Short Report

Now This Is Different

An example of an SS anchor chain exhibiting marked resistance to SRB attack

Many thanks to Luis Fernandes Silva for the sample.

On previous occasions I have reported on the sulphate-rich sea cliffs of Cabo da Roca. They are notorious for the aggressive corrosion of low grade austenitic stainless steels such as 304 (A2). Sulphate reducing bacteria (SRB) are involved, and the deterioration of rock anchors proceeds from hydrogen embrittlement to sulphide stress cracking (SSC) as illustrated below. See my multi-part series “Corrosion at Cabo da Roca, Parts 1 -5” for the background story.

I have discussed in detail why I think this phenomenon is restricted to austenitic SS grades where the nickel content is less than 10%, and, as a way of validation of this hypothesis, I point to the absence of such attack on 316 (A4) with its > 10% nickel content.

Thus, when we come across a corrosion resistant SS fixture on these crags, it is natural to assume it must be composed of 316. However, as we shall see for the case below, it is always worth taking a closer look.

What is different about this chain? It far outlasts any other stainless steel components installed on this crag.

Ignoring the other components of the anchor for now, let’s concentrate on the chain. This chain is used to connect a pair of expansion bolts and fixed hangers.

Its origin has been lost in time but is thought to go back at least twenty years. It is unlikely all the components of this anchor date from the same time. It is not something manufactured for climbing applications, and most likely was acquired from a boating supplier or ships’ chandler.

It has been my experience that dirtbag climbers sourcing cheap components from local hardware stores have been the cause of much grief. But here we have something different. This piece seems remarkably well suited to the corrosive environment.

My first act is always to check such a part with a strong magnet. In this particular case, I went one better and used a magnetic balance technique that is calibrated to return estimates of strain-induced α’-martensite. The results I obtained are illustrated below.

Using a magnetic balance technique, it is possible to estimate the volume percentage of strain-induced martensite at various locations. Cold-forming of low nickel steels results in amounts of α’-martensite as high as 40% which, in turn, makes the formed part vulnerable to hydrogen embrittlement. This chain link shows low levels of martensitic transformation which would explain it resistance to SRB attack.

The low levels of martensite transformation following cold-forming are in keeping with an austenitic steel with nickel content in excess of 10%. Note that the martensite concentration is highest in the bends where cold-forming strain is maximum. However, check the low level (1.8%) at the weld. This is unusual. Welded 316 or 304 would always be magnetic at the weld point given the inevitable formation of the δ-ferrite phase as the weld pool solidifies.

Mystery of low martensite content aside, there is no mystery in the fact that this chain would exhibit very low hydrogen diffusion rates and thus would be resistant to attack by sulphate reducing bacteria. The chart below (source) illustrates the fact that, based on magnetic analysis alone, we could expect this chain to last over 100 years in the environment of Cabo da Roca.

What’s the big deal about the transformation of γ-austenite to α’-martensite Simple, it drastically changes the rate at which hydrogen can permeate the metal and thus inflict damage. A typical 304 climbing anchor has levels of martensite as high as 40%, and thus, in a matter of a few years, the entire fixing is rendered dangerously weak. Contrast this to the fact that this chain would take over 100 years to reach the same level of hydrogen permeation (red ellipse). Source.

Let’s look further into the possible reasons why the strain-induced martensite levels are low in this chain link. At room temperature, the martensitic atomic lattice is more stable than the austenitic form unless there happens to be sufficient nickel to shift the thermodynamic equilibrium towards austenite. Even small levels of cold-working can provide the energy needed to initiate the transformation to martensite. Whilst other elements participate in the austenite stabilisation process, the nickel content becomes critical as the 10% point is crossed.

By critical, I mean really, really critical. The data below from Caskey illustrates how the susceptibility to hydrogen attack increases dramatically when nickel content falls below 10%. This susceptibility is thought to be the result of the increased martensite formation that accompanies the cold-working of alloys where the nickel content falls below 10%.

The tendency of austenitic SS to transform to martensite during cold-forming is critically dependent upon nickel content. And because hydrogen diffusion is enabled by martensite, we find that the susceptibility to hydrogen damage is equally critically dependent on nickel content. In the diagram, we see a drastic reduction in ductility at nickel contents less than 10%. Moreover, this is so for both commercial and high purity alloys indicating that the typical impurities of SS manufacture don’t modulate the role of nickel. Ref.

Bearing in mind the fact that nickel is an expensive component for the manufacturer of stainless steel, it comes as no surprise that grades of stainless steel tend to contain the bare minimum of nickel that they need to make the grade, and it is not uncommon, therefore, to come across “bargain” grades that have crept into the market where the nickel content is less than it should be.

It is obvious that grade 304 at 8% minimum is going to be a problem, and furthermore, that 316 at 10% minimum is right on the edge of acceptability. Although, with regard to the latter, the 2% molybdenum content likely improves the austenite stability.

