... wavelengths

Tables 1 to 5 are only available in electronic form at http://www.edpsciences.org

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...2005

A simplified version of this geometrical prescription has been applied by Misiriotis et al. (2004) to fit the FIR SEDs of bright IRAS galaxies.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... dust

These corrections are only valid for normal disk galaxies and are not applicable to systems with different star/dust geometries such as starburst or dwarf galaxies.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... aperture

We give the average surface brightness in an aperture (in addition to the surface brightness at the centre position of the fitted template) because in practice the brightness is blurred over a point spread function. Therefore it is useful to include a reference aperture which is larger than typical point spread functions, but smaller than the scalelength of the disk.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... inclination

Formally there is a very small dependence on wavelength due to the fact that the scalelength of the stars varies with wavelength in our model. In practice, however, this effect is negligible over the range of inclinations considered here and cannot be seen in the plot.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... inclination

Within the numerical noise induced by the finite numbers of directions for which scattered light is calculated.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... thin)

In our detailed SED modelling of individual galaxies given in Papers I and II we showed that this range in opacity is quite likely to characterise spiral galaxies.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... geometry

For example, depending on the assumed star-formation history, it may be necessary to invoke different star/dust geometries to explain the observed colours of distant disk galaxies (Pierini et al. 2005).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... measurements

The errors of a single scalelength determination were estimated to be $\sim$5% in Möllenhoff, $\sim$3% in Cunow and $\sim$5% in MacArthur et al. Thus the error of the quotients in Fig. 13 can be estimated to be $\sim$7%. A corresponding error bar is plotted there.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.