mnr_10578.tex - BOLmnras06.pdf

Mon. Not. R. Astron. Soc.

370,

1556–1564 (2006)

doi:10.1111/j.1365-2966.2006.10578.x

15-GHz variability of 9C sources

R. C. Bolton,

C. J. Chandler,

G. Cotter,

T. J. Pearson,

G. G. Pooley,

A. C. S. Readhead,

J. M. Riley

and E. M. Waldram

Cavendish Astrophysics, Department of Physics, J. J. Thomson Avenue, Cambridge CB3 0HE

National Radio Astronomy Observatory, PO Box 0, Socorro, NM 87801, USA

Oxford Astrophysics, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH

California Institute of Technology, 1201 East California Blvd, Pasadena, CA 91125, USA

Accepted 2006 May 17. Received 2006 April 20; in original form 2005 September 29

ABSTRACT

We present results from a 3-yr study of the 15-GHz variability of 51 9C sources. 48 of these

sources make up a subsample of a larger one complete to 25 mJy in 9C, and as the sources

are selected pseudo-randomly the results should be representative of the complete sample.

29 per cent of this subsample are found to be variable above the flux calibration uncertain-

ties of

∼

6 per cent. 50 per cent of the flat-spectrum objects are variable whilst none of the

steep-spectrum objects or the objects with convex spectra peaking below 5 GHz are variable.

Nine of the objects studied have convex spectra and peak frequencies above 5 GHz; eight of

these were found to vary at 15 GHz, suggesting that the high-frequency peaking class in this

sample is largely populated by objects with jets aligned close to the line of sight whose emission

is dominated by beamed components.

Key words:

surveys – galaxies: active – radio continuum: general.

1 INTRODUCTION

The 9th Cambridge survey (9C; Waldram et al. 2003) has been car-

ried out at 15 GHz with the Cambridge Ryle Telescope (RT; Jones

1991). The main scientific motivation for 9C was the need to identify

foreground sources that could contaminate cosmic microwave back-

ground (CMB) maps from the Very Small Array (VSA; e.g. Taylor

et al. 2003), a CMB telescope operating at 34 GHz. 9C was carried

out by performing blind raster scans of the sky at 15 GHz and then

following up possible detections from the raster maps with pointed

observations (again at 15 GHz) to confirm the detection and measure

the source flux density or to rule it out.

Extensive radio and optical follow-up of 176 sources from 9C

has been presented in Bolton et al. (2004) (Paper 1 hereafter), and

this provides a simultaneous continuum radio spectrum spanning

1.4–43 GHz for each object, as well as an optical identification for

∼

90 per cent of them. 19 per cent of the follow-up sources in a

sample complete to 25 mJy in 9C had spectra with peak frequen-

cies between 0.5 and 10 GHz, double the fraction found in lower

frequency (e.g. 5 GHz) surveys (O’Dea 1998).

Synchrotron self-absorption models suggest that compact young

radio sources (younger than a thousand years or so) should have syn-

chrotron spectra peaking at a few gigahertz or above (O’Dea 1998;

Snellen et al. 2000). If a significant fraction of the gigahertz-peaked

E-mail: r.c.bolton.97@cantab.net (RCB); guy@mrao.cam.ac.uk (GGP);

julia@mrao.cam.ac.uk (JMR); emw1@mrao.cam.ac.uk (EMW)

spectrum (GPS) and high-frequency peaked (HFP) objects found in

9C are genuinely young objects they will provide insights into the

mechanisms that trigger radio sources and their early evolution.

From the initial follow-up data, it is unclear what proportion

of the sample of HFP objects is made up of beamed objects that

are not necessarily either compact or young. Since beamed ob-

jects are likely to have variable flux density at high radio fre-

quency, whilst genuinely young, unbeamed objects should be stable

(or possibly slowly increasing) in flux density, monitoring of the

flux density of the HFP objects should help to identify beamed

objects.

We also consider whether variability can affect the VSA observa-

tions. For example, in the first VSA observations with the compact

array, for the VSA confusion noise to be significantly lower than

the flux sensitivity (

30 mJy) Taylor et al. (2003) state that they

must subtract all sources brighter than 80 mJy at 34 GHz. In order

to achieve this, during the VSA observations, all sources brighter

than

∼

20 mJy in 9C were observed by the VSA source subtra-

ctor telescopes that operate at the same frequency as the VSA. If

sources have varied in flux density so that they are fainter during

the VSA observations than they were during the initial raster scans

or pointed observations of 9C, then this will not compromise the

CMB data; however, any sources which either were too faint to be

detected at the time of the 9C raster scans, or have flux densities in

9C below the cut-off for study with the source subtractor telescopes,

but whose flux densities have since increased, will add uncorrected

noise to the CMB data. It is therefore desirable to study the variabil-

ity of 9C sources at centimetre wavelengths to establish whether or

2006 The Authors. Journal compilation

2006 RAS

15-GHz variability of 9C sources

1557

not variability is likely to lead to serious incompleteness in the VSA

source subtractor sample.

Subarcsecond resolution radio mapping of the compact 9C

sources from one region of the complete 25-mJy sample of Paper 1

is presented in Bolton et al. (2006) (Paper 2 hereafter). This high-

resolution mapping showed that almost all (16

17) sources with

fitted peak frequencies above 0.5 GHz (in the observed frame) and

60 per cent of the flat-spectrum objects were unresolved at 3-mas

resolution.

