Stress corrosion cracking

Corrosion Science, Vol. 20, pp. 997 to 1016 Pergamon Press Lid, 1980. Printed in Great Britain.

THE INTERACTION OF STRAIN-RATE AND REPASSIVATION RATE IN STRESS CORROSION CRACK

PROPAGATION*

J. C. SCULLY

Department of Metallurgy, The University of Leeds, Leeds LS2 9JT

Abstract--The events occurring at the tip of a propagating stress corrosion crack are analysed with respect to the strain-rate, ~, and repassivation rate, ?, together with the charge required to cause a crack increment. It is argued that the strain-rate and repassivation rate interact to maintain an acidified solution at the tip of a propagating crack in halide solutions just below the pitting potential. If ~ is too small or ? too large, repassivation will occur, resulting in crack arrest. It is further argued that the morphology of cracking in active path cracking is dependent upon the ratio, ~/?. In a-brass and mild steel, lowering ? causes a partial change in fracture mode from intergranular to transgranular. In austenitic stainless steels, raising ? promotes a partial change of the reverse kind. The ratio ~/~ also determines crack velocity for a particular solution. Cracking morphology in systems exhibiting hydrogen embrittlement cracking may not show a change as ? is altered since not all such systems exhibit an alternative active path mode.

I N T R O D U C T I O N

IN 1967, the au tho r first put fo rward the concept that repass ivat ion was a cri t ical fac tor in the p r o p a g a t i o n o f stress cor ros ion cracks. Fo l lowing on direct ly f rom the work o f L o g a n ) the a rgument was advanced tha t the crucial event in the cycle o f events p roduc ing an increment o f crack growth was a cri t ical delay in the repass ivat ion t ime. I f this t ime was too short , insufficient cor ros ion would occur to cont inue crack growth, whereas, if it was too long, too much corros ion would occur, resulting, morpho - logically, in an e longated fissure or pi t ra ther than a na r row crack. Such ideas had come direct ly f rom exper iments 3 at var ious cons tant c rosshead speeds on a T i - 5 A I - 2.5Sn al loy in an aqueous NaC1 solution. These had shown clearly that repass ivat ion contro l led the crack growth process. A t this stage 1 the essential dynamic interact ion of quite separate processes was emphas ized : a creep process p roduc ing fresh metal area, and a repass ivat ion process, together a l lowing crack p ropaga t ion to occur only if repass ivat ion was " inadequa te" . The p ropaga t ion process was considered to occur as the result o f a specific re la t ionship between the two processes.

In 1968, an a t t empt was made 4 to indicate some o f the e lectrochemical aspects o f stress cor ros ion cracking with respect to po la r iza t ion curves, with a view to i l lustrat ing the possible connect ion between al loying and repassivat ion. The impor tance of the rat io o f two different types o f ions at the crack t ip was also stressed. These could be divided into two general classes: act ivat ing and passivat ing, i.e. p r o m o t i n g cracking and inhibi t ing cracking.

Since that t ime the impor tance o f repass ivat ion has been stressed by many workers . Staehle, 5 Vermilyea, 8 Parkins , 7 Engell, 8 Bignold 9 and Beck 1° have all included it in model l ing var ious stress cor ros ion systems o f stainless steels, t i t an ium and magnes ium

*Manuscript received 17 October 1979. 997

998 J . C . SCULLY

alloys and low alloy steels. Some differences have arisen which may in some cases arise from questions of definition, whereas at least one appears to be real, bearing in mind the nature and difficulty of modelling stress corrosion crack propagation processes. The original paper, I for example, implied that cracking stopped when repassivation was complete, as described above, when the crack tip strain-rate in a constant load test had fallen to a very low value. The lower the repassivation rate the lower would be the strain-rate at which crack arrest occurred. In corrosive environ- ments in which repassivation could not occur, instead of crack arrest, cracking would give way to whatever form of corrosion occurred in the absence of stress, as is discussed later. Where crack arrest did occur as a result of repassivation the value of the crack tip strain-rate was designated ~r. 11 For any specimen it was proposed that any threshold stress for stress corrosion cracking, or K~scc value, corresponded to the value of the crack tip strain-rate being ~r. From such considerations, film repair could prevent crack growth on a creeping crack tip surface. Such an idea, for which there is much experimental evidence, is not in agreement with the model of Vermilyea, n who con- siders that crack growth restarts after repassivation when the strain in the surface oxide becomes so great that it fractures. This point is discussed in detail below.

The physical significance of events giving rise to E r was not considered. 11 Whether, for example, the film was ductile and extended and perhaps thickened so that it never fractured to such a depth as to reveal bare metal at such low strain-rates was not discussed. Instead, abundant data were provided that such a concept was realistic whatever might be the precise physical reality of such an occurrence. In a Ti-O alloy, for example, a2 in 3 ~ aqueous NaC1 solution, cracking occurred at a crosshead speed of 8.3 ~tm/s but not at 3.3 l~m/s in SEN specimens. If, after crack initiation at the higher crosshead speed, the crosshead speed was reduced to the lower value, crack arrest occurred. Continued straining at the lower crosshead speed resulted in eventual ductile failure. Stress corrosion cracking was never re-initiated. Similar events of cracks arresting in specimens at low crack tip strain-rates have subsequently been observed in a Mg alloy 7 in a CI-/CrO~ 2- solution, in 70Cu-30Zn brass in a neutral ammoniacal solution in and in Type 304 austenitic stainless steel 14 in MgCI~ solution boiling at 154°C. For these latter two examples the conditions under which ~, can be reached so that cracking is arrested in a straining specimen are such that the crack tip strain-rate is extremely low, a consequence of the low threshold stress values in these alloys. Given the condition, however, that a passivation or filming event is required to observe ~, then any such system where film formation occurs should be capable of exhibiting such an arrest phenomenon.

