In Depth Investigation

Austenite stability – are we missing something here?

The title should be fair warning. This one is going to get geeky.

If an anchor component is at all attracted to a strong magnet, don’t install it at a crag with a reputation for eating hardware.

If you asked your supplier for 316, yet find the product is attracted to a strong magnet, don’t install it. Send it back to the manufacturer and ask them to try harder.

Reasons follow ….


Some time back I promised that I would soon be publishing a post on the brittle fracture phenomenon at Carbo da Roca. However, I wasn’t expecting to launch into a months long process of working through the published literature. What started out as a quick scan ended up a deep dive. Now, some hundred or so papers the wiser, I am struggling to explain what I’ve learnt in terms that will be meaningful to those who just want to go bolt some rock. Regardless of the difficulty, this is something which once understood even in a very basic way, should prove to be of great value in guiding our selection of materials for rock anchors.

What are we missing?:

I believe we have been overlooking a phenomenon of considerable relevance to our understanding of the distinction between good and bad choices for the bolts to be employed on sea cliffs with a history of severe corrosion. There have been hints of the form that 316 is always good, and if it is bad, then testing will show it is not 316. And yes, 304 is always bad except for the fact that sometimes it is not. And yes, sometimes is is not even 304 but a cheaper steel like 301. I’ve always felt there was a story here, but it was getting lost in the community desire to demonise one type of material, one type of bolt or one manufacturer.

I posted a quick note some time ago when it became obvious to me that seemingly identical machine bolts, marked A2, were not in fact identical given that some were magnetic and some were not.

Despite what it says on the label, and despite identical appearances, not all 304 bolts are the same. We need to dig deeper to understand why not, and the possible implications of such differences.

Austenite and Martensite:

The story begins with two phases, or allotropes, common to iron alloys. One, austenite, is based on face-centred cubic packing (fcc) of the constituent atoms, whilst the other, known as martensite, is based on body-centred cubic packing (bcc). Strictly speaking, we should use the term ∝´-martensite, because there is another form, \epsilon-martensite, which involves a hexagonal closest packing (hcp) lattice. However, we will deal with this form but peripherally.

austenite\gamma Fefcc
martensite∝´ Febcc
Fig.1 The two phases of stainless steel that concern us are martensite, which is packed as a body-centred lattice (left), and austenite, which is packed as a face-centred lattice.

Note that in the above diagram, the blue and white spheres do not represent different types of atoms. They are used simply to make the packing arrangement clearer. However, it should also be understood that, when we are dealing with a 300 series stainless steel alloy, whilst the major proportion of atoms are iron, as many as one in five will be substituted by chromium, and as many as one in ten by nickel.

Austenite is not for dirt-bags:

Of the two phases, the austenitic phase, is the one we want, with its moderate strength, good ductility, and high resistance to general corrosion and hydrogen damage. However, at typical ambient temperatures, austenite is not stable unless substantial amounts of nickel are present in the alloy. Given that nickel is an expensive commodity, and given the fact that the majority of route-equippers run on a dirt-poor budget, we would be naive to expect to find high stability austenitic stainless steel on even the most corrosive crags, given the fact that identical-looking, low-cost substitutes will be on offer. Market forces will guarantee that low-cost alloys, or “bargain” alloys of suspect composition, are the ones that will make it onto the cliff. What could possibly go wrong?

And this is exactly what we do find. By far the commonest alloy to be used is 304. To be classified at 304, the nickel content has to lie between 8% and 10%., and you can be very sure that because of the cost element, you won’t find many samples of 304 where the nickel percentage is much higher than the minimum 8%. At this percentage it is classed as a metastable austenitic steel. What this term means is that, whilst it might start life as pure austenite, there are no good thermodynamic reasons why it should continue to exist as austenite, and plenty of reasons why it might, with a wriggle and shake of its atomic lattice, re-arrange to the martensitic form.

Fig. 2 Some common austenitic stainless steels arranged in order of nickel content. I grabbed spot prices for these alloys from the international scrap market as an indicator of their base cost. It’s not hard to see why 304 has been the choice of frugal route-equippers. Increasing the nickel content beyond 10% brings pain, but, as 316 and 904 illustrate, nothing like the pain of adding substantial amounts of molybdenum.

Does any of this matter? For many sea cliffs probably not. But for those with high sulphate loads that feed sulphate reducing bacteria (SRB), I believe it is a very big deal.

A painless look at metastability:

The failure of climbing anchors at Tonsai was never going to be the trigger for a huge investment in fundamental materials research, but, luckily for us, this whole issue of metastability has, in recent years, been the subject of substantial research. Firstly, there is funding driven by the desire to move to hydrogen as a transport fuel. The guys doing this work don’t like metastability any more than we climbers do. Secondly, there is funding driven by the need for high-strength, light-weight alloys at low cost. The guys doing this research, actually love the metastability and are exploiting it to develop special steels known as TRIP alloys. Either way, there is some very cool research happening that is highly relevant to how we specify stainless steels for corrosive sea cliffs.

Let’s kick-off with this informative diagram taken from a review of TRIP steels by

Fig. 3 Effect of temperature on the relative stability of austenite and martensite. For temperatures lower than T0 austenitic steels become metastable.

There is no need sidetrack into the concept of free energy to understand what this diagram is saying; just hold onto the idea that that the state with the lower free energy is the more stable one. Thus, at high temperatures we see that austenite is more stable than martensite, and at lower temperatures martensite is more stable than austenite. It follows on from this observation that there must be a temperature at which the phases are equally stable. This temperature is marked T0 in the diagram.

For a given alloy, the further it is cooled below the point T0, the more it moves into metastability with an ever increasing free energy difference developing between the two states. Eventually a temperature, MS, is reached where transformation from austenite to martensite takes place with an amount of energy, \Delta G_{M_S} ^{\gamma\rightarrow\alpha' } being released.

For very many alloys, the martensite transition temperature, MS , is very cold, maybe -60°C or even colder, and certainly at a temperature way beyond the concerns of us rock climbers. So, who cares if the alloy is metastable, if the temperature for transition is never reached? Well, as it turns out, we do, a lot.

The reason is illustrated in the diagram above. If during manufacture, the part is formed in a way that results in plastic deformation (as will always the case) then it is possible that the energy supplied to achieve deformation, U’, when added to the free energy differential, \Delta G_{T_1} ^{\gamma\rightarrow\alpha' }, relevant to the temperature of the manufacturing process, T1 , will exceed the value \Delta G_{M_S} ^{\gamma\rightarrow\alpha' } required to trigger transformation. This equivalence being represented by the green dashed line.

We can summarise the above by stating that the downside of using a metastable grade of stainless steel is that it is pretty much impossible to cold-form it into a useful shape without triggering at least partial transformation of austenite to martensite.

The folks researching this phenomenon have come up with a good way to get a handle on where a particular alloy sits with respect to the martensite transformation temperature. They use something called the M30/50 transformation temperature. It is defined as the temperature at which 30% strain causes a 50% transformation of austenite to martensite. This parameter has been estimated across a range of alloy compositions, and a number of empirical recipes have been published that relate M30/50 to the composition of the alloy itself.

use the following equation –

M_{30/50} = 551 - 462(C+N)-9.2Si-8.1Mn-\\29(Ni+Cu)-13.7Cr-18.5Mo

where the C, N etc represent the weight percentage composition of the alloy.

Note the impact of nickel on the transformation temperature in the above equation. However, also note that carbon is a very significant austenite stabiliser. Because of the need to weld some components such as ring bolts, the “L” grade of low carbon alloy is nearly always specified. Reducing the carbon composition to say 0.02% guards against chromium carbide precipitation and potential inter-granular corrosion when the part is put in service, but in doing so it reduces the stability margin of the austenite. The difference in the transformation temperature between the normal grade, at say 0.08% max. carbon, and the ‘L’ grade at 0.03% max. carbon is roughly 23ºC . And, because carbon content is difficult to measure at the point of entry of bar stock into the factory, then everything comes down to foundry certification. In reality, we need to be aware that carbon content is a wild card in the production process.

In the diagram below I have applied the above recipe to estimate the transformation temperatures for some of the lower cost steels. The horizontal bars represent the most likely M30/50 for a given alloy on the assumption the manufacturer is going to aim for the lower limit of the nickel specification. The size of the box indicates what could happen should nickel suddenly lose is commercial value to an extent that we start seeing material at the upper nickel limit appearing on the market.

Fig. 4 There is a linear relationship between the nickel content of an austenitic steel and its martensitic transformation temperature. For 304 this temperature is close to ambient, which means we can anticipate that cold-formed 304 will show considerable variation in the proportion of strain-induced-martensite.

This plot was a light-bulb moment for me. It suddenly became obvious that, should martensite conversion be a problem for the product in the field, then there could be no worse choice of alloy than 304 with its minimum nickel content of 8%. At this level, the M30/50 temperature of 26ºC is right in the centre of the ambient range, and thus bound to render the production process very sensitive to small changes in the composition of the feed stock. We are setting up for a situation where, even with good quality control, the product entering the field is likely to contain unpredictable levels of strain-induced-martensite.

Martensite is magnetic, whilst austenite is not, and thus it is possible to gauge the amount of conversion using a magnet. Looking at the diagram above, it is now clear that we shouldn’t encounter magnetic 316 anchor components. If putative 316 is magnetic, then, either it has been cold-formed in Santa’s Arctic workshop or, perhaps more likely, the nickel content is a few percent below specification. On the other hand, it is very likely that all cold-formed 304 will be magnetic to some degree unless it has been annealed as a final step in production, or maybe the nickel content is a percent or so above the minimum, or perhaps the carbon content is on the high side. This assertion aligns with my short note in which I posted this video.

Although heavily worked, the 316 Raumer and CT ring bolts show no magnetism. On the other hand, the Fixe parts are strongly magnetic at points of maximum deformation.

The effect of just a few percent extra nickel, the difference between 304 and 316 at their lower nickel limits, is profound. Does this phenomenon underlie the observation that, despite repeated calls by UIAA Safe Com and by myself, an example of failed 316 hardware is yet to materialise?

What’s so bad about strain-induced martensite?

The short answer is that there is nothing is wrong. Metastable steels are the very thing if you wish to employ a strain-hardening strategy to achieve your product goal.

If, however, your goal lies not so much in ultimate strength, as in maintaining adequate strength in an aggressive, corrosive environment, then you might not want to lose austenite in favour of martensite. This is especially so in a sulphide environment where cathodic charging of hydrogen becomes a real possibility.

