Formulation & Safety

Building a Safe Essential Oil Mosquito RepellentFrom Chemistry to Skin

16–19 min read•Formulation & Safety•Updated July 2026

Reviewed by Quentin Olagne — Certified Aromatherapist (NAHA) — July 5, 2026

Part 2 of the Aroma Compass mosquito series. This article addresses one specific point of intervention in that system: the skin boundary.

Part 1 covered mosquito biology, essential oil chemistry, and the evidence behind repellency mechanisms. This one doesn't hand over a recipe — it shows where and how a formulation system is constrained, and what its outputs tend to look like once those constraints are applied.

Table of Contents▾

Two Points of Intervention The Layering Model The Carrier Phase System Limits Evidence Tiers Decision Framework Formulation Archetypes System Outputs in Practice Application Logic Boundary Ceiling vs. Synthetics Where This Leaves the System Aroma Compass Knowledge Map

Formulating an essential oil mosquito repellent

Two Points of Intervention

Before formulating anything, it helps to name where a repellent strategy actually acts on the mosquito–human system. There are at least two distinct intervention points, and they are not interchangeable:

The boundary layer — the skin surface itself, and the thin film of air immediately above it. This is where topical repellents operate.
The ambient layer — the air volume of a space (a terrace, a room, a seating area), which diffusion-based systems (candles, diffusers, area treatments) act on instead.

One assumption worth correcting before going further, since it shapes how the rest of this article should be read: common interpretation — a diffuser or citronella candle placed in an outdoor seating area creates something like a protective shield around the people sitting there. System reality— ambient treatments reduce mosquito activity within a concentration-dependent radius that falls off with distance, wind, and airflow; it's a local pressure reduction, not a boundary. What this does not mean— someone sitting near the edge of a “treated” table isn't standing just outside a shield; they're already in a zone where effective concentration has dropped, with a correspondingly smaller effect.

This article is scoped entirely to the boundary layer. A future article in this series addresses the ambient layer directly. The two are not competing solutions to the same problem — they're different interventions with different constraints (mobility, group vs. individual coverage, reapplication burden), and the right one depends on the situation more than on which is “more natural” or “stronger.” That comparison belongs in the decision framework below, not resolved here.

The Layering Model, as an Interacting System

A topical blend is often described — including in earlier drafts of this article — as a relay: a fast layer hands off to a medium layer, which hands off to a fixative. That framing is easy to visualize but overstates how cleanly these roles separate in practice. What's actually happening is closer to an interacting system, where the outcome depends on species, environmental conditions, and the specific formulation matrix — not a simple sum of parts.

With that caveat placed up front, the functional roles are still useful as a design vocabulary:

Fast-acting volatiles

Small monoterpene molecules (citronellal, citronellol) evaporate quickly off skin, producing measurable vapor-phase activity in the first 20–40 minutes in controlled test conditions. How this translates to a given person's skin, in a given climate, is variable.

Human expectation	System behavior
Strong scent = strong, ongoing protection	Scent-detection threshold and repellency-effective concentration are different thresholds — scent often outlasts the concentration actually doing any repelling

Moderate-persistence compounds

Somewhat heavier alcohols (geraniol, linalool) show repellency extending further — commonly cited controlled studies report protection in the 60-minute range at test concentrations — but "extending further" describes a tendency observed across studies, not a guaranteed timeline for skin application at safe dilutions.

Slow-evaporating constituents

Woody and resinous compounds (cedarwood and similar) evaporate slowly. In perfumery, this evaporation-rate difference is well established and is the basis for the "fixative" concept — it's real physical chemistry. What is not well established is that this slower evaporation reliably extends repellency specifically, as opposed to simply extending the presence of scent. These are related but distinct claims, and the evidence base for the first (evaporation physics) is much stronger than for the second (fixatives measurably lengthening skin protection time). Both should be held with different levels of confidence.

One evidence distinction worth being explicit about

Some of the strongest data behind geraniol's reputation comes from ambient diffuser trials — geraniol suppressing mosquito activity in air, measured at a fixed distance from a diffuser. That is a boundary-layer-irrelevant result; it describes the ambient layer, not skin. When evidence from one intervention point gets cited to support claims about the other, that's a conflation worth catching rather than repeating.

