Temper Embrittlement

[Jerrod Miller]

This thread is for a discussion on the causes and effects of temper embrittlement.  The topic comes up every now and then, and really deserves a thread to be able to point people to when needed.  I will do my best to keep things as simple as possible, while at the same time covering the important parts.  Besides a number of years as a practicing metallurgist in steel foundries I am also using my 3 most common reference books for this type of topic that I keep within reach on my desk:  The ASM Heat Treater's Guide (2nd ed, 1995), Structure and Properties of Engineering Alloys (2nd ed, William F. Smith, 1993), and Phase Transformations in Metals and Alloys (2nd ed, 1992).  For my own sanity I will not be trying to be overly diligent to indicate which source I am referencing in the details.  It should be pointed out right off the bat that there is still some things that are not known about the mechanisms behind it, but that it actually exists and has negative effects is not under any debate amongst metallurgists.  

 

All temper embrittlement is based on quenched and tempered steel.  The embrittlement affects martensite in steel, no need to worry about normalized steels.  

 

Blue Brittleness

I will admit that I am quite terrible about continuing to use this term as a catch-all for temper embrittlement.  I will try to do better about that.  Blue brittleness is noted by ASTM to form between about 400-700 F, but generally is not a problem with properly deoxidized steels.  Therefore true "blue brittle" problems are not likely to be encountered with modern steels produced in good mills.  Generally when this is referred to (at least by me, and I am sure many others) what is really meant is a general temper embrittlement that will be discussed below.  The term blue-brittle comes from the fact that steel that has fractured due to this phenomenon tends to have a blue hue to the fracture surface due to oxide growth at that temperature.  Other temper embrittlement mechanisms don't show this.  

 

One Step Temper Embrittlement 

Generally formed between 440 and 660F in short temper times.  The embrittlement has experimentally been determined to be related to impurities.  As summarized by Smith:

1. The occurrence of the anomalous impact-energy trough coincides with the beginning of cementite precipitation.  
2. Since the one-step embrittlement causes an intergranular mode of fracture along prior austenitic grain boundaries, it is believed that the segregation of P, N, and possibly S to the austenitic grain boundaries is essential for this type of embrittlement.  
3. Alloying elements such as manganese may have an indirect effect by promoting the segregation of the embrittling elements to the grain boundaries.  
4. The presence of the undissolved carbides at the prior austenitic grain boundaries is thought to accentuate the impurity-induced intergranular fracture, the carbides acting as slip barriers.  

Now, I know that that is a bit of a tough "summary".  Basically it indicates that one should really avoid the 440-660 F range when one has too much (an actual number is not given) of P, N, and S, and the "too much" threshold is probably lowered as Mn goes up.  Also, if one has carbides that have not been properly dissolved prior to hardening then this one-step mode of embrittlement is more likely to be a problem.  

 

Two Step Temper Embrittlement

Smith indicates that it occurs in about the 700-1000 F range.  Every metallurgist I have spoken to actually puts the upper limit in the 800-900 F range, with 900 F being a common temper range.  It is best to spend as little time in this range as possible when tempering; on heating and cooling (more important than the heating speed), and certainly don't hold in this range.  This, too, is an embrittlement mechanism related to impurities and their segregation to the grain boundaries.  As summarized by Smith:

1. The ductile-brittle transition temperature is directly dependent on the grain boundary concentration of the impurities.  This effect [was tested] a nickel-chromium steel doped with Sb, Sn, and P.  The relative effect of these impurities was found to be Sn > Sb > P.  
2. Alloying elements sometimes cosegregate to the grain boundaries with the impurities.  For example, nickel cosegregates with antimony (Sb).  
3. The segregation of impurities to the grain boundaries appears to be an equilibrium phenomenon.  
4. The equilibrium grain boundary concentration of impurities increases with decreasing aging temperature.  Time also is important at lower temperatures.  For example, Increased aging time increases the concentration of Sb in a 3.5% Ni - 1.7% Cr - 0.008%C - 0.06% Sb steel.  

Again, that is a bit dense.  The take-away is that mechanism is again impurity driven and the lower end of the temperature range is more damaging than the upper end of the range.  Also mentioned by Smith is that Ni, Cr, and Mn increase the effects of two step embrittlement.  Mo inhibits the embrittlement by precipitating out as phosphides (thereby tying up the P).  

