NMR Impurity
Solver

How to identify an unknown peak

Five steps, in the order that keeps you from assigning a peak you should have left unassigned.

1. 1

   Reference the spectrum before you look anything up

   Every impurity table assumes a correctly referenced spectrum. Set TMS to 0.00 ppm, or anchor to the residual proton peak of your deuterated solvent — 7.26 ppm for CDCl₃, 2.50 for DMSO-d₆, 7.16 for benzene-d₆. Mis-referencing moves every peak in the spectrum by the same amount, so a lookup against an unreferenced spectrum will confidently return the wrong compound.

2. 2

   Select the solvent you actually ran the sample in

   Impurity shifts are solvent-dependent, and not by a trivial amount. Water appears at 1.56 ppm in CDCl₃, 3.33 in DMSO-d₆, 2.84 in acetone-d₆ and 0.40 in benzene-d₆ — a spread of nearly 3 ppm for the same molecule. Picking the wrong solvent column is the single most common way to get a wrong answer out of an impurity table.

3. 3

   Enter the observed shift, or search by name

   Type the chemical shift of the peak you cannot account for. You can also enter a compound name instead — "EtOAc", "DCM", "grease" — to pull up its full reference spectrum in your chosen solvent, which is what you want when you are checking whether a suspected contaminant fits rather than identifying an unknown.

4. 4

   Constrain the search with multiplicity

   A shift match with the wrong splitting is usually a different compound. Setting the multiplicity filter to a singlet, doublet, triplet, quartet or multiplet hard-excludes candidates whose published splitting disagrees, rather than merely ranking them lower. This is the cheapest way to kill a plausible-looking false positive.

5. 5

   Read the confidence, not just the name

   The solver separates "closest candidate" from "confident match" on purpose. Widening the tolerance surfaces more candidates but does not promote them: a peak more than 0.02 ppm from its literature value cannot be reported as Confirmed, and a multiplicity mismatch caps the score below the Likely threshold entirely. If it says no confident match, the honest answer is that the peak is unassigned.

Residual solvent and water shifts, by solvent

Reference your spectrum to the residual peak in the middle column before looking anything up. The water column is the one that catches people out: it is the same molecule in every row, and it moves by nearly 3 ppm across them.

Values from Fulmer et al., Organometallics 2010, 29, 2176–2179. Toluene-d₈, THF-d₈, chlorobenzene-d₅ and TFE-d₃ have more than one residual proton signal; all of them are listed.

How the confidence score works

Most impurity lookups return whichever tabulated peak is nearest and leave you to judge whether that means anything. This one scores the match and refuses to confirm a weak one.

What the confidence percentage is made of

The score is a sum of four terms, not a distance dressed up as a percentage. Chemical-shift agreement contributes up to 70 points, on a step scale: a ¹H peak within 0.01 ppm of the literature value earns 70, within 0.02 earns 62, within 0.03 earns 52, within 0.05 earns 38, and it decays from there. Multiplicity agreement adds up to 20. Whether the compound has a tabulated peak in the solvent you selected adds 5. How often the species actually turns up at the bench — residual solvent, water and grease rank highest — adds up to 5.

The gates that stop false positives

Two hard rules sit on top of the sum. A ¹H peak further than 0.02 ppm from its tabulated shift is capped at 94 and can never be labelled Confirmed. A multiplicity mismatch caps the total at 59, which is below the Likely threshold, so a compound whose splitting contradicts your observation cannot be presented as an assignment no matter how close its shift is.

Why residual solvent wins ties

At 7.26 ppm in CDCl₃ a naive table lookup will happily return pyridine, because pyridine does have a signal there. The solver injects the residual solvent peak as a candidate in its own right and ranks it above rare isochronous reagents, so 7.26 in CDCl₃ resolves to residual CHCl₃ — which is what it almost always is. The same logic covers 2.50 in DMSO-d₆ and the other eleven solvents.

Bulk lists do not force an assignment

When you paste a full peak list from MestReNova or TopSpin, each peak is scored independently and then table peaks are claimed greedily, so the same literature signal cannot be spent twice. A multi-peak impurity such as ethyl acetate may claim several distinct shifts. Peaks that match nothing stay in an explicit unknown list rather than being pushed onto the nearest candidate.

Where these numbers come from

The database currently holds 60 compounds across 12 deuterated solvents — 943 tabulated ¹H signals and 884 ¹³C signals. 46 of the compounds carry a published citation; the remaining 14 are shown as unverified rather than presented as literature values.

The source

The shift tables come from Fulmer, Miller, Sherden, Gottlieb, Nudelman, Stoltz, Bercaw and Goldberg, "NMR Chemical Shifts of Trace Impurities: Common Laboratory Solvents, Organics, and Gases in Deuterated Solvents Relevant to the Organometallic Chemist", Organometallics 2010, 29, 2176–2179. That paper tabulates CDCl₃, C₆D₆, acetone-d₆, DMSO-d₆, CD₃CN, CD₃OD and D₂O — the last carried over from Gottlieb, Kotlyar and Nudelman, J. Org. Chem. 1997, 62, 7512 — and adds THF-d₈, toluene-d₈, CD₂Cl₂, chlorobenzene-d₅ and TFE-d₃.

Every column names its source, or is marked unverified

Each solvent column and each compound carries a citation field. Values transcribed from the paper are labelled Fulmer 2010; values transcribed from its Supporting Information are labelled separately, because the SI is a different document and readers checking your work will want to know which one to open. Compounds with no recorded published source are shown as Unverified with a plain-language reason attached — either that the compound is absent from Fulmer and its SI entirely, or that the SI does cover it but these particular shifts have not been checked against it.

What was removed, and why

A pyridine-d₅ column previously shipped under the Fulmer citation even though the paper does not tabulate pyridine-d₅. Those values were deleted rather than left sitting under a source that does not contain them. A guard module now enforces the rule: any solvent column added outside the allowlist is reported as uncited instead of silently inheriting the citation from the columns around it.

What this tool is not

It is a lookup and ranking aid for trace solvent and impurity peaks, not a structure elucidation engine and not a purity assay. It will not integrate your spectrum, and a confidence score is not a substitute for checking the original spectra. For formal reporting, cite Fulmer et al. directly rather than this page.