The importance of repassivation was taken further in a later paper 11 which put forward the Constant Charge Criterion. This paper developed the argument that an increment of crack growth by an active path mechanism occurred when a minimum charge, Q,,i,, passed in time, t, following the initial current transient, i, occurring

t upon film rupture, corresponding to Q = ~ i.dt and depicted by the hatched area

in Fig. 1 which is reproduced from Ref. 11'. These ideas coincided with those of Newman 15 who had put the principal idea to use in his investigation of stress corrosion cracking in a Cr-Mo steel exposed to 8M NaOH at 100°C. He had concluded that a constant charge mechanism fitted his analysis of cracking in that system.

Interaction of strain-rate and repassivation rate 999

k, t i m e

FIG. 1. A schematic drawing of the current transient following the fracture of a film under potentiostatic conditions. The hatched area represents the charge that passes

between the fracture of the film and its complete repair, xl

In the discussion below the interaction of a deforming surface and the repassivation event consequent upon film rupture is discussed further, particularly in relationship to the maintenance of a volume of liquid at the crack tip acidified to the necessary condition by hydrolysis. Possible consequences of the relationship between crack tip strain-rate, ~r, and repassivation rate, ~, are then discussed, followed by an analysis of how the chemical composition of the liquid at the crack tip determines the crack morphology in v.-brass, austenitic stainless steers, titanium alloys and zirconium alloys.

THE MAINTENANCE OF THE CRACK TiP SOLUTION COMPOSITION

Galvele 16 and co-workers 17 have emphasized, both from theoretical considerations 1~ and from experimental results, ~7 that incipient pits are stabilized and grow when local conditions are met such that the required amount of hydrolysis can occur at the pit surface at such a rate that it is self-sustaining. The continuation of pit growth depends upon the maintenance of the hydrolyzed volume of solution.

Over the last 10-15 yrs, various workers, particularly Brown, la have demonstrated that the pH of the solutions at the tips of propagating cracks in A1 and Ti alloys and in high strength steels is low and, in each case, it is in the same range of pH values as that found inside pits in the same materials. Cracking commonly occurs just below the pitting potential, as indicated in Fig. 2, at least for passivatable alloys susceptible to stress corrosion cracking, e.g. austenitic stainless steels. I f a similar explanation for the occurrence of low pH values of solutions at the tips of cracks is to be made as is made for the same occurrence in pitting, then for hydrolysis to occur at a sufficiently high rate requires that the solution be provided with metal ions by the action of the crack tip strain-rate in breaking or disturbing the growing film and providing a flux of metal ions, since in the absence of the strain-rate the acidic solution will not be produced nor maintained below the pitting potential. Unless the creep process occurs at or above a certain minimum rate ( ~ ~,) the surface will repassivate completely and simultaneously the pH of the solution will start to rise towards that of the bulk solution since the hydrolysis reaction rate cannot be maintained. For the same reason

l O 0 0 J . C . SCULLY

E

,91__C racking Zone

Y , , ,,,,

log c.d.

F]o. 2. The potential zone in which stress corrosion is observed just below the pitting potential.

the halide ion concentrat ion at the crack tip will start to fall towards that o f the bulk solution. An at tempt is made to depict these in Fig. 3. The main feature which will be discussed later is that upon complete repassivation the crack tip solution starts to become less corrosive. This happens when the strain-rate falls to ~r, which is therefore the strain-rate at and below which equilibrium hydrolysis conditions cannot be maintained at the crack tip.

C I z"

pH~

I

, ~ / pH

C I -

, :

time

FIG. 3. Schematic drawing of the changes occurring in a crack tip solution following repassivation and crack arrest as a result of the crack tip strain-rate falling to ~r. At this point the pH begins to rise and the halide ion concentration begins to fall while the

strain-rate continues to fall.

The relatively simple concept being put forward is that cracking in halide solutions is similar to pitting, with the necessary condition that in the cracking process the creep strain-rate is necessary in order to provide dissolving metal ions f rom newly formed surfaces and thereby maintain the crack tip acidity which otherwise either would not occur or would not be maintained. In the pitting model 16 the important product is the bare metal current density multiplied by the diffusion distance. Since crack velocities can change over several orders o f magnitude in a given system, the influence o f strain-rate in maintaining crack propagat ion is best explained by en-


visaging the crack tip metal surface as being covered by a film of varying composition and, perhaps, morphology. In the extreme case the surface is bare, corresponding to the maximum velocity often observed as a plateau velocity, itself dependent upon pH and solution composition and reaching a maximum at the pitting potential, for example, in Ti 19 and AI ~° alloys. It is this value that is sometimes used, in conjunction with Faraday's Laws, to calculate the required maximum current density for an active path cracking mechanism. In pitting studies current densities have been measured or estimated in various alloys, e.g. 8 A/cm 2 in an austenitic stainless steel, 21 2 A/cm 2 on iron z2 and 0.3 A/cm z on aluminium. "3

The idea that the film on the crack tip surface may exhibit a range of properties which determine the crack velocity is not new. It has been discussed previously, 24 for example, to explain the influence of the I ion concentration on the observed z° plateau velocity on an aluminium alloy. The chemical and physical state of this layer remains to be examined. It may, for example, be a mixed layer of oxide and salt layer, with the latter component being the less protective. A direct comparison can be made with the range of current densities observed on growing pit surfaces. While much emphasis has been placed on the maximum values attained in various alloy/electrolyte combinations, a study of pit propagation rates 25 shows that pits can grow at significantly lower rates at potentials close to the protection potential. This reduction is probably caused by the presence of a film on the pit surface, evidence for which can be deduced from the high values of potential/log i changes that have been recorded for pits, e.g. b 150 mV for A1 z6 and 220 mV for an Fe-Cr alloy, 27 although other interpretations of such results can be made.

