Information on Service 'Yale/San Juan SPM4 Catalog'

Astrometric Reductions

All SPM plates were scanned with the PMM at NOFS. The raw pixel data from the scans were stored and sent to USNO for analysis. The existing StarScan pipeline (Zacharias et al. 2008) was heavily modified by USNO staff to accomodate the SPM pixel data. The overall process included a conversion of the PMM transmission values into density values, smoothing the data for the purposes of image detection and background fitting, and then fitting the unsmoothed 2-d density profiles with an azimuthally symmetric exponential function. (Tests using an elliptical exponential function showed no improvement over the azimuthally symmetric one, even for the slightly elongated 1st and 2nd-order diffraction images.) The derived image positions on each of the 884 PMM CCD footprints required to cover an SPM plate were then transformed into a single global coordinate system using information from the overlap regions of adjacent footprints and the laser interferometric metrology of the footprint centers. As the resulting astrometry from all first-epoch SPM plates was included in the construction of the UCAC3 catalog, further discussion of the PMM data analysis can be found in Zacharias et al. (2010). The USNO-derived centers and image parameters for all detections on the SPM plates, both first and second-epoch plates, were then provided to the Yale SPM team for subsequent reduction.

The SPM CCD frames are corrected for bias, dark (in the case of the Apogee frames, dark current is neglible in the PV), and flatfielding. SExtractor is used to identify detections, give aperture photometry, and provide preliminary x,y centers. Final x,y centers are derived by fitting two-dimensional elliptical Gaussian functions to the image intensities. See Casetti-Dinescu et al. (2007) for further details of the astrometric reduction procedures used with the PixelVision camera data.

In general, similar techniques to those used in constructing previous versions of SPM catalogs were used to build the SPM4. (See Girard et al. 1998, Platais et al. 1998, Girard et al. 2004.) Stars for which both the central-order exposure and first-order grating-image pair were measurable were used to derive and correct each plate's magnitude equation individually. Following the procedures developed for the SPM1 and SPM2 catalogs, all extended sources were given magnitude corrections corresponding to their measured magnitude shifted by -0.7.

In the case of the CCD image centers, there were systematic offsets detected between the position of the central image and the mean of the positions of the grating-order pairs. However, these did not follow the behavior expected for magnitude equation (or charge transfer efficiency effects). Therefore, this offset was corrected as such, a simple offset between the image order systems, instead of as a magnitude equation. (See Casetti-Dinescu et al. 2007.)

Measures from all exposures and all grating-image systems were transformed to a single system for each plate and for each CCD frame. The CCD x,y positions were then corrected for a fixed-pattern geometric distortion believed to be linked to the filter. This correction "mask" was built up from residuals of hundreds of frames at different pointings reduced into UCAC2 coordinates. The corrected CCD x,y positions were then transformed onto the system of the UCAC2 to facilitate pasting together the roughly 50 to 100 frames (depending on whether it had two-fold or single coverage) that comprise a 6 deg x 6 deg SPM field. An overlap method is employed to perform this task, using Tycho2 stars as an external reference system to ensure that systematics from the individual overlap solutions do not accumulate. In this manner, an artificial "pseudo-plate" is built up from CCD frames. This pseudo-plate can then be treated the same as a real second-epoch plate.

In previous versions of SPM catalogs, first- and second-epoch plate pairs were combined to yield relative proper motions per plate pair. These were then corrected to absolute proper motions using external galaxies in the case of SPM1 and SPM2, or Hipparcos star proper motions in the case of SPM3. For the SPM4, instead of combining plate pairs, we have decided to construct the best possible position catalogs at first and at second epoch, over the entire coverage area. This is accomplished by dividing the plates into three groups; first-epoch plates, second-epoch plates, and second-epoch pseudo-plates, then combining plate data within each group using a plate-overlap strategy as follows.

