Dear Graeme,
I'm using Dimer method to search TS, starting from an NEB calculation.
1. Although the force is converged and the curvature is negative, the angle remains high. I'm not sure what the problem is. Here is the DIMCAR file:
Step Force Torque Energy Curvature Angle
1 0.45114 11.19323 -1398.56124 5.48190 19.42364
1 0.45114 4.78785 -1398.56124 1.51078 10.65017
1 0.45114 3.42320 -1398.56124 0.78326 4.99346
1 0.45114 2.07389 -1398.56124 0.41652 3.85528
2 0.65643 1.93980 -1398.53142 0.49433 3.14839
2 0.65643 1.48646 -1398.53142 0.46152 1.85985
2 0.65643 1.13505 -1398.53142 0.40954 1.78550
2 0.65643 1.12482 -1398.53142 0.38578 1.49109
3 0.79237 1.54170 -1398.49193 0.26861 2.98963
3 0.79237 1.21828 -1398.49193 0.21251 1.70908
3 0.79237 0.98630 -1398.49193 0.18685 1.61089
4 0.86832 1.45484 -1398.44735 0.27758 2.95211
4 0.86832 1.08501 -1398.44735 0.22877 1.58099
4 0.86832 0.81023 -1398.44735 0.19323 1.29277
5 0.92620 1.68102 -1398.39967 0.03216 3.57675
5 0.92620 1.10158 -1398.39967 -0.07044 1.80656
5 0.92620 0.95377 -1398.39967 -0.06646 1.90014
6 1.24774 1.44619 -1398.41545 0.17177 1.83025
6 1.24774 0.85205 -1398.41545 0.12536 1.65629
7 1.22778 2.04710 -1398.36422 -0.10739 3.65833
7 1.22778 1.43017 -1398.36422 -0.23238 2.41708
7 1.22778 1.16204 -1398.36422 -0.31265 2.36629
7 1.22778 1.22648 -1398.36422 -0.34686 1.80974
8 0.70248 1.33814 -1398.37509 -0.44058 3.67025
8 0.70248 1.59295 -1398.37509 -0.50387 2.08741
8 0.70248 1.16259 -1398.37509 -0.56195 2.42798
8 0.70248 1.36925 -1398.37509 -0.56950 1.78283
9 1.54707 2.12078 -1398.37501 0.13852 3.48254
9 1.54707 1.66253 -1398.37501 0.05952 2.33920
9 1.54707 1.35159 -1398.37501 -0.00405 1.93227
9 1.54707 1.27630 -1398.37501 -0.01894 1.84871
10 1.17015 1.76483 -1398.36700 0.25052 2.97592
10 1.17015 1.45964 -1398.36700 0.20486 2.09150
10 1.17015 1.21691 -1398.36700 0.14847 1.64902
10 1.17015 1.04072 -1398.36700 0.13423 1.58242
11 1.18746 2.75617 -1398.31128 -0.37915 7.21541
11 1.18746 2.25165 -1398.31128 -0.75861 4.02204
11 1.18746 2.13399 -1398.31128 -0.87397 3.78472
11 1.18746 2.10040 -1398.31128 -0.98024 3.27382
12 0.91145 2.24781 -1398.33159 -1.13529 3.80671
12 0.91145 2.11557 -1398.33159 -1.25602 3.26825
12 0.91145 1.91138 -1398.33159 -1.35754 3.05214
12 0.91145 1.85417 -1398.33159 -1.42591 2.83282
13 0.70858 2.09604 -1398.34526 -1.80094 3.77529
13 0.70858 2.15228 -1398.34526 -1.92852 2.97960
13 0.70858 1.83702 -1398.34526 -2.07103 3.17697
13 0.70858 1.99599 -1398.34526 -2.14796 2.56768
14 0.58288 1.90248 -1398.35109 -2.11271 3.59644
14 0.58288 1.91743 -1398.35109 -2.21093 2.49061
14 0.58288 1.42762 -1398.35109 -2.29544 2.45340
14 0.58288 1.46088 -1398.35109 -2.35162 2.06511
15 0.48707 1.39760 -1398.35373 -2.45347 2.30144
15 0.48707 1.32838 -1398.35373 -2.51432 1.73954
15 0.48707 1.06743 -1398.35373 -2.57693 1.80120
15 0.48707 1.21055 -1398.35373 -2.58372 1.58791
16 0.43537 1.25482 -1398.35789 -2.56329 2.38431
16 0.43537 1.12863 -1398.35789 -2.59006 1.53111
16 0.43537 0.88830 -1398.35789 -2.64503 1.46520
17 0.31612 1.21444 -1398.36121 -2.48818 2.76324
