Fracture Mechanics Assessment of Hydrogen Effects on Steel

TWI Ltd, Granta Park, Great Abington, Cambridge, CB21 6AL

Paper presented at the International Steel and Hydrogen Conference
28 September 2011

Abstract

It has been very well known for many years that hydrogen as an interstitial solute within steel affects the steel properties. Of particular concern is the effect on fracture toughness, with any effects of hydrogen being universally adverse.

This paper describes TWI’s approach for determining the fracture toughness on C-Mn steel in the hydrogen-charged condition (denoted as KIH or equivalent parameters in terms of CTOD (δ) or J). The paper focuses, in particular, on specimen preparation, determination of loading (or strain) rate, expression of results as single-point values or as a resistance curve, associated limitations and practical implications and usage in ECA projects, and recent thinking with regard to development of alternative and improved approaches.

Introduction

It has been very well known for many years that hydrogen as an interstitial solute within steel affects the steel properties. Of particular concern is the effect on fracture toughness, with any effects of hydrogen being universally adverse. Data published by Graville et al (1) on the effect of hydrogen on notch tensile strength (assumed to be somewhat similar to effect on fracture toughness) in 1967 provide a useful summary of the generally observed effects. In Figure 1, from this work, it can be seen that the universally adverse effect on notch tensile strength is also dependent on temperature and strain rate.

In this relatively high strength ferritic steel, the greatest embrittling effect is seen at around normal ambient temperature, and over the full range of temperature, embrittlement is greater at slower strain rate. Thus, if an engineering critical assessment (ECA) is to be performed on equipment in service which introduces hydrogen into the steel, the effect of hydrogen on toughness needs to be taken into account. Such service is most commonly high pressure/high temperature hydrogen containment, or corrosive service, particularly in the presence of H2S (sour conditions). Cathodic polarisation, applied in wet (ground water or sea water) conditions as a corrosion protection measure also introduces some hydrogen into the steel. Furthermore, in considering the effects of hydrogen, due account needs to be taken of temperature and strain rate. It is also generally recognised that higher strength steels are more susceptible to hydrogen embrittlement than lower strength steels. However, although low strength steels may be immune to cracking due to hydrogen (for example, steels of below 250HV hardness are commonly found to be resistant to cracking in sour service), they will still suffer embrittlement.

In 1984, the then Union Oil company suffered the catastrophic failure of an amine absorber tower, with resultant loss of 17 lives. In this failure an unstable fracture developed from a service crack, and the unstable fracture could only be understood in terms of fracture toughness measured on the steel after charging it with hydrogen. Following this failure, Exxon did a large amount of work, including a programme of fracture toughness testing carried out at TWI (2). Care was taken to determine an appropriate slow strain rate, and testing was performed at normal, ambient temperature. Different corrosive media were used to charge the steels with hydrogen, and the hydrogen content of each test specimen was measured on completion of testing.

The results (Figure 2) showed a marked effect on hydrogen on CTOD fracture toughness, with a level of <0.2mm being achieved at 20ºC for all the steels explored once they contained over 2ml/100g diffusible hydrogen.

Significantly higher hydrogen concentrations than this can be anticipated from wet sour service (Figure 3). The fracture surfaces of the hydrogen charged specimens showed a quasi-cleavage morphology, typical of hydrogen embrittlement in a C-Mn steel. The CTOD curves, however, all went to maximum load, such that δm values were recorded, indicating that macroscopic unstable fracture did not occur.

Since the above work, it has been observed on several occasions that hydrogen can induce quasi cleavage fracture, and a significant reduction in fracture toughness (determined in terms of CTOD (crack tip opening displacement, δ) or J-integral, J) in comparison with fracture toughness levels in air or inert environments. Such a reduction in fracture toughness can have a significant influence on the flaw tolerance of hydrogen-containing components, which may in some cases result in smaller flaws than would be accepted according to workmanship criteria.

At present, there are no standard procedures for determining the fracture toughness of steel components in the hydrogen-charged condition. TWI has developed an in-house procedure, in response to the needs of, mainly, the oil and gas sector both in the UK and internationally. This procedure, which has been used in numerous projects, continues to be developed and improved to take into account new findings from work at TWI and elsewhere and to reflect the changing needs of the industry.

This paper describes TWI’s approach for determining the fracture toughness on C-Mn steel in the hydrogen-charged condition (denoted as K_IH or equivalent parameters in terms of CTOD (δ) or J). The paper focuses, in particular, on specimen preparation, determination of loading (or strain) rate, expression of results as single-point values or as a resistance curve, associated limitations and practical implications and usage in ECA projects, and recent thinking with regard to development of alternative and improved approaches.

