You are on page 1of 22

Engineering Failure Analysis 100 (2019) 329–350

Contents lists available at ScienceDirect

Engineering Failure Analysis


journal homepage: www.elsevier.com/locate/engfailanal

Review

A review of underlying reasons for intergranular cracking for a


T
variety of failure modes and materials and examples of case
histories
Stan Lyncha,b,c,

a
Defence Science and Technology, Melbourne, Australia
b
Dept. of Materials Engineering, Monash University, Melbourne, Australia
c
International School of Engineering, Chulalongkorn University, Bangkok, Thailand

ARTICLE INFO ABSTRACT

Keywords: The reasons why cracking and corrosion can occur preferentially along grain boundaries in a
Intergranular fracture wide variety of metallic materials are discussed, along with case histories of failures involving
Environmentally assisted cracking intergranular fracture and corrosion. Failure modes discussed include fast fracture, fatigue (in
Fractography inert and embrittling environments), liquid-metal embrittlement, hydrogen embrittlement, stress
Microstructures
corrosion cracking, and corrosion fatigue, in materials such as steels, Al alloys, Ni alloys, and Cu
Fracture mechanisms
alloys. The features sometimes observed on grain-boundary facets (e.g. fatigue striations, crack-
arrest markings, slip lines, steps/microfacets, nano−/micro-scale dimples, second-phase parti-
cles, and deposits), and how they are formed, are described, as such knowledge helps in diag-
nosing fracture modes and causes. Reasons for intergranular cracking in the cases considered
include (i) preferential precipitation at grain boundaries (sometimes with an associated pre-
cipitate-free zone), (ii) segregation of metalloid impurities, hydrogen, or some alloying elements,
(iii) preferential adsorption of embrittling environmental species (such as some metal atoms and
hydrogen) at intersections of grain boundaries with surfaces/crack tips, and (iv) locally anodic
regions at or adjacent to grain boundaries. Some lessons for failure-analysis and prevention in
general, illustrated by the case histories, are then summarised.

1. Introduction

An understanding of grain-boundary phenomena and characteristics is essential for carrying out failure-analysis investigations
since almost all metallic structures and components are polycrystalline, and intergranular fracture and corrosion are common when
failures occur [1]. There are many reasons why grain boundaries are preferred sites for fracture and corrosion, and all types of failure
can exhibit intergranular (or mixed intergranular and transgranular) crack paths. For example, intergranular failure has been ob-
served for fast fracture and fatigue (in air) in some materials, and is especially prevalent for environmentally assisted cracking modes
such as hydrogen embrittlement (HE), liquid-metal embrittlement (LME), solid-metal-induced embrittlement (SMIE), stress-corrosion
cracking (SCC), and corrosion fatigue (CF) in many materials [2–4]. Failures at high homologous temperatures (T/Tm > 0.5) (where
Tm is the melting point in Kelvin units) also occur predominantly at grain boundaries [5].
In the present paper, some basic aspects of grain boundaries are summarised. Then the numerous features that can be present on


Defence Science and Technology, Melbourne, Australia.
E-mail address: lynch_stan@hotmail.com.

https://doi.org/10.1016/j.engfailanal.2019.02.027
Received 24 November 2018; Received in revised form 11 February 2019; Accepted 14 February 2019
Available online 16 February 2019
1350-6307/ Crown Copyright © 2019 Published by Elsevier Ltd. All rights reserved.
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 1. Schematic diagram illustrating examples of repeating polyhedral structural units at a grain boundary. The closed and open circles represent
atoms on two different planes. (Adapted from [8]). The geometry of the structural units depends on the misorientation of grains and the grain-
boundary plane.

intergranular fracture surfaces, and how they are formed, are described. Being able to recognise and interpret these features is
obviously essential for diagnosing modes and causes of failure, and making recommendations for preventing failure. Examples of
failures associated with problematic grain-boundary microstructures, grain-boundary segregation, and preferential environmental
interactions at grain boundaries, are then outlined. Some general lessons for failure analysis and prevention arising from these cases
are then discussed.

2. Some grain-boundary fundamentals [6–10]

The intrinsic grain-boundary structure (atomic arrangements) depends on the misorientation between adjacent grains and on the
grain-boundary plane. Low-angle tilt grain boundaries, for example, can be characterised as arrays of edge dislocations, while high-
angle grain boundaries (> 10-15o misorientation) are best described as having one or more repeating polyhedral structural units
(Fig. 1). Some high-angle grain boundaries (with particular grain misorientations and grain-boundary planes) have low energies,
which are sometimes associated with a high number of atomic sites coincident with both grain lattices. Some of these ‘special’
boundaries may be more resistant to fracture and corrosion, and efforts have been made to increase the proportion of such boundaries
by thermo-mechanical processing (‘grain-boundary engineering’) so that boundaries with inferior properties do not form a continuous
crack path [11,12].
An under-appreciated aspect of high-angle grain boundaries is that their structure (and properties such as fracture resistance) can
change as a result of changes in temperature and stresses, as well as changes in composition (segregation), for a given misorientation
and plane [13–17]. Thus, when the stress-axis is taken into account along with the lattice orientations and the grain-boundary plane,
there are “eight degrees of freedom” that can affect the grain-boundary structure [18]. Evidence for reversible grain-boundary-phase
transitions owing to changing temperatures includes: (i) transitions from planar to faceted boundaries, (ii) re-arrangement of intrinsic
grain-boundary dislocations, and (iii) discontinuities in the kinetics of grain-boundary diffusion and migration. The wide variety of
other phenomena that can occur at grain boundaries, which all depend on the grain-boundary structure, is summarised schematically
in Fig. 2, and discussed further in the following paragraphs.
Most high-angle grain boundaries (with the exception of ‘special’ boundaries) are preferred sites for fracture and corrosion, owing
to (i) segregation of embrittling metalloid impurities, hydrogen, or some alloying elements at boundaries, (ii) preferential pre-
cipitation (sometimes with an associated precipitate-free zone) at boundaries, and (iii) preferential adsorption of embrittling en-
vironmental species (such as some metal atoms, hydrogen, and some specific ions) at intersections of grain boundaries with surfaces/
crack tips. The somewhat lower co-ordination number for atoms bridging adjacent grains (compared with atoms within grains) does
not, however, generally lead to lower cohesion across grain boundaries (in the absence of other effects) for most materials, although
there are exceptions.
Significant segregation of metalloid impurities, e.g. P, S, Sb, Sn in steels, can occur at grain boundaries even when bulk impurity
concentrations are low, with ‘enrichment factors’ of 103 to105, depending on the particular impurity, alloying elements, and heat-
treatment [19,20]. Increasing extents of segregation increase the extent of embrittlement, and some elements are more embrittling
than others for a given coverage (e.g. Sb > P for some steels). Impurity-induced embrittlement probably involves the formation of
strong bonds between the impurity atoms and substrate atoms such that adjacent metal-metal bonds are weakened, resulting in
‘decohesion’ (atomically brittle fracture) or dislocation emission from crack tips. However, segregation is not always deleterious, and
some elements can increase cohesion across grain boundaries or displace embrittling elements. The relative sizes of segregants and
substrate atoms as well as the nature of the chemical bonding between them probably determines the behaviour.
For quenched and tempered martensitic steels, there are two main types of grain-boundary embrittlement, viz. (i) tempered
martensite (one-step) embrittlement (TME), and (ii) two-step (reversible) temper embrittlement (TE) [19,20]. TME occurs when
steels are tempered in the 300 to 450 °C range for short times (1−2 h), and is manifest as a ‘ductility trough’ in plots of fracture-
energy versus tempering-temperature. Fracture is often (but not always) along prior-austenite grain boundaries, and results from the
combined effects of metalloid-impurity segregation (occurring predominantly during austenitizing) and the precipitation of carbides
with weak interfaces at boundaries during tempering. Two-step TE can occur when steels are tempered above 600 °C and then re-

330
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 2. Schematic diagrams illustrating various phenomena that can occur at grain boundaries (a) segregation of metalloid impurities, hydrogen,
and solute atoms, (b) preferential grain-boundary precipitation and precipitate-free zones, (c) grain-boundary sliding/migration/fast-diffusion, (d)
preferential adsorption or dissolution at grain-boundary/surface/crack-tip intersections (in the absence of oxide films), and (e) grain-boundary/slip
interactions.

tempered or slowly cooled in the 350 to 500 °C temperature range. In this case, segregation predominantly occurs during tempering
at the lower temperatures owing to ‘rejection’ of impurities from growing carbides. It is manifest as an increase in the ductile-to-
brittle transition temperature (DBTT), with intergranular fracture predominating below the DBTT and dimpled transgranular fracture
occurring above the DBTT. Both types of temper embrittlement described above involve fast fracture, but slow intergranular crack
growth can occur in steels and some other materials under sustained (residual or applied) stresses at elevated temperatures owing to
stress-induced diffusion of impurities or alloying elements to grain boundaries just ahead of crack tips [21–23].
The degree of grain-boundary embrittlement of tempered-martensitic steels by hydrogen increases considerably when the
strength (hardness) is greater than about 1200 MPa (38HRC), and is exacerbated by the presence of metalloid-impurity segregation
(as is the case for other alloys) [24,25]. Bulk hydrogen concentrations of only 1–2 ppm (wt%) are sufficient to produce HE in high-
strength steels, and such concentrations can easily be exceeded during electroplating and pickling unless procedures for minimising
hydrogen uptake, or removing hydrogen by ‘baking’ after such treatments, are not followed. Hydrogen can also be produced on steel
surfaces or at crack tips, and then diffuse/segregate along grain boundaries during service in hydrogen-bearing environments,
especially if cathodic-protection is applied or sacrificial coatings are present.
In the absence of oxide films (after oxide rupture or dissolution), adsorption of embrittling species (including hydrogen) at
surfaces/crack-tips can occur preferentially at grain-boundary/surface intersections. This adsorption can weaken surface:substrate
interatomic bonds and thereby facilitate crack growth, without diffusion/segregation being necessary [3]. Indeed, for metal-atom
adsorbates (resulting in LME), crack-growth rates can be so rapid (100 s mm/s) that there is no time for diffusion ahead of inter-
granular (or transgranular) cracks [4]. Some metal adsorbates are more embrittling than others, and some are non-embrittling, e.g.
LME of Ni occurs in Li and Na, but not in other alkali metals. However, the reasons for this ‘specificity’ is not understood. Adsorption
of complex ions may be responsible for SCC in some systems, e.g. Cu alloys where de-alloying occurs (often preferentially at grain
boundary/surface intersections), but the mechanism of SCC in such systems is especially controversial [2].
A variety of grain-boundary/dislocation interactions can occur (Fig. 2(e)), with the detailed behaviour dependent on grain
misorientation, grain-boundary composition, and other variables [26–28]. For example, grain boundaries can act as barriers to slip,
so that a smaller grain size results in increasing strength (via the well-known Hall-Petch σy = σo + kd-½ relationship – valid down to
~100 nm grain size). Dislocation pile-ups at grain boundaries can, however, produce locally high stresses that could lead to initiation
of cracks/voids at boundaries. Strain incompatibilities in adjacent (‘hard’ and ‘soft’) grains can also produce locally high stresses near
grain boundaries, especially around triple points. Stresses owing to dislocation pile-ups are likely to be higher when planar slip

331
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 3. SEM of fracture surface of pure iridium after fast fracture in air, showing relatively smooth intergranular facets and transgranular cleavage-
like area with numerous steps. A mixed mode fracture occurs because some grain boundaries are more fracture-resistant than others owing to
‘special’ misorientations/planes. Pure iridium (with no grain-boundary segregation or precipitation) is unusual in that bonding across grain
boundaries is intrinsically weak, for reasons that are not well understood, although Ir does exhibit more direction bonding than most metals, and has
the highest shear modulus.

(confined to narrow bands) rather than wavy, more homogeneous slip, occurs, and when slip transmission into the adjacent grain is
difficult. Planar slip occurs especially in materials strengthened by coherent precipitates since dislocations can shear the precipitates,
thereby reducing their strengthening effectiveness such that further slip on the same plane is facilitated. Higher stacking-fault energy
materials, and materials containing solute hydrogen (as atmospheres around dislocations) can also be more prone to planar slip.
Stresses at grain boundaries owing to dislocation pile-ups can be relieved by slip transmission across grain boundaries or by grain-
boundary migration, rather than by cracking. Grain-boundary sliding can also result in high stresses, especially at triple points, and
thereby promote cavitation, especially at elevated temperatures [5].
Grain-size and grain-shape are other important variables influencing intergranular fracture and corrosion resistance. For example,
materials with elongated grains resulting from rolling, extrusion, etc. are more prone to intergranular fracture, and can have a much
lower fracture resistance, for short-transverse crack-plane orientations (S-L, S-T), as a result of most high-angle grain boundaries
being oriented normal to the tensile-stress-axis, compared with other orientations (such as L-T and T-L). Decreasing grain size could
result in decreasing impurity segregation at grain boundaries owing to increasing grain-boundary area. The resistance to cleavage
fracture (across grains) may also be increased by decreasing grain size because cleavage cracks growing along low-index crystal-
lographic planes have to re-orient/re-initiate on crossing grain boundaries. Grain boundaries in nanocrystalline materials (grain
size < 100 nm) may behave differently from those in microcrystalline materials, e.g. sliding may occur at lower temperatures [29],
but consideration of such effects is beyond the scope of this paper.

3. Intergranular fracture features and their formation [2–4,30,31]

At low magnifications, intergranular fracture surfaces generally have a smoother appearance with fewer steps than transgranular
cleavage facets (Fig. 3). In addition, when grains are equiaxed (dodecahedral-shaped), the angles between intergranular facets on
fracture surfaces are generally larger than those for cleavage facets parallel to low-index crystallographic planes. Secondary cracks
along grain boundaries intersecting the main fracture path are also more common for intergranular cracking than for cleavage. At
high magnifications, intergranular facets can sometimes be featureless, but more commonly exhibit a variety of fine-scale features,
such as steps, dimples, fatigue striations, and crack-arrest markings, depending on the grain-boundary microstructure, fracture mode,
and environment. Many of these features are associated with localised plasticity occurring during cracking, and this plasticity is also
evident from observations on the pre-polished side surfaces of specimens, polished and etched sections, TEM of foils from just beneath
fracture surfaces, and other techniques (Fig. 4). Deformation becomes more localised as the stress-intensity factor, K, decreases for
sustained-load cracking, and is more localised in higher strength materials. Increasing extents of embrittlement also result in plas-
ticity being more localised, and deformation may not occur to any significant extent when embrittlement is particularly severe.

3.1. Steps on intergranular facets

Wavy or ‘blocky’ steps on intergranular fracture surfaces are sometimes observed because grain boundaries were stepped/faceted
prior to brittle fracture (Fig. 5) [32]. These steps occur, especially when segregation is present, in order to minimise the grain-
boundary energy. Steps on fracture surfaces may also be the result of slip bands impinging on boundaries ahead of cracks to produce a
ledge, followed by brittle fracture. Deformation also occurs just behind crack tips (when blunting occurs at crack tips), so that slip
steps and twin traces are produced on fracture surfaces (Fig. 6). Slip steps can be distinguished from ledges because they do not match
with steps on the mating areas of opposite fracture surfaces (unlike other steps). Terrace and step patterns can also occur on the
surfaces of inter-grain porosity during solidification to minimise surface energies, if there is insufficient liquid to fill cavities between
grains (Fig. 7).

332
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 4. Optical micrographs of (a) pre-polished surface showing extensive slip associated with intergranular cracking in nickel embrittled by liquid
mercury at 20 °C, and (b) polished and etched section of intergranular SCC of a peak-aged Al-Zn-Mg alloy in distilled water at 20 °C, showing
localised slip adjacent to grain boundaries (revealed by ageing after SCC so that precipitation occurred on dislocations, which affected the etching
response). The more extensive slip at A is probably associated with a crack-arrest/blunting event. Lighter etching areas adjacent to grain boundaries
at B are more highly strained (higher dislocation density) than areas further from the boundaries. Variations in strain are presumably associated
with changes in grain-boundary plane/structure because boundaries are curved.

Fig. 5. TEM of replicas of intergranular fracture surface of nickel embrittled by sulphur segregation, showing (a) blocky and (b) wavy steps, owing
to the grain boundaries being faceted prior to brittle fracture [32].

Fig. 6. SEM showing three sets of intersecting slip lines produced by deformation behind crack tips on an intergranular facet of nickel embrittled by
hydrogen charging.

3.2. Particles and precipitates on intergranular facets

Grain-boundary particles/precipitates (GBPs) are most clearly evident on intergranular facets when brittle intergranular fracture
(with minimal plasticity) occurs. Moreover, SEM fractographic observations of small GBPs are sometimes the best way of revealing
their size and distribution, since much larger areas can be examined compared with transmission-electron microscopy (TEM) of thin
foils (although the latter can resolve smaller precipitates and determine compositions using energy-dispersive spectroscopy (EDS).
(Compositions of GBP can also be determined by TEM of fracture surfaces of replicas if the particles are ‘extracted’.) Brittle

333
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 7. SEM of inter-grain porosity in a 2297-T87 Al-Li alloy showing solidification growth patterns, and second-phase particles (one with an
unusual shape) protruding from the surface. Some of the straighter lines in the lower part of the image are slip lines.

intergranular fracture to reveal precipitation can sometimes be promoted by testing at −196 °C (in liquid nitrogen) for some alloys
(Fig. 8), or by exposure of Al alloys to liquid gallium (at ~35 °C) which results in rapid diffusion of gallium along grain boundaries to
produce a liquid film (several nano-meters thick) so that grains separate easily.
In a study of 7xxx Al alloys, the sizes and distribution of GBPs on brittle intergranular fracture surfaces produced at −196 °C were
measured as a function of alloy composition and ageing time, along with TEM thin-foil analyses of GBP composition and electro-
chemical (open-circuit potential) measurements on intergranular fractures. These studies showed that the SCC resistance of 7xxx Al
alloys was determined largely by the composition of GBPs rather than by their size/area fraction, and that overageing increased the
SCC resistance by increasing the Cu content of GBPs (if the alloy contained sufficient copper) [33].

