SUMMARY
The pointing catalog is dominated by the RB co-ordinates from 1950. With Y2K just around the corner, it is time to adjust ourselves to the new epoch.

Contents

• The old catalog - its sources and inconsistencies
• Co-ordinate transformations - comparison with SIMBAD J2000 coordinates
• The B1950 data - comparison with SIMBAD
• Conversations with Patrick Wallace - we use coco correctly
• ICRF - a true inertial reference frame
• Comparison with ICRF -     dRA = -0.05"+0.17",     dDec = 0.01"+0.23" from 100 blazars
• Additional continuum pointing sources - from ICRF lists
• Distribution of continuum pointing sources - and remaining voids
• Conclusions - go with the ICRF system a.s.a.p.

The old catalog
The old catalog mainly contains coordinates described as 'RB'. Loosely this means epoch and equinox 1950.0, and on the FK4 system. In his Examples and Explanations Pat Wallace (PTW) describes how FK4 has problems from which FK5 is essentially free, being a better approximation to an inertial reference frame. A move to FK5 J2000.0 coordinates seems overdue and may itself improve the pointing of the telescope.

Of particular focus in this document are the 107 continuum pointing sources in our current catalog . These are mainly highly compact radio sources - quasars or blazars - bright enough at 1.1 mm to be useful for pointing with our old bolometer, UKT14. The preamble to the catalog says that the coordinates come from one of 4 references :

• Kuhr et al. (1981) Astronomy & Astrophysics Supplement 45, 367
• Perley, R.A. (1982) Astronomical Journal 87, 859
• Hewitt & Burbidge (1987) Astrophysical Journal Supplement 63, 1
• Edelson, R.A. (1987) Astronomical Journal 94, 1150

The positions in Kuhr et al are selected from a heterogeneous list (their Table II) of radio-interferometer measurements made typically in the 1970s, although none of the positions in our catalog correspond precisely to these entries. Perley's radio (1-5 GHz) positions form the basis of our catalog, and while Edelson's data are quoted as being taken from Perley there are occasional minor differences. The Hewitt & Burbidge positions are those of optical counterparts to the radio sources, or are those of nearby galaxies, and only 4 positions in our catalog are precise matches.

Among our 107 continuum pointing sources, CenA (NGC5128) and CygA (3c405) are relatively nearby radio galaxies with discernible small-scale nuclear structures, while 3c111 is not found any of the 4 references above. Of the remaining 104, 54 (52%) are exactly as given by Perley, with 33 (32%) being negligibly different, and 2 being significantly different. 'Negligibly', here and hereafter, means with differences of less than 0.01^s in RA, 0.1" in Dec. The 2 outstanding errors are shown below :

•    3c279      12 53 35.838  -05 31 08.4   RB  JCMT
     3c279      12 53 35.838  -05 31 08.04  RB  Perley
    dRA(JCMT-Perley)   0.00"          0.36"
  

•    1749+701   17 49 03.8    +70 06 39.    RB  JCMT
     1749+701   17 49 03.400  +70 06 39.6   RB  Perley
    dDec(JCMT-Perley)  2.05"         -0.60"
  

Despite these inconsistencies and the heterogenous nature of the coordinates, the implied catalog accuracy suggests that it should provide adequate pointing accuracy for the JCMT. Indeed, the canonical performance statistics quoted over recent years - r.m.s. deviations in (azimuth, elevation) of (1.5", 1.5") - confirm that the limiting factor to the pointing accuracy of the JCMT is not the accuracy of the catalog.

In this light, updating the catalog to J2000 coordinates seems somewhat cosmetic. However, the impending millenium warrants the psychological adjustment and this then offers an opportunity to review the catalog, in order to ferret out inconsistences such as are seen above, as well as to offer the chance of improving the basis of the catalog using recent interferometric measures. As we shall see, such a new foundation is available.

