The formula carried 0.8% phenoxyethanol, well under the 1.0% legal ceiling. The challenge test came back borderline at day 14 and then failed on Aspergillus brasiliensis. The first reaction was to add more. There was nowhere to add it to. The second reaction was to blame the raw material. The raw material was fine, and the assay on the COA was 99.2%.
What failed was not the preservative. It was the assumption that the number on the formula sheet is the number the mould sees. Microorganisms live in water. Only preservative dissolved in the water phase, unbound and undissociated, can cross a cell membrane. Everything else is inventory.
Four ways an added dose becomes a smaller active dose
Between the weigh boat and the cell membrane, a preservative passes through four filters. Each one has its own arithmetic, and only one of them is visible on a specification sheet.
| Mechanism | What it removes | Where it shows up | Can it be calculated? |
|---|---|---|---|
| Micelle partitioning | Hydrophobic preservative solubilised inside surfactant micelles above the CMC | Cleansers, shampoos, high-surfactant systems, solubilised fragrances | Partly, if the micelle-water partition coefficient is known |
| Oil-phase partitioning | Preservative dissolved in the oil or lamellar phase instead of water | Emulsions, creams, cleansing oils | Yes, from an oil-water partition coefficient |
| pH dissociation | Organic acid converted to its anion, which does not penetrate | Any acid preservative above its pKa | Yes, exactly: 1/(1+10^(pH-pKa)) |
| Binding and adsorption | Preservative bound to anionic surfactant, protein, polymer or container wall | Cationic systems, protein-rich formulas, HDPE and PET packs | No, only by measurement |
Micelles: the same 0.1% behaves like two different products
Above the critical micelle concentration, surfactant molecules assemble into aggregates with a hydrocarbon core. Any preservative with meaningful lipophilicity partitions into that core, and the amount that does so is described by a micelle-water partition coefficient, usually written K or Km. The classic paraben measurements are the clearest illustration: reported K values exceed 2 for butylparaben and fall well below 1 for methylparaben in the same polysorbate system, and partitioning rises with the surfactant's hydrocarbon chain length while falling as its polyoxyethylene chain lengthens. Same surfactant, same temperature, four-fold difference in how much preservative is effectively withdrawn from the water.
The practical consequence is that "phenoxyethanol is surfactant-compatible" is a half-truth. It is far more compatible than a long-chain paraben, but nonionic surfactant above the CMC still reduces its free aqueous concentration, which is why phenoxyethanol systems in sulfate-free and high-surfactant cleansers routinely need higher use levels than the same system in a lotion. Published work on benzalkonium chloride points the same way with a different mechanism: the free fraction falls as surfactant concentration rises, and the apparent partition coefficient is highest in polysorbate 80 among the surfactants tested.
Cationic preservatives have a separate failure mode. Quaternary ammonium compounds bind anionic surfactants electrostatically. The result can be charge neutralisation, precipitation, or simply a loss of activity that is invisible in the finished product, because a rinse-off gel with slightly less quat still looks like a gel.
The acid arithmetic, done exactly
Organic acid preservatives work only in the undissociated form, because the neutral molecule crosses the membrane and the anion does not. The fraction that remains undissociated is:
undissociated fraction = 1 / (1 + 10(pH - pKa))
That single line explains most "it worked in the lab batch" failures. Take sorbic acid, pKa 4.76, at the EU ceiling of 0.6%:
| Formula pH | Undissociated fraction | Active sorbic acid from 0.6% added | What it feels like in practice |
|---|---|---|---|
| 4.0 | 85.3% | 0.51% | Robust, wide margin |
| 4.8 | 47.7% | 0.29% | Normal working point |
| 5.5 | 15.4% | 0.09% | Marginal, mould will find it |
| 6.0 | 5.4% | 0.03% | Functionally unpreserved |
From pH 4.8 to pH 5.8 the label does not change by a single digit, and the active concentration falls about six-fold. Benzoic acid, pKa 4.20, is steeper still: 33% undissociated at pH 4.5, 4.8% at pH 5.5. Salicylic acid at pKa 2.97 is 97% undissociated at pH 4.5 and 32% at pH 6.0, which is why it is one of the few acids that still works in a mildly neutral formula, and why its 0.5% legal ceiling in the EU is not generous.
This is the reason a pH drift of 0.4 units during a six-month stability study is a preservation event, not a cosmetic one. It is also the reason a thickener that needs neutralisation to pH 6.5 and an acid preservative are, in combination, a design conflict rather than a formulation choice.