3.3. Dimples on intergranular fractures

Dimpled intergranular fracture surfaces for fast fracture in air occur as a result of preferential nucleation, growth, and coalescence
of micro/nano-voids along grain boundaries in preference to void nucleation and growth within grains. They are observed especially
in precipitation-hardened materials where closely spaced GBPs nucleate voids as a result of localised strains at grain boundaries
causing separation of precipitate-matrix interfaces [34]. The strains causing nucleation and growth of voids are especially localised
when soft PFZs are present adjacent to grain boundaries. PFZs arise owing to either (i) preferential growth of GBPs (vis-à-vis matrix
precipitates) resulting in local depletion of solute required for matrix precipitation, or (ii) grain boundaries acting as vacancy sinks
during quenching so that the vacancy concentration adjacent to boundaries is insufficient to nucleate matrix precipitates [35].
The size and depth of dimples on intergranular fracture surfaces decreases with decreasing size and spacing of GBPs (as would be
expected) and, when PFZs are present, the depth of dimples decreases with decreasing PFZ width, providing GBPs are not too closely
spaced (Fig. 9). In some cases, dimples on intergranular facets are so small and shallow that they can only be resolved by high-
resolution fractographic techniques, e.g. using field-emission-gun scanning electron microscopes (FEGSEM) or TEM of replicas
shadowed at low angles and examined under optimum conditions. Thus, SEM observations of featureless intergranular facets cannot
be claimed as evidence that fracture has occurred by an atomically brittle (decohesion) process, although such claims are surprisingly
common.
Small, shallow dimples on intergranular fractures have also been observed after environmentally assisted cracking (LME, HE,
SCC) in materials where fast fracture in air occurs by a dimpled transgranular mode (large, deep dimples with small dimples within

Fig. 8. SEM of brittle intergranular facets showing MgZn2(Cu) GBPs and dispersoids (and sites where they had been present but remained attached
to the opposite fracture surface) for an overaged 7075-T7 Al-Zn-Mg-Cu alloy tested at −196 °C [33].

334
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 9. SEMs of intergranular facets in an age-hardened Al-Zn-Mg alloy with (a) a wide PFZ and well-spaced GBPs (inset) showing well-defined
dimples, and (b) narrow PFZ and closely spaced GBPs (inset) showing very shallow dimples. (Note differences in scales) [30,31].

the larger dimples) [2–4]. For example, sub-critical crack growth of a tempered-martensitic steel tested in liquid mercury and gaseous
hydrogen produced similar dimpled intergranular fracture surfaces (Fig. 10). Moreover, the size and depth of dimples on inter-
granular facets (and the proportion of intergranular fracture) depended on the steel heat-treatment (tempering temperature), but was
similar for HE (in H2) and LME (in Hg) for each temper condition [2].
The similarities between LME and HE in steels, and in many other materials [2,3], suggest that a common mechanism may be
involved. LME in most systems involves only adsorption at crack tips since crack velocities are high (10s mm/s) and mutual solu-
bilities are negligible, so there is no time or tendency for other reactions (such as grain-boundary diffusion) [4]. Thus, the similarities
between HE and LME (along with other evidence) suggests that HE is due to adsorbed hydrogen at crack tips rather than solute
hydrogen ahead of cracks (for materials where hydrides are not involved) [2,3]. Transgranular as well as intergranular LME and HE
can occur, but the latter is often favoured because adsorption occurs preferentially at crack-tip-surface/grain-boundary intersections.
Mechanisms of HE based on adsorbed hydrogen are applicable not only to hydrogen-environment embrittlement (in H2), but also to
internal hydrogen embrittlement (for materials containing solute hydrogen) since solute hydrogen can diffuse to, and adsorb at
internal crack tips and interfaces (and precipitate as H2 in cracks/voids).
The dimpled intergranular fracture surfaces (plus evidence of high localised strains beneath fracture surfaces) have been ex-
plained by an adsorption-induced dislocation-emission (AIDE)/void-coalescence process. This process involves (i) weakening of
interatomic bonds at crack tips owing to adsorption, (ii) emission of dislocations from crack tips producing crack-opening and
advance, (iii) void nucleation and growth in the plastic zone ahead of cracks, and (iv) preferential growth of the crack towards the
void, resulting in coalescence (Fig. 11). Adsorption at crack tips obviously does not directly affect dislocation activity or void growth
ahead of cracks, but does so indirectly by promoting crack advance by dislocation emission from crack tips. Thus, lower strains are
required for coalescence of cracks with voids compared with inert environments, where dislocation egress from sources in the plastic
zone predominates.
In inert (air) environments, strong surface bonding prevents dislocation emission from crack tips so that crack/void growth/
blunting occurs by egress of dislocations from sources in the plastic zone. Unlike dislocations emitted from crack tips (which produce
crack advance), only a small proportion of dislocations emanating from near-crack-tip sources exactly intersect crack/void tips to
advance the crack/void – with most dislocations egressing behind crack/void tips producing only blunting, or just contributing to the
general strain in the plastic zone. Extensive blunting therefore occurs at crack/void tips so that larger strains are required for
coalescence of cracks and voids than in embrittling environments, and large deep dimples are observed on fracture surfaces (Fig. 12).
The coalescence process also generally involves small voids forming between larger voids (owing to a distribution of particle sizes and

Fig. 10. SEMs showing dimpled intergranular facets produced in a high-strength tempered martensitic steel (tempered at 650 °C) by (a) rapid sub-
critical cracking in liquid mercury, and (b) slow crack growth in hydrogen gas (101 kPa) at 20 °C [2].

335
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 11. Schematic diagram illustrating the Adsorption-Induced Dislocation-Emission (AIDE)/void-coalescence mechanism for sustained-load LME
and HE for some systems, resulting in small, shallow dimples on intergranular fracture surfaces [2]. Dislocation emission from crack tips (owing to
adsorption-induced weakening of interatomic bonds) promotes the coalescence of the crack with voids formed in the plastic zone. Stresses required
for dislocation emission are sufficient for dislocation activity and void formation ahead of cracks. The inset illustrates emission on plane A, and then
on plane B resulting in an increment of crack advance, Δa.

Fig. 12. Schematic diagrams illustrating transgranular ductile crack growth in inert/air environments resulting in large, deep dimples on fracture
surfaces. Dislocation emission from the crack tip does not occur, and only a small proportion of dislocations egressing at crack tips produce an
increment of crack growth, Δa [2].

void-initiation strains) so that small, stretched dimples are observed within larger ones. Ductile crack growth usually occurs normal
to the applied stress-axis (in plane-strain regions) and, hence, is transgranular.
Close similarities between adsorption-induced LME and SCC in aqueous environments have been observed in some materials (for
both intergranular and transgranular crack paths), suggesting that an adsorption-based mechanism is also applicable to SCC in some
cases [3]. Studies of intergranular SCC in pure magnesium are particularly noteworthy in this regard since environmentally assisted
cracking in aqueous environments under dynamic loading occurred at such high crack velocities (v ~10s mm/s) that there would not
have been time for reactions other than dissociation of water molecules and adsorption of hydrogen at crack tips (i.e. no time for
hydrogen diffusion ahead of cracks). For tests that were carried out on Mg bi-crystals first in an aqueous environment and then in dry
air, fracture surfaces examined at low magnifications exhibited a fairly featureless appearance after intergranular cracking in the
aqueous environment, and then a dimpled appearance after intergranular cracking in dry air. At high magnifications, however, the
smooth areas were also dimpled, but on a finer scale than that for fast fracture in air (Fig. 13). Thus, these observations, along with
close similarities between SCC and LME for Mg, support the AIDE/void-coalescence mechanism. Other evidence for an AIDE me-
chanism, and its applicability to some other materials, is discussed elsewhere [2,3]. These references (and references therein) should
be consulted regarding other proposed mechanisms of HE/SCC, and the ongoing controversies surrounding them, which are beyond
the scope of this paper.

3.4. Fatigue striations on intergranular facets

Fatigue striations on intergranular facets (Fig. 14) occur by the same process as usually proposed for striations on transgranular
facets, i.e. plastic blunting/growth during loading, followed by crack-tip re-sharpening by slip behind crack tips during unloading,
such that each striation is produced by one stress cycle (at intermediate ΔK) [36]. Intergranular fatigue sometimes occurs in pre-
ference to transgranular fatigue owing to strain localisation in soft PFZs at grain boundaries, or adsorption of environmental species
preferentially at grain-boundary/surface intersections. Striations generally have a saw-tooth profile with peak matching peak on

336
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 13. SEM of intergranular fracture surface of a Mg bi-crystal cracked first in an aqueous environment (v ~ 10 mm/s) and then in dry air, showing
relatively smooth area produced by environmentally assisted cracking and dimpled area produced in dry air. At high magnifications, (inset) small,
shallow dimples were apparent in the ‘smooth’ area, although some could only be resolved using TEM replica fractography [3].

Fig. 14. SEM of intergranular fracture in a 2024-T4 Al-Cu-Mg alloy component subject to cyclic stresses in an aqueous environment, showing fatigue
striations and etch pits. Black arrows indicate local direction of crack growth.

opposite fracture surfaces, and striation spacings are larger, and profiles are shallower, for corrosion fatigue (for a given ΔK) probably
as a result of the AIDE process (Fig. 15). Striation profiles after corrosion fatigue are sometimes very shallow, especially in high-
strength materials, and striations may not be resolved or may be obscured by corrosion products.

