\documentstyle[11pt,aaspp]{article}

\begin{document}
\title{A new sample of GPS sources\footnotetext{To appear in the
Proceedings of The 2nd Workshop on Peaked Spectrum Radio Sources, Leiden
1996, Snellen et al. (eds).}}
\author{A. Marecki\altaffilmark{1}, J. Niezgoda\altaffilmark{1},
        H. Falcke\altaffilmark{2}, S. T. Garrington\altaffilmark{3}, A. R. Patnaik\altaffilmark{4}}

\altaffiltext{1}{TRAO, N. Copernicus University, Toru\'n, Poland}
\altaffiltext{2}{Dept. of Astronomy, University of Maryland, College Park, USA}
\altaffiltext{3}{NRAL, Jodrell Bank, UK}
\altaffiltext{4}{MPIfR, Bonn, Germany}

\begin{abstract}

Using four published catalogues we have compiled a new sample of compact
sources that {\it apparently} exhibit a spectral turnover at GHz
frequencies. For most of these sources the flux densities at low
frequencies ($\sim 400$~MHz) were not available so the convex shape of
their spectra was based on a crude extra\-polation of L, C, and X-band
data. In order to determine the spectral turnovers accurately and to
exclude any flat spectrum sources we successfully measured fluxes of
100~objects at 408~MHz with MERLIN. So far we knew that 37~sources in the
MERLIN phase calibrator list by Patnaik et al.~(1992) were actually GPS
sources rather than typical flat spectrum sources. With these
observations we increased that number to 117. 

\end{abstract}

\section{Sample selection}

By definition, GPS sources have a {\it steep spectrum} at high frequencies
but due to the spectral turnover in the GHz range, their spectra in this
regime are either flat or inverted
($\alpha>-0.5,\,S\nu\propto\nu^{\alpha}$). The samples of GPS sources
published so far appear to be biased, since they mainly consist of steep
spectrum sources in the classical sense. Consequently, to get a large,
unbiased sample of GPS sources, we have started a programme to identify GPS
sources from the MERLIN calibrator list by \markcite{Patnaik et
al.~(1992)}, containing all compact radio sources in the range $35^\circ <
\delta < 75^\circ$ with fluxes $>200$ mJy at 8.4~GHz and a flat spectrum.
The selection criterion of this catalogue is such that, besides typical
flat spectrum sources, it should also contain a large number of GPS
sources. To determine candidate GPS sources, we took X-band fluxes from
Patnaik's list, added L and C-band fluxes from GB catalogues
\markcite{(White \& Becker,~1992)}, fitted second order polynomials to the
data and selected the sources with convex spectral shapes. 31~of these
sources were strong enough at 365~MHz --- the Texas catalogue \markcite{
(Douglas et al.~1980)} --- and sometimes even at 151~MHz --- the 6C
catalogue --- so that we were able to establish the turnover frequency
satisfactorily. 5~sources out of these 31~(0710+439, 0711+356, 1031+567,
1225+368, 2050+364) are already present in a classic "working sample" of
\markcite{O'Dea et al. (1991)}. The remaining 26~are listed in Table 1. 

\begin{table}[t]
\caption{GPS sources present in Texas catalogue}
\begin{center}
\begin{tabular}{llllllll}

\tableline

&0059+581 &0652+577  &1107+485 &1324+574 &1627+476 &1815+614 &2323+478\\ 
&0102+480 &0700+470  &1239+552 &1428+422 &1630+358 &1946+708 &2356+385\\ 
&0627+532 &0750+535  &1256+546 &1532+680 &1745+670 &2253+417\\
&0652+426 &0828+493  &1311+552 &1602+576 &1755+578 &2310+385\\

\tableline
\end{tabular}
\end{center}
\end{table}

For the majority of sources (132~out of 163) we have no low frequency data
i.e. they seemed to be weaker than the sensitivity limit of the Texas
catalogue (250~mJy). Again, some sources out of these 132 (0108+388,
0153+744, 0636+680, 0646+600, 1333+459, 1843+356) are already listed by
\markcite{O'Dea et al. (1991)}. A crude extrapolation of the spectra
applied to high frequency data (L, C and X bands) of the remaining
126~sources gives us only an unreliable estimate of the frequency turnover
and even the convex shape of the spectrum remains uncertain in some cases.
Particularly some weaker sources could either be GPS or flat spectrum yet
having 365~MHz fluxes below the catalogue limit. Another effect that could
have produced a spurious convex shape in our non-simultaneous data is the
variability typically observed in flat spectrum sources. 

