An egg cooks by heat moving in from wherever it touches something hot, and the shape of that problem changes with the method. Boiled and poached eggs are heated all around, a textbook heat-conduction problem for a sphere that physicist Charles D. H. Williams (University of Exeter) solved in closed form — poaching reuses that same solution without the shell, at a gentler simmer rather than a full boil. Fried eggs are heated from one side only, closer to heat moving into a thin slab than a sphere, so this tool switches to the matching one-sided diffusion solution. Scrambled eggs are different again — constant stirring keeps the mixture at one temperature throughout, so the right model is the same exponential approach to a target temperature used for the site's tea and drink-cooling calculators, not a diffusion equation at all.
A fridge-cold (4°C) large chicken egg (57g) dropped into sea-level boiling water reaches a classic soft-boiled yolk (63°C at the yolk-white boundary) in 4 minutes 32 seconds — matching Williams' own published example almost exactly. The same egg at 1500m altitude, where water boils at only 94.9°C, needs closer to 5 minutes 10 seconds.
Heat has to diffuse all the way to the centre, and that distance scales with the cube root of volume while the formula's mass term scales as mass^⅔ — so a jumbo egg takes meaningfully longer than a small one, but nowhere near "jumbo mass ÷ small mass" longer.
Boiling water at altitude is genuinely cooler — the boiling point drops as atmospheric pressure drops, exactly like the pressure-cooking relationship in reverse. Cooler water means a smaller temperature gap driving heat into the egg, so it takes longer to reach the same target, even though the water is "boiling" just as hard.
A poached egg isn't a compact sphere any more — cracked into water, it spreads out flat, which gives it far more surface area relative to its volume than a shelled egg has. That extra surface lets heat in faster, so this tool applies a calibrated adjustment to the same formula rather than pretending the shape hasn't changed.
A fried egg is heated from underneath only, by a pan, not surrounded by hot liquid — that's a one-sided diffusion problem into a thin layer, not heat converging toward the centre of a sphere. Both are real, standard heat-transfer solutions; they just apply to different shapes of the same underlying physics.
Diffusion physics assumes heat is spreading through a still object. Stirring breaks that assumption on purpose — it keeps the whole mixture at close to one temperature, which is exactly the condition an exponential-approach (lumped-capacitance) model describes well, and diffusion describes badly.
That's Williams' original derivation — it's the point that determines whether the yolk has started to set, which is what "soft" vs "hard" actually describes. A separate, more complex model (Barham's) targets the yolk centre instead; the two agree reasonably well in practice.
Above roughly 77°C, hydrogen sulphide from the white reacts with iron in the yolk to form a harmless but unappetising grey-green ring. This tool won't offer a doneness target above that, on purpose.
Read that as a removal rate, not a survival chance. 99.9998% removed means the bacteria that were present are divided by about 500,000. It does not mean a 99.9998% chance the egg is fine — those are different quantities, and only the first one is what this calculation produces. It is a multiplier on an unknown starting number, and it never reaches certainty at any cook time.
Most eggs carry none at all. The FDA’s risk assessment estimated roughly 1 in 20,000 eggs is contaminated with Salmonella Enteritidis — context for why raw-egg dishes are not constantly causing illness, though outbreaks traced to shell eggs still occur.
The chance of actually falling ill is a different question, and not one this tool answers. It depends on how many bacteria were there to begin with, which turns far more on how the egg was stored than on how it was cooked, and on who is eating it. The joint FAO/WHO risk assessment put it at roughly 0.00002% to 0.00045% per serving — between about 1 in 5 million and 1 in 220,000 — across its modelled scenarios. That is a published figure for ordinary eggs and ordinary handling, not a result of this calculator.
Where the numbers come from. Heat kills bacteria at a rate that rises logarithmically with temperature, described by a decimal reduction time D (the seconds to cut the population tenfold) and a z-value (the temperature rise that cuts D tenfold). Schuman and Sheldon measured these for Salmonella in liquid egg yolk: D = 0.28 min at 60.0 °C and 0.087 min at 62.2 °C, which together give z = 4.33 °C. Those two figures are from the same paper and are consistent with each other, rather than mixed from different studies.
And the part that is usually left out. That paper measured liquid yolk held at one uniform temperature in sealed capillary tubes. It makes no claim about whole shell eggs, about the temperature gradient inside one, or about any cooking time being safe. Carrying its D and z onto a modelled gradient is an extrapolation this page is making — it is not a finding the authors published, and it should not be attributed to them. Kitchen writing quotes pasteurisation tables as if they certified a recipe; they certify a process, at a validated temperature, measured at the coldest point, which a pan of boiling water is not.
Two things deliberately left out of the calculation. Carryover cooking after the egg leaves the water is ignored, and so is the white — which is hotter and a far less protective medium than yolk. Both would only ever add lethality, so leaving them out keeps this an upper bound on the cold spot in the one direction that matters.
Releases in which this page changed, newest last. Derived from the archived copy of every release, not from notes written afterwards — so it reflects what actually shipped. Site-wide passes are left out; they are in the full changelog.