 

The two-step embrittlement is the most likely form we have to worry about with modern steels.  I can speak anecdotally to the effects of temper embrittlement, having either performed or reviewed results for hundreds of tests where this was a potential issue.  When temper embrittlement occurs, everything about ductility (elongation, reduction of area, bending) and impact toughness suffers.  Tangentially related to temper embrittlement is hydrogen embrittlement ("rock candy" failure).  This is when (monatomic) hydrogen causes a problem in steel.  It is often removed at 500-600 F, as that is the fastest temperature at which it will diffuse out of the steel.  This still takes days in many cases.  Manufacturers put in ovens just for heating these parts.  I mention it here because they all have to soak these parts at that temperature for a long time before they harden the parts because they know that they can't do it after the fact due to temper embrittlement.  I have been in a lot of professional heat treating facilities (most often foundries with their own HT operation).  While I have seen a good number of questionable (and down-right horrible) practices, I have yet to see a professional operation that did not avoid this temperature range post hardening.  

 

Please let me know if anyone has any questions.  I can certainly dig a little more if anyone would like to know more.  I have access to several more resources, including potentially newer studies and tests, but I stuck with the 3 books on my desk today and my own experiences for now.  

Edited January 12, 2021 by Jerrod Miller
Correction on blue brittle name.

[Alan Longmire]

Thanks so much for doing this, Jerrod!

 

Since the topic that inspired this thread was folder springs, I'll stick with the "spring temper" angle.  For O1 folder springs (pre-Evenheat oven), I temper with the blade at 350 F for two one-hour sessions, then I (gasp!) use a propane torch with a flame spreader to gently draw the loose end of the spring to a light blue oxide color.  I know, bad form, but I do clean and degrease the springs carefully to get the most accuracy possible from this method.  I then let it air cool.  I have not had any problems with spring breakage doing this.  I suspect I'm getting to the very edge of 550 F this way, but I'm not sure.  

 

In the thread I deleted I also used swords as an example.  I have found that with 5160 I get good results tempering at 560 F.  This is as measured via thermocouple in a horizontal heat treat forge, and incidentally ends up turning the blade a nifty blue.  

 

However, I know different alloys have different temperature ranges in which they suffer embrittlement.  Is there a set of charts somewhere that show this? Larrin over at Knife Steel Nerds has a few for L6 I've seen, and a few for assorted super-stainless alloys, but is there one for, say, 1075 vs 5160 vs O1?

 

Finally, since I got the Evenheat I've been using AEB-L for folder blades and springs, and tempering the springs at 1145 F. I have broken two of those, although not in normal use.  I was nudging them a bit tighter in a three-point jig and they hit the limit of elastic deformation rather abruptly before I thought they would.  I got that temperature from a variety of sources that suggest 1100-1150 F as the ideal for that steel.  Any thoughts there?  AEB-L is relatively low in carbon compared to most stainless blade steels at 0.6%, with a correspondingly low chromium content (13%) to prevent the formation of large carbides.  That's why I chose it, for the fine grain and ease of sharpening compared to, say, CPM-154 or even good old 440C.  Is it possible that it has a temper embrittlement issue at 1145F? I haven't lost a single one I didn't abuse, but the ones that did break seemed a bit brittle compared to low alloy carbon steel springs. 

[Guy Thomas]

Great thread, I'll be curious to see what Jerod says about the AEB-L issue. I remember bringing temper embrittlment up many years ago when Howard Clark was still active on the original forum as I had just read about the phenomenon. I was curious because the tempering range of most swords places them squarely in the range where this could happen, but typically doesn't cause a problem. I can't remember details now but I believe the upshot was that modern industrial tool steels don't necessarily behave the same in our bladesmithing applications as they do in their intended and intensively studied industrial use. Probably helps that we don't typically (but with some obvious exceptions like L6) use high alloy steels.

 

A quick Google search for this on this forum shows it's been brought up many times since then, I have some reading to do!

Edited January 9, 2021 by Guy Thomas

[Jerrod Miller]

I'll dig in a bit deeper on Monday, I generally don't visit the forum on my phone and don't use my computer much on the weekends, but I felt I should check this thread today. 

