Multi-Part Report

Corrosion at Cabo da Roca -5

Hydrogen Embrittlement


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 the bolts that he has extracted during route refurbishment. We now know a lot more about why the Cabo da Roca crags eat stainless steel.

1. Introduction:

We return once more to the brittle fracture phenomenon that plagues the bolted infrastructure installed on the sea-cliffs of Cabo da Roca.

 In Part 3 we looked at the way in which, under the agency of sulphate reducing bacteria (SRB), a tough, resilient material like 304 stainless steel could possibly be transformed into a hazardously brittle one, something that can happen within a few short years. We suggested hydrogen embrittlement (HE) as the likely cause.

In Part 4 we provided evidence for the involvement of an unusual, acidophilic, sulphate reducing bacteria (SRB) that sponsors penetrative corrosion deep into the interior of the bolt. We showed that the prevailing redox chemistry is likely to favour significant levels of cathodic charging of the interior with atomic hydrogen. If true, such would mean that hydrogen embrittlement as a consequence of attack by this bacterium is certain.

The final link in the chain of causality is to provide evidence that the observed fracture does indeed result from the loading of the metal with atomic hydrogen. In this post I will attempt to provide that evidence.

We will need to descend into the world of atoms where things are going to get a little geeky. It is going to be difficult to strike a balance between too much tech-speak on one hand and furious wand-waving on the other. Let’s see what is possible.

2. The Continuum Mechanical World:

First-up, let’s start with the world view where atoms are so small that the mechanical properties of materials are, for all intents and purposes, continuous. In this world we are concerned with bulk properties such as density, hardness, machinability, formability, tensile strength etc. all of which dictate the suitability of a material for a particular application.

Climbers, frequenting the crags of Cabo da Roca, have good reason to want to understand how a ductile material such as that illustrated on the left in the picture below can be transformed into the dangerously brittle material on the right. Both are 304 stainless steel.

There are well established engineering concepts with which to frame such an inquiry, and in Part 3 we took a bit of time to look at the way in which brittle materials differ from ductile materials in terms of the engineering stress/strain relationship. We went on to describe the way in which the uptake of hydrogen by the metal impacts this relationship.

The diagram below is taken from Part 3 and summarises the most obvious impact of hydrogen on the continuum properties of the metal, i.e. a marked reduction in the so called “elongation at break” or EL parameter. This is the most convenient indication of embrittlement at the macroscopic level.

The most striking effect of hydrogen on the engineering properties of a metal is the marked reduction in the elongation at break parameter.

Keeping the key concepts of elastic deformation and plastic deformation in mind, let’s shift down to the atomic level, and see what is happening when a ductile material is stressed beyond its yield point.

3. The World of Atoms:

For a start, we observe that the constituent atoms of the alloy pack together in close cubic symmetry. There are two common cubic packing schemes, namely the body-centered cubic arrangement (bcc) representative of ferritic phases such as δ-ferrite and α’-martensite, and the face-centred cubic arrangement (fcc) observed for the austenitic or γ phase. All three of these phases can be present in a commercial 304 SS climbing bolt.

In the above illustration it should be understood there is no difference between the blue and white spheres. The colour is being used simply to make the difference in the packing arrangement more obvious.


It is easy enough to find an explanation for elastic behaviour if we consider each atom is positioned with respect to its neighbours at a “sweet spot” where the repulsive forces pushing the atoms apart are cancelled by attractive forces pulling them together. Thus, an ever-increasing force needs to be applied should we wish to distort this lattice from its ideal form, and, following this line of thought, it stands to reason that the more the lattice is distorted, the more the potential energy of the system will be raised. Clearly, on removal of the perturbing force, the lattice will return to its preferred arrangement. This is the elastic deformation observed in the continuum mechanical world.

Brittle Fracture

Keeping in mind the fact that the blue and white atoms are equivalent in all regards, then to a first approximation, each atom is hanging onto its respective neighbours to much the same degree, Thus, in the face of a relentlessly increasing external force, the resultant internal forces acting to pull atoms apart will be uniformly distributed. However, we can anticipate that eventually, at some point within the elastically distorting lattice, the stress will be sufficient to break the metallic bond between a pair of adjoining atoms. This initial rupture immediately transfers greater stress to the bonds of neighbouring atoms, which in turn break, thus setting in train a rapidly propagating rupture through the erstwhile intact material. In the continuum mechanical world, we observe a sudden snapping to reveal a brittle fracture.

