How to Thermal Cycle Knife Steel (article and video)

[Larrin]

Experiments and recommendations on how to process steel after forging. Hardness and toughness measurements along with images of the microstructure. There is a YouTube video as well as an article that has more details.

Article: https://knifesteelnerds.com/2021/08/28/how-to-thermal-cycle-knife-steel/

Video:

 

[outlandish710]

good video - thanks

theoretically toughness increase due to grain refinement depends on size&number of grains/total grain surface/total metallic bond strength... in other words tied to basic physical principle, should have completely linear effect unless something else goes wrong like quenchant becoming too slow - smaller grain is more reactive, lower hardenability + faster/lower aust
regarding that also, should ideal GR not include gradually decreasing temp? or mix of normal GR+quench GR? the micrograph at 13:20 shows relatively chaotic/incomplete result at 3x compared to 8x - justifies use of more agressive method (quench GR). something like 2x normal to reduce distortion then 2-4x quench. maybe that would also solve the accidental spheroidization issue in 52100.
forge/oven comparison was depressing, expected it to be bad but not that bad - wonder about details of forge HT.

 

[Larrin]

good video - thanks

theoretically toughness increase due to grain refinement depends on size&number of grains/total grain surface/total metallic bond strength... in other words tied to basic physical principle, should have completely linear effect unless something else goes wrong like quenchant becoming too slow - smaller grain is more reactive, lower hardenability + faster/lower aust
regarding that also, should ideal GR not include gradually decreasing temp? or mix of normal GR+quench GR? the micrograph at 13:20 shows relatively chaotic/incomplete result at 3x compared to 8x - justifies use of more agressive method (quench GR). something like 2x normal to reduce distortion then 2-4x quench. maybe that would also solve the accidental spheroidization issue in 52100.
forge/oven comparison was depressing, expected it to be bad but not that bad - wonder about details of forge HT.

Grain refinement does not affect metallic bond strength. Grain refinement does not necessarily affect toughness in a linear fashion. It depends on the specific test, material, and changes to the microstructure. Generally grain refinement is talked about in terms of strengthening (Hall-Petch) which is inversely proportional to the square root of the grain size. Though this is not usually represented by martensitic steels as martensite has its own effective grain size and is very strong to begin with. Also doing "grain refining" cycles does not necessarily lead to finer grains because the final grain size may be more dictated by the anneal and/or the final austenitize.

I don't see why gradually decreasing temperature would lead to finer grain size. If you want finer grains you want lower temperature where they grow more slowly. You don't have to make the grain size coarser to then make it finer.

As I said in the article the forge heat treated specimens shown on that chart were overheated. The other details aren't particularly relevant. There was a whole range of different processing that he tried with various annealing or normalizing procedures but it was all pointless because it was overheated in austenitizing.

 

[outlandish710]

Grain refinement does not affect metallic bond strength.

Right, of course. I confused grains w molecules/atoms here.

Though this is not usually represented by martensitic steels as martensite has its own effective grain size and is very strong to begin with.

what does that mean?

Also doing "grain refining" cycles does not necessarily lead to finer grains because the final grain size may be more dictated by the anneal and/or the final austenitize.

that sounds hard to believe. maybe I misunderstand you but I doubt 2x normalize/GR + proper austenitize can achieve the same as 8x GR (for example). the respective part in the article does not include micrographs.

I don't see why gradually decreasing temperature would lead to finer grain size.

because afaik reactivity, in regard to phase change, depends on grain size too.
we don't jump from b) to e) in one cycle, partial not instant process.
still, by using the same temp repeatedly on a given grain structure we eventually reach a certain grain size. but now we're still using the temp that was necessary for the coarser grain on the achieved finer grain. obviously a mismatch, the smaller grain transforms faster thus has lower effective aust temp.

As I said in the article the forge heat treated specimens shown on that chart were overheated.

Sorry, only watched the video before and forgot about the article. Fracture grain indeed says it all. Kinda weird, he could have done that test himself. Don't even need a loupe at that scale.

 

[000Robert]

Umm...., what did they say...??

 

[DeadboxHero]

Seems your getting tunnel vision on grains and forgetting about carbides.

Right, of course. I confused grains w molecules/atoms here.

what does that mean?

that sounds hard to believe. maybe I misunderstand you but I doubt 2x normalize/GR + proper austenitize can achieve the same as 8x GR (for example). the respective part in the article does not include micrographs.

because afaik reactivity, in regard to phase change, depends on grain size too.
we don't jump from b) to e) in one cycle, partial not instant process.
still, by using the same temp repeatedly on a given grain structure we eventually reach a certain grain size. but now we're still using the temp that was necessary for the coarser grain on the achieved finer grain. obviously a mismatch, the smaller grain transforms faster thus has lower effective aust temp.

