We Observe Brittle Fracture
The third part of a multi-part series
In conducting this work I have been greatly assisted by Luis Fernandes Silva and Rui Rosado who have provided multiple samples and photos for analysis. Luis has spent a great deal of time measuring spot-sulphate levels, and the pH profiles of bolts he has extracted during route refurbishment. We now know a lot more about why the Cabo da Roca crags eat stainless steel.
In Part 2 of this series, I presented the phenomenon of sulphide stress cracking (SSC) as it applies to the bolted infrastructure installed on the sea-cliffs of Cabo da Roca. Here, in Part 3, I will make a start on what is going to be a long journey, as we explore what is known about the brittle fracture phenomenon we also observe at these cliffs. As it turns out, the two are related. SCC is the end-game of a years long process of hydrogen embrittlement.
2. What do we mean by embrittlement?:
As any equipper knows, extracting a well-placed, 10mm SS bolt from solid rock is a mission. Armed with just a light hammer, it can take an enormous number of hammer blows to bend the bolt back and forth, and to mangle it to such an extent that it embrittles through work-hardening, and ultimately snaps. Most people reach for the grinder long before that happy event occurs.
We know from real world experience, that a properly manufactured rock anchor, fashioned from either 304 or 316 SS, can take a lot of physical abuse before it might be considered dangerously brittle.
Consequently, when I’m presented with a clean, almost shiny 304 bolt that has snapped at just below the rock surface without any sign of putting up a fight, I’m forced to consider that there has to be something unusual going on. Take a look at the bolt illustrated below, and consider that this is not just one freak occurrence, but is a fairly common presentation for an aged bolt at Cabo da Roca. Yes, many might show more corrosion than this, but pretty much all can be dispatched with a few swipes of a hammer.
Brittle Fracture – a quick recap:
If we recall that stress is defined as the force applied per unit cross-sectional area, while strain is the proportional elongation of the part being stressed, then we can formalise the distinction between brittle fracture and ductile fracture according to the stress-strain diagrams below.
Consider what happens as an outward force is applied to a bolt. As the applied tension increases, so the atomic lattice of the metal begins to distort and the part lengthens. However, if the force is removed, the atoms move back to their stable spacing and the part reduces to its original size. This is merely the elastic behaviour that is shared by all metals to varying extents.
However, a distinction between brittle and ductile metals occurs as the material is stretched beyond its elastic limit. For a brittle material, at some limiting value of stress, a crack will suddenly, and very rapidly, propagate through the metal leading to total rupture. Once broken, the material returns to its original dimensions, with the obvious exception that we now have two parts instead of one.
On the other hand, for a ductile material, crack propagation does not occur, and instead, a mechanism exists by which atoms move, one past the other, before stress can build-up to a level sufficient to cause fracture. The value of the stress at which this atomic movement occurs is called the yield point, and is illustrated in the diagram below. The deformation that occurs up to this point can be considered purely elastic. Moving beyond, results in irreversible plastic deformation.
For alloys like 304 and 316, it is possible for the applied stress to be increased well beyond the yield point, and yet the material will continue to elongate in a predictable manner. However, there are limits to such ductility, and, with increasing strain, the material gradually work-hardens, micro-voids begin to form within the bulk material which then coalesce, eventually leading to the rupture of the part. This phenomenon is known as micro-void coalescence failure (MVC), and typifies the ductile case. In the diagram above it can be seen that a point of maximum tensile strength is reached, beyond which MVC progressively weakens the material leading to rupture.
The elongation at break (EL), measured as a percentage of the length of the test specimen, is commonly taken as a measure of the ductility of a material. Furthermore, because the test specimen ‘necks down’ as well as elongates, another measure of ductility we will come across is the percentage reduction in cross-sectional area (RA) at the point of fracture.
The foregoing principles are illustrated below.
This is what we are finding in the field:
We can generalize the mechanical properties of the 300 series austenitic stainless steels as follows.
- Yield strength: 250MPa approx.
- Ultimate tensile strength: 600MPa approx.
