Astron. Astrophys. 358, 572-574 (2000)

2. Instrumental problems?

The MuSiCoS échelle spectropolarimeter (Baudrand & Böhm 1992; Donati et al. 1999) used in Donati & Wade's (1999) investigation uses twin fibers to transfer the double split beam emerging from the polariser, to the spectrograph. The star is imaged directly onto each of the twin fibers. From beam ratios at the detector in the spectrograph, the wavelength-dependent Stokes' parameters can be derived. From an appropriate choice of double ratios (e.g. Eversberg et al. 1998; Tinbergen 1996), the fractional values of , and can attain very high precision in principle, limited mainly to photon statistics and nearly impervious to pixel-to-pixel sensitivity variations.

Eversberg et al. (1998, 1999) used a similar setup in their recently built William-Wehlau (WW) spectropolarimeter. The major difference between the two apparati is that the WW instrument employs two quarter-wave plates (QWPs) simultaneously to measure q, u or v, while MuSiCoS uses either a QWP to measure v or no plate at all, but instrument rotation to measure q and u. In addition, as with MuSiCoS, Eversberg et al. (1998) find a scatter in continuum (i.e. broadband) polarisation from one measure to another, when stars are observed (see their Fig. 4). Typically, the scatter is 1% in q, u or v from one exposure to another independent of the telescope, somewhat worse in the blue. However, the instrument can still be used very well for high precision, relative line polarisation work, since the deviations in a given measurement vary only slowly with wavelength.

The MuSiCoS instrument, being very similar in general concept, leads to a similar scatter of 0.8% in broadband q, u or v (Donati et al. 1999). However, details between the two instruments differ, so that the quantitatively similar scatter in continuum polarisation may be a coincidence. For example, the ratio of analyser splitting angle to fiber core diameter is quite different in the two instruments, leading to possibly stronger chromaticity in the WW instrument; on the other hand, MuSiCoS has fibers that are twice as small in core diameter as those in the WW polarimeter, and thus being more sensitive to positioning errors. Note that this continuum effect is completely independent of the Donati & Wade (1999) magnetic field upper limit and does not affect that result.

While small in the context of transmission at the fiber interface, this 1% scatter is large (even fatal in most cases) for astronomical work on continuum or absolute line polarisation. In the case of the WW spectropolarimeter, Eversberg et al. (1998) have shown that the scatter is most likely due to small inhomogeneities in spatial surface sensitivity across the face of the fibers, combined with varying average position of the star on the fibers from one exposure to the next. This causes the beam intensities (and their ratios) to fluctuate, depending on exactly where the star is focussed on the fibers. It has nothing to do with intrinsic polarisation of the source. Presumably, small guiding errors are the main cause of the star's average position on the fibers changing from one exposure to another. Even a small change in position is apparently enough to cause this effect.

In contrast to stellar light, Eversberg et al. (1998) find that extended sources (e.g. flat fields) do not fluctuate by more than 0.02% (!) and are of much lower residual value (especially in v, where the source polarisation is less likely to be significantly different from zero; see their Fig. 5). Since fiber transmission generally falls off towards shorter wavelengths, the effect becomes gradually worse towards the blue. Note that it is not important whether the extended source is local or at "infinity" (i.e. astronomical); what is important is the uniform illumination of the fibers.

Based on the above experience of Eversberg et al. (1998), it appears very likely that the continuum circular polarisation observed by Donati & Wade (1999) is spurious: The continuum + photospheric-line polarisation (in circular mode v) is a result of the stellar light (i.e. a point-spread function), which with a 2-fiber system is known to fluctuate, as discussed above. As can be seen from the data of Donati & Wade (1999), the continuum circular polarisation in Ori C fluctuates by close to 1% rms, even including the freak 3.8% deviation observed near H 1997 on Feb 20 ¹. The component of narrow emission line flux on the other hand is dominated by the nebular light of the Orion Nebula, which is an extended source. It therefore behaves like a flat field source, which as Eversberg et al. (1998) have shown for their similar apparatus, shows little fluctuation and lower polarisation. Of course, in the case of Ori C, about 1/3 of the light at the peak of H at Donati & Wade's (1999) spectral resolution is stellar (continuum + emission + absorption), so the polarisation does not necessarily fall to zero at line peak. In fact, for zero intrinsic and instrumental polarisation of nebular H (the normal case), one expects the observed circular polarisation, based on the relative line to continuum flux at line peak, to fall to c. 1/3 of the continuum value (regardless of its origin), as seen. Eversberg et al. (1998) failed to see any significant instrumental depolarisation in H probably because of their much lower spectral resolution.

As a first check of the spurious nature of the continuum circular polarisation in Orionis C, one should use MuSiCoS to measure simultaneous circular and linear polarisation across the H line. If the above explanation is correct, as already appears highly likely based on our experience with the WW spectropolarimeter, one would expect the same to occur as in circular polarisation, i.e. one would expect to find (spurious) variable continuum linear polarisation. In fact, Eversberg et al. (1998) already showed the same spurious continuum polarisation in any of q, u and v in their data using the WW spectropolarimeter. A more definitive check would be to compare observations of astronomical sources (rather than local dome-flats) that are uniform and extended (e.g. planets in this context) with point-like stellar light.

© European Southern Observatory (ESO) 2000

Online publication: June 8, 2000
helpdesk.link@springer.de