17 0.31612 1.18815 -1398.36121 -2.50164 1.47953
17 0.31612 0.85189 -1398.36121 -2.52218 1.27482
18 0.30097 1.29938 -1398.36501 -2.73967 1.95878
18 0.30097 1.09061 -1398.36501 -2.78841 1.35930
18 0.30097 0.84672 -1398.36501 -2.81800 1.19374
19 0.33368 1.13503 -1398.36772 -2.48925 1.54438
19 0.33368 0.76274 -1398.36772 -2.49629 1.04696
20 0.31894 0.84380 -1398.37095 -2.53315 1.50909
21 0.32667 0.69634 -1398.37340 -2.62499 1.26267
22 0.34397 0.88502 -1398.37554 -2.58892 1.15199
23 0.30485 0.78630 -1398.37705 -2.54748 1.17704
24 0.23053 1.02537 -1398.37825 -2.48673 1.37917
24 0.23053 0.70709 -1398.37825 -2.50365 1.03112
25 0.23005 0.95815 -1398.37933 -2.60836 1.65419
26 0.23703 0.90785 -1398.38043 -2.62848 1.38944
27 0.20624 1.05322 -1398.38183 -2.57490 1.67736
27 0.20624 0.92232 -1398.38183 -2.58337 1.04248
28 0.21345 0.96904 -1398.38322 -2.70548 1.75291
29 0.19751 1.13416 -1398.38433 -2.53832 1.65268
29 0.19751 0.94398 -1398.38433 -2.55196 1.13769
30 0.19591 0.85044 -1398.38544 -2.59786 1.94910
31 0.17928 0.86720 -1398.38649 -2.70646 1.11163
32 0.15810 0.86793 -1398.38705 -2.42043 1.73660
33 0.13999 0.95967 -1398.38749 -2.66873 1.53949
34 0.14248 0.86666 -1398.38789 -2.50064 1.14696
35 0.12134 1.15507 -1398.38830 -2.39386 1.48295
35 0.12134 0.68546 -1398.38830 -2.38212 1.14780
36 0.12508 0.76793 -1398.38861 -2.44780 1.63544
37 0.10903 1.04904 -1398.38901 -2.36953 1.46019
37 0.10903 0.76140 -1398.38901 -2.43781 0.80662
38 0.10833 0.78474 -1398.38941 -2.56973 1.49178
39 0.11785 0.91634 -1398.38979 -2.42822 1.07332
40 0.12026 0.64350 -1398.39010 -2.46941 1.04753
41 0.11837 0.79424 -1398.39046 -2.34490 1.12659
42 0.13503 0.80650 -1398.39078 -2.45405 1.35260
43 0.12494 1.08862 -1398.39107 -2.58760 1.45344
43 0.12494 0.74630 -1398.39107 -2.61734 1.09925
44 0.10486 0.83252 -1398.39137 -2.60481 2.03775
45 0.09056 0.94665 -1398.39165 -2.49684 1.13570
46 0.08056 0.54506 -1398.39183 -2.43180 0.82003
47 0.07792 0.71322 -1398.39199 -2.54122 1.13136
48 0.07085 0.84059 -1398.39212 -2.46294 0.83558
49 0.06105 0.50912 -1398.39222 -2.42515 1.22853
50 0.05943 0.89873 -1398.39231 -2.37467 2.03590
51 0.05483 0.97034 -1398.39238 -2.48634 1.17470
52 0.04998 0.65463 -1398.39244 -2.39793 0.82764
53 0.04422 0.56342 -1398.39249 -2.43659 1.39070
54 0.03732 0.82046 -1398.39253 -2.48678 0.91642
55 0.03548 0.72584 -1398.39257 -2.47746 1.45306
56 0.03787 0.85847 -1398.39261 -2.40884 1.13555
57 0.03613 0.73042 -1398.39264 -2.36347 1.29412
58 0.03626 0.81899 -1398.39268 -2.35415 1.20436
59 0.03373 0.64676 -1398.39271 -2.31397 0.75804
60 0.03138 0.53452 -1398.39273 -2.37543 0.89830
61 0.02689 0.51127 -1398.39276 -2.34677 0.71652
62 0.02778 0.66537 -1398.39278 -2.37801 0.73711
63 0.02753 0.55214 -1398.39280 -2.37914 1.19052
64 0.02812 0.65795 -1398.39281 -2.39861 0.89210
65 0.02544 0.52703 -1398.39283 -2.39822 0.67673
66 0.02389 0.49058 -1398.39285 -2.39986 1.19510
67 0.02395 0.62535 -1398.39286 -2.47152 0.98492
68 0.02238 0.81567 -1398.39287 -2.50818 0.77664
69 0.02176 0.45314 -1398.39289 -2.50757 0.91963
