> Hello SIESTA users. > I want to optimize the lattice parameters of a tetragonal structure > (TiO2) i.e. a and c/a ratio. > I know SIESTA can optimize it, but I want to do it manually i.e. total > energy vs a and c/a. > I want to follow the following steps. > > 1. First optimized a for fixed c/a. i.e. I will change the value of a > and will keep c/a fixed and then I will plot a vs total energy and I > will get the optimized a. > 2. using the optimized a, I will vary the c/a ratio and will keep a > constant. From this I will get the optimized c/a, after piloting c/a vs > total energy. > Now I will have optimized a and c/a and I will compare this with the > experimental values. > Am I right??
In principle a and c/a are two independent variables, so that - taking the task seriously - you should search for a minimum of the total energy as function of both. What you describe is just two steps (i.e., 1-dim. minimum search) in two perpendicular directions, and there is no guarantee in will bring it to the true 2-dim. minimum. Recipes for a "naive" manual minimization: 1. repete the sequence till convergence, i.e., after optimizing c/a fix it and vary a again, etc.; 2. take an array of sampling points on a piece of 2-dim (a, c/a) plane bracketing the minimum, and make a 2-dim. polynomial (quadratic) fit from which you'll get the minimum. Do you have variable internal coordinates in your system? If your system is TiO2 then you do. Then you should include them in the process, i.e., your search will become multi-dimensional (a; c/a; u for rutile) Best regards Andrei Postnikov