Ductility and Dislocations

So far so good. Such is brittle fracture, but what of ductility and plastic deformation? From, the foregoing it should make intuitive sense that one cannot simply drag layers of atoms one past the other without precipitating a brittle fracture. We need to dig deeper into the atomistic world view if we are explain ductile flow.

It turns out that the answer lies in the presence of imperfections, or defects, in the ordering of the lattice, and we find that ductile materials have large numbers of such defects scattered throughout the bulk of the material. These endow unusual mechanical properties. The structures created by such imperfections in the crystal lattice are known as dislocations.

Dislocations are structures that can vary in scale from that of single atoms, to rows of atoms, to entire layers of atoms, and also can include phenomena such as heterogeneous inclusions. Wikipedia has some good information on dislocations here, but it leaves a lot to be desired as a clear explanation of the phenomenon. Let’s see if I can do better, albeit by resorting to a gross simplification of a very complex subject.

Below we see illustrated an edge dislocation. It is as if I had inserted just half, rather than a complete sheet, into a stack of sheets of paper. In the illustration we are viewing the stack, side-on, with the edges of the pages oriented vertically.

Schematic of an edge dislocation. The inverted ‘T’ indicates the dislocation line, which can be thought of as where the ‘extra’ plane of atoms starts. The line runs into the page. Credit

It is easy to understand that the atoms proximal to the line of the dislocation are displaced from their “sweet spot”, and thus the potential energy of the system must be elevated compared with that of a perfect crystal. This is illustrated below. Notice however, that the dislocation represents a stable, though higher energy state – an alternative, but “not so sweet spot” as it were. It is stable in that a certain amount of energy needs to be supplied if it is to return in the reverse direction of the blue arrow to the “perfect” state.

Simple schematic of transition from a low energy state to a stable higher energy state, such as presented by a defect. Credit

So what happens should sufficient energy to cross the energy barrier be supplied in the form of a persistent external force? Well, as it turns out, something quite remarkable. As a bond begins to release on one side of the dislocation, a new one is formed on the other, as illustrated below. Certainly the energy barrier discussed above needs to be crossed, but, for a ductile material, the energy involved to move a dislocation is much less than that to initiate a crack, and thus we see dislocations migrating at relatively low levels of stress compared with that for fracture.

Schematic of edge dislocation migration due to applied shear force. Credit

Note that, although it is the dislocation structure that is migrating through the lattice under the influence of the applied force, there is a bulk displacement of metal associated with this process, i.e. ductile flow.

A further important point to note is that, following each displacement of the dislocation, the potential energy of the system returns back to its original value and thus, at the continuum mechanical level, we should not see an increase in elastic potential energy. In fact, this proves to be true: we observe plastic deformation, rather than elastic deformation.

It makes intuitive sense that dislocations will migrate preferentially along those planes within the crystal lattice which provide the closest spacing of atoms. Such preferred planes are called “glide planes”, a somewhat confusing term, given that, as I have just explained, planes of atoms most definitely are not gliding one over the other. But, be that quibble as it may, we find that austenite with its fcc structure has more glide planes available to it than bcc ferrite, and thus, it comes as no surprise, that the ductility of austenite is greater than that of ferrite.

Limits to Ductile Flow

However, there have to be limits to the distance that any one dislocation can migrate, and consequently, there are limits to the extents of ductile flow. The most obvious one is that imposed by the intergranular boundaries. A second, obvious limit is the one imposed by the abrupt shift in lattice orientation between twin crystals. However, there are some less obvious structural barriers to dislocation migration, such as those presented by the inclusion of secondary phases within the bulk material, as is the case, for example, with the precipitation harden-able stainless steels like 17-4PH.

The above barriers to the movement of dislocations are easily visualised, but there is one other we need consider. Because dislocations may propagate along a number of glide planes, and because such glide planes have multiple lines of intersection one with another within the lattice, it is inevitable that one dislocation can block the motion of another leading to a dislocation pile-up. As we shall see, this is certainly the case for austenitic stainless steels.

Void Coalescence

So what happens if we continue to stretch a ductile material? Certainly, dislocations will begin to pileup at locations where their free motion is impeded. At such points we’d anticipate the buildup of local stresses and, ultimately, the formation of localised micro-voids as metallic bonds fail. These voids coalesce to create larger voids and lead ultimately to the ductile failure of the part. This is the failure mode we’d expect to see when a climbing bolt is overloaded to the point of failure. It should never occur in normal use, and it certainly differs to that we are seeing at Cabo da Roca.

