[SOLVED] Structural Optimisation and BFGS

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
e.ross1
Posts: 2
Joined: Tue Apr 27, 2010 5:14 am

[SOLVED] Structural Optimisation and BFGS

Post by e.ross1 » Tue Apr 27, 2010 5:27 am

Hello,

when using ionmov=3, optcell=1 (Broyden-Fletcher-Goldfarb-Shanno minimization) on Copper with the pseudopotential 29-Cu.LDA.fhi the following error occurs in multiple Broyden iterations

" findmin : WARNING -
The 2nd degree equation has no positive root (choice=4).

findmin : COMMENT -
Will continue, since the new total energy is lower
than the old. Take a larger step in the same direction.
line minimization, algorithm 4"

The code continues to run, but the length never converges oscillating between about 5 and 9 Bohr (around the experimental value ~6.6).

Using a different pseduopotential .pspnc, I did not come accross this error. Using Gold instead of copper I found that this warning would occur for both pseudopotentials but after a few iterations it would disappear and the cell volume would converge.

Any idea why?

I enlcose my full input

Code: Select all

occopt 4
tsmear 0.01

acell 3*6.6       
rprim  0.0  0.5  0.5   
       0.5  0.0  0.5
       0.5  0.5  0.0

optcell 1
ionmov  2
ntime  20
dilatmx 3
ecutsm .5

ntypat 1       
znucl 29         
                         
natom 1       
typat 1           
xred             
   0.0  0.0  0.0   

ecut  10.0

ngkpt 4 4 4   
nshiftk 4         
shiftk 0.5 0.5 0.5
       0.5 0.0 0.0
       0.0 0.5 0.0
       0.0 0.0 0.5

nstep 25     
toldfe 1.0d-6

(Note the parameters ecut, ngkpt and tsmear have not been converged; this is just an initial test)

Cheers.

P.S.
Actually using ionmov=2 seems to resolve the issue. However, I am still interested in why it occurs.
Last edited by e.ross1 on Thu Apr 29, 2010 5:28 am, edited 1 time in total.

e.ross1
Posts: 2
Joined: Tue Apr 27, 2010 5:14 am

Re: Structural Optimisation and BFGS

Post by e.ross1 » Thu Apr 29, 2010 5:28 am

It appears the main cause of the problem is that I was too far from the minimum; when I improved the convergence parameters there wasn't a problem.

Cheers.

mverstra
Posts: 655
Joined: Wed Aug 19, 2009 12:01 pm

Re: Structural Optimisation and BFGS [SOLVED]

Post by mverstra » Tue May 04, 2010 12:17 am

Indeed: the Broyden depends in some sense on being in the right basin of attraction. If you reach a region where the potential energy surface is convex the Broyden will explode your structure (well, very rarely).

Matthieu
Matthieu Verstraete
University of Liege, Belgium

yaser
Posts: 19
Joined: Wed Mar 10, 2010 6:03 am

Re: [SOLVED] Structural Optimisation and BFGS

Post by yaser » Wed May 12, 2010 5:47 pm

Hi Dear Matthieu,
I have faced with this error.
Give me more explaination about the solotion of it ,please.
Thank you.
Asadi.

Locked