Co-ordinate transformations
Transformations from one system to another are not simple and in any case will generate only mean places , of which FK4 and FK5 are examples. At least for the blazars in our catalog , which are used exclusively for establishing the pointing models using SCUBA , the additional problem of proper motion is not an issue. We shall consider the stellar spectral line pointing sources later.

Our RB coordinates were therefore duly transformed to RJ coordinates using PTW's coco program.

How do these results compare with other reliable sources of positional information - such as the SIMBAD database ? Bearing in mind that SIMBAD quotes coordinates accurate only to 0.01^s and 0.1", the differences between the (RA,Dec) coordinates derived from these two processes are shown below :

Three objects (3c111, QSO 1219+285 and CenA) were omitted from the analysis because the J2000 SIMBAD positions are too coarse, one (QSO 1044+719) because SIMBAD had no coordinates for it at all. Three additional objects showed anomalous differences and these are marked on the plot above. PKS0537 is our southernmost pointing source, while QSO 1749+701 and CygA are 2 of 8 sources with RAs quoted to only 0.1 seconds of time in our current catalog . The omission of trailing zeroes accounts for 4 of these 8 truncations, with CenA and CygA being another two. Perley (1982) does not offer coordinates for the other two (QSOs 0215+015 and 2155-304).

The remaining 100 blazars show the following statistics in regard to the differences plotted above :

• dRA = -0.95 + 0.38 (s.d.) arcseconds
• dDec = -0.04 + 0.21 (s.d.) arcseconds.

• dRA = -1.0" + 0.4"*cos(RA - 2)         where RA is in hours

Are these differences due to errors in our current catalog , or in the SIMBAD database, or in coco ?

The B1950 data
A similar exercise was performed on the B1950 coordinates :

• dRA = -0.93 + 0.41 (s.d.) arcseconds (with similar sinusoid)
• dDec = -0.04 + 0.20 (s.d.) arcseconds.

The differences between our FK5 J2000 coordinates and those in the SIMBAD database are therefore not due to the coordinate transformations. Indeed, applying coco to SIMBAD's B1950 coordinates generates SIMBAD's J2000 coordinates to the accuracy of their display. The difference between our coordinates and SIMBAD's is therefore in the coordinates themselves.

The typographical error described above probably explains the anomalous position of 1749+701 in the above diagram, and CygA looks like a late (uncorroborated) addition to our list. On the other hand PKS0537 is taken verbatim from Perley (1982), so may highlight possible errors of our catalog in the southern hemisphere.

It's not what you know, it's whether you know Pat Wallace that counts !
A conversation with Pat Wallace clarified a couple of points that suggest that the coco-ed B1950 coordinates are accurate.

1. The FK4 system has a rotation (due the rotation of the Galaxy) w.r.t. an inertial frame. Transformation to FK5 requires allowing for this rotation, and coco does this.
2. For objects like blazars, with no proper motion in the inertial reference frame, the correct way to use coco is to enter the B1950 coordinates only, equivalent to implying that proper motion is not known. Fortuitously, I had done it this way. Inputting proper motions of 0.0000^s, 0.000" - i.e. in the (rotating) FK4 system - is incorrect and would lead to errors of 0.1-0.2 arcseconds in RA, and some periodicity such as is seen above. That still leaves 0.8-0.9" of RA error unexplained, of course.

Is there another J2000 database with which to compare our coordinates ?

A true inertial reference frame - the ICRF
The FK5 system was officially replaced on 01 Jan 1998 by the International Celestial Reference Frame (ICRF, see ICRS - the International Celestial Reference System; the foundation of celestial positions). ICRF positional data are derived from Very Long Baseline Interferometry, and the ICRF final orientation is described as being "consistent with the FK5 J2000.0 optical system". I refer you to these documents :

• The ICRS
• The IAU mandate to establish an International Celestial Reference Frame
• The ICRF : including links to papers establishing the ICRF, comparisons with other systems, internal accuracies, etc.
• A paper by Ma et. al. Astronomical Journal Vol 116, p516 including discussions of transformation theory and errors.