4. Embrittlement by Atomic Hydrogen:

At the continuum level we know that, should a steel such as 304 absorb hydrogen, the predominant outcome is a reduction in ductility. Given that ductility depends upon the unrestricted flow of dislocations through the material, it is tempting to assume that hydrogen must act by restricting such flow. Alas, if only life were so simple! Even after decades of research, the reality remains obscure. Let’s explore some possibilities.

HELP, Hydrogen Enhanced Local Plasticity

There are a number of competing theories that attempt to explain hydrogen embrittlement, one of which, hydrogen enhanced local plasticity, HELP, is a leading contender. But, hold-up one minute! What is this talk about “enhanced local plasticity”? Surely it is the very opposite of increased brittleness? Are we seriously suggesting that hydrogen embrittles metal by increasing its plasticity? Err, yes we are, but in a roundabout way.

This subject gets very geeky and, for the masochist, I recommend for an overview of the various HE theories. However, such stuff is seriously heavy going, and for now, at the risk of offending the purists, let’s go for a grossly simplified explanation.

If we start with a block of austenitic stainless steel, then, should atomic hydrogen be produced at its surface via a cathodic corrosion reaction, it will diffuse into the metal to eventually become evenly distributed.

Well, not quite, the hydrogen would become evenly distributed were it not for the many dislocations that confer on the austenite phase its ductility. As illustrated below, we find that hydrogen accumulates in the vicinity of the dislocations because there is more room between atoms at this point.

Illustration of a H Cottrell atmosphere associated with an edge dislocation. The blue circles represent the metallically bonded atoms of the metal, and the red circles the atomic hydrogen present as an interstitial solute.

Perhaps I should replace “more room between atoms” with the more rigorous phrase “thermodynamically-favoured”, because nothing happens in the real world that is not thermodynamically ordained. What happens is that the free energy of the system is reduced when hydrogen dissolves into the metal to form what is known as a Cottrell atmosphere at the dislocation structures – recall that dislocations introduce local stress into the crystal lattice, so the effect of hydrogen is to reduce this stress.

This concept is illustrated below where we see the presence of hydrogen lowering the formation energy associated with a defect.

Simple energy diagram for the migration of a single point defect within a crystal lattice. The effect of dissolved atomic hydrogen is to lower the free energy associated with the inclusion of the defect.

Note that in this diagram it is assumed that the energy barrier to migration is also reduced. There is a hidden subtlety here. For such to be so, the Cottrell atmosphere would need to comigrate with its dislocation, and not only that, be able to do so without thermodynamic penalty. For larger solutes such as that of the carbon dissolved in carbon steel, we know that this most definitely is not the case, but perhaps for the small hydrogen atom things are different. In any event, the basis of the HELP theory is that there is an increase in plasticity that arises courtesy of the hydrogen Cottrell atmosphere surrounding dislocations.

So we have the H, the E and the P of HELP. The next part of the story is the L for Localised. The impact of HELP is localised to those regions of the metal immediately adjacent to where dislocations have become physically locked. Perhaps it is not immediately obvious, but it follows that, as the energy barrier to the migration of dislocations is reduced, the stresses immediately adjacent to the locked regions must rise. Putting this another way, we can agree that most of the applied external force has to end up across the margins of the internal microstructures opposing ductile flow. The greater the plasticity, the narrower this margin, and thus the greater the local stress. In this roundabout way, HELP posits the raising of localised stresses through the agency of increased plasticity.

In the diagram below, illustrate how microvoid formation and growth might occur where slip planes intersect an immoveable barrier.

When a set of slip bands are intercepted by a planar obstacle, such as another slip band, then hydrogen induced plasticity can result in high local stresses, sufficient to break metallic bonds and initiate microvoid formation. From

If the above scenario were to be playing out, we would expect to see some sign of it when we examine the microstructure of a putative hydrogen embrittlement fracture. Let’s take some examples from the literature.

Typical Microstructure for H-embrittled 304 SS – #1

The following SEM images are taken from .

On the left, we see the typical ductile fracture surface of 304, marked by myriad dimples and voids. On the right, a typical brittle failure of H-enriched 304, with quasi-cleavage surfaces being a common feature.

The micro-void coalescence of the ductile failure is understandable, but, given what we have learned about hydrogen embrittlement, how do we explain the uneven, quasi-cleavage fracture surface associated with it?

Looking at a polished section of the HE failure under the optical microscope reveals that cracks have developed in the metal below the fracture surface. The fracture surface itself is uneven, but with some partially straight edges, presumably due to the presence of quasi-cleavage structures. Notice also the cracks/inclusions aligned with the long axis of the bolt. The authors don’t comment on these, but it is reasonable to assume the presence of some δ-ferrite even for a well annealed specimen. Intriguingly, we can see some evidence of anisotropy in the fracture profile with a tendency to step fractures at a number of points.

Optical microscopy of the hydrogen-charged specimen of 304 SS reveals a number of major cracks underlying the fracture surface. The size and disposition of these are taken to correlate with the flattish, quasi-cleavage structures on the fracture proper. From .

The authors investigated the material disposition associated with the major cracks by EBSD (electron back-scatter diffraction), and produced the nice images below.

Electron back-scatter diffraction (EBSD) imagery of a major crack in a hydrogen-charged specimen of 304SS. (a) basic image illustrating the crack orientation is independent of slip band orientation. (b) phase map showing red ∝’-martensite is localised within slip bands that have been created by the deformation of the blue γ-austenite. (c) inverse pole mapping reveals crystallographic orientation of the bulk austenite. It is strikingly obvious that the crack must be following a twinning plane and is terminating at its margins. From .

The basic EBSD image (a) shows a crack gradually terminating at each end, and cutting across slip bands without substantial regard to their orientation. Does the crack actually cut through slip bands? I think not. It is more likely that it has formed along a surface that is acting as a physical barrier to the migration of dislocations along the array of slip planes.

The phase map (b) shows α’-martensite (red) to be associated with the slip bands as we would expect. This formation of strain-induced martensite from austenite need not concern us here, but nevertheless, it is an important subject which I have discussed in some detail here. However, it does show that slip planes within the γ-austenite (blue) phase either side of the putative barrier have sponsored the pileup of dislocations that leads to the α’-martensite formation.

Image (c), which uses inverse pole mapping to reveal crystal orientation, tells us that the austenite either side of the crack differs in orientation, and the authors take this as evidence that cracks form along the boundaries between twins, and it is this process that gives rise to the quasi-cleavage structures.

This sort of micro-structural evidence is in keeping with the HELP hypothesis.

Typical Microstructure for H-embrittled 304 SS – #2

In this example, have presented data from an experiment where they cathodically charged a 304 test-piece for a duration such that only the outer layer region was penetrated by hydrogen, and, at which point, the sample was pulled to destruction in a tensile tester. It should be noted that the specimen previously had been fully annealed, and then pre-strained by 25%.

The point reached by the diffusing hydrogen can be clearly seen by the change in the nature of the fracture.

SEM image of a 304 SS test piece that was cathodically charged with hydrogen for a period of time sufficient for atomic hydrogen to diffuse just several hundred microns into the metal. The test piece was then pulled to fracture in a tensile tester. The impact of hydrogen on the nature of the fracture surface is striking. From .

Looking at the SEM images below, we see, on the left, a view taken from the centre of the specimen. It exhibits the typical dimpled and micro-voided surface of a ductile fracture, while, on the right, an image from the edge, where we see the quasi-cleavage structures of hydrogen-induced brittle fracture.

5. What Did We Find?

Firstly, a quick visual reminder of the six bolts we have been studying. More details were given in Part 3.

Typical Microstructure of Strained 304 SS

Before scrutinizing the microstructure of the above bolts, let’s familiarise ourselves with the typical microstructural features of cold-worked 304 SS.

In the case of the austenitic phase of 304, there are two main types of plastic deformation that might occur, a) slipping with the production of visible slip bands, and b) twinning which shows somewhat differently.

In the image below, taken from the centre of Bolt #5, well away from the fracture, I have highlighted the intergranular boundaries in red. Note their irregular outline. This tends to be the case for a cold-worked product such as this bolt.

Bolt #5. Within the rounded/irregular-shaped grains outlined in red, we see two types of strain-induced dislocation structures, slip-bands and twins. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 12V for 1s.

The dark, irregular, elongate features are δ-ferrite stringers. These are artefacts of upstream processing, and need not concern us here.

