The radio positions of sufficiently compact sources were measured using the Mk IA-Mk II interferometer, also at a frequency of 966 MHz. A position accuracy better than 2" rms has been obtained for most sources.
Optical positions with an accuracy of ~0.5" rms have been measured for candidate identifications, using print copies of the Palomar Sky Survey. The combined radio and optical position errors are sufficiently small for reliable identifications to be made with any type of optical object appearing on the Sky Survey prints.315 of the sources have positive identifications, with an overall reliability of about 90%. Finding charts are presented for new identifications. 1 Introduction For statistical studies of radio sources, it is important to have a large sample selected in an unbiased way, with optical identifications complete to a known magnitude limit. Accurate radio and optical positions are needed to obtain reliable, complete and unprejudiced identifications on the basis of positional coincidence of the optical object with the radio source (see McEwan, Browne & Crowther 1975).
A few such optically identified samples already exist. The 3CR sample, selected at 178 MHz, has been identified by Longair (1965) and Wyndham (1965, 1966). Samples of sources selected at other frequencies include the NRAO 5-GHz survey (Johnson 1974) and the 2700-MHz Parkes +-4 deg survey (Wall, Shimmins & Merkelijn 1971) which was reidentified using better positions by McEwan et al. (1975). The optical identifications of sources selected at higher radio frequencies contain a higher proportion of QSOs, whilst samples selected at low frequencies contain more galaxies.
This paper presents results from a continuing study of a complete sample of 779 radio sources selected from a survey at 966 MHz covering an area of sky between declinations +40d and +71d. Little previous systematic identification has been attempted in this declination range for sources not in the 3CR sample. Caswell & Wills (1967) have compared positions of sources in the 4C catalogue with positions of bright galaxies in the catalogues of Vorontsov-Velyaminov and Zwicky. Bailey & Pooley (1968) used declinations from a fan-boam survey of a strip 47 < DEC < 51 , together with right ascensions from the 4C survey to make identifications of 4C sources in this area. Willson (1972) used the same method to identify sources in a strip of sky 22h<RA<08h, 38.9<DEC<41.5. These three papers relied on positions from the 4C catalogue, and the identifications are incomplete owing to the poor position accuracy.
We have measured accurate radio positions for sufficiently compact sources from the 966-MHz survey (those with angular sizes <60"). A search for optical identifications has been made to the magnitude limit of the Palomar Sky Survey, which is typically 20.0 mag on the red ('E') print, and 21.1 mag on the blue ('O') print. The radio positions and identifications are presented here. A detailed discussion of the identifications will be reserved for a later paper. Studies are being made of the radio spectra and structure of these sources, and optical spectra are being obtained for quasar candidate identifications. One of us (RWP) is measuring radio positions for the extended sources (angular sizes ~60") using the 300-ft telescope at NRAO at a frequency of 5 GHz. Identification work is already in progress for these sources.
The 966-MHz survey was confusion limited, and the lower flux density limit of the sources selected is 0.7 Jy, several times the rms confusion error. 779 of these sources lie in a well-defined area of 1.32 sr nominally between declinations 40.3 and 71 but excluding |b|<10d and a few small areas extending to slightly higher latitudes.
Radio positions determined for these sources from the survey scans have an rms error of about 2 arcmin, which is not sufficiently accurate to be used as a basis for a search for optical identifications. The rms error in the 966-MHz flux density, S. is [(0.095)2+(0.2)2]1/2 Jy, where 0.2 Jy is mainly due to confusion with a very small contribution due to noise.
-------------------------------------------------- Frequency 966 MHz (lambda=0.310 m) Bandwidth 10 MHz Mk IA beamwidth 18 arcmin Mk 11 beamwidth 32.5 x 41 arcmin Polarization used Left circular Baseline length 425.20 m (= 1370 A) RA and dec of baseline = B081455.7+280447 --------------------------------------------------
The observing parameters and a diagram of the 966-MHzreceiving system used for the present observations are shown in Table I and Fig. 1. The computer sets the compensating delay cables every integration period (normally every 50 s), drives the phase switch, and analyses the signals from the receivers. The analogue multiplier gives two output channels in phase quadrature. The outputs from the voltage-to-frequency converters are counted for each half cycle of the phase switch and the counts are passed into the computer. Here the differences between pairs of half cycles are averaged over a quarter of a second. Those samples which lie outside 3D from the mean of the remainder are rejected to remove some of the effects of short bursts of interference.