In this paper, we present a study of the 15-GHz flux density

variability of a sample of 51 radio sources from the 9C survey,

taken from Paper 1. Of the 51 sources studied in this paper, 48 were

selected from the complete 25-mJy sample (sample A) from Paper 1;

no consideration was given to spectral shape but sources close to

each other on the sky were chosen to increase the efficiency of the

observations. Because sources were selected in a pseudo-random

manner, the sample studied here should be representative of the

whole 25-mJy sample.

A further three HFP sources from sample C of Paper 1 were

monitored because they happened to fit well into the observing pro-

gramme. The sources in sample C had originally been selected from

9C on the basis that they were likely to be GPS or HFP sources, but

they either were outside the area covered by the complete sam-

ple or had flux densities lower than the complete sample selection

limit.

The layout of this paper is as follows. The observations and data

reduction are described in Section 2; the method and results of

measuring the variability of each source are shown in Section 3.

In Section 4, the variability results are discussed with reference

to the radio spectral type of the sources and the variability time-

scales are considered. The implications of the results for the VSA

observations and sample completeness are discussed in Section 5.

A summary and discussion of the results from this work and from

Papers 1 and 2 is given in Section 6.

2 OBSERVATIONS AND DATA REDUCTION

The study of the variability of 9C sources was carried out with the

RT, with 51 sources observed frequently, but at irregular intervals,

between 2003 June and 2005 April. The data obtained, in combina-

tion with the original 9C survey data (1999 November–2001 June)

and the multifrequency follow-up of Paper 1 (2001 January–2002

May), give 15-GHz flux density measurements separated by time-

scales of a few days to between 1.5 and 5 yr.

Observations were made in batches; the targets in each observing

run consisted of six or seven sources close together on the sky, and an

appropriate phase calibrator. In each observing run, approximately

15 min was spent on each source and the resulting rms thermal noise

is between 1 and 2 mJy.

The RT has linearly polarized feeds, and so measures the com-

bined Stokes’ parameters I

Q. Flux calibration of the RT is carried

out using 3C286 (

3.5 Jy) and 3C48 (

1.7 Jy). In Paper 1,

we assumed that the day-to-day calibration of the RT flux-density

scale was consistent with 3 per cent rms (Pooley & Fender 1997).

However, as part of this variability study, we checked the accuracy

of this day-to-day calibration by examining the archive data for the

phase calibrator source 2005

403, which was observed on an al-

most daily basis during the epoch of the variability observations.

We looked at the fractional rms deviation in the RT estimates of the

flux density of 2005

403 over short periods of time (

25 d) since

the source is itself variable. These data indicate that the fractional

rms deviations over the epoch of this study are rather larger than

3 per cent, being typically between 4.5 and 6 per cent. Whilst this

may in part be due to real variations in the source flux density on

time-scales of a few days there are some indications that the tele-

scope tracking may be slightly less accurate than it was. In this study,

we take a conservative estimate of the accuracy of the day-to-day

calibration of 6 per cent rms.

The data initially showed correlated variations in the fluxes from

sources from the same observing runs. In particular, there were flux

density drops at times coincident with wet weather. However, poor

weather also resulted in increased noise on the phase calibrators so

this was used to flag bad observations. All data taken during runs

where the rms variation per 32-s sample on the phase calibrator

was greater than 2.5 times the thermal noise limit of 6 mJy were

discarded as this was found to be a reasonable choice for separating

‘good’ and ‘bad’ data.

3 TESTING FOR VARIABILITY

To test for variability a

test was carried out on the ‘good’ data for

each source assuming the null hypothesis (

) that the

th source flux

density measurement is taken from a distribution of constant mean

and standard deviation

(where

is the weighted mean of all ‘good’

data points and

is the uncertainty on the

th measurement, with

the error resulting from the six per cent flux calibration uncertainty

added in quadrature to the rms noise). The

test allows us to

quantify the probability,

, of obtaining the data if

is true. If

is greater than 1 per cent then the null hypothesis is not rejected and

the objects are classed as not variable (NV). Objects with

less

than 1 per cent are classed as variable (V).

The value of

is given by

∑

(

−

)

(1)

where

is the

th measured flux density.

The fractional variation,

≡

√

(

−

)

, was not used

to classify the objects, but the calculated values are shown in Table 1,

along with a summary of the data for each source.

For 34 of the 51 sources studied the

test gives a null result

at the 1 per cent level: there is no strong evidence of variability

above the uncertainties. Examples of these ‘non-variable’ sources

are shown in Fig. 1. The fractional variation of objects in the ‘non-

variable’ class ranges from 2.9 up to 14.7 per cent and the median

value is 6.1 per cent. 17 of the 51 sources are classed as variable,

with convincing evidence for flux density variations. These have

fractional variation between 8.4 and 70 per cent, and a median of

14 per cent. Examples of variable objects are shown in Fig. 2.

Four of the 34 ‘non-variable’ sources have borderline classifica-

tions with

between 0.01 and 0.1 (all of the other ‘non-variable’

sources have

0.1). The sources are J0013

2834 (which

has a fractional variability of 9.9 per cent); J0022

3250 (frac-

tional variability 14.7 per cent); J0927

2954 (fractional variability

7.9 per cent) and J0942

3309 (fractional variability 7.9 per cent).

The data for these four objects are shown in Fig. 3.

4 ANALYSIS OF RESULTS

4.1 Radio spectral type and variability

Sources have been classified by spectral shape by fitting a quadratic

function to the simultaneous observations of Paper 1. Four classes

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

1558

R. C. Bolton et al.

Table 1.

test results. The columns are: source name; observation period; the number of observations used; the value of the weighted mean (in mJy); the

fractional variation (per cent); the

value; the variability class; spectral type and fitted peak frequency (in GHz: see Paper 2). The three sources marked with

∗

are from sample C.

