How much evidence identifies an unknown impurity: from a formula to a structure
A formula is not a structure, and a library match is not an identification. This works through what each layer of evidence actually proves - the formula search, the isotope arithmetic, the fragment that cannot fix a position - and the one step that converts a proposed structure into an identified impurity.
On this page
An identified impurity has a definition, and it is not a library hit
ICH Q3A defines an identified impurity as an impurity for which a structural characterisation has been achieved. The definition is short and it sets the bar. A peak that matches a spectrum in a library, or an accurate mass that fits one elemental composition, is not structural characterisation. The guideline asks for the studies conducted to characterise the structure of impurities above the identification threshold to be described, which means the reasoning has to survive being read by someone who was not in the laboratory.
That distinction carries commercial weight. Everything downstream of identification - the qualification argument, the specification limit, the impurity standard that makes the method quantitative - rests on the structure being right. An isomer assigned in error produces a limit for a compound that is not in the batch, and no amount of later method work repairs it.
The formula narrows the search, it does not end it
Accurate mass is the cheapest evidence available and the most over-read. Take an ion at m/z 455.2111 in positive mode and search carbon, hydrogen, nitrogen, oxygen and sulfur within 40, 70, 8, 14 and 2 atoms, keeping only compositions with an integer ring-and-double-bond equivalent between 0 and 30. At 5 ppm, seven compositions survive. At 2 ppm, three. At 1 ppm, one: C24H30N4O3S, at twelve ring-and-double-bond equivalents.
That looks like an answer and is not one. Twelve degrees of unsaturation in a molecule carrying four nitrogens, three oxygens and a sulfur is compatible with a very large number of connectivities, and every one of them returns the same exact mass and the same isotope pattern. Mass measures composition; identification is a claim about connectivity. The distance between the two is where most of the analytical work sits.
Two cautions on the search itself. The list is only as good as its constraints: widen the sulfur range or allow phosphorus and the 1 ppm window reopens. And 1 ppm is an instrument specification measured on a calibrant, not a promise for a peak sitting at 0.05 per cent of full scale in a dirty matrix, where the mass error is routinely two to three times the error on the tuning solution.
The isotope pattern is a subtraction, not a threshold
The textbook rule is that one sulfur appears as an M+2 of about 4.4 per cent. Used as a threshold the rule misleads, because carbon contributes to M+2 as well.
For C24H30N4O3S the calculated M+1 is 28.7 per cent of the monoisotopic peak and M+2 is 9.1 per cent. Part of that M+1 is not carbon at all: twenty-four carbons give 25.7 points, four nitrogens 1.5 and the sulfur 0.75, leaving under a point to the hydrogens and the 17O. Dividing 28.7 by 1.1 returns 26 carbons, two too many, so the carbon count cannot be read until nitrogen and sulfur have been assigned.
M+2 behaves the same way. Keep twenty-four carbons, drop the sulfur and take the oxygens from three to five: M+2 falls from 9.1 to 4.8 per cent. The 4.3-point difference is the sulfur; the 3.2 points that remain are the 13C pair among twenty-four carbons. At this size the sulfur is still visible, but it is a minority of M+2, and a rule expecting 4.4 per cent and finding 9 would have reported no sulfur at all.
The pattern earns its place for another reason: it is orthogonal to mass accuracy. Sulfur and two oxygens differ by 17.8 mDa, which is 39 ppm at m/z 455. An instrument that cannot resolve those two compositions can still separate them on isotopes, because the sulfur version carries 4.25 per cent of 34S where the oxygen version carries 0.4 per cent of 18O.
Fragments prove a substructure, never a position
Product-ion spectra are where a formula becomes a hypothesis, and where the hypothesis is most often over-claimed. A neutral loss of 28 Da is reported as carbon monoxide or as ethylene depending on what the analyst expected. The two differ by 36 mDa - 27.9949 against 28.0313 - which is 85 ppm on a fragment at m/z 427 and needs a resolving power near 12000 to separate. A nominal-mass instrument returns the same 28 for both, and the assignment is then an assumption wearing the clothes of a measurement.
Even a correctly measured loss proves less than it appears to. A loss of 43.9898 is carbon dioxide and a loss of 44.0262 is C2H4O, and either can arise from more than one position on the same skeleton. What a fragment establishes is that a set of atoms left the ion as a unit, not where they were attached. Positional isomers fragment into the same pieces, which is why a fragmentation tree narrows a structure without ever closing it.
Intensity adds a further caution. Fragment ratios move with collision energy, cone voltage, source temperature and matrix load, so a library match on an MS/MS spectrum compares spectra, not structures. Two laboratories running the same compound on different instruments can return scores that differ by more than the gap between two candidates.
Retention is evidence only when the mechanism changes
Relative retention on the method that found the peak is a consistency check, not an identification. Two isomers separated by 0.3 minutes on one C18 gradient can invert on another, so agreement on a single method excludes very little.
What discriminates is a second mechanism: a different stationary phase, a different mobile-phase pH, or a mixed-mode column, chosen because the candidate isomers are predicted to move relative to each other. If the unknown and the candidate standard co-elute on both, the case is far stronger than two runs on the same method. If they split on the second, the structure is wrong and the finding cost one injection.
The corollary matters when buying: a certificate that reports a retention time and a purity without stating the method cannot support an identification claim, because the retention number does not transfer to any other system.