Within each of these plate groups, all plates are pushed through the software pipeline that performs the preliminary reductions described above. This pipeline combines multiple images of the same star (short/long exposure and diffraction orders), corrects the positions for magnitude equation, and then models these x,y positions into RA,Dec from a subset of the Tycho2 catalog adjusted to the epoch of the plate. (The subset is roughly half the Tycho2 stars, those of better quality.) The plate model consists of classical 5th-order distortion terms plus a general 3rd-order polynomial. Uncertainties as a function of magnitude are derived from the scatter of stars with multiple images measured on that plate.

With each plate having been reduced into RA,Dec on the system of Tycho2, we then make use of the overlapping areas to make further adjustments of each plate. This is done in an iterative approach as opposed to a simultaneous global solution. The procedure we're using is to

1. create an "internal+external" reference catalog by doing a weighted average of mutiply-measured stars' RA,Dec and supplementing this with Hipparcos and Tycho2 positions/proper-motions,
2. model each plate into this ref catalog with a general 3rd-order model plus classical 5th-order distortion, correcting the plate's RA,Decs at the end of the iteration,
3. examine the differences in positions, obtained before and after this iteration. If it's still subtstantial, go back to step 1/.

The presence of the Hipparcos/Tycho2 stars in the reference catalog prevents errors from the overlap solutions from accumulating and causing a reference system drift. The number of iterations required for convergence was from 5 to 9 for the three plate groups.

When all is done, i.e., after sufficient convergence of the iterative solutions, the weighted-average positions for every object on every plate are derived and adopted as the celestial coordinates of that object, at the weighted mean epoch for that particular star.

This procedure was applied to the first-epoch plates, the second-epoch plates, and the second-epoch pseudo-plates that had been pasted together from the PV CCD frame data. For this last group, a second "pasting" of the CCDs was performed using preliminary proper motions derived from a first iteration to update all CCD data within a single pseudo-plate to the same epoch. Also, for the pseudo-plate regions it was realized that there were some "holes" in the sky coverage from several causes. In areas with single-fold sky coverage, inaccurate telescope pointing led to occasional gaps between adjacent PV frames. Additionally, some frames that had passed a quality check at the telescope were later found to have problems that rendered them unusable. Finally, there were a handful of SPM fields for which the second-epoch plates were also unusable and pseudo-plates created from an incomplete number of CCD frames in these fields were constructed in their place. In order to avoid having holes or cracks in the SPM4 sky coverage for want of second-epoch positions in these cases, it was decided to supplement the pseudo-plate fields with second-epoch positions taken from the master input catalog. The additional stars and galaxies added were those with input catalog V estimates less than 17.5 in most areas, but a cutoff of V=16.5 was used in two galactic plane fields. Of course, in order to appear in the final SPM4 catalog, a corresponding detection and measure of the object in the first-epoch plate material must exist. Objects with proper motions derived in this manner can be identified in the catalog, their values of np and nc, the number of second-epoch plate and CCD measures per object, will both be zero.

When completed, first-epoch positions and second-epoch positions on the system of the ICRS were in hand for all detected objects in the 660 fields. Uncertainties in the positions were derived from the (weighted) scatter of multiply measured stars as a function of magnitude and this empirical relation calculated for each object according to its magnitude estimate.

The positions and uncertainties were then combined to yield proper motions and proper-motion uncertainties in a straightforward manner. While in theory these proper motions should be on the system of the ICRS via Hipparcos and Tycho2, and thus absolute, in practice an additional correction is needed. Examining the measured proper motions of galaxies within each field as well as the differences with Hipparcos proper motions at the bright end, it was apparent that a residual magnitude equation remained in the derived proper motions. It was decided to calculate a final correction to absolute proper motion per field that was linear with magnitude, using the mean magnitude of galaxies and of Hipparcos stars on the field. Such a linear correction was derived for all 660 fields. The actual proper-motion correction applied to each star in the catalog was the weighted mean of the corrections for the three closest field centers to the star, weighted by the inverse distance from the field center squared.

