# Is Orca wrongly breaking bonds? Why can it only optimize the molecule with COPT activated?

I would like to ask for help in using Orca (v. 4.0.0) for conducting a simple optimization.

First I have this structure (primarily optimized using MMFF94s force field on Avogadro):

Thus I tried to conduct a simple optimization by using Orca and just as a test I used the following input:

! PAL8 PBE def2-SVP def2/JK RIJONX TightOpt TightSCF CPCM(DMSO) KeepDens NormalPrint Grid4 FinalGrid5
%method
Z_MaxIter 150
end
%scf
MaxIter 150
end

*xyz 0 1
C       4.026085     -1.212051      0.000031
C      -3.313454      0.000008     -0.000131
C       5.413748     -1.205853     -0.000365
C      -4.026091      1.212051     -0.000333
C       6.111884      0.000013     -0.000433
C      -5.413752      1.205847     -0.000315
C       5.413733      1.205871     -0.000018
C      -6.111885     -0.000021     -0.000102
C       4.026071      1.212051      0.000379
C      -5.413728     -1.205875      0.000032
C       3.313452     -0.000005      0.000350
C      -4.026066     -1.212050      0.000016
C       1.893686     -0.000008      0.000523
C       0.678508      0.000000      0.000397
C      -0.678520      0.000009      0.000187
C      -1.893672      0.000016     -0.000043
H       3.483003     -2.148319      0.000143
H       5.951276     -2.145874     -0.000768
H      -3.483012      2.148322     -0.000572
H       7.194627      0.000021     -0.001168
H      -5.951286      2.145866     -0.000334
H       5.951249      2.145899     -0.000149
H      -7.194627     -0.000032      0.000311
H       3.482975      2.148312      0.000762
H      -5.951240     -2.145905      0.000285
H      -3.482967     -2.148309      0.000048


Thus after running it, I saw that somehow Orca was "splitting" all bonds and getting this:

However, my job keeps going without detecting any problem and it just ends because hits the maximum allowed iterations.

I also tried to change the basis set, but without success. Also, I have played a lot with functionals, changed between LDA and GGA ones without success, semiempirical also have not shown any good result.

The only thing I could use to solve this problem was by setting COPT instead of OPT only, which changes Orca to cartesian coordinate (but was I already using cartesian coordinates?).

I just wonder why that is happening. Am I missing something? What is wrong with my input to cause Orca to "split" all atoms apart? And why Orca only can optimize this molecule with COPT activated?

• Does this also happen if you remove CPCM(DMSO)? Jul 13 '17 at 21:30
• About copt: Your input coordinates are transformed to some internal coordinates, which are then used for the optimization. copt is about controlling these internal coordinates and thus independent of the input. Jul 13 '17 at 21:45
• Hi Feodoran, yes. That also happens if I remove CPCM(DMSO). I also thought that could be the reason, but removing that did cause the same result. Jul 13 '17 at 21:54
• If you inspect the output you can see a lot of "RMS(Cart) increases - taking linear step, building new B-matrix and trying again" messages that ends up with a cleat indication that something went wrong: "Not converged!" Since OP mentioned that switching to optimization in Cartesian coordinates solves the issue, it looks like a bug with optimization in internal coordinates. Jul 13 '17 at 22:49
• @Feodoran, COPT in ORCA means that optimization is done directly in Cartesian coordinates without even constructing internal ones. Jul 13 '17 at 22:52

It seems to be a bug with optimization in internal coordinates. As I already mentioned in comments, ORCA couldn't converge at geometry relaxation step when optimization is done in internal coordinates. Switching to optimization in Cartesian coordinates (coordsys cartesian) or old internal ones (coordsys redundant_old) in the %geom section of the input, or using the ! COPT simple input as OP has done, solves the issue.