PBE0 cannot be used in vasp5.2?

Message

yuanxun · #1 Post by **yuanxun** » Mon Aug 24, 2009 6:45 am

DEAR ALL:
why the label "LHFCALC" can not be used in vasp5.2?
The calculation seems to get into endless loop after five interaction without any output any more.
Thanks!

alex · #2 Post by **alex** » Sun Aug 30, 2009 9:57 am

The first 5 iterations are DFT only, then exact exchange is calculated. This is _very_ time consuming. Try a small system.

Cheers

Alex

VASP001 · #3 Post by **VASP001** » Sat Jan 09, 2010 6:28 pm

[quote="yuanxun"]DEAR ALL:
why the label "LHFCALC" can not be used in vasp5.2?
The calculation seems to get into endless loop after five interaction without any output any more.
Thanks! [/quote]
You can check the file of OUTCAR ,to find the LHFCALC=.TRUE. or .FALSE.

Danny · #4 Post by **Danny** » Thu Mar 04, 2010 1:45 pm

the first 5 iterations are the normal speed (as alex mentioned) after this it becomes very slow, we are talking about orders of magnitude:
A small example of a single atom fcc system with 5x5x5 k-points and HSE06 calculation...there are only 10 kpoints in the IBZand the the steps become 100x slower than the first 5:

LOOP: cpu time 8.38: real time 8.38
LOOP: cpu time 2.46: real time 2.46
LOOP: cpu time 2.13: real time 2.14
LOOP: cpu time 2.08: real time 2.08
LOOP: cpu time 3.09: real time 3.09
LOOP: cpu time 311.79: real time 312.30
LOOP: cpu time 288.99: real time 289.54
LOOP: cpu time 285.76: real time 286.52
LOOP: cpu time 284.03: real time 284.25
LOOP: cpu time 282.32: real time 282.33
LOOP: cpu time 285.06: real time 285.34
LOOP: cpu time 281.36: real time 281.36
LOOP: cpu time 279.80: real time 279.81
LOOP: cpu time 283.62: real time 283.74
LOOP: cpu time 288.30: real time 289.55
LOOP: cpu time 287.12: real time 287.44
LOOP: cpu time 285.60: real time 286.15
LOOP: cpu time 286.33: real time 286.51
LOOP: cpu time 286.84: real time 287.50

Cheers Danny

#5 Post by **admin** » Thu Mar 04, 2010 4:30 pm

it is the exact exchange part which makes the calculation slow down.
whenever possible, use a k-mesh that cen be reduced for the exact-exchange contributions so you can make use of NKRED

Danny · #6 Post by **Danny** » Fri Mar 05, 2010 8:57 am

What are the properties of a k-mesh that can be reduced?
eg: k-point mesh Nx*Ny*Nz
with Nx a integer multiple of NKREDX, Ny of NKREDY and Nz of NKREDZ?

Danny

Sonny · #7 Post by **Sonny** » Wed Oct 13, 2010 7:38 am

I dont really undrestand this manual entry for NKRED.
http://cms.mpi.univie.ac.at/vasp/vasp/W ... plied.html
I cant get this to work at all with various entries. =>

k-point set generating an fcc lattice -> 4 k-points in BZ
0
diret
0.5 0.5 0.0
0.0 0.5 0.5
0.5 0.0 0.5
0 0 0

Is it intended that ISEMAR=-5 is used with this set??
Do i need to get the Wavecar first?
Is it suitable for a loopscript?

<span class='smallblacktext'>[ Edited Wed Oct 20 2010, 02:59PM ]</span>