For decades austenitic stainless steel has been valued for its resistance to hydrogen embrittlement (HE). And almost for the same period of time it has been recognized that this resistance is very good when its good, and very bad when its bad; an indicator of threshold mechanism of sorts. And for decades an accusative finger has been pointed at strain induced martensite as the culprit. Yet the hard evidence for such a deleterious impact was never there.

From we have the following –

The role of martensite on hydrogen embrittlement in austenitic stainless steels has not been firmly established. Although generally viewed to be neither necessary nor sufficient to explain susceptibility to hydrogen embrittlement in austenitic stainless steels, ∝´-martensite, is associated with lower resistance to hydrogen embrittlement.

In recent years, the evidence has been accumulating that martensite transformation and hydrogen embrittlement arise not from a dependency of one upon the other, but from a common cause associated with something called the stacking fault energy, eg. see We will circle back round to look at these findings later, but let’s first move back out to the macro world view. What are the practical implications?

A nice experiment showing the interaction between martensite transformation, hydrogen embrittlement and temperature was published by .

Fig. 5 The temperature at which the pre-strain is applied has a marked effect on the level of hardening achieved, and the subsequent susceptibility to hydrogen embrittlement. .

In this experiment, the test pieces were machined from 10mm, 304L plate containing 8.1% nickel. Based on the full elemental composition, the authors calculate its M30/50 temperature to be 21ºC. The test pieces were fully annealed to convert them to austenite, before stretching by 30% of their initial length to impart a constant pre-strain. This process of pre-straining was carried out at three different temperatures. The test pieces were then cathodically charged to introduce atomic hydrogen, before being pull-tested to measure their mechanical properties. The cathodic charging process the authors used is not dissimilar to the mechanism by which our crag nemesis, sulphate reducing bacteria, or SRB, are thought to mediate hydrogen embrittlement.

Firstly, let’s consider just the black curves. Prior cold-working of the test piece almost tripled the value of the yield point, and added perhaps 30% to the ultimate tensile strength. There is a marked reduction in ductility, with elongation at break (EL) being reduced to approx. 30% of the un-worked figure.

What I find surprising, is how sensitive the work-hardening process is to temperature. Yes, I know that our prior discussion of Fig. 4 indicated that this is exactly what we should expect, but, I’m sure I’m not alone when I think of hot-working steel, and conjure up a mental image of a blacksmith’s shop with glowing bars of steel being struck on the anvil amidst a shower of sparks What’s happening here couldn’t be more different.

Looking at the data presented one is forced to conclude that there could be measurable differences in product quality when working 304 steel bar on a cold compared with a hot day.

Now let’s consider the effect of atomic hydrogen, as depicted by the red plots in the above figure. Notice that the predominant effect of hydrogen is to reduce the ductility of the test piece, i.e. reduce the elongation to break (EL) figure, rather than impact either the yield point or the progression from yield to break. Notice also that something approaching true embrittlement occurs only for the test piece cold-worked at 20ºC. Thus we can be certain that if the cold-working temperature was reduced a further 10ºC or that the nickel level was 1% below specification, we would be making a product that was very susceptible to hydrogen embrittlement.

Wang et al used XRD to quantify the volume percentage of martensite versus austenite. Based on the author’s calculated value for M30/50 , we would anticipate the level of martensite transformation to be more like 50%, than the 40% they actually measured, but even that level of agreement seems pretty good to me, and lends support to the utility of Fig.4 for bolt material selection.

Fig. 6 Subjecting 304 test pieces to 30% strain at a variety of temperatures reveals the martensite transformation to be strongly temperature dependent. .

Despite the consensus that martensite transformation and susceptibility to hydrogen embrittlement are not, mechanistically linked as cause and effect, the association between the two phenomenon is striking. Wang et al summarise this empirical association in the graph below. The HE index is calculated as the percentage reduction in the elongation to break (EL) caused by the presence of hydrogen.

Fig. 7 Although susceptibility to HE is not considered by most authors to be a direct consequence of the presence of strain-induced martensite, the association is nevertheless striking. .

Zooming in for a closer look:

Much of what we see happening at the macroscopic level can be explained by events we observe at the microscopic level. In particular, the strain-induced conversion of austenite to martensite has, over recent years, given up much of its mystery. Let’s take a close-up look at this phenomenon.

Firstly, we’ll need a bit of crystallography to put what we are seeing into context. Crystallography is a deep and complex subject, but a couple of bullet points should serve our purpose. Wikipedia has a good overview for those that wish to dig deeper.