We note that the quoted uncertainties, particularly those of the proper motions, are unexpectedly (and possibly unrealistically) low. This may be due to the use of weightings that are themselves uncertain enough that a single measurement dominates the calculated mean, more so than it should. The uncertainties will be studied further and presented in an upcoming paper (Girard et al. 2010). In the meantime, the proper-motion uncertainties should be used with caution.

Photometric Reductions

The B,V photometry in the SPM4 is extremely heterogeneous. In some cases, it is derived from our blue and visual filtered CCD cameras. In some cases, it is derived from the PMM measures of our first-epoch plates. And in the cases where neither of these are available or reliable, it is propogated from the input master catalog. In this latter group one can find relatively good photometry from Tycho2 or less reliable extrapolations to B and V magnitudes from 2MASS J,H,K. As such, it is difficult to estimate uncertainties for much of the B,V photometry listed. Our magnitude errors are as likely to be caused by spurious radius measures or inapproriate extrapolations as they are by signal to noise considerations. Thus, we do not provide individual uncertainty estimates for the B,V photometry listed. We do, however, indicate the source of the B and V values given, be they CCD-based, plate-based, or input catalog values.

For the purpose of identifying which image orders should be searched for within the list of detections on a plate or CCD, a magnitude estimate of each star in the input master catalog is needed. For Hipparcos and Tycho2 stars, the B_Tycho and V_Tycho values (transformed to the Johnson system) were adopted. For almost all other stars, B and V photometry was not available so an approximate extrapolation was derived based on 2MASS J,H,K. Hipparcos and Tycho2 stars, which have both B,V and 2MASS photometry, were used to calibrate each SPM field with a relation of the form

B-J = b0 + b1*(J-H) + b2*(H-K) + b3*J*(J-H) + b4*J*(H-K)

with a similar function for V-J. These were used to provide an approximate estimate of B and V for stars without Tycho2 photometry. For the small fraction of objects without Tycho2 or 2MASS photometry, the objects were assumed to be faint and arbitrarily assigned the magnitude limit of the plate or CCD on which it was expected to fall. Again, these input master catalog B,V magnitude estimates were primarily to aid in identifying the various image orders detected. Only in the case that there was neither SPM plate-based photometry nor SPM CCD-based photmetry did these estimates find their way to the final catalog.

The PV and Apogee CCD frames of the second-epoch SPM survey were reduced in a standard fashion using aperture photometry with calibration into Tycho2 V and B photometry (corrected to the Johnson system). When available, these CCD-based magnitudes are provided in the SPM4 catalog, superseding the other magnitude estimates.

Photographic photometry based on the parameters of the image model fits of the PMM scan data proved to be problematic. Among the various image model fit parameters derived, the fit radius provided the best (although still poor) correlation with external calibrating photometry. For extended sources, the radius was, of course, inappropriate. For such objects the input master catalog's magnitude estimate was retained instead, (unless there existed CCD-based photometry). Also, there was a large, non-gaussian scatter between the radius measures and calibrating photometry, indicating that at times the radius estimate was simply erroneous. Thus, during the SPM4 plate photometric reduction procedure, a comparison was made between the preliminarily derived (radius-based) magnitude and that from the input master catalog. If these differed by more than one magnitude, it was interpreted as evidence that one or the either was in error. Since we could not know which, a somewhat expedient choice was made: the fainter of the two magnitude values was retained, under the assumption that the steepness of the luminosity function implies that it is more likely that the star is faint. Unfortunately, the only relevant flag that was retained per star was whether or not it had passed through the plate photometry portion of the pipeline, not whether the resulting magnitude estimate was truly plate-based or a retention of the input master catalog value. The photographic photometry was disappointingly poor in any event. Thus, the only truly reliable B,V photometry in the SPM4 catalog is that flagged as being CCD-based, i.e., with ib=3 and/or iv=3.