Source

Observation period

Variability

Spectral

Peak

name

(mJy)

class

type

(GHz)

J0010

2619

2001 November 23–2005 March 03

70.1

2.9

5.7

Steep

–

J0010

2650

2001 November 23–2005 March 03

35.4

7.2

27.4

Flat

–

J0010

2717

2001 November 23–2005 March 03

27.3

9.0

33.6

Flat

–

J0010

2838

2001 November 23–2005 March 03

39.9

20.3

201.4

Flat

–

J0010

2854

2001 November 23–2005 March 03

88.2

12.2

102.1

Flat

–

J0011

2928

2001 November 23–2005 March 03

49.8

4.7

12.7

Steep

–

J0012

3053

2001 November 23–2005 March 03

17.4

16.8

65.9

Flat

–

J0012

3353

2000 February 21–2005 March 03

136.1

21.2

208.7

HFP

36.9

J0013

2646

2000 February 21–2005 March 03

28.6

5.2

9.4

Steep

–

J0013

2834

2001 November 23–2005 March 03

27.3

9.9

30.5

Flat

–

J0014

2815

2001 November 23–2005 March 03

42.1

10.5

64.9

Flat

–

J0015

3052

2001 November 23–2005 March 05

38.9

4.2

8.3

Steep

–

J0015

3216

1999 October 18–2005 March 23

127

461.9

6.0

143.1

Steep

–

J0019

2956

2003 July 17–2005 March 05

43.2

5.6

13.0

GPS

1.2

J0019

3320

2003 July 17–2005 March 05

31.5

5.9

12.9

GPS

1.2

J0020

3152

2001 November 27–2005 March 05

24.8

7.4

14.2

GPS

4.9

J0021

3226

2001 November 27–2005 March 05

26.2

6.7

14.1

Steep

–

J0022

3250

2001 November 27–2005 March 05

13.3

14.7

36.4

Flat

–

J0023

2734

2001 November 27–2005 March 06

73.2

6.3

10.9

Steep

–

J0023

3114

2001 November 27–2005 March 06

33.2

3.5

2.6

Steep

–

J0024

2911

2000 February 22–2005 March 06

40.7

28.0

880.2

HFP

13.1

J0027

2830

2001 November 27–2005 March 06

24.4

6.4

8.5

GPS

1.1

J0028

2914

2001 November 27–2005 March 06

76.6

4.9

8.5

Steep

–

J0028

2954

2001 November 27–2005 March 06

26.8

9.4

16.6

GPS

4.4

J0927

2954

1999 November 27–2005 March 31

28.9

7.9

60.5

Flat

–

J0927

3034

2001 July 02–2005 March 31

42.4

5.3

22.4

Flat

–

J0928

2904

1999 November 27–2005 March 31

25.7

7.1

45.9

Steep

–

J0932

2837

2000 August 19–2005 March 31

57.9

4.2

27.6

GPS

2.1

J0933

2845

1999 November 27–2005 March 31

28.5

7.0

38.8

GPS

1.6

J0933

3254

2000 January 06–2005 March 31

21.3

13.0

107.4

Flat

–

J0934

3050

1999 November 27–2005 March 31

43.2

4.8

24.7

Steep

–

J0935

2917

1999 November 27–2005 March 31

44.7

8.7

104.0

HFP

5.3

J0936

3207

1999 November 27–2005 March 31

37.8

10.1

84.8

HFP

16.8

J0937

3206

2000 January 06–2005 March 31

55.3

9.1

74.2

Flat

–

J0939

2908

2002 February 02–2005 March 31

63.2

33.8

877.0

Flat

–

J0940

3015

2003 August 10–2005 March 31

62.0

5.3

33.8

Flat

–

J0942

3309

2002 February 09–2005 March 31

43.0

7.9

53.6

Flat

–

J0945

3003

∗

2000 January 07–2005 March 31

12.6

15.5

85.2

HFP

16.0

J1459

4442

∗

2001 January 17–2005 April 21

166.0

14.4

101.2

HFP

8.4

J1501

4537

∗

2001 January 17–2005 April 21

6.5

70.6

206.2

HFP

11.3

J1502

3753

2003 June 04–2005 April 21

39.3

4.2

9.3

Steep

–

J1502

3947

2003 June 04–2005 April 21

35.5

3.4

5.7

Steep

–

J1506

3730

2002 January 11–2005 April 21

535.5

8.4

77.1

Flat

–

J1516

3650

2003 June 04–2005 April 21

96.7

6.5

23.9

Flat

–

J1517

3936

2003 April 25–2005 April 21

31.4

10.7

52.3

Flat

–

J1519

3913

2002 January 12–2005 April 19

41.1

5.2

8.6

Steep

–

J1520

3843

2002 January 12–2005 April 19

42.0

6.8

15.1

Steep

–

J1526

3712

2002 January 12–2005 April 19

56.4

6.8

13.8

HFP

6.8

J1528

3738

2002 January 12–2005 April 19

86.2

6.2

14.8

Steep

–

J1528

3816

2002 January 12–2005 April 19

93.9

14.0

82.5

HFP

60.7

J1530

3758

2002 January 10–2005 April 19

59.6

6.2

11.4

GPS

3.4

are defined in Paper 2, these are: GPS objects with convex spectra

peaking between 0.5 and 5 GHz; HFP objects with convex spec-

tra peaking above 5 GHz; flat-spectrum objects with fitted spectral

index at 10 GHz,

0.5 (we take

∝

−

), and steep-spectrum

objects with

0.5. Full details of the classification scheme are

given in Paper 2. The assigned spectral type for each source, and,

for the GPS and HFP sources, the fitted peak frequency, are shown

in Table 1.