Objective of fracture toughness testing on hydrogen-charged material

The testing is aimed at determining the fracture toughness of a component made from C-Mn steel in the hydrogen-charged condition, i.e. which contains absorbed hydrogen. The hydrogen source may, for example, be chemical reactions associated with a wet sour environment that the component is exposed to on one or several surfaces. Results from the testing are used determine the significance of flaws in the component with regard to failure, under quasi static loads, associated with crack extension through hydrogen-charged material. Thus, such results (referred to as KIH, if the driving force is expressed in terms of the stress intensity factor, K) apply to flaws, such as embedded flaws, that are exposed to hydrogen absorbed within the steel, but which are not exposed directly to the hydrogen charging source. Flaws that are exposed directly to the hydrogen-charging source, such as surface breaking flaws on the inside surface of pipelines carrying sour products, are normally assessed using a measure of the fracture toughness referred to as K_ISSC, which represents the threshold stress intensity factor above which cracks will extend by a sulphide stress cracking mechanism.

Specimen preparation

Normally, standard single-edge-notch-bend (SENB) specimens are utilised, but alternative specimens such as single-edge-notch-tension (SENT) specimens may be used (provided that possible effects of crack tip constraint in the specimen and structural component are adequately considered). Specimens of appropriate dimensions (normally with a thickness as close as possible to that of the actual component) are prepared according to standard procedures such as those of BS 7448. After notching (into the parent metal, weld metal or heat affected zone), the specimens are fatigue pre-cracked. Fatigue pre-cracking procedures are important since it is necessary to ensure that the maximum K during pre-cracking is significantly below KIH expected in the tests; otherwise the determination of KIH may be affected. The specimens are then immersed in a chemical environment, intended to be representative of the actual service environment, to allow the specimen to be charged with hydrogen. In the case of wet sour service, this may require a CO₂/H₂S mixture to be bubbled through a brine/acetic acid solution at a given temperature. A range of environments from NACE TM0177 solution A, which is very severe, to a solution with a composition matching that experienced in service can be used (TWI experience suggests that exposure of a 25mm thick specimen for about a week is required to give the maximum atomic hydrogen levels). During exposure, the crack tip must be protected from the environment to avoid any corrosion crack tip blunting, typically with a thin strip of tape or mastic.

After exposure to the chemical environment, the specimens are stored at low temperature (for example, in liquid nitrogen) to prevent the hydrogen diffusing out prior to testing. The level of hydrogen within the specimens can be measured at this stage if an additional dummy specimen had been included during the chemical environment exposure.

Test loading rate

Prior to testing, each specimen is warmed up to the test temperature. Testing is normally conducted at room temperature since this is believed to provide the maximum embrittling effect in ferritic steels (see Figure 1). The specimen is then loaded at a very slow strain rate, to enable the hydrogen to diffuse to the crack tip during the test. The strain rate should be at least an order of magnitude lower than that conventionally used in fracture mechanics tests. However, there are two competing effects with regard to determining the influence of strain rate, as the maximum embrittlement effect requires a sufficiently long time to enable diffusion of hydrogen to the crack tip, but a short enough test period to minimise escape of hydrogen from the specimen. An optimal (or compromise) strain rate could be determined by conducting preliminary tests on a number of similar specimens (same dimensions, notch and fatigue pre-crack location and size) under a range of strain rates. The strain rate that leads to the lowest fracture toughness result is then adopted in subsequent tests.

Figure 4 illustrates schematically force vs. displacement curves for three nominally similar specimens that were tested under decreasing strain rates (highest for Specimen A and lowest for Specimen C). It can be seen that although the three force vs. displacement curves each reach a different maximum force, that for Specimen C was the lowest and was reached much earlier, i.e. at a significantly lower displacement, than for Specimens A and B.

