Good day!

During the last few weeks I tried to do some research on the magnitudes and dependecies of zero-point vibrational corrections on NMR shieldings. I have, however, hit a wall when I realized the following:

A calculation's results are dependent upon the orientation of the input molecule.

The predicted shielding, corrected by the ZPV can be altered by up to 3ppm for a small molecule just by changing its input orientation.

As the effects I am looking for (environmental dependency of ZPVC) are supposedly in the area of 1ppm, research seems to be impossible at this point in time.

See the following example: A simple amide, called 1a (HCONH2) is used. The level of theory is HF/6-31G**. Dalton2015 is used. The input scripts are copypasted from the Dalton2015 manual and only slightly altered.

- Geometric optimization of 1a.mol is performed using refgeo.dal
- Minimum structure is extracted as 1a-ref.mol
- Effective structures for 0K and 298K are calculated using effgeo.dal and extracted as 1a-eff000.mol and 1a-eff298.mol
- 3 different rotations are generated for 1a-eff298.mol, rotating by 45° around each of the axes.
- Shielding calculations are performed using the shield298.dal on the original and the three new orientations.
- The vibrationally corrected shielding is calculated by taking the "Vibrationally corrected" values of Bxmx, Bymy and Bzmz from the "Nuclear magnetic shielding constants (ppm)"-subsection of the output and dividing by three.

Code: Select all

```
shield298_1a-eff298.out: 186.40225970
shield298_1a-eff298_x45.out: 186.00186449
shield298_1a-eff298_y45.out: 186.36358421
shield298_1a-eff298_z45.out: 188.29636215
Maximum difference: 2.3 ppm
```

The same phenomenon appears when the rotation is applied to the initial structure or the reference structure.

What I played around with:

- All the threshholds and steplength
- .ACCURA (an undocumented five point derivation scheme that a colleague found while digging through the code
- different molecules
- sequential/parallel
- more stuff that I can't remember :p

To me it seems as though there are problems with the numerical approximations.

I also realized that the displacements (necessary for calculation of derivatives) are not performed along the normal modes but rather in x, y and z. Could this have an influnce?

Well, I'd be glad to get an answer, a reaction on this. Did I simply do something wrong? Underthunk something?

Thank your for reading! Have a nice day!

Best Regards,

Tim