 

I could be mis-remembering the blue brittle being blue on the fracture surface. It has been a while. I'll definitely search for that. I want to say that it may be from the blue oxide forming on the fracture surface because it often cracks during the temper, but this, too, could be wrong. 

 

The more complex alloys, like AEB-L, often start seeing a rise in hardness and drop in ductility in a certain range (often 900-1100 F) due to that range being the point where there is enough energy to precipitate out more carbides. After these carbides precipitate out, then you generally see the the hardness drop like a rock. 

 

Again, I'll dig deeper into this on Monday. I'll also address a few things from KSN.  There is some great stuff there, but some pretty misleading stuff, too. 

[Jerrod Miller]

This, like many things metallurgical, is hard to get into without going into many other things.  It is all related.  The following will be a bit overly simplified for the sake of getting to the important stuff for this discussion.  Other posts have more information on grain refinement and phase transformations.  

 

Grain refinement is necessary in steels because pretty much everything that goes badly with steel is related to grain boundaries.  This is why we normalize as we do to get the refined grain structure.  These refined grains go from austenite (FCC) to martensite (BCT) upon cooling during quench (it doesn't all always convert, we'll get there, but go with me for now).  When this happens we have a very stressed lattice, making it very hard and brittle.  When we temper the martensite we relieve the stresses a little bit.  This happens via diffusion, which only happens when there is enough energy (heat) to allow atoms to move in the lattice.  A good deal of the stress relief is from the carbon atoms being able to move from one interstitial location to a new one that allows the BCT martensite to get a little closer in shape (and stress level) to BCC ferrite.  Given enough time a temperature it will in fact fully convert.  While these carbon atoms are moving, they will even get a chance to get out of an interstitial location and fully bond with a metal atom (Fe or otherwise) to form a carbide.  See list below for carbide formation during tempering of martensite.  

 

Carbide - Formation temp (deg C)

epsilon carbide (Fe2-3C) - 100-250

Cementite (Fe3C) - 250-700

VC-V4C3 ~ 550

Mo2C ~ 550

W2C ~ 600

Cr7C3 ~ 550

Cr23C6 ~ 700

M6C (Fe3Mo3C, Fe3W3C) ~ 700

Edited January 12, 2021 by Jerrod Miller

[Jerrod Miller]

Carbides in the steel.  Ideally speaking, blade makers want very fine, evenly distributed carbides.  We do this by dissolving everything (temp and time is alloy dependent) so all the elements are evenly distributed.  And by "we" I am including the mill that made the steel.  Then upon cooling the carbides form (not that this is not the same mechanism detailed in my previous post).  These fine carbides form in austenite, and continue to form through the phase change to ferrite until the steel is too cool to allow for the formations to continue, or when all carbon is tied up, whichever comes first (generally we don't super slow cool the metal, so the energy to form the carbides ends first).  When heat treating a blade we try to keep the fine carbides from dissolving or ensure that they are thoroughly dissolved (when we have extended soak times before cooling).  These steps control the carbides present upon hardening, which in turn controls what precipitation of carbides can happen during tempering.  All this is very dependent on alloy composition.  When things get too complex, we have to start playing with retained austenite.  Speaking of which, that is one of the reasons why retained austenite goes away during a temper cycle.  The chemistry of the matrix changes due to the formation of carbides, thus destabilizing the austenite which converts to martensite.  

 

I will continue to add to this as I get time, but no more today.  But if there are any specific questions please feel free to ask and I will work them in to my next post.  

[John Page]

Great stuff Jerrod! I really appreciate your knowledge, there's so much to learn and most of the time it's difficult to digest from the text books. Although not specified, would you consider the ranges of Phosphorous in relation to Manganese found in knife steels (i.e. 1084 at <.4%P and .04%Mn) be above the danger threshold for One Step Temper Embrittlement? 

 

Also unrelated and more on a historic perspective, are the potential levels of Sulphur in smelted ore (and follow-on addition of Phosphorous [might be Mn, don't remember off hand] to mitigate Hot-shorting) be much more of an issue? In a quest to better understand early steel, there are a lot of metallurgical technicalities which, without any means to determine alloyed percentages, give raise to a lot of 'tribal knowledge' that was lost over time but undoubtedly gives a great deal of technical and mechanical approach to  how steel was hardened and tempered... 

