Colour or mineral powder that settles out before the balm has a chance to set

Dense powders settle by Stokes law while the melt is still thin, so colour ends at the base. Yield stress, pour temperature and the changes that hold it up.

Cut a set puck in half and the fault announces itself: a dense band of colour in the bottom two or three millimetres, a pale and nearly clear layer above it, and a product that pays off weakly at first use and then suddenly too strongly. Nothing has gone wrong chemically. The powder simply fell through a thin liquid for several minutes while you were waiting for the balm to set, and this page is about closing that window.

• Mica 2.7 g/cm3
• Titanium dioxide 4.2
• Zinc oxide 5.6
• Rate scales with diameter squared
• Fix is yield stress, not stirring

Confirm it is settling before you change anything

Three different faults look like uneven colour, and they need opposite fixes, so spend two minutes on a diagnosis. Take one finished unit, ideally the worst one, and cut it vertically with a knife warmed under a tap and dried. A clean vertical cut face through a tin or a pushed-up stick shows you the whole depth at once.

Settling gives you a gradient: continuous, oriented, always denser at the bottom, and the boundary is diffuse rather than sharp. Wipe the cut face on white paper from top to bottom and the smear darkens smoothly along its length. If instead you see bright specks in a duller ground, or swirls with no vertical direction to them, the pigment was never properly wetted and you are looking at the wetting fault covered in streaky or mottled colour. If the top few millimetres are a different material rather than a paler version of the same one, that is phase separation of the fats, dealt with under balm separated in layers.

One further check grades the severity. Scrape the top 2 mm into one dish and the bottom 2 mm into another, melt both, and smear each on paper side by side. If the two smears look like different products, the batch settled completely and needs remelting; if the difference is small, a process change alone will fix it.

Stokes law, and why the numbers are so unforgiving

A particle falling through a fluid reaches a terminal velocity where drag balances the weight it has in excess of the fluid it displaces. For a small sphere in the creeping flow regime, Stokes gave that velocity in 1851 as the density difference times gravity times the square of the diameter, divided by eighteen times the viscosity. Two things in that expression decide everything: the density difference, which you choose when you choose a pigment, and the diameter, which enters squared.

The viscosity term is the one people expect to save them, and it does not, because a molten balm above its setting range is thin. Values for a lip balm melt at 70 C sit in the region of 10 to 30 mPa s depending on how much wax and butter the formula carries, far too thin to hold a mineral particle. The calculations below take 20 mPa s as a working middle value.

That table is the whole problem stated numerically. A tin filled to 10 mm and left liquid for two minutes has lost its mica to the base. The same tin carrying zinc oxide loses it in under a minute, which is why a barrier formula such as a nappy rash balm is the hardest case on the site and why magnesium balm, carrying a dense salt at a high loading, behaves the same way. Colourants at the light end of the density range are correspondingly forgiving, which is one reason the organic-substrate and starch-extended tints in colourants and tints give less trouble than the oxides in iron oxides and mineral pigments.

Size is the lever that squares

Because diameter is squared, halving it slows settling roughly fourfold. A 40 micron mica crossing 10 mm in about two minutes becomes a 20 micron mica taking eight, and a 10 micron grade taking half an hour, which is far longer than any balm stays liquid. This is why a satin-grade pearl sits evenly in the same formula that a coarse sparkle abandons, and it is the cheapest fix available: change the grade rather than the formula. The size bands and what each one looks like are set out in mica and pearl pigments.

Two cautions attach to that. First, the number that matters is not the median. A distribution quoted as D50 25 microns can carry a D90 of 60 microns, and it is the top decile that settles first and forms the visible band, so ask for the full laser diffraction result under ISO 13320 rather than a single figure. Second, do not create your own fine grade by grinding. Milling an effect pigment shatters the platelets that produce the pearl, so you trade a settling problem for a dull product. Mineral absorption pigments such as the oxides tolerate milling, and in fact benefit from it, but only in oil, using a glass muller or a three-roll mill, never dry in a coffee grinder.

The third dimension of the same lever is agglomeration. A poorly wetted powder does not settle as individual particles at all. It flocculates into clusters of 100 microns and upward, which by the square law fall twenty-five times faster than the 20 micron primary particles they are made of. A great many "my pigment sank" batches are really dispersion failures wearing a settling costume, and the tell is a gritty rather than smooth sediment. Slurrying the powder into a small quantity of castor oil and working it to a lump-free paste before it goes anywhere near the melt is the single highest-value habit in coloured balm making.

Yield stress is the only thing that holds a particle up

Viscosity slows a particle. It never stops one. A fluid with viscosity alone lets every denser particle reach the bottom eventually, and the only question is whether it arrives before the balm sets. What stops a particle permanently is a yield stress: a threshold shear stress below which the material behaves as a solid and does not flow at all. Once the yield stress of the medium exceeds the stress the particle exerts on it, the particle stays exactly where it is, indefinitely.

The arithmetic is encouraging. The stress a particle exerts scales with the density difference times gravity times the diameter, which for our 40 micron mica is about 0.7 pascal, and the accepted criterion for static suspension in a yield stress fluid puts the requirement at roughly a tenth of that, so under 0.1 Pa. A set balm has a yield stress in the hundreds to thousands of pascals. There is no contest at all. The problem is not that the requirement is hard to meet, it is that a molten balm meets none of it, and the entire fault lives in the minutes between those two states. The transition itself, and the wax network that creates it, are covered in rheology and yield stress.