REPASSIVATION

Since repassivation at crack tip surfaces occurs in a specific type of solution and over a narrow range of potentials, it is important that repassivation experiments that are designed to examine this aspect of cracking should replicate both the solution conditions and the potential values. If, for example, repassivation in scratching or straining electrode experiments is examined well below the pitting and cracking range of potential in solutions quite different from those known to exist at crack tips, then significant amounts of hydrolysis will not occur, repassivation will be relatively rapid, and the film growth kinetics can be expected to obey a log i ~.-- log t behaviour, typical of passive film formation, and frequently observed. 28 3o Such results may have little or no application, however, to crack tip repassivation. At the crack tip the fall in pH can be expected to lower the repassivation rate, as will the increasing C1- ion concentration. As already discussed, the type of film forming may be different. Film growth kinetics under such conditions have not been widely investigated although a slowing down as the potential approaches the pitting potential has been observed in aluminium 32 and in austenitic stainless steels.31, z3 A computer simulation of a re- passivating scratch on a stainless steel surface, 3z however, over a wide range of potentials up to the pitting potential, has shown that near to the pitting potential the log //log t decay is not a more accurate description of the decay process than an exponential decay (log i ~ -- t) during the early stages (100 ms) of the film repair process, even with no allowance being made for the significant amount of hydrolysis occurring in that region of potential. Practical results ~4 for Fe in chloride/molybdate

1002 J.C. SCULLY

solutions in the pitting region of potential have indicated a similar conclusion: during the early stages of repassivation (250 ms) the results fit two exponential decay lines of slightly different slope and log/ / log t decay describes the process only after the early stage has passed. It is the early stage of repassivation that is of importance since cracking can be viewed as a series of current transients recurring at short time intervals. The repassivation rate between such transients is what needs to be known accurately.

The type of decay law operating during repassivation has a direct relevance to any electrochemical analysis of crack propagation, as has been described previously, ix I f an exponential decay is operative then the current, i, flowing from a surface deforming at a constant strain-rate, ~:, is related: log i ~ ~, under constant potential conditions. I f a different decay behaviour is operative then clearly a different dependency between strain-rate and current will be observed. This is a very important point since by the association of the creep rate with the repassivation process what is being emphasized is that the relationship between i and t during an increment of crack growth is both mechanically and electrochemically dependent.

The sequence of current transients emanating from a crack tip and observable hypothetically under ideal experimental conditions is depicted in Fig. 4 where the

log i

time

FIG. 4. Schematic drawing of the effect of repeated identical crack tip strain transients occurring at the tip of a propagating crack on the current/time relationship under a condition of constant extension rate. The same charge flows in each increment which

nucleate each other successively.

conditions are intended to be a constant extension rate in a specimen under potentiostatic control. The overall creep rate is considered to be constant, giving rise to a series of identical crack tip strain transients. The minimum charge, Qm;,, flows before a further increment is generated and complete film repair does not therefore occur before the next increment is initiated, i.e. ~ never falls to ~r. A general formula can be advanced:

v = constant Qrnin (1) t


where v is the crack velocity and t is the time interval between successive slip events which are initiated by the propagation of the crack. If v = 1 mm/h, for example, and the slip line separation is 0.1 ~m, then 3 transients will occur approximately every second. The repassivation kinetics over the first 330 ms will determine the charge flowing during that time.

Under constant load conditions, drawn in Fig. 5, successive creep transients will occur under conditions of gradually increasing stress. Each increment of crack growth will occur in successively shorter time intervals. This sequence has been extensively described previously, n The creep rate increases progressively with the propagation of each increment because the load is being supported by a reducing cross section. The maximum current attained at the beginning of the transient will gradually increase and will reach a maximum determined by electrochemical factors. Beyond that point the current will be at a maximum value for an increasing proportion of the transient lifetime until it occupies it completely, a situation corresponding to the maximum diffusion-controlled plateau velocity, as described at length previously, n Since the current flowing at any creep rate depends upon the repassivation characteristics of the solution, the time for Qmin to pass is both a mechanical and an electrochemical dependent.

log i

f f

t ime

FIG. 5. Schematic drawing of the effect of consecutive slip transients occurring at the tip of a propagating crack on the current/time relationship under conditions of constant load. The reducing cross-section causes the stress to rise, causing an increasing crack tip strain-rate, 6, and thereby increasing the current, i, so that the same charge passes in

decreasing time intervals, i.e. the crack velocity increases.

If repassivation occurs before Q,,i, has passed then crack arrest will occur. This situation, previously described, xl is drawn in Fig. 6. For reasons already described above and drawn in Fig. 3, the solution composition will then begin to change, Under constant load conditions with the strain-rate falling and the crack arrested, no further transient can be expected. If, however, the strain in the film becomes too large to be sustained and it fractures, the consequent charge passed will be relatively small since

1004 J.C. SCULLY

log i

1__

t i m e

FIG. 6. Schematic drawing of crack arrest arising because the crack tip strain-rate, /:, has fallen to ~, before Q,~n has passed. The current/time relationship shows that the

current falls to the original background value upon repassivation.

the repassivation characteristics of the changed solution will be such that film repair will be much more rapid to an extent that will depend upon the time interval between repassivation and film fracture. In the model of Vermilyea 6 the important event is the fracture of the film after repassivation. For many systems it is difficult to see that such an event will re-initiate cracking. Where repassivation has already occurred cracking will only be restarted by the production of a fresh transient large enough to produce sufficient hydrolysis immediately, as must always happen, in general, whenever cracks are initiated. This type of intermittent propagation is referred to as Interrupted Loading Stress Corrosion Cracking and can occur as a result of increasing or oscillating the applied load, both of which actions can have the effect of causing transient high strain-rates. This seems a much greater disturbance of the crack tip than that occasioned by the fracture of a film at a low strain-rate, which is the implication of the Vermilyea model. In the present state of knowledge no further distinction can be made. There is no intrinsic reason that the Vermilyea model should not apply to some systems of stress corrosion cracking, particularly in those systems where the crack tip solution may not be markedly different from the bulk solution, e.g. in nitrate and hydroxide cracking of steels perhaps, although direct evidence about the crack tip solution composition in these solutions is surprisingly lacking. For cracking in halide solutions with the attendant hydrolysis, solution changes after repassivation will render crack re-initiation more difficult. The two types of transients after repassivation in a constant load test and in an oscillating load test are shown in Fig. 7. The implication is that the transient in the constant load test is too small to be of much significance, if any. Much, of course, depends upon the mechanical properties of the film, which in turn depend upon its structure. The implication of the description given above concerning k, is of a layer adjusting continually to an extending surface which below the maximum velocity is permanently covered with some oxide/salt film probably similar to that occurring in pitting corrosion below the maximum rate. If such a picture is realistic


log i 1

film fractures

time

FIG. 7(a). Schematic diagram of crack arrest as shown in Fig. 6. Subsequently a small transient occurs adding very little charge to that already passed before crack arrest since it occurs in a solution that has already undergone the changes indicated in Fig. 3 to an extent that will be dependent upon the time lapsed between the occurrence of ~, and the

occurrence of the transient.