- Elongation at break: 70% approx.
- Reduction in area: 60% approx.
Although it is reasonable to assume all anchors installed at Cabo da Roca started life possessing the typical properties above, after a number of years in service, many of them acquired new, dangerous, and quite unexpected characteristics.
The bolt illustrated below was fractured by a small number of hammer blows, all applied in the same direction. The stand-out feature is that there is almost no perceptible deformation of the form of the bolt. That is, the fracture occurred without yield. The failure appears to have been initiated by a small crack forming at the point of stress concentration within the thread root. In this particular case, there is no evidence that the fracture initiated at a point of prior SSC corrosion. Indeed, this example seems to be completely free of sulphide corrosion products. However, as will be discussed in a subsequent post, microscopic and chemical examination inevitably reveal the presence of sulphide.
A further illustration of the fact that fracture occurs before yield is shown for the bolt below. Here we see the same initiation process with a small crack forming within the thread root. However, in this particular case we have ample evidence of sulphate reducing bacteria (SRB) activity with mackinawite coating the thread flanks, and crystals of greigite exposed on the fracture surface. These sulphide corrosion products will be discussed in detail in a subsequent report. The fracture surface also reveals the small irregular cracks and multiple voids that are typical of the brittle fractures we are finding.
If we further examine the bolt illustrated above, it is apparent that to some extent it has bent in the direction of the hammer blows. Not all bolts bend in this manner when snapped, but the fact it occurs at all indicates that in some cases part of the original ductility is retained.
By measuring the thread pitch (nominally 1.5mm) at different points along its length, we can see that one side of the bolt has been elongated by as much as 20% at a point some 3mm to 6mm out from the fracture. It is to be expected that, should SRB be having a deleterious effect on the metal structure, such effects will be restricted to the anoxic regions of the assembly beneath the hanger, and thus the outer sections of the bolt may well retain its original ductility. At this stage we lack the evidence to draw a firm conclusion.
Clearly, a material that started out as tough and ductile, has become embrittled. Why?
3. Hydrogen Embrittlement (HE):
There are many interwoven threads to this tangled story, and it is difficult to identify just that one starting thread, the one, should we begin to pull on it, that will untie the knot. To make matters worse, ultimate explanations have to be sought at the atomic level, so there’ll be no avoiding a visit the esoteric world of atoms, before we are done. However, that’s a subject for later posts, for now let’s see what can be seen at the macroscopic level.
I have previously argued elsewhere that there are very sound reasons why an anchor constructed of 304 will be vulnerable to degradation by atomic hydrogen should it be located in an environment where cathodic charging of the metal surface is a possibility. The presence of SRB in the anoxic spaces associated with the mechanical structure of the anchor is an example of such an environment.
The connection between the presence of SRB and the potential for cathodic charging of the metal surface with hydrogen is accepted by many authors in the field, eg see . Indeed, taking the phenomenon piece-wise, the thermodynamics are sound, plausible equations can be written, and it looks all good until we attempt to build the bigger picture. I previously described my initial findings in Part 2 of this series, and there I pointed out that they didn’t align with the few published studies where sufficiently rigorous steps were taken to establish the complete MIC mechanism. However, not being able to explain the complete mechanism doesn’t mean hydrogen embrittlement isn’t part of what we see happening. The potential for HE is clear.
So, with a bit of hand-waving for now, we can state that, where a section of metal presents a cathodic potential of sufficient magnitude, hydrogen ions at the surface accept electrons from the metal to form hydrogen atoms. As the process proceeds, hydrogen atoms pair up to form hydrogen gas. Because the paired atoms are more stable than the individual atoms on this cathodic surface, the concentration of single atoms never rises to a significant level, i.e. they pair-up before the surface concentration can rise to any real extent.
Well that is until we add SRB with their sulphide metabolites into the mix, and then the equilibrium of the above cathodic reaction is disturbed. The pairing of atoms to form hydrogen gas is inhibited. (For this reason, sulphide is called a hydrogen recombination poison.) With pairing suppressed, the atomic hydrogen concentration of the surface increases, and hydrogen begins to diffuse into the metal itself.