Single-point vs. R-curve fracture toughness tests

For non-hydrogen-charged materials, conventional fracture toughness testing produces single-point values of the fracture toughness, which depend on the highest force recorded during the test. The results are expressed in terms of CTOD or J and different subscripts are commonly used to denote the type of fracture behaviour. For instance, the subscript 'c' (result classed as δc or J_c) denotes unstable fracture with limited tearing (< 0.2mm), while the subscript u (result classed as δu or J_u) denotes tearing greater than 0.2mm but failure prior to reaching a maximum force plateau on the force; clip gauge displacement curve (typically when the material is on the lower shelf of transition region). The subscript 'm' indicates that a maximum force was reached on the force vs. clip gauge displacement curve, usually after a certain amount of crack extension by ductile tearing, and fully-ductile behaviour (result classed as δm or Jm). In the latter case, fracture toughness can be expressed in terms of a resistance curve (or R-curve), where fracture toughness is given as a function of the tearing length. R-curves can be determined using a multiple-specimen technique. This requires that several (generally at least six) specimens to be tested to different load/strain levels (i.e. giving different amounts of tearing). An R-curve can then be fitted to the data points. Alternatively, R-curves can be determined using single specimen methods requiring that crack growth be estimated during the test. The most commonly used procedure is the unloading compliance method, where the specimen is repeatedly partially unloaded during the test to measure the specimen compliance and infer the amount of crack growth.

For hydrogen-charged material, conventional fracture toughness testing, where the specimen is loaded until it reaches a maximum load or fails prior to that, produces single-point values of the fracture toughness. Normal practice is to conduct testing on at least three nominally similar specimens notched into the parent metal, weld metal and/or HAZ. (Owing to the narrowness and inhomogeneous microstructures of most HAZs, it is usual to test more than three specimens to achieve three reliable results for this region.) Typically, the load-displacement curves for these tests (hydrogen-charged) show that a maximum force was reached during the test. This implies, according to the approach utilised to classify results for non-hydrogen-charged steel (outlined above), that the results can be classed as maximum load values (and denoted as δm or J_m). However, such classification is not strictly-speaking correct since the fracture surface normally shows a quasi-cleavage appearance, i.e. the failure is significantly different from ductile failures observed when maximum load values are obtained from tests on non-hydrogen-charged steel. An example is shown in Figure 5. An interpretation of this failure mode and its practical implications are given in the discussion section below.

As an alternative to single-point values, fracture toughness testing on hydrogen-charged steel can be conducted to generate R-curves. The approach would be identical to that outlined above, except that a multiple-specimen technique is preferred to the alternative unloading compliance technique. The latter involves repeated partial unloading and re-loading steps, which when a test is conducted at a very low strain rate would lead to a much longer test duration and potentially significantly greater hydrogen loss during the test.

Figure 6 shows schematically an idealised approach for generating a resistance curve using six specimens. The test denoted as Specimen 6 would be identical to that denoted as ‘C’ in Figure 5 if the testing conditions for the latter, mainly the loading or strain rate, were judged to have produced the lowest fracture toughness results. Specimens 1 to 6 are nominally identical, except that they are tested to six different load/strain levels such that each specimen produces a fracture toughness result at a different amount of crack extension. An R-curve can then be constructed as a curve-fit to the data points. However, as will be shown in the discussion below, since such an R-curve is both strain and time-dependent, its use may lead to unsafe assessments. Until a better understanding of relevant parameters is developed, it is recommended that the fracture toughness at initiation of crack extension (normally defined, in the absence of hydrogen, by the point at which the R-curve intersects the 0.2mm offset blunting line) be utilised in practical flaw assessment.

Discussion

When single-point fracture toughness testing is conducted on hydrogen-charged specimens, the load displacement curve will typically look ductile, in so far as it curves over maximum load smoothly, but much earlier (at lower displacement) than for a similar uncharged specimen (see Figure 5). However, this apparent ductility is misleading and the fracture face will commonly have a quasi-cleavage morphology (which is very different from the dimpled appearance resulting from growth and coalescence of microvoids in ductile steel in the absence of hydrogen).

A likely explanation for the smooth curve is that the fracture mechanism involves small steps of subcritical crack growth, through a hydrogen-charged plastic zone at the crack tip, which advances with the growing crack. Hydrogen will accumulate in a tensile strained region due to the higher solubility of hydrogen in such regions, thus embrittling this region to a greater extent than the surrounding material. Following a step of crack growth, hydrogen will accumulate in the new plastic zone until there is a sufficient hydrogen concentration to cause a further subcritical crack growth step. This explains the effect of strain rate on measured fracture toughness and the approach outlined above consisting of testing at the slowest possible strain rate. It must be remembered, however, that a low strain rate will result in a long overall test duration, which will enable hydrogen to continually escape from the steel throughout the test. Thus, by the time the maximum load is reached, the hydrogen content, and hence the embrittling effect, will be reduced. This potential problem can be remedied if the level of hydrogen in the specimen is maintained by keeping the hydrogen-charging source in direct contact with the specimen (but not with the crack face region) throughout the test.

