Recommendations
This page contains a summary of the methods we recommend for good performance and cost when calculating chemical shifts and coupling constants. If you are experienced in running these sorts of calculations, you may simply refer to the information below for each method. For further information and help running calculations and processing data, please see the instructions page. For a complete list of empirical scaling factors, see the scaling factors page.
Contents:
Recommendations from Tantillo, et al. for 1H and 13C computed chemical shifts
Recommendations from Bally & Rablen for 1H computed chemical shifts
Recommendations from Bally & Rablen for 1H-1H computed coupling constants
Recommendations from Benassi for 1H and 13C computed chemical shifts
Recommendations from Tantillo, et al. for 1H and 13C computed chemical shifts:
The following information applies to calculations run using G03G03a or G09.G09a
Solvent modeling: We strongly recommend utilizing an implicit solvent model in the NMR single-point portion of your calculations. Doing so consistently provides significant gains in accuracy compared to gas-phase NMR calculations and does so at negligible cost in terms of effort and CPU time.
High accuracy methods: Among the scaling factors we've determined, those from the mPW1PW91 and PBE0 functionals paired with the 6-311+G(2d,p) basis set for NMR single-point calculations generally provided the lowest RMSD values for both nuclei. Both of these functionals perform well, with no significant difference between them. There is also not a significant difference between the two basis sets used for the geometry optimizations (6-31+G(d,p) and 6-311+G(2d,p) with the B3LYP or M062X functionals). Depending on the computational resources available, these methods may likely be quite affordable and we recommend their use where possible.
G03 Methods |
Scaling Factors |
Performance (RMSD = root mean square deviation (ppm)) |
|||||
Geometry |
NMR |
1H |
13C |
1H Test Set |
1H Probe Set |
13C Test Set |
13C Probe Set |
B3LYP/6-31+G(d,p) |
mPW1PW91/6-311+G(2d,p) |
slope: -1.0823 |
slope: -1.0448 |
RMSD: 0.1128 |
RMSD: 0.1446 |
RMSD: 1.8038 |
RMSD: 2.4528 |
B3LYP/6-311+G(2d,p) |
mPW1PW91/6-311+G(2d,p) |
slope: -1.0821 |
slope: -1.0365 |
RMSD: 0.1146 |
RMSD: 0.1455 |
RMSD: 1.6818 |
RMSD: 2.6074 |
B3LYP/6-31+G(d,p) |
PBE0/6-311+G(2d,p) |
slope: -1.0844 |
slope: -1.0447 |
RMSD: 0.1159 |
RMSD: 0.1476 |
RMSD: 1.8206 |
RMSD: 2.4497 |
B3LYP/6-311+G(2d,p) |
PBE0/6-311+G(2d,p) |
slope: -1.0844 |
slope: -1.0365 |
RMSD: 0.1180 |
RMSD: 0.1485 |
RMSD: 1.7008 |
RMSD: 2.5991 |
G09 Methods |
Scaling Factors |
Performance (RMSD = root mean square deviation (ppm)) |
|||||
B3LYP/6-31+G(d,p) |
mPW1PW91/6-311+G(2d,p) |
slope: -1.0936 |
slope: -1.0533 |
RMSD: 0.1180 |
RMSD: 0.1610 |
RMSD: 2.0561 |
RMSD: 2.4900 |
B3LYP/6-311+G(2d,p) |
mPW1PW91/6-311+G(2d,p) |
slope: -1.0933 |
slope: -1.0449 |
RMSD: 0.1169 |
RMSD: 0.1597 |
RMSD: 1.9114 |
RMSD: 2.5949 |
B3LYP/6-31+G(d,p) |
PBE0/6-311+G(2d,p) |
slope: -1.0958 |
slope: -1.0533 |
RMSD: 0.1204 |
RMSD: 0.1640 |
RMSD: 2.0768 |
RMSD: 2.4913 |
B3LYP/6-311+G(2d,p) |
PBE0/6-311+G(2d,p) |
slope: -1.0956 |
slope: -1.0450 |
RMSD: 0.1196 |
RMSD: 0.1628 |
RMSD: 1.9347 |
RMSD: 2.5905 |
#M062X/6-31+G(d,p) |
mPW1PW91/6-311+G(2d,p) |
slope:
-1.0938 |
slope:
-1.0446 |
RMSD:
0.1233 |
RMSD: 0.1604 |
RMSD:
1.9544 |
RMSD: 2.4674 |
#M062X/6-311+G(2d,p) |
mPW1PW91/6-311+G(2d,p) |
slope:
-1.0951 |
slope:
-1.0379 |
RMSD:
0.1227 |
RMSD: 0.1639 |
RMSD:
1.8311 |
RMSD: 2.3399 |
Good accuracy at low cost: For a lower-cost method that still provides very good results, we recommend the B3LYP/6-31+G(d,p)//B3LYP/6-31G(d) level of theory (including implicit solvent in the NMR single-point calculation).
Geometry |
NMR |
1H |
13C |
1H Test Set |
1H Probe Set |
13C Test Set |
13C Probe Set |
B3LYP/6-31G(d) |
B3LYP/6-31+G(d,p) |
slope: -1.0472 |
slope: -0.9600 |
RMSD: 0.1190 |
RMSD: 0.1398 |
RMSD: 2.2640 |
RMSD: 2.8937 |
Specific for 1H NMR: For proton chemical shifts, the WP04/aug-cc-pVDZ//B3LYP/6-31+G(d,p) method (which was specifically-parameterized to reproduce proton chemical shifts in chloroform solvent) performs exceptionally well.