Table 2 shows the numbers of variable and non-variable sources in

each spectral class. No steep-spectrum or GPS sources are variable

whilst 50 per cent of the flat-spectrum sources and eight out of the

nine HFP objects have varied over the study period.

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

15-GHz variability of 9C sources

1559

52000

52500

53000

Date (MJD)

Flux density (mJy)

J0010+2619

J0932+2837

J1502+3947

Figure 1.

Examples of the variability study data for three stable sources.

Data were taken between 2000 February 22 and 2005 March 06.

52000

52500

53000

Date (MJD)

100

Flux density (mJy)

J0024+2911

J0939+2908

J1501+4337

Figure 2.

Examples of the variability study data for three highly variable

sources.

Peng at al. (2000) studied a complete, flux-limited sample of flat-

spectrum sources (with flux density at 5 GHz greater than 1 Jy and

0.5 between 2.7 and 5 GHz) and found that nine (69 per cent)

of the 13 sources had

7 per cent over a 5-yr period. The 32

flat-spectrum GPS and HFP sources from the flux-limited sample

presented here would probably all fall into Peng’s ‘flat-spectrum’

class. Of these, 23 (72 per cent) have

7 per cent over

∼

3yr,

in good agreement with the findings of Peng et al. in spite of the

much lower flux limit. This comparison should be treated with some

caution because the flux calibration uncertainty of the RT means that

variability at the 7 per cent level is not necessarily significant.

4.2 Variability time-scales

The sources were observed at irregular intervals over the 2-yr period

from 2003 June to 2005 April. There were also data available from

the original 9C survey and the multifrequency follow-up. The

analysis in Section 3 allowed us to look at flux variations on time-

scales greater than a few weeks over a period of between 1.5 and 5 yr.

The data are summarized in Fig. 4 in which the fractional variation

for each source is plotted against the time interval spanned by the

observations of that source. This shows that the proportion of sources

which are significantly variable increases as the time-span increases

and that this increase can be attributed entirely to variability of the

J0013+2834

Flux density (mJy)

J0022+3250

J0927+2954

52200 52400 52600 52800 53000 53200 53400

Date (MJD)

J0942+3309

Figure 3.

Examples of the variability study data for the four sources that

have 0.1

0.01.

Table 2.

Numbers of variable and non-variable sources in each spectral

class; the three objects from sample C, all of which are variable and fall into

the HFP class, have been included.

HFP

Flat

GPS

Steep

Variable

Not variable

flat-spectrum and HFP sources. Variability appears to occur at a

higher level and to be much more common in the HFP sources

than in the flat-spectrum ones (see also Table 2); however, given

the dependence on time-span, the results for the two groups are not

directly comparable as seven of the nine HFP sources have been

observed over time-scales of more than 4 yr compared with only

three of the 18 flat-spectrum sources.

To investigate the variability time-scales further, we have per-

formed a structure function analysis (e.g. Simonetti, Cordes &

Heeschen 1985) on the data for the individual variable sources.

Following the procedure adopted by Hughes, Aller & Aller (1992),

we plotted log

against log

, where

is the first-order structure

function and

is the time lag, for each source. Although the data

are sampled too sparsely and irregularly to enable us to comment

in detail on the variability of each source, the analysis does provide

some information on the variability time-scales up to about 2 yr.

(Although values of

were obtained for longer time lags these de-

pended largely on single flux density values obtained in the original

survey or the follow-up and were therefore unreliable statistically).

In the majority of cases, the plots of log

against log

have a

plateau at short time lags consistent with the measurement noise

and then show a steady increase in log

after a time lag of between

50 d and 1 yr. In all but two cases log

is still increasing after

∼

2 yr indicating the variability time-scales are therefore longer than

2 yr. The plots for the sources 0024

2911 and 0939

2908 (the

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

1560

R. C. Bolton et al.

Figure 4.

The fractional variation of each source plotted against the time interval spanned by the observations of that source. The symbols indicating the

different spectral classes are, steep spectrum (

), flat spectrum (

), GPS (filled triangles) and HFP (open crosses).

two sources with the most extreme values of

in Table 1) show

plateaus in log

after a time lag of

∼

1.5 yr indicating slightly shorter

variability time-scales in these two cases. There is no significant ev-

idence for flux variations in any of the sources at a level greater than

∼

6 per cent on time-scales less than

∼

50 d.

5 IMPLICATIONS FOR SAMPLE

COMPLETENESS

The variability data show that of the 48 sources selected from the

25-mJy flux-limited sample A, seven have flux densities that, for a

significant fraction of the time, are less than 25 mJy (see the data

for J1501

4337 in Fig. 2 for example). This suggests that there

are likely to be a number of variable sources fainter than 25 mJy

when the 9C observations were made which now have 15-GHz flux

densities greater than 25 mJy. This will presumably not affect the

statistics of the sample – the sample complete to 25 mJy in 9C should

have the same properties as another 25-mJy sample selected from a

blind survey at 15 GHz conducted at any other time. However, the

high levels of variability displayed by some HFP objects confirm that

care must be taken when using high-frequency surveys to identify

possible contaminating foreground sources for CMB experiments

at higher frequencies and later times.

