Polarization Widgets--SUMMARY Decision: All ALMA receivers will have linearly polarized feeds. Discussion: SM: The use of linear feeds means that for accurate I measurements in the presence of linearly polarized sources (and calibrators), then full polarization measurements are required! Note that dual polarization is necessary as I = XX + YY (Q = XX-YY, U&V from XY-YX and XY+YX), which has data rate implications. Furthermore, there is a difference now in measuring Q and U (the gain uncertainty problem affecting Q as was stated in the ASAC report). Before accepting the linear decision, it might be good to see a detailed workup of the real problems of both schemes including response and aberrations over the bands, calibration uncertainties, etc. Im not sure the circular option has been given a fair shake. ******************What are the reasons pro and con having an insertable widget which can produce circular polarization? Advantages: 1) 'roughly' circular polarization 2) SM: Q and U are derived homogenously from the RL and LR cross-products and thus leakage is more easily calibrated. 3) SM: for the bulk of observations that want high-quality I measurements, a single polarization product RR or LL will suffice as a proxy in the absence of V(=L-R, thus poorly measured) signal. With only linear polarization, both linears must be measured to estimate I. 4) Some advantages for VLBI observations. L D'A: There are various ways to handle the VLBI case. If many ALMA antennas are to be "phased together" for VLBI, as is now done with the VLA, then a linear-to-circular conversion can be done in analog hardware at baseband, before recording. It will be a long time before VLBI recording technology is expanded to ALMA bandwidths. We can also consider doing it digitally, by programming some of ALMA's FIR filters to do Hilbert transforms and then adding/subtracting the results. Finally, the simplest option of all is to allow the ALMA data to remain in linear polarization even though other stations are circular, but to require that the full 4-element polarization matrix be produced by the correlator; then no sensitivity is lost, but accurate polarimetry requires calibration that is arguably more difficult. Thus, I do not believe that the need to accomodate VLBI should affect the design of the initial optics ("widgets") at all, nor do I think it should be allowed to have a large impact on the signal processing at IF. Disadvantages: 1) lossy (QWP perhaps twice as much as reflection-type wave plate; latter is 0.14 db in NRO instantiation) 2) narrow bandwidth (compared to receiver band; but we have no definition of 'narrow'--Crystal noted that the interesting Zeeman lines tend to lie at the edges of the receiver bands). (QWP much more so than reflection-type wave plate--NRO instantiation covers 35-250 GHz, though it has not been used above 110 GHz and stability may limit this). 3) it is a widget L D'A: A QWP enables us to produce a beam that has *roughly* circular polarization, as opposed to one that has *accurate* linear polarization. For a point source, the issue is how well the actual polarization of the instrument can be known. For an extended source, the main issue is how much the polarization varies over the beam. Both issues can only be made worse by inserting the QWP. Here I am assuming that we always have a dual-polarization receiver and that we measure the full 2x2 polarization matrix. *******************What should the design of this widget be? The reflection type wave plate is less lossy and provides substantially more bandwidth. It probably presents engineering challenges, however, which have not been assessed. OVRO: Akeson and Carlstrom ApJ 491, 254-266 (1997) used a tunable grid plus mirror on OVRO for their mm polarization program. This would alleviate some of the bandwidth problems, though I dont see how one could be incorporated into our system without tertiaries and sidecabs (I dont picture it easily fitting into the beam, let alone on in the optics for every band). Lamb: For all our polarization measurements we use a grid/plane mirror combination which can be used to produce any polatization by adjusting the mirror-grid spacing and grid angle. It works pretty well on the whole, but it seems that the repeatability may not be quite good enough for 1-mm observations. It needs to be repeatable at the 1 um level. It could probably be engineered to this level, but the fragility of the grids will probably be an issue. The repeatability requirement might not be as bad for ALMA. We have to change polarization state on each antennas to go through all the combinations of LL, LR, RL, and RR. On ALMA that would not be necessary since ther are two polarization channels available simultaneously. BIMA use grooved dielectric plates which are lossier and narrow band, but probably more repeatable. SM: The OVRO design, by tuning the grid wrt the mirror, can cover a band at least. It may be that one polarizer could cover a couple of bands, but Im not sure as the wire grid spacing matters to some extent. Action items: Crystal noted that stability of the parallel hands V=RR-LL or Q=XX-YY is enhanced by having a transfer switch in the front end that lets one swap the signal paths of RR and LL or XX and YY through the electronics - the VLA does can do this (though I wasnt aware of that mode) and seems to be necessary for high-sensitivity Zeeman work (which needs V). Has this been considered for ALMA? I will ask Larry about this but though I'd bring it up with you also. (AW mentioned it to LD'A). What does BIMA do? A detailed description of the instrument is due in Rao, R., Crutcher, R., Girart, J. M., Lai, S.