Case study 3 · Location
A carrier that holds the preservative: cyclodextrin in a water-based serum
A water-based serum needed a replacement for sodium benzoate. The real constraints were where each preservative would sit at the finished pH, and whether a cyclodextrin carrier in another raw material would hold it out of action.
The problem
The serum was water-based, with potassium azeloyl diglycinate and an encapsulated salicylic acid raw material whose INCI listing includes a methylated cyclodextrin. It was preserved with phenoxyethanol and ethylhexylglycerin plus sodium benzoate, and the formulator wanted to replace the sodium benzoate.
The pH could not simply be lowered to help the preservative: at pH 4.8, free azelaic acid came out of solution. The serum therefore had to sit at about pH 5 or above.
What an HLB-based reading checks
There is no oil phase and no emulsifier, so an HLB-based reading has nothing to calculate. This is not a failing of HLB; the question simply lies outside what a selection framework was built to answer. It does show that a formulation review organised around emulsifier choice can pass over a problem that has nothing to do with emulsification.
What a placement (TDA Logic) reading asks
TDA Logic asks one question of every ingredient, oils or not: at the finished pH, where does this ingredient sit, and in what form [1]? For an ionisable ingredient that means its logD at the formula pH rather than its logP [2]. For a preservative the answer decides whether it works, because microbes are exposed to the free, available fraction in the water phase rather than to the percentage on the formula sheet.
Is any preservative held inside a carrier?
Cyclodextrins are cyclic oligosaccharides with a hydrophobic cavity, and they form inclusion complexes with many preservatives, including phenoxyethanol, benzyl alcohol, parabens and, to a lesser degree, sorbic acid. A preservative inside a cavity is not available to microbes. If the cyclodextrin in the salicylic acid raw material is not fully occupied, its free cavities can take up preservative from the rest of the serum.
What fraction of each weak-acid preservative is active at the finished pH?
Weak-acid preservatives act mainly in their undissociated form. By Henderson–Hasselbalch, the undissociated fraction is 1/(1 + 10pH − pKa).
| Preservative | pKa | pH 5.2 | pH 5.5 | Cyclodextrin note |
|---|---|---|---|---|
| Benzoic acid (from sodium benzoate) | ≈ 4.2 | ≈ 9% | ≈ 5% | Can be complexed |
| Sorbic acid (from potassium sorbate) | ≈ 4.76 | ≈ 27% | ≈ 15% | Complexed to a lesser degree |
| Benzyl alcohol | Not ionisable at these pH values | pH-independent | pH-independent | Can be complexed |
At pH 5.2–5.5, benzoic acid is just 5–9% undissociated, which is why sodium benzoate is weak here. Sorbic acid keeps about 22% undissociated at pH 5.3, a workable fraction. Benzyl alcohol does not depend on pH at all, though it can be complexed.
Does the pH hold?
pH is a cross-cutting constraint. Here it sits between a lower limit set by the azelaic acid and an upper drift that would erode the weak-acid preservatives. A small buffer capacity is part of the design, and its salts count towards the ionic load.
Practical fix
- Set the pH at 5.2–5.5 with a little buffer capacity, for example citrate (its second and third pKa values, about 4.8 and 6.4, bracket this range), so the azelaic acid component stays in solution.
- Replace sodium benzoate with potassium sorbate at 0.15–0.2%, benzyl alcohol, or both. Sorbate can oxidise and discolour over time, so protect from light and check colour on stability.
- Keep disodium EDTA at 0.1%.
- Ask the supplier how much cyclodextrin the salicylic acid raw material contains and whether it is in excess.
- Regard preservative-efficacy testing as decisive. Because free cyclodextrin can bind preservative, no calculation of the free fraction substitutes for the test.
Takeaway
Location decides activity, not percentage. Asking where every ingredient sits at the finished pH, inside a carrier, in the water phase, ionised or not, is the question that catches a preservative quietly held out of action.
Discussed publicly
The formula problem behind this analysis was discussed in a public forum thread. See In the field for the link and context. Other cases: case-study index.
References
- Pensive Beauty® Nanoscience Labs (2026). TDA Logic: The Tri-Domain Placement Logic Succeeding 1940s HLB Logic in Nanoscale Cosmetic Formulation (Version 1.0) [Preprint]. Zenodo. doi:10.5281/zenodo.23384911
- Leo A, Hansch C, Elkins D (1971). Partition coefficients and their uses. Chem Rev 71(6):525–616. doi:10.1021/cr60274a001