Preservatives in an oil phase: solubility, dose and the products that need one

Phenoxyethanol at up to 1 percent, benzyl alcohol, caprylyl glycol and the pH dependent acids compared on oil solubility, dose and whether a balm needs them at all.

A preservative kills things that live in water, so putting one into a product that contains no water is a strange act on the face of it. Sometimes it is the right one anyway, and sometimes it is money set on fire. This page sorts the materials that will actually dissolve in an oil phase, gives the doses and the legal caps, and works out with numbers how much of a dose ever reaches the place where an organism could be.

• EU cap: phenoxyethanol 1.0%
• Sorbic acid 0.6% as acid
• Only works in water
• ISO 11930 challenge test

Which anhydrous products actually need one

The general case is settled elsewhere: a product with no free water has nothing for bacteria, yeast or mould to grow in, which is the argument laid out in do balms need preservatives. What remains is the list of exceptions, and they divide cleanly into products that contain water because of an ingredient and products that meet water because of how they are used.

Two rows deserve a second look. Honey is stable in its own jar because its water activity is low, but that self-protection does not survive dispersion through an oil phase, as honey in balms explains. And a scrub is worse than a cleansing balm, not better: sugar is hygroscopic and a nutrient, and dissolved at the right dilution it is a growth medium.

Where a preservative has to be to do anything

Micro-organisms live in water. A preservative kills or inhibits them by reaching a concentration in that water above its minimum inhibitory concentration for the organism in question. Everything else about the molecule is logistics.

In a two phase product the molecule distributes itself between oil and water according to its partition coefficient, usually written K, the ratio of its concentration in oil to its concentration in water at equilibrium. The concentration that matters is then straightforward algebra. If the product is a fraction Fw water and Fo oil, and the nominal dose is C, then the concentration in the water phase is:

C_water = C / (F_w + K F_o)

An anhydrous product contaminated in use is roughly 99 percent oil and 1 percent water. Put those numbers in and the result is uncomfortable.

Read the top two rows and the bottom two together, because the result is the opposite of the folk rule. The preservative you want in an oil-continuous product is a water-loving one, since it concentrates thirty to fifty fold in the small water phase where the organisms are. The oil-loving one dissolves beautifully into your formula and then stays there, a hundred to three hundred times too dilute in the only place it could act. "Use an oil-soluble preservative in an oil product" gets the physics exactly backwards.

The trap is that the hydrophilic salts will not enter an oil phase at all. Potassium sorbate and sodium benzoate are crystalline solids with water solubilities near 60 g per 100 mL and no meaningful oil solubility. Added as powder they give you grit and false confidence, because a particle that never dissolves protects nothing a millimetre away from itself. What will and will not dissolve in a balm is set out in solubility in anhydrous formulas.

Phenoxyethanol, the default

Phenoxyethanol is the single most common preservative on a cleansing balm label, and with reason. It is a clear liquid, miscible with most cosmetic esters and vegetable oils, effective across pH 3 to 10 so it does not care what the arriving water is like, and capped at 1.0 percent in the EU and UK under Annex V of Regulation 1223/2009. The SCCS reaffirmed that 1.0 percent is safe for all consumers, including children, in its opinions of 2016 and 2021.

Its water solubility of about 2.7 g per 100 mL at 20 C is the useful figure: it is why phenoxyethanol partitions imperfectly rather than hopelessly. It is weak alone against Pseudomonas, so it is nearly always sold blended with ethylhexylglycerin, caprylyl glycol or a sorbate.

Now the arithmetic. At the legal maximum of 1.0 percent in a 99 percent oil product, the table above gives about 0.07 percent in the water phase. Published minimum inhibitory concentrations for phenoxyethanol against common test organisms sit in the region of 0.25 to 0.5 percent. The dose that reaches the water is therefore three to seven times short, and you cannot fix it by adding more, because 1.0 percent is the ceiling.

One practical note: phenoxyethanol dissolves readily in esters and polar vegetable oils but is less happy in a pure hydrocarbon phase, and in a mineral oil rich cleansing balm it can haze. Add it below 40 C, and if it hazes, bring in a polar ester from cosmetic esters to carry it.

Benzyl alcohol and the pH dependent acids

Benzyl alcohol behaves much like phenoxyethanol: capped at 1.0 percent as a preservative, oil miscible, water soluble to about 4 g per 100 mL, indifferent to pH. It carries one extra obligation, since it is one of the fragrance allergens that must be named separately on an EU or UK ingredient list above 0.001 percent in a leave-on product, which is covered in fragrance allergen labelling.

The acids are a different proposition. Sorbic acid, benzoic acid and dehydroacetic acid only work in their undissociated form, the neutral molecule that can cross a cell membrane. The dissociated anion does almost nothing. How much of the acid is undissociated is set entirely by pH.

Two consequences follow, and the second is the one that ends the argument. First, an anhydrous product has no pH. pH is a property of an aqueous solution, and a meter pushed into a balm reads nothing meaningful, so there is no acidity to adjust and no buffer to add. Second, whatever water does arrive brings its own pH with it, and tap water is neutral to alkaline. Stack that on the partition loss and sorbic acid at its 0.6 percent cap delivers around 0.03 percent to the water, of which under one percent is in the active form: something like a few parts per million against a requirement one to two orders of magnitude higher.

The acids also have poor solubility on both sides. Sorbic acid dissolves to only about 0.16 g per 100 mL in water and needs warming to go into oil; dehydroacetic acid is similar. Their sodium and potassium salts are freely water soluble and effectively oil insoluble. There is no version of this family that both mixes into a balm and works in the water that reaches it. The same logic applies to the marketing-friendly organic acid blends sold as natural preservatives, whichever acid they are built on.

