Dear Cyrille, Nicola, and Serge Thanks for your helps. My main problem comes back to the computational method section in scholarly papers. In adsorption of molecules on surfaces, authors say that they cleave a surface from fully optimized (vc-relax) unit cell, enlarge it 2*2 or 3*3 times in u and v directions, and then define 15-20 angstrom vacuum in the z direction. Then all of them take 2-4 bottom layers fixed to represent bulk, and only relax (and not vc-relax) other atoms containing adsorbed molecule. If I do according to the published works, I get large stress, as it was discussed in previous long discussion. If I do according to Nicola's advice, the stress eventually reduces, but the bottom layers positions will not be the bulk ones. Which one is correct? Serge, maybe in the links you have provided this issue has been addressed. I will read them.
Regards David Foster Ph.D. Student of Chemistry -------------------------------------------- On Fri, 12/5/14, Serge Nakhmanson <s...@ims.uconn.edu> wrote: Subject: Re: [Pw_forum] Slab optimization with -24kbar stress!!! To: "PWSCF Forum" <pw_forum@pwscf.org> Date: Friday, December 5, 2014, 5:24 PM On 12/5/14 5:11 AM, David Foster wrote: 1- Cyrille, I fixed both in-plane and z-direction lattice constants. I need to fix the 15A vacuum in the z to prevent periodic interactions for studying adsorption. Dear David, Just my 2 cents to add to this. (a) Residual surface stress is a perfectly "legal" quantity. Some people hunt for it (as well as for the surface elastic constants) specifically with DFT. You can check it out here: http://dx.doi.org/10.1103/PhysRevB.71.094104 or here http://dx.doi.org/10.1016/j.cma.2010.09.007 It is a lot of fun! :) (b) You have to be *extremely careful* when attaching something (a molecule?) to your surface, as this will open a new can of worms. You will need to check if your combined system is likely to have a dipole moment pointing perpendicularly to the slab -- and it might. Then you may have to keep an xy mirror plane running through the middle of your slab [adding many more atoms to the calculation] or apply a dipole correction [makes calculation convergence much slower] to get rid of it. If you don't, you run a chance of having a spurious electric field in vacuum due to interaction between the polarized slab and is periodic images. There is an old paper that comments on the situation: http://dx.doi.org/10.1103/PhysRevB.63.205426 but much more had been published since then. The really graceful way of studying slabs involves using a code that can do 2D PBC in plane and break periodicity along the perpendicular direction. But then you have to abandon the PW basis -- i.e., codes like VASP and QE -- and use something else. HTH, Serge -- Serge M. Nakhmanson Dep-t of Materials Science & Engineering, and Institute of Materials Science University of Connecticut Storrs, CT 06269-3136 Phone: (860) 486-5252 http://satori.ims.uconn.edu/ -----Inline Attachment Follows----- _______________________________________________ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum _______________________________________________ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum