Strange energy curve

Message

yjwang · #1 Post by **yjwang** » Tue Aug 18, 2009 8:31 am

I don't know whether anybody encountered this strange energy curve.
The energy curve is the result of the shell command "grep F OSZICAR"
grep F OSZICAR
1 F= -.42529140E+03 E0= -.42525207E+03 d E =-.425291E+03 mag= 74.5744
2 F= -.44616080E+03 E0= -.44613709E+03 d E =-.208694E+02 mag= 69.9987
3 F= -.44616133E+03 E0= -.44613756E+03 d E =-.208699E+02 mag= 69.9997
4 F= -.44617504E+03 E0= -.44615141E+03 d E =-.208836E+02 mag= 69.9998
5 F= -.44620021E+03 E0= -.44617708E+03 d E =-.209088E+02 mag= 69.9996
6 F= -.44620957E+03 E0= -.44618588E+03 d E =-.209182E+02 mag= 69.9998
7 F= -.44621368E+03 E0= -.44618986E+03 d E =-.209223E+02 mag= 69.9998
8 F= -.44621582E+03 E0= -.44619199E+03 d E =-.209244E+02 mag= 69.9998
9 F= -.44621692E+03 E0= -.44619320E+03 d E =-.209255E+02 mag= 69.9998
10 F= -.44621759E+03 E0= -.44619394E+03 d E =-.209262E+02 mag= 69.9998
11 F= -.44621804E+03 E0= -.44619436E+03 d E =-.209266E+02 mag= 69.9998
12 F= -.44621852E+03 E0= -.44619483E+03 d E =-.209271E+02 mag= 69.9998
13 F= -.44621889E+03 E0= -.44619519E+03 d E =-.209275E+02 mag= 69.9998
14 F= -.44621924E+03 E0= -.44619551E+03 d E =-.209278E+02 mag= 69.9998
15 F= -.44621954E+03 E0= -.44619564E+03 d E =-.209281E+02 mag= 69.9998
16 F= -.44621992E+03 E0= -.44619611E+03 d E =-.209285E+02 mag= 69.9998
17 F= -.44622428E+03 E0= -.44620057E+03 d E =-.209329E+02 mag= 69.9998
18 F= -.44622484E+03 E0= -.44620105E+03 d E =-.209334E+02 mag= 69.9998
19 F= -.44622503E+03 E0= -.44620125E+03 d E =-.209336E+02 mag= 69.9998
20 F= -.44622529E+03 E0= -.44620162E+03 d E =-.209339E+02 mag= 69.9998
21 F= -.44622545E+03 E0= -.44620179E+03 d E =-.209340E+02 mag= 69.9998
22 F= -.44622582E+03 E0= -.44620206E+03 d E =-.209344E+02 mag= 69.9998
23 F= -.44622595E+03 E0= -.44620230E+03 d E =-.209345E+02 mag= 69.9998
24 F= -.44622618E+03 E0= -.44620252E+03 d E =-.209348E+02 mag= 69.9998
25 F= -.44622632E+03 E0= -.44620265E+03 d E =-.209349E+02 mag= 69.9998
26 F= -.44622664E+03 E0= -.44620295E+03 d E =-.209352E+02 mag= 69.9998
27 F= -.44622680E+03 E0= -.44620312E+03 d E =-.209354E+02 mag= 69.9998
28 F= -.44622706E+03 E0= -.44620338E+03 d E =-.209357E+02 mag= 69.9998
29 F= -.44622720E+03 E0= -.44620354E+03 d E =-.209358E+02 mag= 69.9998
30 F= -.44622741E+03 E0= -.44620374E+03 d E =-.209360E+02 mag= 69.9998
31 F= -.44622756E+03 E0= -.44620390E+03 d E =-.209362E+02 mag= 69.9998

forsdan · #2 Post by **forsdan** » Tue Aug 18, 2009 9:28 am

Please make it your habit to actually elaborate a bit more on an issue when you think something is strange. Which particular feature is off in your opinion and why?

Are you referring to the first point with respect to other ones and that there is a near 20 eV difference? Then I guess that either some position was off in your input and that the ionic relaxation algorithm managed to correct for this rather quickly, or that the electronic minimization in the first ionic step hit the NELM setting before it had time to converge fully. To check the latter please check the convergence in the first ionic step. To eliminate the first possiblity a geometry check might be in order.

Hope this helps

Cheers,
/Dan

<span class='smallblacktext'>[ Edited Tue Aug 18 2009, 06:29PM ]</span>

yjwang · #3 Post by **yjwang** » Wed Aug 19, 2009 7:49 am

Sorry! I should describe my question more elaborately.
The 20eV difference is not surprise to me. It is the case that the electronic minimization in the first ionic step did not converge fully. What confuses me is that after the second (trial) step there are so many correction steps. I don't know why.
<span class='smallblacktext'>[ Edited Wed Aug 19 2009, 09:51AM ]</span>

yjwang · #4 Post by **yjwang** » Mon Aug 24, 2009 12:20 am

Sorry! I should describe my question more elaborately.
The 20eV difference is not surprise to me. It is the case that the electronic minimization in the first ionic step did not converge fully. What confuses me is that after the second (trial) step there are so many correction steps. I don't know why.

[quote="forsdan"]Please make it your habit to actually elaborate a bit more on an issue when you think something is strange. Which particular feature is off in your opinion and why?

Are you referring to the first point with respect to other ones and that there is a near 20 eV difference? Then I guess that either some position was off in your input and that the ionic relaxation algorithm managed to correct for this rather quickly, or that the electronic minimization in the first ionic step hit the NELM setting before it had time to converge fully. To check the latter please check the convergence in the first ionic step. To eliminate the first possiblity a geometry check might be in order.

Hope this helps

Cheers,
/Dan

[/quote]

#5 Post by **admin** » Mon Oct 12, 2009 11:25 am

if for (any of the ) steps electronic convergence is not reached, the forces of that step are not reliable. Usually it is better to re-start the run with an increased number of electronic steps (NELM; default NELM=60)