Assuming that the number of sources increasing in flux density,

but missing from 9C, is the same as the number decreasing in flux

density and present in 9C, there are likely to be about 10 sources in

the VSA fields which are now brighter than 25 mJy at 15 GHz, but

are missing from the 9C catalogue.

The 15-GHz variability of some sources is as high as

0.71. It is therefore possible that a source could be missing from

9C because its flux density was just less than 25 mJy whilst the 9C

observations were being made, but has almost doubled at 15 GHz to

43 mJy by the time of the VSA observations; with a spectral index

−

0.75 between 15 and 34 GHz (the lowest spectral index seen in

the objects studied here), it would have a flux density of 79 mJy at

34 GHz. This is very close to the 80-mJy flux limit for VSA source

subtraction discussed in the introduction. However, only the most

extreme spectral index and the highest fractional variability com-

bined in one source would allow a source fainter than 25 mJy in

9C to be as bright as 80 mJy at 34 GHz a year or two later. Thus al-

though the completeness of 9C is affected by variability, it is unlikely

that point sources brighter than 80 mJy are contaminating the VSA

data.

6 DISCUSSION

As mentioned in Section 1, the ‘youth scenario’ predicts that young,

compact radio sources should be optically thick at low frequency;

they should therefore have radio spectra peaking at gigahertz or

few-gigahertz frequencies, with younger sources peaking at higher

frequency than older, less-compact objects. For samples of GPS

and compact steep-spectrum (CSS) sources, Snellen et al. (2000)

find correlations between the peak frequency, peak flux and angular

size which provide strong evidence for this picture. Additionally,

there are examples of GPS objects with ages inferred from mea-

sured expansion speeds. These measurements indicate that they are

genuinely young objects and lend considerable support to the youth

scenario. For example, 0710

439 (

0.518) is a

∼

1000-yr-old

source with an observed spectrum peaking at

∼

2 GHz and a physical

size of

∼

100 pc, and 2352

495 (

0.238) has an observed spec-

tral peak at

∼

1 GHz, a size of 100 pc and an inferred age of

∼

2000 yr

(Owsianik, Conway & Polatidis 1999, and references therein).

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

15-GHz variability of 9C sources

1561

However, despite the evidence from the continuity in proper-

ties between GPS and CSS sources and between CSS sources and

large-scale radio galaxies that these objects form an evolutionary se-

quence, there is as yet little convincing evidence that the HFP objects

represent the earliest stage in this evolution. In particular, concerns

have been raised about the contamination of GPS and HFP samples

by beamed objects. Tinti and coworkers (Tinti et al. 2003, 2005) find

that their sample of HFP objects (with peak frequencies above a few

GHz) is largely contaminated by blazar objects, and, especially, that

the majority of HFP sources associated with quasars are highly vari-

able, probably beamed objects that happen to have emission dom-

inated by a flaring, strongly self-absorbed component. They report

a positive correlation between the peak frequency and the level of

variability, in agreement with the results reported here. Torniainen

et al. (2005) have also studied the variability on time-scales of about

20 yr of GPS and HFP sources associated with quasars. They find

that all but five of the 35 inverted-spectrum sources they observed

are strongly variable. The radio spectra of these sources change with

time: in some the spectra remain convex whereas in others the spec-

tra may at times be flat with an inverted spectrum only during flares.

Torniainen et al. conclude that genuine, unbeamed GPS and HFP

quasars are extremely rare and that, in order to find them, long-term

flux-density monitoring is essential.

We now review the findings of all the 9C follow-up work and

consider the results with reference to the nature of the 9C GPS and

HFP objects in particular. The NASA/IPAC Extragalactic Database

(NED) archives have been searched for all objects in Paper 1

and

these data have been used to help improve our understanding of each

source.

6.1 Variability in flat-spectrum, HFP and GPS objects

18 of the 51 objects studied here have flat radio spectra, and one

half of these are variable. It is generally accepted that flat-spectrum

objects owe their spectral shapes to relativistic beaming of different

components in a jet closely aligned with the line of sight (LOS), and

so variability in flat-spectrum objects is common.

The most striking result of this variability study is the differ-

ence in the levels of variability displayed by objects with observed

spectral peaks above and below 5 GHz. Excluding the flat-spectrum

objects, there are 24 GPS and steep-spectrum objects that peak be-

low 5 GHz, if they peak at all, none of which are variable. However,

eight of the nine objects that peak above 5 GHz are variable. The

one source peaking above 5 GHz and not found to be variable in this

study is J1526

3712; however, this shows evidence for long-term

variability (especially at 5 GHz) in the NED archive data, shown in

Fig. 5.

Tornikoski et al. (2001), in their long-term study of variability in

GPS and HFP objects, find that objects exhibit more pronounced

and higher frequency spectral peaks when they are in a flaring state

than when they are in a quiescent phase. It is therefore possible that

an object in a quiescent, non-variable phase will display a gigahertz-

peaked or perhaps a flat spectrum and when it enters an active stage

its spectrum peaks at high frequency. Although variability provides

evidence for beaming the converse is not true – the absence of

variability in an object does not imply that no beaming is occurring.

Thus, some of the GPS sources in the sample studied here, which

Plots of the spectra recovered from NED are available from the ftp site

linked from http://www.mrao.cam.ac.uk/surveys/9C/

110

Frequency (GHz)

100

Flux density (mJy)

RT variability data

data from paper 1

NED data

Figure 5.