\section{Observations and data reduction}

In order to investigate the low frequency part of the spectra of our
sources we have set up a programme of flux density measurements at low
frequency with a high resolution facility (VLA or MERLIN). We decided to
use MERLIN at 408~MHz. The observations were carried out from November
1994 until January~1995. The typical length of our snapshots was
15~minutes. Interference at the Cambridge antenna restricted the useful
bandwidth from that antenna, so it was not used in the subsequent
analysis. Interference and system problems at other sites resulted in the
loss of data for more than 20~sources from our sample. 

Since we know from the VLA observations in A configuration at 8.4~GHz
\markcite{(Patnaik et al.~1992)} that our sources are pointlike we used
the {\it Fringe-frequency-Delay Search} --- (FFD) technique (see eg.
Walker,~1981). The chief advantage of this technique is that it allows the
contribution from confusing sources (which is significant at 408~MHz) to be
eliminated by making a form of low-resolution, wide-field map.
It can be highly automated and therefore it is also fast and
efficient. During the reduction process the scans were divided into
subscans consisting of 128~raw 4~second integrations. For each subscan we
calculated an FFD array and took a value from its very centre i.e.
rejecting all other sources in the beam.

\section{Results}

Having reduced our data we found that 20~sources have just flat spectra.
The remaining 80~sources make a new sample of GPS sources. They are listed
in Table 2. 

Adding those 26~sources which already had low-frequency data available,
(cf. Table 1) we can state that 106~out of the 823 (12.9\%) flat spectrum
point sources in the MERLIN phase calibrator catalogue by
\markcite{Patnaik et al.~(1992)} are indeed GPS sources rather than
typical flat-spectrum sources. (Together with 11~sources already quoted by
\markcite{O'Dea et al.,~(1991)} we have 117~of them.) Two sources from
Table 2. (0513+714 and 1622+665) are also found in WENSS based sample (cf.
Snellen, these proceedings). The well defined selection criterion and a
large number of "new" sources makes our catalogue ideally suited for
statistical surveys of GPS sources. 

\begin{table}[h]
\caption{GPS sources found using the MERLIN survey at 408~MHz}
\begin{center}
\begin{tabular}{lllllllll}
\tableline

 &0001+478 &0148+546 &0514+474 &0758+594 &1125+366 &1320+394 &1544+398 &1941+413\\
 &0015+529 &0153+389 &0537+392 &0849+675 &1138+644 &1321+410 &1607+563 &2000+472\\
 &0046+511 &0251+393 &0559+422 &0851+719 &1157+532 &1337+637 &1614+466 &2005+642\\
 &0051+679 &0307+380 &0621+446 &0924+732 &1206+416 &1338+381 &1622+665 &2013+508\\
 &0058+498 &0335+599 &0651+410 &0925+745 &1226+638 &1357+404 &1724+609 &2014+463\\
 &0102+511 &0336+473 &0708+742 &0939+620 &1232+366 &1403+411 &1753+648 &2151+431\\
 &0123+731 &0338+480 &0718+374 &1019+429 &1239+606 &1436+445 &1812+560 &2248+555\\
 &0129+431 &0412+447 &0732+755 &1035+430 &1245+676 &1454+447 &1828+399 &2300+638\\
 &0129+560 &0424+414 &0753+373 &1055+433 &1300+485 &1533+487 &1839+389 &2310+724\\
 &0140+490 &0513+714 &0753+519 &1101+609 &1308+471 &1534+501 &1937+630 &2341+697\\

\tableline
\end{tabular}
\end{center}
\end{table}

\begin{references}

\reference{Douglas J.N. et al. 1980, {\it The University of Texas Publications,
Publications in Astronomy}, {\bf 17}}
\reference{O'Dea C.P. et al., (1991), \apj, {\bf 380}, 66}
\reference{Patnaik A.R. et al. 1992, \mnras, {\bf 254}, 655}
\reference{Walker R.C. 1981, \aj, {\bf 86}, 1323}
\reference{White, R. L. \& Becker, R. H. 1992, \apjs, {\bf 79}, 331}

\end{references}

\end{document}