Annu. Rev. Astron. Astrophys. 1992. 30: 653-703 Copyright © 1993 by Annual Reviews. All rights reserved

4.2 Confusion by Galactic Synchrotron Emission

The group at the University of Durham has carried out a detailed study of the implications of Galactic synchrotron emission for microwave background radiation observations (Lawson et al 1987, Banday et al 1991, Banday & Wolfendale 1991). Similar studies have been carried out independently by Lasenby et al (1991) and us (in preparation). All of these studies indicate clearly that at frequencies below ~ 15 GHz fluctuations in the diffuse Galactic synchrotron emission are expected at the level of the anisotropy detected by Davies et al (1987), and that this anisotropy could be mainly due to the Galaxy and not to fluctuations in the microwave background radiation. We discuss here the published work by the Durham group. The methods used by the three separate groups were very similar and were based on the same low frequency radio surveys.

The main problem in this area is that there are no accurately calibrated extensive sky surveys at frequencies above 1420 MHz. The dependance of the intensity of the Galactic synchrotron emission, I()_Gse upon frequency may be approximated by a power law over a fairly wide range in frequency - i.e. I()_Gse = I(₀)_Gse x ( / ₀), where lies between -0.5 and -1.1. It is well-known that the power law steepens at high frequencies. The microwave background radiation, on the other hand, is known to follow a Planck black-body curve to within the precision of the COBE measurements (Mather et al 1990). The ratio of the intensities of these two emission processes is, therefore, a strong function of frequency:

(10)

The drop in intensity of the Galactic synchrotron emission with increasing frequency makes it far more difficult to measure accurately the Galactic synchrotron emission at high frequencies than at low frequencies. For this reason, and because of the more complex instrumentation required, the 19 GHz balloon surveys discussed above are not sufficiently sensitive to detect Galactic emission away from the plane. To estimate the Galactic synchrotron emission at 10 GHz and above, one therefore has to rely upon extrapolation from low frequencies and this introduces significant uncertainties in the predicted levels of synchrotron radiation.

Lawson et al (1987) used many of the published maps of large regions of the sky at frequencies between 38 MHz and 1420 MHz to determine the variations in spectral index of the Galactic synchrotron emission over the sky and over this range of frequencies. Their detailed analyses and spectral index maps indicate that the best surveys for the purpose of estimating the Galactic synchrotron emission as a function of frequency are those at 408 MHz by Haslam and his co-workers (Haslam et al 1970, 1974, 1981); and at 1420 MHz by Reich (1982) and by Reich and Reich (1986).

Banday et al (1991) and Banday and Wolfendale (1991) have used these observations in an attempt to predict the expected level of Galactic synchrotron emission at the frequency of the Davies et al (1987) observations and with the same experimental setup as used by these authors. They carried out a number of simulations which reproduce fluctuations at the same level as those observed by Davies et al, but do not reproduce the observed fluctuations in detail. The predicted level of anisotropy due to the Galactic synchrotron emission in a double-differencing experiment is shown in Figure 6, together with the signal at 10.46 GHz measured by Davies et al (1987).

Figure 6. Levels predicted by Banday et al (1991) of T/T due to Galactic synchrotron emission as a function of observing frequency, for double-switched observations. The circle indicates the anisotropy detected by Davies et al (1987) at 10.4 GHz.

Lasenby et al (1991) have carried out similar simulations, with the same results. An example of one particular simulation by Lasenby (private communication) is shown in Figure 5. Also shown is a running Gaussian mean of the Davies et al data. It is clear that Galactic synchrotron emission can account for fluctuations of the level observed by Davies et al, but that there is not good agreement in detail. The lack of good agreement in detail is not surprising since it is well known that the index varies with frequency, and, with an extrapolation over such a wide range in frequency as this, small variations in , coupled with uncertainties in the absolute levels of the low frequency observations and other systematic effects, could easily account for the differences between observations and predictions.

This result by Davies et al (1987) was the first convincing case in which the sensitivity of the observations was limited by diffuse Galactic synchrotron emission and it exemplifies the difficulties that now face observers in this field. In order to account accurately for the effects of foreground fluctuations, these must be not simply detected but also measured with greater accuracy than the level at which intrinsic fluctuations are being studied if they are to be removed with the requisite precision. This raises the question of whether it is reasonable to expect to be able to subtract out foreground effects with the required accuracy.

The strong frequency dependance of the specific intensity of the Galactic synchrotron emission compared with that of the microwave background radiation, given by the relation in Equation 10, means that multi-frequency observations can, in principle, be used to subtract out the effects of the Galactic synchrotron emission.

Over the last two years we have been involved in a detailed study of the contamination caused by the Galactic synchrotron emission (in preparation). The 408 MHz (Haslam et al 1970, 1974, 1981, Lawson et al 1987) and the 1420 MHz (Reich 1982, Reich & Reich 1986) maps were kindly supplied to us by Banday. The resolution on these maps was 1°. We scaled the maps at these two frequencies by the same factor, such that the extrapolation to ~ 30 GHz in the Galactic plane is consistent with the COBE results (Wright et al 1991). The scaled maps then give the correct intensity and the level of fluctuations at 26-40 GHz. We then selected 10 x 10° regions of sky and searched for regions of this size in which the fluctuations were a minimum. Over the whole sky we found ten 10 x 10° patches for which the rms variations due to synchrotron and dust emission were less than 9 µK.

Clearly, one can have no confidence about the details of the synchrotron emission after extrapolating over this large range in frequency, but this does give a reasonable estimate of the intensity and fluctuations in intensity in this frequency range. We also added in our best estimate of the contribution from the infrared cirrus (from IRAS data scaled to fit the COBE results) which is considerably lower in intensity than the synchrotron component at these frequencies, and we included the microwave background radiation contribution with the addition of a randomly varying component. We then simulated the extraction of the microwave background radiation signal from this ``fake sky'' data assuming different levels of random noise. The results are shown in Figure 7. As the instrumental noise per channel goes to zero the error in extrapolation of the microwave background radiation fluctuations approaches the level of fluctuations in the dust model we have used (i.e. ~ 1.5 µK). We see that, for example, fluctuations in the microwave background radiation can be extracted with an accuracy of 1 µK if the noise in each of the seven channels is 2 µK. Given only data in this frequency range, it is not possible with achievable sensitivities, to subtract out the contribution due to dust.

Figure 7. Simulated extraction of image of the microwave background radiation at 26-40 GHz, by observations in 7 frequency channels. Galactic synchrotron emission and thermal emission from dust have been included, but only the former is subtracted out in the simulation. The limiting error, as the noise per channel -> 0, of ~ 1.5 µK is due to dust. Clearly the microwave background emisson can be extracted with accuracy T/T ~ 10-6 if the noise per channel is < 2 µK, provided that free-free emission is not a serious contaminant at this level.