Atomic hydrogen diffuses along the “highways” of strain-induced martensite that are to be found in most 304 products that have been cold-formed during manufacture. The basis for this assertion can be found in a previous post.
Although the rate of diffusion is low, it can be demonstrated that it is nonetheless sufficient to reach into, and thus possibly degrade a 10mm climbing bolt over a period of a decade or so.
In the laboratory:
There is an abundance of experimental evidence to show that as the hydrogen content of the metal increases there is a concomitant increase in its brittle nature – most notably a decrease in the elongation to break .
The nature of the impact of hydrogen is interesting. From our current macroscopic viewpoint it looks as if the mechanism that allows atoms to glide past one another, and thus give rise to ductile flow, is unaltered, but, the extent to which such flow can be tolerated before propagating a fracture, reduces with increasing hydrogen concentration. In a later post I’ll descend into the world of atoms, and the difficulties of giving a a single coherent explanation for what we see happening here. For now let’s stay with the macroscopic engineering world-view.
An obvious question to ask is what happens should SRB activity cease? Surely such an event is entirely possible given natural fluctuations in the levels of available sulphate (see previous post). Such a consideration leads to two further inter-related questions. Firstly, will hydrogen diffuse back out of the metal, and secondly will the material’s mechanical properties return to the original state should it do so?
If the answer to both questions is yes, then, on this reckoning, if hydrogen embrittlement were to be a problem, there would need to be a persistent hydrogen presence for a significant burden of the gas to be accrued, and more critically, such a burden would need to be present at the time a stress demand happened to be applied, i.e. at the time someone chose to hit the bolt with a hammer.
The above considerations beg the important question of whether or not “hydrogen damage” can accrue over time. At the most general level of inquiry, both theory and laboratory evidence answer in the negative. The hydrogen burden presents a risk that may never manifest in harm, and essentially is transient in nature.
The real world, however, has a habit of inserting sufficient exceptions in our well-crafted generalisations to render them of limited utility. For starters, the movement of atomic hydrogen in and out of the bulk metal is highly asymmetric, and very dependent on a whole pile of real world considerations. For example, the degree of work-hardening of the material, the temperature at which such processing took place, and the nickel content of the material are all factors that will significantly impact the hydrogen burden that accrues, as well as the potential of that burden to cause damage. I have partially covered some aspects of the forgoing assertion in a previous post.
The second intrusion of reality arises because all bolts harbour residual stresses from the manufacturing process. Simply put, if a bolt is forced into yield by the production process, as nearly always will be the case, then it will contain internal stresses greater than the nominal yield stress unless efforts are made to include a stress-relieving step – something that is rarely done. The more highly strained the material, the greater the level of residual stress. At the lower end of the range, sitting somewhere above the nominal yield stress for the material, will be the relatively gentle bends of a glue-in bolt. At the other extreme, harbouring stresses close to the ultimate tensile stress we have the rolled threads of expansion bolts. The cut edges of pressed components such as fixed hangers will be somewhere intermediate with respect to the aforementioned extremes.
So what happens if the hydrogen content is sufficient to reduce ultimate tensile strength right down to that of the yield point as we see for the material illustrated below?
It is likely that micro-cracks will form to an extent sufficient to relieve internal stress fields. At this point I’ll ignore the issue of whether such cracks will coalesce or propagate, and simply note that, at least in this case, the damage will persist after hydrogen is withdrawn.
The fact that we commonly encounter stress-cracking of bolts at Cabo da Roca, gives some support to the idea that hydrogen embrittlement might reduce metal strength to such an extent that residual manufacturing stresses are relieved by a process of crack initiation.
However, as we shall see as this tale unfolds, I am yet to be convinced that the accrual of damage is a necessary condition for the brittle fracture that occurs when a bolt is dispatched with a few swings of a hammer.