Reality check: this is part of why geraniol's reputation reads as stronger than boundary-layer use alone would produce — the ambient-layer result is real, it's just answering a different question than "how long does this last on skin."

The Carrier Phase: A Second Design Axis

Everything above concerns the volatile, aromatic fraction of a blend. But a topical formulation has a second, independent axis: the non-volatile carrier or fixed-oil phase, which doesn't evaporate on the same timescale at all and instead sits on the skin as a film.

This phase isn't inert. Neem oilis a useful example of why. Neem functions here purely as a fixed-oil carrier — its role in this article is physical: it forms a film on the skin, slows the departure of the volatile actives sitting in it, and behaves as an occlusive layer that changes how the whole system evaporates. Neem's separately documented biochemical interaction with insects (covered in Part 1) is a different property entirely, operating through a different mechanism, and this article isn't the place to expand on that claim or attach a formulation percentage to it. The point of mentioning neem here is narrower: it illustrates that “carrier” is a design decision with its own weight, not a neutral vehicle standing in the background of the volatile phase that gets all the attention.

This is the underlying reason two blends with identical essential oil ratios can behave differently on skin — the carrier phase they're suspended in is doing real work.

System Limits

Every formulation choice above operates inside a set of hard constraints. These aren't advisory suggestions — they're the thresholds the rest of the system has to be designed within.

Dermal tolerance is not uniform across the body or across individuals. A concentration unremarkable on a forearm is not the same input on facial skin, which has different absorption and reactivity characteristics.
Aldehyde-rich constituents (citral, citronellal — found in lemongrass, citronella, melissa) carry the highest sensitization risk in this oil category. Sensitization is cumulative and, once crossed, does not reverse. This sets a hard ceiling on how much weight the “fast-acting volatile” role in Section 2 can be given before the system becomes unsafe rather than just less effective.
Population-specific thresholds exist.General aromatherapy safety convention places children, pregnant users, and reactive-skin populations at roughly half the general adult dilution ceiling. This is a system boundary, not a strength preference — going above it isn't “more protection,” it's outside the tolerance envelope for that population.
Study concentration and safe skin concentration are two different numbers. Several published repellency trials use high concentrations to produce measurable lab protection times — geraniol, cited as a high-evidence reference in the section below, was tested at 25%in Barnard & Xue (2004), putting it 8–25× above this system's 1–3% ceiling. That figure describes an experimental input, not a boundary this system can operate within for repeated daily skin contact. The gap is structural, not a rounding difference — and for geraniol specifically, we don't actually know what protection time to expect at 1–3%, because no one has published that trial at skin-safe concentrations.

The catnip exception

Catnip (Nepeta cataria) is the one oil in this series where tested study concentrations and the safe skin ceiling actually overlap. Zhu et al. (2006) measured 1–6 hours of protection at 2–4% against Aedes albopictus — concentrations that sit at or just inside this system's 1–3% ceiling. For geraniol and citronella, we don't know what their real-world efficacy looks like at safe dilutions; for catnip, we have a directly usable data point. That's rare in this category — and it's why it holds a central position in the evidence table below.

Common interpretation:because these are natural ingredients, there's effectively no ceiling — more essential oil simply means safer, stronger protection. System reality:the thresholds above are chemistry properties of specific molecules, not a synthetic-versus-natural property; they apply the same way regardless of origin. Raising concentration past these ceilings doesn't move the system toward more protection — it moves it toward the sensitization threshold described above.

Personalized formulation

The framework is general; your situation isn't. A consultation adapts it to you.

