[gmx-users] energ minimization: cg and I-bfgs

maggin maggin.chen at gmail.com
Fri Jun 7 08:06:40 CEST 2013


Hi,

protein have 2 positive chage, I use PRODRG built .gro of CH3C00H, and add
two molecular CH3C00H, four molecular NA into system.

At energy minimization step, after steep, when I use cg, Fatal error:
DD cell 1 4 0 could only obtain 14 of the 16 atoms that are connected via
constraints from the neighboring cells. This probably means your constraint
lengths are too long compared to the domain decomposition cell size.
Decrease the number of domain decomposition grid cells or lincs-order.

I search the ways to sovle it in the GMX forum, there are some similar error
exist, I also try as the answers suggest, but I can not fix it. I don't
known what the cause.

things should go on....I want to do enery minimization more better in order
to avoid protein crash during simulation.

So, I try another way to do energ minimization. After steep, I use I-bfgs

steep .mdp:
title               =  bovin
cpp                 =  /usr/bin/cpp
define              =  -DFLEXIBLE
constraints         =  all-bonds
integrator          =  cg
dt                  =  0.002    ; ps !
nsteps              =  1000
nstlist             =  10 
ns_type             =  grid
rlist               =  1.0
coulombtype         =  PME
rcoulomb            =  1.0
vdwtype             =  cut-off
rvdw                =  1.4
fourierspacing                =  0.12
fourier_nx                =  0
fourier_ny                =  0
fourier_nz                =  0
pme_order                =  4
ewald_rtol                =  1e-5
optimize_fft                =  yes
emtol               =  100.0
emstep              =  0.01
; GENERATE VELOCITIES FOR STARTUP RUN
gen-vel                  = no
gen-temp                 = 293
gen-seed                 = 173529

process:
Steepest Descents:
   Tolerance (Fmax)   =  1.00000e+03
   Number of steps    =         2000
Step=    0, Dmax= 1.0e-02 nm, Epot= -3.49891e+05 Fmax= 2.19267e+04, atom=
21003
Step=    1, Dmax= 1.0e-02 nm, Epot= -3.66773e+05 Fmax= 1.02238e+04, atom=
3054
Step=    2, Dmax= 1.2e-02 nm, Epot= -3.86198e+05 Fmax= 5.97878e+03, atom=
1146
Step=    3, Dmax= 1.4e-02 nm, Epot= -4.03308e+05 Fmax= 3.61163e+03, atom=
1146
Step=    4, Dmax= 1.7e-02 nm, Epot= -4.19782e+05 Fmax= 1.80657e+03, atom=
29071
Step=    5, Dmax= 2.1e-02 nm, Epot= -4.38870e+05 Fmax= 1.43308e+03, atom=
29071
Step=    6, Dmax= 2.5e-02 nm, Epot= -4.53627e+05 Fmax= 3.63104e+03, atom=
923
Step=    7, Dmax= 3.0e-02 nm, Epot= -4.58169e+05 Fmax= 4.16424e+03, atom=
923

Step 8, time 0.016 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 0.004145, max 0.106308 (between atoms 923 and 924)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
    921    923   32.4    0.1385   0.1445      0.1340
    923    927   33.0    0.1382   0.1444      0.1340
    923    924   39.9    0.1386   0.1482      0.1340
    924    925   34.2    0.1014   0.1068      0.1000
Step=    8, Dmax= 3.6e-02 nm, Epot= -4.60414e+05 Fmax= 5.82027e+03, atom= 31