3.5. Crack-arrest markings on intergranular facets

Crack-arrest markings (CAMs) on intergranular fracture surfaces look similar to fatigue striations, but are produced by sustained-

Fig. 15. Schematic diagrams illustrating fatigue crack growth and striation formation involving crack-advance/plastic blunting during crack
opening, and resharpening of the crack tip during unloading (a) in embrittling environments, and (b) in inert/dry-air environments. Nano-voids (not
shown) may be present just ahead of crack tips in both cases. The larger crack-growth increments for a given crack-tip-opening displacement (CTOD)
in embrittling environments occurs as a result of the adsorption-induced dislocation-emission (AIDE) process in many cases [36].

337
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 16. SEM of intergranular fracture produced by SCC of a 7079-T6 Al alloy in moist air, showing crack-arrest markings [37]. The arrows indicate
local directions of crack growth, which may correspond to the crystallography (slip planes) of underlying grains. Some GBPs are also evident.

load cracking modes such as SCC, HE, and LME (Fig. 16) [37–39]. They occur as a result of repetitive sequences of rapid crack jumps,
crack-arrest, and crack blunting and, hence, have a somewhat different profile to fatigue striations. For hydrogen-assisted SCC, the
sequence of events could involve: (i) generation of hydrogen and oxide-film formation at crack tips, (ii) hydrogen diffusion along
grain boundaries, (iii) film rupture and rapid crack growth in regions of high hydrogen concentration, and (iv) crack-arrest and
blunting when regions of low hydrogen concentration are encountered (Fig. 17). Other possible reasons for intermittent crack jumps
and arrest are discussed elsewhere [39]. There is no correlation between crack-growth rates (and K values) and the spacing of CAMs
(typically 0.2–1 μm), and it is the time between crack jumps that increases with decreasing crack-growth rate [37–39].

3.6. Films and deposits on intergranular fracture surfaces

Films and deposits are present on both intergranular and transgranular fracture surfaces after SCC, CF, and LME, but intergranular
cracking is more common for these fracture modes. These films and deposits often obscure features such as nano-dimples, striations,
and CAMs, and can sometimes make it difficult to diagnose the mode of fracture and elucidate mechanisms of cracking. In some cases,
however, the type, thickness, and colour of the films can provide useful information. The film characteristics depend not only on the
material and environment, but whether the film forms after cracking or in the environment within cracks (which is usually oxygen-
depleted and has a different pH compared with the bulk environment).
For high-strength steels, oxide films on fracture surfaces after SCC and CF in mildly corrosive environments are thin, and produce
interference colours, e.g. blue and yellow tints. The thickness of the oxide (and interference colour) depends not only on the time in
the environment, but also on the strain beneath fracture surfaces, since the strain can influence oxide microstructure, and thereby
oxide growth rate. Thus, for variable-amplitude loading, bands of different colours are sometimes observed (Fig. 18). For more
corrosive environments, films are thicker, and black magnetite films (Fe3O4) are formed within cracks. Coloured or black films are
also present on the fracture surfaces of ‘quench cracks’ owing to oxide-film thickening during tempering, providing the cracks are
surface-breaking. If steel fracture surfaces are exposed to humid or aqueous environments after final fracture, then they can go rusty.
For materials such as Al alloys in aqueous environments, thick, hydrated oxide films are often produced during SCC/CF, and sub-
sequent drying and contraction result in ‘mud-crack’ patterns.
For LME failures, films of embrittling metal are present on fracture surfaces as a result of liquid metal flowing into cracks under
capillary action during crack growth. The embrittling metal usually solidifies when components cool, and thin oxide films on the
surface of the embrittling metal films can result in interference colours. For LME of steels by copper, films are either a reddish Cu-
colour or black if the copper has been oxidised sufficiently. When LME is due to mercury, globules of liquid mercury are observed as a
result of ‘de-wetting’ after fracture. If components are only partially cracked and contain solidified embrittling metal, then breaking
open cracks will involve fracture through the embrittling metal as well as through the remaining ligament of the substrate. If the
embrittling metal is ductile, then dimple or vein patterns are observed on fracture surfaces (Fig. 19). Metal films can also be present
on fracture surfaces if components have undetected cracks prior to electroplating (or re-plating) as a result of plating solution
penetrating cracks.

Fig. 17. Schematic diagram illustrating the formation of crack-arrest markings (CAMs) on intergranular fractures, involving rapid crack followed by
blunting.

338
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 18. Macroscopic view of fracture surface of a high-strength steel specimen tested in distilled water showing bands of intergranular SCC and
transgranular fatigue exhibiting interference colours. The arrow indicates the direction of crack growth. Note the unusual curvature of the crack
fronts owing to the side grooves. The rougher texture of intergranular SCC areas compared with the fatigue areas is apparent even at this low
magnification.

Fig. 19. SEM of a high-strength steel specimen partially cracked by LME in Cd at ~350 °C, then by fast fracture in air at 20 °C, showing intergranular
facets with veins of Cd resulting from fracture through solid Cd within the cracks.

4. Case histories involving intergranular fracture and corrosion

4.1. Intergranular failures of Ni-alloy bolts from a helicopter shaft [40]

An analysis of 7 mm diameter bolts manufactured from a precipitation-hardened Ni-based superalloy (Waspaloy) that fractured
(from the first thread root) below specified torque levels showed that the composition (18.8Cr,12Co,3.8Mo,2.9Ti,3.0Al,1.5Fe) and
hardness (~37HRC) conformed to specifications. However, optical microscopy of metallographic sections revealed that grain
boundaries were ‘decorated’ with quite large particles. Fracture surfaces (examined by SEM) were completely intergranular and
covered with dimples, which was not unexpected given the large area-fraction grain-boundary precipitates (probably carbides) that
acted as void-nucleation sites (Fig. 20).
Waspaloy does not generally exhibit grain-boundary particles, and overload fractures occur by transgranular dimple rupture, as
was demonstrated by testing specimens cut from a turbine blade (used in a previous research study) (Fig. 21(a)) [41], confirming
suspicions that the bolt material had been incorrectly heat-treated. Excessive grain-boundary precipitation in age-hardened materials
is usually caused by an excessively slow quench after solution-treatment. However, details of the heat-treatment process could not be
obtained from the material supplier to check this hypothesis. The transgranular dimpled fracture surface for Waspaloy without grain-
boundary precipitation is also of interest because the alloy is strengthened by coherent Ni3(Al,Ti) matrix precipitates that are easily
sheared by dislocations, leading to marked slip planarity (and dislocation pile-ups at grain boundaries) (Fig. 21 (b)) [41]. Thus, these
observations suggest a planar slip mode per se (i.e. without additional microstructural features that promote intergranular fracture)
may not be an important factor promoting intergranular fracture, contrary to popular opinion.
Recommendations to prevent further cracking of bolts included checking whether bolts from shafts in other helicopters were from
the same batch of material as the suspect ones and, if so, to replace them with bolts from a different batch. Checking the micro-
structures and fracture modes of several bolts from other batches to make sure that the material had been correctly heat-treated was
also advised.

339
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 20. (a) Macroscopic view of a helicopter engine shaft showing location of failed bolts, (b) optical micrograph of the Waspaloy microstructure,
showing grain-boundary network of precipitates, and (c) SEM of fracture surface showing dimpled intergranular facets [40].

Fig. 21. (a) SEM of fracture surface of Waspaloy (without precipitation at grain boundaries) showing dimpled transgranular fracture, and (b)
Optical micrograph of polished surface of Waspaloy after deformation showing planar slip bands [41].

4.2. Cracking and corrosion of Al-Mg alloy plate [42]

Al-Mg based alloys, such as 5083 (Al4.5 Mg,0.7Mn) are strengthened primarily by Mg in solid solution, and are commonly used
for construction of ship hulls and other structures where lightweight and good weldability are required. The alloys are given pro-
prietary ‘stabilisation’ heat-treatments to minimise the formation of anodic Al3Mg2 β-phase precipitates at grain boundaries, which
promote intergranular corrosion and SCC [43]. On one occasion, however, material was inadvertently supplied in a ‘sensitised’
condition with a large area-fraction of β-phase at grain boundaries. Numerous failures of structural members of ships and fast ferries
occurred after only short times in service as a result of intergranular corrosion and SCC, and many boats had to be scrapped (costing
the insurers USD 75 million to settle claims!) [43]. Even when the alloys are supplied in a ‘stabilised’ condition, precipitation of β-
phase can occur after welding or after long times in service at temperatures commonly experienced by structures in hot climates, e.g.
20 years at 50 °C, leading to corrosion/SCC (Fig. 22).
The β-phase at grain boundaries occurs as thin plates, as shown on the fracture surface of a sensitised 5083 alloy embrittled by
liquid gallium (Fig. 23), or revealed by thin-foil TEM. Intergranular corrosion and SCC occur primarily by anodic dissolution of the β-
Phase, although SCC can also involve hydrogen embrittlement of areas between the precipitates. Hydrogen is produced by the
cathodic reaction balancing the anodic-dissolution reaction, and diffuses along and segregates at boundaries, thereby weakening
interatomic bonds.

340
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 22. Plate cut from 5083-H112 Al-Mg-Mn alloy ship-plate, where cracking had occurred around a weld (where a guard-rail post had been
attached), then parallel to the surface of the plate near the centre line owing to the elongated grain structure of the plate [42].

Fig. 23. SEM of fracture surface of a sensitised 5083 Al-Mg alloy embrittled by pre-exposure to liquid gallium, showing plate-like Al3Mg2 β-phase
precipitates.

Tests for ‘sensitisation’ after heat-treatments (involving determining the mass-loss of specimens owing to grain detachment after
immersion in concentrated nitric acid) are now required as part of the acceptance criteria. This does not, of course, address sensi-
tisation in service, and novel alloying additions are being investigated (using computer-aided thermodynamic calculations in some
studies) to try to produce alloys with improved resistance to intergranular corrosion/SCC.

4.3. Intergranular corrosion and SCC in aerospace Al alloy structures [3,42]

Precipitation-hardened 2xxx series (Al-Cu-Mg) and 7xxx series (Al-Zn-Mg-Cu) alloys in peak-aged conditions are highly suscep-
tible to intergranular corrosion and SCC, owing to dissolution of anodic grain-boundary precipitates and hydrogen embrittlement [3].
A major difference between these alloys and sensitised 5xxx alloys is that hydrogen effects dominate for SCC, as the precipitates are
more globular so that grain-boundary areas between precipitates are greater. In addition, dissolution of precipitates is not always
involved, as sufficient hydrogen can be generated by the reaction between water molecules and aluminium surfaces. SCC can
therefore occur in air even at low relative humidity, albeit more slowly than in high-humidity air or in aqueous environments.
SCC failures of peak-aged 7075 (and 7079) alloys plagued the aerospace industry in the 1960s, 70s and 80s, but the incidence of
SCC subsequently decreased when overaged (and retrogression and re-ageing) heat-treatments were introduced. These ageing
treatments increase the size and spacing of grain-boundary (and matrix) precipitates, increase precipitate-free zone widths, and
increase the copper contents of MgZn2Cu grain-boundary precipitates. Until recently, there were controversies regarding which factor
(or factors) were responsible for increasing SCC resistance. As mentioned earlier, it is now widely accepted that increasing copper
contents of grain-boundary precipitates is the key factor because precipitates are then less anodic with respect to the adjacent
material, thereby decreasing the rate of generation of embrittling hydrogen [33]. This sensitivity to grain-boundary composition
needs to be considered, and rigorous SCC testing carried out when newer, higher-strength-to-weight ratio, 7xxx series alloys, with
higher zinc contents than in previous alloys, are used in aircraft structures, since grain-boundary precipitates are likely to be more
anodic when zinc contents are higher.
SCC failures are still occurring in ageing aircraft either because peak-aged 7075 alloys are present or peak-aged alloys have
mistakenly been used instead of overaged alloys [42]. For SCC in aqueous environments, fracture surfaces are usually corroded and
exhibit mud-crack patterns. An interesting example, where this was not the case, occurred recently in wing spars of a 40-year-old
military aircraft, where the spars were protected by a (still-intact) thick epoxy coating and paint scheme (Fig. 24(a)) [42]. Such
protective coatings do not, however, prevent diffusion of water molecules through to the underlying Al alloy after long times, and

341
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 24. (a) Macroscopic view of epoxy-coated aircraft wing spar, with position of longitudinal cracks arrowed, and (b) SEM of fracture surface after
breaking open cracks showing crack-arrest markings [42].

numerous stress-corrosion cracks were detected by eddy-current inspection (and subsequent destructive examination). Fracture
surfaces exhibited only superficial corrosion, and CAMs with ~500 nm spacing were clearly evident on intergranular facets
(Fig. 24(b)). These CAMs, associated with hydrogen diffusion ahead of cracks as discussed previously, do not give any indication of
crack-growth rates, and there were no crack-growth-rate data for the particular circumstances. Thus, it was unclear whether the
cracks would pose a structural-integrity risk over the remaining life of the aircraft. A safety-by-inspection regime was therefore
instigated.

4.4. Intergranular fracture of Al-Li alloy aircraft structures [41,44–46]

Precipitation-hardened Al-Li based alloys underwent extensive development in the 1970's and 80's with the aim of replacing
conventional Al-Cu-Mg and Al-Zn-Mg-Cu alloys in order to save weight. For example, the 8090 alloy (Al2.3Li, 1.2Cu,0.7 Mg, 0.1Zr)
(wt%) (which equates to ~10 at.%Li) is about 10% lighter than conventional Al alloys. However, low fracture toughness, associated
with intergranular fracture, limited their use to a few niche applications – one of which was for the structure and skin of a helicopter
(where weight-saving was critical). Several helicopters crashed during service (for reasons not related to poor fracture toughness),
but catastrophic damage occurred during crashes (which was associated with low fracture resistance) (Fig. 25).
Subsequent research showed that the initially low fracture toughness (considered to be just adequate at manufacture) had de-
creased during service owing to additional ageing as a result of high levels of lithium remaining in solid solution and structural
temperatures reaching ~80 °C for cumulative times of at least several months in hot climates. The reasons for the initially low
fracture toughness and its decrease during secondary ageing were subject to considerable research and controversy. Proposed hy-
potheses for the low fracture toughness included (i) planar-slip characteristics owing to the easily sheared, coherent Al3Li δ' age-
hardening precipitates promoting dislocation pile-ups and associated stresses at grain boundaries, (ii) alkali-metal-impurity (AMI)
phases (liquid at 20 °C) at grain boundaries resulting in liquid-metal embrittlement, and (iii) embrittlement due to Li segregation at
grain boundaries.
A number of observations suggested that Li segregation was primarily responsible for brittle intergranular fracture, since the other
hypotheses could be discounted for various reasons. For example, abrupt ductile-to-brittle intergranular fracture transitions with
decreasing temperature were observed in very underaged 8090 alloys (and binary Al-Li alloys), where there were (i) no GBPs or PFZs,

Fig. 25. Macroscopic view of part of the wreckage from a crashed helicopter showing fractures with no signs of plastic deformation, owing to low-
energy intergranular fracture (inset) of the Al-Li-Cu-Mg 8090 alloy used.

342
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 26. Ductile to brittle intergranular fracture transitions with decreasing temperature for (a) an Al-Li-Cu-Mg 8090 alloy underaged for various
times as indicated, and (b) Cu-Sb alloys with varying Sb contents. The ‘y’ axis in (a) is the % Ductile Transgranular Fracture and 0% corresponds to
100% brittle intergranular fracture. Low impact energies in (b) corresponds to 100% brittle intergranular fracture. [46].

(ii) low levels of AMIs (and any phases would be solid above and below the transition temperatures and not embrittling in any case),
and (iii) no significant difference in slip characteristics above and below the transition temperature. In addition, Li segregation has
been detected at grain boundaries (albeit with difficulty owing to the low atomic number of lithium), and is predicted to be em-
brittling based on thermodynamic calculations. No detectable impurity segregation was present in the 8090 alloy but, in other
materials such as Cu with Sb impurities and steels with metalloid impurities, similar ductile-to brittle intergranular fracture tran-
sitions with decreasing temperature have been observed (Fig. 26), supporting a segregation hypothesis for Al-Li alloys.
Segregation would be present above and below the transition temperatures, but it was proposed that segregation-dependent grain-
boundary structural (2D) phase changes occurred at the transition temperatures, which resulted in brittle behaviour. Other ob-
servations indicated that Li segregation was responsible for brittle intergranular fracture in near-peak-aged alloys as well as un-
deraged alloys, and that segregation levels increased during secondary ageing resulting in decreasing fracture toughness. Strain
localisation at grain boundaries owing to the presence of PFZs and GBPs would also make some contribution for near-peak-aged
alloys, but these features are present in other precipitation-hardened Al alloys that have reasonable toughness, so they are not
primarily responsible for low fracture toughness of Al-Li alloys.
Regardless of the underlying mechanisms of embrittlement, a decision was taken to stop using Al-Li alloys for manufacturing the
helicopters, and use conventional Al alloys, with weight-savings achieved in other ways, e.g. lighter avionics. The experience with the
high-Li second-generation Al-Li alloys such as 8090 led to the development of third-generation alloys with lower lithium contents
(generally < 1 wt%) with reasonable toughness, and these alloys are finding gradually increasing application.