The liquid window is the number to measure

Everything useful reduces to one measurable quantity: how many seconds the poured balm spends thin enough to let particles fall. Measure it once, in your own room, with your own formula. Pour a tin, and every fifteen seconds touch the surface with a cocktail stick until it leaves a mark that does not flow back. That time, from pour to immobile surface, is your liquid window.

Typical bench values are worse than makers expect. A 12 mm deep tin poured at 72 C in a 22 C room, left standing on a wooden bench, commonly runs four to eight minutes. Against a mica falling 4.8 mm per minute, that is total settling with time to spare. The same tin poured at 3 to 5 C above the set point onto a cool metal tray can close in 45 to 90 seconds, in which the same flake moves three to seven millimetres in a 12 mm depth: visible drift, but not a band. Get the window under about 60 seconds and most colourants at 25 microns or less become a non-issue.

Three things lengthen the window, and all three are common. Pouring hot, because you are afraid of lumps, adds the whole cooling span from the pour temperature down to the set point. Holding a pigmented melt hot on a hotplate while you fill fifty tins means the last twenty are pouring out of a batch whose colour is already at the bottom of the jug. And filling into containers that sit in a warm draught-free corner, or worse on a warm tray fresh from the oven, doubles the cooling time. The workable ranges per container and wax system are in pour temperatures, and the effect of cooling rate on the crystals themselves is in crystallisation and cooling rate.

Rescue the batch in front of you

Remelting is safe here, with two exceptions noted below. Mineral pigments are thermally stable well past any temperature a balm sees, the fault is physical rather than chemical, and a correctly reprocessed batch is not a compromised one. What you must not do is reheat unit by unit. Settled product is not uniform, so twenty tins remelted separately give you twenty different shades.

1. Scrape the whole batch into one vessel. Every tin, including the pale tops. The consolidated layer at the base of a dense-pigment batch may need a spatula and some force, because settled powder compacts under its own weight into a cake rather than a loose sediment.
2. Melt gently and check for a cake that will not redisperse. Melt to about 70 C using a water bath as described in melting methods, then stir hard and look at the vessel base with a torch. If a granular layer persists after two minutes of vigorous stirring, the powder was never wetted and no amount of suspension aid will fix it. Strain the melt through a 200 micron mesh, discard the residue, and treat the batch as a partial loss.
3. Add the suspension aid to the hot melt. Fumed silica at 0.5 to 1 percent of the batch, sifted in slowly with the stirrer running, or 1 to 3 percent hydrogenated castor oil, or an extra 1 to 2 percentage points of your structuring wax taken off the liquid oil. Do not add all three at once.
4. Hold, then recheck the shade. Give a fumed silica addition ten minutes of stirring at temperature to build its network. Smear a sample on white paper and compare it with a retained sample of the original before you commit fifty containers.
5. Cool to 3 to 5 C above the set point before you pour. Stir slowly as it comes down. The melt should look faintly cloudy and drag on the spoon. That is the correct pouring state, and it is thicker than most people find comfortable.
6. Pour fast onto a cool surface and leave it alone. A metal tray at room temperature, or one that has spent ten minutes in the fridge, removes heat far faster than wood. Do not lid until the surface is fully set.
7. Cut one unit the next day and inspect the face. Record the pour temperature and the liquid window, because this fault recurs the moment someone changes the pouring routine.

What to add, and what each addition costs

Fumed silica deserves the top row because it is the only one that structures the melt independently of the wax. Everything else works by making the wax network arrive earlier or the melt thicker, so everything else also changes how the finished balm feels. Silica and the other particulate structurants are covered in starches and silica, and hydrogenated castor oil in castor wax. If you take the wax route, put the change through the wax ratio calculator rather than adding to the top of a formula that already totalled 100.

Prevent it in the next batch

The ordering of levers matters. Choose a finer grade first, because it costs nothing and works by a square law. Disperse properly second, because agglomerates defeat everything else. Shorten the liquid window third, by pouring cooler and cooling faster. Add a structurant only fourth, when the first three have been done and the pigment is still moving. Reaching for fumed silica while still pouring at 75 C onto a warm tray is treating the symptom of a habit.

When the answer is a different pigment or a different container

Some combinations cannot be rescued by process. A 100 micron sparkle in a 20 mm deep jar has to travel a long way through a melt that cannot set in time, and no realistic amount of silica will hold a particle that heavy and that large without turning the balm into a paste. Depth is part of the problem and rarely gets mentioned: the same formula that settles visibly in a deep jar looks fine in a 5 mm lip tin, because there is not enough distance for a gradient to develop. If a shimmer must go in a deep container, split the load between a fine grade carrying the background sheen and a small fraction of the coarse grade for flash, and accept that the flash will be slightly bottom-weighted.

The honest decision rule is this. If the pigment is below about 25 microns and the liquid window is under a minute, settling is solved and any remaining unevenness is a wetting problem. If the pigment is above about 50 microns, or the density is above roughly 4 g/cm3 at a high loading, you need a genuine yield stress in the melt and should expect to change the texture of the product to get it. And if neither is acceptable, change the pigment. A tinted lip balm built on finely milled oxides and a satin mica will behave; the same formula built on coarse sparkle will not, however well you pour it. Hard particles left sitting proud of the surface also cause the abrasive feel described in drag and poor glide, so a settled batch usually feels wrong as well as looking wrong, and both faults close with the same fix.

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