Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.atnf.csiro.au/people/Enno.Middelberg/pubs/middelberg2004b/middelberg2004b.ps.gz Äàòà èçìåíåíèÿ: Mon Sep 27 06:29:25 2004 Äàòà èíäåêñèðîâàíèÿ: Wed Dec 26 14:25:03 2007 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ngc 4261

Proceedings of the 7th European VLBI Network Symposium
Bachiller, R. Colomer, F., Desmurs, J.F., de Vicente, P. (eds.)
October 12th­15 2004, Toledo, Spain
VLBI detections of a source weaker than 100 mJy at 86 GHz
E. Middelberg # , 1 A. L. Roy, 1,2 R. C. Walker, 3 and H. Falcke 4
1 Max­Planck­Institut f˜ur Radioastronomie, Auf dem H˜ugel 69, D­53121 Bonn, Germany
e­mail: enno.middelberg@csiro.au, aroy@mpifr­bonn.mpg.de
2 Geod˜atisches Institut der Universit˜at Bonn, Nussallee 17, D­53115 Bonn, Germany
3 National Radio Astronomy Observatory, P.O. Box 0, Socorro, NM, 87801, USA
e­mail: cwalker@aoc.nrao.edu
4 ASTRON, P.O. Box 2, 7990 AA Dwingeloo, The Netherlands
e­mail: falcke@astron.nl
Abstract. We use a new phase­calibration strategy to calibrate the phase of 86 GHz VLBI observations of the FR I
radio galaxy NGC 4261. Instead of switching between a calibrator source and the target source, the target was
observed while rapidly switching between the target frequency and a lower reference frequency. Self­calibration at
the reference frequency yielded phase corrections which were multiplied with the frequency ratio and applied to
the target frequency visibilities. The resulting detection of NGC 4261 is, to our knowledge, the first of NGC 4261
with 86 GHz VLBI, and it is also the weakest source so far detected with VLBI at that frequency.
1. Introduction
The scientific interest in VLBI observations at 86 GHz in­
cludes spatially resolved imaging of extragalactic radio jets
on the linear scales on which the jets are launched. The
process is not yet understood and only millimetre VLBI
can resolve this region even in the closest objects. Interests
also include investigations of internal jet structure, jet
composition and collimation (e.g., Doeleman et al. 2001,
G’omez et al. 1999), and imaging of the event horizon in,
e.g., Sgr A # (Falcke et al. 2000). 86GHz observations of
the FR I radio galaxy NGC 4261 provide the opportunity
to eventually resolve its jet collimation, because its com­
bination of proximity and black hole mass yields a linear
resolution of only 200 Schwarzschild radii.
Higher frequency observations involve a number of se­
rious problems. The sources are usually weaker, the aper­
ture e#ciencies of most radio telescopes used for VLBI
drop to 15 % or less, the receiver performances become
disproportionately worse, the atmospheric contribution to
T sys becomes larger, and the atmospheric coherence time
scales decrease as 1/#. In cm VLBI, the coherence time can
be extended using phase­referencing, but in mm VLBI,
phase­referencing is not commonly used, although a suc­
cessful proof of concept exists (Porcas & Rioja 2002).
Problems arise from the need for a suitable, strong phase
calibrator in the vicinity of the target source and relatively
long telescope slewing times.
We outlined the technique of fast frequency switching,
in Middelberg et al. (2002), in which an object is observed
while rapidly switching between the target frequency and
a lower reference frequency. We demonstrated there the
feasibility of the technique with detections of a calibrator
# Current address: Australia Telescope National Facility, PO
Box 76, Epping NSW 1710, Australia
and made tentative detections of a weak target, the AGN
in M81. We have since made considerable improvements to
the observing strategy and present here the first detection
of the AGN in NGC 4261 with 86 GHz VLBI.
2. Principle of phase correction
The main source of phase noise in VLBI observations at
frequencies higher than about 5 GHz is turbulence in the
troposphere causing refractive inhomogeneities. The re­
fractive screen is non­dispersive, and one can self­calibrate
the visibility phases at the reference frequency, # r , and
use the solutions to calibrate the target frequency visibil­
ity phases, # t , after multiplying them by the frequency
ratio, r = # t /# r . However, the lag between the two mea­
surements must not exceed half the atmospheric coher­
ence time. This is possible with the VLBA 1 because fre­
quency changes need only a few seconds and because the
local oscillator phases return to their original settings af­
ter a frequency switch. After multiplying the phase so­
lutions by the frequency ratio and applying them to the
target­frequency phases, there remains a constant phase
o#set between the signal paths at the two frequencies, ##,
which must be calibrated. It can be monitored with obser­
vations of achromatic, strong calibrators, and must be sub­
tracted from the high­frequency visibility phases. Thus,
the true high­frequency visibilities are phase­referenced
to the source's low­frequency visibilities, and so the tech­
nique can prolong coherence and can measure the posi­
tion shift of cores in active galactic nuclei (AGN) with
frequency.
