Well work hardening is caused by dislocations forming so that crystals jam up against each other, so in copper at least annealing must be removal of these.
Tempering is not the same as annealing though, here's an extract from the excellent article by Richard Rex in MEW 224 January 2015:
Hardening carbon steels
Roughly speaking, the hardening process today is only a refinement of what I learned back at the forge: heat then quench. The back story is a little more complicated. First, the metal must be in the all-austenite state, meaning above the UCT. Second, you need to know that quenching forms an entirely new phase called martensite. Yes, another ‘ite’ word. There are at least eight in the steel literature; this article talks about only five. Martensite is responsible for tool steel's useful qualities.
Consider the case of an ordinary tool steel like O1 or W1. Looking at the Fe-C phase diagram, which depicts equilibrium conditions, we know that the metal must be entirely austenitic above the UCT, with all of its carbon (say 1%) dissolved into the wide-open spaces of FCC iron. Slow cooling through the UCT and below changes the composition to a mix of austenite and cementite. With further slow cooling below the LCT this is transformed again into cementite and pearlite. These processes take time.
But what if you don't allow the time? Suppose you quench it, cool it very rapidly, such as 2000OF (1100 C) per second. Unsurprisingly, this causes a violent, near-instantaneous transformation from austenite to something else entirely, namely martensite. Theoretically, martensite is all there should be in just-quenched steel at room temperature; in practice the martensite content may be as low as 60% or so, the remainder being austenite that didn't make the cut.
The change to martensite is dramatic indeed, like the shattering of safety glass. In a fraction of a second we have gone from the 14-atom FCC austenite to a stretched version of the 9-atom BCC, yet without losing any carbon, thus preventing the formation of pearlite. The new ‘martensitic structure’ is called a Body Centered Tetragonal, BCT (Fig.6).
Martensite is indeed a phase, but it is highly unstable and therefore doesn't appear on the phase diagram. It is a very hard and brittle form of steel with a hardness close to 65 RC (Rockwell ‘C’ scale), about the same as high speed steel. Martensite comes with a lot of internal stress and strain which, as you've guessed, we can relieve to the desired degree by cooking at a low temperature – in other words, by tempering.
What does tempering actually do?
This is a question that gets evasive answers, if any, mainly because the metallurgy is very complex. At the practical level we know that tempering makes the workpiece a usable tool, trading its hardness/brittleness for toughness. Tempering is a slow process, taking an hour or more in the oven. The short version of the story is as follows. Tempering stabilizes the steel in three stages: 1 By forming in the martensite very small ‘intermediate’ carbides (relatives of cementite, but not quite the same), then; 2 By decomposing retained austenite into ferrite and intermediate carbides, and; 3 Finally, by replacing the intermediate carbides with their more stable counterpart, cementite, Fe3C.
Now, back to the shop
Do you really need to know all that austenite, cementite, martensite stuff for practical heat treatment? No, but it helps to talk the talk when your project calls for additional background.
<oops – I missed out the important bit>
Neil
Edited By Neil Wyatt on 21/09/2015 19:02:46