Kalymnos – catastrophic anchor failure

Preliminary report on the St Savvas anchor, which failed with lethal consequences on 27^th March 2026

I would like to extend my condolences to all those who were touched by this terrible event. It is my earnest intent, through the pages of Crag Chemistry, to shine a light on such wild and unexpected failures as come at us out of left-field. I am certain that we can learn how to tame them.

Introduction:

It is not my intention to detail the events of the fatal accident, or to get drawn into what we, as a community, should or should not be doing in response to such events.

I have months of work ahead of me investigating various aspects of how such an anchor came to fail. At this stage I have not looked beyond the two anchor bolts themselves. However, I have discovered enough to state that this is a very unusual mode of failure – it is certainly the first time I have encountered it.

Further to this, I now know enough to state that there are other dangerous bolts of this type out there waiting to trap the unwary and thus wish this short report to serve as a warning.

Key Findings:

This is what I know with reasonable surety. Each point will be elaborated further down the page. The anchor was comprised of two bolts, and thus, where necessary, I’ll differentiate between them as lefthand (LH) and righthand (RH).

1. Bolt type: Wedge style expansion bolt 10mm x 60mm.
2. Material: Stainless steel 316L.
3. Percentage Austenite: close to 100%.
4. Brand: Petzl Goujon.
5. Age: approx. 24 years.
6. Failure Point: roughly transverse, some millimetres below the rock surface.
7. Failure Type: stress cracking –
   1. not Cl mediated stress corrosion cracking (Cl-SCC): there is no evidence.
   2. not SRB mediated sulphide stress cracking (SRB-SSC): there is no evidence.
   3. but likely prime cause is hydrogen assisted cracking (HAC): extreme cold-working of the rolled threads has caused a transition from chloride SCC to hydrogen‑assisted SCC.

Details:

1. Bolt Type:

Note: the damage to the surface is a consequence of extracting them from the rock.

Note the absence of corrosion products, even after 24 years in the field.

2. Material:

We have had samples of both bolts analysed by optical emission spectroscopy (OES). The major components fell within the 316L specification, and we were able to eliminate the possibility of embrittlement by the likes of sulphur and phosphorus.

You can view the lab report here.

3. Percentage Austenite:

Magnetic susceptibility was estimated by a micro-balance technique. These values were then converted to % austenite using a calibration curve derived from XRD measurements. Many thanks to Luigi Pisani for the XRD work.

It is a property of 316 to maintain its austenitic state during cold-working. However, considering the degree of cold-working revealed by the micrograph down page, and considering the nickel content is just on the 10% threshold, it is surprising that there has been so little conversion from austenite to martensite during processing of these parts.

4. Brand:

All the bolts on St Savvas appear to be the same make, and the nuts, where we have them, are stamped Petzl Goujon. I don’t have nuts for the failed anchor bolts, but the body of the bolt matches others I have.

This what they look like. Note that this an example from elsewhere in Kalymnos

This bolt would have been put on the market sometime in early 2000. However, attempts to match it to Petzl literature have been confusing. We have not been successful in locating a manufacturer’s specification for it.

5. Age:

We have adequate records to assert the bolts were placed in 2002, and thus the age must be approx. 24 years.

6. Failure Point:

The lefthand anchor, post failure, is pictured below. The righthand anchor presented in a similar fashion. This manner of snapping just below the surface is similar to that we see with the majority of SRB mediated fractures where we know anoxic conditions are a strict requirement.

Of course, this need not be SRB attack, and, whatever the mechanism is, the preference for anoxic conditions is worth noting.

7. Failure Type:

We are certainly looking at a progressive cracking phenomenon that is being driven by the stress field of the preload applied during installation. However, for the austenitic stainless steels, there are a number of quite different possible causes for such a thing, and we will consider each of them in turn. Firstly, let’s take a closer look at what is actually present before getting influenced by that we might expect to see.

Despite the fact the thread crests are a bit mangled as a result of being grabbed during extraction from the rock, we can see that smaller cracks seem to cluster into major crack systems. On the LH specimen, the crack system originating from a thread root is obvious. Some of the other systems may be originating from the primary failure crack itself. Remember we at looking at a pattern that has had decades to develop.

However, this simple view becomes more complex when we look closer. Note that I say complex and not chaotic because the underlying metallic microstructure seems to be exerting some tendency toward order.

Note the large number of fine secondary cracks and note also that often they are branching from the main cracks with large angular deviation. Their direction is certainly not conformal with the general direction of the stress along the length of the bolt.