  • Starting from the annealed state at approx. 1050ºC, water-quenching causes the rapid formation of individual crystals that grow outward against one another to form a solid mosaic. The crystals are all identical in that they are of the austenite phase, and differ only in crystallographic orientation. Minor components and impurities, that have been excluded from the atomic lattice, crystalise out as small granules along the inter-crystal boundaries.
  • Crystals are never perfect, and thus contain irregularities in an otherwise well organised atomic lattice. Defects can consist of a single point, or they may present as a linear structure, called a dislocation, that extends across a particular plane of the lattice. A third class of defect is the stacking fault which is one that applies to an entire plane of the lattice.
  • Metals that are ductile owe that property to the presence of dislocations located throughout the atomic lattice. When stress is applied, there will be preferred planes of the lattice that will glide one over the other by virtue of the migration of dislocations. It seems a bizarre notion, but when you see a bolt deforming under load, the fact it is doing so, and not snapping, is because, hidden from plain sight, dislocations are migrating through the lattice. For now, hold onto the fact that when a metal is being plastically deformed, then dislocations are migrating, and that means they must end up somewhere!
  • Because dislocations distort the lattice from its perfect geometry, they have a tendency to rearrange into lower energy forms whenever possible. A common rearrangement seen during deformation of austenitic stainless steels is the creation of a stacking faults from prior edge dislocations.
  • Stacking faults, like all dislocations, move through the metal thereby endowing plastic flow. However, there are physical limits to such migration. For instance, limits imposed by grain boundaries or by intersection with other slip planes. Thus, when we examine austenite that has undergone plastic deformation, we observe “pile-ups” of stacking faults that cause a feature called a slip-band.

Let’s go back to Wang et al , to take a microscopic view of their strain-hardening data which we presented in Figs. 5, 6 & 7.

In Fig. 8, below, we see a typical presentation of a pure austenite phase. Without paying close attention to the heat-treatment regime, the observed heterogeneous mix of grain sizes is likely typical. Furthermore, dark inclusions of what are presumably carbides can be seen, not only in the inter-granular zones, but also in bands within individual crystals. Thermal twins are also visible. Twins are a type of stacking fault where the layer sequence for the lattice is the mirror image on one side compared with the other. We can see that the starting material used by Wang et al is not well ordered, and thus must contain a certain level of internal stress. However, it would be optimistic to expect better than this for a raw commodity like 304 stainless steel.

Fig. 8 Annealed 304 material comprised primarily of the austenite phase. Despite the fact it was fully annealed then water-quenched, it is pretty heterogeneous with poor grain uniformity, thermal twinning, and dark inclusions on the inter-granular boundaries, possibly carbide..

The authors pre-strained the above material by 30% at 80ºC and measured its mechanical properties. Recall 80ºC is well above the martensite transformation temperature, and whilst this level of pre-strain resulted in some degree of work-hardening, no martensite was formed. Under the microscope, we observe that, as a result of this deformation, the original granular structure is now difficult to recognize, and very many slip-bands have become visible. Sets of slip-bands can be seen terminating along lines that are either grain boundaries or intersecting stacking faults.

Once stacking faults “pile-up” at a particular point, plasticity in that localised region is exhausted. Thus the presence of a great many slip-bands is an indicator of the depletion of easy options for plastic flow, and, consequently, at the macroscopic level, we see a marked increase in yield strength, as was illustrated in Fig. 5.

Fig. 9 The same material as in Fig 8 above, but this time following the application of 30% strain at 80ºC. Note that we have plenty of slip-bands, but not much in the way of martensite which would show up blue-grey in this image because of the colorant employed. By reference to Fig. 6 we know this material contains but marginally more strain-induced-martensite that the original annealed starting point. .

The above is no more than you would expect for a typical austenitic steel being worked within its stable temperature range, but what happens if we perform the same operation within the metastability range at 20ºC?

Fig. 10 The same material as in Fig 8 above, following the application of 30% strain at 20ºC. The granular structure is partially lost, and large numbers of dark grey slip bands can be seen in parallel arrays crossing individual grains. A colorant has been used to highlight the martensite in blue-grey. It can be seen to be associated with the slip bands. By reference to Fig. 6 we know this material is comprised of 40% strain-induced-martensite..

In Fig. 10, we see that many, but not all, slip-bands are associated with blue-grey regions of martensite. The identification of martensite is dependent upon the colouration produced by Beraha’s reagent which the authors used for optical contrast.

Recall that stacking faults are planar structures, and thus, when we see an array of slip-bands traversing a single crystal of austenite, we need to be aware that we are viewing sheets layered into the crystal edge-on, not needle-like structures. An understanding of how martensite is interwoven within the residual austenite will become important when we come to consider the effect of hydrogen.

Another example of just how thoroughly 304 becomes criss-crossed by slip bands if it is strained at a temperature within its metastable range is provided by . In Fig. 11 we can see the impact of increasing metastability as the temperature of deformation is lowered. The authors recorded high levels of martensite formation associated with this process

Fig 11. Optical photo-micro-graphs of 304 following acid-etching to reveal underlying structure. (a) annealed at 1050ºC, (b) strained 57% at 23ºC and (c) strained 51% at -50ºC.

A further example of how slip-band formation and martensitic conversion totally dominate the low grade stainless steels when they are cold worked is shown below. From .

Fig. 12 An example of 301 stainless steel that has been progressively strained by amounts a) 0, b) 0.025, c) 0.05, d) 0.10, e) 0.15 and f) 0.20

How sure can we be that what we are viewing here is in fact martensite associated with the slip-bands? Well, as it turns out, optical microscopy is not the only imaging tool that can be brought to bear on the matter.

The MFM image below is taken from . The authors imaged 304 subsequent to imparting four levels of increasing strain at 22ºC. This image is not the output of any sort of focusing device, but rather, comes from a nano-scale, mechanical, piezo-electric scanner that skims over the surface of the sample measuring the magnetic force acting on a tiny sensor. Thus, the image is a totally alternative way of viewing putative martensite by virtue of its magnetic properties. It is very cool just how similar these images are to the optical images presented above.

