The variational approach to linear scaling is based on (i) the reformulation of the original constrained variational principle into an unconstrained one in terms of the density matrix directly or in terms of localised orbitals and (ii) the truncation of the density matrix or localised orbitals in physical space. Because of the truncation, the number of variables is proportional to the size of the system and the numerical minimisaton can be achieved with O(N) efforts. The linear scaling algorithm of Ordejón is presented as described in the original papers [Ordejón et al., 1993] [Ordejón et al., 1995] [Ordejón et al., 1996]. An application to a biological system is discussed [Lewis et al., 1997].