Marcus P. Corrosion mechanisms in theory and practice

Подождите немного. Документ загружается.

age very rapidly when out of contact with the solution, losing the ability to trigger

cleavage in thin foils (Fig. 30).

Cracking of silver alloys has been reported by Galvele and others in a series

of papers [139–141]. Interesting SCC phenomena were reported in both aqueous

and gaseous environments, including halogen vapor. Clearly, all these systems

involve selective reaction of the silver in the alloy, which in halide solutions or

atmospheres generates a composite of silver halide and nearly pure noble metal

(Au or Pd); this layer should behave mechanically much like a dealloyed layer

formed under conditions of free silver dissolution, so the mere occurrence of SCC

does not help to distinguish between SCC mechanisms. Foil-breaking experiments

[33,46–51] should be performed in some of these environments to test the generality

of the film-induced intergranular fracture mechanism.

Type D SCC involves surface films that are neither oxides nor dealloyed layers.

The cracking of high-strength 4340 steel in dry Cl

[102,103] or dissociated N

[142]

432 Newman

Figure 31 Continued

appears to be a film-induced process; Sieradzki [103] proposed that FeCl

acted

as a stiff layer, causing a ductile-to-brittle transition due to the modulus mismatch

with the substrate. Later the lattice parameter was emphasized as a casual factor

[143]. A detailed rationalization of SCC was made for iron exposed to anhydrous

ammonia-methanol, where anodic oxidation of ammonia leads to interstitial

penetration of nitrogen [69,100]. According to the film-induced cleavage model,

such very thin, brittle layers are sufficient to allow cleavage through several μm of

a body-centered cubic (bcc) material, whereas in fcc systems a nanoporous metallic

layer 10–100 nm thick is a specific requirement as it is the only way of epitaxially

coupling a brittle reaction product of sufficient thickness to the fcc substrate. A

special case may be the SCC of pure copper or silver, where a micropitted or

tunneled zone is a possible brittle layer that could nucleate cracking [144,145].

There are several other film-induced SCC processes, such as cracking of Zr

alloys in gaseous iodine [101]. In no case has monolayer adsorption been validated as

the cause of SCC—every known case involves a 3D film. However, liquid metal

embrittlement often occurs, even in fcc systems, without alloy or compound

formation [15]. Possibly there is a regime of very rapid adsorption-induced cracking

that is rarely accessed in aqueous systems as it requires the adsorption process

to be faster than plastic relaxation processes at the crack tip; hence in aqueous

environments this is likely to occur only in very reactive metals (Al, Ti, Mg).

Type E SCC (embrittlement of steels by cathodic hydrogen in the active

state) is quite well understood except for the atomistic action of the hydrogen, which

is still debatable. Briefly, hydrogen atoms are produced by water reduction during

corrosion in neutral solutions, and some of them enter the steel, especially if their

recombination to H

is poisoned by adsorption of S, As, P, or Sb. Once in the steel, the

Stress-Corrosion Cracking Mechanisms 433

Figure 32 Measurements of double-layer capacitance (proportional to pore surface area;

inversely proportional to pore radius) as a function of time during potentiostatic aging of

dealloyed layers on Au-Ag, showing the effects of potential and chloride ions [48].

hydrogen can move with an effective diffusivity (D

eff

) of 10

–8

to 10

–4

/s,

depending on the microstructure as well as temperature [21] (e.g., small carbides

in tempered martensite act as traps and greatly reduce (D

eff

). The lattice (interstitial)

hydrogen solubility is elevated ahead of stressed cracks or notches, and these sites

accumulate higher hydrogen concentrations. The same hydrostatic tension that

increases the H solubility in the lattice also enhances hydride phase precipitation in

434 Newman

Figure 33 (a) Crevice potential and pH and (b) hydrogen permeation rate, as a

function of potential for carbon steel crevices in NaCl solution. (From Ref. 154.

