Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.atnf.csiro.au/people/Matthew.Whiting/thesis/ch4.ps.gz Дата изменения: Wed Mar 29 06:51:53 2006 Дата индексирования: Sun Dec 23 05:09:09 2007 Кодировка: Поисковые слова: m 63

Chapter 4
Sources and Data
This chapter serves two purposes. Firstly, we introduce the Parkes Half-
Jansky Flat-spectrum Sample, a catalogue comprising 323 radio-loud at-
spectrum objects rst published by Drinkwater et al. (1997). This catalogue
forms the basis for the modelling and analysis that will be done in this thesis.
Secondly, we present the optical photometry that will be used in the actual
modelling. This data has already been published (Francis, Whiting, and
Webster 2000), but it is presented here also, together with additions and
alterations specic to the modelling performed later in the thesis.
Additionally, a note for completeness: whenever names of sources from
the sample are presented in this thesis, they are specied in the standard
form HHMMDDD, where \HHMM" is the right ascension of the source in
hours and minutes, and \DDD" is the declination to the nearest tenth of a
degree. Both right ascension and declination are in B1950 coordinates. In
the literature, sources from the Parkes Catalogue (as all these sources are)
have the prex \PKS". For the purposes of brevity, however, this has been
omitted from all the source names for the remainder of this thesis.
4.1 The Sample
During the 1960s and 1970s, the 64m radio telescope near Parkes in New
South Wales surveyed the entire sky (as visible from the telescope) at 2.7
GHz, generating a catalogue that numbered nearly 11,000 sources. Many of
these sources are AGN or quasars, and so this catalogue represents a huge
resource of possible candidates for radio-selected quasars.

50 Sources and Data
This was taken advantage of in the creation of the Parkes Half-Jansky
Flat-Spectrum Sample (hereafter PHFS) (Drinkwater et al. 1997). This is a
sample of bright radio sources, taken from the machine-readable version of
the Parkes Catalogue (PKSCAT90, Wright and Otrupcek 1990), that have
a at radio spectrum based on the uxes from the catalogue. The criteria
for selection are:
Observed: Fluxes at both 5.0 GHz and 2.7 GHz exist in the Parkes
Catalogue (PKSCAT90).
Radio loud: S 2:7GHz > 0:5Jy
Flat spectrum: Radio spectral index  2:7=5:0 > 0:5, where S  /   .
Position: Galactic latitude: jbj > 20 ф
Position: Declination (B1950.0): 45 ф < ф < +10 ф
Note that the spectral index is derived from the 2.7 and 5 GHz uxes
from the catalogue. However, these uxes were taken at dierent times {
sometimes up to more than a year apart { and so, if the source in question
had varied in ux between these observations, this spectral index will not be
the true radio continuum slope of the quasar. More recent contemporaneous
observations 1 using the Australia Telescope Compact Array, have shown
that  10% of the sources are incorrectly included in the PHFS (a greater
number than predicted by Drinkwater et al. (1997)), as their true radio
spectral index is steeper than 0:5. Additionally, a similar number will
have been excluded for having an apparently steep index.
The Galactic latitude requirement was included to exclude those sources
close to the Galactic plane, where increased optical extinction by dust, as
well as the increased density of sources, would pose a problem in the identi-
cation and study of the optical counterparts of the sources. The declination
limits were imposed purely to limit the size of the sample. These criteria
resulted in a total of 323 sources.
The resulting sample of radio sources is quite heterogeneous. The sources
included range from low redshift radio galaxies (including narrow-line radio
galaxies and steep (radio) spectrum galaxies), through the spectrum of tra-
ditional quasars { both optically bright and optically faint { with a number
1 These observations are currently ongoing, and have involved many people at the Uni-
versity of Melbourne.

4.1 The Sample 51
Figure 4.1: Distribution of redshifts for sources from the PHFS.
of BL Lac objects, up to some very high redshift quasars. The redshift dis-
tribution for those sources with measured redshift (a total of 294 sources
out of 323) is shown in Fig. 4.1. The 29 sources that do not have a mea-
sured redshift generally fall into two distinct classes: BL Lac objects, with
no measurable emission lines from which to deduce the redshift; or optically
faint sources that are too faint to get a suitable spectrum from which to
measure a redshift.
While the radio properties of the objects are all very similar (by virtue
of the selection process), the optical properties vary considerably. The most
obvious eect, rst pointed out by Webster et al. (1995), is that there is
a very large dispersion in their B K colours (that is, the ratio of ux in
the K band (infrared: 2:2m) to the B band (blue: 0:4m)). This range,
2 < B K < 8, is much greater than that seen in typical optically-selected
samples of quasars (for instance, the Large Bright Quasar Survey, or LBQS,
has a range of 2 . B K . 4).
Webster et al. interpreted this range of optical-NIR colours as being due
to the presence of large amounts of dust, that obscures the quasar at optical
wavelengths. This had the implication that, if similar amounts of dust were
present in radio-quiet quasars (as suggested by X-ray observations), then
optical surveys of bright quasars may be missing up to 80% of quasars for a