Adapt this framework to me

Evidence Tiers, With Intervention Point Marked

Reprising Part 1's evidence review, but marking which intervention layer each result actually describes:

Tier	Compounds	Evidence type
High-evidence	Geraniol	Strong ambient-layer diffuser data; moderate boundary-layer (skin) protection-time data — primarily sourced from 25% concentration studies, well above the safe ceiling
High-evidence	Citronellal / citronellol	Boundary-layer protection-time data, concentrated in first-hour window — study concentrations often 5–25%, well above the safe ceiling
High-evidence — safe-dilution data available	Catnip (nepetalactone)	1–6 h protection measured directly at 2–4% against Aedes albopictus (Zhu et al., 2006) — the only oil in this table tested at or below the system's 1–3% ceiling
Moderate	Linalool	Boundary-layer data, generally shorter duration than geraniol in comparative trials
Support / contextual	Cineole-rich eucalyptus, woody oils	Limited independent repellency evidence; contribute primarily through evaporation-rate modulation and odor masking, not standalone activity

Two structural patterns matter more than any individual oil's ranking:

Volatility accounts for most real-world variation— an oil's evaporation rate under actual outdoor conditions (heat, humidity, movement) matters as much as its raw repellency ceiling in a still-air test.
The delivery system changes the outcome independently of the active oil.The same citronella oil has produced substantially different repellency measurements as a candle, a diffuser, and a skin-applied cream in separate trials — meaning “citronella” alone is not a complete description of what's being tested.

Decision Framework: Where to Intervene

This is the layer missing from a pure ingredient-and-example approach. The question this framework answers isn't "which blend do I use" — it's "where in the system does it make sense to intervene at all."

Axis A — Exposure duration & intensity

Brief/incidental exposure ↔ Extended, active exposure (gardening, outdoor dining at dusk)

Axis B — Mosquito pressure context

Low-density urban ↔ High-density Mediterranean evening conditions

Axis C — Population tolerance

General adult ↔ Sensitive skin / child / pregnant (general-guidance tier only; defers to a qualified practitioner)

Axis D — Reapplication tolerance

Comfortable reapplying hourly ↔ Wants a single low-maintenance intervention

Axis E — Intervention layer

Skin (boundary) ↔ Ambient (diffusion) ↔ Combined

Axis E is the one that keeps this from collapsing back into “which recipe.” A reader positioned toward high mosquito pressure, extended exposure, and low reapplication tolerance is being told something about where to intervene, not just whatto apply: a skin-only boundary-layer approach is a poor fit for that combination regardless of which oils are in it, because the constraint is structural (evaporation and reapplication burden), not a matter of picking a better blend. That same profile might point toward combining a boundary-layer layer for the individual with an ambient-layer treatment for the group space — a combined intervention rather than a stronger single one. A reader positioned toward brief exposure and high reapplication tolerance may find boundary-layer intervention alone sufficient, and doesn't need the ambient layer at all.

Reality check: moving toward “combined” here doesn't mean layering on more topical concentration — the ceiling in Section 4 doesn't move just because a second layer is in play. It means letting a different layer of the system (Article 3) carry the part the skin can't safely take on.

This framework doesn't resolve the ambient layer's specifics — that's Article 3's scope — but it establishes that the choice of intervention point precedes the choice of ingredients.

Formulation Archetypes: What the System Converges Toward

The framework in Section 6 tells you where to intervene. This section names what a boundary-layer intervention tends to look like once the axes have been worked through — not as formulas, but as recognizable behavioral patterns. Think of these as convergence points, not blends: no ratios, no exact instructions, no step-by-step construction. Each archetype is defined by which functional role from Section 2 dominates, and why that dominance is the correct output for a given axis position — the "why" is the actual content here, not the ingredient list.

The fast-response archetype (citronellal/citronellol-dominant)

This is what the system converges toward when Axis A points toward short exposure and Axis D shows high reapplication tolerance. The dominant behavior is front-loaded activity: strong initial vapor-phase effect, acknowledged short duration, and an assumption that the user will simply reapply rather than needing the blend to last. This archetype trades duration for immediacy on purpose — it isn't an incomplete version of a longer-lasting blend, it's the correct shape for a short-exposure use case.

The balanced archetype (geraniol-centered)

This is what the system converges toward at moderate-to-high pressure with moderate exposure duration and moderate reapplication tolerance — the most common real-world position on the axes. The dominant behavior here isn't the fastest onset or the longest tail, but the flattest curve: a steadier presence across the middle of the exposure window, at the cost of the sharp initial peak the fast-response archetype has. This is usually the default a boundary-layer-only strategy lands on when no single axis is extreme.