Fig. 13 A magnetic force microscopy (MFM) image of a 304 test piece that has been strained by amounts, (a) 20%, (b) 30%, (c) 40% and (d) 45% at 22ºC. The image is not optical, but is the result of a scan by a very small sensor that records the magnetic force of attraction of the probe to the surface. Thus we see magnetic martensite in strong contrast to non-magnetic austenite. .

Yet another alternative source of imagery comes from the electron microscopy technique of EBSD phase mapping. have published the nice image below of a 304 forged steel sample subsequent to being strained at -50ºC. The martensite resolves as the yellow regions against the blue austenite bulk. This technique very precisely locates the martensite within the slip-bands.

Fig. 14 An electron back-scatter diffraction (EBSD) phase map of a 304 test piece at -50ºC . This electron microscopy technique is used to highlight the location of distinct crystalline phases within a conventional EM image. Here we see martensite rendered in yellow. It is very exactly located within the dark slip bands traversing bulk austenite which is rendered in blue.

A quick look at possible mechanisms:

I am optimistic that without necessarily turning the dial right up to full-geek, it should be possible to give some sort of explanation that ties martensitic transformation to nickel content and temperature. A well cited paper by underpins much of the arm-waving that follows.

When a stacking fault occurs in the body-centred cubic lattice of austenite, a single plane of atoms in hexagonal closest packing forms, bound by two partial dislocations. Such an arrangement can be considered an atomic slice of \epsilon-martensite. The width of this structure is not random, but represents a contest between two opposing sets of forces, one pushing the edges apart, the other pulling them together. The resultant width is an inverse measure of something called the Stacking Fault Energy (SFE). If the SFE is low, the stacking fault tends to expand across the plane, and if it is high, the edges are kept bound closely together.

It has been amply demonstrated that the greater the nickel content of an austenitic stainless steel, the greater the SFE, and consequently the narrower the width of the stacking fault. Likewise it is well demonstrated that the SFE increases with temperature. Simply put, low nickel content and low temperature is a recipe for wide stacking faults that will present an expanse of \epsilon-martensite.

When alternate layers of stacking faults pile-up, one over the other, provided the faults have sufficient width, we observe the spontaneous transformation of \epsilon-martensite to ∝´-martensite, and an expansion of the width of the overall structure. The diagram below comes from the modeling of , and illustrates the critical nature of the stacking fault energy in driving the proliferation of ∝’-martensite during deformation. Thus, when SFE energies are as low as 10mJ/m2, it takes an overlap of but two faults for a layer of ∝´-martensite to propagate across the crystallographic plane. At room temperature, a specimen of 304 at its lower nickel limit has an SFE close to that critical limit, and herein lies the fundamental reason for the instability of 304.

Fig. 15 The width of a set of stacking faults, overlapping on alternate crystallographic planes, is unstable for materials with a low SFE (\gamma), and the fault will propagate across the entire extent of the available plane, spontaneously converting to ∝´-martensite as it does so. .

How do we go forward from here? Obviously, 304 is a poor choice for those crags where sulphate availability guarantees that SRB will thrive. The need for martensitic-free steel becomes paramount, yet, what are the alternatives?

If this atomic level stuff seems like magic, then it is well understood magic, thanks to the generations of rigorous physical chemistry it is built upon. It is possible to make well-informed choices for bolt materials once you understand what the application demands. From Fig. 15, it is obvious we should be choosing a material with an SFE in excess of 20 mJ/m2 , if we are to be well clear of metastability.

And because this stuff is well understood, we can look up a fairly rigorous theoretical relationship between nickel content and SFE. The diagram below is taken from , who have investigated both nickel and nickel/manganese stabilized chromium steels. By eye-balling the theoretical relationship presented (solid red markers) we can see that we would expect to be clear of metastability problems if we chose an alloy with at least 15% nickel.

Fig. 16 Theoretical and experimentally determined dependency of stacking fault energy (SFE) on the nickel content of chromium steels. .

Let’s see what that means by calculating the SFE for a range of low-cost stainless steels and marking the values up on Fig. 15.

Fig. 17 I have calculated the SFE for a range of low-cost stainless steels based on their minimum nickel content and added the values to the stability chart of Fig. 15. It is immediately obvious why 304 is in the unstable region, 316, on the other hand, is on the very edge of stability, 309 is better, but ideally 310 is required to be sure strain-induced martensite formation will be suppressed.

It is immediately obvious, why 304 with a SFE of 14mJ/m2 is going to be a problem. Likewise, we can see that 316 with SFE of 16mJ/m2 may often be serviceable, but is too close to the stability margin to be considered ideal. Type 309, would be an improvement, but would not be as certain as 310.

The 2% molybdenum content of the 316 would further increase its SFE, but I know of no experimental data to guide the quantification of this effect. It maybe that 316 ends up being similar to 309 in terms of SFE.

What’s so bad about martensite? – reprise

So, if the consensus among many authors is that martensite is “neither necessary nor sufficient to explain susceptibility to hydrogen embrittlement in austenitic stainless steels” why are we so obsessed with austenitic stability?

From a practical point of view, what matters is the fact that metastable steels have a long history of poor hydrogen resistance. This is not news. For example, from 37 years ago, in George R Caskey’s, Hydrogen Compatibility Handbook for Stainless Steels” we have –

The composition range from 8 to 14% nickel, where the marked improvement in resistance to hydrogen damage occurs, corresponds to increased austenite stability and decreased formation of martensite during plastic deformation.

