High-Resolution Mass Spectrometry: How Much Resolving Power and Mass Accuracy You Actually Need

Instrument specifications advertise a single resolving-power number and a single mass-accuracy number, and both are close to meaningless without a job attached to them. "240,000 resolution" is a number measured at one m/z, under one definition, on a scan whose duration you may not be able to afford. "Sub-1 ppm mass accuracy" describes an error budget, not an identification.

The useful questions are arithmetic ones: how much resolving power does it take to separate this pair of interfering ions, and how many candidate molecular formulas survive at this mass with this ppm window? Both have exact answers, and both are computed below. This page is about the analyser and the mass measurement. The ion source is covered in tuning electrospray source parameters, targeted quantitation on a triple quadrupole in building and optimising MRM transitions, and the discovery-proteomics workflow in DDA vs DIA sample prep and run QC.

Two numbers, and why neither is portable

Resolving power is definition-dependent

Resolving power is R = m/Δm. Everything depends on how Δm is defined, and there are two conventions in routine use:

• FWHM (full width at half maximum) — Δm is the width of a single peak at 50% of its height. Standard for quadrupole, time-of-flight, Orbitrap and FT-ICR instruments, and the number vendors quote.
• 10% valley — Δm is the separation at which two equal-intensity adjacent peaks are divided by a valley falling to 10% of their height. The older magnetic-sector convention.

For Gaussian peaks the 10%-valley figure is approximately half the FWHM figure for the same physical performance. That factor is not a textbook curiosity — it is written into regulation. Commission Implementing Regulation (EU) 2021/808, governing confirmatory methods for residues of pharmacologically active substances, sets the HRMS requirement as resolution "typically… greater than 10 000 for the entire mass range at 10% valley or 20 000 at full width at half maximum (FWHM)". One instrument, one performance level, two legitimate numbers differing by 2×. Always state which definition you are quoting, and never accept a bare resolving-power figure in a method comparison without it.

Resolving power is also m/z-dependent, and the dependence differs by analyser

A second reason a single number does not travel: resolving power is not constant across the mass range, and how it varies is an architectural property.

• Orbitrap and FT-ICR resolve by measuring an image-current transient and Fourier-transforming it. Frequency resolution improves with transient length, and the ion oscillation frequency falls as m/z rises. For an Orbitrap, resolving power therefore falls as roughly 1/√(m/z) — which is why the specification is always quoted at a stated m/z, conventionally 200.
• Time-of-flight resolves by flight-time dispersion down a fixed path. Resolving power is set by the flight path, reflectron and detector timing, and is roughly constant across the useful mass range rather than decaying with it.

Applying the 1/√(m/z) scaling to three common Orbitrap specifications:

Quoted R at m/z 200 (FWHM)	m/z 400	m/z 600	m/z 1000
60,000	42,400	34,600	26,800
120,000	84,900	69,300	53,700
240,000	169,700	138,600	107,300

A 120,000 instrument is a ~54,000 instrument where a tryptic peptide or a lipid actually sits. Check the analyte’s m/z, not the datasheet’s.

Mass accuracy is a different quantity from resolving power

Mass accuracy is the deviation between the measured m/z and the theoretical value, conventionally expressed in parts per million:

error (ppm) = (mmeasured − mtheoretical) / mtheoretical × 106

Because ppm is relative, the same ppm budget is a very different absolute window at different masses:

m/z	1 ppm	3 ppm	5 ppm	10 ppm
100	0.10 mDa	0.30 mDa	0.50 mDa	1.0 mDa
200	0.20 mDa	0.60 mDa	1.0 mDa	2.0 mDa
300	0.30 mDa	0.90 mDa	1.5 mDa	3.0 mDa
500	0.50 mDa	1.5 mDa	2.5 mDa	5.0 mDa
1000	1.0 mDa	3.0 mDa	5.0 mDa	10 mDa

This is why EU 2021/808 expresses its identification tolerance in both units: the mass deviation of all diagnostic ions must be below 5 ppm, or below 1 mDa where m/z < 200. The dual form is not redundancy. At m/z 100, 5 ppm is a 0.5 mDa window; permitting 1 mDa instead is effectively a 10 ppm allowance at that mass, acknowledging that a relative tolerance becomes physically unreasonable as the denominator shrinks.