70 0.02111 0.42534 -1398.39290 -2.46693 0.54307
71 0.02075 0.43255 -1398.39291 -2.45636 0.75348
72 0.02013 0.69498 -1398.39292 -2.48938 1.38313
73 0.02115 0.93933 -1398.39293 -2.51415 1.12696
74 0.02073 0.55634 -1398.39295 -2.39111 0.80593
75 0.02173 0.63299 -1398.39296 -2.30273 1.55951
76 0.02176 0.84482 -1398.39297 -2.39120 1.50319
77 0.02021 0.65473 -1398.39298 -2.48756 1.15834
78 0.01886 0.75952 -1398.39299 -2.40926 1.18743
79 0.02073 0.59438 -1398.39300 -2.37535 0.71174
80 0.01932 0.56736 -1398.39301 -2.41589 0.88766
81 0.01791 0.71952 -1398.39301 -2.34578 1.30160
82 0.01908 0.77951 -1398.39302 -2.32344 0.91346
83 0.01754 --- -1398.39303 --- ---
2. I set up dimer run from a NEB run by using neb2dim.pl. In some cases, the neb calculation can not give the good MEP after several hundreds steps, for example, the energy and force of images in the neb is as below,
0 0.02817000 1398.48542200 0.00000000
1 0.02903400 1398.48618900 0.00076700
2 0.03529400 1398.69457500 0.20915300
3 0.20550800 1398.78036500 0.29494300
4 1.09499500 1398.63408500 0.14866300
5 0.75971100 1398.93578200 0.45036000
6 3.71795400 1399.25563900 0.77021700
7 1.56527800 1397.30267500 -1.18274700
It seems that the image 5 is the highest point, may I use this image as the starting of the dimer run and generate MODECAR between image 5 and 6?
3. In some cases, the initial guess looks reasonable, however, the TS in dimer run goes crazy (or become another structure), should I change some parameters in input file to avoid it?
4. Last question, After getting TS, how to link this TS to the real initial state and final state?
Thanks a lot.
Problems in Dimer calculations
Moderator: moderators
Re: Problems in Dimer calculations
1) A small rotation of 1-2 degrees is quite reasonable and not something to worry about.
2) Actually, isn't image 6 the highest energy? The neb2dim.pl should interpolate to find the highest energy geometry along the band and calculate the tangent between the neighboring images. Of course, you are welcome to do this manually, if you wish. Also, I notice that we have some formatting problems of the numbers -- the negative sign on the energy values are being cut off. I've updated the printing in svn.
3) If your guess is far from a saddle point a dimer search may take a long time to converge and it may converge to a saddle which is not near to your initial guess. If you are suspicious that there really is a nearby saddle, you can increase the maximum number of rotations per translation so that the negative mode is well converged. Also, reducing the MaxMove value for the optimizer will make sure that small translational steps are taken towards the saddle. If you do this and (for example) the dimer finds a positive mode during the search, it is most likely that you are not starting near a saddle.
4) To find the minima connecting a saddle, you can make small displacements along the negative mode at the saddle and minimize these two geometries. The script dimmins.pl can be used to create these initial images to minimize.
