Dear all,
I was trying to optimize geometry of two hydrogen-bonded complexes, thiourea - 2-acetylthiazole and thiourea - 2-acetylpyrimidine, and I got three different errors; in first case, after first iteration all bonds are elongated ca. 2 times, then optimization fails because “Step is far too large” error occurs. (http://pastebin.com/YbABm99h) In second case, optimization is not aborted, but psi4 is continuously performing backsteps (http://pastebin.com/0YKFL35d). Sometimes torsion value calculation fails (http://pastebin.com/efSLLqNs). I have tried updating to 1.1a2, but this doesn’t solve these problems. By any means, I couldn’t obtain optimized geometries of these complexes. Any help will be appreciated.
Actually, I didn’t check my last attempt last night. I did a MM opt of the initial input structure in Avogadro, then started the Psi4 opt from there. It didn’t totally converge, but forces down considerably.
Input:
memory 11 gb
set basis aug-cc-pVDZ
molecule {
0 1
C -7.41666 -0.96465 0.61701
C -7.59032 -2.35556 0.66745
S -6.06660 -2.99683 0.71804
C -5.22732 -1.57243 0.67776
N -6.10503 -0.53859 0.62338
H -8.24430 -0.26945 0.57586
H -8.52901 -2.89236 0.67195
C -3.76285 -1.41061 0.69583
O -3.27957 -0.28989 0.66132
C -2.85850 -2.60460 0.75616
H -1.79307 -2.28920 0.76111
H -3.03385 -3.25189 -0.12837
H -3.06303 -3.18214 1.68173
N -5.82761 3.05785 -0.02506
C -4.68512 3.88915 -0.15860
N -3.38712 3.34109 0.01419
S -4.84243 5.28592 -0.46857
H -6.77999 3.43488 -0.14657
H -5.71364 2.05966 0.19718
H -3.27460 2.34220 0.23629
H -2.54365 3.92737 -0.07853
}
optimize('b3lyp-d3')
Optimization trace:
>>> grep '~' input2.out
--------------------------------------------------------------------------------------------- ~
Step Total Energy Delta E MAX Force RMS Force MAX Disp RMS Disp ~
--------------------------------------------------------------------------------------------- ~
Convergence Criteria 1.00e-06 * 3.00e-04 * o 1.20e-03 * o ~
--------------------------------------------------------------------------------------------- ~
1 -1269.97090326 -1.27e+03 2.96e-01 3.42e-02 o 2.78e-01 4.46e-02 o ~
2 -1270.04233162 -7.14e-02 6.92e-02 9.58e-03 o 1.46e-01 2.65e-02 o ~
3 -1270.05422298 -1.19e-02 3.03e-02 3.93e-03 o 1.53e-01 2.21e-02 o ~
4 -1270.05691474 -2.69e-03 1.07e-02 2.00e-03 o 2.55e-01 3.78e-02 o ~
5 -1270.05525907 1.66e-03 2.87e-02 5.22e-03 o 7.24e-02 1.79e-02 o ~
6 -1270.05767451 -2.42e-03 1.42e-02 2.15e-03 o 1.68e-01 2.60e-02 o ~
7 -1270.05757073 1.04e-04 6.71e-03 1.58e-03 o 2.22e-01 4.05e-02 o ~
8 -1270.05590037 1.67e-03 3.02e-02 5.38e-03 o 9.58e-02 2.03e-02 o ~
9 -1270.05669972 -7.99e-04 2.16e-02 3.53e-03 o 5.08e-02 1.31e-02 o ~
10 -1270.05844960 -1.75e-03 5.79e-03 9.09e-04 o 4.62e-02 1.31e-02 o ~
11 -1270.05875091 -3.01e-04 3.02e-03 6.99e-04 o 6.91e-02 1.61e-02 o ~
12 -1270.05868273 6.82e-05 2.60e-03 6.34e-04 o 4.95e-02 1.36e-02 o ~
13 -1270.05881995 -1.37e-04 1.30e-03 3.44e-04 o 6.24e-02 1.02e-02 o ~