Of greater interest are the features with perfectly straight lines mirroring the underlying ordering of the crystal lattice. The bolder lines are the edges of twinning planes which lie on the bounding surface separating two distinct orientations of the crystal lattice. They are the result of spontaneous rearrangements of sections of the lattice when it is strained during cold-working. Although less distinct, notice also the slip bands that lie in parallel sets. These arise from the migration of dislocations along the preferred slip planes of the crystal lattice as it undergoes plastic flow. Sometimes you will see two sets of slip bands that intersect at approximately 70º. Notice that, as would be expected, both twinning boundaries and slip bands don’t cross intergranular boundaries.

The Fracture Is Uneven and Transgranular

Once we turn up the magnification, it is clear we are looking at a mechanism that is entirely transgranular in nature. The image below is from Bolt #4. Note the fracture edge cuts across individual grains in a chaotic manner. There is no indication that the boundary between grains is sponsoring this fracture. Nor is there any indication of cleavage along preferred planes within the crystal lattice.

Notice also that the fracture seems to have occurred without any displacement of the material of each grain. At the local, level it’s as if the material did not put up much of a fight – there are no signs of deformation. This fits with observations from the climber’s worldview where we know it is trivially easy to snap such a bolt with a hammer blow.

Bolt #4. Longitudinal section intersecting the plane of the fracture. The fracture edge appears not to be guided by either intergranular boundaries or crystal lattice orientation. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 1.6V for 50s.

Let’s take a closer look at that fracture surface. The image below is a progressive-focus composite that allows us to visualise the fracture surface by looking down through the clear, epoxy resin encapsulant.

Bolt #4. Longitudinal section intersecting the plane of the fracture. This is a progressive focus composite that allows us to image the fracture below the surface of the epoxy resin embedment. The fracture appears to be uneven in all directions. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 1.6V for 50s.

For the avoidance of doubt, I have marked in the actual fracture edge in red in the diagram below. This is the line where the fracture cuts across the polished surface of the specimen. The features to the right of this line are those of the fracture surface as viewed down through the encapsulant.

Bolt #4 as above, but with the fracture’s intersection with the polished surface marked in red.

Although there are some relatively straight edges to the fracture, there is nothing to support the idea that there are preferred cleavage planes. The fracture seems to be uneven in all directions. If there is any structure to be observed, then it is of the order of 2um to 3um in size.

Before looking at more pictures, it is worth pausing to reassess the interpretation of such images.

Firstly: The ground and polished surface has been differentially etched by choosing a set of conditions that contrasts the material of the grain boundaries with that of the bulk material. It also differentiates inclusions such as δ-ferrite stringers and a putative sulphide. To a lesser extent it differentially etches dislocation structures within the crystal lattice, such as strain-induced twins and slip-bands.

Secondly: Whilst the surface on view is etched, the fracture boundary on the right most definitely is not. The irregular edge of the fracture meets the epoxy resin without any sign of a gap. That is to say, the wetting of the fracture surface by the resin during the embedment process is pretty much perfect (it is done under vacuum). Notice that the resin is transparent and it is possible to see some distance down into the fracture surface until limited by the focal range of the microscope. This surface genuinely is the original fracture surface, and given the close contact of the resin, it is difficult to see how it could have been altered during the sample preparation.

Thirdly: Perhaps stating the obvious, but for the avoidance of doubt, let’s state it anyway: the fracture surface we are viewing side-on is the result of stresses applied by a hammer. This is how the metal pulled apart. Whilst SRB are involved in changing the chemical environment that immerses the bolt, in no way did these organisms directly contribute to the observed unevenness of the fracture edge by dissolution of metal.

Transverse and Longitudinal Fractures Don’t Differ

If we look at the surface of the longitudinal steps that are a common feature of this brittle fracture phenomenon, then the story remains unchanged. Clearly there is some anisotropy in play for the transverse fracture to occasionally propagate as a perfectly longitudinal one. We see these step-structures over the size range of 0.1mm to 2mm, but at smaller scales there is nothing in the microstructure to distinguish a longitudinal from a transverse fracture.

The image below is a polished, but unetched longitudinal section of the fracture presented by Bolt #2.

Bolt #2. Longitudinal section intersecting the plane of the fracture. Several longitudinal steps can be seen in what is otherwise a predominantly transverse fracture. At this scale, there is a clear anisotropy at play, but what? Is the matching alignment of the δ-ferrite stringers a cause, or merely another view into such anisotropy? Optical microscopy following polishing but no etch.