Sorry, only watched the video before and forgot about the article. Fracture grain indeed says it all. Kinda weird, he could have done that test himself. Don't even need a loupe at that scale.

 

[outlandish710]

Seems your getting tunnel vision on grains and forgetting about carbides.

hm maybe. problem with carbides here is, in aust grain refinement is partially (?) driven by forces (higher energy potential -> lower renucleation temp, Verhoeven page 71) that don't apply to carbides - dislocations and martensite. (In this regard I'm focussing on low carbide steels like 52100, not considering more elaborate HT like >Cm solution treatment. [doesn't matter much since I misunderstood the context, but normalizing temps in the article are effectively high enough for CR in 52100] I.e. regarding Ac1 being affected by carbide forming elements, should in this case be only ~+25°C. GR temp not necessarily = final quench temp, more on that later)
I expected, perhaps falsely, the carbides to simply 'go along' with GR - either unaffected if aust temp is too low to dissolve them or dissolving and re-precipitating hopefully smaller/more numerous due to now increased total grain boundary surface area. [carbide size should not be affected but aust temp matters in regard to carbide/cementite volume]
Particular temp+hold time sets represent certain % of carbon in solution in a given steel (at a given grain size...), theoretically a range from 0-100%. In addition to renucleation temp manipulation mentioned earlier, it should be possible to use the lower end of that range for GR, simulating a full aust of higher temp by repetition or longer hold time while getting smaller grain made possible by lower temp, no? [not relevant here since disadvantage from higher carbide volume is higher than possible advantage from further GR. there might be more exotic HT regimens where this isn't the case...]

 

[Larrin]

what does that mean?

Look up martensite blocks and packets.

that sounds hard to believe. maybe I misunderstand you but I doubt 2x normalize/GR + proper austenitize can achieve the same as 8x GR (for example). the respective part in the article does not include micrographs.

Fortunately reality does not have to conform to our beliefs and doubts.

because afaik reactivity, in regard to phase change, depends on grain size too.
we don't jump from b) to e) in one cycle, partial not instant process.
still, by using the same temp repeatedly on a given grain structure we eventually reach a certain grain size. but now we're still using the temp that was necessary for the coarser grain on the achieved finer grain. obviously a mismatch, the smaller grain transforms faster thus has lower effective aust temp.

No the transformation temperature for austenite largely doesn't change with grain size. We don't need a higher temperature before lower temperatures.

 

[Larrin]

hm maybe. problem with carbides here is, in aust grain refinement is partially (?) driven by forces (higher energy potential -> lower renucleation temp, Verhoeven page 71) that don't apply to carbides - dislocations and martensite. (In this regard I'm focussing on low carbide steels like 52100, not considering more elaborate HT like >Cm solution treatment. I.e. regarding Ac1 being affected by carbide forming elements, should in this case be only ~+25°C. GR temp not necessarily = final quench temp, more on that later)
I expected, perhaps falsely, the carbides to simply 'go along' with GR - either unaffected if aust temp is too low to dissolve them or dissolving and re-precipitating hopefully smaller/more numerous due to now increased total grain boundary surface area.
Particular temp+hold time sets represent certain % of carbon in solution in a given steel (at a given grain size...), theoretically a range from 0-100%. In addition to renucleation temp manipulation mentioned earlier, it should be possible to use the lower end of that range for GR, simulating a full aust of higher temp by repetition or longer hold time while getting smaller grain made possible by lower temp, no?

Carbides grow and nucleate just like grains do. The mechanisms are somewhat different but a lot of things are similar.

With cycling steel there is a limit to how many cycles you can do before refinement no longer occurs. Eventually the grains grow faster than you can form new ones.

 

[Nathan the Machinist]

Fantastic article. I always learn something from your articles.

 

[samuraistuart]

Another issue with excessive cycling is that it reduces hardenability, if memory serves. I wonder if you could cycle a steel like W2 (already very low hardenability) so many times that no quench medium would be fast enough to achieve full hardness.

 

[Larrin]

Another issue with excessive cycling is that it reduces hardenability, if memory serves. I wonder if you could cycle a steel like W2 (already very low hardenability) so many times that no quench medium would be fast enough to achieve full hardness.

Finer grain size and finer carbide size both mean reduced hardenability. So for the oil study for example using steel from the as-received condition meant hardenability was the best case scenario.

 

[A.McPherson]

Hey Larrin , very cool article! Thanks for helping us out!
Again.

 

[outlandish710]

Look up martensite blocks and packets.