The vocabulary that gets conflated

• Nominal mass — the integer sum of integer nucleon counts. C₈H₁₀N₄O₂ = 194.
• Exact (monoisotopic) mass — the calculated sum of the most abundant isotope masses. C₈H₁₀N₄O₂ = 194.08038 Da.
• Accurate mass — the measured value. "Exact mass" is what theory says; "accurate mass" is what your instrument reported. Accurate-mass measurement is the act; exact mass is the reference it is compared against.
• Average mass — isotope-abundance-weighted, and the wrong reference entirely for an HRMS measurement of a resolved monoisotopic peak.

The electron trap. The theoretical value you compare against must be the ion’s mass, not the neutral’s: for [M+H]⁺ add a proton (1.007276 Da), not a hydrogen atom (1.007825 Da). The difference is one electron mass, 0.000549 Da. As a relative error that is 5.5 ppm at m/z 100 and 1.1 ppm at m/z 500 — so below roughly m/z 110, neglecting the electron alone consumes an entire 5 ppm budget before any instrument error is counted. Most software handles this; homemade calculation spreadsheets frequently do not.

What resolving power actually buys: separating a named pair

Resolving power does exactly one thing — it separates two ions of nearly equal m/z. So the only sensible way to specify it is to name the pair you must separate, compute their mass difference, and divide. Below, Δm is computed from IUPAC isotope masses, and R is the minimum FWHM resolving power at which the two peaks are just resolved.

Isobaric pair (same nominal mass)	Δm	R needed at m/z 200	at 500	at 800
N2 (28.00615) vs CO (27.99491), at m/z 28	11.23 mDa	R ≈ 2,500 at m/z 28
CH4 vs O substitution (16.03130 vs 15.99491)	36.39 mDa	5,500	13,700	22,000
34S vs 2×13C on the M+2 peak	10.91 mDa	18,300	45,800	73,300
13C vs 15N on the M+1 peak	6.32 mDa	31,600	79,100	126,600
C3 vs SH4 substitution (36.00000 vs 36.00337)	3.37 mDa	59,300	148,300	237,300

Read down that column and the point makes itself. The classic CO/N₂ separation that defines "high resolution" in an introductory course needs about 2,500 — achievable on a modest TOF. The CH₄/O substitution, which is the most frequent genuine interference in small-molecule work, needs 5,500–22,000 depending where it falls. Separating ¹³C from ¹⁵N in isotopic fine structure needs 31,600 at m/z 200 and 79,100 at 500. The C₃/SH₄ doublet that dominates petroleum and dissolved-organic-matter analysis needs ~148,000 at m/z 500 — the reason that field runs FT-ICR rather than Orbitrap or TOF.

Three qualifications that matter in practice:

• "Just resolved" is not "quantifiable". R = m/Δm at FWHM is the threshold at which two equal-height peaks become distinguishable. If the interferent is 100× the analyte, or if you need accurate peak areas rather than detection, budget substantially above the computed minimum.
• Unequal intensity is the harder case. A large peak’s tail sits under a small neighbour long after their maxima are separated, biasing both the measured mass and the area of the minor component.
• Charge state has its own requirement. To assign a charge from isotope spacing you must resolve peaks separated by 1.00336/z Th, so R ≥ z × (m/z) / 1.0034. A 5+ ion at m/z 1000 needs ~5,000; a 20+ ion at m/z 1200 needs ~23,900. This, not formula confirmation, is what sets the MS1 resolution floor in most protein and peptide work.

