There are two fundamental factors to be considered here:
The intrinsic spectrum depends on the star formation history and the metallicity evolution of a galaxy. We will not expand on this here since this is covered by the review talk of Leitherer . The propagation of photons through the ISM depend on:
The neutral gas affects the line emission component of SEDs; this topic is covered in detail in the talk of Kewley et al. . The optical properties of grains, which mean their absorption and scattering properties throughout the UV-submm range, depend on the composition and size distribution of the grains. This topic is covered in detail by the review talk of Dwek  (see also the talks by Li  and by Gordon ), and is only briefly touched upon here, in Sect. 4. The amount of dust and its distribution with respect to the stars has received little attention until recently, even though the distribution of dust is the single most important factor affecting the propagation of photons in galaxies. To illustrate this statement one only has to recall that a fixed amount of dust distributed on a scale of a few parsec around stars will have the same effect as 1000 times more dust distributed on kpc scales in the disk. Therefore this review will mainly focus on the effect of the relative geometries of stars and dust on the SEDs. Previous reviews related to this topic include: Calzetti , , Popescu & Tuffs , Kylafis & Misiriotis .
We will start with very simple geometries and increase the complexity until we have identify a minimum degree of complexity that can account for observed broad-band UV/optical/FIR/submm SEDs. We will also consider the effect of these models on the UV/optical/FIR/submm surface brightness distributions. In terms of the dust emission, we will place most emphasis on the FIR emission, rather than, for example, on the MIR emission. This is firstly because most of the energy absorbed by grains is re-radiated in the FIR, and secondly, because the FIR colours of a galaxy depend on the strength of the radiation fields in a galaxy, and therefore more directly constrain the propagation of photons in the disk.
Our entry point is to consider geometries with cylindrical symmetry, which are essential for the description of disk galaxies. Spherical symmetry is a more reasonable approximation for the description of dwarf galaxies (e.g. Galliano et al. ) and starburst galaxies (e.g. Witt et al. , Gordon et al. ). The simplest model is the infinite slab/sandwich. A sandwich model, not incorporating scattering, was used by Disney et al.  to investigate the attenuation-inclination behaviour of spiral galaxies. This work first emphasised the strong effect of the relative scaleheights of stars and dust on the attenuation. Another version of the sandwich model, this time including scattering, was used to calculate the energy balance between the emission and re-emission of light in the pioneering work of Xu & Buat . In the Xu & Buat formulation there is only one free parameter, the face-on optical depth, which is adequate to account for the energy balance. Apart from its application to the integrated emission from galaxies (Buat & Xu , Xu et al. ), this particular model was used by Xu & Helou  in the modelling of the large-scale dust heating and cooling in the diffuse medium of M 31. A common drawback, though, of models involving infinite slab/sandwich geometries is that they cannot predict the shape of the observed FIR/submm SEDs. Understanding the FIR colours is not only a matter of academic concern, but also provides a further dimension to the predictive power of models, since the FIR colours directly probe the strength of the radiation fields, and, as we shall see, strongly depend on physical quantities of interest, such as SFRs.
In order to also fit the FIR colours, one needs to consider more realistic geometries, where by realistic we mean incorporation of finite disks, bulges and small scale structures:
Finite disks are usually described by double exponentials in both radial and vertical direction. Bulges can be described by a variety of forms: de Vaucouleurs, truncated Hubble, spherical with King profile/exponential. We note here that the exact choice of bulge geometry has little effect on the shape of the globally integrated SED, provided that the bulk of the luminosity of the bulge is emitted within an area much smaller than the disk. Small scale structures have been described as small scale dust clouds/clumps (depending on terminology) which, according to the model, may or may not be physically associated with young stars. As we shall see, this different treatment has strong consequences on the prediction of the FIR colours. All or combinations of these geometrical components have been employed by the models introduced in the following papers: Kylafis & Bahcall , Bianchi et al. , Bianchi et al. , Xilouris et al. , Silva et al. , Granato et al. , Kuchinski et al. , Popescu et al. , Matthews & Wood . All these models were used to account for the optical SEDs, but only three of them (Bianchi et al. , Silva et al. , Popescu et al. ) were used as the basis for a self-consistent calculation of both the stellar and dust emission SEDs. Recently, self-consistent calculations of attenuation and re-emission by dust grains in galaxies have also been done in the work reported by Baes et al.  in this volume, and have started to be incorporated in population synthesis models, such as those of Piovan et al.  and Rocca-Volmerange , also as reported in this volume.