The low-sensitization archetype (linalool-type, aldehyde-excluded)

This is what the system converges toward whenever Axis C points toward a sensitive population, regardless of where the other axes sit. The dominant behavior is a deliberately narrowed ceiling: shorter duration and a milder peak are accepted outcomes, not failures, because the population constraint from Section 4 overrides pressure or duration considerations entirely. This archetype exists to demonstrate that Axis C can dominate the other four axes rather than being averaged against them.

The combined-intervention archetype

This is what the system converges toward when Axis E resolves to "combined" rather than "boundary-layer only" — typically high pressure, extended exposure, and low reapplication tolerance stacked together, the exact combination Section 6 flagged as a poor fit for boundary-layer intervention alone. Its defining behavior is that the boundary-layer component is deliberately not pushed toward its own ceiling to compensate; instead, it holds at a moderate, sustainable profile (closer to the balanced archetype) while the ambient layer — outside this article's scope — carries the burden the skin layer can't safely take on alone. This archetype is where Article 3 picks up.

These four aren't an exhaustive taxonomy — they're the recognizable shapes the five-axis framework tends to produce, useful as a reference point for recognizing which behavior a given context calls for before any specific oils or carriers are chosen.

System Outputs in Practice (Illustrative Micro-Formulations)

The axes and archetypes above are the system's working parts. This section runs two concrete households through them, to make the mapping from variable to lived situation less abstract. What follows describes what a system output would tend to look like — not a recipe, not a ratio, not a step-by-step instruction.

Hugo

Ground-floor apartment with a small garden, peri-urban area. No known skin sensitivities. He works from home some days, so has easy access to products and time to reapply. His bedroom window stays cracked open through summer for heat relief.

Scenario A — Saturday dinner, garden terrace, dusk onward

Moderate pressure, defined concretely: mosquitoes appear intermittently after sunset; repeated landing attempts, uncomfortable if ignored, but not constant enough to make sitting outside for the evening unworkable.

AA few hours at dusk outdoors, moderate.

BModerate — repeated landing attempts, not yet a swarm.

CNo constraint, general adult.

DHigh — he's hosting from home, product on hand.

EIndividual boundary layer likely enough for him; the shared table is a separate question.

System convergence

Balanced archetype (geraniol-centered) — moderate pressure and duration, high reapplication tolerance — no axis is extreme, exactly the position Section 7 defaults to for this archetype.

Micro-formulation output (illustrative only)

A geraniol-centered profile carrying the steady middle of the evening, with a short citronellal/citronellol opening phase to cover the first few minutes after sitting down, before geraniol's flatter curve takes over.

Intervention decision

Boundary-layer sufficient for Hugo himself. The open question isn't about strengthening his blend — it's whether the shared table, as a space rather than a set of individuals, is being addressed at all. That's an ambient-layer question this article doesn't resolve.

Scenario B — Asleep, window cracked, Culex pipiens active overnight

Goal is uninterrupted sleep through a 6–8 hour window. Long exposure, no reapplication possible once asleep.

ALong — the full night.

BModerate-to-high, Culex pipiens active after dark.

CNo constraint for Hugo.

DEffectively zero — nobody reapplies at 3am.

EExactly the combination Section 6 flags as a poor fit for boundary-layer alone.

System convergence

No archetype closes this gap — the layer's own depletion timeline (roughly 60–90 minutes, Section 9) is structurally shorter than the exposure window; this isn't a potency problem, it's a duration mismatch.

Micro-formulation output (illustrative only)

Describing what the best possible boundary-layer blend would look like isn't productive here: even a version pushed toward its longest, most realistic tail — a geraniol-centered profile stretched as far as it goes — still lands far short of the exposure window. It isn't the blend's structure that's missing here; it's that the structure of the problem exceeds what this layer can offer.