Data from the NED archive, Paper 1 and this paper for the source

J1526

3712.

Table 3.

Numbers of sources in each spectral class for samples A and B

from Paper 1. The numbers in brackets are of sources resolved with the

VLA and greater than 2 arcsec in angular extent.

HFP

GPS

Flat

Steep

Total

Sample A

15 (1)

16 (0)

40 (7)

53 (31)

124

Sample B

13 (0)

11 (5)

23 (7)

23 (10)

are currently showing no evidence for variability, may become HFP

objects during times of activity.

6.2 The nature of the GPS and HFP objects

In this section, we examine the nature of the GPS and HFP objects.

We use the data from the complete 9C samples presented in Paper 1,

the angular size distribution of the sources in the 9C subsample

discussed in Paper 2, and the variability data presented here along

with the data from NED which provides information on variability

over longer time-scales. We omit from the discussion any sources

from the incomplete sample C.

The results of classifying all the sources by their spectral shape

are shown in Table 3 for the two flux-limited samples from Paper 1

(sample A, complete to 25 mJy in 9C containing 124 objects and

sample B, complete to 60 mJy and containing 70 objects). The angu-

lar sizes, or limits on the angular sizes, on arcsecond scales of all the

sources in both samples have been measured with the Very Large

Array (VLA) (see Paper 1); the numbers of sources with angular

sizes larger than 2 arcsec in each spectral class are given in brackets

in Table 3.

If we ignore the flat-spectrum objects, which we assume are

beamed, it can be seen from Table 3 that there are roughly equal

numbers of HFP and GPS sources in both samples, and that to-

gether they make up 37 per cent of sample A and 50 per cent of

sample B. Snellen et al. (2000) suggest that the GPS phase in the

life of a radio source lasts until it reaches a size of 1 kpc or so.

In their model for this phase, the linear size

of a source grows

in time

∝

and the peak frequency scales with time as

peak

∝

−

. With these assumptions, the measured sizes and ages

of the GPS sources 0710

439 and 2352

495 (Owsianik et al.

1999) mentioned earlier indicate that sources should be

∼

–10

old when they reach 1 kpc in size; this is about one-thousandth of

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

1562

R. C. Bolton et al.

the lifetime of a powerful radio source which is believed to be typ-

ically

∼

–10

yr. Also, if HFP sources expand to become GPS

sources, the HFP phase must last for

yr. According to this

analysis, the proportion of HFP and GPS sources in samples A and

B seems large given their lifetimes relative to those of extended

sources; additionally, from their lifetimes relative to each other, we

would not expect to get equal numbers of HFP and GPS sources.

These results therefore suggest that a number of the HFP and GPS

objects in the samples are not part of an evolutionary sequence from

HFP to GPS to CSS and eventually extended steep-spectrum ob-

jects. (More precise predictions of the relative numbers of sources

in each phase for a flux-limited sample at a given frequency requires

detailed modelling of the evolution of an individual source, the fre-

quency at which it becomes optically thick and the cosmological

evolution of the radio luminosity function).

Further support for this conclusion comes from the angu-

lar size distribution of the subsample of 61 sources studied in

Paper 2 with subarcsecond resolution using the Multi-Element

Radio-Linked Interferometer Network (MERLIN) and the Very

Long Baseline Array (VLBA). This subsample is itself complete to

25 mJy in 9C and consists of all the sources in sample A from the

9C field; there are seven GPS sources and nine HFP sources in

this subsample. Using lifetime arguments, many of the GPS sources

would be expected to be of the order of 1 kpc in size – or at least

100 sin

mas in projected angular extent (where

is the angle be-

tween the jet and the LOS). The majority of such objects would

be resolved with the VLBA. However, all the GPS sources in the

sample in Paper 2 were unresolved even at 3-mas resolution (whilst

all the steep-spectrum objects and one-third of the flat-spectrum ob-

jects were resolved with either the VLA, MERLIN or the VLBA).

In order for these objects to remain unresolved with the VLBA,

they must have projected linear sizes of less than 50 pc (

∼

160 light-

years) – half the projected sizes of 0710

439 and 2352

495. This

requires that either they are younger than

∼

500–1000 yr (by com-

parison with 0710

439 and 2352

495) or they are aligned closely

with the LOS. It is statistically highly improbable that all the GPS

objects in the sample in Paper 2 are aligned close to the LOS – unless,

of course, they are beamed, which would simultaneously increase

the chances of seeing such objects and remove the need to inter-

pret their spectral shapes as signatures of youth. In the youth scen-

ario, the sizes of the HFP sources are expected to be considerably

smaller than those of the GPS sources. For example, extrapolat-

ing the data given by Snellen et al. (2000) sources with spectral

turnovers at

5 GHz are expected to be typically less than

∼

10 mas

in extent. Nevertheless, if these sources did represent exclusively

a population of very young sources, it is again surprising that the

nine HFP sources in the sample in Paper 2 are also all unresolved

at 3-mas resolution.

There is additional evidence pointing to the fact that most of the

GPS and HFP sources are not young self-absorbed objects. We look

first at radio variability. In addition to the 15-GHz variability ob-

servations presented here, variability at 1.4 GHz, over a time-scale

of about 8 yr, has been investigated comparing the data from the

NRAO (National Radio Astronomy Observatory) VLA Sky Survey

(NVSS; Condon et al. 1998) with those given in Paper 1. Variability

at 4.8 GHz, over a time-scale of about 15 yr, has also been examined

using data from the GB6 catalogue of radio sources (Gregory et al.