So, this chain is most likely 316, yet …. the lack of magnetism at the weld is a clear warning not to jump to conclusions. So, let’s look a bit closer.

Wet chemical analysis reveals that this material is in fact 304 grade. Its nickel content is higher than one would normally expect, but nonetheless, it is 304. This finding was a surprise.

ElementWeight Percentage
Nickel 9.3 +/- 0.1
Chromium18.75 +/- 0.1
Molybdenum< 0.5
Analysis of a sample of the chain reveals it to be 304 not 316 as suspected.

Based on this composition alone we would expect much higher levels of α’-martensite from the cold-forming process, and furthermore, the weld point should be quite magnetic. The fact we don’t see this suggests that the part has been annealed following manufacture.

In keeping with the above possibility is the observation that the material is softer than I would anticipate for a cold-formed part in 304.

Hardness (LEEB D)AverageSD
this sample16525
typical cold-formed 30444020
Testing with a LEEB rebound tester reveals the material of the chain to be softer than would be expected for cold-worked 304. Has it been annealed post manufacture?

In fact, the figure of HLD165 seems too soft to be credible, and, whilst I have reservations about the accuracy of the tester for physically small samples such as this chain link, I remain confident that it is unusually soft for a cold worked sample of 304.

At this point in the investigation, I was fully expecting that a polished section, were I to take one, would reveal the microstructure typical of a solution-annealed sample.

What do I mean by that? Well, at temperatures just over 1000ºC, phases such as α’-martensite and δ-ferrite arising from cold-working and welding will rapidly transform to the more stable γ-austenite form, a process which then leads to the recrystallisation and uniform grain growth of the pure γ phase.

In the event, I found something different. It is neither representative of a cold-worked part, nor of a fully annealed part. In the picture below you can see the distorted and broken grain boundaries you’d expect from cold-working. However, the crisp, parallel slip-bands one would anticipate in cold-worked 304 are now diffuse and distorted, with their parallelism corrupted.

Longitudinal section of chain reveals distorted and broken grain boundaries indicative of cold-working, and yet, what I presume to be the original slip bands resulting from such a process, are diffuse and distorted. It looks like the specimen has been annealed at a high enough temperature to reverse martensite to austenite, but not high enough to promote recrystallisation and grain growth. Optical microscopy following electrolytic etching in 50% nitric acid at 2V for 20s.

My first thought was that my sample preparation was defective, but on reflection, it occurred to me that I might be observing a partially annealed state. I really wasn’t sure what that might look like.

A search through the published literature came up with the following rather nice paper from Naghizadeh and Mirzadeh . These authors have looked at this reversion-annealing process in some detail for both 304 and 316. The pictures below illustrate the rapid restructuring of a sample of 316 that has undergone 70% size reduction and then is heated to either 1000ºC or 750ºC.

The authors concluded that the process of reversion-annealing must involve three stages. Firstly, comes the transformation of α’-martensite to γ-austenite. Secondly, such austenite as is not altered by the cold-forming, begins to recrystallise. Thirdly, this later process leads to grain growth which then sweeps all the newly converted austenite up into a new homogeneous grain structure. This final step they conclude is diffusion-limited and will not proceed unless the activation energy is provided. Thus, they find that whilst a temperature of 750ºC will provide for conversion of martensite to austenite, the temperature has to be increased to 1000ºC for growth and resizing of the grain boundaries to occur.

Notice that that at 1000ºC the original distorted and broken grain-boundaries are replaced by new grain growth within 3 minutes. However, at 750ºC the process fails to proceed beyond martensite conversion, and no new austenite crystallization occurs. I’m guessing that because martensite is associated with slip-bands, then the conversion of this martensite to austenite without recrystallisation results in what I take to be poorly defined slip-band pseudomorphs within the original grain boundaries.

In any event, the similarity to what we are observing with the chain link makes me more confident to assert that it is likely annealed, but only to a temperature sufficient to convert the martensite to austenite, and not sufficient to promote grain growth.


I feel we are looking at something of significance here. Hitherto, when I have pointed to the apparent resistance of 316 to SRB attack, people quite reasonably have raised the objection that such resistance might be due to molybdenum content rather than martensite-mediated hydrogen transport.

Such an objection is certainly not fatuitous. There is considerable evidence, e.g. see , that molybdenum salts inhibit the enzymes responsible for sulphate reduction, and thus could contribute to suppression of SRB activity. However, without doing a Pourbaix diagram for MoS2 I can’t ascribe electrochemical significance to such a possibility. I will investigate this side story another day. For now, I am happy that we have an example of a low molybdenum alloy showing SRB resistance. It allows us to conclude that martensite development is the critical component of SRB vulnerability.

