Since we are taught by experience and traditional wisdom that steel contracts when cooled we tend to get stuck in that rut. Steel does indeed contract while cooled and expands when heated, but...
Much more major expansion and contractions will occur when there are changes in the atomic arrangements. Steel expands when heated until it reaches the transformation temperature, then it contracts since austenite has a denser atomic stacking (face centered cubic) than normal body centered iron (alpha iron). But more importantly for this discussion, steel contracts when cooled until it makes a transformation from austenite to something else, where it will go to a lesser density atomic stacking and expand. Normally this will want to be around 1000F when pearlite forms, but if we cool fast enough to avoid this it will occur at around 450F when martensite will form. Martensite is a distorted body centered tetragonal, so it is the most expansive of all. So too the contrary of what we often hear steel actually expands heavily when it is hardened. The reason for preheats by industry and my fiddling with marquenching is to compensate for all this expanding and contracting in differing cross sections.
Inducing a curvature in a blade quenching can be real tricky thing due to the critical and complicated timing that must take place. For sori to be induced water works best because of its timing with simple steels. The spine must cool slow enough through the pearlite range in order to form that product, but not so slow that it does not form it before the edge can reach 400F. or below.
If the spine cools properly to form pearlite, there will be an expansion that will momentarily cause the tip to go down before rebounding and coming back up, when the edge begins to harden. The pearlite formed, it will act as an anchor when the martensite begins its massive expansion, pulling the blade into a sori curve.
The reverse curve in oil has been a mystery for some time and I have worked for a while now to get to the bottom of it. I had been sitting on my conclusions for use in an article, but I have noticed at least one smithing savvy metallurgists on another forum (Bertie Laroux) has figured it out, so what the heck. For some time smiths assumed it was just the timing of oil was off and that the pearlite was too complete and pushed the edge down for the less traumatic oil-formed martensite to recurve, but I noticed that it would also happen with O1 blades quenched point down in oil or salts- no pearlite going on there, and I could overcome the problem on bare blades by quenching spine down, so I doubted traditional wisdom (not like me at all
), and began studying the issue.
This is what I came up with; Oil it seems cannot cool the spine quick enough to lose the more ductile austenite in the spine before martensitic expansion begins. Under these conditions the spine doesn't anchor anything but instead gets pulled along with the expansion, causing no significant curvature. Then when the elongated spine does complete its expansive transformation the rigid edge is instead the anchor and the blade curves in reverse. *
Phillip, #50 is not fast enough to induce the curvature that water will, so you may want to do some preforming and hope that it will remain about the same. Many who do the Japanese style quenching also use much lower austenitizing temperatures to facilitate quicker cooling curves.
*This is based upon the best data I have gathered so far, as always things could change.