The cost of resolution differs by analyser

On an FT instrument, resolving power is bought with transient length: doubling R roughly doubles the acquisition time for that scan. That trades directly against chromatographic fidelity — a 6-second-wide UHPLC peak needs roughly 10–12 points across it for reliable area integration, i.e. about 2 Hz of usable MS1 scans, and a high-R setting plus a data-dependent MS/MS duty cycle can quietly drop you below that. On a TOF, resolving power is set by the flight geometry rather than by dwell time, so raising it does not cost scan rate in the same way. This asymmetry, not the headline resolution figure, is usually what decides which analyser suits a fast-gradient method. The chromatographic side of that trade-off is covered in GC and HPLC column selection, and the coupling itself in how LC and MS are coupled.

What mass accuracy actually buys: counting the surviving formulas

The common claim is that accurate mass "gives you the molecular formula". It does not, above trivial masses — and the failure is quantifiable. Below is an exhaustive enumeration of neutral CHNOPS compositions whose monoisotopic mass falls inside a given ppm window around three real compounds. Raw counts every arithmetically valid composition. Filtered applies elementary chemical plausibility: ring-plus-double-bond equivalent between 0 and 40, and H/C ratio between 0.2 and 3.1, with S ≤ 3, P ≤ 3, N ≤ 20, O ≤ 30.

Compound and neutral monoisotopic mass	±1 ppm	±3 ppm	±5 ppm	±10 ppm
Caffeine, C8H10N4O2, 194.08038	3 raw / 1 filtered	5 / 2	9 / 2	20 / 7
Verapamil, C27H38N2O4, 454.28316	49 / 5	167 / 19	275 / 34	556 / 69
Reserpine, C33H40N2O9, 608.27338	177 / 35	529 / 105	898 / 183	1,788 / 360

At 194 Da, 1 ppm leaves a single plausible formula — accurate mass really does determine the composition. At 454 Da it leaves five, and at 608 Da it leaves thirty-five. The candidate count grows steeply with mass because the number of compositions in a fixed relative window grows with both the absolute window width and the combinatorial space. Tightening from 5 ppm to 1 ppm at 608 Da only takes you from 183 candidates to 35 — a real gain, and nowhere near an answer.

The practical consequence is the one Kind and Fiehn made explicit in the title of their 2006 analysis, Metabolomic database annotations via query of elemental compositions: mass accuracy is insufficient even at less than 1 ppm (BMC Bioinformatics 7:234), and codified the following year in the Seven Golden Rules heuristic filter set (BMC Bioinformatics 2007;8:105). Formula assignment above a few hundred daltons is a multi-evidence problem, and the additional evidence is roughly in this order of power:

1. Isotopic pattern. The relative intensities of M+1 and M+2 constrain carbon count, and the presence of S, Cl or Br is often unmistakable from the M+2 abundance. This was the central finding of the Kind and Fiehn analysis cited above: isotopic abundance pattern is a more powerful constraint on the candidate list than further improvement in mass accuracy.
2. MS/MS fragments. Each fragment’s own accurate mass must be a sub-formula of the candidate, and neutral losses are highly diagnostic.
3. Retention behaviour and adduct consistency. Seeing the same neutral mass as [M+H]⁺, [M+Na]⁺ and [M+NH₄]⁺ at one retention time is strong evidence the neutral mass is real rather than an artefact.

The self-check that catches most bad assignments: if a search returns exactly one formula at 5 ppm for a compound above ~400 Da, the window is not doing the work — an element restriction or a database restriction is. Re-run it unrestricted and see how many candidates the mass alone actually admits before reporting a confident identification.

What HRMS buys that a triple quadrupole cannot: retrospective analysis

The most consequential difference between a high-resolution full-scan acquisition and a targeted MRM method on a triple quadrupole is not sensitivity, and it is often not even specificity. It is that a full-scan accurate-mass file contains the compounds you were not looking for.