From a practical point of view, does this finding indicate that rock anchors manufactured from 304 could be rendered suitable for use in locations of high sulphate by annealing post-manufacture?

It is definitely possible. Afterall, we have the evidence right before us. However, is it practical? I feel that there is nothing trivial about the reversion-annealing process and can think of a number of ways in which one might merely swap hydrogen susceptibility for intergranular corrosion. From where I stand, I see manufacturers struggling to implement adequate quality control without adding the burden of a process that requires even higher levels of QC.

Naghizadeh, M. & Mirzadeh, H. Microstructural Evolutions During Reversion Annealing of Cold-Rolled AISI 316 Austenitic Stainless Steel. Metall Mater Trans A 49, 2248–2256 (2018).
Nemati, M., Mazutinec, T. J., Jenneman, G. E. & Voordouw, G. Control of biogenic H2S production with nitrite and molybdate. Journal of Industrial Microbiology and Biotechnology 26, 350–355 (2001).
Caskey, G.R. Hydrogen effects in stainless steels. in R.A. Oriani, J.P. Hirth, M. Smialowski (Eds.), Hydrogen Degradation of Ferrous Alloys, Noyes Publications, Park Ridge, NJ, pp. 822-862. (1983).

13 replies on “Now This Is Different”

Is it possible that the annealing is confused with a hot forging process in some very peculiar conditions of temperatures and cooling?

There is no doubt hot-forging brings its own problems with the management the heating/cooling regime. However, in this particular case, it is pretty clear that the links are cold-formed and welded.

I was given a hot-forged part (industrial height safety anchor) that looks like there was a problem as you suggest. It’s labelled 316, and I can confirm it is absolutely non-magnetic and spot tests for molybdenum, yet it has developed some hideous inter-granular mode cracks. The environment it was exposed to was outdoors but pretty mild. When I get time, I’ll do a proper analysis and write up the results.

Hi Dave, thank you so much for your detailed research. It’s incredible to see an example of 304 showing resistance to SRB attacks after decades.

I’ve read all your posts, but I still haven’t understood one thing. Why is being close to the sea apparently a requirement for SRB activity? Are there any cases of SSC or HE happening inland with 304SS? I couldn’t understand why being close to the sea is important for this phenomenon to happen. Do you have any insights related to that?

Hi Renato,
I have suspected for many years that annealed 304 must be resistant to SRB attack. But I do love to see confirmation from the field.

SRB will attack any cold-worked 304-bolt provided there is sulphate available. It is sulphate availability that has the dependence upon distance from the sea. I am working on a post that takes a formal look at why we have this dependency. My theory is that the sulphur source is elemental sulphur of volcanic origin.

If you have a look back to my early posts, you will see examples where I show that sulphate levels fall-off very rapidly with distance from the sea at Railay.

Another place to look is on the wall-wash-data-archive page. At the bottom of the page there are a couple of graphs that show the sulphate and chloride distribution for the entire 100 or so samples I have taken. You’ll see that sulphate rich sites, should they occur, are all close to the sea.

However, there are some exceptions. For example, the Secret Garden and Iannis sectors of Kalymnos, although inland, exhibit very elevated sulphate levels compared with the sea cliffs. And yes, they have SRB corrosion problems. My explanation for this is that they both show hydrothermal activity which is carrying sulphide to the surface.

I am yet to encounter any other occurrences of sulphate inland but am certain it is a possibility especially around geothermal activity.

Thanks for the details Dave, I’m excited for your future posts!

How closely in contact are you with UIAA? Given the discoveries you made I was surprised to see that the latest updates to the 123 UIAA standard focus only on chlorine induced stress corrosion cracking, without mentioning sulphate reducing bacteria. I hope your work can trigger changes in the standard so that we’ll have safer bolts for use worldwide.

I presented a summary of my work at the last UIAA Safe Com meeting in May. They have certainly been following my research, and I am very grateful for the funding they provided as I was trying to get the project off the ground.

From my point of view, it seems fair enough to take a cautious approach in making further recommendations. In the early days when bolts started failing in Railay, I feel they were too quick to jump to conclusions. Once a recommendation goes out, it is very hard to pull it back in.

Yeah, being cautious makes complete sense. Particularly when their recommendations coincidentally also solve the issue with SRB (even if going a bit overboard). There’s no reason to be too quick now.

One other thing, are there any bolts that were confirmed to have broken through SCC? Or was it wrong initial assumptions, with SSC being more likely? (like what happened with broken “316” bolt examples that turned out to not be 316)

I guess my question is whether all failures where SCC was deemed the culprit were actually SSC in disguise or if there’s a chance both things are happening. At least from digging in old Mountain Project threads, a lot of pictures of bolts that supposedly failed through SCC seemed to have telltale signs of SRB activity (if I learned one thing or two from your blog!)

