Due to periodic boundary conditions, you have 2 interfaces (repeat your cell along the growth direction and you will see immediately), that may be equivalent or not according to their details (crystalline structure, crystallographic direction, composition of the interface itself...).
Coming to your question: If the in-plane lattice constant is fixed, you should optimize c (='whole z lattice constant'). A good starting point is determined by calculating the z lattice constant separately of bulk1 and bulk2 constrained on that in-plane lattice constant , using the macroscopic theory of elasticity and using the average at the interface region. Then, you calculate the corresponding stress and if the z-component is non negligible you will slightly modify c (rescaling the internal atomic position; check forces on the atoms; if they increase too much, you will need also to re-relax them) and recalculate the stress. The optimal c could be found by linear interpolation. You could look at a review article and references therein: M. Peressi, N. Binggeli, and A. Baldereschi, Band engineering at interfaces: theory and numerical experiments. Phys. D: Appl. Phys. 31, 1273 (1998). Maria > > Dear all PWSCF users, > > I am trying to do some relaxation calc. of interfaces with PWSCF, but > these's a question wrt relaxations: > > Suppose we have a supercell without vacuum layers and only one > interface (i.f.), which might look like this: > > ------- > | | > |bulk1| > | | > | i.f.| > | | > |bulk2| > | | > ------ > There are strong inplane constraints, so we only relax distances between > atomic layers at the interface. So here comes my problem: if we use > "relax", the atomic layers at the interface will be relaxed, but since the > whole z lattice constant won't change, the structure is supposed to be not > fully relaxed. Is this right? > > Otherwise, if we use "vcrelax", how could one set up inplane constraints? > > Thanks a lot. > > Bests > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum >
