Dairy and Heat: Why Cream Splits and Butter Burns

Butter is not pure fat

Butter is about 80% milk fat plus 16-18% water and milk solids. That is why it sizzles in a pan: the sound is water boiling off.

The milk solids are the problem. Once the water has gone, they start to burn at around 150C. Clarified butter — ghee — has both removed and takes 250C. Indian cooking uses ghee for a temperature reason, not a traditional one.

• Gentle heat and butter flavour: butter.
• High heat needed: ghee, or oil with a knob of butter.
• Brown butter: you are deliberately browning those solids, and the next stage is burnt.

With cream, the fat percentage decides

Whether cream splits when boiled is largely set by its fat content. 35% and above survives boiling; 30% and below splits easily. That is usually the difference between what is sold as whipping cream and as cooking cream.

Three things split it:

• Boiling. Proteins clump and the fat separates out.
• Acid. Add lemon, tomato or wine and boil, and it goes much faster.
• Salt plus prolonged heat, both of which help proteins coagulate.

So cream pasta never boils after the cream goes in, and in any acidic sauce the cream goes in off the heat, last.

The same number decides whether it will whip

That percentage arrives here as a survival threshold for a pan, and then as the name on the carton — "the difference between what is sold as whipping cream and as cooking cream." What the whipping one is for is never said. Cream much under 30% will not hold air at all, and for the job that number is really sold for, everything above inverts.

• The temperature requirement runs the other way. Everywhere above, the risk is getting cream too hot. Here it has to be cold, because the fat must be solid to build anything. Prinsesstårta cites this guide for it: "Cold cream, cold bowl. The fat has to be cold to whip firm. Lukewarm cream never reaches the stiffness a dome needs, however long you beat it". Eton mess says the same in one clause — "Cold is not optional either: the fat has to be solid to build that network at all."
• And what holds the foam up is the fat, which is what ruins the other foam in this kitchen. Eton mess writes the mechanism down: "Whipped cream is a foam held up by fat globules partly stuck to one another". Cooking eggs has a section headed Keep fat away from whipped whites, and it opens "A trace of yolk in the whites and they will not whip: the fat destabilises the protein film." Two foams on opposite principles, and chocolate mousse builds both of them in one bowl.
• It also splits with no heat at all, which puts a bound on the claim further down this page. Over-whipping is that fat network taken too far: Eton mess — "whipping past soft peaks pushes them further into sticking — which is how cream becomes butter", and acid then finishes it, "Folding acidic strawberry juice into over-whipped cream finishes the job, and it turns grainy." Medovik starts from the acid and cites this guide while doing it: "Stop the soured cream at firm peaks. It is already acidic, so whipping past that splits it". So two of the three causes listed above — acid, and proteins and fat aggregating — run perfectly well in a cold bowl once a whisk supplies the energy that heat was supplying. The heat those three need is the heat that dairy standing still needs, which is the case šaltibarščiai is making.
• Where to stop is set by what the cream has to do next, not by preference. Soft peaks where it is going to be folded into something (chocolate mousse: "to soft peaks — don't overwhip"); firm where it has to hold a shape, which is why prinsesstårta's dome "is not decoration, it is the structure the marzipan will lie on" and goes into the fridge for an hour, and why medovik can stack eight layers on it. Past firm is butter. Soufflé gives the same soft-peak instruction for the opposite reason — over-whipped whites "have already stretched as far as they go, so in the oven they tear instead of expanding".

So the percentage above is read as what cream can survive. The same number in a cold bowl is a capability rather than a tolerance, and at the top of its range the failure is not splitting but butter.

Yoghurt needs all three conditions

Yoghurt is already acidic and high in protein, making it the most fragile dairy of all. Three rules, all of them:

• Room temperature. Cold yoghurt into a hot pan sets on contact.
• Off the heat, with the pan given a moment to cool.
• In stages, stirring. Poured in at once, the local temperature cooks it.

This is why rogan josh nails all three into its steps. Once it floats as curds, no amount of stirring recovers it (emulsions and pan sauces).

Dishes that cannot obey all three handle the problem differently instead. Dovga beats an egg and two spoons of flour into a whole litre of yoghurt first, putting protein and starch between the casein, and then boils it. Šaltibarščiai goes the other way and never applies heat at all, so the conditions never arise — all three causes of splitting need heat, in dairy that is standing still — which is also why its potatoes are served beside the bowl rather than in it. Koldskål puts raw yolk into acidic buttermilk, and since yolk dropped into cold acid seizes into threads, the buttermilk is let down into the yolk rather than the reverse.

And one dish breaks the second rule outright, because the list has an unstated premise: that the hot thing gets stirred in. Çılbır pours butter well past 100°C straight onto room-temperature yoghurt and nothing separates — butter is fat with almost no water, so it cannot carry heat into the protein; it is less dense, so it floats instead of mixing; and the contact lasts seconds. Stir that same butter into the yoghurt and it splits on the spot. So "in stages, stirring" is the safe way to combine something hot with yoghurt, and not stirring at all is safer still. The first rule still binds: the yoghurt has to be at room temperature.

