13C NMR Predictor

In short: Draw a molecule or paste its SMILES. You get a predicted ¹³C shift for every numbered carbon, DEPT-90/135 multiplicities, symmetry-equivalent carbons merged into one signal, and a stick spectrum. It runs in your browser with no sign-up.

The shifts come from textbook additive increments, and each value shows the terms it was built from. They are predictions, not measured data. On our test set of 15 molecules the mean absolute error is 1.3 ppm; see accuracy.

How to use the 13C NMR predictor

1. Draw the molecule, or paste a SMILES string and press Enter. Pick an example to try it.
2. Each carbon gets a red number on the drawing. The table and the spectrum use the same numbers.
3. Carbons that are equivalent by symmetry are merged into one signal (for example the two ortho carbons of toluene).
4. Read the multiplicity: CH₃, CH₂, CH or C (quaternary). DEPT-90 shows only CH. DEPT-135 shows CH and CH₃ up and CH₂ down. Quaternary carbons disappear in both.
5. Open "terms" in any row to see how the value was built. Copy the list, or download CSV or the spectrum as SVG.

How the shifts are estimated

Every carbon starts from a base value for its type. Then the effect of each nearby atom or group is added. These additive increments are the classic textbook method for estimating ¹³C shifts by hand.

• sp³ carbons: δ = −2.3 + Σ Z(α, β, γ, δ) for every group up to four bonds away (Grant–Paul and Pretsch increments), plus Grant–Paul steric terms for branched carbons and a ring term taken from the parent cycloalkane (cyclohexane 26.9 ppm, cyclopentane 25.8 ppm).
• Benzene-type carbons: δ = 128.5 + ipso, ortho, meta and para increments for each substituent. Pyridine-type nitrogen and the heteroatoms of furan, thiophene and pyrrole rings add position terms taken from the parent heterocycles.
• Alkene carbons: δ = 123.3 + increments for groups on the same carbon and on the other carbon of the C=C.
• Carbonyl, nitrile and alkyne carbons: a base value for the class (ketone ≈ 207, aldehyde ≈ 200, acid ≈ 177, ester ≈ 171, amide ≈ 172, nitrile ≈ 117 ppm) with corrections for extra carbons on the α-carbon, conjugation and ring size.
• Equivalent carbons: carbons with the same topological symmetry class (OpenChemLib) are merged and averaged.

Sources: E. Pretsch, P. Bühlmann, M. Badertscher, Structure Determination of Organic Compounds, 4th ed. (Springer, 2009); D. M. Grant & E. G. Paul, J. Am. Chem. Soc. 1964, 86, 2984; D. F. Ewing, Org. Magn. Reson. 1979, 12, 499; R. M. Silverstein, F. X. Webster, D. J. Kiemle & D. L. Bryce, Spectrometric Identification of Organic Compounds, 8th ed. (Wiley, 2014).

Accuracy: measured against experimental spectra

We compared the predictions with experimental ¹³C shifts in CDCl₃ for 15 common molecules. Values for ethanol, acetone, toluene, acetic acid, ethyl acetate, pyridine and acetonitrile are from Fulmer et al. (2010). The others are literature values (SDBS, AIST, and standard texts), rounded to 0.1 ppm. This table is computed live in your browser from the same code that makes the predictions.

Read these numbers with care. The benzene increments were first derived from monosubstituted benzenes such as toluene, anisole, benzaldehyde and nitrobenzene, so their near-zero error flatters the method. We also checked the test molecules while writing the code. The second set (methanol, ethers, THF, DMF, triethylamine, alkanes, nitromethane and others, all from Fulmer et al. 2010) is a fairer guide: about 2 ppm on average. Expect 2–5 ppm for typical drug-like molecules and 5–10 ppm or more for some carbons in heterocycles (caffeine C4 and C5 are off by about 10 ppm), crowded rings and molecules with several heteroatoms on one carbon. Database (HOSE-code) and machine-learning predictors usually reach about 1.5–2.5 ppm.

G. R. Fulmer, A. J. M. Miller, N. H. Sherden, H. E. Gottlieb, A. Nudelman, B. M. Stoltz, J. E. Bercaw, K. I. Goldberg, Organometallics 2010, 29, 2176–2179. SDBS: Spectral Database for Organic Compounds, National Institute of Advanced Industrial Science and Technology (AIST), Japan.

What this tool does not do

• It does not simulate a measured spectrum. Peak heights in ¹³C NMR depend on relaxation and NOE, so the sticks only show position and multiplicity.
• No solvent, temperature, concentration or pH effects. Values are close to CDCl₃. Acids, amines and phenols can move by several ppm in DMSO-d₆ or D₂O.
• No stereochemistry. Diastereotopic carbons (for example the two CH₃ of an isopropyl group next to a stereocentre) are merged, and cis/trans or axial/equatorial effects are ignored.
• No conformational, tautomer or ring-strain effects beyond the simple ring terms. Large errors are likely for fused heterocycles, organometallics, carbocations and molecules with unusual elements.
• No ¹³C–¹⁹F or ¹³C–³¹P couplings are shown.

For a database-based prediction that also gives ¹H spectra, try Structure to NMR. To look up typical ranges, use the NMR chemical shift table.