2 Sample selection, observations, and data reduction

We selected 15 sources from the O'Dea et al. (1991) working list of powerful GPS sources, with either an "EF'' or undetermined identification, or without a proper redshift determination. To this sample were added several GPS objects from various sources in the literature. Taken from White (1992): 0316+161, 0437-454, 0507+179, 1540-077, 1648+015, and 1815-553; and from O'Dea et al. (1990): 1045+019. The source 1245+676 is an extended radio source with a GPS core. Work on this source is in progress by De Bruyn et al. Furthermore, the spectra of 1848+283 and 2050+364, while observed quite some time ago, are presented here for the first time. Table 1 summarizes the objects, their origin, and the observational details.

In addition, we present 5 minute K-band images taken with the Keck-I telescope of 1942+722, 2121-014, 2128+048, and 2322-040.

2.1 Observations

The first observations of this program were taken at the ESO 2.2 m telescope at La Silla (Chile) in November 1993. This telescope was equipped with an imaging spectrograph (EFOSC2), so immediately after a successful identification a spectrum can be taken, making this setup ideal for our program. Based on our experience from this run in 1993 (Paper I), it became clear large apertures and long exposures are needed to obtain redshifts for most of these faint objects. We obtained imaging and spectroscopy of the southern part of the sample using EFOSC1 on the ESO 3.6 m, and spectroscopy of the northern objects using CRYOCAM on the 4 m Mayall telescope at Kitt Peak (U.S.A.). A few additional observations were obtained at Keck and Lick observatories. The next couple of paragraphs describe the instrumental setups at the respective observing runs.

The Kitt Peak run of January 1996 was on the 4 m telescope with the CRYOCAM instrument. We used grism 810 in combination with the Schott GG 455 blocking filter to suppress the 2$^{\rm nd}$ order below 9000 Å. The unresolved lines had a FWHM of $\sim$30 Å, or equivalently, a 1500 km s^-1 resolution. The spatial resolution of the CRYOCAM is 0.84 arcsecond/pixel, and we used a 2 $\hbox{$.\!\!^{\prime\prime}$ }$5 wide slit. The typical seeing through the slit was on the same order, so a narrower slit would have resulted in light loss.

Our March 1996 run was on the ESO 3.6 m telescope in La Silla. Like the 2.2 m, this telescope was equipped with an imaging spectrograph, EFOSC1, an instrument similar to EFOSC2. The installed chip was ESO CCD #26, which has a 0.61 arcsecond/pixel spatial scale. In combination with the R300 low dispersion grism, this setup resulted in a spectral resolution of $\sim$20 Å (FWHM), quite comparable to both previous runs. Seeing permitting, we used the 1 $\hbox{$.\!\!^{\prime\prime}$ }$5 slit, but on some occasions we had to use the 2 $^{\prime\prime}$ slit. The seeing hovered around the 1 $\hbox{$.\!\!^{\prime\prime}$ }$2 mark on both nights.

Observations of 1848+283 date back to July 1990, and were taken at La Palma's INT telescope in service mode. The telescope at the time was equipped with the Faint Object Spectrograph which simultaneously dispersed the wavelength ranges of 5000-10000 Å in first order and 3000-6000 Å in second order. This setup resulted in resolutions of 20 Å for the red and 10 Å for the blue arm of the spectrograph respectively. The spectrum of the source 2050+364 was taken in October 1991 at the 3 m Lick telescope. The resolution, as measured from sky-lines, was around 15 Å.

The K-band images were obtained on July 6 1998 with the NIRC instrument on the Keck-I telescope, the near-infrared camera (Matthews & Soifer 1994). NIRC re-images the telescope focal plane onto a 256² InSb detector at 0 $.\!\!^{\prime\prime}$15 per pixel, giving a field of view of 38 $.\!\!^{\prime\prime}$4 square. The night was photometric, with a seeing of 0 $.\!\!^{\prime\prime}$4 FWHM. Five dithered 60-second images were obtained on each field.