This statement is supported by the figure below. The message, “when it is good it is good, but when it is bad it is terrible” could not be clearer.

Fig. 18 A warning from 37 years ago: things go wrong fast in a hydrogen atmosphere, even at room temperature, if the nickel content drops below 10%. .

From 12 years ago we have this data-set for 304 and 316 over a wide range of nickel content . Note that this data was obtained at -50ºC, so the critical nickel composition for martensite transformation is more like 12% to 13%, rather than the 8% to 9% typical of room temperature. (Refer to Fig. 4). Note also that the scaling of the x-axis is different in this diagram compared with the previous one, so the transition does not appear as abrupt.

Fig. 19 Hydrogen embrittlement measured by the relative degree of “necking-down” of the test piece at the point of fracture. Ductile materials will reduce diameter by as much as 80% before fracture. Brittle materials may fracture with no diameter reduction whatsoever. Note that the experiment was conducted at -50°C, so we can expect 316 to be metastable over much of its nickel composition range. .

If the correlation between low nickel content, subsequent martensitic transformation, and hydrogen damage is so strong, what is the evidence that martensite is not the direct causative agent of this damage? Quite simply, it is the number of instances where hydrogen embrittlement can be shown to occur in the absence of martensite transformation.

Consider Fig. 20 below. This data is from and clearly illustrates the reduction in ductility in 310 with increasing hydrogen burden. Note that this experiment was conducted at room temperature, so, given the 19% nickel content of 310, it is unlikely there would have been pre-existing martensite, or, for that matter, any strain-induced martensite transformation under such conditions. Yet we see a substantial reduction in the ductility of this material. However, we should note the fact that the test piece is a foil no more than 63um thick. This will prove to be a critical observation.

Fig. 20 Hydrogen embrittlement in the absence of martensite. A thin foil of the stable, austenitic alloy, 310, clearly loses ductility with increasing hydrogen burden. The fact the test piece is just 63um thick is critical to an understanding of what is happening here. .

Another example, this time with 316. utilised 1.0mm thick test pieces that were fully annealed before being subject to a pre-strain of 20% at room temperature. The test-pieces were cathodically-charged for periods varying from 12 to 48 hours, before being pull-tested. The nickel content of the test material was 14%, and, as would be expected, no transformation to martensite occurred; this fact being confirmed by XRD. Yet, we clearly see from Fig. 20 below, that both a significant increase in the yield point and a reduction in elongation-at-break occurred when longer periods of cathodic charging were employed. Once again we have unambiguous evidence of hydrogen embrittlement of a fully austenitic sample. However, once again, it has to be noted that the thickness of the test specimen was just 1.0mm.

Fig. 21 Hydrogen embrittlement of 20% pre-strained, but fully austenitic, 316. The numbers in the boxes are the hours of hydrogen pre-charging applied before tensile testing. Judged by two measures of embrittlement, increased yield point and decreased elongation-to-break, the deleterious effect of the increasing hydrogen burden is clear. The fact that the test piece is just 1.0mm thick is significant. .

We can take a closer look at what is happening here thanks to the work of . These authors used a 1.5mm diameter wire composed of 316 containing 12% nickel. It was annealed, then hydrogen charged for a period of 24 hours before being subject to a slow strain rate tests. As we have come to expect, the elongation-to-break was substantially reduced by the presence of hydrogen. However, what is more interesting from our point of view is the micro-structural study carried out by these authors.

The surface layers of the specimen were very clearly embrittled as can be seen in the image below, with predominantly inter-granular cracking accompanied by some trans-granular cracking.

Fig. 22 Hydrogen embrittlement at the surface of a 316 stainless steel wire that has been hydrogen charged and then gradually pulled to fracture. We see predominantly intergranular cracking with some trans-granular cracks. This brittle mode of fracture transitioned to predominantly ductile fracture some 100um below the surface. .

However, moving from the outside to the inside, the brittle fracture rapidly gives way to fracture mechanisms that are increasingly ductile in nature. Thus, only the outer 100um of the specimen could be said to to be significantly impacted by hydrogen, although the authors observed some secondary cracking as far as 700um into the specimen. The situation is represented diagrammatically below.

Fig. 23 Diagrammatic representation of the depth to which hydrogen effects are observed in a a purely austenitic stainless steel wire. Thus for a 10mm bolt, the effect of hydrogen on its mechanical properties should be largely insignificant. .

The reason for this marked surface effect undoubtedly lies with the low diffusion rate of atomic hydrogen through austenite.

For purely austenitic steels it seems likely that only mechanically thin components can be significantly impacted by hydrogen embrittlement. This is in distinct contrast to metastable steels where the embrittlement of much thicker components is associated with substantial martensite transformation.

We will take a closer look at the role of diffusion next.

The role of atomic hydrogen diffusion:

It has long been postulated that the resistance of austenitic steels to hydrogen attack is the result of the very low diffusion rate of atomic hydrogen through the atomic lattice. It is orders of magnitude lower than that for the bcc ferritic steels such as martensite. Thus it is reasoned that, whilst hydrogen might embrittle the outer 100um, it can never diffuse inward to a sufficient degree to cause structural damage.