1 The VLBA is an instrument of the National Radio
Astronomy Observatory, a facility of the National Science
Foundation, operated under cooperative agreement by
Associated Universities, Inc.

Middelberg et al.: VLBI detections of a source weaker than 100 mJy at 86 GHz
However, in our project, unmodelled ionospheric path
length changes, which are dispersive, limited the coherence
to half an hour in the worst case. Therefore, we here only
show how fast frequency switching can be used to calibrate
the short­term phase fluctuations. To remove the remain­
ing long­term phase drifts and remaining phase o#sets,
we used one extra step of self­calibration at the target fre­
quency with a half­hour solution interval. This extra step
of self­calibration prevented us from making a core­shift
measurement.
3. Observations
We observed NGC 4261 as a fast frequency switching tar­
get and 3C 273 and 3C 279 occasionally for the phase o#­
set, ##, and to test the technique on strong sources.
Dynamic scheduling allowed us to observe during a period
of superb weather, using 256Mbps to record a bandwidth
of 64 MHz with 2­bit sampling.
In this article, a pair of two integrations at # r and # t
is called a ``cycle'', each integration of which is called a
``half­cycle'', and a sequence of cycles on the same source
is called a ``scan''.
Several considerations influenced the experiment de­
sign. The target frequency should be an integer multiple
of the reference frequency to avoid having to unwrap phase
wraps. We chose a reference frequency of 14.375GHz
since the third and sixth harmonics at 43.125 GHz and
86.25 GHz lie within the VLBA receiver bands. For con­
venience, we will refer to these frequencies as ``15 GHz'',
``43 GHz'' and ``86 GHz'', respectively.
We chose a cycle time of 50 s, of which 22 s were spent
at the reference frequency of 15 GHz and the remaining
28 s were spent at the target frequency, either 43 GHz or
86 GHz. An average time of 7 s per half­cycle was lost in
moving the subreflector between the feed horns, resulting
in net integration times of 15 s at # r and 21 s at # t . The
integration times are a compromise, depending on source
brightness, antenna sensitivity and expected weather con­
ditions. This setup yielded a 5 # detection limit of 89 mJy
in 15 s at 15GHz for the VLBA on a single baseline.
4. Data reduction
Data reduction was carried out in AIPS. The amplitudes
were calibrated using T sys and gain measurements, and
amplitude corrections for errors in the sampler thresholds
were performed using autocorrelation data. The instru­
mental delays and inter­IF phase o#sets were corrected
using a fringe­finder scan and were found to be stable
over the experiment. A correction for the dispersive de­
lays introduced by the ionosphere was attempted using
GPS­based, global maps of the total electron content.
Unfortunately, the error in these maps can be quite high
(typically 10 % to 20 %, but up to 50%), causing resid­
ual phase rates after the transfer of phase solutions be­
tween frequencies. These errors prevented us from making
a core­shift measurement because self­calibration was still
required to calibrate the residual rates.
From the start of a new half­cycle, 5 s to 8 s are re­
quired to position the subreflector. Data during that time
should be flagged. Compromise flagging times for each an­
tenna and frequency at the start of each half­cycle were
applied using the AIPS task QUACK.
We fringe­fitted the 15 GHz data using the AIPS task
FRING. We made a 15 GHz image that we used as a source
model in a second run of FRING, so that the phase solu­
tions did not contain structural phase contributions. The
solution interval was set to 1 min, yielding one phase, delay
and phase rate solution per half cycle. The detection rate
was # 90 %. The solution table was written to a text file
to do the phase scaling outside AIPS. A Python program
was used to read in the table, to generate timestamps to
coincide with the target frequency half­cycles, and to cal­
culate a solution consisting of a phase, a phase rate, and
a delay. These solutions were imported to AIPS to update
the most recent calibration table at the target frequency.
5. Results
5.1. 43 GHz
NGC 4261 was detected on most baselines at all times af­
ter scaling the 15 GHz solutions to 43 GHz, with correlated
flux densities of 30 mJy (800 M#) to 160mJy (30 M#).
Here, the term ``detected'' means that by inspecting the
phase­time series by eye one can see that the phases are
not random. The 43 GHz half­cycle average visibilities on
baselines to Los Alamos are shown in Fig. 1. The short­
term fluctuations introduced by the troposphere are al­
most perfectly calibrated, but residual phase drifts remain
on longer time­scales, especially at the beginning and at
the end of the experiment, when the sun was setting and
rising at Los Alamos and the source elevation was low at
most stations.
