|
We now proceed to an examination of issues directly pertinent to the other component of an interconnect scheme, the interlayer dielectric (ILD). The first requirement for an ILD is a low dielectric constant k, but in our general discussion in Sec. 2.2 we have listed other materials properties that are necessary for a low-k ILD to be viable. Let us mention two specific areas of concern: mechanical stability and thermal conductivity of the ILD. In the context of low-k ILD candidates to replace SiO2, one should keep in mind that polymers typically are much less strong than SiO2 and less adherent to themselves or to other materials. Furthermore, the thermal conductivity of polymers is typically a factor of 5 lower than that of SiO2, and the coefficient of thermal expansion (CTE) of polymers is typically an order of magnitude larger than that of Si. Analogous differences in materials properties would apply to silicate-based materials with a lower density than SiO2.
From our earlier discussion it is also clear that, ideally, the metal conductors should be surrounded completely by low-k ILD [Eqs. (4) and (5)]. If that cannot be realized, then substantial benefits can still be achieved if the spaces between high-aspect-ratio metal lines and vias are filled with low-k ILD.
A few possible architectures to implement these goals are illustrated schematically in Fig. 31, showing metal structures (two vias, a via with a metal line on top, and a metal line by itself) that may occur in different parts of the interconnect layer. In order to be specific, we take the layers shown to represent metal 1 and ILD 1, but the same principles apply to the upper levels of the interconnect scheme. Note that, for simplicity, barrier layers are not shown in Fig. 31. Also, the reader should review Figs. 5 and 6 for an abbreviated sequence of processing steps leading to the situations described in Fig. 31.
|
