EVIDENCE LIBRARY
Does a Cannabis Breathalyzer Miss THC Vapor? Measured Pressure Versus Modeled Breath Fractions
Direct THC/CBD/CBN vapor-pressure measurements and modeled breath aerosol fractions show why particle-only THC breath capture needs validation—not proof of impairment.
Answer: An aerosol-only cannabis breath collector could miss some vapor-phase THC, but the size of that loss has not been established in actual people by the 2026 study examined here. Beuning and colleagues directly measured the vapor pressures of pure THC, cannabidiol (CBD), and cannabinol (CBN) at 364–424 kelvin under laboratory conditions. They then extrapolated to an assumed exhaled-breath temperature of 307.15 K and modeled how molecules might divide between vapor and suspended organic aerosol. With an assumed aerosol concentration of 1.7 µg/m³, the model places about 22% of THC mass in aerosol and 78% in vapor. At ten times that assumed aerosol concentration, the calculation shifts to 74% aerosol and 26% vapor. These are predictions, not collections of breath from drivers, clinical diagnostic performance, or individual impairment determinations.
What was actually measured, and why does that distinction matter?
The direct experimental endpoint was equilibrium vapor pressure of reference cannabinoid materials, expressed in pascals (Pa). Vapor pressure describes the tendency of a condensed substance to generate molecules in the gas phase at a specified temperature. It is a thermodynamic property, not an individual’s THC concentration in a breath specimen. A gas-phase sampling instrument might need to account for that property, but a vapor-pressure result cannot by itself reveal how much THC a person has exhaled.
The authors then used a fitted vapor-pressure relationship and an organic-aerosol partitioning model to calculate a possible distribution at conditions resembling exhaled breath. Experimental vapor pressure and calculated breath-phase fraction therefore occupy two different evidentiary levels. The first involves collected and chemically analyzed cannabinoid vapor in the laboratory. The second depends on assumptions about temperature, organic aerosol concentration, activity coefficient, and equilibrium behavior in a respiratory environment that was not sampled in this study.
Who or what constituted the sample?
There were no human participants, no animals and no authentic exhaled-breath specimens. The experimental units were series of measurements on chemically characterized cannabinoid materials. For each of THC, CBD and CBN, investigators ran six temperature series—three days in each of two weeks—over seven nominal temperatures separated by ten kelvin. Replicated runs improve characterization of apparatus performance and measurement scatter; they are not six independent cannabis users and cannot be converted into clinical sensitivity or specificity.
At the lowest temperature during the first three days, the authors repeated the measurement after completing a heating series to check for detectable thermal deterioration. The laboratory also validated its modified instrument against an independently characterized reference substance, n-eicosane. These safeguards matter for physical metrology, but they do not replicate aerosol generation in a human lung, a roadside collection procedure, or the effect of recent smoking on an actual breath sample.
How did dynamic vapor microextraction isolate cannabinoid vapor?
The study used dynamic vapor microextraction (DVME), a gas-saturation approach rather than a commercial roadside cannabis breathalyzer. Pure cannabinoid material coated small glass beads inside a heated saturator. A controlled flow of high-purity helium passed over the material, carrying a small amount of vapor out of the saturator. The vapor entered a one-meter capillary collection trap cooled to approximately 263 K. That cryogenic trap captured cannabinoid for later extraction; it did not collect natural aerosol particles from a person.
After a controlled helium volume passed through, investigators eluted trapped molecules with methanol and quantified the collected cannabinoid by gas chromatography with flame-ionization detection (GC–FID), using desipramine as an internal standard. Vapor pressure was calculated using the measured collection mass, helium volume, saturator pressure and chemically assessed sample purity. The apparatus thus linked mass of vapor recovered from a calibrated flow to pressure at a defined temperature. A commercial or forensic breath device would face additional collection, selectivity, contamination and specimen-identification requirements beyond those established by this laboratory apparatus.
Which temperatures, quantities and analytes were tested?