A + B = Q . mln

log i ' ~

C < Q min

oad increased

time Figure 7(b). Schematic diagram of crack arrest as shown in Fig. 6. Additional or oscillating loading causes a large increase in ~ and a substantial amount of additional charge flows which, when added to that passed before crack arrest, exceeds Q. . . and causes an additional crack increment. Crack arrest recurs because ~, is reached before

Q..,, has passed during the further period of crack growth.

1006 J .c . SCULLY

then fracture after repassivation might depend upon a change in the mechanical properties of the film, reflecting, perhaps, the development of a more resistant thicker, passive film.

The value of 6, determines Kl,.cc and any other mechanical threshold between failure and non-failure. It is environment-sensitive since the competition between metal surface production rate, hydrolysis and neutralization kinetics will depend, inter alia, on the nature of the bulk solution. For the same reason it will be potential- dependent, as is K I . . . . since as the potential is lowered to more active values the repassivation rate will increase. The establishment and maintenance of the acidic solution at the crack tip will be more difficult, i.e. it will require a higher rate of metal surface production. This general effect is drawn in Fig. 8(a) which shows the boundaries of cracking in relation to potential and strain-rate and which also shows how these boundaries are moved by alterations to the solution. The boundaries indicate the joint role of strain-rate, 6, and repassivation rate, f, dependent upon the value of the potential, as already shown for aluminium a2 and austenitic stainless steels,31, 3s in determining whether cracking occurs in an environment in which repassivation is possible and in one in which it is not possible. The lower boundary value of k is considered first.

E

pitting * SCC increasing: inhibitor

pitting ~ l k ~ halide potential ~ viscosity

inhibitor~ ~. ~d l¢ '~ inhibitor h,

"viscosity

Ductile Fracture

;- _._~

Fro. 8(a). Schematic diagram of the relationship between electrode potential and crack tip strain-rate, ~, for a solution in which passivation is possible. The boundaries are determined by solution composition, particularlyinhibitors, halide ion concentration, pH

and solution viscosity.

The dependence of 6, upon pH can be contrasted with the general independence of pitting potential upon pH. In the former case the occurrence of acidity is entirely dependent upon the creep rate generating fresh metal area or disturbing the surface film, while in the latter case the breakdown of the film and development of acidity is entirely electrochemical. As the pH is lowered the boundary falls to lower values of


g,, until when passivity is no longer possible, it is vertical, corresponding to where the crack velocity is similar in rate to the corrosion rate (Fig. 8(b)). Below that value o f the cracking will give way to whatever form of corrosion occurs in unstressed specimens. This is frequently intergranular corrosion. It can be expected that the boundary will be affected by passivating and film-forming inhibitors and, possibly, by adsorpt ion- type organic inhibitors too, a l though the effect o f the latter may not be continuous in situations where the crack tip moves more rapidly than the rate o f diffusion of the organic molecules.

E 2

Corrosion + SCC

SCC Ducti le

Fracture

Fic. 8(b). Schematic diagram of the relationship between electrode potential and crack tip strain-rate, ~, for a solution in which passivity cannot occur. At very low values of ~, corresponding to tests lasting for a long period of time, corrosion occurs as on unstressed specimens, e.g. general corrosion or intergranular corrosion. This is also likely to be observed as the potential is raised in the noble direction. At active potentials cathodic protection may be achieved. Over a wide range of potentials the crack propagation rate will be constant, lg,~° Because of the corrosiveness of the solution the boundary between corrosion and scc may be difficult to discern. The upper boundary may have a slight

positive slope.

In Fig. 8(a), the upper boundary of ~, above which cracking does not occur, is drawn. This is sometimes referred to as depassivation, the implication being that much or even all o f the metal surface undergoes corrosion under these conditions and crack initiation cannot occur. Ductile failure takes place in a relatively short time. Cracking will be sustained to a higher value of ~ as ~ decreases with increasingly noble potential. For this reason the line is not vertical. The addition o f inhibitors under controlled potential conditions will lower the corrosion rate and move the boundary to lower values o f 6. Increasing the solution viscosity will have the same effect. Increasing the halide ion will increase the maximum value of/;. The upper boundary of

corresponds to the conditions for maximum crack velocity, observed as a constant velocity in some systems. It is lowered by lowering the potential over a range o f pH values19, 2° above a minimum value below which crack velocity is potential-independent in some systems, e.g. A1 and Ti alloys,19, 2° as is shown in Fig. 8(b).

1008 J. C, SCULLY

The upper potential boundary in Fig. 8(a) corresponds to the pitting potential, or possibly the protection potential, if that is different, since below that value pits do not grow. This boundary is also affected by inhibitor and halide additions, as well as by solution viscosity, 32 but it is generally independent of pH. The potential at which pits nucleate and grow in straining aluminium specimens has, however, been designated the pitting potential. ~7 At and above the pitting potential there is usually a region of pitting and cracking which gives way to general disintegration as the potential is raised. This may not be the situation for those systems exhibiting a pitting inhibition potential where, perhaps, a region of non-cracking may occur, giving ductile failure. For a-brass in neutral ammoniacal solutions, raising the potential to the value at which the tarnish film dissolves lowers the crack velocity and eventually leads to a region of non-susceptibility. Anodic polarization of mild steels also leads to eventual non-susceptibility.