Milk is vulnerable to acid and salt

Add tomato or lemon to a milk-based sauce and it curdles. Reordering usually fixes it: reduce the acid first, add the milk last and gently.

Evaporated milk is the exception — already heat-concentrated, its proteins are stabilised and it takes boiling well. That is part of why poo pad pong curry uses it rather than coconut milk.

Barfi does that concentration at home over two hours — and it is not simply thickening. Lactose and milk protein brown against each other, so what is left smells of caramel and nuts rather than of milk.

Some dishes make the curdling you are avoiding here the objective. Rasgulla breaks milk with acid deliberately and lifts the curd out. Douhua uses a calcium salt on soy milk and stands the whole pot up as a gel with nothing to lift — acid gives lumps, a salt gives a gel.

Cheeses melt in different ways

• Young, high-moisture cheese (mozzarella) stretches, because the casein structure is still loose — and why it is loose comes down to how it was set: rennet cuts one spot on the casein surface and leaves the calcium inside, and that calcium is what later lets the protein slide.
• Long-aged hard cheese (parmigiano) does not stretch; it melts in — but releases its fat readily.
• Mid-aged cheese like cheddar goes oily and clumpy if the heat is too high.
• A cheese that does not melt at all — and there are three ways to get there. The three lines above sort by age and moisture; this one runs on a different axis. The kefalotyri or graviera that Saganaki actually uses is low in moisture with a tightly knit casein network, so the surface browns while the inside only softens. The halloumi that dish names separately was cooked in its own whey near 90°C during manufacture, denaturing the proteins in advance. And paneer is acid-set, so almost no calcium bridging is left — the exact inverse of the mozzarella line above, where the calcium stays inside and lets the protein slide. Which is why Rasgulla lifts that curd out as the point of the dish, and why Jāņu siers has to force it to flow with bicarbonate and egg.

Two applications sit at either end of that list. Aligot makes a stretching cheese stretch inside a starch, so the two networks become one material, and Jāņu siers takes an acid-set curd — which by the rule above can never flow — and makes it flow anyway with bicarbonate and egg. Brunost sets nothing at all: it is boiled down from the whey the other three throw away.

The fix for a clumping cheese sauce is lower heat plus a little starch, which gets between the casein and stops it knitting. That is exactly what the cornstarch and white wine in cheese fondue are doing — starch preventing clumps, the wine's acid keeping the casein loose.

And that same fix can arrive already inside the cheese

Everything above describes cheese as it comes off the ageing shelf, where how it melts was settled by moisture and time. Processed cheese is the melting problem solved during manufacture, which puts it on the list as an entry in its own right rather than as an inferior version of a third.

The fix is the one in the paragraph above, moved. Where a fondue adds starch in the pot to keep the casein from knitting, processed cheese carries emulsifying salts — citrates and phosphates — that do it from the start: they take the calcium off the casein and nudge the pH up, so the protein disperses and stays dispersed. Which is why an American slice melts smooth at a heat that makes the mid-aged cheddar two lines above go oily and split, and why Grilled Cheese can name cheddar and American in one breath while the rest of its method exists to keep the cheddar out of trouble.

Jāņu siers is the same move made at the stove: its page says outright that emulsifying salts "give a smoother result than soda, on the same principle: raise the pH and the casein disperses. That is how sliced cheese is made." So there are three distances at which this is done — in the pot (fondue's starch), at the stove to a curd that has already refused (soda and egg), and in the factory, before you buy it. Same problem, one answer, three places to put it.

Melting is not a switch you can flip back

Everything above describes a cheese on the way up. A dish built on melted cheese is usually decided on the way down, and they do not all arrive at the same place. What decides where it arrives is not a property of the cheese but what you melted it into.

• Melted as itself, it comes back as itself — and that can be the structure. Quesadilla is one disc folded over with nothing else holding the halves together, so the minute before cutting is doing real work: "a minute takes the temperature back below the point where it flows, and the wedge holds its shape. The rest is not cooling it down, it is setting it."
• Softened but never flowed, it knits back tighter than it started. The kefalotyri in Saganaki never ran, so there is nothing to re-form, and its own step text says so — "as it cools the softened casein knits together again and it turns rubbery." Reheating makes it worse rather than better. That is the price of having been heated at all, and it is why the dish is carried straight to the table.
• Knitted into a starch, the state is a passing one. Aligot's ribbon stands up only between 60 and 70°C, and left alone "it sets into a block you have to cut". Reheating does not bring the stretch back: two networks were stirred into one material, and cooling takes the fat out of it first.
• Dispersed into a liquid, there is no cheese to come back to. Cheese fondue is an emulsion held by the starch and the wine's acid exactly as the paragraph above describes, and cooling does not return it to cheese — "the emulsion breaks into strings and puddles". That, and not presentation, is why the burner is part of that recipe.

Serving temperature carries the clock for saganaki, aligot and fondue alike, filing them as dishes with no grace period — and it already asks this same question of two other materials, where a starch set loosens again when it is warmed and one set with egg has no way back. Cheese holds both answers, and which one you get was settled before the pot went on the heat.

In short

Dairy breaks for one of three reasons almost every time: too hot, acid present, or added too early. Moving it later and turning the heat down solves most of it (heat control).