52 Sources and Data
given intrinsic magnitude.
Serjeant and Rawlings (1996), however, suggested the alternative in-
terpretation that the red optical colours are due to the presence of a red,
optical synchrotron component. This component would likely be an exten-
sion of the beamed radio emission. This idea will be tested by the modelling
in Chapter 5.
Finally, we note that a further alternative explanation was oered by
Benn et al. (1998). They investigated a sample of radio-selected quasars
(the B3-VLA quasar sample, Vigotti et al. (1997)), and found that many
of them exhibited an excess of light in the K band. They argued that this
light mostly came from star-light from the host galaxy. The B3 sample
is dominated by steep radio-spectrum sources, and, according to unied
schemes, the central AGN source should be subject to more obscuration than
for at-spectrum sources, making it easier to detect the galactic emission.
This explanation, however, was tested by Masci et al. (1998), and rejected
as an explanation for the spread in optical colours of the PHFS sources.
4.2 Observational data
To attempt to solve the puzzle of the red optical-NIR colours, it was decided
to make observations of the PHFS quasars at seven wave-bands, covering
the optical and NIR wavelength ranges: B (  0:44m), V (  0:55m),
R (  0:7m), I (  0:88m), J (  1:24m), H (  1:65m), and K
(  2:19m). Having this many data points in the optical and NIR range
of the spectrum makes it far easier to accurately model the spectral energy
distributions (SEDs) of the Parkes quasars, and hopefully determine the
nature of the reddening mechanism. The results of these observations have
been published by Francis et al. (2000). However the method and results of
observations are discussed here, as well as the choice of data that is used in
this thesis.
4.3 The new observations
Over the course of 26 nights in 1997, a total of 157 sources from the PHFS
were observed, using the Australian National University's 1 metre 2 and 2.3
2 Or, as it is traditionally known, the \40 inch".

4.3 The new observations 53
metre telescopes. The optical observations were made using either the 1m
or the imager on the 2.3m telescope, while the NIR observations were made
with the CASPIR 256  256 InSb array camera (McGregor et al. 1994) on
the 2.3m telescope. The methods of the observations and data reduction
are explained in detail in Francis et al. (2000), so they will not be repeated
here.
To minimise the eects of variability, the optical and NIR observations
for each object were made as close as possible in time to each other (the
largest spread in time for a single object was six nights, and most sources
had a spread much shorter than this).
It should be noted that variability has been seen on timescales less than
this. So-called \intra-day variability" (IDV) has been commonly seen at
radio frequencies in many radio loud quasars, including many of the objects
observed here (see, for example Kedziora-Chudczer (1998), or the review
of IDV properties by Wagner and Witzel (1995)). Such variability has also
been seen at optical and NIR wavelengths (Heidt and Wagner 1996; Romero,
Cellone, and Combi 1999), mainly in sources such as BL Lac objects and
at-spectrum radio quasars (i.e. the sort of object found in the PHFS).
While the radio variations could be due to scintillation of an intrinsically
small source by the interstellar medium, this cannot explain the variability
at infrared and optical frequencies. Instead, these variations are thought
to be due to relativistic shocks propagating down jets, or the interactions
of these shocks with features in the jets. This model is preferred to one
where the variations are due to ares or hot spots on the accretion disk,
primarily because of the low incidence of optical intra-day variability seen
in radio-quiet quasars (which either have very weak jets or are without them)
(Jang and Miller 1997; Romero et al. 1999). However, see Chapter 8 for a
result that challenges this. Amplitudes of this variation can be up to 0.1
magnitudes over the course of one night.
For the purposes of modelling and plotting the data, the magnitudes need
to be converted into uxes. The assumed uxes for objects of zero magni-
tude are listed in Table 4.1. In the optical, the lter sets used approximate
the Johnson & Cousins system (the calibration was done using Graham stan-
dards, which also approximate this system), and the zero magnitude uxes
are taken from Bessell, Castelli, and Plez (1998). The NIR observations were
made using the CASPIR lter set, calibrated with IRIS standards, and the