Intervention decision

Boundary layer not sufficient. This falls entirely outside boundary-layer scope — not a “stronger blend” problem but an intervention-layer problem, pointing toward the ambient layer Article 3 covers.

What this teaches us

In Scenario A, the dominant constraint for Hugo is soft: a coverage-scope question (himself alone versus the whole table), not concentration. In Scenario B, it's structural: a duration gap that reapplication can't close during sleep. A "stronger blend" isn't the solution to Scenario B, because more essential oil doesn't extend the exposure window — the same short-duration ceiling (Section 9) applies no matter how the blend is composed. What matters here is recognizing which layer of the system the problem belongs to, not formulating it better.

Delphine

House near a canal in a high tiger-mosquito-density part of Southern France, mature garden. Reactive skin herself, and a young child in the household — so the population constraint (Axis C) applies at the stricter end for at least one person present at all times. No air conditioning, so bedroom windows stay open through summer nights. Often hosting or managing a child single-handedly in the evening, which leaves little bandwidth for repeat applications.

Scenario A — Saturday dinner, garden, dusk onward

Mosquitoes are present continuously; landing attempts start within minutes of sitting down, and staying outside unprotected for the evening isn't realistic.

AExtended, the full evening.

BHigh — continuous presence.

CSensitive population at the table (herself, plus the child nearby).

DLow — hosting or managing the child alone.

EThe same axis stack Section 6 names directly as a poor fit for boundary-layer-only.

System convergence

Combined-intervention archetype — high pressure, extended exposure, and low reapplication tolerance stacked together — the exact combination flagged in Section 6.

Micro-formulation output (illustrative only)

Boundary-layer coverage held to a moderate, low-sensitization profile — linalool-leaning, with a restrained aldehyde load due to skin sensitivity constraints — rather than a geraniol blend pushed toward its own ceiling to try to compensate for pressure. The population ceiling doesn't move just because mosquito pressure is high.

Intervention decision

Boundary layer alone not sufficient. The table itself, as shared space, is where an ambient-layer treatment would carry the load the skin layer structurally can't — intensifying the topical blend here runs directly into the Section 4 ceiling rather than solving the pressure problem.

Scenario B — Delphine and child asleep, windows open, high-density night biting pressure

ALong, the same as Hugo's sleep scenario.

BHigh — strong night density, tiger-mosquito area.

CNarrows independently of duration: the child is present, which lowers the ceiling before duration is even considered.

DZero — everyone is asleep.

ETwo independent constraints point to the same answer, rather than a trade-off between them.

System convergence

No topical archetype closes this gap — nothing is being traded off between axes here, since duration and population each independently rule out boundary-layer-only.

Micro-formulation output (illustrative only)

Describing a blend shape here would suggest an archetype could be picked to fit — when the actual finding is that none can. This scenario is precisely why the ambient layer exists as a separate design category.

Intervention decision

Ambient layer necessary; boundary layer not sufficient, regardless of which composition is chosen. Details aren't covered here — that's Article 3's scope.

What this teaches us

The dominant constraint for Delphine is the population ceiling (Axis C), reinforced rather than created by duration in Scenario B. A "stronger blend" isn't the solution, because the ceiling in Section 4 doesn't move with pressure — intensifying concentration to fight high mosquito density would move Delphine, and her child, toward the sensitization threshold rather than toward more protection. What the system does here is refuse to treat pressure as a problem the skin can solve on its own.

A systems perspective spends less effort picking "the best archetype" and more effort identifying which axis is actually the binding constraint — the one no amount of adjusting the others will move. The same physical space can also demand different answers for different people in it: a host's own blend can sit at the balanced archetype while the shared table they're sitting at remains a separate, unresolved coverage problem. This is offered as a way of reading the system, not a ranking of who "needs it more."

Key takeaway

This system's outputs aren't recipes — they're structural patterns the five axes produce once applied to a real situation.
Hugo illustrates a soft constraint (coverage scope) followed by a hard one (duration/reapplication) that no blend alone can close.
Delphine illustrates a population constraint that dominates the other axes, reinforced rather than created by duration in her sleep scenario.
In both households, "strengthen the blend" is never the answer once the binding constraint is structural rather than chemical.