Each pair of 1 s averages forms the Cartesian components of a vector which is then rotated by an angle (do of Section 2.3) calculated from the phase equation using an assumed position for the radio source. The rotated vectors are added together until the end of the integration period, when the mean vector is rotated by an amount dependent on the particular delay cables in circuit during the integration. The amplitude and phase of the resulting vector are printed out and a new integration is begun.
For each source, fringe Risibilities at each hour angle were calculated relative to the 966-MHz flux density determined from the survey. A correction was made for position offset of the source from the centre of the interferometer fringe envelope due to the error in the starting position.
Errors in fringe Risibilities are mainly due to confusing sources, noise errors being negligible. The expected rms confusion error on an interferometer fringe amplitude can be estimated by assuming a distribution of confusing sources (S<0.7 Jy) in space and flux density, and integrating vectors due to such sources over the instantaneous interferometer spatial response. The value obtained is 0.11 Jy rms giving an error in fringe visibility of [0.112+(0.09S)2+0.22]1/2/S, where [(0.09S)2+0.22]1/2 is the contribution due to the error in the 966-MHz flux density, S (see Section 2.1). For example, the error on the fringe visibility of a source with S= I Jy would be about 0.24, most of which is due to the error in S.
Flux density range Percentage of IF sources resolved ------------------------------------------- Flux density range Percentage of IF sources resolved 0.7 < S < 1.6 Jy 35 8.8d 1.6 < S < 3.0 Jy 21 3.9d S > 3.0 Jy 22 2.1dAssuming that the confusion error follows approximately a Rayleigh distribution, about 13% of the source measurements will suffer a contribution to up greater than 2sigma_CF. Thus, 13% of the sources in the range 0.7 < S < 1.6 Jy, for which lack = 8.8, have a contribution to ap greater than 17. This is sufficient to make these sources appear resolved, and accounts for the excess number of 'resolved' sources in the lowest flux density range. However, of the stronger sources (So 1.6 Jy and OF < 3.9), less than 0.2% have such a large contribution to up due to confusion, so a large value of op for one of these sources would truly indicate an angular size greater than about 60". The percentage of such sources (about 20%) is in excellent agreement with the results of Mackay (1971) for the 3CR sample.
(i) If the object appeared brighter on the blue print than on the red, it was classed 'blue'. Such objects have mb-mr < 1.1 mag. For 'neutral' objects, appearing equally bright on the two prints, mb-maw 1.1 mag, and for'red' objects mb-mr > 1.1 mag.
(ii) The object was assumed to be a galaxy if its image on either print appeared noticeably extended, otherwise it was classed as 'stellar'. Very faint objects were not classified in this way, however, as it is difficult to make this distinction for very small images. The very faintest 'objects' examined on the prints may sometimes be no more than statistical fluctuations of grain density. An inspection of plate copies of the Sky Survey or deeper photographs would help to verify these identifications. The classification scheme adopted is shown in Table 5.
BSO,NSO,RSO Blue, neutral or red stellar object respectively. QSO Confirmed by optical spectroscopy (also includes three 'BL Lac-type' objects with continuous optical spectra). BO,NO,RO Blue, neutral or red object whose image is too faint to class as stellar or extended. G Galaxy (extended image on at least one print). DG Double galaxy or two very close galaxies. GCL A galaxy in a cluster. CL Cluster. E Empty field. NI No identification. Either the optical field is very crowded or totally obscured. : A less certain identification (see Section 3.6) or an extremely faint object.
Twelve quasars from the 3C sample which were measured during this programme also have accurate optical positions measured relative to the AGK3 reference system by Argue & Kenworthy, AK (1972). Twenty of the identifications in the present sample were also measured by Wills, Wills & Douglas, WWD (1973) using the SAO star catalogue. The mean and rms positional differences in ~ and ~ between present optical positions (J) and positions from these two papers are shown in Table 6.