Seven nominal temperature settings were 364, 374, 384, 394, 404, 414 and 424 K, corresponding roughly to 91–151 °C. The reference substances were parent Δ9-tetrahydrocannabinol, cannabidiol and cannabinol; the paper did not test urinary THC-COOH or 11-hydroxy-THC as vapor samples. The analyzed purity of the pretreated standards was approximately 96.0% for THC, 97.7% for CBD and 99.1% for CBN, with purity used in calculating pressure. The authors did not administer any amount of these compounds to a participant.
Heating well above breath temperature was essential because cannabinoid vapor pressures were too small for similarly low-uncertainty direct measurement at approximately 307 K with this arrangement. The helium throughput varied to recover sufficient analyte mass. CBN, which had particularly low vapor pressure at 364 K, needed about 6,000 standard cubic centimeters of carrier gas at that temperature. These operating conditions are precisely why direct measurement at 364–424 K must not be silently represented as an observed 307 K breath result.
What were the observed vapor-pressure ranges?
In the original primary paper, pressure rose strongly as temperature increased. Across 364–424 K, measured THC vapor pressure ranged from 0.0459 to 7.833 Pa; CBD from 0.0826 to 13.44 Pa; and CBN from 0.0199 to 5.678 Pa. These are vapor pressures of condensed reference compounds at elevated experimental temperatures, not mass concentrations measured in air sampled from lungs. A compound may have a low pressure in pascals and still partition between phases depending on aerosol loading and other environmental conditions.
The laboratory reported a combined standard relative uncertainty for each pressure determination. For THC it ranged from about 5.2% to 9.5% over the tested temperatures, for CBD 2.9% to 5.0%, and for CBN 3.0% to 5.3%. These are metrological uncertainties on pressure readings, with coverage convention k = 1; they are neither population confidence intervals for drivers nor an accuracy claim for a field breathalyzer. The table below preserves the source’s direct measured temperature span and representative endpoint uncertainty values.
| Measured reference analyte | Temperature range | Directly measured pressure range | Combined relative standard uncertainty, k = 1 |
|---|---|---|---|
| Parent Δ9-THC | 364–424 K | 0.0459–7.833 Pa | 5.2% at 364 K; 9.5% at 424 K |
| CBD | 364–424 K | 0.0826–13.44 Pa | 3.4% at 364 K; 5.0% at 424 K; minimum 2.9% at 374 K |
| CBN | 364–424 K | 0.0199–5.678 Pa | 5.3% at 364 K; 5.1% at 424 K; minimum 3.0% at 384 K |
Did the authors check the measurement system against an independent reference?
Yes. Before using the new apparatus configuration for cannabinoids, investigators tested n-eicosane, a comparatively stable molecule with a well-established reference vapor-pressure relationship. At 373.91 K they obtained three replicate results of 5.427 ± 0.137 Pa, 5.766 ± 0.146 Pa and 5.399 ± 0.137 Pa, compared with a reference correlation of 5.636 ± 0.192 Pa. Those results help establish the apparatus’s suitability for low-volatility molecular metrology in that experimental setting.
Validation by n-eicosane is important because many cannabinoids can deteriorate under sustained heat. It does not mean a cannabis-user breath specimen was run, that the chemical method can distinguish every possible matrix interferent in collected breath, or that environmental smoke particles cannot influence collection. A forensic analyst should specify exactly which system was validated: the gas-saturation pressure-measurement process, not a real-life roadside drug detection workflow.
What do measured pressure and calculated pressure at breath temperature mean?
The original curves were fit to a thermodynamic correlation and then carried down from the 364–424 K measurement window to hypothetical exhaled-breath temperature 307.15 K and normal human body temperature 310.15 K. The source’s Table 5 lists projected cannabinoid vapor pressures at those lower temperatures. For exhaled-breath temperature, the fitted pressures were approximately 4.9 × 10⁻⁵ Pa for THC, 1.0 × 10⁻⁴ Pa for CBD and 2.3 × 10⁻⁵ Pa for CBN. None is a direct GC–FID observation made at 307.15 K.