Application Logic

Because this article is scoped to the boundary layer, reapplication is a structural property of that layer specifically — it is the cost of intervening at the skin rather than the ambient layer. Volatile actives leave the skin faster than they leave a treated air volume, which is precisely why ambient-layer approaches (Article 3) don't carry the same reapplication burden, at the cost of not moving with the individual.

Sweat, wind, and activity level accelerate the boundary layer's depletion beyond what still-air lab conditions predict. This isn't a formulation flaw to solve — it's an inherent property of intervening at this particular layer of the system.

Common interpretation: one application in the morning provides protection for the rest of the day, the way sunscreen is sometimes (also incorrectly) assumed to work. System reality: at boundary-layer-safe dilutions, active volatiles typically deplete within roughly 60–90 minutes under real exposure conditions — considerably faster than synthetic options built specifically to resist evaporation. What this does not mean:it doesn't mean the blend has failed after 90 minutes. It means the boundary layer is behaving exactly as this intervention point is expected to, and reapplication is the normal maintenance cost of operating here — not a sign something went wrong.

Boundary-Layer Ceiling vs. Synthetic Reference Points

DEET, picaridin, and PMD remain the reference standards for boundary-layer protection duration, with the largest evidence base for extended or high-pressure exposure. Essential-oil-based boundary-layer systems, even well-designed ones, operate within a shorter and more variable duration ceiling, and carry a sensitization consideration synthetic options largely don't share in the same form. This isn't a case to be argued around — it's a structural difference in what each system is built to do, and it's the reason Axis E in the decision framework exists at all: sometimes the right answer to “how do I get more protection” isn't a better essential oil blend, but a different intervention point entirely.

Reality check: "natural" is often read as "automatically gentler," but the sensitization risk from aldehyde-rich oils (Section 4) is a real, documented tradeoff with nothing to do with synthetic-versus-botanical origin — it's a property of specific molecules, present in this system the same way it would be in any other.

Where This Leaves the System

This article has only addressed one layer of a larger structure. It hasn't addressed how ambient-layer interventions change the constraints above, and it hasn't addressed what happens after the boundary layer fails and a bite occurs — a different problem, operating on a different part of the system entirely. It also hasn't fully corrected the intuition flagged in Section 1 — that a treated space works like a shield. Article 3 takes that on directly.

What this article has established: the boundary layer has real, evidence-supported tools, a hard set of tolerance limits that cap how much can be asked of it, and a specific point at which its structural limitations (reapplication burden, duration ceiling under high pressure) mean the right move is to look at a different layer of the system rather than intensify this one.

This article describes formulation logic, not personalized guidance. Population-specific thresholds referenced above (children, pregnant users, sensitive skin) are general system boundaries, not a substitute for a qualified aromatherapist or healthcare provider's assessment of an individual case.

Series Map

Aroma Compass Knowledge Map

This article is one layer of a structured series on mosquitoes and essential oils. To help you navigate:

Article 1 — Mechanism & Evidence Layer

Covers mosquito biology, the chemistry behind repellency, and how to read lab vs. field research.

Read Article 1 →

Neem Oil and Mosquitoes — Botanical Case Study

A focused look at one specific plant oil, its chemistry, and its evidence base — including its role as a fixed-oil carrier, discussed again in Section 3 here.

Read the case study →

This article — Formulation & Safety Layer

Building a Safe Essential Oil Mosquito Repellent. Translates the chemistry from Article 1 into a decision framework, archetypes, and safety limits for the boundary layer — no ratios or recipes.

Calculate a dilution →

Future: Diffusion Systems (forthcoming)

Will look at the ambient layer in depth: device types, room dynamics, and outdoor limitations — the layer that closes the gaps identified in this article's sleep scenarios.

Read in this order — mechanism, case study, formulation, diffusion — for the clearest path through the topic.

A framework for repellents, a formula for you

This article gives you a decision framework and general safety limits — but your skin, age, and real exposure change what actually makes sense to use. A consultation applies this same framework to your specific, human situation.

Adapt this framework to me

Next step