Courtesy of HMSO.)

metals such as Ti and Nb [101]. Hydrogen segregrates to interfaces, including

grain boundaries, and may weaken these in combination with other segregants

such as P and Sb [146]. Once it reaches its site of action, and provided it is in a strong

microstructure, the hydrogen causes local fracture by cleavage [147], intergranular

separation [146], or enhanced microplasticity [148], or some combination of these

processes. In low-strength steels, dynamic loading is required for cracking. If the

input fugacity of hydrogen is high enough (hundreds of atmospheres in media such

as acidic H

S), recombination at internal sites such as nonmetallic inclusions may

cause blistering or cracking (hydrogen-induced cracking, or HIC). This is minimized

in pipeline steels by careful microstructural control.

The altered solution chemistry within cracks is an important aspect of SCC

and corrosion fatigue in structural and high-strength steels. Turnbull [149–151]

has shown that crack acidification, as proposed by Brown [152], is not the norm

for steels corroding freely in NaCl solutions or seawater. Only in Cr-containing

steels with very short cracks could any acidification be predicted [153]. Possibly some

of Brown’s historic measurements were flawed by oxidation of Fe

by atmospheric

oxygen; more recent work shows that deep cracks normally become net cathodes

and reach a pH of about 9 [154] (Fig. 33). Nevertheless, with assistance from the

IR-induced isolation of the crack enclave, the potential can remain below that of a

reversible hydrogen electrode, and in sufficiently strong steels embrittlement will

occur above some K

ISCC

value. An excellent correlation exists between hydrogen

uptake on the walls of simulated cracks and SCC of high-strength steels in the

same solution [154] (Fig. 34). Cathodic protection, within limits, can reduce the

hydrogen uptake from cracks by increasing the local pH [149–151]; however, this

is rarely used for high-strength steels that are sensitive to hydrogen, even though

major benefits can be demonstrated in the laboratory. The problem in practice is

that most cathodic protection systems locally reduce the potential to values that

are dangerously low for high-strength steels.

Stress-Corrosion Cracking Mechanisms 435

Figure 34 Band of results showing the effect of potential on the rate of SCC, for a range

of high-strength steels in seawater [152]; compare with Figure 33b.

Hydrogen effects often show the same kind of strain rate sensitivity as

slip-dissolution processes. Low-strength steels are immune to SCC under static

loads in salt water, but steels of all strengths can suffer hydrogen-assisted fatigue

crack growth [105,155]. High strength continues to be detrimental, but relatively

less so than under static loading.

STRESS-CORROSION TESTING IN RELATION TO

MECHANISMS OF CRACKING

Because standard methods of SCC testing have been excellently reviewed by

Sedriks [156], we focus on the implications of mechanisms for testing and vice versa.

To rationalize or predict service performance, and to test models of SCC, the

slow strain rate test [26,27,113,157] is convenient as it enables a large number of tests

to be conducted at a large number of potentials within a defined time; the maximum

duration is simply the failure time in air (Fig. 35). The maximum load, elongation to

failure, and percent reduction in area at fracture are measured in the test environment

and normalized to values measured in an inert environment. The disadvantages are

that the mechanical condition is relatively undefined when there are multiple cracks,

the crack velocity cannot be measured continuously, and failures occur in materials

that would never fail by SCC in service. An elastic slow strain rate or ultraslow cyclic

test [67] is a useful compromise that maintains the dynamic loading without gross

plastic straining (Fig. 36), but these tests are difficult to carry out in large numbers. It

has been suggested that slow strain rate and cyclic-loading tests are part of a

continuum [28–30,113,158] and that in some systems SCC and corrosion fatigue are

also a continuum with a single mechanism (Fig. 37); certainly the fractography of

slow strain rate SCC and low-frequency fatigue crack growth can be very similar,

e.g., in low-strength steels. The most notorious cases in which the slow strain rate test