54 Sources and Data
Band Mean  Zero magnitude ux
(m) (F  ; W m 2 m 1 )
B 0.440 6:32  10 8
V 0.550 3:63  10 8
R 0.700 2:18  10 8
I 0.880 1:13  10 8
J 1.239 3:11  10 9
H 1.649 1:15  10 9
K 2.192 4:10  10 10
Table 4.1: Assumed uxes of a star with zero magnitude, for each of the wave-bands
used.
zero magnitude uxes used are those calculated by P. McGregor 3 assuming
that the star Vega is well represented in the NIR by a black body of temper-
ature 11200 K and normalisation F  (555nm) = 3:44  10 12 W cm 2 m 1
(Bersanelli, Bouchet, and Falomo 1991). The observations were made with
the CASPIR K n lter, but were calibrated using the K magnitudes of the
IRIS system without a colour correction term applied. They have thus been
normalised using a K-band zero point (Table 4.1).
4.4 Photometry
The photometry conrms the result presented in Webster et al. (1995),
namely that there is a large spread in B K colours in the PHFS quasars.
This distribution is shown in Fig. 4.2, along with the distribution of colours
for a small representative sample of quasars from the LBQS. As can be
clearly seen, the PHFS distribution is much broader and contains many
quite red objects. Also note that the lower envelope of the PHFS distri-
bution is approximately the same as that of the LBQS quasars { i.e. there
are not really any sources in the PHFS that are signicantly bluer than the
optically selected quasars. (The one outlying source in Fig. 4.2 is 0232 042,
which has a very blue SED: B K = 1:45.)
In Fig. 4.3 is plotted a histogram of the absolute magnitudes of the
PHFS sources with simultaneous photometry. The absolute magnitude is
3 CASPIR manual, Mount Stromlo & Siding Spring Observatories, Australian National
University. http://msowww.anu.edu.au/observing/teldocs/2.3m/CASPIR/

4.4 Photometry 55
Figure 4.2: Distribution of B K colours for the PHFS sources with quasi-simultaneous
photometry. The top panel shows a small representative sample of sources from the LBQS
{ the photometry is taken from Francis (1996) and Francis et al. (2000).
calculated by
M V = m V 5 log 10 (d L ) + 2:5(1 + ) log 10 (1 + z) 25 (4.1)
where m V is the apparent V magnitude, dL is the luminosity distance, in
Mpc, to redshift z, and  is the spectral index (F  /   ). In calculating
M V , we used the best-t spectral index to the optical photometry (that is,
B, V , R, and I magnitudes). The term K(z) = 2:5(1 + ) log 10 (1 + z) is
the K-correction term (Peterson 1997).
One can use the absolute magnitude to discriminate between sources of
quasar and Seyfert luminosity. The dividing line is traditionally given by
M V = 21:5 + 5 log 10 h (4.2)
where H 0 = 100h km s 1 Mpc 1 . As can be seen in Fig. 4.3, there are
16 sources in this sub-sample of the PHFS that are of Seyfert luminosity.
These sources are either low-redshift galaxies, or optically faint sources (that
are not necessarily at a low redshift). We note that the distribution from
quasar to Seyfert luminosities is continuous, demonstrating that the division

56 Sources and Data
Figure 4.3: The distribution of absolute magnitudes for sources in the PHFS with
quasi-simultaneous photometry. In calculating MV , we used H0 = 75 km s 1 Mpc 1 and
q0 = 0:5, and the spectral index  was calculated from the optical photometry. The open
histogram shows those sources of quasar luminosity, while the hatched histogram shows
those with Seyfert luminosity (see text for denition).
between the two classes is largely an arbitrary one (although we also note
that even the low-redshift sources here are not true Seyferts, due to their
strong radio emission { Seyfert galaxies are not generally radio-loud objects).
4.5 Errors
When the observational data was published (Francis et al. 2000), the quoted
error for each object was the sum (in quadrature) of random errors (deter-
mined from the root-mean-squared pixel-to-pixel variation in the sky re-
gions) and an assumed 5% systematic error in the photometric zero points
(determined from the scatter in the zero points from dierent standard star
measurements over the course of a night). The rms scatter in the zero points
was typically  3%, so a somewhat conservative value of 5% was adopted
as the systematic value.
However, for the purposes of the modelling to be done in this thesis,
there are a number of potential systematic errors not considered in Francis
et al. (2000) that may be important. These include:
¬ The optical zero magnitude uxes (Bessell et al. 1998) were derived