An MRM method records only the transitions you programmed. If a new contaminant of concern is identified two years later, an archived MRM dataset cannot answer whether it was present — you must re-run the samples, if they still exist. An archived HRMS full-scan file can be re-interrogated for the new compound’s exact mass without touching the sample. For long-running environmental, food-safety, doping-control and biobanked clinical programmes, that retrospective capacity is frequently the whole reason the instrument was bought.

The honest counterweight, and the reason the two architectures coexist:

• A triple quad in MRM often wins on sensitivity and dynamic range for a defined panel. It spends the entire duty cycle on your transitions rather than dividing it across a full spectrum, and typically delivers better limits of quantitation and wider linear range for known targets — which is why regulated bioanalysis and therapeutic drug monitoring remain overwhelmingly MRM. See calculating LOD and LOQ and building a calibration curve for how those figures are actually established.
• HRMS wins on specificity against unanticipated interference, and on everything unknown. A nominal-mass MRM transition can be reproduced by a co-eluting matrix component; a 3 ppm accurate mass plus matching isotope pattern plus MS/MS is far harder to counterfeit.
• The decision is workload-shaped, not quality-shaped. Defined recurring panel, tight LOQ, regulated quantitation → triple quad. Unknown identification, suspect screening, retrospective re-analysis → HRMS. The procurement version of this trade-off is worked through in QTOF vs triple quadrupole, and the ionisation-source dimension in GC-MS vs LC-MS.

Suspect screening and the confidence-level vocabulary

Non-targeted work needs a way to state how sure an identification is, and the field has largely converged on the five-level scheme proposed by Schymanski and colleagues (Environmental Science & Technology, 2014): Level 5 is an exact mass of interest; Level 4 an unequivocal molecular formula; Level 3 a tentative candidate structure or compound class; Level 2 a probable structure from library or diagnostic evidence; Level 1 a confirmed structure matched against a measured reference standard. Only Level 1 requires an authentic standard, and no amount of mass accuracy substitutes for one. Report the level explicitly; "identified by accurate mass" without a level is not a claim a reviewer can evaluate.

Which numbers to specify, by job

Job	Resolving power (FWHM, at the analyte’s m/z)	Mass accuracy
Confirming a known small molecule against a reference standard	10,000–20,000	≤5 ppm
Regulated residue confirmation under EU 2021/808	>20,000 FWHM (or >10,000 at 10% valley)	<5 ppm, or <1 mDa below m/z 200
Suspect screening / retrospective non-targeted analysis	30,000–70,000	≤3 ppm, with isotope-pattern scoring
Peptide charge-state assignment (MS1, discovery proteomics)	z × (m/z) / 1.0034; ~24,000 for 20+ at m/z 1200	≤5 ppm precursor tolerance is typical
Resolving 13C from 15N isotopic fine structure	≥79,000 at m/z 500	not the limiting factor
C3/SH4 doublets (petroleomics, dissolved organic matter)	≥148,000 at m/z 500 — FT-ICR territory	<1 ppm

Regulatory note and sourcing caveat: the EU 2021/808 figures above are quoted from the text of Commission Implementing Regulation (EU) 2021/808 as reported in secondary sources; EUR-Lex refused automated retrieval during preparation of this page, so verify the operative wording against the Official Journal text before relying on it in a validation dossier, and check for consolidated amendments. For pesticide residues in food and feed, the parallel document is guidance SANTE/11312/2021, which applies a ±5 ppm mass-accuracy criterion; it is revised periodically, so confirm the current version number with the EU Reference Laboratories rather than citing an older revision.

