How to open a real GC/MS report, find the four things that matter, and know where the paper stops telling you anything.
~13 min read
Every claim about purity in this industry ends, if you follow it far enough, at a piece of paper. Somebody drew a sample, an instrument ran it, and it produced a list of compounds with a number against each. Everything said afterwards summarises that list, and few of the people repeating it have opened one.
The paper is a GC/MS report: gas chromatography and mass spectrometry, two instruments joined end to end, doing two jobs and printing one page. Between them they say which volatile compounds were in the sample and roughly how much of each. Purity and Quality lists the tests that make up a standard; this is the first of them.
One piece of honesty first. A report characterises a sample — one sample, drawn on one day, from one batch. It does not certify a company, and a clean report on one batch says nothing about the next off the still. The value of a testing programme is that every batch produces one.
Photograph to come
A working gas chromatograph–mass spectrometer on a laboratory bench, the way it actually looks: a plain boxy oven unit with the mass spectrometer bolted to its side, a tangle of gas lines, an autosampler tray of small amber vials, a monitor showing a live chromatogram. Unglamorous and real — not a stock shot of a scientist holding a beaker to the light. The point is that the machine behind every purity claim is ordinary bench equipment.
An essential oil is a crowd — a few dozen to several hundred volatile molecules, most of them terpenes and terpenoids, in wildly different amounts. No instrument reads that crowd at once: separate first, identify second.
A fraction of a microlitre of oil is injected at around 250 °C and flashes to vapour. An inert carrier gas, normally helium, sweeps it onto the column: a fused silica capillary thirty to sixty metres long and a quarter of a millimetre across inside, its wall coated with a stationary phase perhaps a quarter of a micrometre thick. That thin film is the entire mechanism.
Every molecule alternates between two states on the way down. In the gas it moves at the speed of the carrier; dissolved for a moment in the coating, it is not moving at all. One that interacts strongly arrives late, one that barely interacts arrives early. Nothing is filtered — separation is purely a difference in how long each kind of molecule dawdles.
The oven does not sit still. A typical run holds at 50 to 60 °C then ramps a few degrees a minute to around 250 °C over forty to ninety minutes, and rising temperature persuades progressively heavier compounds out of the coating — light monoterpenes first, sesquiterpenes much later. Each reaches the far end as a narrow band, the detector registers a rise and fall, and that becomes a peak; the moment of arrival is its retention time. Peaks with times and sizes, no names.
Hence the first real limitation. Two compounds can interact with the coating to almost exactly the same degree and arrive together. They co-elute, and the instrument cannot tell: one peak where there should be two, the smaller added to the larger, a figure quietly inflated. Anything added deliberately is easier to hide under an existing peak. The fixes — a different coating, a longer column, a slower ramp — all cost time.
Diagram to come
A cutaway of a short length of capillary column drawn wide enough to see inside: the inner wall lined with a thin stationary-phase film, carrier gas flowing left to right. Three differently coloured molecules travel along it — one barely touches the coating and is well ahead, one dips in and out and sits in the middle, one spends most of the drawing stuck in the film and trails. Below the tube, the same three arriving at a detector at three separated moments, producing three peaks on a time axis. The point is that separation is a difference in dawdling, not a filter: draw the molecules moving and pausing, never being sorted by size.
A retention time is a fact about one instrument on one day. It depends on the column’s length, the thickness of its coating, the carrier flow rate, the temperature programme, even how much has been trimmed off the column head during maintenance. The same compound might arrive at 12.41 minutes in one laboratory and 13.08 in another, and both are correct. Comparing raw retention times between reports is close to meaningless.
So the run carries its own ruler. A series of straight-chain alkanes is injected under the same conditions and assigned fixed numbers by definition — the ten-carbon alkane is 1000, the eleven-carbon 1100 — and every real compound is placed between two of them by interpolation. That number is its retention index. Because it is a position relative to markers that travelled the same column under the same programme, it holds steady across instruments and years: limonene sits close to 1030 on a standard non-polar phase almost anywhere. Published tables exist for virtually every constituent, so a report giving indices is one you can check yourself.
Separation on its own gives arrival times and no identities. The mass spectrometer answers the other question: what is this thing that just arrived.
As each separated compound leaves the column it enters the ion source, under vacuum, where a beam of electrons at a standard 70 electronvolts strikes it. The collision knocks an electron off, leaving the molecule charged and badly destabilised, and it breaks along its weakest bonds into smaller charged fragments, which are sorted by mass-to-charge ratio and counted.
What comes out is a bar chart: fragment masses along the bottom, how many of each up the side. The value of that mass spectrum is that it is reproducible — the same molecule broken at the same energy shatters the same way every time, on two instruments on two continents. That is the fingerprint. Identification is then a match against a library of tens of thousands of spectra, each recorded from an authenticated pure compound.