1996). The results are presented in Table 4 which summarizes the

properties of all the GPS and HFP sources in samples A and B. At

both 1.4 and 4.8 GHz, any source for which the difference between

the flux densities at the two epochs is more than three times their

combined errors (

) have been classed as variable (V); if the dif-

ference is between 2

and 3

, the source is classed as possibly

variable (PV) and if it is less than 2

it is classed as non-variable

(NV). The relatively large beamsize of the GB6 survey (

≈

3.6

3.4 arcmin

) means that confusion may be a problem; consequently

any source for which the GB6 flux appears significantly greater than

the flux density measured in Paper 1 may not be a true variable –

where this is the case the class is marked with an asterisk in Table 4.

(There are six sources with single epoch data only at either 1.4 or

4.8 GHz: three of these were not observed at 1.4 GHz in Paper 1 and

three were too faint to appear in the GB6 catalogue).

It can be seen that in spite of the failure to detect variability in the

GPS objects at 15 GHz, a significant fraction (10 out of 24) of them

show evidence for variability at 1.4 or 4.8 GHz. It may be that GPS

objects are more variable at 1.4–4.8 GHz than at 15 GHz (although

this would counter the expected trend that objects are generally more

variable at higher radio frequency than at lower frequency, down to

a gigahertz or so); it is, however, plausible that 3 yr is too short a

time interval over which to detect GPS variability – we see them as

variable at 1.4 or 4.8 GHz simply because of the longer time-scales

provided by the NVSS and GB6 surveys.

As discussed in Section 6.1, in the variability study presented

here, five of the six HFP sources from the 9C complete sample are

found to be variable at 15 GHz and the sixth shows longer term

variability in the NED archive data. In addition, seven other HFP

sources show variability at 1.4 or 4.8 GHz. Thus at least 13 of the

18 HFP sources are variable.

A source which has a peaked spectrum is likely to be strongly

beamed rather than intrinsically young if it has any of the following

properties: radio flux density variability over time-scales of a few

years, association with a quasar, a flat or steep (rather than a rising)

spectrum at frequencies below 1 GHz and structure on arcsecond

scales. For each GPS and HFP source in samples A and B, NED

has been used to provide lower frequency data and some optical

information, and additional radio structural information has been

obtained where possible from FIRST (Faint Images of the Radio

Sky at Twenty centimeters; Becker, White & Helfand 1995). These

data are summarized in Table 4. If a source has any of the proper-

ties listed above which are consistent with strong beaming rather

than the youth scenario, it is labelled Y in Column 9 of Table 4 –

Y? indicates that the source has possible evidence for beaming. Tak-

ing all these indicators into account the data show that 16 (possibly

21) of the 24 GPS sources and 15 (possibly 17) of the 18 HFP

sources have properties which are consistent with strong beaming.

Thus the evidence presented here indicates that the majority of the

GPS and HFP objects present in samples selected at 15 GHz are ob-

jects dominated by emission from a strongly beamed self-absorbed

component and are not necessarily intrinsically young objects.

7 CONCLUSIONS

A study of the 15-GHz variability of 51 9C sources over a 3-yr

period has found no evidence for any variability (above

∼

6 per cent

flux calibration uncertainties) in steep-spectrum objects. Half of the

18 flat-spectrum objects were found to be variable. The HFP objects

have been found to be highly variable – the majority (eight out of

nine) exhibit variability at 15 GHz over a 3-yr period; archive data

show that the ninth source is variable over slightly longer time-

scales. None of the GPS objects were seen to vary at 15 GHz over

3 yr, but several show evidence of changing flux density over longer

(

∼

8–15 yr) time-scales in NED.

The majority of the GPS and HFP sources in the 9C survey are

not resolved with resolutions of 3 mas with the VLBA. In particular,

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

15-GHz variability of 9C sources

1563

Table 4.

Properties of the GPS and HFP sources in samples A and B. Column 1: source name. Column 2: sample. Column 3: fitted peak frequency (see

Paper 2). Columns 4–6: variability classes at 1.4, 4.8 and 15 GHz as defined in the text –

∗

indicates that there may be a problem with confusion. Column 7:

optical type

:G–g

alaxy; G? – point-like object with

−

1.6 (see Paper 1); Q – confirmed quasar; Q? – point-like object with

−

1.6 (see Paper 1).

Column 8: redshift. Column 9: evidence for beaming (properties as defined in the text)

: Y – yes; Y? – possibly. Column 10: notes.

Source

Sample

Peak Variability Variability Variability Opt. Redshift Beamed Notes

name

(GHz) (1.4 GHz)

(4.8 GHz)

(15 GHz)

type

GPS sources

J0019

2956

1.2

0.0244

J0019

3320

1.2

–

J0020

3152

4.9

–

J0027

2830

1.1

–

J0028

2954

4.4

–

J0029

3244

1.0

–

J0032

2758

4.0

–

J0925

3159

0.5

–

Steep spectrum below 1.4 GH

z – 5 arcsec in extent

J0925

3127

3.3

∗

–

Flat spectrum below 1 GHz?