The lower potential boundary in Fig. 8 may be narrow or broad, depending upon the nature of the solution, in particular the ratio of activating to passivating species at the crack tip. As the solution becomes more aggressive the boundary widens until, when passivation is no longer possible, it extends across the full range of k values and downwards to that value of potential where cathodic protection becomes possible. An attempt is made to depict the same diagram for a very acidic solution (e.g. pH = 0) in Fig. 8(b). Since repassivation is not possible, the cracking range of ~ is independent of potential over a wide range.

Figure 8(a) illustrates many of the principal features of stress corrosion cracking systems: the cracking range of / ; is both environment- and potential-dependent; the highest value of ~ depends upon the maximum crack velocity of any system; the cracking range depends upon potential, strain-rate and solution composition.

THE RELATIONSHIP BETWEEN STRAIN-RATE AND REPASSIVATION RATE

The general model so far described1, n is based upon the crack tip strain-rate, ~, and the repassivation rate, ~, being in a state of relative imbalance. Kinetically, this would arise as a result of the relationship between the two which might be generally described as a ratio: ~/f(i')wheref(~)is some unknown function. In a previous paper la discussion centered around the ratio, R, of repassivation rate to strain-rate. The inversion has been made only because it provides an easier presentation of the results in references 13 and 41. Since the strain-rate generates area and the repassivation rate is also concerned with area it would be easy to take, as a simple assumption, the ratio ~./L but the generalized ratio is preferred until such time as the precise nature of the ratio is determined. The concept of the importance of the ratio between the two rates must be qualified since there will be a limit, possibly quite narrow, to the values of and ~ over which the ratio will apply and stress corrosion crack propagation occur. Thus it must be supposed x > ~; > y and a > ~ > b, where x and a are maximum values and y and b are minimum values of ~ and ~ respectively. It may also be expected that even within these ranges high values of ~ and ~ (~ x and o) and low values of and ~ ( ~ y and b) which give the same value ratio cannot necessarily be expected to yield the same effect. With such general conditions it becomes possible to examine the significance of the ratio for a limited number of reported experiments. Much


work has been repor ted for crack velocity as a function of crosshead speed (and, indirectly, s t rain-rate) but little work has been repor ted on crack velocity as a function of repass ivat ion effect. Several workers 35 7 have used repass ivat ion rat ios, e.g. the values of current a t some t ime interval after the ini t iat ion of repass ivat ion for two different potent ia l values, or the t imes taken for the current to fall to some fract ion o f the t ransient max imum, as indicators of repass ivat ion significance but it is not possible to use such da ta for the present discussion. Effects can only be general ly and qual i ta t ively inferred, e.g. that # increases as the potent ia l is lowered or as the con- centra t ion of an oxidizing inhib i tor is added under cons tant potent ia l condi t ions .

a -BRASS

Crack velocity measurements in Johnson and Leja solut ions at p H 6.8 have been repor ted recently 13 as a function of crosshead speed. Over a wide range o f values (0.66 nm-33 ~m/s) cracking was observed to be in tergranular in a solut ion con- ta ining 6 g/l of dissolved copper . Fi lm format ion and rupture appea r to be an impor t an t par t o f the crack p ropaga t i on process since a crack arrest phenomenon was detected in the cons tant crosshcad speed tests 13 and also in notched cant i lever beam experiments , 3s which can be a t t r ibu ted directly to a film forming process. In the lat ter case, removal of the film resulted in renewed crack p ropaga t ion in the same solution.

If the crack tip s t rain-rate developed in such tests is assumed to be directly pro- por t ional (and this poin t has been discussed previously) 13 to the imposed crosshead speed, then the results previously repor ted la for velocity vs. c rosshead speed can be depicted schematical ly as shown in Fig. 9 in which the strain-rate, ~, is d iv ided by the repassivat ion rate, ~, for tha t par t icu la r solut ion, which is not known, a l though a

log v

r % ductile f ra~re, I 100

S~IG+TG //7..

oo. . P " / h . < . ° . , , , , ; ,

. . . . . . . • ° ° t"'lG *'I 'G ~.! > r2

FIG. 9. Schematic diagram of the relationship between the ratio ~/l(~) and crack velocity for a-brass in neutral ammoniacal solutions, taken from ref. 13. The solid line - - is for one solution (f (constant) - ?0. The dashed line . . . . is for one crosshead speed with ~ being lowered, the broken line - - - - - suggests how the velocity falls at lower values of the ratio based upon a lower constant value or ~(~). The cracking/ductile fracture barrier is also drawn on the diagram. This corresponds to the upper value of

in Fig. 8. The lower the crack velocity the earlier is ductile fracture encountered.

1010 J .C . SCULLY

growth rate of 90 nm/s has been reported a9 for a similar solution. In the same solution any change that lowers t ~ results in (i) a fall in crack velocity, and (ii) a transition from a completely intergranular fracture to a mixed mode fracture, intergranular/transgranular. Such changes are caused ~3 by (a) anodic polarization, (b) cathodic polarization, (c) the addition of halide ions, (d) lowering the dissolved copper concentration, and (e) raising or lowering the pH from 6.8. The common effect of each of these changes is to lower i. Most of these changes were examined at only one crosshead speed. These results are drawn in Fig. 9. The solid line corresponds to a ratio derived from a varying ~ and a constant ~ (il) with completely intergranular cracking. When

is lowered as a result of any of the changes listed in (a)-(e) the velocity falls and the ratio increases. This change is drawn as a dashed line showing a falling velocity as ~/f(i) increases as a result of lowering ~ at a constant crosshead speed (~ constant). Additionally, the same general shape of the solid line can be expected to be found for solutions with a lower ~ value (t~2) but with lower values of velocity. An example of such behaviour is drawn as a lower broken line. There is some support for such an assumption. Crack velocity measurements in 15N ammoniacal solutions 4° have shown that altering the potential or lowering the concentration of reactive species (a) suppresses the complete log v/K curve, which is of the same form as Fig. 9, (b) lowers the slope of the K-dependent velocity section, and (c) does not affect the transition point of the two lines. The reduced slope drawn in Fig. 9 would be expected from consideration of simple proportions. If the denominator in the ratio, ~, is lowered, the values of the ratio will be further apart over the same range of crosshead speeds (~ values).