Table 6. Comparison of optical positions ---------------------------------------------------------------------- No. of dRA(") rms_dRA(") dDEC(") rms_dDEC(") positions +/- +/- compared A K-J 12 -0.02 0.15 0.52 +0.19 0.14 0.50 W W D-J 20 +0.16 0.09 0.40 0.21 0.18 0.82 ---------------------------------------------------------------------- N.B. dRA and dDEC are mean position differences, and rms_dRA and rms_dDEC are rms differences in RA and DEC respectively. ----------------------------------------------------------------------
We find no significant systematic difference in either comparison, and the rms differences are consistent with errors of 0.5" rms in each coordinate for J and WWD, and about 0.15" quoted for AK.
Hunstead (1971) and Hoskins et al. (1974) find a systematic difference between positions based on the SAO and AGK3 catalogues. However, no such difference is found in comparison WWD-J, or in a comparison by Wills et al. of WWD positions with 13 positions based on AGK3 measured by Kristian & Sandage (1970).
The photometric standard stars in the M3 globular cluster (Table 1 of Johnson
& Sandage 1956) were used as a calibration sequence, including the stars Fl
to F12 in the faint extension. The image of each identification was compared
visually with images of stars in M3 on both red and blue prints. The apparent
magnitude of the identification on the blue print, mb, was interpolated from
Johnson & Sandage's B magnitudes for stars most similar in appearance to the
identification. Similarly, the apparent red magnitude, mr, was interpolated
from values of R calculated for the stars in M3 using the equation
R= V-0.3(B-V) (Perek 1958).
To assess the accuracy of the magnitude scales, mr and mb were estimated for
38 quasars from the 3C sample, and compared with photometric magnitudes
measured for these quasars by Sandage (1972). A systematic error of +0.3 mag
in both mr and mb was revealed, which may be due to the difference between
the sensitivity of the Sky Survey print showing the M3 cluster, and the mean
sensitivity of all the other prints used. A correction for this effect has
been applied to all values of mr and mb. After this correction, the rms
difference between estimated magnitudes and Sandage's photometric magnitudes
was 0.5 mag. This includes any effects due to optical variability of the
quasars, and the true rms errors in mr and mb may be less than this.
The rms error in the colour, mb-mr, was estimated in a similar way by comparison with B-R values derived from Sandage's data. The value obtained was 0.4 mag rms, considerably less than the combined errors in mr and mb.
The comparison of diffuse galaxy images with stellar images of the calibration sequence requires some integration by eye over the extended parts of the galaxy, and is less accurate. For very bright galaxies, reference was occasionally made to the sequence of galaxies near Selected Area 57, which have R magnitudes estimated by Abell (1958).
The errors in galaxy magnitudes were assessed by comparing estimates of mr and mb obtained for 20 radio galaxies from the 3C sample with red and blue magnitudes measured for these galaxies by Sandage (1973). For fainter galaxies only (mr> 15.5 mag), both mr and mb were found to be systematically numerically high by between 0.5 and 1.5 mag. This error may be due to the relative insensitivity of the prints to faint extended nebulosity, which is thus omitted in the estimation of the brightness of the image. The size of the error will then depend on the fraction of the optical emission of the galaxy which comes from the extended regions. This is small for N-type galaxies which have a bright compact nucleus, or for very bright elliptical galaxies with burned-out images. However, the systematic errors in mr and mb seem to be roughly equal for an individual galaxy, and so its colour, mb-mr, will be approximately correct.
It is generally difficult to tell whether a very faint image is stellar or extended, and in this case it is not clear whether the magnitude is likely to have been over-estimated. For these reasons, and also because the size of the systematic error is not well known, a correction has not been applied to mr and mb for galaxies or for any faint objects.
The surface densities of different types of optical object were obtained by counting and classifying objects detectable on the prints within 54 randomly chosen areas in the declination range of interest (40 < DEC < 70) between galactic latitudes-20 and +80. Regions with |b|<10d were not included. A total area of 405.5 arcmin2 was searched. This was done during the identification programme so that the classification scheme and limiting magnitude adopted would closely resemble those applying to the identifications themselves.