As you say, a side effect of the new UIAA corrosion standard is that in guaranteeing SCC resistance it will also seem to cover SSC resistance. I say “seem” because there is no theoretical basis for such a conclusion.

The use of the G30 test means that in practice we are going to see just titanium products carrying the SCC stamp. If there is a good reason why SRB attack doesn’t lead to hydrogen embrittlement for titanium, I’m unaware of it, but I haven’t done anything more than cursory research.

We do know that controlling hydrogen embrittlement is a major issue for titanium fabrication, and we have fundamental reasons to suspect the resistance of titanium to a reducing environment such as that presented by the SRB anaerobic environment, yet the evidence of the survival of large numbers of titanium bolts at Railay and Cayman Brac says the magic of titanium extends to sulphate cliffs. It’s hard to argue with success, but I distrust “magic” as an explanation.

As for the question as to whether there are any real word examples of SCC corrosion out there, or whether they are all SSC, I’m open minded that the SCC might occur. However, I don’t see the requirements for SCC being easily met. Yes, a number of authors have published examples of anchor failure they have attributed to SCC, but none of these have eliminated the possibility of SSC by testing for metallic sulphide. As you say, many of the images look rather like SSC.

In the long run, I think the uptake of titanium anchors will be slow, and people will keep installing stainless steel. Provided the shift from 304 to 316 continues, then my guess is we will see the problem of aggressive corrosion fading into a bad memory. That certainly seems to be the case at Long Dong, where the evidence of a regular testing régime seems to point to the resistance of quality 316.

That makes sense. Thanks for the detailed explanation. The reason I’m so interested in this subject is because 316 is nearly impossible to find in my country (Brazil), and 304 equipment is just starting to become commonly available. I’m seeing lots of folks moving from using plated steel to using 304 in many crags inland since 316 can’t be found.

Given the standard discouraging the usage of 304 everywhere outdoors, it’s not clear yet to me whether we can really trust this new equipment. Maybe we could test for SCC or SSC vulnerability in our crags with your wall wash/spot tests? I’d definitely be willing to do that.

I’m worried we might start having problems in 5-10 years like in other places around the globe. Although I know this risk is low with inland crags, the standard update is scary since it explicitly mentions it can occur inland particularly in karst limestone + tropical climate which is exactly what we have here in many places.

Yes, I’ve heard about the problems of sourcing 316 in Brazil. If the supply was no problem, then by all means apply a rule that excludes the use of 304, but in reality, you guys need reasons more than you need rules.

There are huge numbers of 304 bolts giving excellent service in coastal locations across the world. There have to be specific reasons to declare this material cannot be used. For sure, the presence of sulphate is one such reason. Currently I know of no other. I have no evidence to indicate that “tropical karst” presents a peculiar problem.

I’ve sampled the sea cliffs at Campo Escola da Barra da Lagoa where 304 has been in use for many years. Have a look back through the blog to find my two posts on this crag. Apart from one exception I haven’t had time to investigate further, there is no evidence for SRB attack, and the metal work shows little more than the superficial chloride attack one would expect.

I have samples from a sea cliff immediately south of Laguna, Santa Catarina. Here we see 304 glue-ins exhibiting brittle failure. I can confirm that this is not SRB attack but haven’t had time for a more thorough metallurgical examination.

Sulphate crags are all associated with access to volcanic activity because the sulphur has to come from somewhere. I currently have just one crag that seems to defy that rule, but it can’t be magic, the sulphur has to come from somewhere 🙂

Have a look at the sulphate spot test kit in the “Sampling Kits” menu pick. I reckon this is a very good way of evaluating if a potential crag development is going to give you a problem. I’d gladly send you one, but I know we would have dramas with the carrier and customs. They are very easy to make if you have access to a university lab.

Hey Dave, re-reading this comment of yours, do you mean you have a sample of brittle failure of 304 that did not happen through sulphide stress cracking? So this means there’s another issue with 304 at play around the world apart from SSC?

I’m always on the lookout for a case of brittle failure which tests negative for the presence of sulphide. I currently have two such samples, but I cannot be sure about the failure mechanism at this stage until I get time to take a closer look at the metallurgical structure.

Some years ago, Fixe 304 chain sets were failing through fracture of the top link of the chain. Although consultants to Fixe attributed this to SSC, others, including myself, attributed the failure to poor quality control of the material used to form that top link. It was way harder than should be the case for moderately strained 304, and analysis showed the phosphorus content to be higher than specification.

My current position is that 304, provided it is within specification, is a very robust material, and its only weakness is its susceptibility to SRB attack where the environment contains large amounts of sulphate. However, I am working my way through a huge pile of failed anchors, and it may be that in a year or so I’ll come up with an exception to that rule.

Leave a Reply

Your email address will not be published. Required fields are marked *