Also shown in Fig. 9 is k r which for this alloy systems corresponds to a lower crosshead speed than the lowest employed in the work ~z from which the figure is derived. Crack arrest was observed in experiments in which the crosshead movement was discontinued. This would correspond to some low, unmeasured crack velocity. This initial part of the log v/log CHS graph was not determined and for that reason a gap is left in the figure.

The upper limit of ~, corresponding to depassivation, can be expected to fall with decreasing corrosion rate, as discussed in the previous section. As ~ increases, however, cracking does not occur and the boundary of the cracking/ductile failure modes will go towards lower ~/f(i') ratios, as is drawn. Thus x/a causes a higher crack velocity than y/b, even when the two ratios have the same value.

AUSTENITIC STAINLESS STEELS

In austenitic stainless steels in MgCl~ solutions at 154°C log v/log CHS results have been reported by Takano. 41 These are reproduced in Fig. 10. Work under potentiostatic conditions has yielded similar results34 Takano observed that fracture was intergranular at lower temperatures as did Stalder and Duquette 4~ (which would also include a lower chloride concentration). It has also been reported 4a that additions of Mo promote intergranular cracking at 154°C. All this indicates that as the repassivation rate, t ~, increases, the cracking becomes, at least partly, intergranular. The results, drawn in Fig. 11, are similar to s-brass, except that there is a morphological reversal: the higher velocity is transgranular.


FIG. 10.

-I

log

V

mm/s -2 / ~

-3 i i i 3 - 2 -1 0

log C, H.S. mm/s

Reproduction of the results of Takano4*: Type 304 steel in MgCI., solution at 154°C.

log v

Er c o n s t a n t

FI~. 11. Schematic diagram of the relationship between the ratio ~/f(i-) and crack velocity for Type 304 austenitic stainless steel in a MgCi~ solution boiling at 154°C. Constant f refers to experiments done 14 at a constant potential; constant g refers to

experiments dorte t4 at a fixed crosshead speed under conditions of increasing f.

Figures 9 and 1 1 are based, as far as is currently possible, upon experimental work. They place particular emphasis upon the repassivation rate in controlling the crack propagat ion rate and crack morphology, together with the crack tip strain-rate. The potential is indirectly a factor; its importance lies in its effect upon f. Where there is no repassivation the crack propagat ion rate is constant (Fig. 8(b)).xg. 2° For any value o f potential within the cracking range the repassivation rate will depend upon the crack tip solution composition. As has already been emphasized with reference to Fig. 1, the factor controlling propagat ion is the charge and the calculation o f this incorporates the repassivation rate.

1012 J .C. SCULLY

CRACK TIP SOLUTION COMPOSITION AND CRACK PATH

The tentative descriptions of the relationship between stress corrosion crack velocity and the ~/f(f) ratio are consistent with the observed cracking mode changes. The type of fracture observed appears to depend upon the repassivation process at the crack tip which, in turn, depends upon the value of the potential and the composition of the crack tip solution. In ~-brass the filming characteristics control the crack path in neutral ammoniacal solutions. 18 Rapidly tarnishing solutions cause intergranular cracking and any change that lowers this rate produces transgranular cracking, as has already been described, which propagates more slowly. This may be the consequence of the increase in the number of reactive sites as a result of the slower film formation rate: the same current is distributed over a higher number of sites. It may also arise because the transgranular crack paths are less reactive than the grain boundaries, although if this were true, the first argument would still need to be invoked to explain why the crack followed the less reactive path.

In austenitic stainless steels the reverse change is seen. The commonly observed transgranular fracture becomes intergranular in solutions that are less aggressive, i.e. allow repassivation of slip steps, than those that cause transgranular cracking. In such solutions, developing near to the crack arrest potential, 14 and in solutions con- taining inhibitors 14 and as the chloride ion concentration and temperature are lowered, or as the Mo content is increased (which can be expected to increase the repassivation rate) the emergent slip steps are repassivated but the grain boundaries are not and intergranular cracking is observed. As in 0~-brass, intergranular cracking occurs when the crack tip solution allows slip step repassivation. The explanation for these changes is not clear. The less aggressive solution causing intergranular stress corrosion cracking in the stainless steels may have undergone less hydrolysis since it is associated with relatively low corrosion rate conditions. There may therefore be a greater deposition rate from the solution. It is at least conceivable that different types of film are formed in the two solutions. Compositional effects, e.g. P in the grain boundaries, may also be of importance. There are many such possible causes but currently there is a lack of experimental evidence to support any one of them.

The attempted analysis that the less aggressive, more rapidly filming solutions cause intergranular cracking while the more aggressive, less rapidly filming solutions cause transgranular cracking seems to apply to other systems of active path stress corrosion cracking. Armco iron, for example, cracks intergranularly 44 in phosphate solutions of pH 9, probably in a similar way to steels in HCOa-/CO32-, O H - and NO3- solutions with the formation of thick protective magnetite scales, possibly associated with other compounds, 45 yet if tested at pH 4 it cracks partly transgranularly. 44 In the acidic solution it can be supposed that the crack tip solution is sufficiently aggressive to be less rapidly film forming and cause transgranular cracking. Plain carbon steels also crack transgranularly in gaseous CO/CO2 mixtures 4e in which the rapid formation of a protective three-dimensional film cannot be expected.

The circumstances under which such fracture mode transitions may be seen are likely to be confined to a relatively small number of examples. Altering the solution pH can be expected either to promote general corrosion, as a result of lowering f, or to promote passivity, as a result of raising f, both of which effects will tend to prevent stress corrosion crack propagation. Fracture mode transitions probably


require small changes in solution causing small changes in t ~. Transitions are rarely complete. The changes observed in x-brass ~s and austenitic stainless steel ~4 are to mixed modes, not complete transitions, indicative of very small changes in solution, possibly not homogeneous along a propagating front.