 

Anyway, thanks for the info! 

 

John

[Alan Longmire]

Thanks again for doing this, Jerrod!  

[jimmyh]

In the 1095 temper chart I was able to find here, the unnotched charpy shows no evidence of embrittlement, but it shows up quite dramatically in torsion, rapidly dropping off at a fairly low temperature.

 

What causes this difference?

 

[Jerrod Miller]

15 hours ago, John Page said:

would you consider the ranges of Phosphorous in relation to Manganese found in knife steels (i.e. 1084 at <.4%P and .04%Mn) be above the danger threshold for One Step Temper Embrittlement? 

I think you have those backwards.  <0.04% P and about 0.40% Mn is fairly common.  Phos is bad.  It drives me nuts when people talk about adding extra P... "for contrast".  That is something I would say is never worth doing.  0.0000% P is the ideal amount of P in steel.  That being said, when specs call out 0.04% as a max P, they generally are producing steel at a much lower level than that.  And it generally seems to be the case that these levels are low enough to not be too problematic (though I'm certain some embrittlement is occurring, just not debilitating amounts).  

 

15 hours ago, John Page said:

Also unrelated and more on a historic perspective, are the potential levels of Sulphur in smelted ore (and follow-on addition of Phosphorous [might be Mn, don't remember off hand] to mitigate Hot-shorting) be much more of an issue? In a quest to better understand early steel, there are a lot of metallurgical technicalities which, without any means to determine alloyed percentages, give raise to a lot of 'tribal knowledge' that was lost over time but undoubtedly gives a great deal of technical and mechanical approach to  how steel was hardened and tempered... 

It is important to remember that early steel was utter garbage compared to modern steel.  I'm certainly no expert on early steel production, but what I gather to be the case is that they certainly didn't know about elements (atoms) and strictly worked with things more based on bulk materials.  E.G. When using this ore, you do x, y, and z, then you will get a certain final product.  If you do the same steps with ore from the other source then you will get completely different final product.  But if you use that other ore and you do a, b, and c instead of x, y, and z, then it is not as bad.  Some places got quite good with their trial and error, then repeating processes.  But I don't think they really knew that much about what was really happening.  Not that they needed to.  It is a good thing for us today that they were doing what they were doing though; it was an important step to get to where we are now.  

[Jerrod Miller]

37 minutes ago, jimmyh said:

In the 1095 temper chart I was able to find here, the unnotched charpy shows no evidence of embrittlement, but it shows up quite dramatically in torsion, rapidly dropping off at a fairly low temperature.

 

What causes this difference?

 

I would really like to know a bit more about what went into making that chart before commenting on it.  But given some of the other info on the page that came from, I don't think anyone should reference that site for good information.  

[Jerrod Miller]

On 1/9/2021 at 7:16 AM, Alan Longmire said:

Now that you've defined "blue-brittle" to refer to the color of the broken surface rather than the oxide color I always thought it was, I feel a bit more confident in that.  

I found a source, Principles of Metallurgy, L. Carl Love, that explicitly states that the name Blue Brittle does indeed come from the surface oxide layer that forms at that temperature.  I have corrected my initial post.  Apologies for any confusion.  

[jimmyh]

Unfortunately I don't know the ultimate source of that data or anything about the experiments that went into it, but it it isn't something he just made up. He traced over this one in pretty colors, which I believe originally came from some metallurgy book Alvin Johnston (of rec.knives) had.

[Gerhard Gerber]

I'll just echo, thank you for taking the time Jerrod.
I understand very little, but I read it all several times hoping something sticks for the day the light comes on.

[Brian Dougherty]

Thanks Jerrod.  I've come back and read through this a few times with some time in between to digest. 

 

I've been tempering 1075 folder springs at 550F.  I haven't seen any break (yet) and am now wondering if it is just luck, or if it is because the 1075 I have is clean enough to  not fall into the "one step temper embrittlement" trap.  The max S is 0.05% and max P is 0.04%, but as you say, the mill probably tries to run below that.

