[CPMD-list] lattice constant of magnesium-oxide

[Axel Kohlmeyer]

On Mon, 14 Mar 2005 Nico.Zobel at TU-Berlin.DE wrote:

NZ> Dear CPMD-community,

dear nico,

[...]

NZ> lattice constant 			total energy
NZ> in angstroem				in eV
NZ> 
NZ> 4.14					    	-1092.00484447
NZ> 4.15    					-1092.01032496
NZ> 4.16    					-1091.62546596
NZ> 4.17   					-1092.01537806
NZ> 4.18    					-1092.01386046
NZ> 4.19    					-1092.01213459
NZ> 4.20    					-1092.00728452
NZ> 
NZ> The value for 4.16 Angstroem is somewhat strange and I don't know why. In
NZ> all cases I used the same input-file (except the value for the lattice
NZ> constant). The important parts of the input file is provided below.

it seems, that cpmd has converged to the wrong state. try converging
tighter, e.g., CONVERGENCE ORBITALS 1.0e-7...

NZ> I also encountered this penomenon when I calculated the lattice constant
NZ> of Li2O. I usually was able to get a better accordance of the results by
NZ> changing the cutoff-energy slightly (1 to 5 Ry).

...and try a higher real space density, e.g. by setting DUAL to 6.
the many default settings and most recommended parameter values
aim at a reasonably fast MD with not too much loss of accuracy.
for accurate single point calculation, you usually need to tighten
the convergence and use tighter grids. you should also consider using
GC-CUTOFF 1.0e-6 or even 5.0e-6 if you run into problems converging
the single point wavefunction.

NZ> But I'm not sure whether this procedure is really correct. That's why I
NZ> would like to know if this is a common behaviour of CPMD or if this is due
NZ> to unrealistic parameters of the USPP I use (please see appendix) ?

i have seen similar behavior with other codes as well for a
variety of reasons. some are listed above, others are:
insufficient k-point sampling, bugs in the code, or compounds
that are difficult to treat with DFT (e.g. actinides).

one thing you may want to try out with your pseudopotential
is to add 'partial core correction' or to include semi-core
states. see for example the sodium or aluminum pseudopotentials
in the uspp library or the existing norm-conserving mg 
pseudopotentials. 


best regards,
	axel.

NZ> 
NZ> Thank you very much for your time.
NZ> 
NZ> Nico Zobel.
NZ> 
NZ> P.S.:
NZ> I used both of the CPMD-patches provided by Axel Kohlmeyer in this
NZ> mailing-list on Feb 25th.
NZ> 
NZ> P.S. to Axel Kohlmeyer:
NZ> My Mg-USPP was CPMD-compatible, my Li-USPP was not.
NZ> 
NZ> ---------------------------input-file---------------------------------
NZ> 
NZ> &CPMD
NZ>   OPTIMIZE WAVEFUNCTION
NZ>   PCG MINIMIZE
NZ>   CONVERGENCE ORBITALS
NZ>     1.d-6
NZ> &END
NZ> 
NZ> &DFT
NZ>   FUNCTIONAL PBE
NZ> &END
NZ> 
NZ> &SYSTEM
NZ>   ANGSTROM
NZ>   SYMMETRY
NZ>     8
NZ>   CELL
NZ>     8.4 1.0 2.0  0  0  0
NZ>   CUTOFF
NZ>     40.0
NZ>   TESR
NZ>     4
NZ>   SCALE
NZ>   DUAL
NZ>     5.0
NZ> &END
NZ> 
NZ> &ATOMS
NZ> *012-Mg.uspp BINARY NEWF
NZ>   LMAX=D
NZ>    64
NZ> ....
NZ> *O_VDB_PBE.psp BINARY NEWF
NZ>   LMAX=D
NZ>    64
NZ> ....
NZ> &END
NZ> 
NZ> --------------------------mg-pseudopotential-------------------------
NZ> 
NZ>    ============================================================
NZ>    |  pseudopotential report: version  7.3.5 date  3-10-2005  |
NZ>    ------------------------------------------------------------
NZ>    |  mg (Vol06)           PBE - GGA           exchange-corr  |
NZ>    |  z =  12.00  zv =   2.00  exfact =   5.00000             |
NZ>    |                            etot  =  -1.65416             |
NZ>    |  index    orbital      occupation    energy              |
NZ>    |    1        300           2.00        -.35               |
NZ>    |  keyps = 3     ifpcor = 0                                |
NZ>    |  rinner =      1.43     for L=    1                      |
NZ>    |  rinner =      1.43     for L=    2                      |
NZ>    |  rinner =      1.43     for L=    3                      |
NZ>    |    new generation scheme:                                |
NZ>    |    nbeta =  4     kkbeta =  583     rcloc =    3.1000    |
NZ>    |    ibeta     l     epsilon   rcut iptype                 |
NZ>    |       1      0        .00   3.00      2                  |
NZ>    |       2      0       -.35   3.00      2                  |
NZ>    |       3      1        .00   3.00      2                  |
NZ>    |       4      1       -.35   3.00      2                  |
NZ>    |  npf    = 6  ptryc =   8.000                             |
NZ>    |  lloc   = 2  eloc   =    .000                            |
NZ>    |  ifqopt = 3  nqf    = 6  qtryc =   8.000                 |
NZ>    |  all electron calculation used koelling-harmon equation  |
NZ>    |         ************logarithmic mesh************         |
NZ>    ============================================================
NZ> 
NZ> 
NZ> 

-- 

=======================================================================
Dr. Axel Kohlmeyer   e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum  http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.