Dear Mic: I don't quite understand how you put it, and your attached .ppt does not make it simple, either. To my opinion it is easier - sorry for advertizing one my old publication not related to TiO2, but it contains several figures which explain the search of minimum, by hand, of a function of two variables. The paper is PRB 48, 5910 (1993), see Figs.3 and 7. Which values of a and c/a, and how many, to choose - is fully up to you. You should bracket a minimum, first (by trial and error).
> Then I can also optimize u for TiO2. "Then" won't work because, as you change the "u" value, your otimized a and c/a might be gone. If you are after high precision, you should optiize all three parameters simultaneously... Note that searching for a multi-dimensional minimum by a sequence of orthogonal steps (optimize a; optimize c/a; optimize u; optimize a again; optimize c/a, etc.) is VERY inefficient (as you can see from the abovementioned figures). Consult "Numerical Recipes". Good luck, Andrei Postnikov > What I understood is that. > I will optimize c/a for a series of calculations i.e. I will > optimized c/a for a1,a2,a3,a4,.. this will give me an optimal value.this > will consist of about 16 calculations > Then I will optimize a for ,(c/a)*0.98, (c/a)*0.99 ,(c/a)*1.0, > (c/a)*1.01, ,(c/a)*1.02 > This series of calculations will give me an optimal a. > So now I have Final optimized c/a and a. > Then I can also optimize u for TiO2. > Please see the attach ppt file. > Mic