Caprylyl glycol and the multifunctionals

Caprylyl glycol, ethylhexylglycerin, glyceryl caprylate and 1,2-hexanediol are the materials most often described online as natural preservatives for anhydrous products. None of them is a preservative in the legal sense. EU Regulation 1223/2009 defines a preservative as a substance intended exclusively or mainly to inhibit micro-organisms, permits only substances on Annex V to be used as one, and none of these four is on Annex V.

They are not useless, though. All four are amphiphilic: an oil-friendly chain attached to a small polar head. That structure lets them sit at an oil and water interface and destabilise microbial membranes there, and it lets them make a conventional preservative more effective at a lower dose. The technical name for the effect is potentiation, and it is why a phenoxyethanol blend at 0.8 percent can outperform straight phenoxyethanol at 1.0 percent.

Dose bands and legal caps

Annex V of Regulation 1223/2009 is a positive list: if a material is being used to preserve a cosmetic in the EU or UK, it has to be on that list and used within its stated conditions. The list is amended regularly, so check the current consolidated text rather than a copy in a supplier catalogue.

The United States runs the same materials under a different system. There is no positive list of permitted preservatives; a cosmetic ingredient other than a colour additive needs no pre-approval, and the manufacturer carries the duty to substantiate safety. Since the Modernization of Cosmetics Regulation Act of 2022, that duty is explicit and comes with facility registration, product listing and adverse event reporting, set out in MoCRA duties. In practice American formulators use the CIR safety assessments as their numbers, and those conclusions broadly track the EU caps, so a 1.0 percent phenoxyethanol product is normal on both sides of the Atlantic. Individual states go further, and several preservatives banned by California's cosmetics legislation are still lawful federally. The labelling differences are in labelling cosmetics in the US.

Proving it: ISO 11930 and what a challenge test costs

None of the reasoning above substitutes for testing. The standard method is a preservative efficacy or challenge test to ISO 11930, or USP chapter 51 in the United States. A laboratory inoculates the finished product with a defined panel, Staphylococcus aureus, Pseudomonas aeruginosa, Escherichia coli, Candida albicans and Aspergillus brasiliensis, at around 10⁵ to 10⁶ colony forming units per gram, and counts survivors at 7, 14 and 28 days.

ISO 11930 sets two outcomes. Criterion A is the full pass: at least a 3 log reduction in bacteria by day 7, held to day 14, with no increase at day 28, and at least a 1 log reduction in yeast, with no increase thereafter. Criterion B is a lower bar that may be accepted only when the risk assessment shows other protective factors, typically packaging that excludes contamination. A cleansing balm in an airless pump can rest on criterion B in a way that the same formula in an open jar cannot.

Practicalities. Send 50 to 100 g of the exact finished formula in its intended packaging, because the pack is part of what is being assessed. Expect four to six weeks. As a 2025 planning figure in the UK and EU, a single ISO 11930 challenge test costs in the low hundreds of pounds or euros per formula, and any change of oil, fragrance or supplier means testing again. Enumeration to ISO 17516, which counts what is in the product rather than what survives a challenge, is a cheaper routine test and it still applies to a product that needs no challenge test. Where all this sits in a UK or EU dossier is covered in safety assessment and the CPSR, and the commercial steps around it in selling balms in the UK and EU.

ISO 29621 is the companion standard that excuses a genuinely low risk product from the challenge test, on a documented justification rather than a category name. A plain balm qualifies on water activity. A cleansing balm sold in a jar does not.

Where the money is wasted, and what to do instead

A preservative in a lip balm, a tin of salve, a lotion bar or a body butter is wasted, and the waste has three parts: the cost of the material, the loss of a clean ingredient list, and the false confidence that stops someone fixing the real problem. In every one of those products the actual failure mode is oxidation, not microbiology.

The rest of the alternative to preserving is unglamorous and effective. Fill hot, above 65 C where the formula allows it, which ISO 29621 recognises as a supporting factor rather than a standalone one. Dry botanicals to brittleness before infusing, since residual plant moisture is the classic route to mould in a balm. Keep the workshop dry and the equipment dry, which is the whole point of workshop hygiene and of moisture control. Sell smaller pack sizes so the product is finished before ambient humidity matters. Give a real storage instruction that names humidity as well as heat. And run shelf life testing against rancidity, which is what will actually end the product.

Where a preservative is not wasted, use it properly rather than tokenistically: a blend at the top of its dose band, an appropriate pack, and a challenge test on the finished article. A cleansing balm at 0.3 percent phenoxyethanol in an open jar has the worst of every option, a compromised ingredient list and no demonstrated protection. The formula that most often needs this decision made honestly is in cleansing balm, and the alternative of a stick format is in cleansing balm stick.

The honest limits of this reasoning

The partition calculation on this page is a bulk two phase model at equilibrium, and a contaminated balm is neither bulk nor at equilibrium. Water arrives as droplets with an enormous surface to volume ratio, several of these molecules are surface active and concentrate at that interface, and a cleansing balm contains emulsifiers that change the picture again the moment it is used. That is the honest reason well-formulated cleansing balms pass challenge tests despite the arithmetic looking hopeless: the preservative is doing its work at an interface the model does not represent.

So the conclusion is not that preservatives never work in oil. It is narrower and more useful. You cannot predict preservation in an anhydrous system from a log P value, in either direction, so the log P table is a tool for excluding obviously futile choices, not for approving a system. Three rules survive that caveat. If there is no water and none will arrive, do not preserve. If water arrives on every use, preserve at the top of the band, choose the pH independent materials, and prove it with ISO 11930 on the finished pack. And if an ingredient brought water in with it, no preservative available to a small maker will rescue the formula: take the water out or build an emulsion and accept everything that follows.

Reviewed and updated 6 September 2026. Spotted an error? Tell us and we will fix and log it.