Several longitudinal steps can be seen in the overall transverse fracture. The inclusions that can be seen aligned on the longitudinal axis are δ-ferrite. We will discuss the possible relevance of this phase at a latter point. An intriguing observation is that it is only for those bolts taken from an SRB environment that the δ-ferrite phase is visible without prior etching, i.e. we are observing some sort of δ-ferrite alteration product.

The image below is from a longitudinal step. This can be deduced by the orientation of the δ-ferrite. The fracture edge is clearly transgranular and uneven on the same scale as that for the transverse ones we looked at above.

Bolt #2. Longitudinal section intersecting the plane of the fracture. If there is a tendency for fracture to occur along this axis, as appears to be the case at the macro-scale, then there is zero evidence for it at the micro-scale. This uneven, transgranular fracture would be indistinguishable from similar transverse fractures were it not for the orientation of the δ-ferrite stringers. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 12V for 1s.

The anisotropy associated with these fractures is something we have commented upon in earlier posts. It is a common enough feature to be considered a distinguishing characteristic of SRB induced embrittlement, yet micro-structural examination doesn’t provide a clear explanation for the phenomenon. Maybe the longitudinally oriented δ-ferrite stringers are introducing anisotropy by pinning dislocations and thus impacting the preferred direction of ductile flow? I hesitate to jump to such a conclusion without a better understanding of how they sit within the austenite matrix. EBSD imagery would help us here.

Taking a closer look at the fracture mechanism

In preparing the images that follow, I used a more aggressive etching process to highlight dislocation structures. My reason for doing so lies in the fact that the literature abounds with conjecture implicating these structures in the phenomenon of hydrogen embrittlement. It would be interesting to see if we can find any visual correlation between such strain-induced dislocation structures and the transgranular fracture we are observing.

In the case of Bolt #3 shown below, twins and slip bands are plentiful, but, we also see sections of otherwise solid metal showing a network of fine cracks. I have marked one such area with a red arrow. There does seem to be some ordering within this mass of tiny cracks, and when I compare this area of micro-cracking, alongside the other examples, it does seem there is some sort of alignment with the direction of the slip bands. However, it is hardly the most striking of correlations.

Bolt #3. Detailed view of the fracture edge. A network of fine cracks can be seen at several points adjacent to the fracture surface. One such is marked with a red arrow. Sets of fine slip bands, often representing two distinct axes for any given member of a set of mechanical twins can be seen in abundance. The slip band spacing is commensurate with that of the micro-cracks, but the relative orientations match less well. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 12V for 1s.

Looking at another region of the Bolt #3 fracture in the image below, we see once again sections traversed by myriad micro-cracks (A). This is especially the case for material in immediate proximity to the fracture itself.

Bolt #3. Detailed view of the fracture edge. Note micro-cracking (A) associated with the edge. Note the section (B) where we see an array of putative micro-voids formed on the intersection of two families of glide planes. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 12V for 1s.

However, even more interestingly, there is a section exhibiting what could be interpreted as an array of micro-voids (B) forming on the intersection between two families of glide planes. The periodicity of this structure is similar to that of the network of microcracks, and it is tempting to speculate that the microcracks originate via a HELP mechanism.

The section from Bolt #2 below, tells the same story. That is, micro-cracked regions are to found on the margin of the fracture (A). However, being a progressive-focus composite, it is possible to look down the fracture surface below the encapsulant, where we see identical micro-crack patterns on the very fracture surface itself (B).

Bolt #2. This is a progressive-focus composite image. Not only do we see the same network of fine cracks adjacent to the fracture edge (A), we can also see similar patterns on the fracture surface as we look down through the epoxy resin embedment (B). Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 12V for 1s.

Some General Conclusions

I have mentioned at several points that the fracture is uneven, and appears not to be governed by underlying crystallographic structure. However, just because we are not seeing quasi-cleavage structures doesn’t mean that the mechanism isn’t one driven by the HELP mechanism.

If we etch the surface with the intent of revealing slip-bands, then it is often the case that we see that every grain is comprised of closely packed arrays of such bands. Their spacing, one to another, tends to be uniform and of the order of 1um. The image below is taken from Bolt #2 some 2mm back from the centre of the fracture.

