Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 344, 402-408 (1999)

Previous Section Next Section Title Page Table of Contents

2. Observations and data reduction

We obtained 8.4-GHz observations with the VLBA in December 1994 (1994.95). A total of eight baseband converters (BBCs) were used, with four recording right-circular polarization (RCP) and four recording left-circular polarization (LCP). Each BBC had a recording bandwidth of 8 MHz, with 2-bit 16-MHz sampling. In order to study the distribution of the very high rotation measure of 3C 119 on milliarcsecond scales, we spread the different sky frequencies for the four sets of BBCs across the available 8.4-GHz band for the VLBA antennas, which encompasses roughly 840 MHz. By having the individual BBCs observe at widely spaced frequency bands rather than at adjacent frequency bands, we were able to obtain simultaneous polarization images at different frequencies in the 8.4-GHz band, which could then be used to study the rotation measure distribution on milliarcsecond scales.

Due to technical limitations of the VLBA and BBC distribution system, it was not possible to set the BBCs to observe simultaneously at four arbitrary frequencies in the available band. We overcame this limitation as follows. In order to maximize the total bandwidth over which the BBC frequencies were distributed, to make our rotation measure determinations as reliable as possible, we switched between two local oscillators, obtaining pairs of scans for each "observation" of each source. Each source was observed first using a local oscillator at 7604 MHz (upper side band; LO1), then using a local oscillator at 9404 MHz (lower side band; LO2). The time allowed between scans for the LO switching and general system checks was three minutes. Technical details of our setup, including the local osillator frequencies, BBC frequencies, and resulting sky frequencies are given in Table 1. After concatenating the datasets obtained using the two LOs, the maximum difference between the individual BBC frequencies was about 700 MHz. The integrated rotation measure of 3C 119 (1728 rad/m2; Kato et al. 1987) corresponds to a rotation of roughly 20o over this bandwidth.


Table 1. Frequency setup

Note that we obtained data in the frequency range 8508.49-8524.49 MHz during the measurements for both of the LOs, making it possible to calibrate the right-left phase difference between LO1 and LO2. There are two pairs of adjacent sky-frequency bands for each LO in the third column of Table 1: 8153.49-8169.59 MHz and 8508.49-8524.49 MHz for LO1 and 8508.49-8524.49 MHz and 8850.49-8866.49 MHz for LO2. We will refer to these four bands as IF12, IF34, IF56, and IF78, as indicated in the last column of Table 1 ; IF34 and IF56 cover the same frequency range.

We also had observations with the VLA for four hours during our VLBI experiment, only in the frequency range 8508.49-8516.49. This provided information about the integrated total intensity and polarization of the sources observed. We obsered two sources as VLBI calibrators: the bright unpolarized source 3C84, for use in calibrating the polarization cross talk, and the compact polarized source 0300+470 (Gabuzda et al. 1994), for determining the absolute position angle [FORMULA] of the polarized emission.

The data were calibrated, imaged, and analyzed using the AIPS package. During the entire reduction process, the data at each of the four separate frequencies were processed independently, bearing in mind possible variation in the instrumental polarization across the overall bandwidth. The fringe fitting was done in three steps: (1) manual phase calibration using a short time segment on 3C84, due to lack of phase-cal information in the data; (2) global fringe fitting; and (3) calibration of the right-left delay difference using the AIPS procedure CROSSPOL on scans of 3C84 and 3C 119. In the latter case, we assumed that the phase shift across each individual 8-MHz frequency band due to the source rotation measure was negligible.

After the initial calibration of the data, the data for each source were imaged via self-calibration using the usual AIPS techniques. The instrumental polarization parameters ("D-terms") were determined in two ways: (1) using the model of the total intensity structure of 0300+470, using a linear D-term approximation in the task LPCAL, which allows for the presence of polarization structure in the calibrator source; and (2) using the self-calibrated data for 3C84 using an ellipticity-orientation approximation using the task PCAL. The two D-term solutions obtained were similar; both clearly improved the quality of the polarization map of 0300+470, while the linear D-term solution obtained using LPCAL yielded a higher dynamic range. The D-terms were typically of the order of 1% or less, and the largest D-terms were of the order of 2% (for the Hancock and St. Croix antennas) We adopted the LPCAL D-term solution for our calibration of the 3C 119 data.

The final calibration step-determining the absolute polarization position angle-is very important if we wish to accurately map the magnetic fields on milliarcsecond scales. After the data have been calibrated up to this point, there remains an arbitrary offset in the polarization position angles [FORMULA], which is the same for all sources. This offset can be determined using data for a source in which a high fraction of the integrated polarization is present on milliarcsecond scales, by comparing the integrated [FORMULA] with the [FORMULA] for the sum of the polarized flux density on milliarcsecond scales, assuming they should be equal. We used the compact source 0300+470 for this purpose.

As a check, we derived the integrated polarizations of 0300+470 and 3C 119 in two ways. One way made use of the VLA data obtained during the VLBI experiment. In addition to observing the three VLBI sources, we observed the primary flux-density and polarization position-angle calibrators 3C48, 3C138, and 3C286 with the VLA. The VLA data for these sources were used to determine the absolute polarization position angle calibration for the VLA data in the usual way. After applying this calibration, the integrated [FORMULA] values measured for 0300+470 and 3C 119 were [FORMULA] and [FORMULA], respectively. We also derived the integrated polarization position angles for these two sources using multi-channel polarimetric data obtained by Inoue et al. (1995) using the Nobeyama 45-m telescope. The integrated [FORMULA] values for 0300+470 and 3C 119 yielded by these observations were [FORMULA] and [FORMULA], respectively, in very good agreement with our VLA measurements.

The total polarized flux density of 0300+470 on milliarcsecond scales was calculated from the sum of the Q and U flux densities in the images at each of our four frequencies in the VLBA 8.4-GHz band. In order to remove any offsets between frequencies, we applied appropriate rotations to the uv data of all the sources, so as to align the [FORMULA] values of 0300+470 at each of the frequencies with respect to the value at IF12. This procedure assumes that the Faraday rotation across the observed bandwidth is negligible, as is reasonable given the low rotation measure of 0300+470 (11 rad/m2) reported by Rudnick & Jones (1983). We then were able to determine the overall rotation needed to calibrate the VLBI [FORMULA] values by comparing the VLBI [FORMULA] for 0300+470 with the integrated values indicated by the VLA and Nobeyama measurements indicated above.

Following vector averaging of the two polarization images at IF34 and IF56, which corresponded to the same frequency, we obtained a data cube for 3C 119 at our three observing frequencies. We were then able to map the rotation measures using the AIPS task RM, which performs a weighted fit of the position angle to the [FORMULA] dependence expected for Faraday rotation.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1999

Online publication: March 18, 1999