Looked it up. So the martensite grain is composed of packets, which each have their own orientation (bain variant) and are themselves composed of blocks (same orientation inside a packet) which in turn are composed of laths. Blocks are the effective grain of martensite but their orientation in bordering packets can be similar which goes against the Hall-Petch mechanism of slip interruption, i.e. the effective grain size is a bit higher than block size.
Wikipedia on Hall-Petch: Yield strength increases until 10nm (where grain boundary sliding begins), for subgrain at 100nm (=ASTM grain size ~23-24) [in iron], so I guess either the higher limit applies or there's an equilibrium point in between were losses from the latter offset gains from the former.
Wikipedia again:"[..] by plotting both the volume fraction of grain boundary sliding and volume fraction of intragrain dislocation motion as a function of grain size, the critical grain size could be found where the two curves cross. "


... still don't understand why you said Hall-Petch is "not usually represented by martensitic steels".
"It has been shown that the higher the density of the subgrains, the higher the yield stress of the material due to the increased subgrain boundary. The strength of the metal was found to vary reciprocally with the size of the subgrain, which is analogous to the Hall–Petch equation. " ?!?
So subgrain density depends on GR, nucleation sites/dislocations from cold work, maybe something else?

Another issue with excessive cycling is that it reduces hardenability, if memory serves. I wonder if you could cycle a steel like W2 (already very low hardenability) so many times that no quench medium would be fast enough to achieve full hardness.

Super Quench has (anecdotal) speed of ~250C/s. General quenchant speed (severity, H value) can be increased by agitation/flow, vibration/sonication (mechanical breakup of vapor phase) or combination of both ("wave technology"). Not likely to max that out.
Beyond that... wonder if heat pipe+compressor setups could be used in plate quenching.


Another question; from the original article

Pearlite is a finer structure where transformation can occur more rapidly. Martensite (quenched steel) essentially has perfectly evenly distributed carbon. If you temper high enough and long enough you do get a structure that starts to look more like a fast DET anneal, however.

So for practical purpose, repeated quenching could replace DET anneal?

 

[Larrin]

Looked it up. So the martensite grain is composed of packets, which each have their own orientation (bain variant) and are themselves composed of blocks (same orientation inside a packet) which in turn are composed of laths. Blocks are the effective grain of martensite but their orientation in bordering packets can be similar which goes against the Hall-Petch mechanism of slip interruption, i.e. the effective grain size is a bit higher than block size.
Wikipedia on Hall-Petch: Yield strength increases until 10nm (where grain boundary sliding begins), for subgrain at 100nm (=ASTM grain size ~23-24) [in iron], so I guess either the higher limit applies or there's an equilibrium point in between were losses from the latter offset gains from the former.
Wikipedia again:"[..] by plotting both the volume fraction of grain boundary sliding and volume fraction of intragrain dislocation motion as a function of grain size, the critical grain size could be found where the two curves cross. "

I was pointing out that Hall-Petch isn't linear not specifically point out that it breaks down at extremely small grain sizes.

... still don't understand why you said Hall-Petch is "not usually represented by martensitic steels".
"It has been shown that the higher the density of the subgrains, the higher the yield stress of the material due to the increased subgrain boundary. The strength of the metal was found to vary reciprocally with the size of the subgrain, which is analogous to the Hall–Petch equation. " ?!?
So subgrain density depends on GR, nucleation sites/dislocations from cold work, maybe something else?

Because martensite has its own effective grain size you wouldn't be correlating strength with the prior austenite grain size (the grain size measurement we are talking about). Higher carbon martensite has a finer effective grain size.

So for practical purpose, repeated quenching could replace DET anneal?

Repeated quenching would still result in martensite. The temper anneal needs only one quench plus high temperature tempering.

 

[Storm W]

Experiments and recommendations on how to process steel after forging. Hardness and toughness measurements along with images of the microstructure. There is a YouTube video as well as an article that has more details.

Article: https://knifesteelnerds.com/2021/08/28/how-to-thermal-cycle-knife-steel/

Video:

Loved the audio dude. You are the man. If I ever get back to work I will throw some cash your way. I owe it to you. I can't believe that you aren't putting this behind a pay wall. I hope karma does it's job and keeps things ok for you because this let's noobs like me make some nice stuff too. Thanks again for all your hard work. I'm not that far away from you. If you ever need something welded up feel free to hit me up and I will do what I can. Or I will buy you a beer and lunch if you are ever up here in Syracuse. I don't know if you ever come up to Crucible. 369.305.6225 if you are up.

 

[Tom Lewis]

Larrin, thanks for sharing!

 

[outlandish710]

after re-reading the 52100 article among other things I think I now (finally) understand the issue in #4/8 and by extension #6/7.
Lowest aust temp dictates lowest grain size. One cycle at that temp can, at longer holding time, replace several shorter cycles.
A lower temp than discussed in the article would be possible but was not used since it would increase carbide (cementite) volume, leading to worse toughness.
Most of what I've wrote in #7 is wrong, I'll correct that later.