That extrapolation crosses physical regimes. The 307.15 K and 310.15 K values are below the reported triple-point temperatures for CBD and CBN, so their calculated liquid-phase values represent supercooled-liquid conventions used for aerosol modeling. The investigators explained why a fluid respiratory environment can make that treatment useful for modeling, but it remains conditional. Importantly, the k = 1 relative uncertainty quoted for direct high-temperature measurements must not automatically be attached to the extrapolated pressure at exhaled-breath temperature.
How does vapor–aerosol partitioning translate pressure into a breath hypothesis?
In the atmospheric-chemistry framework the authors applied, an effective saturation concentration C* converts the modeled vapor pressure into a mass concentration of organic aerosol at which equal fractions of the cannabinoid would be in vapor and particle-associated phases. At 307.15 K, using assumed activity coefficient one, they obtained C* of about 6.0 µg/m³ for THC, 13 µg/m³ for CBD and 2.7 µg/m³ for CBN. These are modeled partition parameters, not observed breath THC concentrations or device reporting limits.
The equilibrium model represents aerosol-associated fraction as Faerosol = COA / (C* + COA), where COA is the assumed organic aerosol mass concentration. Increasing aerosol loading raises predicted particle association. Conversely, under low organic aerosol loading the model leaves a greater fraction of each cannabinoid in the gas phase. The calculation helps evaluate whether a filter that only captures particles might recover less than the total volatile and particle-associated molecule mass that was theoretically present.
What did the two aerosol-loading scenarios actually predict?
For a presumed background of 1.7 µg/m³ organic aerosol in exhaled breath, calculated particle fractions were THC 0.22, CBD 0.12 and CBN 0.38; complementary vapor fractions were 0.78, 0.88 and 0.62. A different hypothetical aerosol loading of 17 µg/m³, ten times higher, shifted predicted particle fractions to THC 0.74, CBD 0.57 and CBN 0.86. The corresponding gas fractions become THC 0.26, CBD 0.43 and CBN 0.14. Both sets are model results based on assumed aerosol and equilibrium conditions.
A breath collector consisting only of an aerosol filter could plausibly miss some gas-phase molecules if the actual respiratory system behaved like the low-loading model. But the source did not test a collection device side-by-side on the same human exhalation, measure real phase-specific mass in subjects, or quantify how often a real breathalyzer would report a false negative. The table compares projections; it should not be read as evidence of actual 22% or 74% cannabis detection efficiency.
| Analyte | Calculated pressure at 307.15 K | Calculated C* at 307.15 K | Modeled aerosol fraction at COA = 1.7 µg/m³ | Modeled aerosol fraction at COA = 17 µg/m³ |
|---|---|---|---|---|
| THC | 4.9 × 10−5 Pa | 6.0 µg/m³ | 0.22 (vapor 0.78) | 0.74 (vapor 0.26) |
| CBD | 1.0 × 10−4 Pa | 13 µg/m³ | 0.12 (vapor 0.88) | 0.57 (vapor 0.43) |
| CBN | 2.3 × 10−5 Pa | 2.7 µg/m³ | 0.38 (vapor 0.62) | 0.86 (vapor 0.14) |
Why does ambient particle loading change the interpretation?
The assumed amount of exhaled organic aerosol is not a universal constant. The study used 1.7 µg/m³ as an illustrative typical endogenous breath-aerosol concentration based on previous work, then explored a tenfold higher scenario because inhaled ambient particles could add to the organic aerosol pool. Ambient smoke, other air pollution, breathing maneuvers, respiratory physiology and sample handling could plausibly change the available particle material. This paper did not measure such variables in specific human breath collections; it deliberately tested their consequences in a mathematical scenario.