---------------------------------------------------------------------- Type of object Mean density (10-4"^-2) (a) For |b|>= 20 deg BSO 0.21 Faint G (m_r>18.0) 2.58 Bright G (m_r<=18.0) 0.05 RSO + NSO 1.02 Very faint objects 1.30 Total 5.16 (b) For 10 deg <=|b|< 20 deg BSO 0.81 RSO + NSO 8.71 Very faint objects 2.22 Galaxies (as above) 2.63 Total 14.37 ----------------------------------------------------------------------Table 7 shows the mean densities found for various types of object. For low galactic latitudes, 10 deg<=|b|< 20 deg, the densities of stellar objects and faint objects are much higher than for |b|>=20 deg. The density of galaxies is not significantly different, however, and a constant value of background density was used for these objects.
To allow for the possibility of finding such identifications in the 966-MHz sample, a three-stage identification procedure was followed:
(i) Initially the assumption was made that the radio-optical positional
differences, X and Y. are due only to random Gaussian position errors. In
this case an optical identification is acceptable if it lies within the
ellipse described by
(X/3 sigma_X)^2 + (Y/3 sigma_Y)^2 = 1
where sigma_X and sigma_Y are the combined rms optical and radio position errors in
right ascension and declination respectively, and are typically 1.8".
Only 1.2% of the identifications should lie outside this ellipse if
the initial assumption were true. For accurate radio and optical positions,
the area of this 3a ellipse (typically ~ 100 arcsec^2) is sufficiently small
that the number of random objects expected is very low. For galactic
latitudes |b|>20d, the number expected is about 5 x 10-2 objects per search
area. Thus identifications can be made with confidence with objects of any
magnitude, colour or type. If two objects appear within the 3a ellipse (this
happened in only a very few cases), the object that is brighter, or closer to
the radio position, or of a more rare type was accepted as the
identification. Objects known to be galactic stars were not accepted.
(ii) If no identification was found within the 3 ellipse, the criterion was relaxed and objects within 15" of the radio position were considered. Sources with high interferometer visibilities have angular sizes < 30", and are unlikely to have an identification further than ~15" from the radio centroid (see McEwan et al. 1975). The number of chance objects expected in this extended search area is, of course, much greater, and only particular types of objects with low sky surface densities such as BSOs or bright galaxies may be accepted as reliable identifications.
(iii) If no acceptable identification was found in either search area, the field was generally classed 'empty'. However, five noticeably extended sources have been identified with objects outside the 15" limit. The possible number of identifications missed because of such large radio-optical separations will be discussed in Section 3..
A tentative identification, 'CL', is suggested if the optical field contains a cluster, but if it is not obvious which galaxy is the most likely candidate.
The accepted identifications are listed in Table 3 using the classification scheme of Table 5. Finding charts for new identifications are given in Plate 1.
It can be seen from the predicted numbers of individual classes that the misidentifications will be mostly RSOs, NSOs, faint galaxies and other faint objects; the BSO and bright galaxy identifications are more reliable than this (possibly 99% reliable). Over half of the BSOs have already been confirmed as QSOs (see notes on individual optical fields).
(ii) Table 8(b) shows a similar comparison for objects found in the 293 extended search areas. The area of the 3a ellipse has been omitted from the calculation of numbers of random objects, since it was considered in
Table 8(a). A column has been included to show the number of identifications expected to lie outside the 3D ellipse because they have radio-optical position differences in the tail of the error distribution.
Table 8. Comparison of the number of identifications with expected number of chance objects.
