NMR predictor

The structure is read as a graph and every carbon and proton is placed in an environment. Each shift starts from a parent value and adds increments for the groups around it.

• Saturated carbons: Grant–Paul α, β, γ and δ increments with steric corrections, plus substituent terms for heteroatoms and functional groups. Ring carbons start from the cycloalkane and add terms for ring heteroatoms, carbonyls and double bonds.
• Aromatic rings: benzene, or a reference heteroaromatic (pyridine, furan, thiophene, pyrrole, azoles, azines, naphthalene, indole, quinoline and others), plus ipso, ortho, meta and para increments for each substituent.
• Alkenes: ethylene plus α and α′ increments for ¹³C, and the Pascual–Meier–Simon gem, cis and trans increments for ¹H. Cis or trans comes from the double-bond geometry in the SMILES.
• Carbonyls and nitriles: a base value for each class (ketone, aldehyde, acid, ester, amide, anhydride, urea, carbamate, carbonate) corrected for the groups on either side and for ring size.
• CH₃, CH₂ and CH protons: Shoolery-type additivity of α- and β-substituent terms, damped when several electronegative groups sit on one carbon.
• OH, NH and SH are shown as broad signals at a typical position with the range they are usually found in. They depend on concentration, water and temperature, and they disappear in CD₃OD and D₂O.

Chemically equivalent atoms are found from the symmetry of the molecular graph. Multiplicities are first order, from typical coupling constants: 7 Hz for vicinal aliphatic protons, 7.7 and 1.6 Hz for ortho and meta aromatic protons, about 10.8 Hz cis and 16.8 Hz trans across a double bond, and couplings to ¹⁹F. Where two coupled signals are closer than six times their coupling constant the multiplet is labelled m, because the real pattern is second order.

On 40 reference compounds measured in CDCl₃ the mean absolute error is 0.04 ppm for ¹H and 1.0 ppm for ¹³C. Those compounds include the parents the tables are built from, so the figure flatters the method. On 11 substituted compounds the tables were not read from (ibuprofen, aspirin, vanillin, cinnamaldehyde and others) it is 0.07 ppm for ¹H and 1.2 ppm for ¹³C.

For an unfamiliar molecule expect ±0.2 ppm for ¹H and ±3 ppm for ¹³C, and more where several functional groups crowd one another. Use the prediction to decide which signal is which, not to prove a structure.

• Strained three- and four-membered rings, bridged and caged skeletons.
• Heavy atoms and metals, and charged species such as salts and zwitterions.
• Tautomers (amide–iminol, keto–enol, azole NH) and fused heterocycles without a reference ring, such as purines: caffeine is off by 9 to 11 ppm at C-4, C-5 and C-6.
• Conformational effects: diastereotopic CH₂ protons next to a stereocentre appear as one signal here, and axial and equatorial ring protons are averaged.
• Crowded substitution: ortho-disubstituted benzenes and polyfunctional carbons deviate from additivity.
• Solvent: the increments are for CDCl₃. Other solvents move only the exchangeable protons here, although aromatic solvents and DMSO shift other signals by up to 0.3 ppm.

• E. Pretsch, P. Bühlmann, M. Badertscher, Structure Determination of Organic Compounds, 4th ed., Springer, 2009: aliphatic, aromatic, alkene and carbonyl ¹³C increments and aromatic ¹H increments.
• R. M. Silverstein, F. X. Webster, D. J. Kiemle, Spectrometric Identification of Organic Compounds, 7th ed., Wiley, 2005: substituent increments for ring carbons and proton shifts α and β to functional groups.
• D. M. Grant, E. G. Paul, J. Am. Chem. Soc. 1964, 86, 2984: alkane increments.
• C. Pascual, J. Meier, W. Simon, Helv. Chim. Acta 1966, 49, 164: alkene proton increments.
• H. E. Gottlieb, V. Kotlyar, A. Nudelman, J. Org. Chem. 1997, 62, 7512, and G. R. Fulmer et al., Organometallics 2010, 29, 2176: residual solvent signals and the reference shifts of common solvents.
• Where the printed tables have no row for a group, its increment is taken from the simplest parent compound, for example dimethyl sulfoxide for sulfoxides.