The ceiling is legal, not chemical
Once the free fraction is the problem, the instinct is to raise the dose. In Europe that door is closed by Annex V of Regulation 1223/2009, which caps the total, not the free, concentration.
| Preservative | CAS | EU Annex V maximum | Practical note |
|---|---|---|---|
| Phenoxyethanol | 122-99-6 | 1.0% | Broad spectrum but weak on moulds; commonly paired with polyols or acids |
| Benzoic acid and sodium benzoate | 65-85-0 | 2.5% as acid | Only useful below about pH 5; rinse-off biased |
| Sorbic acid and potassium sorbate | 110-44-1 | 0.6% as acid | Mould and yeast oriented, pH sensitive |
| Salicylic acid | 69-72-7 | 0.5% | Works at higher pH; not for children under three except shampoo |
| Dehydroacetic acid | 520-45-6 | 0.6% | Often supplied as a blend with benzoic acid or benzyl alcohol |
| Chlorphenesin | 104-29-0 | 0.3% | Fungal activity, frequently formulated with phenoxyethanol |
| IPBC | 55406-53-6 | 0.02% | Not in oral or lip products; not in products for children under three except bath and shampoo |
| MIT | 2682-20-4 | 0.0015% | Rinse-off only at this level; leave-on use withdrawn |
Read that table as a design constraint, not a shopping list. When 0.6% is the ceiling and only 5% of it is active at your pH, the formula has to be redesigned rather than re-dosed.
Five levers that raise the free fraction without raising the dose
- Buffer the pH into the acid's working range. A citrate buffer at pH 4.5 to 5.0 costs nothing and can multiply the active fraction of an acid three to five times. Citric acid anhydrous (CAS 77-92-9) plus a small amount of sodium citrate is the cheapest version of this lever; the buffer capacity, not the pH number, is what holds it through ageing.
- Choose low-partition preservatives for high-surfactant systems. Polyols sit at the other end of the scale from parabens: 1,2-hexanediol (6920-22-5) has a logP near 0.6 and stays overwhelmingly in water, which is exactly why the polyol-plus-acid systems dominate sulfate-free cleansers. They are weaker per gram and more predictable per gram, and predictable is what a challenge test rewards.
- Reduce micelle density rather than adding preservative. Lower total surfactant, or shift to a mixed surfactant system with a higher CMC, and the same preservative dose gains free fraction. This is a cost engineering decision, not a microbiology one, and it is usually available.
- Add a chelator to lower the required free concentration. EDTA and its alternatives strip the divalent cations that hold the Gram-negative outer membrane together, which potentiate many preservatives at unchanged dose. It is a potentiation effect, not a preservative effect, and it does not appear in the total-preservative number at all.
- Stop pairing cations with anions. A cationic preservative in an anionic surfactant base is a neutralisation reaction running inside your product. If both are required, the surfactant system has to change.
Prove it in the finished formula, not in water
Preservative efficacy is judged by ISO 11930, and the judgement is made on the final formula with its real surfactant load, real pH, real oil phase and real packaging, not on a dilution of the preservative in water. Criterion A is the demanding one: bacteria must fall by at least three logarithms by day 7, yeasts by at least one logarithm by day 7, and moulds by at least one logarithm by day 14, with no regrowth through day 28. Criterion B relaxes the timing to day 14. A system that passes Criterion A in water and fails on mould in the emulsion has not changed its preservative; it has changed its free fraction.
If a failure is contested, the free concentration can be measured rather than argued about. Equilibrium dialysis and ultrafiltration both separate the free fraction from the micelle-bound fraction, and the published o/w emulsion gel work used exactly that method to show that free phenoxyethanol in the water phase, not total phenoxyethanol, correlates with the result against Pseudomonas aeruginosa, Staphylococcus aureus, Candida albicans and Aspergillus brasiliensis.
What to put on the purchase order
| Field | Why it matters | Typical value to demand |
|---|---|---|
| Active assay | Total dose is calculated from assay, not from the drum label | Stated minimum, with the analytical method named |
| Composition of blends | Blends carry a Mixture CAS and hide the active ratio | Percentage of each active, not just INCI order |
| Recommended pH window | An acid blend used at pH 6 has already lost most of its activity | Explicit range, with the pKa logic behind it |
| Recommended use level in high-surfactant systems | The lotion figure does not transfer to a cleanser | A separate figure for surfactant-rich bases, or a statement that none exists |
| Heavy metals and 1,4-dioxane | Ethoxylated actives can carry both | 1,4-dioxane at or below 10 ppm for cosmetic grades |
Supply
We supply both the actives and the systems built around them. Single actives: phenoxyethanol (CAS 122-99-6), 1,2-hexanediol (6920-22-5), 1,3-butylene glycol (107-88-0) and citric acid anhydrous (77-92-9) for the buffer. Ready systems: YS-801, YS-802, YS-803, YS-804 and YS-805 for personal care; YS-806, YS-808 and YS-812 for detergent systems; YS-701 for coatings and construction; YS-MK (2634-33-5) and YS-HEY (3811-73-2) for metalworking fluids. Polysorbate 80 (9005-65-6) is listed here as the surfactant most often responsible for the micelle effect described above, not as a preservative.
For two orthogonal angles on the same problem, read how to choose a preservative system (which one) and water activity and polyol preservation (how much water is available). This article is the third question: of what you added, how much is still free. For blend systems we can supply the composition table, the recommended pH window and a separate use-level recommendation for surfactant-rich bases. Samples and the full technical dossier are available through the enquiry form on any product page.