Taking a closer look at the fine cracks, we see straight sections, orientated, not randomly, but in a reduced set of preferred directions. We also see zigzagging where a developing crack has jumped from one preferred orientation to another.

A final point to consider is the question of whether the fracture is running between grains (intergranular) or across them (transgranular). This can point to one mechanism rather than another.

And here we come to another important observation. The metal, even at the centre of the bolt, has been so work-hardened, presumably by the final dimensional reduction of the bar stock, that it is hard to discern actual grains. Numerous mechanical twins confuse the boundaries.

Guessing from the extent of some of the twin structures, it seems the average grain size is of the order of 80um. This is perhaps some five times greater than I would expect for a production bolt given the fact that smaller grain sizes favour higher tensile strengths.

If we look at a fracture edge where the microstructure is obvious. It seems the fracture is transgranular, even though the grain boundaries themselves are not easy to discern.

Let’s now consider some possible mechanisms.

mode 1: Cl-mediated stress corrosion cracking (Cl-SCC):

This is the star demon of the climbing bolt world. Everyone knows about it. Everyone attributes failure to it, and yet I am still looking for a single example that meets my standards of proof.

In the coming months, watch this failure at St Savvas get labelled as a Cl-SCC event. And, just like that, the matter will be all over – “sea cliffs, stainless steel, I told you so … blah blah blah … time to move on”.

Such is the danger of not knowing what you don’t know. As I will reveal, there is a hazardous, bolt-production issue right here, that we risk losing sight of in the rush to apply the Cl-SCC label.

But first, we must give SCC its due. It is most definitely a real phenomenon. It’s not magic. Its key thermodynamics are understood, and it is by working with these, that I can develop an argument that says it is unlikely to manifest on a climbing crag. If you like pain, then I detail the case for Tonsai/Railay here. The case for St Savvas would differ somewhat from this, but rather than sidetrack, I’ll go straight for the evidence.

SCC requires very high chloride levels at the attack site. By very high, consider that sea water is as nothing in the matter, and that saturated sodium chloride also would be hopelessly weak. However, if by some process, we could magic sufficient chloride into the spaces surrounding those anchor bolts, then 24 years later, the evidence of attack would be obvious. The SCC process is initiated by the development of manifold corrosion pits, some of which will develop the correct geometry to raise the stress of the tightened bolt to a level that exceeds the critical stress intensity for the steel. A crack will be initiated.

I don’t need to go on to discuss the mechanism of crack propagation. It is enough to know that a bolt that has failed by SCC, after 24 years, will show three things, a) corrosion products, b) a pitted surface and, c) any cracks under development will arise from a pit on the surface.

If we look at the photo below, we see stress cracks that have opened without initiation by a corrosion pit. In fact, there are no pits anywhere, and there are no corrosion products.

Whatever is initiating this cracking, it is not Cl pitting. This is something else going on. Look at the initiation of the major crack system at the thread root in the picture below.

mode 2: sulphide stress cracking mediated by sulphate reducing bacteria (SRB-SSc):

This is by far the most common agency of the aggressive corrosion of climbing bolts. I contend that most published attributions to Cl-SCC would instead have been assigned to SRB-SSC if checks for it had been made.

What are those checks? Well, these are three I use.

a) I look for the presence of corrosion products filling the first thread root past the fracture, and also within corrosion cavities of the fracture surface close to the outside. The distinctive iron sulphide, greigite, is absolutely diagnostic.

What we find is just clean, bare metal.

b) The SRB organism cannot thrive without leaving metal sulphides behind it. The Iodine-Azide test is a key diagnostic in this regard.

Both fracture surfaces recorded negative responses, making the active presence of the SRB organism very unlikely.

c) Normally, under the microscope, the signs of H embrittlement at the fracture are distinctive and are diagnostic of SRB attack. However, in our particular case, matters are not so straightforward.

All cases of SRB-SSC that I have examined have involved cold-worked 304 stainless steel with its substantial content of α’-martensite. This material provides for ready diffusion of atomic hydrogen, and consequently the possibility of catastrophic embrittlement over a wide area.

However, in our particular case we have measured the material as almost completely austenitic, and consequently hydrogen diffusion will be close to non-existent. Thus, no catastrophic hydrogen embrittlement would be expected. I believe this is why I have never verified SRB attack on a specimen of true 316.

Thus, we could reject the possibility of SRB-SSC simply on grounds that the steel is purely austenitic. However, there is no need to go out on that particular limb, given we have checks a) and b) as detailed above.

And just when we thought we had clarity, it becomes necessary to explain the morphology of those fine cracks I illustrated above. They point to a role for hydrogen in their propagation. In the photo below I have highlighted a number of such features.

