Appendix C

When to use ``super'' orientation versus single-band orientat ion?

Definitions:

mag(r=10) == Integrated flux mag for radius = 10.

sup(ecc) == (b/a) axil ratio using "super" J+H+K coadd.

sup(phi) == Position angle of ellipse using "super" J+H+K coadd.

maglim == Limiting magnitude at which the "super" orientation is used instead of those obtained from the individual images.

super coadd == Simple J+H+K addition of the individual band images (with background subtracted). No weights or calibration fudge factors are applied. The typical improvement in the noise is around sqrt(2) (because the J and H sensitivity is similar, while the K sensitivity is considerably less than J) for most of the sky (excluding the galactic plane).

Description:

For faint galaxies, it is very difficult to measure an elliptical orientation due to the poor signal to noise ratio. For this reason it is prudent to use the super-coadd derived orientation (straight J+H+K image), which should have a lower noise by sqrt(2) or so depending on the extinction and color of the galaxy. However, our limits must be set carefully so that we do not bias the sample from a true ``K band'' sample (note: by using ``super'' orientation values, we are effectively using an average between the J and H orientation values, with K only important when the galaxy is either very red or very extincted by dust). The current method for switching to ``super'' orientation values is to compare the measured flux of the source using radius = 10 with a preset limiting magnitude (a tunable parameter, placed in the ``galworks'' namelist). If mag(r=10) is greater than the maglim then the ``super'' orientation values (sup(ecc) & sup(phi)) are applied otherwise the individual band elliptical orientation values (ecc & phi) are used. The current limiting magnitudes for this procedure are:

K(maglim) = 13.5

H(maglim) = 13.9

J(maglim) = 14.2

These values were chosen to reflect an approximate SNR of 5 to 15 for radius = 10; further refinement is required when real 3-channel data becomes available for analysis. An alternative method to using limiting magnitudes is to directly use the measured SNR for the R=10 integrated flux. Here we could apply a limiting SNR of 10, for example, to set the orientation switch from an individual measure to a ``super'' measure. This method will be explored with future analysis.