overestimates the susceptibility of a material are the hydrogen-induced cracking of

low-strength steels under cathodic protection [159], the SCC of commercial-purity

titanium in salt water [160], and the hydrogen-induced cracking of duplex stainless

steels [161]. In every case the K

ISCC

value, if it exists, is extremely high, at least

50 MPa m

1/2

. Sometimes this hydrogen-induced cracking, which would not occur in

practice, overshadows a “genuine”-SCC phenomenon that occurs at a lower velocity

but has a much lower K

ISCC

value, e.g., SCC of carbon steel in CO-CO

-H

solutions [162,163]. One approach to this kind of problem (without changing the test)

is to study the distribution of secondary cracks on the failed tensile specimen.

Cathodic hydrogen embrittlement is confined to the necked region [159,164], but the

genuine SCC is distributed as secondary cracks along the whole length of the tensile

specimen [162] (Figs. 38 and 39).

Having classified CO

-induced cracking as possibly an artifact, we note that

this is a favored mechanism for a transgranular cracking phenomenon seen in

high-pressure gas transmission pipelines. Clearly, the hydrogen uptake in CO

NaHCO

solution must be high compared with cathodic protection in salt water,

where similar steels do not crack, or else the coexistence of localized corrosion

and hydrogen entry helps to maintain the crack tip strain rate. Dynamic service

stresses are also an important factor.

436 Newman

Having reviewed some of the flaws of slow strain rate testing, we must stress

that this is an outstanding useful test. It is always desirable to have the slow strain

rate information, even if it is not used for direct prediction of service performance.

Neither the pipe cracking in boiling water reactors [59] nor the SCC of high-pressure

gas transmission lines [27,67] (both type A systems) could have been predicted by

static-load tests on smooth specimens.

Loading rate, or dK

/dt, affects SCC initiation and growth in precracked

specimens [113] (Fig. 37) and a good appreciation of the cracking mechanism is

Stress-Corrosion Cracking Mechanisms 437

Figure 35 Typical fractures of C-Mn steel obtained in slow strain rate tests: (a) in air;

(b) in anhydrous ammonia-methanol. (Courtesy of I. M. Hannah and W. Zheng.)

essential to avoid surprises in service, as not all variants of K

(t) can be tested in

the laboratory. Attempts to rationalize all such data with the slip-dissolution model

(for high-temperature aqueous environments) have been more in the nature of a

sophisticated fitting exercise than a first-principles approach [29], and there is a

great divergence of views on the cracking mechanisms, especially in pressurized

water reactor (PWR) pressure vessel steels, where everything from slip dissolution

[28–30] to hydrogen embrittlement [165] to film-induced cleavage [42] to surface

mobility [166] has been suggested.

438 Newman

Figure 36 Various types of elastic slow strain rate test.

Figure 37 The effect of crack tip strain rate, via loading rate, on crack growth in a

high-strength aluminum alloy [54,158]. (Courtesy of NACE.)

In some systems, SCC is initiated at a very large number of sites, and the

crack tip strain rate is greatly reduced at each individual site, leading to near

arrest of the cracks (another reason for crack arrest might be discontinuous

segregation or precipitation; recall Fig. 13). Coalescence of two or more cracks is

then required to produce a dominant crack that causes final failure [27,167]

(Fig. 40). Statistical physicists are very interested in similar problems [168],

ranging from mud cracking to fracture of random composite media. Multiple crack

interactions are a major growth area in fracture mechanics. Such phenomena are

peculiar to smooth specimens and provide a strong argument for carrying out several

kinds of laboratory test.

A major area of development in SCC testing is the role of crevices in crack

initiation. We have discussed the essential role of localized corrosion in Cl-SCC of

austenitic stainless steels [68,111,112], and topical problems with duplex stainless

steels necessitate extension of the same approach to these more resistant materials.