2) Actually, isn't image 6 the highest energy? The neb2dim.pl should interpolate to find the highest energy geometry along the band and calculate the tangent between the neighboring images. Of course, you are welcome to do this manually, if you wish. Also, I notice that we have some formatting problems of the numbers -- the negative sign on the energy values are being cut off. I've updated the printing in svn.
3) If your guess is far from a saddle point a dimer search may take a long time to converge and it may converge to a saddle which is not near to your initial guess. If you are suspicious that there really is a nearby saddle, you can increase the maximum number of rotations per translation so that the negative mode is well converged. Also, reducing the MaxMove value for the optimizer will make sure that small translational steps are taken towards the saddle. If you do this and (for example) the dimer finds a positive mode during the search, it is most likely that you are not starting near a saddle.
4) To find the minima connecting a saddle, you can make small displacements along the negative mode at the saddle and minimize these two geometries. The script dimmins.pl can be used to create these initial images to minimize.
Re: Problems in Dimer calculations
Dear Graeme,
Thank you for your prompt reply.
I still have queries for Q3:
I'm searching TS with initial guess from NEB calculation. When the forces in the neb caluculation are low enough (as below), I use the image 5 as intial guess and carried out dimer calculation.
0 0.04431500 1398.46409800 0.00000000
1 0.08946200 1398.48875900 0.02466100
2 0.07267100 1398.75528700 0.29118900
3 0.07275100 1398.53586100 0.07176300
4 0.10557800 1397.51326200 -0.95083600 (image 5)
5 0.25335700 1397.90002600 -0.56407200
6 0.18457900 1397.80146500 -0.66263300
7 0.16486000 1397.82825300 -0.63584500
8 0.08915000 1397.87303800 -0.59106000
9 0.07324900 1397.91091500 -0.55318300
10 0.06226000 1397.94085300 -0.52324500
11 0.06152500 1397.96698400 -0.49711400
However, I checked the energy and structure during dimer calculation and found the energy goes down to -1399 something, which is more stable than both initial state and final state in NEB cal. That means the guess sturcture does not go towards saddle point, but converge to another local minimum. The corresponding structure indicated that the C-C bond is broken (which is expected) accompanying with C-H bond broken (which is unexpected).
In this case, if I keep image 5 as intial guess, only increase DrotMax and decrease DNFMax, does it help the convergence?
Or I need run more steps for neb calcualtion, then start dimer cal with a new intial guess?
Many thanks
Thank you for your prompt reply.
I still have queries for Q3:
I'm searching TS with initial guess from NEB calculation. When the forces in the neb caluculation are low enough (as below), I use the image 5 as intial guess and carried out dimer calculation.
0 0.04431500 1398.46409800 0.00000000
1 0.08946200 1398.48875900 0.02466100
2 0.07267100 1398.75528700 0.29118900
3 0.07275100 1398.53586100 0.07176300
4 0.10557800 1397.51326200 -0.95083600 (image 5)
5 0.25335700 1397.90002600 -0.56407200
6 0.18457900 1397.80146500 -0.66263300
7 0.16486000 1397.82825300 -0.63584500
8 0.08915000 1397.87303800 -0.59106000
9 0.07324900 1397.91091500 -0.55318300
10 0.06226000 1397.94085300 -0.52324500
11 0.06152500 1397.96698400 -0.49711400
However, I checked the energy and structure during dimer calculation and found the energy goes down to -1399 something, which is more stable than both initial state and final state in NEB cal. That means the guess sturcture does not go towards saddle point, but converge to another local minimum. The corresponding structure indicated that the C-C bond is broken (which is expected) accompanying with C-H bond broken (which is unexpected).
In this case, if I keep image 5 as intial guess, only increase DrotMax and decrease DNFMax, does it help the convergence?
Or I need run more steps for neb calcualtion, then start dimer cal with a new intial guess?
Many thanks
Re: Problems in Dimer calculations
If you have found an intermediate minimum, the best thing that you can do is to break up your NEB into two bands; initial -> intermediate and intermediate -> final. Then you can focus on the two saddles, or the highest saddle, and use the dimer to find them accurately.