 

What do you think?

[Bruno]

Great info.   Thanks Jerrod for taking the time.

[Jerrod Miller]

Sorry for the delay.  There were very high winds here yesterday with power outages and such.  Didn't make it online yesterday.  

 

On 1/13/2021 at 5:24 AM, Brian Dougherty said:

I've been tempering 1075 folder springs at 550F.  I haven't seen any break (yet) and am now wondering if it is just luck, or if it is because the 1075 I have is clean enough to  not fall into the "one step temper embrittlement" trap.  The max S is 0.05% and max P is 0.04%, but as you say, the mill probably tries to run below that.

 

What do you think?

I think we, as bladesmiths, are quite fortunate in the quality of steel being produced for us these days.  Anything that is a good blade material is now made with all this knowledge in mind and the end HT known to be pushing the limits.  The steel specs call out certain ranges for the P and S, but most producers (mills) I am sure have their own internal specs that are lower still.  I would be a bit more worried about anything that was made "small batch" versus a large mill run.  The reason being is that there is a greater chance that a small batch furnace may not have the capability of refining.  For example, we have 2000 pound induction furnaces where I work, and we cannot do any meaningful refining; we have to buy good materials to begin with (stuff that was already refined in a larger production facility).  

 

[John Page]

Thanks again for the info Jerrod! When thinking about early steel, I often forget that there were much more dramatic localizations in production and the following localized knowledge and specialty that comes with it doesn't translate quite in the way that our modern understanding does. Much food for thought!

[Jerrod Miller]

15 hours ago, John Page said:

there were much more dramatic localizations in production and the following localized knowledge and specialty that comes with it doesn't translate quite in the way that our modern understanding does.

Fun anecdote:  Last week I needed to find something out about how exactly N works out in tool steels.  I sent an email to a guy I know and got sent back 36 scientific papers about 4 hours later.  That is an element that historic smiths never knew existed, and I was easily able to obtain a veritable mountain of research from around the globe in less than a day.  Now I just have to read it, understand it, and figure out how to apply it.  

[Gerhard Gerber]

Bumped my head against this badly......

 

My first folder was a 14C28N kit, I followed the advice on BF, all I could find, toe spring temper this steel 1000F-1100F, I tempered at 540C and it worked as a spring.

 

Folder #4 is also 14C28N, did the same HT and shattered it trying to adjust the spring slightly.

 

Local makers advised 300C for 2 hours, I'll give that a try next.

[Alan Longmire]

For AEB-L, which isn't that different from 14C28N, I temper for two hours at 620 C per the advise I found.  The one spring done that way I broke was due to trying to get too much bend.  I don't think 300 C would soften this stuff enough to be a spring?  But I am a total noob with stainless steels. 

[Jerrod Miller]

These super fancy steels are going to be made in either an AOD or VOD process and refined very well.  Therefore temper embrittlement should be a very remote concern due to the lack of impurities that tend to be a major contributing factor.  

[Gerhard Gerber]

15 hours ago, Alan Longmire said:

For AEB-L, which isn't that different from 14C28N, I temper for two hours at 620 C per the advise I found.  The one spring done that way I broke was due to trying to get too much bend.  I don't think 300 C would soften this stuff enough to be a spring?  But I am a total noob with stainless steels. 

That's my feeling as well, the advice was from a local maker that only started with the folders inspired by my efforts end of last year.  With all due respect to him, his folder haven't lived and worked for very long.
 

 

15 hours ago, Jerrod Miller said:

These super fancy steels are going to be made in either an AOD or VOD process and refined very well.  Therefore temper embrittlement should be a very remote concern due to the lack of impurities that tend to be a major contributing factor.  

 

Thanks for that Jerrod......every bit of information helps, that points at too short a temper perhaps?

[Jerrod Miller]

9 hours ago, Gerhard Gerber said:

points at too short a temper perhaps?

And/or too cold.  While I am certainly not an expert in the complex super alloy steels, what I do know about heat treating steels leads me to think that the more complex alloys should probably be tempered longer than the simpler alloys.  

[Alan Longmire]

The datasheets from Uddeholm-Bohler say temper for two hours and call it good.  I think Sandvik says the same thing.