4.5. Intergranular fracture of a Cadmium-plated steel screw from a helicopter [47]

Investigation of the wreckage of a helicopter that crashed and caught fire revealed that one of two cadmium-plated steel screws
from a cover-plate on an engine fuel-control unit had fractured, raising the question of whether the screw had failed before crashing
(and possibly caused the crash) as a result of HE or SCC, or fractured during the post-crash fire by LME. (The screws had a higher-
than-specified hardness of ~42HRC making them susceptible to HE/SCC). The fracture surface of the screw was intergranular (along
prior-austenite grain boundaries of the tempered martensitic structure), and was covered by a cadmium film, as was evident from
SEM/EDS analysis, and the blue/yellow tints typical of oxidised cadmium observed optically. Examination of metallographic sections
(before etching) showed that there were networks of secondary intergranular cracks completely filled with cadmium (Fig. 27). These
observations unequivocally show that the screw failed by LME during the post-crash fire, since cadmium from the melted surface
coating would not flow into tight pre-existing cracks, but must have been drawn in by capillary flow during cracking. For the Steel:Cd
system, LME is due to adsorption-induced weakening of interatomic bonds at grain boundaries, and the observation of films of
embrittling metal within secondary intergranular cracks is not evidence that embrittling metal has diffused into materials along grain
boundaries prior to fracture, as sometimes suggested. A requirement for LME to occur is that the embrittling metal must ‘wet’ the
substrate metal, i.e. be in intimate contact without any intervening oxide film. For electroplated coatings, tests showed that this was
not always the case, and this could explain why only one of the screws failed.

4.6. Intergranular fracture of a Cadmium-plated steel aircraft propeller bolt [48]

As for the previous example, fracture surfaces of the cadmium-plated steel bolt were intergranular (along the prior-austenite grain
boundaries of the tempered martensitic structure). Intergranular facets near the origin of cracking exhibited cadmium deposits, but
the majority of the fracture surface was ‘clean’, i.e. did not exhibit any films or corrosion products (Fig. 28). Other relevant

343
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 27. (a) Macroscopic view of fracture surface of failed Cd-plated screw (which cracked at the head-to-shank radius, and intersected a locking-
wire hole at A) showing interference colours typical of oxidised Cd. SEM showed that fractures were intergranular but were obscured in many areas
by a film of Cd. Small areas of dimpled fast fracture were observed at the positions arrowed. (b) Optical micrograph of polished section showing
secondary cracks filled with Cd. Inset shows the EDS analysis confirming the presence of Cd in cracks.

Fig. 28. SEMs of intergranular fracture surfaces of high-strength steel propeller bolt (a) near origin showing Cd deposits, and (b) away from origin
showing ‘clean’ facets.

information included: (i) the bolt was manufactured from a low-alloy steel with a hardness of ~44HRC, (ii) it had been electroplated
with ‘bright’ non-porous cadmium, (iii) the plating had been removed for non-destructive inspection for cracks at some stage, and
then re-plated, and (iv) a standard baking treatment of 23 h at 190 °C had been applied after plating to remove hydrogen which
diffused into the steel during plating.
Based on the above background information, and the fact that high-strength steels with a hardness of > 38HRC are very sus-
ceptible to hydrogen embrittlement (with susceptibility increasing with increasing hardness), it was concluded that cracking was due
to hydrogen embrittlement (HE). The cadmium deposits on intergranular facets near the origin were thought to have arisen because
there had been small, undetected HE cracks present during the re-plating process. The re-plating process would also have led to more
uptake of hydrogen, and the dense cadmium coating would have inhibited egress of hydrogen from the steel during the baking
process. It was therefore recommended that other bolts should be re-plated with a porous cadmium coating which would allow egress
of hydrogen during baking, while still providing sufficient corrosion protection.

4.7. Intergranular fractures in high-strength steel automotive bolts [42]

Failures of automotive rear-axle differential bearing-cap bolts occurred after short times and distances (as low as1000 km) in
service, and sometimes resulted in considerable damage to the differential gears. The bolts (Grade 12.9) had a tempered-martensitic
structure (tempered at 400 °C for 90 min.) with a hardness between 38 and 44 HRC, and had a phosphate coating to provide some
measure of corrosion protection before and during service (in the rear-axle oil environment). Failures usually initiated at the first
thread root (common for threaded fasteners in general), and laps, corrosion pits, and decarburisation were observed at this location,
all of which would have facilitated crack-initiation.
Optical microscopy and SEM examination of fracture surfaces showed that there were (unusually) six distinct zones. First, there
were small areas of transgranular cracking (exhibiting a few progression markings), which then transitioned to corroded

344
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 29. (a) Macroscopic view of one of failed bolts showing initiation sites (arrowed), intergranular areas (I) separated by a region of ductile
fracture (D), and (b) schematic diagram illustrating the various regions in more detail: (i) small transgranular areas near origins, (ii) corroded
intergranular areas, (iii) intergranular regions without significant corrosion, (iv) similar to iii but with an irregular crack front, (v) dimpled
transgranular area, (vi) intergranular facets, and (vii) shear lip.

intergranular facets, and then to intergranular facets that had not suffered from corrosion, and exhibited tear ridges and isolated
dimples. There was then a band of dimpled transgranular fracture followed by another intergranular area, also exhibiting tear ridges,
and finally a transgraular shear fracture (with shear dimples) near the surface diametrically opposed to the initiation site (Figs. 29,
30).
It was concluded, based on the above observations, tests on new bolts, and data in the literature, that crack growth occurred first
by CF (producing the transgranular and initial intergranular areas) then by intergranular SCC (based on a more irregular crack front

Fig. 30. SEMs of (a) regions (i) and (ii), (b) region (iii), (c) region (v), and (d) region (vi), respectively, shown in Fig. 29. Intergranular facets in
region (i) exhibit corrosion pits, and exhibit tear ridges and isolated dimples in regions (ii) and (vii).

345
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

than expected for CF). It was assumed that the oil around the bolts contained traces of water, and it had become acidic over time such
that hydrogen was generated at the steel surface. After cracking by SCC/CF had reached a critical crack length, final failure occurred
initially by ductile tearing, and then by intergranular fracture. This conclusion was based on tests on new bolts containing a saw-cut
about half-way through the bolt and bending them rapidly in air, resulting in some initial ductile tearing then intergranular fracture.
This intergranular cracking must therefore have been the result of temper-embrittlement, owing to the presence of metalloid im-
purities such as P, Sb, Sn, which weaken interatomic bonds across grain boundaries. Temper embrittlement is promoted at high strain
rates, and with high constraint, which would explain why the initial fast fracture occurred by ductile tearing and then transitioned to
intergranular fracture as strain rate increased. The final ligament failed by shear when triaxial constraint decreased.
Another important observation (based on photographs of the bolts before their removal from the component) showed that the
initiation sites were always in the same position with respect to the overall component geometry, suggesting that there was a
consistent misalignment issue that increased cyclic and sustained stresses. Recommendations to avoid failures therefore included (i)
addressing the misalignment issue, (ii) using lower strength bolts (Grade 10.9 with HRC < 37) that are less susceptible to hydrogen-
assisted CF and SCC, and (iii) ensuring bolts were clean and dry before installation.

4.8. Intergranular fracture of α-brass valve from a military aircraft engine [42]

The choice of a free-machining brass (10%Zn2%Pb) for a valve housing, which experienced temperatures up to140oC, is another
example of poor materials selection since cracking should have been anticipated. There are lead particles throughout the micro-
structure which facilitate machining since the particles melt and reduce friction between the tool and workpiece, and result in
fragmentation of machining chips owing to LME. However, the lead particles can result in SMIE (slow, sub-critical intergranular crack
growth) at temperatures well below the melting point of Pb (327 °C) under low sustained loads during service. For the brass valve,
cracking initiated from the inside surface near a crimped rim, where tensile residual stresses would have been present, and then
slowly progressed to about a quarter of the way through the wall thickness before transitioning to transgranular fracture exhibiting
large deep dimples.
The slow, sustained-load cracking involves (i) nucleation of cracks around lead particles at grain boundaries, (ii) surface-diffusion
of lead atoms to crack tips, (iii) adsorption of lead atoms at crack tips, (iv) weakening of substrate interatomic bonds, (v) crack
growth by dislocation emission from crack tips, (vi) linking up of cracks with nanovoids produced by dislocation/grain-boundary
interactions ahead of crack tips, and (vii) coalescence of cracks (with other cracks nucleated from other lead particles) by localised
necking of ligaments between them. Intergranular fracture surfaces therefore exhibited large, flat-bottomed dimples centred on sites
of lead particles (Figs. 31, 32), with nanoscale dimples within the large dimples apparent at high magnifications. The fracture
surfaces would be covered by thin films of lead (perhaps only several atoms thick), but such films are difficult to detect except by
high-resolution techniques such as Auger electron spectroscopy or secondary-ion mass spectroscopy.
SMIE can occur in other materials, such as free-machining steels and free-machining Al alloys, which contain embrittling phases.
The general lesson from the failure of the brass valve is that free-machining alloys (containing low-melting-point embrittling phases)
should not be used when service temperatures exceed about 0.6 of the melting-point (Kelvin units) of the embrittling phase. SMIE
could also occur in materials with coatings such as cadmium if service temperatures were sufficiently high. Concerns have also been
raised about the possibility of SMIE of titanium-alloy components contaminated by cadmium from the use of cadmium-plated tools,
although actual examples have apparently not been reported. The rate-controlling step for SMIE is surface self-diffusion of the
embrittling metal from its source to the crack tip, so that crack growth rates decrease with (i) decreasing temperature, and (ii)
increasing distance from the source to the crack tip.

Fig. 31. SEM of fracture surface a leaded brass valve showing flat-bottomed dimples centred on sites of Pb particles, typical of failure by solid-metal-
induced embrittlement (SMIE) from internal phases. Slip lines produced by deformation of the fracture surface are also evident.