A collector designed to preferentially trap particles could therefore respond differently under circumstances in which particle load is high versus low—even if the amount of an analyte in a person’s respiratory system did not follow a simple fixed ratio. Sampling design should ultimately be evaluated with actual gas-phase and aerosol-phase measurements under controlled human conditions, including blanks, positive controls, environmental backgrounds and relevant periods after ingestion. Until that evidence exists, an exact percentage from this model is a premise for testing rather than a forensic correction factor.
Why is comparing cannabis breath with ethanol breath scientifically risky?
The authors’ Figure 6 places fitted THC, CBD and CBN pressure curves next to literature vapor-pressure curves for ethanol and linalool. At simulated breath temperature, estimated cannabinoid vapor pressures lie approximately eight to nine orders of magnitude below ethanol vapor pressure. That difference is one reason a THC measurement in human breath cannot be assumed to work like a mature evidential ethanol breath measurement. The chemistry and collection challenges are different, as are analyte levels and potentially the fraction associated with generated particles.
A breath-alcohol concentration is an analytical concentration in collected exhaled air; it can be subject to distinct legal and instrument-calibration rules. A pure-material cannabinoid saturation pressure in Pa is not a BAC and cannot be converted to a driver’s statutory whole-blood THC result by inserting a simple blood-to-breath ratio. Likewise, a calculated C* in µg/m³ is an equilibrium partition parameter, not an observed individual subject’s exhaled THC dose or a validated cannabis breath detection threshold.
Could CBD or CBN complicate a THC-specific breath measurement?
CBD and THC share a molecular formula and are structural isomers. CBN differs by degree of unsaturation. Similar molecules can create selectivity challenges for a field instrument unless its detection technology and chemistry discriminate them at practical concentrations and in relevant mixed matrices. The Beuning study measured these three reference substances separately and produced distinct vapor-pressure correlations, but it did not measure analytical cross-reactivity, device misidentification, or a clinical diagnostic confusion matrix in authentic breath.
The model even projects different gas-versus-aerosol fractions for the three analytes under the same assumed environmental loading. An instrument sensitive to particle-associated molecules may therefore have a different collection bias for CBD, CBN and THC. Whether that theoretical asymmetry materially biases any particular commercial apparatus requires independent paired-phase sampling, extraction recovery checks, chemically selective identification and blinded validation using actual human breath. The study provides a rationale for such tests, not their result.
What would a defensible breathalyzer validation have to measure?
A clinically or legally useful validation would first specify its intended analyte, specimen fraction, minimum detectable mass, collection duration, temperature, flow profile, storage stability and interference control. It would need independently characterized positive and negative human breath collections, preferably with paired particle-plus-gas recovery and chemically selective confirmatory methods. Blanks should represent ambient aerosol and recent environmental exposure as well as instrument carryover. Repeat collections should assess within-person and between-person variation and capture time course after known product administration.
Forensic validation would then have to separate analytical presence from the legally distinct propositions of recency, consumption route, blood concentration and behavioral impairment. A device could accurately detect a chemical species without predicting performance on a divided-attention task, accident responsibility or criminal mens rea. Conversely, a negative particle-only collection might not establish absence of all vapor molecules if its vapor recovery were not measured. This physical chemistry paper supports evaluating the sample pathway but contains no patient-level diagnostic accuracy or risk-prediction evidence.
Can these values show when someone last used cannabis?
No. The experimental cannabinoid was placed in a heated glass-bead saturator, not inhaled by a human participant. There was no controlled dose in people, no observed last-use time, no venous whole-blood sampling series and no real exhaled-breath longitudinal time course. A pressure–temperature correlation lacks the distribution, metabolism, redistribution and sample-contamination data required to estimate time since administration in a specific person. Even simultaneous measures of gas and aerosol in breath, if later obtained, would need independent pharmacokinetic validation before use as a last-use clock.
The article’s 307.15 K model should never be quoted as an individual’s measured level; the 364–424 K saturation results are not a measured route of inhalation. A real legal proceeding would require correct source specimen identification and independent case-specific toxicological data, including whether the submitted test detected THC itself or a different metabolite in a different matrix. Equations about equilibrium physical vapor pressure cannot fill gaps in those records.