Before making an image from the 43GHz visibili­
ties calibrated with fast frequency switching, the resid­
ual phase o#sets and phase rates were removed using
fringe­fitting with a solution interval of 30 min so that
one solution per scan was obtained. The resulting dirty
image (Fig. 2) has a peak flux density of 79 mJy beam -1
and an rms noise of 4.4 mJy beam -1 , yielding a dynamic
range of 18:1. After several cycles of phase self­calibration
with a solution interval of 30 s and one cycle of amplitude
self­calibration with a solution interval of 12 h, the peak
flux density was 95 mJybeam -1 and the rms noise was
0.78 mJybeam -1 , so the dynamic range improved to 122:1
(Fig. 3). The theoretical rms noise at 43 GHz was expected
to be 0.46 mJy beam -1 . Thus, fast frequency switching al­
lowed us to use a half­hour solution interval in fringe fit­
ting, and so gain much improved detection sensitivity.
The performance of the calibration technique can be
demonstrated by comparing the measured phase noise to
the expected phase noise. We have measured the rms of
the cycle­to­cycle variations of the calibrated 43 GHz visi­

Middelberg et al.: VLBI detections of a source weaker than 100 mJy at 86 GHz
bilities after a) only applying the scaled­up 15 GHz phase
solutions and b) after correcting for the residual phase er­
rors with a cycle of self­calibration using a 30 min solution
interval. In case a), the rms was found to be 50 # in the be­
ginning and end of the experiment (coinciding with dusk
and dawn at the stations in the south­western US and
with predominantly low elevations), and was 33 # in the
middle of the observations (at night at most stations and
at high elevations). In case b), the rms dropped slightly
to 44 # and 31 # , respectively.
The expected phase noise in the visibilities calibrated
with fast frequency switching consists of three parts (aside
from long­term changes in electron content, the ionosphere
does not contribute noticeable phase noise, and errors in
the source model at # r are negligible). a) thermal phase
noise at the reference frequency scaled by the frequency ra­
tio, b) thermal phase noise at the target frequency and c)
tropospheric phase changes during the two integrations.
We estimate the three terms as follows, assuming that
NGC 4261 has an average compact flux density of 200mJy
at 15 GHz and 100mJy at both 43 GHz and 86 GHz. a) On
a single baseline, the expected signal­to­noise ratio (SNR)
of a detection at 15 GHz is 11.2 when averaging over the
band. In fringe­fitting, this is increased by # N , where N
is the number of baselines. N # 8, so the SNR of a de­
tection at 15 GHz increases to 31.7, corresponding to a
phase error of 1.8 # . This error is scaled by the frequency
ratio to 5.4 # at 43GHz and 10.9 # at 86 GHz. b) The ther­
mal noise contributions at 43 GHz and 86 GHz are 21.4 #
and 54.7 # , respectively. c) We estimated the tropospheric
phase noise within the switching cycle time using obser­
vations of 3C 273, which are essentially free of thermal
noise and any phase changes during and between the half­
cycles are due to changes in the troposphere. We found the
median rms phase noise after fringe­fitting with a 30 min
solution interval to remove the residual long­term phase
drift to be 13.3 # at 43 GHz, or 26.6 # at 86 GHz.
Adding those three noise components in quadrature
yields 25.8 # at 43 GHz and is in good agreement with the
measured rms phase noise of 31 # .
5.2. 86 GHz
Following the same data reduction path as for the 43 GHz
data, we obtained good detections of NGC 4261 at 86 GHz
on baselines among the four stations FD, KP, LA, and
PT and only weak detections on baselines to NL, OV and
MK. The corrected visibility phases are plotted in Fig. 4,
and an image is shown in Fig. 5. The peak flux density is
59.3 mJy. To our knowledge, this is the first VLBI detec­
tion of NGC 4261 at 86 GHz, and is probably the weakest
continuum object ever detected with VLBI at this fre­
quency.
With only delay calibration applied, the median rms
phase noise of the baselines during the best 25 min scan
is 104 # , after applying the scaled 15 GHz phase solutions
is 70 # and after fringe­fitting with a 30 min solution inter­
100
0
­100
BR ­ LA ( 1 5
100
0
­100
HN ­ LA ( 3 5
100
0
­100
KP ­ LA ( 4 ­ 5
Degrees 100
0
­100
LA NL 5 ­ 7 )
100
0
­100
LA OV ( 5 8
100
0
­100
LA PT ( 5 9 )
TIME (HOURS)
1/00 1/02 1/04 1/06 1/08 1/10
100
0
­100
LA SC ( 5 ­ 10 )
Fig. 1. NGC 4261 calibrated 43 GHz visibility phases on base­
lines to Los Alamos. Calibration used ionospheric corrections
and scaled­up phase solutions from fringe­fitting with a clean
component model at 15 GHz. The weather at HN and SC was
worse than at the other stations, hence the detections have
lower SNR. Each data point is an average over a half­cycle.