346
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 32. Schematic diagram illustrating processes involved in solid-metal-induced embrittlement (SMIE) resulting in flat-bottomed dimples centred
on sites of Pb particles.

4.9. Intergranular fracture in Nickel-Aluminium-Bronze submarine valves and couplings [49,50]

Nickel-Aluminium-Bronze (NAB) (Cu 9.2Al-4.3Fe-4.0Ni-0.16Mn)(wt%) is commonly used for hydraulically operated valves and
couplings in submarines owing to its good corrosion/SCC resistance. However, numerous failures have been encountered in these
components in air (interior submarine atmosphere) and exposed to sea water and, for the former, no significant corrosion was
observed on exterior surfaces or fracture surfaces. The microstructures of the failed components (machined from extruded bar)
exhibited numerous κ-phase particles (which is usual for NAB), although the size and distribution of these particles and the grain size
was unusually variable from one component to another. However, this variability did not appear to be related to the failures, as
components with all microstructures suffered similar failures. An unusual aspect of the failures was the diversity of crack locations,
and both circumferential and longitudinal cracking were observed in couplings (with some couplings exhibiting four longitudinal
cracks at 90° to each other along with a circumferential crack) (Fig. 33(a)). SEM and TEM of replicas of fracture surfaces revealed
numerous κ-particles on intergranular facets, along with faint striations, (which were more clearly apparent by TEM (inset)
(Fig. 33(b)).
For one of the valves, separation occurred into three pieces (Fig. 34(a)), and fracture surfaces were completely intergranular, and
exhibited faint progression markings (when illuminated obliquely with fluorescent lighting under a binocular microscope). At high
magnification using SEM, a few progression markings were also evident, but intergranular facets more generally exhibited numerous
intersecting slip lines (Fig. 34(b)). There was no signs of dimpled fast fracture or shear lips adjacent to the free surfaces. Testing small
specimens cut from the valve (such that the crack-plane orientation was the same as for the primary fracture) showed that (i) fast
fracture was dimpled (as for the couplings), (ii) fatigue at intermediate ΔK (crack-growth rate > 10−5 mm/cycle) produced
transgranular fractures covered with fatigue striations, and (iii) fatigue at low ΔK (crack-growth rate ~ 2 × 10−6 mm/cycle) pro-
duced mixed intergranular and transgranular fractures, with the former exhibiting numerous slip lines as observed for the failed
valve.
Based on the above observations, and literature indicating that transitions to intergranular cracking could occur for fatigue of
copper-based alloys at very low crack-growth rates, it was concluded that failures occurred by very-high-cycle fatigue (prob-
ably > 1010 cycles) at extremely low stress levels, and that resonant excitation (from an unknown source) produced very high

Fig. 33. (a) Macroscopic view of nickel-aluminium-bronze hydraulic coupling showing fracture surface of circumferential crack which was inter-
sected by four longitudinal cracks (two of which are arrowed) diametrically opposed to each other, and (b) SEM of intergranular fracture surface
showing numerous κ-particles: Inset is a TEM of a replica showing fine striations (arrowed) between particles. Fast fracture in air was transgranular
and dimpled.

347
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

Fig. 34. (a) Macroscopic view of nickel-aluminium-bronze valve from a submarine which (remarkably) separated into three pieces, and (b) SEM of
intergranular fracture surface (observed in all areas) showing slip lines.

frequency (> 100 Hz) cyclic stresses. Occasional higher-than-usual loads would have produced the occasional progression marking or
striation, but cyclic loads generally were too small to produce observable fatigue striations. The slip lines on the intergranular facets
would have been produced by high-cycle fatigue deformation of the surface just behind the crack tip.
Attempts to monitor cyclic stresses during various submarine operations, using strain gauges and accelerometers, aimed at
eliminating the source of the stresses were unsuccessful. It was therefore recommended to use a different material with a higher giga-
cycle fatigue endurance limit than NAB for the components, e.g. a super-austenitic stainless steel, to try to prevent failures. It was also
pointed out that data on giga-cycle fatigue limits were limited, and that any change to another material must consider its galvanic
compatibility with other materials in electrical and electrolyte contact.
Why fatigue at low ΔK in Cu-based alloys (and some other materials) is intergranular is not known. Indeed, the mechanism of
fatigue crack growth in general (regardless of fracture path) is not well understood at low ΔK. Fatigue striations are sometimes
observed on transgranular facets at low ΔK but, unlike at intermediate-to-high ΔK, not every stress cycle produces an increment of
crack growth and striation marking, so the striation spacing is larger than the crack-growth rate. Sometimes it is difficult to dis-
tinguish striations (crack-front markings) from slip markings produced behind the crack tip. For fatigue at low ΔK, microstructural
changes, such as formation of specific dislocation cell structures, may be necessary before an increment of crack growth occurs, and
such microstructural changes may occur preferentially adjacent to grain boundaries.

5. General lessons from the failures described

General procedures and protocols for conducting failure-analysis investigations have been documented in many books/hand-
books) and papers, but some general lessons from the case histories summarised in this paper are worth listing.

1. Do not jump to conclusions, especially when failures are associated with grain boundaries, since a diversity of failure modes (and
causes) can result in intergranular fracture and corrosion. For the failures of the NAB couplings, some failure analysts drew the
incorrect conclusion that SCC had occurred, just because SCC in Cu-based alloys is common and often intergranular.
2. Obtaining all the relevant background information is obviously critical (but often easier said than done), and it is not only major
factors, such as material composition, microstructure, strength, environment, and stresses (including residual and fit-up stresses)
that are important. ‘Secondary’ factors such as porosity of electroplated coatings, and whether coatings had been removed for
inspections and then re-plated, as illustrated for the failure of the cadmium-plated propeller bolt, are also important.
3. Taking photographs of failures and the surroundings before dismantling equipment is important, as illustrated by the failure of the
automotive rear-axle differential bolts where initiation sites were consistently oriented with respect to the overall geometry in the
same way, indicative of misalignment. Photographs also provide documented evidence regarding, for example, whether com-
ponents were missing, such as split-pins or spacers, as has occurred and been implicated as a cause of some major disasters.
4. Fractographic examination is obviously critical, but it is important to have examined the microstructure first to determine, for
example, if there are grain-boundary features, such as a necklace of second-phase particles as for the Waspaloy bolt, that would
promote intergranular fracture or corrosion. Fracture surfaces should also be examined optically before SEM examination to see if
colours are present, e.g. oxide-film interference colours, or copper or gold films. Using oblique fluorescent lighting to determine if
faint progression markings are present, as mentioned for the NAB valve failure, is worth emphasising again.
5. Being able to recognise fracture-surface characteristics indicative of particular failure modes is also obviously vital, as is being
aware that different features can look similar, e.g. fatigue striations, crack-arrest markings produced by SCC, and slip lines
produced by deformation behind crack tips. When in doubt about fractographic features, viewing stereographic pairs of fracto-
graphs, and examining mating areas of opposite fracture surfaces can be useful.
6. Testing specimens cut from failed components is sometimes necessary if information about fracture paths and modes for particular

348
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

materials/microstructures are not available in the literature or from prior experience. When failures are intergranular, it is always
useful to know whether or not fast fractures in air produce intergranular fracture, e.g. as a result of temper embrittlement or
precipitation at grain boundaries. This is easy to do simply by bending a small notched specimen using a vice and a lever. Fatigue
testing is more time-consuming, but was vital for the NAB valve and coupling failures in showing that intergranular cracking only
occurred at low ΔK.
7. Inadvisable material selection was an issue for some of the cases discussed, and people making such decisions need to have a
greater awareness of previous failures for the materials selected. For example, they should know that selecting high-strength
tempered martensitic steels with strength levels above ~1200 MPa (38 HRC) greatly increases the risks of failure by HE/SCC/CF,
as illustrated by the failures of the propeller and rear-axle bolts. Another example of ‘asking for trouble’ is to use free-machining
alloys (with low-melting-point phases) in applications where temperatures are above about half the melting-point (degrees
Kelvin) of the phases, as illustrated by the failure of the leaded-brass valve. Material selection should also consider whether or not
deleterious changes in fracture or corrosion resistance may occur in service, as illustrated by the case involving poor crash-
worthiness of Al-Li based alloys used in helicopters, and Al-Mg-based alloys used in ships.
8. Finally, it is worth emphasising again that having a sound understanding of the mechanisms of all types of fracture and corrosion
and, for intergranular failures, a knowledge of grain-boundary fundamentals, are essential for analysing failures and making
recommendations to prevent further failures.

Acknowledgements

Some of the failures discussed in the present paper were investigated by a number of former colleagues at Defence Science and
Technology Organisation as well as myself. Rohan Byrnes and Darren Edwards merit a special mention in this regard.