Another feature of the above fracture mechanism, is that a rising load can be accommodated at the new crack tip position, reached after a subcritical crack growth step, since the crack will not advance by hydrogen embrittlement until a sufficient amount of hydrogen has accumulated at the new crack tip to cause a further crack extension. This implies that if the load were to be held constant after a subcritical crack growth step, further propagation would be anticipated eventually. Thus, the fracture toughness value derived from such a test at maximum load may be non-conservative, in that the increase in load (and corresponding recorded fracture toughness) is influenced by the displacement rate employed in the test and the corresponding crack tip loading rate and associated hydrogen embrittling rate. Consequently, the resistance to crack extension observed in the laboratory (characterised by a rising R curve) may be significantly higher than that which would occur in practice under quasi constant loads. For these reasons, and until superior testing procedures are developed (for example consisting of conducting the test while the specimen is continually charged with hydrogen) it is recommended to use the toughness at initiation of crack extension rather than the full R curve. [It is worth noting that fracture toughness results classed as δm or J_m, and obtained for hydrogen-charged or non-hydrogen-charged materials, are not material properties since they depend on specimen geometry and represent the fracture toughness after a certain amount of crack extension that cannot be determined from single point tests.]

In considering a way forward, it is important to attempt to simulate the loading conditions being assessed, with regard to loading rate, as closely as possible but in a conservative way. It may be argued that when the actual maximum loading on the component is constant or is applied very slowly, the relevant fracture toughness is that determined as described above, i.e. a value corresponding to initiation of crack extension determined at the lowest possible loading rate. During testing, the specimen should either be continually charged with hydrogen and/or measures should be implemented to prevent escape of hydrogen during the test.

A different approach for assessing components subjected to rapidly increasing loads could be justified. In this case, the use of fracture toughness results obtained under a loading rate significantly lower than that experienced by the component can be overly conservative. As an alternative, the assessment could be conducted using a fracture toughness determined under a loading rate just below that experienced by the component. The fracture toughness could be defined in terms of an R-curve provided that uncertainties are allowed for, for example via safety factors.

Summary and Conclusions

• Fracture toughness testing on hydrogen-charged specimens, is normally conducted at room temperature, under a very low loading (or strain) rate. This can be specified as a compromise satisfying two competing effects, as the maximum embrittlement effect requires a sufficiently long time to enable diffusion of hydrogen to the crack tip, but a short enough test period to minimise escape of hydrogen from the specimen.

• The fracture surfaces of hydrogen-charged specimens typically show a quasi-cleavage morphology, but the load vs. displacement curves show that a maximum load is reached which implies a ductile response.

• It is likely that the ductile appearance of the load vs. displacement curve from a test specimen is due to the fracture mechanism, which involves small steps of subcritical crack growth, associated with a hydrogen-charged plastic zone at the crack tip, which advances with the growing crack. After a subcritical crack growth step, a rising load can be accommodated since the crack will not advance by hydrogen embrittlement until a sufficient amount of hydrogen has accumulation at the new crack tip to cause a further crack extension.

• It is believed that if the load were to be held constant after a subcritical crack growth step, further propagation would be anticipated eventually. However, a higher load can be accommodated in a rising load test, since the crack will not advance by hydrogen embrittlement until a sufficient amount of hydrogen has accumulated at the new crack tip to cause a further crack intrusion. Thus, the fracture toughness value derived based on the maximum load reached in the test may be non-conservative.

• As an alternative, at least for assessing components under quasi-static loads, it is recommended that the relevant fracture toughness is that corresponding to initiation of crack extension determined from an R curve derived from tests conducted at the lowest possible loading rate or even under a constant load.

• Components subjected to rapidly increasing loads may be assessed using fracture toughness results determined under a loading rate just below that experienced by the component. The fracture toughness could be defined in terms of an R-curve provided that uncertainties are allowed for, for example via safety factors.

• Hydrogen loss during testing should be minimised and/or accounted for in all testing, particularly at very low strain rate and under static loading conditions.

 

References

1. Graville B A, Baker R G and Watkinson F: ‘Effect of temperature and strain rate on hydrogen embrittlement of steel’, British Welding Journal 14(6) June 1967.

2. Humphries, M J, McLaughlin, J E and  Pargeter, R J: ‘Toughness characteristics of hydrogen charged pressure vessel steels’, Int Conf on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel Service, Paris, France, 28-30 March, 1989.