However , should martensitic transformation occur, then there is a real possibility that high diffusion rate channels might open up, with the potential to reach into even quite thick structures. This is the so called “Hydrogen Diffusion Highway Effect”. The illustration below is taken from

Fig. 24 The “Hydrogen Diffusion Highway Effect” whereby interlinking sections of martensite provide rapid diffusion paths for atomic hydrogen deep into the structure of a part.

The diffusion rate should not be confused with solubility which is another well-studied characteristic. The latter figures prominently in hydrogen embrittlement work, but, right now, we are narrowly concerned with Fick’s Law, and the rate of movement of hydrogen from an area of high concentration to one of low concentration. From a practical point of view, we are narrowly interested in answering the question “how long does it take atomic hydrogen to diffuse from the outside to the inside of a 10mm climbing bolt?

A number of research groups have made estimates of the diffusion rate of hydrogen through austenitic steels containing varying amounts of strain-induced martensite. A consistent finding is that all steels, both stable and metastable, share very similar, and very low hydrogen diffusion rates, provided no martensite is present. The degree of work hardening doesn’t matter provided martensitic transition has not been triggered. Thus the apparent diffusion rate is purely a function of the amount of martensite present. Fig. 25 below, taken from , illustrates this point.

Fig. 25 The apparent diffusion constant of an austenitic steel depends upon the fraction of strain-induced martensite it contains. In this case we see a linear relationship between the log of the diffusion constant and the log of the ratio of martensite to austenite. .

I have taken datasets from three research papers, which after correction for the different diffusion temperatures employed, allow us to answer the question, how long does it take for hydrogen to penetrate to a specific depth, given the martensitic content of the material. The three papers I have used for reference are .

The diffusion constants were adjusted to 20ºC using the fact that diffusion processes obey the Arrhenius relationship, while the times required to diffuse 2mm into the material were calculated using the Fick’s second law simplification 4\sqrt{D.t}. A plot of all three data sets is shown below.

Fig. 26 Based on three published datasets, I calculated the rate of diffusion of hydrogen, expressed as the time required to penetrate to 2mm depth, for steels containing various volume fractions of martensite. This plot accords well with real world experience, where materials either seem to last “forever” or are “consumed” within a decade.

I’ve used 2mm as the depth criteria because it is significant with respect to the dimensions of climbing anchor components. For example, a 10mm diameter bolt would be reduced to a 6mm effective diameter, and as a first approximation, its strength reduced to 36% of its design strength. A 4mm fixed hanger would be reduced to half that thickness if hydrogen diffused into the metal from the underneath surface.

For those crags where hydrogen attack is an issue, it is clear that for any component, wherein cold-working has converted more than 30% of austenite to martensite, there is potential for field failure within a decade or so of installation. On the other hand, materials with low levels of martensite transformation, say less than 10%, will exhibit a long service life under the same conditions.

The nature of hydrogen damage:

Up till this point in the discussion, we have been saying that hydrogen is bad because it causes embrittlement. The majority of published work on the subject takes embrittlement as an easily measured metric of hydrogen damage. However, the spectrum of hydrogen damage goes far wider than this simple metric.

I have purposely avoided discussing mechanisms of hydrogen damage because that would be a truly deep dive into a very complex field which is by no means free of contention. And. as it turns out, there is no need to go there provided we have the knowledge to a) recognize when hydrogen is likely to be a problem and b) choose the correct materials for just those cases.

However, a few comments are in order with references for those that want to dive deeper. From a practical point of view we need to make two distinctions. a) There are cases where a material can be subject to a hydrogen environment, and, provided no stress is applied, no damage will ensue. Thus, when the hydrogen source is removed, atomic hydrogen will diffuse back out of the metal, and the material will regain its normal mechanical properties, and b) there will be many cases where the presence of hydrogen results in damage that persists after the hydrogen source is removed.

If we consider the environment of anchors installed in a cliff where sulphate is prevalent, then, as I have alluded to in previous posts, there is a possibility that conditions favourable to SRB will come and go with sulphate availability, and, over a period of years, we can anticipate that hydrogen-charging processes within that micro-environment will likewise cycle. Thus, if we are to consider damage mechanisms, then it is the permanent damage mechanisms we need to be looking at.

Here is a list of mechanisms all of which may be playing a part in deteriorating the mechanical properties of an anchor.

  1. HELP (hydrogen-enhanced localised plasticity): Remember, up-page, we took a look at stacking fault energy and martensite transformation, well, the HELP mechanism operates at this level. It’s complex, contradictory and contentious, but nevertheless might well be pointing in the direction of one of the most fundamental mechanisms. is a good starting point.
  2. HID (hydrogen-induced decohesion). This mechanism seems to be falling out of favour. It attempts to explain a role for hydrogen in reducing the energy necessary to separate the atomic lattice at the tip of an advancing crack. Once again, is a good starting point.
  3. SSC (sulphide stress corrosion): Most authors attribute the propagation of stress cracks in sulphide environments to the action of atomic hydrogen. It is possible that mechanism like 1. and 2. above are involved.
  4. Hydrogen blistering: This mechanism can give rise to ‘blisters’ that scale from fractions of a millimetre to several metres across. It is thought to involve the recombination of atomic hydrogen to gaseous hydrogen within the bulk metal. provide a good description of the likely mechanisms.
  5. SRO (short range ordering): This mechanism asserts that hydrogen damage occurs due to a hydrogen-mediated increase in the short range ordering of the austenitic phase. Such ordering is accompanied by physical shrinkage of the austenitic phase, and the subsequent creation of intra-granular tension that leads to both inter-granular and trans-granular cracking. The authors attempt to sweep pretty much all existing theories into this one grand theory. It is a fascinating read, and being a recent publication we’ll need to wait to see what other workers in the field make of it. .