Bolt #2. This section is 2mm along the length of the bolt from the fracture. The substantial etching applied to this section highlights the extensive slip-bands which are stacked with a 1um to 2um spacing through all grains. This is a sure testament to the extensive cold working applied during production of this bolt. Oil immersion optical microscopy following electrolytic etching in 50% nitric acid at 12V for 1s.

I believe this micro-structure is the result of cold-working of the bolt during production, not a consequence of the facture itself. We would anticipate such cold-working would sponsor the formation of substantial amounts of strain-induced martensite (see this post for reasoning). In fact we measure about 20% martensite at this point (see part #3 for method).

I have argued elsewhere that metastability could well be a major determinant of corrosion resistance of stainless steel in an SRB environment. In particular, my argument was based on the elevation of the diffusion rate of atomic hydrogen that occurs when the production process results in martensite levels in excess of 10% in the finished product. However, I now see here another possibility for elevated vulnerability to hydrogen attack.

Metastable stainless steels, such as 304, exhibit low stacking fault energies (SFE) and for this reason tend to convert austenite to martensite when cold-worked. The mechanism of stacking fault pile-up that leads to this conversion is also one that will give rise to a 3D array of glide locking-points at the intersections of the families of glide planes. Looking at the image above, it can be supposed that this material will present a 3D array of points at which local stress concentration could initiate micro-void formation under the HELP mechanism.

Thus when stress is applied in the form of a hammer blow, we would anticipate micro-void propagation from the most highly stressed locking-point toward its neighbour, such that an uneven fracture surface forms broadly transverse to the bolt’s axis. On this basis, I would contend that the fracture surface with its 2um to 3um featured unevenness is consistent with our current understanding of the HELP mechanism.

6. Can Hydrogen Levels Persist Long Term?

Although there exist multiple threads of evidence that lead us to the conclusion that we are observing hydrogen embrittlement, one question remains unanswered.

We need to rationalise the following –

It happens not uncommonly that, after many years in service, and on some random day, climbers abseil down a route and take out every bolt with a few hammer blows. For this to happen we are forced to conclude that hydrogen levels must have been sufficiently high for the bolts to present in a weakened state. Yet, it makes little intuitive sense that every bolt on a crag could be dangerously weakened at the same point in time. If hydrogen can diffuse into the metal, surely it can also diffuse out again? Why should we see such persistence?

We have good reason to believe that the activity of SRB varies from year to year simply because the supply of sulphate is dependent upon the weather. And thus we need ask, what happens to hydrogen levels under such a scenario? Surely there are times when they are high and times when they are low?

Maybe, as an alternative, we could argue that, despite hydrogen levels varying, a bolt might gradually accrue direct damage from phenomena such as sulphide stress cracking (SSC). And, yes, there is no shortage of evidence of damage via SSC. However, there is also plenty of evidence of bolts displaying brittle behaviour in the absence of stress cracking. In writing this series of reports, I have deliberately pushed consideration of SSC to one side to allow focus on HE proper. Why are we seeing brittle fracture where SSC is not directly involved?

The answer must lie with the fact that atomic hydrogen can be ‘trapped’ within the metal. For example, the formation of a hydrogen Cottrell atmosphere at a dislocation relies on such entrapment. Let’s take a closer look.

If hydrogen can diffuse into the metal, surely it can also diffuse out again?

Yes, of course it can. The diffusion coefficient of hydrogen through the austenite matrix is the same regardless of diffusion direction. However, this doesn’t mean hydrogen will depart the bolt as rapidly as it is acquired. If traps exist, then the activation energy of those traps is the major determinant for the persistence of atomic hydrogen. It is possible for the activation energy to be sufficiently high such that hydrogen is, from a practical point of view, permanently bound.

The basic thermodynamic relationship is simple enough, and I have cut and pasted a section from for those interested in a more formal description.

As one might expect, the rate of loss of hydrogen from a trap is temperature dependent, and thus, thermal desorption analysis (TDA) can be used to estimate the magnitude of the activation energy. The TDA spectrum below is from .

Thermal Desorption Analysis: 25ºC to 600ºC at 100ºC/h. Sample thickness 0.5mm. Sample cathodically charged as indicated. From .

The authors employed a 0.5mm thick sample of 316. They pre-strained it at -196ºC to induce a substantial degree of α’martensite conversion, and cathodically charged it to varying degrees. The spectrum reveals two types of trapped hydrogen, one broadly centred on 200ºC and the other on 300ºC.