For ~-brass and mild steel it is possible to observe that intergranular cracking is associated with the rapid formation of relatively thick films. Under such circumstances rapid slip step repassivation may be achieved by the small amount of surface revealed. An attempt is made to draw this in Fig. 12. I f the film is relatively thick and

A

C

FIG. 12. Schematic diagram indicating how the amount of fresh surface exposed after rupture of a surface film is reduced if the film grows at a significant rate. The area exposed initially will be determined by the thickness of the film at the time of creation of the step. The rate at which the exposed area is reduced will be determined by the growth rate of the film. The diagram depicts a film growing inwardly. A similar situation

arises if the film grows outwardly.

growing rapidly, e.g. linearly at 90 nm/s over the first 3 s for ~-brass, sa then the charge emanating from the freshly revealed step will be relatively small. For austenitic stainless steels, intergranular cracking at elevated temperatures, c a . 300°C in chloride- contaminated steam, may be attributable to the same cause--the formation of thick films, and as with ~-brass, by deposition from solution, which can produce rates that do not diminish rapidly with increasing thickness. In MgCI~ solutions at 154°C the explanation is not so obviously applicable. There are no data on the relative thick- nesses of films formed on slip steps and grain boundaries in this system but they are not likely to be as thick as those formed on brass. Nor do they exhibit a linear growth rate. Rather than thickness it is the relative rates of repassivation that are important. Intergranular cracking occurs in less aggressive solutions in which slip steps are more rapidly repassivated. In more aggressive solutions slip steps do not repassivate and a more rapid transgranular cracking ensues. While an active path has been assumed, it must also be pointed out that an alternative explanation is available. Recent workS~, 53 has illustrated similarities between transgranular stress corrosion fractures and hydrogen embrittlement fractures in austenitic stainless steels. This may indicate that stress corrosion cracking in this material falls within the second category: inter-

1014 J.C. SCULLY

granular dissolution cracking and transgranular hydrogen cleavage cracking, with the latter occurring in more aggressive solutions.

Inherent in this discussion is that in a-brass, austenitic stainless steels and plain steels the transition in cracking mode is dependent upon crack tip chemistry causing rapid film growth and not directly upon any mechanical variable such as strain, strain-rate or stress, which is why it is possible to discuss it in terms of the ratio ~/f(i'), independently of stress.

The three systems described above are commonly thought to crack by an active path mechanism. For the more mechanical types of stress corrosion cracking, exhibiting cleavage modes of failure attributable to embrittlement by absorbed hydrogen, e.g. Ti alloys 12 and zircaloy-2, 47 the transition from intergranular dissolution to transgranular cleavage in corrosive mixtures of CH3OH/HC1 is associated with a cleavage initiation stress. The cleavage failure is still observed even if the grain boundaries are repassivated by corrosion-inhibition additions of water to the corrosive mixture or if experiments are conducted in neutral aqueous NaCI solutions in which intergranular fracture is not seen. In such examples the less aggressive passivating solution does not prevent the occurrence of transgranular cleavage. These are not examples of one crack path being preferred to the other because of differences in repassivation rate. They are examples of hydrogen being absorbed to a sufficient degree to cause cleavage fracture during the repassivation of slip steps at a ~ > ~r- For zircaloy-2 anodic polarization is required to obtain transgranular cleavage in aqueous NaCI solution which suggests that Ecor~ is below the lower boundary in Fig. 9.

It appears that a simple distinction can be made between active path stress corrosion cracking and that caused by hydrogen absorption. In the former type the crack path is independent of the stress and is determined by the crack tip solution composition. In the latter type cleavage occurs at a certain stress which probably depends, at least in some alloys, upon the amount of hydrogen absorbed. What happens below that value of stress depends upon the type of corrosion occurring on unstressed specimens, which is commonly either no corrosion or intergranular corrosion, which means that either only cleavage is seen or cleavage is preceded by intergranular cracking.

CONCLUSION

A detailed analysis of events occurring at the tip of a propagating crack is difficult to make, particularly of a quantitative nature, because there are so many unknown factors. Some of these are rarely discussed. Two general examples can be quoted. First, it can be supposed that the corrosion of a stressed metal may release dislocations as the result of the dissolution of a dislocation barrier. 4a If this happens, and it would seem to be more likely to occur as the amount of corrosion occurring increases, then locally ~ will be increased. This effect coutd be developed as an argument that increases as ~ decreases and thereby allows more corrosion to occur. This dissolution of dislocation barriers is implicit in the description provided for Figs. 4 and 5. Corrosion periodically releases dislocations and causes high local strain-rates. A more specific explanation is required, of course, in order to delineate why such an effect gives rise to a crack, and this must be derived from morphological studies. Why, for example, do transgranular cracks in austenite, 49 a-brass and Al-alloys 5°


follow (110) p lanes? Secondly, a selective adso rp t ion o f species f rom solu t ion on to the emerging metal surface may occur and thereby al ter the crack t ip so lu t ion significantly, sx Such adsorp t ion effects, if they occur, would depend upon a range o f surface topograph ica l and compos i t iona l factors.

Several o ther areas o f equal poss ible impor tance (e.g. de-al loying) cou ld be cited, but little or no direct evidence can be cited and these areas cannot be pursued current ly . The discussion above has therefore been a relat ively simple descr ip t ion o f wha t are undoub ted ly complex interact ions. A n a t t empt has been made to depic t c rack ing as occurr ing when the re la t ionship between the crack t ip s train-rate, ~, and repass iva t ion rate, ~, is such tha t an acidic solut ion can be main ta ined at the crack t ip a n d comple te repass ivat ion is delayed. It has been argued tha t in active pa th cracking the solut ion compos i t ion determines the crack path , with rapid ly fi lming solut ions giving inter- g ranu la r cracking, and slowly filming solut ions giving t r ansgranu la r cracking. M u c h o f the behav iour o f stress cor ros ion cracks can be explained by cons ider ing the re la t ionship between ~ and t ~.

