Creep: Slow Failure at High Temperature

A failure that only needs patience

Guide 4 showed that fatigue can break a part far below its static strength if you cycle the load enough times. This last failure is stranger still: it needs no cycling at all. Hang a steady weight on a metal bar that is nowhere near its yield strength, keep it hot, and come back weeks later to find the bar visibly longer — and if you wait long enough, snapped. This slow, time-dependent stretch under a constant load is creep, and it is the reason a jet engine, a power-station boiler, or a furnace bolt has a service life measured not in a single overload but in thousands of hours.

The one number that decides whether creep matters is not the temperature in degrees but the temperature relative to melting — the homologous temperature, written T/T_m with both in kelvin. Below about 0.4 of the melting point a metal barely creeps; above it, creep grows fast and cannot be ignored. That fraction is why the same rule bites wildly different materials at wildly different temperatures.

The creep curve: three acts of a slow death

The standard experiment is brutally simple: hang a fixed load on a specimen inside a furnace held at a fixed temperature, and record its length against time. Plot strain versus time and you get the creep curve — the single most important picture in this subject. After the instant elastic (and maybe a little plastic) jump when you first apply the load, the curve unfolds in three distinct stages, like three acts of a play whose ending you already know.

strain
  |                                            X  <- RUPTURE
  |                              tertiary     /
  |                              (necking,  _/
  |                              voids)   _/
  |   secondary = STEADY STATE        __--
  |   (minimum, constant slope)   ___--
  |                          ___---   slope = minimum creep rate
  |                    __----
  |   primary      _--
  |   (slows    _--
  |    down)  _/
  | e0 ___--/   <- instant strain when the load is first applied
  |__________________________________________________ time
     I           II  (most of the life)          III

In primary creep the strain rate starts high and steadily falls: the metal is work-hardening, its dislocation tangles jamming further flow, much as kinking a paperclip stiffens it against the next bend. In secondary or steady-state creep the curve becomes a straight line — hardening from ongoing deformation is now exactly balanced by softening from the heat (recovery), so the strain rate holds at a constant minimum. This flat stretch is the longest act and the one you design around. In tertiary creep the rate climbs again as real damage accumulates — the section necks, internal voids open along boundaries and link up — and the curve sweeps upward to rupture.

One honest warning: the shape of this curve is not fixed. Raise the temperature or raise the stress and the whole curve steepens — the steady-state rate climbs and rupture arrives far sooner. A creep curve is therefore always labelled with its stress and its temperature; a single unlabelled creep number means nothing, in the same spirit that a single ceramic 'strength' means nothing without knowing its worst flaw.

What is actually moving down there

At room temperature a dislocation that runs into an obstacle just stops, and the metal holds. Heat changes the rules, because heat switches on diffusion — atoms hopping one lattice site at a time. That single ingredient is what makes creep possible, and it acts through a few competing mechanisms. The dominant one at high stress is dislocation climb: a blocked dislocation no longer has to wait, because atoms diffuse to or away from its core and let it step up and over the obstacle onto a new plane, then glide again. Deformation that was permanently jammed at room temperature quietly resumes.

At low stress and very high temperature a second family takes over — diffusional creep — where the grains simply change shape as atoms migrate from faces under compression to faces under tension, either straight through the crystal or, faster, along the grain boundaries themselves. Alongside it, whole grains slide past one another at their boundaries. Notice the theme: every creep mechanism is powered by atoms moving, so every one of them speeds up exponentially as temperature rises. That is why the melting point sets the clock.

All of this is captured, roughly, in one workhorse relation for the steady-state rate: creep-rate = A times sigma^n times exp(-Q/RT). Two numbers give the game away. The stress exponent n is small for diffusional creep (about 1) but large for dislocation creep (typically 3 to 8), so raising the stress a little can multiply the creep rate enormously. And Q is an activation energy that, for the dominant mechanism, comes out remarkably close to the energy for self-diffusion — a direct fingerprint that diffusion is the engine. The exp(-Q/RT) term is the temperature time-bomb: a swing of 50 degrees C can shorten a component's creep life by a large factor.

Stress rupture, and trading time for temperature

Sometimes you do not care about the slow creep strain at all — you only need to know when the part will actually break. Run the same test but push it to failure and record the time to rupture at a given stress and temperature, and you have a stress-rupture test. Plot the applied stress against the time-to-rupture and you get the design chart an engineer really wants: at 600 degrees C this alloy survives 100,000 hours at 150 MPa but only 1,000 hours at 250 MPa. That is how a boiler-tube designer picks a safe working stress for a demanded lifetime.

Designing against creep: the great grain-boundary reversal

Here is the twist that ties this whole ladder together, and it is worth pausing on. Back on the strengthening rung you learned the Hall-Petch relationship: at room temperature, more grain boundaries make a metal stronger, because a boundary is a wall that blocks a gliding dislocation. Creep flips that lesson on its head. At high temperature the grain boundaries become the weakest part of the metal — they are where atoms diffuse fastest, where grains slide, and where creep voids nucleate and link into cracks. The very seams that were your friends in the cold become the enemy in the heat.

So the fix for a jet-engine turbine blade is astonishing once you see the logic: if grain boundaries cause creep, get rid of them. Blade-makers first grew columnar grains all aligned along the blade (directional solidification), eliminating the weak boundaries that run across the load; then went the whole way and cast each blade as a single crystal with no grain boundaries at all. Combine that with a nickel-based superalloy stuffed with tiny, stable precipitates that pin dislocations even when red-hot, add internal air cooling and a ceramic thermal barrier coating to keep the metal below its creep zone, and you have the reason a modern engine can run its gas hotter than the blade metal's own melting point.

Keep three honest distinctions straight as you leave this rung. First, creep is a high-temperature problem for that material, set by T/T_m, not by any absolute thermometer reading — which is why lead creeps in your hand and nickel does not. Second, creep resistance is its own property: an alloy that is beautifully strong and tough at room temperature can be hopeless at 800 degrees C, because raw yield strength and diffusion-controlled creep are governed by different physics. Third, the boundary reversal is real and counter-intuitive — fine grains for room-temperature strength, coarse or no grains for creep. Master those three and you understand why the hottest parts of the machines around us are quietly some of the most sophisticated materials ever made.