A carbonara is not cooked at a temperature, it is cooked down a slope

Every carbonara recipe, including the one in this index, tells you to take the pan off the heat before the eggs go in. None of them tells you what temperature you are then working at, because until recently nobody had written it down.

The number turns up in an odd place. In 2020 a pair of food microbiologists in Brazil, Lopes and Tondo, wanted to know whether a carbonara made with raw yolk could carry Salmonella through to the plate. To find out they had to instrument the dish, and in doing so they published the thermal profile of a carbonara.¹

Drained pasta hit 86.0 degrees. Four and a half minutes after the yolk mixture went in, the pan was already below 60. The sauce they described as creamy and silky.¹ Their conclusion about Salmonella was that this is not a reliable kill step, which is the answer they were after. The by-product is the more interesting number for a cook.

You are not holding the sauce at a temperature. You are dropping through a window, fast, and the whole technique is a race between the heat leaving the pan and the proteins deciding to set.

The window has no single edge

I wanted to give you the temperature at which egg yolk sets, and the literature will not supply one. Depending on what is being measured, the figures run from 62 to 65 degrees for the unfolding of the apolipoprotein in yolk's low-density lipoprotein, 65 to 70 for the practical yolk window, around 69 to 70 for the onset of gelation, and 75 for a solid gel, with calorimetry peaks scattered between 74 and 85.

That spread is not sloppiness. Yolk is a suspension of several protein populations with different thermal habits, and diluting it with water, fat and acid moves the whole set upward. The direction is well established. The size of the shift gets quoted with a precision that traces back to blogs rather than measurements, so I will give you the direction and not a number.

What the Brazilian trace tells you is that the window is being crossed downwards in a few minutes, from well above it to well below. Stir a carbonara slowly and the pan passes through the setting range while the sauce is still thin. Stir it briskly and you drag the emulsion together while there is still heat to work with.

What actually holds it together

Ask why egg yolk emulsifies and you will be told lecithin, and this is where the story gets more interesting than the correction usually offered.

Yolk's surface activity is dominated by its low-density lipoprotein particles, which are around 30 nanometres across. What they do at an oil and water boundary is dramatic: the LDL micelles break down when they come into contact with the interface, spilling their phospholipids and apoproteins across it.² The particle is a delivery vehicle that destroys itself on arrival, spreading a mixed film of protein and phospholipid where a simple surfactant would have arranged a single layer.³

The tempting move is to declare lecithin a myth and LDL the whole answer, and I am not going to, because I could not stand a source up for the stronger claim. What the fetched work supports is narrower and still worth having: the surface-active job is done by lipoprotein particles and the phospholipids and apoproteins they unload at the interface, not by free lecithin arriving on its own.²³ How the credit divides between yolk's several lipoprotein fractions is a live question, and lecithin alone is a shorthand rather than an account.

The physicists did not study carbonara

In 2025 a group of physicists published a phase diagram for an Italian pasta sauce and won an Ig Nobel Prize for it that September. It has been reported everywhere as the physics of carbonara. It is not. The dish is cacio e pepe, and there is no egg in it at all.⁴

That is worth getting right, because the paper is genuinely useful and it is useful about a different problem. Its variables are pecorino, water, starch and temperature, and what it explains is why grated hard cheese seizes into strings instead of dissolving into sauce. Below about one per cent starch relative to cheese, the system falls into what the authors name the Mozzarella Phase. Between two and three per cent it behaves. Clumping sets in around 65 degrees, and with enough starch present the sauce stays stable up to 80 or 90.⁴

They measured that phase diagram by photographing cheese lumps against a backlight and using the length of the major axis of the lump as their order parameter.⁴ A phase diagram of a pasta sauce, built from pictures of clots.

The pasta water problem

Which brings us to the most repeated piece of advice in Italian cooking, and the one the same paper quietly demolishes. Save a ladle of the starchy pasta water, everyone says, and it will bring the sauce together.

The mechanism is right. The dose is not. In the authors' words, the pasta water alone does not contain enough starch to stabilize the sauce effectively.⁴ Put numbers on it and the gap is large: total cooking loss from dried pasta is about 0.037 grams per gram, roughly three and a half per cent of the pasta's dry weight, and that is all solids and not only starch.⁵ Ordinary pasta water lands somewhere near half a per cent starch by weight, against a target of two to three per cent of the cheese mass.

This is why the traditional fix is to cook the pasta in far less water than usual, or to finish it in the pan risotto-style, so the starch concentrates instead of dispersing. The physics vindicates the technique and faults only the dosage that most recipes imply.

There is a second finding in the pasta-cooking work that deserves more attention than it gets. Cooking loss stayed essentially constant as the water was reduced from twelve litres per kilo down to two.⁵ The big pot of water, defended for generations, is not protecting the pasta from anything. It is diluting the one thing you wanted to keep.

Why the cheese fights you

Pecorino and parmigiano are hard, aged and low in moisture, and their casein is held in a network cross-linked by calcium. Heat that network in a watery sauce and the proteins pull tighter on each other rather than dispersing, which is the seizing the physicists were photographing.

There are two ways out and they work by different routes. Starch gets in the way sterically, physically obstructing the protein aggregates from finding each other. Citrate goes after the cause, chelating the calcium that is doing the cross-linking.⁴ That second trick is not a novelty: sequestering calcium with citrate and phosphate salts is exactly how processed cheese is made to melt smoothly, and has been industrial practice for a century.⁶

Four ways an emulsion dies, and the one that is innocent

Emulsions are thermodynamically doomed from the start. Oil and water have no interest in each other, and every droplet you make is a temporary arrangement paid for with energy and held by whatever crowds the interface. There are named ways for the arrangement to end.

Creaming is separation by buoyancy, the droplets floating without merging. Flocculation is droplets clustering while keeping their own skins. Coalescence is the fatal one, where the films between droplets fail and they fuse into larger droplets and eventually into a pool of fat. Ostwald ripening is the growth of large droplets at the expense of small ones by diffusion of the oil through the water phase.

In a split carbonara you are watching coalescence, driven and accelerated by protein coagulation robbing the interface of its stabiliser. Ostwald ripening is the one that is not happening: it requires the dispersed phase to be at least slightly soluble in the continuous one, and triglycerides in water are not.

The slope

So the received advice is sound and the reasons attached to it are mostly ornamental. Off the heat, because you want to be falling through the window rather than sitting in it. Work fast, because the Brazilians measured how little time you have. Keep some pasta water, but understand that ordinary pasta water is too dilute to do the job the recipes credit it with, and that cooking the pasta in less of it is the actual fix.

And no cream, which the Roman purists insist on for reasons of tradition. It happens to also be the position the chemistry supports, since cream is a way of making an emulsion that a properly handled yolk was already going to make for you.

I like that the best number in all of this came from two microbiologists who were not trying to help anybody cook. They wanted to know if the dish could make you ill. On the way they wrote down the one figure every recipe leaves out, and then described the result as creamy and silky, which is not a phrase you expect in a paper about Salmonella.