(a) For 534 unobscured 3-sigma search ellipses. Identification class Number Number also Expected number identified found but not of background accepted as objects identifications BSO+QSO 91 1 1.7 RSO+NSO 7 8 13.6 Faint galaxies (m_r>18.0) 66 1 13.8 Bright galaxies (m_r<18.0) 27 0 0.3 Faint objects (RO+NO+BO) 46 3 7.8 Total 237 13 37.2 CL 2 NI 2 E 293 ------------------------------------------------------------------------ (b) For the 293 extended search areas. Identification class Number Number also Expected number Number expected identified found but not of background in the tail of the accepted as objects error distribution identification BSO + QSO 20 6 6.1 1.1 RSO + NSO 0 26 47.0 0.1 Bright galaxies (m_r<18.0) 7 0 0.9 0.3 Faint galaxies and other faint objects 32 80 73.5 1.4 Total 59 112 127.5 2.9 CL 12 E 222 --------------------------------------------------------------------------The high number predicted for random RSOs, NSOs, faint galaxies and other faint objects prevents reliable identification with any such object having a large radio-optical separation. However, comparing the predicted number of chance RSOs and NSOs with the number found, it seems unlikely that any such identifications have been missed. Similarly, comparing the number of faint galaxies and faint objects found (a total of 112) with the number expected by chance (74) it is evident that the proportion of true identifications among them is low. Thirty-two of the most likely candidates in this class have been suggested as identifications, namely those which are just fainter than mr = 18.0, or lie only just outside the 3a ellipse. These identifications are generally bracketed in Table 3. Some of them may be incorrect, and some true identifications will have been overlooked, but from the number in Table 8(b) we conclude that these cannot be more than 10-20 objects. Measurements of radio angular structure may help in these cases.
In contrast, the numbers of BSOs and bright galaxies found in the extended areas are many more than predicted by chance or expected in the tail of the error distribution. Only two or three of these can be expected to be misidentifications, implying a reliability better than 90%. Again a number have already been confirmed as QSOs. This result is in agreement with Wills et al. (1973), who also found galaxies and BSOs between 5" and 10" from their radio positions in excess of the numbers expected by chance. Some of their BSOs have also subsequently been confirmed as QSOs (Wills & Wills 1974).
Since reliable BSO and bright galaxy identifications can be made up to 15" from the radio position, no such object should have been overlooked.
(iii) Only five identifications were made with objects further than 15" from the radio position (a QSO two BSOs, and two bright galaxies). To estimate the possible number of BSOs missed because they lay more than 20" away, the number of BSOs appearing between 20" and 30" from the radio positions were counted. Only six were noted, compared to an expected chance number of 5.9, so probably very few BSO identifications have been missed in this way. Radio sources sufficiently extended and asymmetric to exhibit such a radio-optical separation would generally have been heavily resolved with the interferometer, and would therefore be excluded from this sample.
Table 9. Identification statistics. Identification Number Percentage of all classification identified 'unresolved' sources -------------------------------------------------- G+GCL+DG 117 21.7 BSO+QSO 114 21.2 BO 24 4.5 RO 35 6.5 NO 4 0.7 NSO 4 0.7 RSO 3 0.6 CL 14 2.6 Total positive IDs 315 58.6 E 217 40.3 NI 6 1.1 Sub-total 538 100.0 'Resolved' sources 241 Total 779 --------------------------------------------------The identifications for the 241 resolved sources will complete the sample.
KEY TO TABLE Table 3. Column 1: Source name, using Parkes-type truncated RA-dec. This is followed by an asterisk if |b|<20d. 2: Catalogue name from other survey lists: 3C Edge et at (1959), Bennett (1962). 4C Gower, Scott & Wills (1967). 4CP (Pencil beam) Caswell & Crowther (1969). OA-OZ Kraus (1964); Kraus, Dixon & Fisher (1966); Brundage et at (1971). NRAO Pauliny-Toth, Wade & Heeschen (1966). LHE Long, Haseler & Elsmore (1963). V McLeod ef at (1965). DA Galt & Kennedy (1968). 3: 966-MHz flux density in Jy, measured in the survey. 4,5: Maximum and minimum fringe visibilities measured with the interferometer, relative to the survey flux density. A blank in these columns indicates that the source position was not measured with the Mk IA-Mk II interferometer. These sources are mostly position calibrators which are included in Table 2 with references to previously published positions and quoted position errors. 6: RA (1950) in h, min, s. 7: Rms error in RA, in arcsec. 8: Declination (1950) in degrees, arcmin, arcsec. 9: Rms error in dec, in arcsec. 10: Identification type (as in Table 5). 11,12: Magnitudes estimated on the red and blue prints respectively. Rms error ~0.5 mag (see Section 3.3). 'NV' indicates that the object is not visible on that print. Bracketed values are very rough estimates only. 13,14: Optical position of the identification expressed as offsets in arcsec in RA and dec respectively, in the sense 'optical position minus radio position'. Rms error ~0.5" in each coordinate.