At the crag:
If we should occasionally come across a bolt showing brittle fracture, say as often as one case in many hundreds, then it is likely we’d write such failures down to rogue stainless steel that escaped quality control. However, when confronted with a crag where most, if not all, bolts can be snapped with a few hammer blows, then we are forced to consider that a common mechanism might be at play.
In the above picture, we see clear evidence of that common mechanism. When struck with a hammer there is a tendency to break at the same point in the anoxic zone below the fixed hanger. There is very little evidence of deformation before fracture, and there is a tendency for the fracture to be stepped (red arrows). How does this phenomenon relate to the classic symptoms of brittle fracture observed in the laboratory, except perhaps to assert that the elongation to break is very much reduced in these examples? We need to look more closely.
4. Taking a closer look:
Below are six bolts that I have chosen for detailed examination.
Specimen #2, #3, #4 and #5 show stepped fractures, whilst #1 is somewhat conchoidal. Specimen #4 exhibits some deformation at a point approx. 6mm remote from the point of fracture, whilst the others fractured with no obvious deformation. Specimen #6 exhibits brittle fracture of the solid portion that remained subsequent to the development of a field of stress cracks.
Note that for #4, #5 and #6 we have wall-wash analyses for the adjacent rock surface. You can look up the results here.
Because strain-induced martensite (∝’-martensite) plays a significant role in hydrogen embrittlement (see this post for details), I have started estimating the percentage martensite composition of the bolt samples I receive by means of a simple magnetic balance. I will publish details of this method, and its calibration, in a future post.
For the bolts feature above, I obtained the following estimates of martensite composition.
|Specimen||Near Fracture (%)||Away From Fracture (%)|
|#2||20, 15||60, 55|
|#3||70, 85||50, 30|
Apart from #4, all samples showed moderate to high levels of martensitic conversion. It should be noted that, although #4 stands out as low, it nothing like as low as that exhibited by 316 which is typically around 2% to 3%.
Although I am attributing ferro-magnetism to the presence of strain-induced martensite, i.e. the phase which forms from austenite as a result of the deformation of the initial bar stock used in production, we should be aware of a potential wild-card in the form of \delta-ferrite. This magnetic phase certainly is present to some extent because we can see it in in polished sections. The quantity of \delta-ferrite is a feature of the quality of the bar stock used by the manufacturer, and depends upon the soak-annealing time applied by the foundry. It is typically a few percent, but I suspect that, on occasion, it may exceed this figure. Like all ferritic phases it is a conductor of atomic hydrogen, so, to a first approximation, we can lump its impact in with that of martensite. However, I hesitate to dismiss \delta-ferrite as insignificant in the face of superior quantities of ∝’-martensite, and we will swing back round to take a closer look as the subject of a later post.
Given that is is difficult to avoid at least partial transformation of austenite to martensite when forming a 304 stainless steel product at room temperature, the question arises as to how much martensite is acceptable when the product is to be used in a potentially aggressive hydrogen environment? I published a detailed analysis of the issue here. The figure below is taken from that post. My conclusion was that any steel showing more than 15% converted martensite was potentially a problem.
I have used this relationship to derive some worst case (ie most aggressive) estimates of how long it would take atomic hydrogen to penetrate these bolts to a depth of 40% of their diameter.
Given that all these bolts had been in service for something like 20 years, give or take 5 years, it is interesting to note that, with the exception of #4, this service duration would have been sufficient for hydrogen to migrate well into the interior of the bolt. For the sake of comparison, I’ve added an estimate for 316 SS. The difference is striking.
It is intriguing that, for 5 out of 6 of the samples, the magnetic data support the observed lifespan of 304 at Cabo da Roca. In addition it provides a possible explanation for the observation that 316 lasts very much longer under the same conditions.
It should be interesting to see how well this relationship holds up in the harsh light of future data. If it proves to be sound, then we have the basis for a simple but effective bolt classification tool.
In my next post I will examine the six bolts in greater detail and reveal that, uncorroded as they might seem, SRB have deeply penetrated the metal leaving tell-tale sulphide deposits. And in the post subsequent to that, I’ll be taking a close-up view of the fracture mechanism.