Geometry |
NMR |
1H |
13C |
1H Test Set |
1H Probe Set |
13C Test Set |
13C Probe Set |
B3LYP/6-31+G(d,p) |
WP04/aug-cc-pVDZ |
slope: -1.0410 |
-- |
RMSD: 0.0959 |
RMSD: 0.1137 |
-- |
-- |
*G03 keyword is nmr=giao (default NMR method) or nmr=csgt as indicated.
SCRF for G03 methods refers to CPCM implicit solvent model with chloroform and uaks radii.
G03 keyword: scrf=(solvent=chcl3,cpcm,read) with the specifications: radii=uaks and nosymcav read in at the end of the file.
SCRF for G09 methods refers to smd implicit solvent model with chloroform.
G09 keyword: scrf=(solvent=chloroform,smd).
WP04 is invoked using an iop statement and the BLYP functional in G03.
G03 keywords: BLYP/BasisSet and iop(3/76=1000001189,3/77=0961409999,3/78=0000109999).
#int=ultrafine was included in all calculations involving M06 functionals.
Recommendations from Bally & Rablen for 1H computed chemical shifts:
Good accuracy at low cost:
G03 Methods |
Scaling Factors |
Performance (RMSD = root mean square deviation (ppm)) |
|
Geometry |
NMR |
1H |
1H Test Set |
B3LYP/6-31G(d) |
WP04/6-31G(d,p) |
slope: -1.0332 |
RMSD: 0.119 |
Slightly better accuracy at somewhat greater cost:
Geometry |
NMR |
1H |
1H Test Set |
B3LYP/6-31G(d) |
WP04/cc-pVDZ |
slope: -1.0205 |
RMSD: 0.115 |
Money (CPU time) is no object...:
Geometry |
NMR |
1H |
1H Test Set |
B3LYP/6-31G(d) |
WP04/aug-cc-pVDZ |
slope: -1.0544 |
RMSD: 0.103 |
If you don't want to use the WP04 functional:
Geometry |
NMR |
1H |
1H Test Set |
B3LYP/6-31G(d) |
B3LYP/6-31G(d,p) |
slope: -1.0552 |
RMSD: 0.129 |
If you don't want to include a simulated solvent: (This can be useful because sometimes SCRF causes convergence problems)
Geometry |
NMR |
1H |
1H Test Set |
B3LYP/6-31G(d) |
WP04/6-31++G(d,p) |
slope: -1.0140 |
RMSD: 0.129 |
If you don't want to use WP04 or SCRF:
Geometry |
NMR |
1H |
1H Test Set |
B3LYP/6-31G(d) |
B3LYP/aug-cc-pVDZ |
slope: -1.0554 |
RMSD: 0.133 |
B3LYP/6-31G(d) |
B3LYP/6-31++G(d,p) |
slope: -1.0407 |
RMSD: 0.153 |
*G03 keyword is nmr=giao (default NMR method) or nmr=csgt as indicated.
SCRF refers to PCM implicit solvent model with chloroform.
G03 keyword: scrf(solvent=chloroform)
WP04 is invoked using an iop statement and the BLYP functional in G03.
G03 keywords: BLYP/BasisSet and iop(3/76=1000001189,3/77=0961409999,3/78=0000109999).
Recommendations from Bally & Rablen for 1H-1H computed coupling constants:
The simplest recommended procedure below for computing proton-proton coupling constants with gas-phase calculations has been shown to produce results with an RMS error of 0.5 Hz for a large set of organic molecules and at a relatively affordable computational cost.Rab11a
- Optimize the geometry with B3LYP/6-31G(d)
- Run an NMR single-point calculation with the following route section in
GAUSSIANG03a,G09a
#n B3LYP/6-31G(d,p) nmr=(fconly,readatoms) iop(3/10=1100000)
At the end of the molecule specification (separated by a blank line) read in: atoms=H
A sample input file for chloroethane can be viewed by clicking here. - From the resulting log file, extract the desired Fermi contact J values, and scale them by a factor of 0.9117
Note that the iop statement in the above route section is equivalent to the NMR=mixed command in GAUSSIAN. However, this command does not inherently work with the fconly specification for the NMR calculation. Using fconly allows for calculation of just the Fermi contact terms, which are the dominant factor for proton-proton coupling constants (the other factors are spin-dipole and spin-orbit terms). As described in our paper,Rab11a this procedure not only significantly reduces the computational cost (by a factor of approximately 0.5), but the results obtained this way are in fact superior to those obtained by consideration of all terms.
It is possible to further reduce the computational cost for this type of calculation by restricting the basis set enhancement (invoked by iop statement above) to only the hydrogen atoms. However, this requires manual specification of the modified basis set directly in the input file. Detailed instructions for this procedure can be found on the instructions page.
A recent parameterization approach for computing 1H-1H coupling constants with dramatically reduced computational expense has been reported.Kut14a
Recommendations from Benassi for 1H and 13C computed chemical shifts:
When paired with a triple-ζ basis set, the WP04 DFT functional represents a good compromise for both 1H and 13C NMR predictions. mPW1LYP performs well for 1H chemical shifts, while APF performs well for 13C chemical shifts.