The incorporation of crevices into slow strain rate specimens is not at all trivial, as

we discovered during 2 years’ work by Suleiman at UMIST [169]. Tamaki et al. [68]

showed that a fine notch or precrack could be used to establish a reproducible

condition in a fracture mechanics specimen, but this too is difficult work, especially

if one attempts to carry out the work potentiostatically. Tamaki et al. found that,

for 316L stainless steel at 80°C, SCC sometimes occurred only within a few tens of

mV above the repassivation potential of the crevice (E

) (Figs. 25 and 26). Without

accurate prior knowledge of E

, or in a complex service environment, such work can

be very time consuming, so we have proposed a galvanostatic variant in which a

low anodic current is applied that automatically establishes a steady potential just

above the lowest possible value of E

before starting the tensile machine (see Fig. 7c).

This galvanostatic technique is also applicable to the rapid screening test used

by Shinohara and colleagues [112], in which two sheets of austenitic steel are spot

Stress-Corrosion Cracking Mechanisms 439

Figure 38 Stress-strain curves for “real” SCC (cracks initiated soon after yield) and

cathodic hydrogen embrittlement (where cracking is confined to the necked region), for

low-strength steels.

welded together and polarized in a chloride solution. All these procedures are

recommended for further development except perhaps the use of the creviced slow

strain rate test.

NOTES ON OTHER PROPOSED SCC MECHANISMS

Magnin’s model of SCC and corrosion fatigue [16,17] proposes that the motion of

crack-tip dislocations is impeded by an obstacle ahead of the crack tip, which in

440 Newman

Figure 39 Fractography corresponding to Figure 38 [162], for C-Mn steel in CO-CO

-H

solution: (a) genuine SCC; (b) cathodic hydrogen embrittlement. (Courtesy of I. M. Hannah.)

the purest case might be a Lomer-Cottrell lock. Dissolution down a slip band toward

the obstacle leads to the achievement of K

at the obstacle and the propagation

of a cleavage crack on the slip plane back toward the crack tip; in addition (or

alternatively), this decohesion of the microfacet can be facilitated by hydrogen

adsorption on the fracture plane (Fig. 41). Such a model is consistent with certain

observations in corrosion fatigue cracking, and with fractographic studies of

Dickson and others [66,170] on transgranular SCC surfaces, but is not yet adapted

to explain the multitude of special environmental factors in SCC. Application of

this model to alloy 600 in hydrogenated high-temperature water [17] might be

considered arbitrary in comparison with Scott and Le Calvar’s [60] highly focused

investigations of intergranular oxidation. Magnin’s model is elegant and plausible

but can be tested only by fractography, which is bound to be inconclusive.

The model of environmental cracking proposed by Lynch [19] will be

considered only briefly as it accounts for none of the special environmental effects

in SCC. This does not mean that the model is wrong, only that it is untestable

except by microscopic means. The concept of enhanced local plasticity leading to

plastic microfracture is an important aspect of hydrogen embrittlement [148] and

has interesting implications in liquid metal embrittlement. Most authorities reject

Lynch’s adsorption-based approach to hydrogen effects preferring to appeal to

effects of internal hydrogen on plasticity, yet Lynch has a powerful argument

based on the similarity between liquid metal–induced and hydrogen-induced

fractures in certain systems. There is no doubt that plastic microfracture will form

part of the complete SCC spectrum, should this ever be elucidated, but we must

reject the notion that all SCC can be explained by such mechanisms.

“Universal” models of environmental fracture rarely survive for long, and SCC

specialists such as Parkins [27,171] have devoted much effort to promoting a

spectrum of mechanisms from mainly chemical (intergranular slip-dissolution) to

mainly mechanical (hydrogen-induced SCC of high-strength steel). Nevertheless,

Stress-Corrosion Cracking Mechanisms 441

Figure 40 Crack coalescence of {110} microcracks leading to a dominant slip-band

crack for a Cu monocrystal dynamically strained in NaNO

solution [65].