We have disposed of the possibility of hydrogen embrittlement via SRB attack, only to find evidence for hydrogen assisted cracking (HAC). I’ll develop on this finding in the next section. However, note for now that, because a fully austenitic steel is almost impervious to atomic hydrogen, what we see here is the result of hydrogen diffusion along a developing crack up to the tip to facilitate its advance. This is very different to what happens when hydrogen is free to diffuse through the metal, as is the case for cold-worked 304 under attack by SRB.

mode 3: Hydrogen assisted cracking (hac) sustained by An extreme cold-worked surface:

The hypothesis here is that severe cold work at thread roots creates conditions in which cathodic hydrogen‑assisted cracking becomes the dominant failure mechanism, rather than classic anodic slip‑dissolution SCC.

That sounds good, until you need to piece all the diverse literature together to justify it. I won’t attempt to do so here. This is a preliminary report.

Instead, let’s stay with that we can see. We’ve already shown the fine cracking we suspect to be HAC, so how about this extreme cold-worked surface at the thread root?

Today, good engineering practice preferences the cold rolling of threads rather than cutting them. And that is what we have here. Most old bolts I examine have cut threads, and to find rolled threads on these 24-year-old Petzl Goujons came as a surprise.

Cold-rolling, as a properly controlled process, leaves the thread root with a compressive residual stress. This is considered beneficial for SCC resistance. However, when incorrectly controlled, the very opposite of that which is desired, ensues. The material of the root and the thread flanks not only becomes extremely hardened, but a high residual tension is imparted. And this, even before we consider the impact of this treatment at the microstructural level.

So, here we have a couple of thread roots from the LH anchor. The stress field tracks around the thread profile as would be expected.

And the RH anchor displays the same features.

Looking at what happens to the microstructure, it is clear all larger features are smeared and lost by the extreme strain.

The same applies to the RH anchor.

Given the fact that the investigation is at a preliminary stage, I’m happy to accept this HAC mechanism as the most likely cause. I believe we have sufficient evidence to treat all Petzl Goujons with suspicion until they are proven innocent.

Conclusions:

Until proven otherwise, all installed 10mm Petzl Goujon expansion bolts of vintage, say 2000 to 2005, should be treated as dangerous. It should be emphasized that this condition arises from the nature of the bolt itself, and is independent of its installed environment.

---

Addendum:

17th June 2026

Bartek raises a good point in the comments.

He argues that this picture –

raises a problem for the central failure hypothesis. If this section of the bolt has not been cold-worked by thread-rolling, why do we see these stress cracks forming. Good question. I went and took some more photos.

Let’s look at the bolt again.

I would argue that the entire profile (ignore the clip) has been manufactured by cold rolling. Maybe in multiple passes, but plastic flow of metal from low to high points is how it got that shape.

The original micrograph was taken from an area roughly corresponding to the point A. You can see the start of the upsweep to the clip retainer on the left. Let’s take a closer look, but this time after the surface has been etched to reveal microstructure.

It is immediately clear that metal in this section has been every bit as aggressively cold-worked as the threaded sections we were looking at earlier. Ok, we might not have the stress-riser of the thread root, but clearly there are enough microstructural defects to initiate cracking in response to what looks like the same residual tension as we saw with the threads.

Evidence for extreme cold work extends right along the non-threaded section, exactly as you’d expect if it had been formed by cold rolling. The photo below is from the region marked B.

Notice the complete absence of the original microstructure, and the multitude of microcracks initiating from local deformities such a inclusions where they break the surface.

So, it is easy to see crack initiation occurring, even without the stress-riser of the thread root, but I wonder whether the conditions are right to propagate such cracks beyond the work-hardened zone. This seems to require conditions offered by the thread root. Maybe the lower oxygen potential, and/or lower pH is necessary?

---

Please read:

This sort of investigation is slow, technical, and expensive. The independent laboratory fee for chemical analysis of two St Savvas bolts alone was AUD $770, and further testing is expected. Crag Chemistry will continue to provide the time, scientific interpretation, technical reporting, and publication of findings free of charge, but we cannot keep personally absorbing the cost of every serious failed-hardware investigation.

If you value proper testing over rumour and speculation, please consider donating to the –

*Crag Chemistry Independent Bolt Failure Testing Fund*: https://gofund.me/1b3f3aed2.

The purpose is narrow and transparent: independent testing, interpretation, and public reporting of failed climbing bolts, anchors, and relevant rock environments, so that the whole climbing community can learn from these failures.