcename}/source.stats:

a FITS file containing keyword information about the source

{sourcename}/source.psf:

a FITS file containing a multi-ObsId PSF image

{sourcename}/neighborhood.evt:

a FITS event list covering the immediate neighborhood of the source

{sourcename}/source.evt:

a FITS table containing the events extracted for the source

{sourcename}/{sourcename}.pi:

the multi-ObsId source spectrum (not grouped)

{sourcename}/{sourcename}_bkg.pi:

the multi-ObsId background spectrum (not grouped)

{sourcename}/{sourcename}.arf:

the multi-ObsId ARF

{sourcename}/{sourcename}.rmf:

the multi-ObsId RMF

{sourcename}/{sourcename}.sequenced_lc.ps:

a PostScript plot of the light curves and median energy time series for each observation shown on a single ``broken'' time axis.

{sourcename}/{sourcename}.stacked_lc.ps:

a PostScript plot of the light curves and median energy time series for each observation shown on separate time axes.

{sourcename}/source.photometry:

a FITS file containing a table of the photometry values described in §5.10

{sourcename}/extract.reg

A DS9 region file showing the extraction polygons for all observations.

cartoon.img:

a FITS image containing a cartoon of the field

Several interactive plots show characteristics of the merged extraction; see our AE recipes (§7.1) for descriptions of each.