What degrades the numbers you paid for

• Detector overfilling (space-charge effects). On an Orbitrap, injecting too many ions shifts and broadens peaks; automatic gain control exists precisely to hold the ion population in the calibrated regime. A saturating matrix peak degrades mass accuracy for its neighbours, not only for itself — a mass error that appears only in real samples and not in standards usually points here or at co-elution.
• Calibration drift. External calibration decays over hours to days with temperature and pressure. Internal calibration — a lock mass or continuously infused reference — is what sustains low-single-digit ppm across a long batch, and its absence is the most common reason a specified 3 ppm instrument delivers 10 ppm in routine use.
• Co-elution. An unresolved interferent within the peak-detection window pulls the centroid toward itself. Chromatographic separation is a mass-accuracy control, not just a throughput concern; sample cleanup does the same job upstream.
• Low ion counts. Mass accuracy degrades at the limit of detection because centroid estimation is noise-limited. Quoting a ppm figure without the intensity regime it was measured in is meaningless — specify accuracy at the concentrations you actually work at.
• Scan-rate compromise. As above: selecting the highest available resolution on an FT instrument reduces scans per second, which degrades chromatographic peak definition and therefore quantitation. High resolution and good quantitation compete for the same duty cycle.

Frequently asked questions

What counts as "high resolution" mass spectrometry?

There is no universal threshold, and any source giving one is compressing a definition-dependent, m/z-dependent quantity into a single number. Operationally, an instrument is high-resolution for your purpose when its resolving power at your analyte’s m/z, under the definition you are quoting, exceeds m/Δm for the closest interference you must separate. Regulatory texts do set floors for specific uses — EU 2021/808 uses >20,000 FWHM — but those are minima for a defined task, not a definition of the term.

Is 5 ppm mass accuracy enough to confirm a molecular formula?

Below roughly 200 Da, usually yes. Above ~400 Da, no: the enumeration above returns 34 chemically plausible CHNOPS formulas within 5 ppm at 454 Da and 183 at 608 Da. Isotopic pattern and MS/MS fragments are required, and above ~500 Da they contribute more than further tightening the mass window does.

What is the difference between exact mass and accurate mass?

Exact mass is calculated — the sum of the constituent isotope masses for a proposed formula. Accurate mass is measured. You compare an accurate mass to an exact mass and express the difference in ppm. Comparing a measured monoisotopic peak to an average mass is a common and large error, growing with molecular size.

Does an Orbitrap or a TOF have better resolving power?

At m/z 200 an Orbitrap’s specified figure is typically the higher one, but it decays as roughly 1/√(m/z) while a TOF’s stays approximately flat, so the gap narrows with mass. The more decisive difference is cost: Orbitrap resolution is bought with transient time and competes with scan rate, while TOF resolution is fixed by flight geometry and does not. For fast chromatography with many co-eluting analytes, that asymmetry often outweighs the headline number.

Why does my mass error look fine in standards but not in samples?

Almost always matrix: either a co-eluting interferent inside the peak-detection window shifting the centroid, or a high-abundance matrix ion overfilling the detector and degrading accuracy for neighbouring peaks. Check the mass error against injected amount and against retention time before suspecting calibration.

Can HRMS replace a triple quadrupole for regulated quantitation?

Increasingly for some assays, but not by default. A triple quad in MRM generally retains an advantage in limit of quantitation and linear dynamic range for a defined analyte panel, because it commits the whole duty cycle to those transitions. HRMS is chosen where unknowns, suspect screening or retrospective re-analysis are part of the requirement — and many laboratories run both rather than treating one as an upgrade of the other.

How the numbers on this page were produced

Resolving-power requirements were computed as R = m/Δm from IUPAC monoisotopic isotope masses (¹²C = 12 exactly, ¹H = 1.0078250321, ¹³C = 13.0033548378, ¹⁴N = 14.0030740048, ¹⁵N = 15.0001088982, ¹⁶O = 15.9949146196, ³²S = 31.9720710, ³⁴S = 33.9678669), under the FWHM convention. Candidate-formula counts come from exhaustive enumeration of CHNOPS compositions within the stated ppm window, with the plausibility filter described above. Both are reproducible arithmetic rather than cited figures, and the constraint set is given so the counts can be checked; changing the element set or the filters will change them.

This guide sits in the lab equipment cluster alongside related analytical-instrumentation material, including isotope ratio mass spectrometry, accuracy vs precision in measurement, and reading an IR spectrum.