INTERSTELLAR POLARIZATION FROM 3-MICRON TO 5-MICRON IN REDDENED STARS

Measurements on infrared interstellar linear polarization are reported in the spectral range 2-5-mu-m for 18 highly reddened stars. The observations covered lines of sight with wavelengths of peak polarization lambda(max) ranging from the blue-ultraviolet (0.35-mu-m) to the near-infrared (0.78-mu-m) and thus sampled a range of environments. The data are used to address two related questions: (1) what is the appropriate empirical description of the wavelength dependence of continuum polarization, and (2) what is the physical basis (in terms of grain size distribution and the dielectric functions of various materials) for this behavior? We show that excess polarization occurs at 3-5-mu-m relative to the standard Serkowski/Wilking empirical formula describing the wavelength dependence of polarization in the optical and near infrared out to 2.2-mu-m: specifically, we confirm the results of Nagata and Jones that significant excess exists at 3.6-3.8-mu-mm, and we show that it persists to 5-mu-m. This systematic behavior clearly indicates that extrapolation of the Serkowski/Wilking formula cannot adequately represent the infrared continuum polarization at these wavelengths. The form of the infrared polarization is better represented by a power law p(lambda) is-proportional-to lambda(-beta), with beta congruent-to 1.6 as a typical value of the index. However, a single-index power law is probably inadequate for polarization, even in a single line of sight; statistical examination of the local power-law index shows a degree of flattening toward longer wavelengths. We find that neither the excess polarization relative to the Serkowski/Wilking formula nor the value of the power-law index is correlated with changes in lambda(max). Hence, the infrared polarization does not respond to changes in the properties of the grains which give rise to dramatic variations in the wavelength dependence of polarization at shorter wavelengths, suggesting a degree of invariance (which we term "universality") in the form of the infrared polarization law. The infrared continuum polarization of molecular cloud sources (see work by Martin & Whittet) is quite similar to that of the reddened stars studied here, which supports the concept of universality. The power-law behavior of infrared polarization is notably similar to the invariant spectral dependence of extinction in the same wavelength interval which has been discussed previously by several authors. This similarity is not so definitive as to force the view that the grains responsible for the infrared extinction and polarization are one and the same. However, since there is some dispersion in both extinction and polarization about the mean infrared behavior, it might be possible to address this issue by investigating whether these deviations are correlated. The most robust conclusion relating to size is that major changes in the larger-sized portion of the population(s), say by mutual coagulation, are quite constrained. When it comes to achieving invariance in the infrared, in spite of evolution of some populations of grains, the fact that different populations generally produce distinct spectral dependences, which must sum to that observed for extinction or polarization, directly constrains changes in the mixture of populations, as well as the distinct populations themselves. No extant grain model reproduces the 3-5-mu-M polarization well, and it will be interesting to see whether model parameters can be adjusted to take into account these new observations. Of particular interest would be new calculations involving various proposed carbon-bearing grain components (graphite, amorphous carbon, HAC, and organic refractories) which along with silicates might contribute via absorption and scattering to the infrared polarization.