Comparison with ICRF
The ICRF catalog (Ma et al 1998, AJ 116, 516) lists (in their Table III) 212 defining extragalactic radio sources, whose positions are known to better than 1 milliarcsecond. 28 of these are in common with our pointing catalog. A comparison of the J2000 coordinates for these 28 is shown below :

The RA residual for QSO 1749+701 is again in error - due to what I believe is a typograhical error as described above - and if this is corrected, the following differences result :

• dRA = -0.02 + 0.26 (s.d.) arcseconds
• dDec = +0.04 + 0.29 (s.d.) arcseconds.

The ICRF further lists 294 candidate sources and 102 other sources, all of which still have positions determined to better than 6 milli-arcsecond. (Note that errors in the ICRF catalog also increase in the south, where baseline coverage is poorer). Our catalog has 101 sources in common with the combination of these 3 lists. (The objects that did not make the cut are QSOs 3c111, 0954+556, 1219+285, 1716+686, CygA and CenA). The remaining 101 J2000 coordinates compare with those in the ICRF as follows :

After correcting the typo for QSO 1749+701 and omitting PKS0537, the statistics of the agreement from 100 objects are

• dRA = -0.05 + 0.17 (s.d.) arcseconds
• dDec = +0.01 + 0.20 (s.d.) arcseconds.

Additional continuum pointing sources
The ICRF potentially provides other pointing sources on this same system and to similar accuracy. Their usefulness to JCMT/SCUBA is then a question of source brightness. Edelson (1987) provides 2.7mm brightnesses and millimeter spectral indices for 176 compact sources, the brighter ones of which already appear in our current catalog . An additional 81 sources appear in the ICRF lists and we should now make use of them. Their usefulness at the telescope (JCMT) depends upon their 0.85mm brightness (which we calculate from Edelson's numbers and show in the table ) and upon the weather conditions. It is quite feasible to point with SCUBA on 0.3Jy sources - even if it takes a few minutes - and the option of spending such time on a faint pointing source close to a target should be available to an observer.

There are useful, i.e. bright, if less accurate, pointing targets in the current catalog that may still be retained. Our position for CenA, for instance, may be judged by comparison with recent citations (e.g. Turner et al, 1997 Astrophysical Journal 475 118, or Hui et al, 1995 Astrophysical Journal 449 592) as being correct to within about 1" in each coordinate, which is still tolerably useful for JCMT. Consequently we will retain CenA as a pointing source, with appropriate caveats.

Distribution of continuum pointing sources
The distribution of the 101 blazars in our current catalog , (coco-ed to J2000) and the 81 additional, generally fainter, sources in the ICRF-Edelson overlap, are shown below :

There are about a dozen or so identifiable voids of diameter 15-20 degrees in this distribution. It might make sense at some time to identify sufficiently bright ICRF sources in these regions and add them to our catalog.

Conclusions

1. The errors inherent to our current catalog or to a new generation thereof are about (0.2",0.2") in (RA,Dec).
2. The differences between our system and SIMBAD is left unexplained - PTW has a few theories about the neglect of the rotation of the FK4 system that might explain some of it.
3. There are no significant systematic differences between the J2000 coordinates of our pointing sources and the ICRF coordinates of these same sources.
4. A new (blazar) pointing catalog for JCMT could be created using the coco-ed B1950 values, and this would retain the errors seen above , or we could take our coordinates directly from ICRF and avoid these (admittedly small) errors.
5. I would suggest we adopt the ICRF coordinates - recall that these are essentially J2000 coordinates.
6. Following the central bearing replacement, there is to be a period during which the pointing is to be "recovered". I suggest that the previous system be used until we pass through this period, but that the ICRF-based pointing list be tested as soon as possible thereafter.