How do we know that the hydrogen released by thermal desorption has anything to do with hydrogen embrittlement? The authors provide that evidence. The tensile test data below is for both cathodically charged and uncharged samples. Both samples are pre-strained at -196ºC before being pulled to failure at room temperature.

Tensile test of samples of 316 SS. The labels refer to cathodic hydrogen charging method and match the TDA spectra presented above. The material is loaded with hydrogen and pre-strained at -196ºC before being pulled to fracture at 25ºC. The hydrogen charge has the expected effect of reducing EL from 0.5 down to 0.3. From .

It can be seen that the effect of high hydrogen loadings (“modified”) reduces the elongation at break (EL) from 0.5 down to 0.3.

When the samples were subjected to a thermal desorption step, comprised of heating at 200ºC for 24h, prior to the final tensile test, the embrittlement effect of hydrogen was entirely negated with EL being about 0.5.

Tensile test of samples of 316 SS. Identical to diagram above, but with a dehydrogenation step inserted before the final pull test. The affect of the dehydrogenation step was to restore EL to its normal value. From .

We can assume that the above behaviour of 316 pre-strained at low temperature will be representative of the 304 bolts on the cliffs at Cabo da Roca. The materials science behind such an assumption I have covered in great detail in a previous post. However, for those that have doubts, here is the TDA spectrum for a 304 specimen taken under similar conditions .

Thermal Desorption Analysis: Conditions similar to the TDA spectrum above, but using a 304SS sample. From .

So where does this leave us? I believe that we now have an additional reason to distrust metastable stainless steel installed on a sulphate, sea cliff. In particular, low nickel steels such as 304 that are highly likely to have acquired significant levels of martensite during cold-forming are at risk. Not only does the martensite content drastically accelerate the diffusion of hydrogen into the metal of the bolt, it also would seem to sponsor the strong trapping of such hydrogen.

If the trapping were sufficiently strong, then it seems feasible that harmful levels of atomic hydrogen could be acquired over time despite SRB activity coming and going with seasonal changes. How likely is such a scenario? I don’t think I can get close to quantifying this, but reality keeps reminding me that the phenomenon is one of steadily acquired vulnerability to brittle failure, and unless someone can come up with another theory, it sits there as number one.

Coming Next:

In the previous five parts of the Cabo da Roca analysis I have chased a trail of evidence from environmental sulphate levels, to sulphate reducing bacteria and finally to the sponsoring of hydrogen embrittlement. That leaves a lot of stuff still unsaid. For example, where does the sulphur come from? I will cover that next.

And, of course, there is that persistent question of 316 versus 304. I’m working through samples and hope to have firm guidance to offer on that.

Yoshioka, Y., Yokoyama, K. & Sakai, J. Role of Dynamic Interactions between Hydrogen and Strain-induced Martensite Transformation in Hydrogen Embrittlement of Type 304 Stainless Steel. ISIJ International 55, 1772–1780 (2015).
Lee, K. Y. & Lee, J.-Y. A new method of the calculation of the hydrogen trap activation energy in metals. J Mater Sci Lett 2, 538–540 (1983).
Nicho, K. & Yokoyama, K. Marked Degradation of Tensile Properties Induced by Plastic Deformation after Interactions between Strain-Induced Martensite Transformation and Hydrogen for Type 316L Stainless Steel. Metals 10, 928 (2020).
Yang, N. Y. C., Marchi, C. S., Headley, T. J. & Michael, J. Hydrogen-Assisted Fracture of Low Nickel Content 304 and 316l Austenitic Stainless Steels. 9.
Wang, Y., Wang, X., Gong, J., Shen, L. & Dong, W. Hydrogen embrittlement of catholically hydrogen-precharged 304L austenitic stainless steel: Effect of plastic pre-strain. International Journal of Hydrogen Energy 39, 13909–13918 (2014).
Barrera, O. et al. Understanding and mitigating hydrogen embrittlement of steels: a review of experimental, modelling and design progress from atomistic to continuum. J Mater Sci 53, 6251–6290 (2018).
Jackson, H., San Marchi, C., Balch, D., Somerday, B. & Michael, J. Effects of Low Temperature on Hydrogen-Assisted Crack Growth in Forged 304L Austenitic Stainless Steel. Metall and Mat Trans A 47, 4334–4350 (2016).

Leave a Reply

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