Neither is much use alone: in series, the chromatograph hands the spectrometer one compound at a time.
| Answers | Blind to | |
|---|---|---|
| Gas chromatography alone | How many components, in what proportions, arriving when | What any of them is |
| Mass spectrometry alone | What a pure compound is, from how it fragments | Anything in a mixture — spectra overlap into noise |
| The two in series | One compound at a time, named and measured | Anything that will not vaporise |
The similarity score is usually reported out of 1000. Above about 900 the match is strong; between 800 and 900 it is a reasonable suggestion; below 800 it is the software’s best guess. Analysts write “tentatively identified” against those rows, and that phrase is the laboratory being honest rather than careless.
Two things make a match a probability. A library holds only what somebody put in it — hand the software a compound that is absent and it returns the nearest thing it has, with a confident-looking score. And some molecules genuinely fragment almost identically. Thymol and carvacrol are the best example in this field: positional isomers, the same atoms, the same mass of 150, differing only in where one group sits on the ring, breaking into nearly the same fragments in nearly the same proportions. Thymol defines thyme and carvacrol defines oregano, so a laboratory that confuses them has confused two oils.
They are separated by the chromatography, not the spectrometry — carvacrol elutes slightly later on a non-polar phase and their published indices differ by a reliable amount. So the rule of identification is that two independent criteria must agree: the spectral match, and a retention index within a few units of the published value, confirmed for anything critical against an authentic standard. An identification is a conclusion drawn at a stated confidence, not a reading off a dial.
Reports differ in typography and almost not at all in structure. The same parts come in the same order, and you can work through them in two minutes.
Area per cent is not weight per cent
Detectors do not respond equally to equal masses. A flame ionisation detector responds in rough proportion to how much carbon burns, so oxygenated compounds — alcohols, esters, oxides — read slightly low against hydrocarbons of the same weight, part of their mass being oxygen that contributes nothing. Which is why a careful laboratory runs the sample twice: one detector for measuring, one for naming. Read 32.4% as “about a third”.
Diagram to come
The core visual of the lesson: one full page of a GC/MS report, drawn accurately and annotated. Top third is the header block as a small grid of fields. Middle third is the chromatogram — time along the bottom, response up the side, two or three very tall peaks early, a dense run of small peaks after them, flat baseline, and one pair of barely-resolved peaks drawn touching at the shoulder and called out as co-elution. Bottom third is the peak table, six columns and about eight visible rows, with one row highlighted and traced by a leader line up to its own peak so the reader sees that the row and the peak are the same object. Callouts point at the lot number, the retention index column, the area per cent column, a tentative identification flag, and the long tail of sub-one-per-cent rows. Must be legible on a phone: large clean type for the labels, and let the report itself be suggested rather than fully typeset.
What you do with the page is narrower than it looks. Find the three or four peaks that make up most of the oil and check they belong to that species. Check each figure falls inside the published range. Then scan the rest for anything with no business being there. The names only mean something once you can place them in their functional groups.
Which oils are available differs across the 65 markets this app serves, and so does what a label may say. Read your own market’s documentation.
Keep oils away from eyes and inner ears, and out of reach of children. If you are pregnant, nursing, under medical care or taking medication, speak to your healthcare practitioner before use. Nothing here is intended to diagnose, treat, cure or prevent any disease.
Two reports on the same oil, from the same farm, in consecutive years will not match. A plant is not a factory: it builds its compounds in answer to the season, the soil, the altitude and the hour it was cut.
So a published profile is always a range, and the ranges are wider than people expect. Lavender and lemon show it.
| Oil | Published range | What the range is telling you |
|---|---|---|
| Lavender | Linalyl acetate 25–45%, linalool 20–47%, ocimene 0.3–10% | Both main constituents move twenty points; the third can be almost absent or a tenth of the oil |
| Lemon | Limonene 60–75% | Tighter, because the peel is doing one thing — still fifteen points wide |
| Wild Orange | Limonene 80–97% | One compound dominates; the character sits in the few per cent left |
| Bergamot | Limonene 20–55%, linalyl acetate 10–45%, linalool 3–20% | Two chemical families in one oil, the balance shifting a long way |
A range is a description, not a recipe
The range is what the plant was observed to do, written down — not a target the producer aims at. Which is why a run of batch reports landing on the same figures year after year is a signal rather than a reassurance.
A figure outside the published range is not automatically fraud. It can mean an unusual season, an early harvest, or a region whose range has not caught up. What it reliably means is a question, and the answer lives in the sourcing record.