Does a predicted missed vapor fraction mean a person was impaired—or not impaired?
Neither. This original study includes zero directly observed driving outcomes, zero psychomotor tests, zero human intoxication ratings and zero measured human whole-blood concentrations. The 0.22 and 0.74 particle fractions do not represent a percentage of impaired drivers, a false-negative rate, a legal safe level or a probability that someone’s drug use caused an incident. They also cannot show that a laboratory detector fails in 78% of people. Those would require entirely different evidence designs with participant denominators, method-specific outcomes, blinded comparisons and independently evaluated uncertainty.
A reported laboratory THC value from a future breath device may be relevant to exposure if the analyte, collection phase and assay are validated. It would not automatically establish individual impairment at a specific earlier event. Relevant forensic interpretation depends on verified timing, matrix, analytical method, uncertainty, potential contamination, clinical and behavioral observations, and applicable legal standards rather than on a single thermodynamic constant.
How should experts distinguish the physical matrices in a report?
Use the proper nouns for what was physically collected: pure compound in a heated saturator; vapor trapped from helium; aerosol particles hypothetically present in exhaled breath; actual human exhaled vapor; actual human breath aerosol; separately centrifuged plasma or serum; and unseparated forensic venous whole blood. They are not interchangeable. The present source directly generated cannabinoid vapor from a laboratory saturator and modeled partitioning; it did not receive any human breath or blood. Neither plasma concentrations nor urine metabolite results can be inferred from a pressure curve.
The same caution applies to units. Pressure is Pa, C* and aerosol loading are µg/m³, modeled fractions are dimensionless, and forensic blood testing frequently reports ng/mL. A numeric equality between these units without a valid physical and clinical model is meaningless. If a breath-related exhibit claims a 307 K THC pressure was measured directly or that gas-only THC signifies a legal whole-blood concentration, the original source does not support that claim. Describing each stage honestly preserves what the physical chemistry contributes without overstating it.
What are the strongest limitations of the original study?
The strongest strength is direct empirical high-temperature vapor-pressure data for three chemically similar cannabinoids under controlled conditions, with explicit standard-uncertainty accounting and an independent reference compound check. The central limitations are the absence of people, real breath collection and any comparison of two actual collection technologies in the same specimens. Correlations extrapolate well below the measured temperature interval, including supercooled-liquid assumptions for CBD and CBN; organic aerosol activity coefficient is assumed to be one rather than determined in lung lining fluid.
The model does not independently quantify a person’s organic aerosol loading, gas–particle equilibrium kinetics, sampling flow, ambient pollution, phase-selective collection efficiency or material losses in a real breathalyzer. No observed group null p-value for driving or clinical detection is available because such experiments were not part of the paper. Future evidence should test real paired gas-and-aerosol phases, how much a collector retains from each, and whether phase changes alter measured cannabinoids in clinically relevant conditions. Such validation cannot be supplied by adding precision to a model’s assumed values.
What is the scientifically bounded conclusion for forensic reporting?
The empirically supported conclusion is that THC, CBD and CBN have measurable saturated vapor pressures when evaluated as laboratory reference substances at 364–424 K, and that fitted lower-temperature pressures place them in a region where aerosol–vapor partitioning is worth examining. In the authors’ illustrative partition model, a low assumed organic aerosol load gives more THC in vapor than on particles; a higher load reverses that prediction. It is therefore scientifically reasonable to investigate whether collection systems that capture only breath aerosol miss a share of molecules present in the gas phase.
The legally and clinically crucial boundary is equally clear: there is no directly measured human breath THC fraction, validated device false-negative rate, individual venous whole-blood THC equivalent or impairment conclusion in this article. The original full paper is Beuning et al., “Vapor pressure measurements on Δ9-tetrahydrocannabinol, cannabidiol, and cannabinol to inform cannabis breathalyzer development” (doi:10.1088/1752-7163/ae3794). The new evidence helps frame better experiments; it does not replace them.