4.5 Errors 57
for an A0 star, and so will be slightly incorrect for a quasar spec-
trum. This has the eect of introducing small colour terms into the
photometry, the size of which will depend on the spectral index of the
object being observed. A similar eect will of course be present in
the NIR. Bersanelli et al. (1991) found that spectral shape dierences
could produce systematic errors of at least a few percent in the NIR
ux.
The zero point uxes in the optical and the NIR are taken from dif-
ferent references (see above). These dierent zero points may not be
exactly equivalent, which will produce small osets between the optical
and NIR parts of the SED.
® If the sky conditions at Siding Spring Observatory were not completely
photometric for all observations for a given source (particularly if the
transparency changed between dierent bands), then the measured
photometry will have small band-to-band errors present.
# The observations for each source were taken quasi-simultaneously (in-
dicating all observations were made within at most a six-day period),
to minimise the eects of variability. However, as discussed above, a
number of the sources in the PHFS have been found to exhibit intra-
day variability in the optical (Heidt and Wagner 1996; Romero et al.
1999), and a larger number no doubt have similar properties to these.
Therefore, variability on timescales of the order of those separating
our observations is likely for some of the sources. Such variability can
be up to 0.1 mag over the period of a night.
A possible example of a source that has undergone variability between
the two sets of observations is the BL Lac 0537 441. This source has
been shown to be quite variable on very short timescales (amplitudes
of  0:3 mag on timescales of 2 days were seen by Romero et al.
(2000)). Its SED is shown in Fig. 4.4, where it appears as if the
NIR emission (which was observed after the optical) has been boosted
{ note that the slope of the NIR is approximately that of the optical
(excluding the B band point), but with a dierent normalisation. This
dierence corresponds to a variation of  0:5 mag over the course of
4{5 nights. For further examples (although not as extreme) of such

58 Sources and Data
Figure 4.4: The SED of the BL Lac object 0537 441. Note that the three NIR points
appear to be displaced upwards from the extrapolation of the optical points. Error bars
are 1.
\micro-variability", see Chapter 8.
° Finally, the presence of strong emission lines in the quasars' spectra
could boost the ux of a band above the level of the continuum. The
amount of this eect is investigated for a small number of objects in
Section 5.9.
To account for all these extra eects, which will be present in all the
objects in dierent amounts, we have increased the systematic error from
5% to 10%, while leaving the random error the same. These updated errors
are shown in the tables of photometry (Tables B.1 & B.2).
4.6 Sources used
For the purposes of the modelling done in this thesis, not all of the sources
that were observed will be used. The sources that are not are excluded for
a number of reasons.
Firstly, we have excluded those sources that were not observed or de-
tected in at least six bands. The reasoning behind this was to make sure
that there were more points than free parameters in the model tting (i.e.
that the number of degrees of freedom in the tting is signicantly more

4.6 Sources used 59
Figure 4.5: Redshift distribution, showing sources with photometry (\PHFS + Data")
and those sources selected for modelling (\Modelling") (see text for criteria). The entire
PHFS sample is shown for comparison (\PHFS").
than zero).
Secondly, sources without a measured redshift were excluded. The red-
shift is needed to accurately obtain the observed shape of the model (in
particular those parts that are not a power law). Also, it is preferable to
be able to use rest-frame quantities (such as wavelength), as these are in-
trinsic to the source and not dependent on the observer. One source that
had an unknown redshift in Drinkwater et al. (1997), PKS 0829+046, has
a published redshift of z = 0:18 (Falomo 1991) (found by tting to the host
galaxy emission), which is used herein.
Finally, also excluded from the main analysis were those sources with
prominent additional spectral components. One class of these objects were
the low-redshift galaxies, which show a prominent 4000  A break between the
B and V bands. These sources show strong evidence for signicant ux from
the underlying galaxy in their spectra (Masci, Webster, and Francis 1998),
and thus would need an additional galaxian component to be modelled ac-
curately { this is beyond the scope of this thesis. The other class of sources
to be excluded were the high-redshift (i.e. z > 3) quasars. These sources
have strong Ly breaks present between the B and V bands.
Once all these exclusions were made, the number of sources decreased

60 Sources and Data
from 157 to 117. For completeness, the redshift distribution of the selected
sources, compared to that of the PHFS as a whole, and the sources with
any optical/NIR photometry from Francis et al. (2000), is shown in Fig. 4.5.
Note the large number of sources with small redshift that have been excluded
(due to the criterion for removing galaxy-dominated sources). The only
source with z < 0:1 is the BL Lac 1514 241 (z = 0:0486) { this source
is dominated by its smooth continuum, so that no 4000  A break from the
host galaxy is visible. The next highest redshift source is 0829+046 at
z = 0:18. The photometry is listed in Table B.1, with the new errors, while
the photometry for the excluded sources is listed in Table B.2, together with
the reason for each source's exclusion.