Heat treatment isn't one process with one dial. Annealing, hardening and tempering do three different — sometimes opposite — things to the same material, and specifying a hardness number without saying which one, and how deep, is how a part ends up harder than intended and more brittle than needed.
Steel's internal crystal structure changes depending on how hot it gets and how fast it cools. Heat it above its critical temperature and cool it fast enough (quench), and the atoms get frozen into a strained, needle-like structure called martensite — hard, but brittle. That brittleness is exactly why hardened steel is almost never left as-quenched: it's reheated to a lower temperature afterward (tempering) specifically to trade away some of that hardness for the toughness the part actually needs to survive an impact instead of shattering under one.
The quenching medium sets how fast the part cools, and that speed has to outrun the material's own transformation curve to reach full hardness — too slow, and softer, weaker structures form before the quench gets there; too fast, and the thermal shock can crack or distort the part.
| Process | What it does | When you'd call for it |
|---|---|---|
| Annealing | Softens, relieves internal stress, restores ductility | After cold forming, before machining hardened stock, after welding |
| Hardening (quench) | Converts to martensite — high hardness, low toughness | Before tempering — almost never specified alone |
| Tempering | Trades some hardness for toughness by reheating below critical temperature | Always follows hardening; the tempering temperature sets where on the hardness/toughness trade-off the part lands |
Some parts need both a wear-resistant surface and an impact-tolerant core — a gear tooth is the classic example. Case hardening (carburising, nitriding, induction hardening) hardens only a shell at the surface while leaving the interior in a tougher, more ductile state. A case-hardened gear tooth commonly runs 58–62 HRC at the surface with a much softer, ductile core — a through-hardened alternative sits at a uniform 45–55 HRC, trading some surface wear resistance for simplicity and cost. Case depth has to be specified alongside hardness, and the two common definitions — effective case depth (to a 50 HRC cutoff) and total case depth — aren't interchangeable; a drawing that gives hardness without case depth is missing the number that most often causes premature failure in service.
“Harden it” isn't a spec — it's an invitation for the shop to guess which process, how hard, and how deep. I specify the process by name, a hardness range (never a single number — testing and process control both have natural scatter), and a case depth with the standard it's measured to, because that's the difference between a part that does what I designed it to do and one that's technically “hardened” and wrong for the job.