What a reference standard adds that no spectrum does
A standard does two things a spectrum cannot. It fixes the retention time, which turns a proposed structure into a specific peak in a specific method, and it supplies a response factor, which is what converts a peak area into a percentage.
ICH Q3A is explicit about the second. Impurity levels can be estimated against the response of the drug substance itself, and where the response factors of the substance and the impurity are not close, the practice remains appropriate provided a correction factor is applied or the impurities are being overestimated. Read plainly, a laboratory without the impurity standard has two honest options: apply a measured relative response factor, or declare that the reported figure is an upper bound. A figure published with neither is a number with no stated basis.
The same guideline asks that reference standards used to control impurities be evaluated and characterised according to their intended use. For an identification standard that means the certificate should state how the structure was established, not only what the purity is.
Four claims, and the evidence each one needs
| Claim | Evidence held | What can be written | What is still missing |
|---|---|---|---|
| Peak detected | Retention on one method, UV spectrum | An unknown at a stated relative retention and area per cent | Everything above it |
| Composition assigned | Accurate mass inside 1 ppm, isotope fit | A formula, with the search space stated | Connectivity |
| Structure proposed | Formula, product ions assigned, a route consistent with forced degradation | A proposed structure, described as tentative | An independent confirmation |
| Structure confirmed | Candidate standard co-injected, ions and retention matching on two mechanisms, or NMR connectivity | An identified impurity, in the sense the guideline uses | Nothing for identity; the response factor for quantitation |
The table reads downward as a cost table too. Each row costs more than the one above it, and each supports a claim the row above cannot. A submission reporting a structure at the third row is not wrong; it is reporting a hypothesis, and the strongest version of that submission says so in those words.
Where NMR settles what mass spectrometry cannot
Nuclear magnetic resonance is the only routine technique that reports connectivity directly: which proton sits on which carbon, and through two or three bonds, which fragments are joined. For positional isomers that share a formula and a fragmentation pattern it is usually the only in-house answer.
The constraint is amount. An impurity at 0.10 per cent in a one-gram batch is one milligram, enough for a proton spectrum on a modern instrument and marginal for the two-dimensional experiments that establish connectivity. That arithmetic, not the cost of instrument time, decides whether an unknown earns a confirmed structure or a proposed one, and it is worth doing before the isolation campaign is planned rather than after it has been paid for.
Buying a standard before the structure exists
Impurity work carries a sequence problem: the standard cannot be ordered until the structure is known, and the structure is usually confirmed with a standard. The way out is not to buy early but to buy deliberately, and the questions that decide the purchase are the ones a catalogue entry does not answer.
Ask whether the lot is the isomer proposed rather than a mixture, since positional and stereo isomers share a formula and often a mass spectrum. Ask how the structure of that lot was established, and whether the evidence sits on the certificate or only in a technical pack. Ask what the assigned content is expressed against - as is, dried, or free base - and what the uncertainty is at the stated coverage factor. A certificate that answers none of these supports a purity figure and nothing else.
What to keep in the file
The record that survives review is the reasoning rather than the spectra: the search space and the mass error, the isotope fit with the arithmetic shown, the fragments assigned and the ones left unexplained, the second mechanism used for retention, and the standard lot with the sentence from its certificate that states how the structure was established. An unexplained fragment is not a failure to hide. It is the honest edge of the claim.
The guides below cover the threshold that triggers the obligation, the standard that closes it, and the stress work that generates the peaks in the first place.
what an impurity reference standard is | the three impurity thresholds | forced degradation study design | impurity profile comparison | chiral and stereoisomeric standards | reading a certificate of analysis
Frequently asked
Is an accurate mass inside 1 ppm enough to identify an impurity?
Can an impurity be called identified on a library match alone?
We do not have the impurity standard. Can we still report a percentage?
Two isomers give the same accurate mass and the same fragments. How are they separated?
How much material does a confirmed structure need?
A supplier certificate states a retention time and a purity. Does that prove identity?
Sources and further reading
Every document below is published by the body that owns it; none of them is a summary written by us. Accessed 2026-11-20.
- ICH - Q3A(R2) Impurities in new drug substances — the definition of an identified impurity as one for which structural characterisation has been achieved, the requirement that studies characterising the structure of impurities above the identification threshold be described, and the allowance to estimate levels against the drug substance response where a correction factor is applied or the impurity is overestimated
- NIST - Atomic weights and isotopic compositions for all elements — the isotopic abundances behind the arithmetic in this article: 1.07 per cent for 13C, 4.25 per cent for 34S and 0.205 per cent for 18O, together with the exact masses of the isotopes
- ISO - ISO 17034:2016, General requirements for the competence of reference material producers — the requirements a producer is assessed against when assigning property values and stating the uncertainty of a certified value, which is what an identification standard has to carry before it can close an identification
- ISO - ISO 33407:2024, Guidance for the production of pure organic substance certified reference materials — the production guidance specific to pure organic substance reference materials, the category an impurity standard belongs to when its content is assigned by mass balance
- EDQM - FAQ: are drying conditions required before use — that where no drying conditions are stated the material is used as received and an assigned content is expressed as is, which is the basis an identification standard must be used on
Send us a CAS number
Availability, documentation and lead time — answered by the people who handle the batch, not a ticket queue.