Warping is the print shrinking as it cools

Thermal contraction, and why large flat parts lift at the corners.

Warping happens because hot plastic occupies slightly more space than the same plastic once it has cooled, an ordinary case of thermal expansion running in reverse, and if a printed part's corners cool and lock into their smaller size before the centre has finished shrinking too, the whole shape pulls itself upward at the corners to relieve the mismatch.

What is actually happening

Plastic extruded from the nozzle emerges close to its melting temperature, often well above the temperature of boiling water, and like almost every material it takes up slightly less volume once it has cooled to room temperature than it did while hot. A single thin line of plastic shrinks by an amount too small to notice, but a wide, continuous printed floor, the first solid layer of a large flat part, is shrinking along its whole length and width at the same time, and if the surface it is printed onto grips the plastic firmly while the plastic wants to pull itself inward as it cools, something has to give. The middle of that floor, held from every side by material still trying to shrink, mostly stays flat, but the corners are anchored on only two sides, and a corner pulled inward from two directions at once has nowhere to go except up and away from the plate. An edge running along one side of the part is only being pulled from one direction and so lifts far less, if at all, which is why warping shows up almost exclusively at corners rather than spread evenly around a part's whole outline.

The bacon comparison

A strip of bacon in a hot pan curls for a related reason. The outer edge of the strip, thinner and more exposed to the pan, cooks and renders its fat first, firming and drawing in as it does, while the thicker middle is still soft and has barely begun to change. The firming edge pulls against the still-soft centre, and because the strip cannot shrink and stay flat at the same time, it lifts and curls at the ends instead. A warped printed part behaves the same way in reverse order, since the material at the edges cools and shrinks first while the plate has already pinned the shape underneath, and the corners lift for exactly the reason the ends of the bacon do, one part of the material contracting faster than the part it is still attached to. Thicker bacon curls less for the same reason a thicker printed base warps less, since more material at the centre takes longer to catch up to the shrinking edge, giving the whole piece more time to even out before the mismatch becomes large enough to lift anything.

One figure worth keeping in mind

Common printing plastics shrink by only a small fraction of a percent of their length as they cool from printing temperature to room temperature, a change far too small to notice in a single line of plastic, but stretched across a printed floor a few hand-spans wide, that same small fraction adds up to a pull measured in whole millimetres, concentrated almost entirely at each corner, which is more than enough to lift a corner clean off the plate.

What this changes in practice

Because the pull comes from cooling rather than from anything wrong with the print settings themselves, the fixes all work by slowing or evening out that cooling rather than fighting it directly: keeping the surrounding room free of draughts, printing an outer wall wide enough to anchor the corners more firmly, or rounding a sharp square corner into a gentler curve so the pull is spread along a longer edge instead of concentrated at a single point. None of these change how much the plastic shrinks, only how evenly it is allowed to do it. A part enclosed on all sides by a warm, still box of air cools at close to the same rate everywhere on its surface, and a part left to cool in a draught from an open window cools fastest wherever that draught happens to land, which is exactly why the same design can print flawlessly on one attempt and warp badly on the next for no reason connected to the print settings at all.