14 -1270.05886665 -4.67e-05 9.37e-04 2.30e-04 o 3.73e-02 5.48e-03 o ~
15 -1270.05888446 -1.78e-05 5.04e-04 1.10e-04 o 2.19e-02 3.34e-03 o ~
16 -1270.05889107 -6.61e-06 3.14e-04 6.86e-05 o 5.81e-02 9.68e-03 o ~
17 -1270.05890147 -1.04e-05 4.36e-04 1.25e-04 o 6.20e-02 1.01e-02 o ~
18 -1270.05891103 -9.56e-06 5.08e-04 1.47e-04 o 9.34e-02 1.54e-02 o ~
19 -1270.05892239 -1.14e-05 6.64e-04 1.39e-04 o 6.28e-02 1.05e-02 o ~
20 -1270.05892775 -5.36e-06 4.33e-04 8.99e-05 o 7.62e-03 2.04e-03 o ~
21 -1270.05892954 -1.79e-06 2.57e-04 * 3.74e-05 o 7.54e-03 1.40e-03 o ~
Last Geom (in Bohr):
C -4.2317024449 -2.3215137394 0.2522322179
C -4.5730181965 -4.8725139019 0.6186802265
S -1.6811457557 -6.3592833416 0.8575676747
C -0.1920319875 -3.4072792131 0.4439779730
N -1.7814829732 -1.5269652086 0.1588552076
H -5.7533117524 -0.9556737942 0.0414611869
H -6.3336551109 -5.9194899989 0.7555646104
C 2.6188752077 -3.1028641271 0.4239560715
O 3.5463697210 -1.0141881221 0.1362706564
C 4.2254984920 -5.4381311959 0.7671157379
H 6.2197212809 -4.8998565193 0.7053008516
H 3.8263824010 -6.8211729422 -0.7321772768
H 3.8024715572 -6.3495055532 2.5866235666
N -1.3772208600 4.2540636152 -0.5755469096
C 0.6706685217 5.7852985542 -0.7662869441
N 2.9402696024 4.6101904476 -0.6171991595
S 0.4000422914 8.9533652980 -1.1639183357
H -3.0854933810 5.0836467274 -0.6881944720
H -1.2495982903 2.3481804711 -0.3486363051
H 3.1087952886 2.7171451574 -0.3685275806
H 4.5002104909 5.6911028386 -0.7442859774
With loosely bound structures like this, a lot depends on the initial structure. In the first case, a bond angle contained within a torsion was going linear (or zero) I think, causing the torsion to become undefined.
From the initial structure tried (http://pastebin.com/YbABm99h), I was able to make progress by adding:
set optking {
dynamic_level = 1
}
With this turned on, when an error such as an undefined dihedral is thrown, the optimization algorithm will become more conservative and continue. I got RMS forces ~10^-5 before getting impatient and quitting. In any event, it was going downhill. If you follow the output closely, you can learn what techniques are being tried, e.g.,
At level 3: Red. Int. + XYZ, RFO, backsteps, smaller trust.
Means redundant internals + cartesians, with backsteps as necessary and a smaller trust radius. Instead of using the ‘dynamic_level’ one can control all these directly. (I’m a fan of ‘opt_coordinates = BOTH’ for hard cases).
This dynamic_level will not necessarily be very efficient (getting a good geometry like Lori did is much better for that) but it is intended to help you make progress toward a minimum for tough cases. Eventually, if needed to continue downhill, it executes steepest-descent in cartesian coordinates.
I also like to try, when struggling with an optimization:
intrafrag_step_limit = 0.1
interfrag_step_limit = 0.1
though a smaller initial step may not help you much for these particular problems.
Thanks for advice, i’ll try again at earliest opportunity.