Coming up for air:

Even the shallowest dive we could make into this subject, has ended up pretty deep, so let’s catch our breathe, and summarise were we have got to.

  • cold-working of metastable austenitic stainless steel has the potential to transform large fractions of austenite to martensite.
  • for a given level of induced strain, which is something that depends upon the desired form of the product, the fraction of austenite transformed to martensite depends crucially upon the nickel content of the alloy, as well as the temperature at which the part is formed. Any alloy containing less than 10% nickel that is worked at less than 25ºC is going to be a problem. Type 304 has that that potential, type 316 not so much.
  • for crags where attack by hydrogen is a possibility, i.e. all crags with high sulphate levels, then any anchor components comprising more than 30% strain-induced martensite look likely to fail some 10 years after installation. On the other hand, anchor components with less than 10% strain-induced martensite will have a service life in excess of 50 years. Put another way, 304 anchors will be short lived, 316 anchors could last many decades.
Cios, G. et al. The Investigation of Strain-Induced Martensite Reverse Transformation in AISI 304 Austenitic Stainless Steel. Metall and Mat Trans A 48, 4999–5008 (2017).
Wang, Y., Zhou, Z., Wu, W. & Gong, J. Warm Pre-Strain: Strengthening the Metastable 304L Austenitic Stainless Steel without Compromising Its Hydrogen Embrittlement Resistance. Materials 10, 1331 (2017).
Mati, D., Takasaki, A. & Uematsu, S. Hydrogen Susceptibility of Pre-strained Type 316L Austenitic Stainless Steels in Aqueous Solutions. IOP Conf. Ser.: Mater. Sci. Eng. 229, 012006 (2017).
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).
Kanezaki, T., Narazaki, C., Mine, Y., Matsuoka, S. & Murakami, Y. Effects of hydrogen on fatigue crack growth behavior of austenitic stainless steels. International Journal of Hydrogen Energy 33, 2604–2619 (2008).
Mine, Y., Narazaki, C., Murakami, K., Matsuoka, S. & Murakami, Y. Hydrogen transport in solution-treated and pre-strained austenitic stainless steels and its role in hydrogen-enhanced fatigue crack growth. International Journal of Hydrogen Energy 34, 1097–1107 (2009).
Lu, J. et al. Stacking fault energies in austenitic stainless steels. Acta Materialia 111, 39–46 (2016).
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).
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).
Haušild, P., Davydov, V., Drahokoupil, J., Landa, M. & Pilvin, P. Characterization of strain-induced martensitic transformation in a metastable austenitic stainless steel. Materials & Design 31, 1821–1827 (2010).
Nohara, K., Ono, Y. & Ohashi, N. Composition and Grain Size Dependencies of Strain-induced Martensitic Transformation in Metastable Austenitic Stainless Steels. Tetsu-to-Hagane 63, 772–782 (1977).
Soleimani, M., Kalhor, A. & Mirzadeh, H. Transformation-induced plasticity (TRIP) in advanced steels: A review. Materials Science and Engineering: A 795, 140023 (2020).
Talonen, J. & Hänninen, H. Formation of shear bands and strain-induced martensite during plastic deformation of metastable austenitic stainless steels. Acta Materialia 55, 6108–6118 (2007).
Zhang, L., An, B., Fukuyama, S., Iijima, T. & Yokogawa, K. Characterization of hydrogen-induced crack initiation in metastable austenitic stainless steels during deformation. Journal of Applied Physics 108, 063526 (2010).
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).
Ren, X. C. et al. A Nucleation Mechanism of Hydrogen Blister in Metals and Alloys. Metall and Mat Trans A 39, 87–97 (2008).
Whiteman, M. B. & Troiano, A. R. Hydrogen Embrittlement Of Austenitic Stainless Steel. CORROSION 21, 53–56 (1965).
Kim, Y., Kim, S. & Choe, B. The Role of Hydrogen in Hydrogen Embrittlement of Metals: The Case of Stainless Steel. Metals 9, 406 (2019).
San Marchi, C. Hydrogen embrittlement of stainless steels and their welds. in Gaseous Hydrogen Embrittlement of Materials in Energy Technologies 592–623 (Elsevier, 2012). doi:10.1533/9780857093899.3.592.
Herms, E., Olive, J. M. & Puiggali, M. Hydrogen embrittlement of 316L type stainless steel. Materials Science and Engineering: A 272, 279–283 (1999).

2 replies on “Austenite stability – are we missing something here?”

I enjoyed your article! You should read up on AM355 alloy, a “semi-austenitic” stainless steel. It exploits the very phenomena you describe with respect to carbon as an austenite stabilizer and the resulting metastability. If you keep the 0.1% carbon in solution it behaves as an austenitic alloy allowing extensive cold work. However, if you precipitate the carbon out, it raises the Ms temperature of the matrix and it hardens like a martensitic!

Leave a Reply

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