Step 9, time 0.018 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 0.008160, max 0.128456 (between atoms 34 and 38)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
     38     40   38.9    0.1410   0.1558      0.1390
     34     38   39.9    0.1412   0.1569      0.1390
     31     34   39.4    0.1414   0.1559      0.1390
     31     32   38.6    0.1413   0.1545      0.1390
     32     36   38.3    0.1412   0.1551      0.1390
     36     40   38.2    0.1409   0.1552      0.1390
   1020   1021   30.0    0.1397   0.1475      0.1390
   1106   1110   32.9    0.1400   0.1499      0.1390
   1103   1106   32.8    0.1401   0.1492      0.1390
   1103   1104   32.1    0.1401   0.1488      0.1390
   1104   1108   32.7    0.1400   0.1497      0.1390
   1110   1112   32.4    0.1400   0.1496      0.1390
   1108   1112   32.8    0.1400   0.1499      0.1390

Step 9, time 0.018 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 0.003367, max 0.128456 (between atoms 34 and 38)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
     34     38   39.9    0.1412   0.1569      0.1390
     38     40   38.9    0.1410   0.1558      0.1390
     36     40   38.2    0.1409   0.1552      0.1390
     32     36   38.3    0.1412   0.1551      0.1390
     31     32   38.6    0.1413   0.1545      0.1390
     31     34   39.4    0.1414   0.1559      0.1390

Step 9, time 0.018 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 0.008437, max 0.134299 (between atoms 373 and 377)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
    368    371   39.6    0.1451   0.1559      0.1390
    368    369   39.7    0.1452   0.1565      0.1390
    369    373   41.3    0.1446   0.1576      0.1390
    371    375   40.6    0.1446   0.1566      0.1390
    373    377   42.3    0.1433   0.1577      0.1390
    375    377   41.7    0.1432   0.1572      0.1390
    921    923   31.5    0.1445   0.1442      0.1340
    923    927   31.4    0.1444   0.1447      0.1340
    923    924   42.0    0.1482   0.1496      0.1340
    924    926   36.9    0.1065   0.1080      0.1000
    924    925   37.3    0.1068   0.1082      0.1000

Step 9, time 0.018 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 0.004632, max 0.116116 (between atoms 923 and 924)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
    921    923   31.5    0.1445   0.1442      0.1340
    923    927   31.4    0.1444   0.1447      0.1340
    923    924   42.0    0.1482   0.1496      0.1340
    924    926   36.9    0.1065   0.1080      0.1000
    924    925   37.3    0.1068   0.1082      0.1000
Step=   10, Dmax= 2.1e-02 nm, Epot= -4.63786e+05 Fmax= 3.33273e+03, atom=
1112
Step=   12, Dmax= 1.3e-02 nm, Epot= -4.65368e+05 Fmax= 3.16597e+03, atom=
518
Step=   13, Dmax= 1.5e-02 nm, Epot= -4.66743e+05 Fmax= 4.18187e+03, atom=
1112
Step=   14, Dmax= 1.9e-02 nm, Epot= -4.68028e+05 Fmax= 4.61798e+03, atom=
518
Step=   15, Dmax= 2.2e-02 nm, Epot= -4.68799e+05 Fmax= 5.19261e+03, atom=
522
Step=   16, Dmax= 2.7e-02 nm, Epot= -4.68896e+05 Fmax= 5.79535e+03, atom=
518
Step=   18, Dmax= 1.6e-02 nm, Epot= -4.72864e+05 Fmax= 1.66650e+03, atom=
522
Step=   19, Dmax= 1.9e-02 nm, Epot= -4.72901e+05 Fmax= 5.83070e+03, atom=
584
Step=   20, Dmax= 2.3e-02 nm, Epot= -4.75581e+05 Fmax= 4.25066e+03, atom=
160
Step=   22, Dmax= 1.4e-02 nm, Epot= -4.77224e+05 Fmax= 3.08540e+03, atom=
160
Step=   24, Dmax= 8.3e-03 nm, Epot= -4.78336e+05 Fmax= 1.29199e+03, atom=
160
Step=   25, Dmax= 1.0e-02 nm, Epot= -4.79314e+05 Fmax= 3.73890e+03, atom=
160
Step=   26, Dmax= 1.2e-02 nm, Epot= -4.80459e+05 Fmax= 2.21770e+03, atom=
160
Step=   27, Dmax= 1.4e-02 nm, Epot= -4.80600e+05 Fmax= 4.68351e+03, atom=
160
Step=   28, Dmax= 1.7e-02 nm, Epot= -4.81834e+05 Fmax= 3.60914e+03, atom=
160
Step=   30, Dmax= 1.0e-02 nm, Epot= -4.82982e+05 Fmax= 1.80313e+03, atom=
160
Step=   31, Dmax= 1.2e-02 nm, Epot= -4.83315e+05 Fmax= 4.15172e+03, atom=
160
Step=   32, Dmax= 1.5e-02 nm, Epot= -4.84337e+05 Fmax= 3.19667e+03, atom=
160
Step=   34, Dmax= 8.9e-03 nm, Epot= -4.85244e+05 Fmax= 1.44565e+03, atom=
160
Step=   35, Dmax= 1.1e-02 nm, Epot= -4.85711e+05 Fmax= 3.89936e+03, atom=
160
Step=   36, Dmax= 1.3e-02 nm, Epot= -4.86652e+05 Fmax= 2.48984e+03, atom=
160
Step=   38, Dmax= 7.7e-03 nm, Epot= -4.87292e+05 Fmax= 1.57431e+03, atom=
160
Step=   39, Dmax= 9.3e-03 nm, Epot= -4.87783e+05 Fmax= 3.04757e+03, atom=
160
Step=   40, Dmax= 1.1e-02 nm, Epot= -4.88394e+05 Fmax= 2.62054e+03, atom=
160
Step=   41, Dmax= 1.3e-02 nm, Epot= -4.88589e+05 Fmax= 3.92708e+03, atom=
160
Step=   42, Dmax= 1.6e-02 nm, Epot= -4.89062e+05 Fmax= 3.94266e+03, atom=
160
Step=   44, Dmax= 9.6e-03 nm, Epot= -4.90116e+05 Fmax= 9.30801e+02, atom=
160