J0932

2837

2.1

0.3033

J0933

2845

1.6

3.4214

J0936

2624

4.2

–

Steep spectrum below 1 GHz; core-jet 45 arcsec in extent

J0945

3534

4.0

–

1.2721

Flat spectrum below 1 GHz; 16 arcsec in extent

J0949

2920

2.1

–

J0952

3512

3.9

–

1.8764

28-arcsec core-dominated double

J0953

3225

3.3

∗

–

1.57

Flat spectrum below 1 GHz; core-jet(?) 30 arcsec in extent

J0958

2948

2.2

–

2.75

Core-jet(?) 14 arcsec in extent

J1530

3758

A/B

3.4

0.1519

J1531

4356

2.7

–

0.4520

J1547

4208

A/B

2.7

–

2.7442

J1550

4536

3.5

∗

–G–

J1553

4039

2.0

–

2.7526

Steep spectrum below 1.4 GHz

J1554

4348

3.7

–

1.4393

J1556

4259

A/B

4.4

–

1.7462

HFP sources

J0003

2740

A/B

5.5

–

J0003

3010

A/B

10.7

–

J0012

3353

A/B

36.9

–

J0024

2911

A/B

13.1

–

J0919

3324

10.0

–

J0931

2750

8.5

–

J0935

2917

5.3

–

J0936

3207

16.8

1.151

J0955

3335

5.9

–

2.500

J1506

4359

A/B

11.1

–

0.9198

J1506

4239

A/B

14.3

–

0.587

Flat spectrum below 1 GHz; core-jet (?) 20 arcsec in extent

J1510

4138

5.8

–

J1521

4336

A/B

6.1

–

2.180

J1526

3712

A/B

6.8

–

Steep spectrum below 1 GHz; 3.5 arcsec in extent

J1526

4201

A/B

7.9

–

J1528

3816

A/B

60.7

∗

VG?–

J1540

4138

7.8

–

2.5171

J1554

4350

10.7

–

1.4393

none of the 11 HFP objects observed with the VLBA has been

resolved. The GPS objects seen in the 9C survey have been found to

be either very compact objects (less than 3 mas in projected angular

extent) or core-dominated objects with extended envelopes with flat

or steep spectra at low frequency (not true GPS candidates).

If the flat-spectrum objects are excluded, the GPS and HFP objects

make up around 40–50 per cent of the sources in our 15-GHz flux-

limited samples. If they are all compact young radio sources, they are

much more common than simple lifetime arguments would suggest

they should be.

Thus, individually and as a whole, the high-resolution observa-

tions, variability properties and lifetime arguments suggest that the

GPS and HFP objects present in samples selected at 15 GHz are ob-

jects dominated by emission from a strongly beamed self-absorbed

component and are not necessarily intrinsically young objects. This

does not oppose the youth scenario but suggests that, although it

is possible that genuinely young HFP and GPS objects might be

found in high-frequency radio surveys, such surveys are not a par-

ticularly good starting point for searching for young objects because

the contamination from beamed objects is so high.

ACKNOWLEDGMENTS

We thank the referee, Merja Tornikoski, for helpful comments.

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564

1564

R. C. Bolton et al.

This work has made use of the NED, which is operated by the JPL,

California Institute of Technology, under contract with the National

Aeronautics and Space Administration.

The NRAO is a facility of the National Science Foundation oper-

ated under cooperative agreement by Associated Universities, Inc.

REFERENCES

Becker R. H., White R. L., Helfand D. J., 1995, AJ, 450, 559

Bolton R. C. et al., 2004, MNRAS, 354, 485 (Paper 1)

Bolton R. C., Chandler C. J., Cotter G., Pearson T. J., Pooley G. G., Readhead

A. C. S., Riley J. M., Waldram E. M., 2006, MNRAS, 367, 323 (Paper 2)

Condon J. J., Cotton W. D., Greisen E. W., Yin Q. F., Perley R. A., Taylor

G. B., Broderick J. J., 1998, AJ, 115, 1693

Gregory P. C., Scott W. K., Douglas K., Condon J. J., 1996, ApJS, 103, 427

Hughes P. A., Aller H. D., Aller M. F., 1992, ApJ, 396, 469

Jones M. E., 1991, in Cornwell T. J., Perley R., eds, ASP Conf. Ser. Vol. 19,

Radio Interferometry – Theory, Techniques and Applications. Astron.

Soc. Pac., San Fransisco, p. 395

O’Dea C. P., 1998, PASP, 110, 493

Owsianik I., Conway J. E., Polatidis A. G., 1999, New Astron. Rev., 43, 669

Peng B., Kraus A., Krichbaum T. P., Witzel A., 2000, A&AS, 145, 1

Pooley G. G., Fender R. P., 1997, MNRAS, 292, 925

Simonetti J. H., Cordes J. M., Heeschen D. S., 1985, ApJ, 296, 46

Snellen I. A. G., Schilizzi R. T., Miley G. K., de Bruyn A. G., Bremer

M. N., R ̈

ottgering H. J. A., 2000, MNRAS, 319, 445

Taylor A. C. et al., 2003, MNRAS, 341, 1066

Tinti S., Dallacasa D., Stanghellini C., Celotti A., Publ. Astron. Soc. Aust.,

2003, 20, 110

Tinti S., Dallacasa D., De Zotti G., Celotti A., Stanghellini C., 2005, A&A,

432, 31

Torniainen I., Tornikoski M., Ter ̈

asranta H., Aller M. F., Aller H. D., 2005,

A&A, 435, 839

Tornikoski M., Jussila I., Johansson P., Lainela M., Valtaoja E., 2001, AJ,

121, 1306

Waldram E. M., Pooley G. G., Grainge K. J. B., Jones M. E., Saunders

R. D. E., Scott P. F., Taylor A. C., 2003, MNRAS, 342, 915

This paper has been typeset from a T

X/L

X file prepared by the author.

2006 The Authors. Journal compilation

2006 RAS, MNRAS

370,

1556–1564