Fig. 2. Naturally weighted, full­resolution image of NGC 4261
at 43 GHz, calibrated with scaled­up phase solutions from
15 GHz. Fringe­fitting has been used to solve for one residual
phase and rate solution per 25 min scan before exporting the
data to Difmap. No further self­calibration has been applied.
The image noise is 4.4 mJybeam -1 and the dynamic range is
18:1. Contours start at 12.6 mJy beam -1 and increase by fac­
tors of 2.
val is 80 # (the increase in rms phase noise after removal
of phase rates has an unknown cause). We estimate the
expected phase noise to consist of the scaled­up 15 GHz
noise, 10.9 # , the thermal 86 GHz noise, 54.7 # , and the tro­
pospheric phase errors, 26.6 # . Their quadrature sum is
61.7 # , in agreement with the measured noise levels.

Middelberg et al.: VLBI detections of a source weaker than 100 mJy at 86 GHz
Fig. 3. Data and imaging parameters as in Fig. 2, but sev­
eral cycles of phase self­calibration with a solution interval of
30 s and one cycle of amplitude self­calibration with a solu­
tion interval of 12 h have been applied. The image noise is
0.78 mJy beam -1 and the dynamic range is 122:1. Contours
start at 1.9 mJy beam -1 and increase by factors of 2.
100
0
­100
FD ­ KP ( 2 ­ 4
100
0
­100
FD ­ LA ( 2 ­ 5 )
100
0
­100
FD ­ OV ( 2 8 )
100
0
­100
FD ­ PT ( 2 ­ 9 )
100
0
­100
KP ­ LA ( 4 ­ 5
Degrees 100
0
­100
KP ­ OV 4 ­ 8 )
100
0
­100
KP ­ PT ( 4 ­ 9 )
100
0
­100
LA ­ OV ( 5 8 )
100
0
­100
LA ­ PT ( 5 ­ 9 )
TIME (HOURS)
1/00 1/01 1/02 1/03 1/04 1/05 1/06 1/07 1/08 1/09
100
0
­100
OV ­ PT ( 8 ­ 9
Fig. 4. 86 GHz visibility phases on baselines among Fort
Davis, Kitt Peak, Los Alamos, Owens Valley and Pie Town.
Calibration has been done with scaled­up phase solutions from
fringe­fitting at 15 GHz using a clean component model. Good
detections were made during almost every 25 min scan observed
at night between 2:00 h UT and 7:00 h UT.
6. Summary
We have demonstrated the feasibility of using fast fre­
quency switching as a phase calibration method for
86 GHz VLBI, yielding the detection of the faintest source
so far at that frequency. Although ionospheric e#ects pre­
Fig. 5. Naturally weighted, full­resolution image of NGC 4261
at 86 GHz, calibrated with scaled­up phase solutions from
15 GHz. Fringe­fitting has been used to solve for one resid­
ual phase and rate solution per 25 min scan before export­
ing the data to Difmap. No further self­calibration has been
applied. The image noise is 8.4 mJy beam -1 and the dy­
namic range is 7:1. Contours are drawn at 19 mJybeam -1 and
38 mJy beam -1 .
vented purely phase­referenced images and hence the de­
tection of a core shift, the technique allows the detection of
much fainter sources at 86 GHz than is possible with con­
ventional VLBI. This includes nearby sources which can
be imaged with unprecedented linear resolution and allow
one to study the jet formation and collimation processes.
References
Doeleman, S. S., Shen, Z.­Q., Rogers, A. E. E., Bower,
G. C., Wright, M. C. H., Zhao, J. H., Backer, D. C.,
Crowley, J. W., Freund, R. W., Ho, P. T. P., Lo, K. Y.,
& Woody, D. P. 2001, AJ, 121, 2610
Falcke, H., Melia, F., & Agol, E. 2000, ApJ, 528, L13
G’omez, J., Marscher, A. P., & Alberdi, A. 1999, ApJ, 521,
L29
Middelberg, E., Roy, A. L., Walker, R. C., Falcke, H.,
& Krichbaum, T. P. 2002, in 6th European VLBI
Network Symposium on New Developments in VLBI
Science and Technology, ed. E. Ros, R. W. Porcas,
A. P. Lobanov, & J. A. Zensus (Max­Planck­Institut
f˜ur Radioastronomie), 61
Porcas, R. W. & Rioja, M. J. 2002, in 6th European
VLBI Network Symposium on New Developments in
VLBI Science and Technology, ed. E. Ros, R. W. Porcas,
A. P. Lobanov, & J. A. Zensus (Max­Planck­Institut f˜ur
Radioastronomie), 65