writing lowest energy coordinates.

Back Off! I just backed up em.gro to ./#em.gro.1#

result:

Steepest Descents converged to Fmax < 1000 in 45 steps
Potential Energy  = -4.9011588e+05
Maximum force     =  9.3080109e+02 on atom 160
Norm of force     =  7.3772133e+01

then I do I-bfgs,

Low-Memory BFGS Minimizer converged to machine precision in 175 steps,
but did not reach the requested Fmax < 100.
Potential Energy  = -5.38089209528220e+05
Maximum force     =  1.94381568185624e+06 on atom 10842
Norm of force     =  1.61130387384634e+04

I chang emstep=0.005,  result also not chang.

the .mdp is
title               =  bovin
cpp                 =  /usr/bin/cpp ; the c pre-processor
define              =  -DFLEXIBLE
constraints         =  all-bonds  
integrator          =  l-bfgs
dt                  =  0.002	; ps !
nsteps              =  2000
nstlist             =  1
ns_type             =  grid
rlist               =  1.0
coulombtype         =  PME
rcoulomb            =  1.0
vdwtype             =  cut-off
rvdw                =  1.4
fourierspacing        = 0.12
fourier_nx		=  0
fourier_ny		=  0
fourier_nz		=  0
pme_order		=  4
ewald_rtol		=  1e-5
optimize_fft		=  yes
;
;        Energy minimizing stuff
;
emtol               =  100.0
emstep              =0.005
; GENERATE VELOCITIES FOR STARTUP RUN
gen-vel                  = no
gen-temp                 = 293
gen-seed                 = 173529

So, I have the question that why for cg, it take place fatal error? and
another question for steep, if the result is reasonable? 

Thank you very much!

maggin

 



--
View this message in context: http://gromacs.5086.x6.nabble.com/energ-minimization-cg-and-I-bfgs-tp5008882.html
Sent from the GROMACS Users Forum mailing list archive at Nabble.com.



More information about the gromacs.org_gmx-users mailing list