Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 344, 668-674 (1999)

Previous Section Next Section Title Page Table of Contents

4. Comparison with polarization models

4.1. PTEAR mechanism

It has been pointed out by Onaka (1995) that dust grains illuminated anisotropically will have different temperatures depending on the cross-section they present to the radiation source. This can produce a net polarization without grain alignment, because the grains oriented with a large surface area towards the source will be hotter, and thus emit more. This mechanism of polarized thermal emission by anisotropic radiation (PTEAR) can also modify the polarization characteristics when grains are aligned (Onaka 1996). The effects are predicted to be more significant at shorter wavelengths than in the submillimetre.

Our 800 [FORMULA] observations of W3-IRS5 (Greaves et al. 1994) showed a four-point centro-symmetric polarization pattern that could be explained by PTEAR, with IRS5 as the illuminating source. To test this hypothesis, we have made another four-point polarization map around the IRS4 source in W3, which is very similar to the IRS5 core in luminosity and extinction (Oldham et al. 1994). Thus if PTEAR is a significant effect, IRS4 should also show a centro-symmetric polarization pattern.

As shown in Fig. 1, the polarization vectors around IRS4 are not centro-symmetric. Three of the four offset points have vectors pointing roughly towards IRS4, and only the northern point has a centro-symmetric orientation. By considering the Stokes parameter Q (= p cos2[FORMULA]) expected in PTEAR, we can constrain the magnitude of the effect. For the north and south points, Q should be negative, while for the east and west points, Q should be positive (and the parameter U = p sin2[FORMULA] should be zero at all positions). In fact, Q has the wrong sign at three of the four points, so another effect, presumably magnetic alignment, must provide a larger Q component of opposite sign.

Further, we can compare the IRS4 results with the Q and U observed for the four points at the same offsets around IRS5 (Greaves et al. 1994). If we assume that all the polarization there is due to PTEAR, and exactly the same effects are produced around IRS4, then p(PTEAR) = [FORMULA]Q(IRS5)[FORMULA] and p(magnetic) = [FORMULA]. Then the average p(PTEAR) and p(magnetic) are 2.4% and 3.4% respectively, giving p(PTEAR)/p(magnetic) [FORMULA] 0.7. If our initial assumption that p(IRS5) [FORMULA] p(PTEAR) is invalid, then this ratio will be an upper limit. These results indicate that magnetic alignment dominates over PTEAR in the submillimetre, even for cores with very luminous sources (L(IRS4) = 105 [FORMULA], Oldham et al. 1994).

4.2. Suprathermal rotation

It was originally suggested by Purcell (1979) that grains could be spun up to suprathermal speeds, when H2 molecules were ejected after forming on the grain surface. This spin-up enhances alignment because the grains are less disoriented by gas-grain collisions. However, free H atoms are needed to adsorb onto the grain surfaces, whereas in molecular clouds nearly all of the hydrogen is in the form of H2. Recently, Sorrell (1995a,b) has proposed a modified model, in which cosmic rays strike grain mantles, ejecting accreted H2 molecules from the surface and providing the necessary torque for spin-up. Lazarian & Roberge (1997) have argued that this torque will not be significant for typical cosmic ray fluxes.

A key prediction of the Sorrell model is that the efficiency depends on the mantle composition. In particular, H2 molecules evaporate rapidly on pure CO mantles, faster than the timescale for a cosmic ray to arrive, so these grains will not spin up, whereas grains with H2O or CO2 mantles will. Thus p should be much less where CO mantles dominate over H2O/CO2 mantles.

We have tested this model using the observations towards the two sources in Mon R2. The line of sight to IRS2 has a deep 4.67 [FORMULA] CO absorption feature which is not seen towards IRS3 (Geballe 1986). Thus, if the grains around IRS2 cannot align, we expect a much lower p than towards IRS3. In fact, we find that at 800 [FORMULA], p is very similar in the two sources, and at 1100 [FORMULA] the values are consistent within the errors. Further, for IRS2 p(800) is detected at the 95% confidence level (method of Clarke & Stewart 1986), inconsistent with the hypothesis that this source is unpolarized. Thus, at the moderate confidence levels of our data, it appears that CO-mantled grains are aligned in Mon R2, and the model predictions are not supported. A related result is polarization of the 4.67 [FORMULA] CO ice feature towards W33A (Chrysostomou et al. 1996), which also shows alignment of CO-mantled grains.

