An ice lolly freezes from the outside in and the flavour follows the water

Ice rejects almost everything that was dissolved in the water

When water crystallises, the growing lattice accepts water molecules and excludes nearly everything else. Sugar, acid, colour and flavour compounds are all pushed ahead of the advancing crystal front rather than being incorporated into it. This is a general property of freezing rather than a quirk of desserts, and the same principle has been used industrially to concentrate juices by freezing off part of their water and discarding the ice.

In a lolly mould the consequence is a gradient. Freezing starts at the walls, where the cold arrives, and works inward. The outer shell that forms first is depleted of sugar and flavour, while everything rejected accumulates in the shrinking liquid core. That core, now heavily concentrated, has a much lower freezing point and is the last thing to solidify — which is why the centre of a home-made lolly is intensely flavoured and the outside can taste of very little.

You can see the shape of the freezing in the finished lolly

A lolly frozen slowly often shows visible structure: a cloudy, coarse-crystalled outer region and a denser, darker core. Sometimes the concentrated syrup never freezes hard at all and leaks out when the lolly is bitten. Commercial products avoid this partly through much faster freezing and partly by formulating with stabilisers and enough dissolved solids that the gradient is compressed rather than eliminated.

Speed is the variable most available to a domestic kitchen. The faster the whole volume passes through the freezing range, the less time the rejected solutes have to migrate anywhere, and the more evenly distributed they remain. Small moulds freeze more evenly than large ones for exactly this reason, and a mould placed against the coldest surface of the freezer rather than in the middle of a crowded shelf will freeze more uniformly.

Crystal shape follows the same rejection process

As the front advances it becomes unstable, because any small protrusion into the liquid finds itself surrounded by less-concentrated water and grows faster still. The result is branching, needle-like ice rather than a smooth advancing face. Those needles are what makes a slow-frozen lolly crunchy and coarse, and they are the same structures that ruin an unchurned sorbet. The chemistry that concentrates the flavour and the physics that coarsens the texture are one process.

Slushy, granular textures can of course be wanted. Granita is made by deliberately allowing large crystals to form and then breaking them up with a fork at intervals, and its appeal is precisely the coarse, glittering crystal structure that a sorbet works to avoid. The difference between the two is not the ingredients, which are often nearly identical, but the decision about what to do while it freezes.

Formulating around the problem rather than fighting it

A lolly with more dissolved solids freezes more evenly, because the difference in concentration between the outer shell and the core is proportionally smaller to begin with. Adding a little glucose syrup, a fruit purée rather than a clear juice, or a small amount of dairy all move in that direction. Fruit purée is particularly effective, since the suspended pulp physically obstructs crystal growth in a way that dissolved sugar alone does not.

There is a ceiling. Too much sugar and the lolly will not set hard enough to come out of the mould, or will melt off the stick before it can be eaten. The workable range is narrower than for a sorbet because a lolly is expected to be handled at room temperature rather than served from a cold cabinet, so the formulation has to survive several minutes in a warm hand.

Layers, sticks and other practical consequences

The outside-in progression is what makes layered lollies possible. Pouring a first layer, letting it set against the mould walls, and then adding a second gives clean bands, because each addition freezes against an already solid surface rather than mixing. Rushing the layers produces a blur. The same mechanism explains why a stick inserted too early drifts and too late will not go in — there is a window when a soft core still surrounds a firm shell.

Unmoulding is the last piece. A brief dip in warm water melts the thin outer film against the mould, which is the layer that was closest to plain ice anyway, so very little flavour is lost. Doing it slowly under running water melts further in and takes some of the flavoured region with it. A few seconds is enough, and pulling steadily rather than twisting keeps the stick where it belongs.