Why some metals have a fatigue limit and others do not
Why steel can last indefinitely and aluminium cannot.
A steel part loaded gently enough, below a specific stress that can be measured for that particular steel, can in principle survive an unlimited number of load cycles without ever developing a fatigue crack, while an aluminium part loaded just as gently keeps accumulating fatigue damage no matter how light the load is, meaning it will eventually crack given enough cycles, only the number of cycles keeps changing rather than ever reaching a true safe limit.
The mechanism behind a fatigue limit
Every metal accumulates some fatigue damage under repeated loading, the earlier articles in this set have already covered why, but not every metal accumulates that damage at every stress level. In many steels and a handful of other ferrous alloys, the microscopic process responsible for fatigue damage effectively stalls out once the applied stress drops below a certain threshold specific to that alloy, the individual dislocations and slip bands that would otherwise nucleate a crack simply failing to move far enough, cycle after cycle, to add up to anything. Below that threshold, the number of cycles a part can survive stops shrinking as the number of cycles applied to it rises, which is a genuinely unusual property for a fatigue process to have, since almost everything else about accumulated damage in this set has behaved as though more repetitions always cost something. Aluminium and most other non-ferrous structural metals do not show this stalling behaviour at any stress level tested, however gentle, so their fatigue strength keeps declining as the number of cycles considered keeps rising, meaning a fatigue-strength figure quoted for aluminium is only meaningful once it states the specific number of cycles it was measured against, since a different number of cycles produces a genuinely different answer rather than converging on one fixed floor the way steel's does.
The bicycle-spoke-and-cable-tie comparison
An old steel bicycle spoke, flexed by every pothole and every ordinary bump in the road for years on end, can go on doing exactly that indefinitely provided the flex stays gentle enough, the spoke showing no accumulating sign of fatigue no matter how many more years of ordinary riding are added on top of the years already behind it. A plastic cable tie loops around a bundle of wires and gets nudged, brushed or lightly flexed by everyday handling in a way that, taken cycle by cycle, looks every bit as gentle as the spoke's flexing, and yet the tie eventually turns brittle and snaps regardless of how carefully it was ever handled, the difference being that nothing about the cable tie's material ever stops accumulating damage from repeated flexing the way the spoke's steel does below its own threshold. Both objects are being loaded repeatedly and gently, and both are made from a material chosen because it seemed durable enough for the job, but only one of the two materials genuinely stops accumulating harm once the load drops low enough, and the other one simply keeps a slower clock running that eventually still reaches zero.
Why the fatigue curve flattens for one material and not the other
Plotting how many cycles a material survives against how hard each cycle loads it produces a curve that slopes steadily downward for every material at first, higher stress buying fewer survivable cycles the way the earlier articles in this set would predict. For steel, that downward curve genuinely flattens out into something close to a horizontal line once the stress drops low enough, and once a part is being loaded below the stress where that flat section begins, the curve is effectively saying the part will survive any number of cycles a person could realistically apply to it. Aluminium's equivalent curve keeps sloping gently downward for as far as it has ever been tested, never levelling off into a genuine floor, which is why an aluminium part's fatigue strength is conventionally quoted against a specific, stated number of cycles rather than presented as a single number the way a steel part's flat-section fatigue limit can be. This is not a minor labelling difference, it reflects a real difference in what is physically happening inside the two metals under repeated gentle loading, one of them reaching a genuine mechanical stalemate and the other simply taking longer to lose.
The number that matters here
An aluminium alloy's fatigue strength, unlike a comparable steel's, is conventionally reported against a stated cycle count, commonly somewhere in the range of several hundred million cycles, because there is no lower stress at which testing could simply be stopped early on the confident assumption that the material would go on surviving forever, the way a steel specimen tested below its own flat-line threshold reasonably can be.
What this changes in practice
A part expected to see an enormous number of load cycles over its working life, a rotating shaft, a vibrating bracket, a spring flexed on every use, is a very different design problem depending on which of these two behaviours its material shows. A steel version of that part can genuinely be designed for an unlimited service life by keeping its working stress safely below the measured fatigue limit, and once that margin is established the part's fatigue risk effectively stops growing no matter how long it stays in service. An aluminium version of the identical part can never make that same promise, since its fatigue strength keeps falling as the expected number of cycles keeps rising, so a designer working in aluminium has to commit to an actual expected service life in cycles up front and design against the fatigue strength that specific number implies, accepting that the part is being sized for a long life rather than an unlimited one. Choosing aluminium for its lighter weight without making this substitution consciously is one of the more common ways a part that performed well in steel quietly becomes a fatigue liability once rebuilt in a lighter metal that never offered the same flat, forgiving floor to design against.