 

 

Table 1: Source list and observations

	Radio position (2000.0)		1 $\sigma$			Imaging		Spectra
Name	RA	DEC	err ( $^{\prime\prime}$)a	ref	date	exp (s)	ID	exp (s)	Sampleb
0018+729	00 21 27.38	73 12 41.9	0.03	pa	Jan. 96	--	G	3600	O
0159+839	02 07 14.19	84 11 18.7	2.0	sp	Jan. 96	--	Q?	1800	O
0316+161	03 18 57.76	16 28 32.3	2.0	wr	Jan. 96	--	G	3600	W
0437-454	04 39 00.85	-45 22 22.4	0.02	ma	Mar. 96	300	G	1800	W
0500+019c	05 03 21.18	02 03 04.5	0.05	pe	Nov. 93	1200	G	3600	O
0507+179	05 10 02.37	18 00 41.6	0.02	jo	Jan. 96	--	Q?	3600	W
0554-026	05 56 52.62	-02 41 04.5	1.4	me	Jan. 96	--	G	1800	O
0602+780	06 10 24.63	78 01 34.9	2.0	sp	Jan. 96	--	G	3600	O
0703+468	07 06 48.04	46 47 56.2	2.0	do	Jan. 96	--	G	3600	O
0742+103	07 45 33.06	10 11 12.7	0.05	pe	Mar. 96	2400	G	--	O
0802+212	08 05 38.46	21 06 51.4	2.4	do	Mar. 96	600	G	3600	O
0904+039	09 06 41.05	03 42 41.5	1.5	me	Mar. 96	600	G	--	O
0914+114	09 17 16.39	11 13 36.5	2.0	do	Jan. 96	--	G	3600	O
0941-080	09 43 36.90	-08 19 30.9	0.4	wr	Jan. 96	--	G	1800	O
1045+019	10 48 22.84	01 41 47.7	2.0	do	Mar. 96	600	EF	--	O2
1245+676	12 47 33.33	67 23 16.5	0.03	pa	Jan. 96	--	G	1800	O
1433-040	14 35 40.05	-04 14 56.1	1.2	do	Mar. 96	300	G	3000	O
1540-077	15 43 01.69	-07 57 06.8	2.0	wr	Mar. 96	120	G	3000	W
1543+005	15 46 09.58	00 26 24.8	0.4	wr	Jan. 96	--	G	1800	O
1601-222	16 04 01.69	-22 23 41.5	1.3	wr	Mar. 96	240	G	1800	O
1648+015	16 51 03.68	01 29 23.8	0.4	wr	Mar. 96	300	G	3000	W
1732+094	17 34 58.37	09 26 58.2	0.4	dr	Mar. 96	300	G	1500	O
1815-553	18 19 45.40	-55 21 20.8	0.02	jo	Mar. 96	240	G	2400	W
1843+356	18 45 35.11	35 41 16.7	0.03	pa	Jul. 90	--	G	900	O
1848+283	18 50 27.52	28 25 13.1	0.05	pe	Jul. 90	--	Q	900	O
1942+772	19 41 26.87	72 21 43.7	0.4	sn	Jun. 98	300	G	--	S
2050+364	20 52 52.06	36 35 35.3	0.03	pa	Oct. 91	--	G	2400	O
2121-014	21 23 39.17	-01 12 33.9	1.4	me	Jun. 98	300	G	--	O
2128+048	21 30 32.88	05 02 17.5	0.02	jo	Jun. 98	300	G	--	O
2322-040	23 25 10.21	-03 44 46.7	1.4	sp	Jun. 98	300	G	--	O

^a Error in the radio position. To this error one has to add the optical accuracy ( $0\hbox{$.\!\!^{\prime\prime}$ }5 - 0\hbox{$.\!\!^{\prime\prime}$ }7$) in quadrature to obtain $\sigma^2_\alpha$ and $\sigma^2_\delta$ of Eq. (1).

^b Taken from original sample of: O = O'Dea et al. (1991), O2 = O'Dea et al. (1990), S = Snellen et al. (1998a), and W = White (1992).

^c Source is part of the Paper I sample, and is only included again because of the discussion in Sect. 4.

REFERENCES -- (do) Douglas et al. 1996; (dr) Drinkwater et al. 1997; (jo) Johnston et al. 1995; (ma) Ma et al. 1998; (me) McEwan et al. 1975; (pa) Patnaik et al. 1992; (pe) Perley 1982; (sn) Snellen et al. 1998b; (sp) Spoelstra et al. 1985; (wr) Wright & Otrupcek 1990.

2.2 Data reduction

Images were reduced within IRAF. Standard optical reduction procedures were applied to remove the detector signature. Astrometry was performed using HST guide stars, of which typically 2-3 could be identified in the source fields. We used the APPHOT package, available in IRAF, for the V, R, I, and K-band photometry.

Spectral reduction was done with the NOAO longslit reduction package in IRAF. He-Ar reference spectra, plus observations of several standard stars were used to perform the wavelength and flux calibrations.