The DASCH photometry pipeline calibrates “colorterms” for each plate that it analyzes. These numbers describe the color-dependent corrections needed to best align the raw plate photometry with the photometric reference catalog in effect. The dominant factor affecting the derived colorterm is which photographic emulsion was used.

Background

Different photographic emulsions are sensitive to different wavelengths of light. For the same reasons that you might place place different filters in front of a CCD camera, photographic astronomers used different emulsions for different observations.

The vast majority of HCO plates use one of three emulsions, which we can refer to by their rough bands of color sensitivity: “blue“, “red”, and “yellow”. Most HCO plates were blue-sensitive.

This inhomogeneity in the plates is especially important to keep in mind when looking at DASCH images. If you have a source that suddenly seems to have brightened massively, what you may really be seeing is a switch of emulsions from blue to red — if your source is red too.

For instance, here is a partial lightcurve of TYC 9504-35-1, one of the southernmost DASCH stars in the entire sky:

Around 1972, the lightcurve seems to split into several groups at different brightnesses. This can’t be physical, right?

Indeed it is not. These measurements come from three sets of “Damons South” plates, which used three different emulsions: the “blue“, “red”, and “yellow” mentioned above. (These are plate series dsb, dsr, and dsy.) This star is relatively red (Gaia DR3 T_eff of around 4700 K), and apparently the APASS catalog has a missing or low-quality color measurement for this star. The different groups of measurements correspond to plates using the different emulsions, with different errors as the calibration scheme attempts to adjust all of the magnitudes to the main APASS system (Johnson B) using the information available.

By examining the colorterms associated with each plate, you can check for, or remove, color-dependent systematics from your lightcurves.

Plate-Level Data Products

For each plate being analyzed, the DASCH pipeline may compute dozens of colorterms:

• Two sets of colorterms are computed for each of the two reference catalogs in use: APASS and ATLAS. The colorterms derived for each of these catalogs are systematically different because for APASS, the calibration is against Johnson B and V, while for ATLAS it is against SDSS g and r.
• For multiple-exposure plates, colorterms are calculated independently for each exposure.
• Within each exposure, colorterms are calculated for as many of the nine annular (“spatial”) calibration bins as possible.

All of these colorterms may be obtained from the photometric calibration data product files. TODO: these need to be made available and documented.

Additionally, per-plate colorterms are available in exposure tables as the median_colorterm_apass and median_colorterm_atlas columns. These numbers are derived by taking the median of the annular-bin colorterm results for the first WCS solution for each refcat. (Inspection of the data indicates that colorterms do not generally have any significant systematic variation across the annular bins.) Because each reference catalog is analyzed separately, a plate may have an APASS colorterm but not an ATLAS one, or vice versa.

Aggregate Behavior

The figure below shows a histogram of the DR7 per-plate median colorterm values derived for the APASS reference catalog:

The data underlying this histogram are available as the file apass_colorterms.txt. This is a line-oriented text file where each row contains three tab-separated values: the location of the left edge of the histogram bin, the location of its right edge, and the number of plates in the bin.

The main peak around a colorterm of ~0.1 corresponds to the blue plates, since the APASS catalog calibrates to Johnson B magnitudes. There is a sub-peak at exactly 0.0 because the pipeline sometimes defaults to this value if it cannot obtain a good colorterm solution.

A secondary histogram peak around a value of -0.95 corresponds to plates using “yellow” emulsions. The peak around -1.4 corresponds to red plates. The additional bump between the main yellow and blue peaks, around a value of -0.65, is real, and corresponds to plates using both yellow emulsion and a yellow filter in the optical path.

The following figure shows the same kind of histogram for the ATLAS colorterms:

The data underlying this histogram are available as the file atlas_colorterms.txt. They have the same format as the corresponding APASS file.

The pipeline-induced peak around 0.0 is more obvious here, because the blue plates have ATLAS colorterm values clustering around 0.7. Relative to the ATLAS catalog's g / r magnitudes, yellow and red plates have colorterms around -0.5 and -1.0, respectively. The yellow-filter plates have ATLAS colorterms of around -0.2.

The following figure shows a two-dimensional histogram relating APASS and ATLAS colorterms for the (large) subset of plates that have both values available.

There is an approximately linear relationship between the colorterms derived for the two refcats. The histogram above includes a line obeying the equation:

colorterm_ATLAS = 1.166 * colorterm_APASS + 0.604

This equation was not derived rigorously — it comes from fitting a polynomial to the manually-identified locations of the main histogram peaks in each axis. Using the same inputs, a quadratic fit yields:

colorterm_ATLAS = 0.0915 * colorterm_APASS² + 1.274 * colorterm_APASS + 0.605

However, this solution is not obviously superior to the linear fit. We recommend adopting the linear relation.

Inferring Emulsions

If your analysis depends on the color response of a plate, we recommend basing it on colorterm values, rather than any discrete categorization of plate types. However, if you wish to categorize plates by emulsion type, we recommend the following breakpoints:

• For APASS:
  • Colorterm < -1.13: red
  • Colorterm > -0.53: blue
  • Otherwise: yellow
• For ATLAS:
  • Colorterm < -0.70: red
  • Colorterm > -0.07: blue
  • Otherwise: yellow

In both cases, colorterms that are exactly zero should be treated with suspicion, but might be valid.

To isolate a set of “good” blue plates in the APASS calibration, select those with colorterms between -0.164 and 0.212. This cut selects about 95% of the blue plates, which works out to a sample of around 317,000 if no other filtering is performed. The median colorterm in this subset is 0.055, because the distribution of colorterms in the blue peak is asymmetrical.

To do the same in the ATLAS calibration, select those with colorterms between 0.357 and 0.878. This will again select about 95% of the blue plates, or a sample of around 332,000 plates. The median colorterm in this subset is 0.674.

Other Indicators of Emulsion Type

For borderline cases, there are other clues that can indicate which emulsion a plate used. Some hints are:

• Membership in the dny, dnr, dsy, or dsr series (which stand for Damons North Yellow, Damons South Red, etc.)
• A plate class of “L1”, “L2”, “L2S”, “L3”, or “L4”.
• Writing on the plate envelope, viewable in Starglass may manually annotate the emulsion.

Be aware that DASCH plate-class metadata are often missing or incorrect. In general, pipeline-derived colorterm data should be preferred if quantative analysis is the goal.