Seven frequently asked questions
Did the researchers analyze THC in human exhaled breath?
No. They measured pure cannabinoid standards in a heated gas-saturation apparatus, recovered vapor from helium with a cooled trap and quantified it by GC–FID. No volunteers exhaled into that system. The exhaled-breath fractions in the paper are computed predictions, not observed clinical specimens. It is inaccurate to describe the study as validating an actual handheld roadside breathalyzer or as establishing performance in a population of cannabis users.
Is the 78% predicted THC vapor fraction a measured device miss rate?
No. The number is the complement of the calculated THC particle-associated fraction of 0.22 when organic aerosol is assumed to be 1.7 µg/m³. With a different assumed organic aerosol concentration of 17 µg/m³, the predicted vapor fraction is only 0.26. Neither value comes from counting successful or failed tests, and neither includes instrument collection efficiency, chemical discrimination, laboratory limits of detection or real subjects. Therefore neither is an actual false-negative probability.
Did the authors really measure THC vapor pressure at body temperature?
No. The measured temperatures were 364–424 K. The source’s THC vapor pressure at 307.15 K, 4.9 × 10⁻⁵ Pa, is a correlation-based extrapolation. Its value may be useful for a model, but it should be explicitly labeled projected. Measured-pressure k = 1 uncertainties of several percent over 364–424 K cannot be casually transplanted onto an extrapolation at lower temperature.
What was the difference between THC, CBD and CBN in the model?
At assumed exhaled-breath temperature, the model estimated effective saturation concentrations of 6.0 µg/m³ for THC, 13 for CBD and 2.7 for CBN. Under the same 1.7 µg/m³ aerosol assumption, the calculated aerosol fractions were respectively 0.22, 0.12 and 0.38. These differences suggest analyte-specific phase recovery deserves study. They do not establish which molecule a particular analyzer confuses with THC or its specificity in real breath.
Does detecting THC in breath prove recent smoking or impairment?
A properly confirmed breath THC result might inform exposure in a separately validated context, but this paper supplies neither a human breath drug clock nor a test of driving performance. It did not administer THC to persons or measure blood THC. Even a chemically correct measurement of an analyte should not be transformed into person-specific impairment, incident causation or a legal concentration threshold without independent relevant evidence.
Can an ethanol blood-to-breath ratio be applied to these cannabinoid pressures?
No. The tested endpoint was cannabinoid saturated vapor pressure in Pa, with low-temperature modeling and hypothetical aerosol loading. That is a different concept from end-expiratory ethanol concentration and its legally governed interpretation. The authors estimate cannabinoid pressures eight to nine orders of magnitude below ethanol at breath temperature. A simple ethanol-style conversion to legal whole-blood THC would disregard matrix, phase, partitioning and drug-specific physiology.
What experiment would most directly test the proposed sampling issue?
A controlled study collecting actual human breath using paired phase-selective methods, one measuring retained aerosol THC and another measuring gas-phase THC, would directly address recovery. It should measure environmental backgrounds, temperature, flow rates, validated analytical selectivity, storage stability, matrix effects, person-level denominators and time after documented administration. Only then could investigators estimate how often an aerosol-only design misses actual breath THC under particular circumstances. That experiment was not part of Beuning et al. 2026.
Source and related reading
Original primary research: Beuning CN, Berry JL, Paulechka E, Huber ML, Jeerage KM, Widegren JA, Lovestead TM (2026). Vapor pressure measurements on Δ9-tetrahydrocannabinol, cannabidiol, and cannabinol to inform cannabis breathalyzer development. Journal of Breath Research 20(1):016010. https://doi.org/10.1088/1752-7163/ae3794.
For additional background on why physical specimen identity matters, see biological matrices in toxicology. For chemical confirmation and selectivity concepts, consult confirmatory drug testing. For the collection-stage distinctions at issue here, see specimen collection in toxicology.