4.3. Radiative alignment

Irradiation can affect grain alignment, as molecules may desorb from the grain surface, affecting the spin rate (Lazarian 1995). Also, irradiation of asymmetric grains can cause suprathermal spin-up, and anisotropic radiation fields can provide torques that directly align grains (Draine & Weingartner 1996, 1997).

In Fig. 2, we present a possible example of radiative alignment. The data shown are 800 [FORMULA] polarization observations of the SM1/VLA1623 core in [FORMULA] Oph A (Holland et al. 1996). It is noteworthy that the perpendiculars to the six observed polarization vectors point near the star VSSG27. The perpendiculars are within 8o of the star's direction, on average (with a mean error in the [FORMULA] values of 12o), so the convergence on VSSG27 is significant.

[FIGURE] Fig. 2. 800 [FORMULA] polarization data and dust emission map of [FORMULA] Oph A, from Holland et al. (1996). Perpendiculars to the polarization vectors are shown (dotted lines), and they converge near the T Tauri source VSSG27 (star symbol).

This convergence is unlikely to be a chance effect, since there are only a few stars associated with the cloud that could produce aligning radiation. From the K-band survey of Strom et al. (1995), we find only three stars that are (a) bright at 2.2 [FORMULA] (K-magnitude [FORMULA] 12), (b) located to the east of the ridge where the lines converge, and (c) within a projected distance similar to VSSG27. This star is 74" from the (0,0) position of the dust ridge, and there are only two other stars within 150" which fit our criteria, neither of which is near the convergence point. VSSG27 is a Class II classical T Tauri (André & Montmerle 1994); it is not the most luminous object in the region (Strom et al. 1995), but is an embedded source and therefore likely to be closely associated with the molecular cloud core. These results for [FORMULA] Oph suggest that stellar radiation could affect grain alignment in star-forming clouds.

4.4. Wavelength dependence of p and [FORMULA]

It is expected that the degree of polarization in the submillimetre will be independent of wavelength (e.g. Hildebrand 1988), since the dust emission is usually optically thin, and thus p depends only on alignment efficiency and field morphology. Previous polarimetry observations have suggested p may sometimes be wavelength-dependent, but often such individual cases can be explained by differences in resolution, where a region of complicated magnetic field is observed.

To determine if p can intrinsically depend on wavelength, we have observed a region where the field is very close to linear, the DR21 cloud core (Minchin & Murray 1994). Observations were made at two wavelengths towards three points, using matched beam sizes of 19" at both wavelengths. Any variation in p([FORMULA]) can thus not be attributed to field structure within the beam. The results (Table 1) show that p does differ with wavelength. For the (0,0) position, p(800) exceeds p(1100) by 0.95 [FORMULA] 0.34%, where the error is the sum in quadrature of the errors in each of the two measurements. However, for the offset points p(1100) exceeds p(800), by 1.97 [FORMULA] 0.72% (south) and 0.69 [FORMULA] 0.58% (north). A wider study of 115-2000 [FORMULA] polarization in DR21 (Glenn et al. 1997) has shown some additional evidence for wavelength dependence, but the range of beamsizes is large (14-42") which makes the interpretation more complex.

In DR21, the polarization position angles also vary with wavelength by a small amount. The mean value at 800 [FORMULA] is [FORMULA] = 25 [FORMULA] 3o for the three points (our data and Minchin & Murray 1994), while at 1100 [FORMULA] it is 11 [FORMULA] 5o, significantly lower. These changes in both p and [FORMULA] suggest that the two wavelengths are not equally tracing the same grains, and hence the field structure deduced is somewhat dependent on the choice of wavelength. These results are discussed further below.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1999

Online publication: March 18, 1999