Sometimes the difference between two reports is not variation inside a range but a different oil out of the same species: a population in one place dominated by one compound, the same species elsewhere, visually indistinguishable, dominated by another. These are chemotypes.
Almost none of it is crude. Anyone making a cheap oil pass a test aims at precisely the figures the test checks, so the obvious things — a thin profile, a missing constituent — are what you never see.
A figure that has been lifted
A batch came in below the published range, so a quantity of its main constituent, made in a factory, was added. The main peak now looks right; the tell is in the small peaks around it. A plant building a compound also makes a family of related minor ones off the same pathway, in consistent proportion. A drum of the purified compound brings only itself, and those ratios stop making sense.
A cheaper oil extended into an expensive one
Lavandin, *Lavandula Ă— intermedia*, yields far more oil per hectare than true *Lavandula angustifolia* and overlaps it heavily, so most of the table still reads as lavender. What gives it away is a compound belonging to the other plant: lavandin carries several per cent camphor, true lavender very little.
An isolate that is too pure
A real oil is a crowd: dozens to hundreds of peaks, most of them small. Six clean peaks and an empty baseline is not a purer oil but a reconstruction, built from the compounds anybody checks. The practical face of The Oil Effect — the long tail is not noise.
A nature-identical compound
The same molecule, built in a factory. Identical structure, mass spectrum and retention index. GC/MS cannot see it at all — there is nothing to see. It is caught only by asking about the atoms: which mirror form, which isotopes.
Chiral gas chromatography asks which mirror form. Many constituents exist in two mirror-image forms — enantiomers — identical in mass, in fragmentation and in behaviour on an ordinary column. A plant’s enzymes are themselves shaped objects, so they build one form strongly in preference to the other: the linalool in true lavender is heavily weighted to a single form, while a synthesis has no such preference and gives the two in close to equal amounts. Run the sample on a column whose coating is itself chiral, the pair separate, and the ratio answers the question.
Isotope ratio mass spectrometry asks which atoms. Carbon occurs as a light form and a slightly heavier one, and carbon fixed by a living plant carries a different ratio of the two from carbon out of a petrochemical feedstock; hydrogen carries a parallel signal, its ratio set by the water the plant drew up, so geography leaves a trace inside the molecule. The instrument measures those ratios rather than identifying the compound at all. It is slow, expensive, and only as good as the reference measurements of authentic material it is compared against, which is why it runs on a sampling basis rather than on every batch.
There is also a category GC/MS is structurally unable to detect. Extend an oil with a fixed vegetable oil and the triglycerides never vaporise, never enter the column, never register — and because area per cent is normalised to 100, the volatile fraction still adds to 100 and the profile looks untouched. That one is caught by weighing what is left when the volatiles are driven off, which is why several methods sit together in the purity panel. No single method is a verdict.
That is what the page can establish. The more useful discipline is knowing its edges: a document trusted past them does more harm than none.
It cannot tell you whether the plant was grown well. It sees molecules, not fields, and says nothing about soil, about the water the crop was irrigated with, or about whether the harvest was cut at the right moment or a convenient one — a sourcing question.
It cannot tell you whether the oil is still what the report describes. A report is a photograph of a sample on one date, and oils change in the bottle. Keep bottles closed, cool and out of direct light, and buy sizes you will finish. Separately from age: the expressed citrus oils — lemon, lime, grapefruit, bergamot and wild orange — are photosensitising. After applying any of them to skin, avoid direct sunlight and UV for up to twelve hours.
What ages an oil is oxygen, light, warmth, and the growing gap of air above the liquid as the bottle empties. Monoterpene-rich oils change fastest, and the citrus oils are monoterpene-rich almost to the exclusion of everything else — lemon at 60–75% limonene, wild orange at 80–97%. Limonene takes up oxygen readily and forms products that were not there when the bottle was filled, and an oxidised citrus oil is more likely to cause skin sensitivity than a fresh one. A report on the batch, however clean, says nothing about a bottle that has been open on a windowsill for two years.
And it cannot tell you that the sample tested is the oil in your hand — the point that matters most and gets least attention. The only thread from a laboratory bench to your shelf is the lot number: on the report, on the bottle, the same characters in both places. A report with no lot number is about an oil somewhere in the world.
What the page proves is narrow: that on a given date, a given sample contained these volatile compounds in approximately these proportions, and that nothing the method could see was there that should not have been. That is all — and still more than almost any other claim made about an essential oil. A report is a tool for asking better questions, not a certificate.
The skill only becomes useful when it meets an actual oil.
This content is for education only. Nothing here is medical advice, and none of it is intended to diagnose, treat, cure or prevent any disease. Speak to a qualified healthcare practitioner about your own situation, particularly if you are pregnant, nursing, under medical care or taking medication.
Lesson 10 of 18 · Science