-P., Wright, M. C. H., Plambeck, R. L., Lugten, J. B. in preparation. From Rao (ApJ 502, L75): Quarter wave plates are used in front of the linearly polarized BIMA feeds to obtain left (L) or right (R) circular polarizations. Since only a single polarization is received, L and R are time-multiplexed on each antenna using a fast Walsh function switching pattern in order to sample all possible cross correlations (LL, LR, RL, RR) on every baseline. The data were averaged over the Walsh cycle to produce quasi-simultaneous dual polarization measurements. Beam smearing resulting from this averaging process is negligible. The instrumental polarization response, or "leakage," for each antenna was calibrated by observing a strong point source (3c273 or 3c279) over a wide hour angle range to provide good parallactic angle coverage. For antennas with alt-az mounts and orthogonal circular feeds, the fringe phases for a linearly polarized source vary with parallactic angle, while the instrumental leakages remain constant, so one can solve simultaneously for the leakages and the source polarization (Sault, Hamaker, & Bregman 1996). Typical leakage amplitudes were 1% at 3.3 mm and 5% at 1.3 mm. Leakages measured on different days or on different calibrators were consistent within 0.3% rms at 3.3 mm and 0.5% rms at 1.3 mm. Test observations also showed that the instrumental polarization does not vary strongly across the primary beam. The QSO 3c286 was used to check the position angle calibration. Heiles opinion. The ATCA could describe their experiences with linear feeds. It sounds like no useful Zeeman observations have come out of that instrument, perhaps due to the choice of optics. Ask Miller? WSRT used to have linear feeds, but have they changed during the upgrade? Ask Miller?? Shinnaga used what at Nobeyama? We have designed and constructed a tunable polarimeter to cover frequencies from 35 GHz to 250 GHz (8.6 mm and 1.2 mm in wavelength) for the 45-m telescope at Nobeyama Radio Observatory. Both circular and linear polarizations can be measured by the polarimeter. The reflection-type wave plate is assembled by mounting a free-standing wire grid in front of the reflecting surface of a plane mirror with both surfaces parallel to each other (Howard et al. 1986; Prigent et al. 1988). The grid is made of a plane array of parallel wires with circular cross section. The insertion loss was measured to be 0.14 +/- 0.05 dB in the 100-GHz band, about half that of the transmission-type polarimeter with a stack structure made of Teflon in the same frequency band. The overall instrumental polarization of the system in the 100 GHz band is as low as <= 3%. The performance of the polarimeter in astronomical observations was tested by simultaneously measuring the linear polarization of the J = 2--1 transition of SiO in the v = 0 and 1 states at 86 GHz toward VY Canis Majoris. The observation revealed that the J = 2--1 emission in the v = 0 state of the object is highly linear polarized, which suggests that the emission originates through maser action in the circumstellar region. The details of the design, construction, and tests are presented. \bibitem[Shinnaga, Tsuboi \& Kasuga(1999)]{1999PASJ...51..175S} Shinnaga, H., Tsuboi, M.\ \& Kasuga, T.\ 1999, \pasj, 51, 175 Instantaneous Bandwidth of the Polarimeter with the Half-Wave Plate The instantaneous bandwidth of the polarimeter set as HWP, df/f (f is the observation frequency), depends on the repeatability of the measurements with the system. It was found that there was an error of 1% in the relative intensity for repeated measurements of the same point of a source using the telescope system. In observing linear polarization, this error introduced a 2% uncertainty in measuring the degree of polarization P, which means 100% linear polarized radiation is observed to be P=98%, using the system. The degradation in P of 2% is also estimated to be caused by the phase-shift error at the band edge, f +/- df ~ 1 +/ -0.08. Thus, the bandwidth df/f of the polarimeter is considered to be ~+/-0.08. Insertion Loss A low insertion loss is essential for the good performance of a polarimeter. The insertion loss of our polarimeter is due to the surface resistance and irregularities of the wave plate. The loss at 110 GHz was measured to be 0.14+/-.05 dB, in good agreement with the estimated value of 0.14 dB at 110 GHz, calculated from the loss of surface resistance (0.06 dB) and that of irregularity (0.08 dB). The losses at other frequencies estimated using the same assumption were 0.11 dB at 35 GHz and 0.17 dB at 250 GHz. The insertion loss of our system is very low compared to a transmission-type polarimeter made of Teflon stack with a QWP configuration. For instance, the calculated loss of our polarimeter at 90 GHz was 0.13 dB, about half of the value reported by Watanabe (1981, 0.3 dB in the same frequency band). For a system at a noise temperature of 200 K, a loss of 0.14 dB corresponds to an increase of noise by ~15 K, compared to an increase of ~35 K at a loss of 0.3 dB.