References

[1] ASM Metals Handbook, Failure Analysis and Prevention, Vol. 11 ASM International, Ohio, USA, 2002.
[2] S.P. Lynch, Hydrogen Embrittlement Phenomena and Mechanisms, Chapter 2, in: V.S. Raja, T. Shoji (Eds.), Stress Corrosion Cracking: Theory and Practice,
Woodhead Publ., Cambridge, UK, 2011, pp. 90–130 and references therein. (Republished in Corrosion Reviews, 30, 105–123, 2012).
[3] S.P. Lynch, Mechanistic and fractographic aspects of stress-corrosion cracking, Chapter 1, in: V.S. Raja, T. Shoji (Eds.), Stress Corrosion Cracking: Theory and
Practice, Woodhead Publ., Cambridge, UK, 2011, pp. 3–89 and references therein. Republished in Corrosion Reviews, 30, 63–104, 2012.
[4] S.P. Lynch, Failures of structures and components by metal-induced embrittlement, J. Fail. Anal. Prev. 8 (2008) 259–274.
[5] H. Reidel, Fracture at High Temperatures, Springer Verlag, Berlin-Heidelberg-GmbH, 1987.
[6] A.P. Sutton, R.W. Balluffi, Interfaces in Crystalline Solids, Oxford University Press, 1995.
[7] B.C. Muddle (Ed.), Interfaces II, Mater. Sci. Forum (1995) 189–190.
[8] C.L. Briant, On the chemistry of grain boundary segregation and grain boundary fracture, Metall. Trans. A. 21A (1990) 2339–2354.
[9] C.L. Briant, Grain boundary structure, chemistry, and failure, Mater. Sci. & Tech. 17 (2001) 1317–1323.
[10] P. Lejcek, M. Sob, V. Paidar, Interfacial segregation and grain boundary embrittlement: an overview and critical assessment of experimental data and calculated
results, Prog. Mater. Sci. 87 (2016) 83–139.
[11] V. Randle, Grain boundary engineering: an overview after 25 years, Mater. Sci. & Tech. 26 (2010) 253–261.
[12] T. Watanabe, Grain boundary engineering: historical perspective and future prospects, J. Mater. Sci. 46 (2011) 4095–4115.
[13] K. Sickafus, S. Sass, Grain boundary structural transformations induced by solute segregation, Acta Metall. 35 (1986) 69–79.
[14] C. Rottman, Theory of phase transitions at internal interfaces, J. Phys. 49 (1988) C5-313–C5-326 Colloque C5.
[15] E.I. Rabkin, L.S. Shvindlerman, B.B. Straumal, Grain boundaries: phase transitions and critical phenomena, Int. J. Mod. Phys. B 5 (1991) 2989–3028.
[16] T.E. Hsieh, R.W. Balluffi, Observations of roughening/de-faceting phase transitions in grain boundaries, Acta Metall. 37 (1989) 2133–2139.
[17] M. Aramfard, C. Deng, Mechanically enhanced grain boundary structural phase transformation in Cu, Acta Mater. 146 (2018) 304–313.
[18] D.P. Field, H. Weiland, K. Kunze, Intergranular cracking in Aluminum alloys, Can. Metall. Q. 34 (1995) 203–210.
[19] C.L. Briant, N. Lewis, Effect of tempering on fracture mode in high-strength phosphorus-doped Ni-Cr steels, Mater. Sci. & Tech. 2 (1986) 34–41.
[20] J. Olefjord, Temper embrittlement, Int. Metals Rev. 23 (2013) 149–163.
[21] C.A. Hippsley, H. Rauh, R. Bullough, Stress-driven solute enrichment of crack tips during low-ductility intergranular fracture of low-alloy steel, Acta Metal. 32
(1984) 1381–1394.
[22] J. Shin, C.J. McMahon, Mechanisms of stress relief cracking in a ferritic steel, Acta Metal. 32 (1984) 1535–1552.
[23] E.V. Barrera, M. Menyhard, D. Bika, B. Rothman, C.J. McMahon, Quasi-static intergranular cracking in a Cu-Sn alloy: an analogy of stress relief cracking of steels,
Scripta Metall. Mater. 27 (1992) 205–210.
[24] K. Yoshino, C.J. McMahon Jr., The cooperative relation between temper embrittlement and hydrogen embrittlement in a high strength steel, Metall. Trans. A. 5
(1974) 363–370.
[25] R.H. Jones, A review of combined impurity segregation-hydrogen embrittlement processes, in: R.M. Latanision, T.E. Fisher (Eds.), Advances in the Mechanics
and Physics of Surfaces, 3 1986, pp. 1–70.
[26] T.C. Lee, I.M. Robertson, H.K. Birnbaum, TEM in situ deformation study of the interaction of lattice dislocations with grain boundaries in metals, Philos. Mag. A
62 (1990) 131–153.
[27] L. Priester, “Dislocation-interface” interaction – stress accommodation processes at interfaces, Mater. Sci. & Engng A301-310 (2001) 430–439.
[28] Z. Shen, R. Wagoner, W. Clark, Dislocation and grain boundary interactions in metals, Acta Metall. 36 (1988) 3231–3242.
[29] K.S. Kumar, H. Van Swygenhoven, S. Suresh, Mechanical behaviour of nanocrystalline metals and alloys, Acta Mater. 51 (2003) 5743–5774.
[30] S.P. Lynch, Mechanisms of Intergranular Fracture, in: G. Was, S. Bruemmer (Eds.), Volume on Grain Boundary Chemistry and Intergranular Fracture, 46 Trans.
Tech. Publ., Materials Science Forum, 1989, pp. 1–24.
[31] S.P. Lynch, Ductile and Brittle Crack Growth: Fractography, Mechanisms and Criteria, Materials Forum, 1988 Special Issue, Vol. 11 (1988), pp. 268–283
Republished in “Engineering Materials 2000”, Ed. J.D. Watson, Trans. Tech. Publ., Zurich, 1988.
[32] G. Henry, J. Plateau, La Microfractographie, Editions Metaux, France, 1968.
[33] S.P. Knight, N. Birbilis, B.C. Muddle, A.R. Trueman, S.P. Lynch, Correlations between intergranular stress corrosion cracking, grain-boundary microchemistry,
and grain-boundary electrochemistry, Corros. Sci. 52 (2010) 4073–4080.
[34] A.K. Vasudévan, R.D. Doherty, Grain boundary ductile fracture in precipitation hardened aluminium alloys, Acta Metall. 35 (1988) 1193–1219.
[35] P. Unwin, G. Lorimer, R. Nicholson, The origin of the grain boundary precipitate free zone, Acta Metall. 17 (1969) 1363–1377.
[36] S. Lynch, Some fractographic contributions to understanding fatigue crack growth, I. J. Fatigue 104 (2017) 12–26 (and references therein).
[37] S.P. Knight, Stress Corrosion Cracking of Al-Zn-Mg-Cu Alloys, Ph.D Thesis Monash University, 2008.

349
S. Lynch Engineering Failure Analysis 100 (2019) 329–350

[38] G.M. Scamans, Evidence for crack-arrest markings on intergranular stress corrosion fracture surfaces in Al-Zn-Mg alloys, Metall. Trans. A. 11A (1980) 846–850.
[39] S.P. Lynch, Progression markings, striations, and crack-arrest markings on fracture surfaces, Mater. Sci. Eng. A A468-470 (2007) 74–80.
[40] Defence Science and Technology Organisation, Melbourne, Australia. Unpublished work.
[41] T. Pasang, N. Symonds, S. Moutsos, R.J.H. Wanhill, S.P. Lynch, Low-energy intergranular fracture in Al-Li alloys, Eng. Fail. Anal. 22 (2012) 166–178.
[42] S. Lynch, Failure of metallic components involving environmental degradation and material selection issues, Corr. Rev. 35 (2017) 191–204.
[43] N.J. Holroyd, G.M. Scamans, Environmental degradation of marine aluminum alloys – past, present, and future, Corrosion 72 (2016) 136–143 (and other papers
in this issue).
[44] S.P. Lynch, R.J.H. Wanhill, R.T. Byrnes, G.H. Gray, Fracture Toughness and Fracture Modes of Aerospace Aluminium-Lithium Alloys, in: N. Eswara Prasad, Amol
A. Gokhale, R.J.H. Wanhill (Eds.), Aluminium-Lithium Alloys: Processing, Properties, and Applications, Elsevier, 2014, pp. 415–455.
[45] S.P. Lynch, B.C. Muddle, T. Pasang, Ductile-to-brittle fracture transitions in 8090 Al-Li alloys, Acta Mater. 49 (2001) 2863–2874.
[46] S.P. Lynch, B.C. Muddle, T. Pasang, Mechanisms of brittle Intergranular fracture in Al-Li alloys and comparison with other alloys, Philos. Magazine 82 (2002)
3361–3373 (and references therein).
[47] S.P. Lynch, D.P. Edwards, A. Crosky, Failure of a screw in a helicopter fuel-control unit: was it the cause of a fatal crash? J. Fail. Anal. Prev. 4 (2004) 39–49.
[48] S.P. Lynch, Failure of structures and components by environmentally assisted cracking, Eng. Fail. Anal. 1 (1994) 77–90.
[49] S.P. Lynch, D.P. Edwards, R.B. Nethercott, J.L. Davidson, Failure of Nickel-Aluminum-Bronze hydraulic couplings, Pract. Fail. Anal. 2 (2002) 50–61.
[50] S.P. Lynch, An unusual failure of a Nickel-Aluminium-Bronze (NAB) hydraulic valve, Eng. Fail. Anal. 49 (2015) 122–136.

350

You might also like