Acknowledgement--It is a pleasure to acknowledge that much of this paper was written while the author was a Visiting Professor in the Department of Metallurgy and Materials Science of the University of Florida, in Gainesville, Florida.

R E F E R E N C E S 1. J. C. SCULLY, Corros. Sci. 7, 197 (1967). 2. H. L. LOGAN, J. Res. nat. Bur. Std48, 99 (1952). 3. J. C. SCULLY and D. T. POWELL, Corros. Sci. 10, 371 (1970). 4. J. C. SCULLY, Corros. Sci. 8, 513 (1968). 5. R. W. STAIBHLE, The Theory of Stress Corrosion Cracking in Alloys, p. 223 (Ed. J. C. SCULLY).

NATO, Brussels (1971). 6. D. A. VERMILYEA, J. electrochem. Soc. 119, 405 (1972). 7. W. R. W~ARMOUTH, G. P. DEAN and R. N. PARKINS, Corrosion 29, 251 (1973). 8. H. J. ENGELL, ref. 5, p. 68. 9. G. BIGNOt.D, Corrosion 28, 307 (1972).

10. T. R. BECK, J. electrochem. Soc. 115, 890 (1968). 11. J. C. SCULLY, Corros. Sci. 15, 207 (1975). 12. J. C. SCULLV and T. A. ADEPOJU, Corros. ScL 17, 789 (1977). 13. M. K~RMANI and J. C. SCULLY, Corros. Sci. 19, 89 (1979). 14. M. TALEBIAN and J. C. SCULLY, to be published. 15. J. F. N~WMAN, C.E.R.L. Report RD/L/N 120/78 November (1978). 16. J. R. GALVEL~, J. electrochem. Soc. 123, 434 (1976). 17. S. B. DE WEXLER and J. R. GALWLE, J. electrochem. Soc. 121, 1272 (1974). 18. B. F. BRow~, ref. 5, p. 186. 19. M. J. BLACKnURN, J. A. FEENi~Y and T. R. BECK, Advances in Corrosion Science and Technology

(Eds. M. G. For,rrANA and R. W. STAEHLE), Vol. 3, p. 67. Plenum Press, New York (1973). 20. M. O. SPEIDEL, ref. 5, p. 289. 21. N. SATO, T. NAKAOAWA, K. KUDO and M. SAKASHITA, Trans. Japan Inst. Metals 13, 103 (1972). 22. K. J. V~T'rER and H. H. STRI~HBLOW, Ber. Bunsenges. Physik. Chem. 74, 1024 (1970). 23. H. KA~c~m, Z. Physik. Chem. N.F. 34, 87 (1962). 24. J. C. SCULLY, Third International Congress on Fracture, Munich, 1973. PL-1V-222, Proo~dings,

Dusseldorf (1973). 25. B. C. S ~ T r , Corrosion 33, 221 (1977). 26. H. KAI~SCHE, Z.Physik. Chem. NF34, 87 (1962). 27. N. SATO, T. NAKAOAWA, K. KUDO and M. SAKASmTX, Localized Corrosion, p. 447, N.A.C.E.,

Houston (1974). 28. D. M. BRASh~R and A. H. KI~GSBURY, Trans. Faraday Soc..54, 1214 (1958). 29. N. SATO and M. COHEN, J. electrochem. Soc. 111, 512 (1964). 30. G. M. BULMAN and A. C. C. TSEUNG, Corros. Sci. 12, 415 (1972).

1016 J . C . SCULLY

31. P. ENGSETH and J. C. SCULLY, Corros. Sci. 15, 505 (1975). 32. W. J. RUDD and J. C. SCULLY, Corros. Sci. 20, 611 (1980). 33. M. BARBOSA and J. C. SCULLY, Environment-Sensitive Fracture of Engineering Materials (Ed. Z. A.

FOROULIS), p. 91. A.I.M.E., New York (1979). 34. T. KODAMA and J. R. AMBROSE, Corrosion 33, 155 (1979). 35. Y. S. PARK, J. R. GALVELE, A. K. AGRAWAL and R. W. STAEHLE, Corrosion 34, 413 (1978). 36. J. RXMBERT and J. PAGETTI, Corros. Sci. 20, 189 (1980). 37. R. N. PARKINS, Corros. Sci. 20, 147 (1980). 38. M. KERMANI and J. C. SCULLY, Corros. Sci. 19, 111 (1979). 39. J. A. S. GREEN, H. D. MENGELBERG and H. T. YOLKEN, J. electrochem. Soc. 117, 433 (1970). 40. M. KERMANI and J. C. SCULLV, Corros. Sci. 18, 883 (1978). 41. M. TAKANO, Corrosion 28, 332 (1972). 42. F. STALDER and D. J. DUQUETTE, Corrosion 33, 67 (1977). 43. H. OKADA, Y. HOSOI and S. ABE, Corrosion 27, 424 (1971). 44. J. FLIS, Corros. Sci. 19, 151 (1979). 45. R. N. PARKINS, N. J. H. HOLROYD and R. R. F'ESSLER, Corrosion 34, 253 (1978). 46. A. BROWN, J. T. HARRISON and R. WILKINS, Corros. Sci. 10, 547 (1970). 47. P. M. MAJUMDAR and J. C. SCULLY, Corros. Sci. 19, 141 (1979). 48. J. C. SCULLY, The Physical Basis of Yield and Fracture, p. 119. The Institute of Physics, London

(1966). 49. G. M. SCAMANS and P. R. SWANN, Corros. Sci. 18, 983 (1978). 50. J. L. NELSON and J. A. BEAVERS, Met. Trans. 10A, 658 (1979). 51. J. C. SCULLY, Brit. Corros. J. 1, 355 (1966). 52. D. ELIEZER, D. G. CHAKRAPANX, C. J. ALSTETTER and E. N. PUGH, Met. Trans. IOA, 935 (1979). 53. H. HANNINEN artd T. HAKKARAINEN, Met. Trans. 10A, 1196 (1979).

Stress corrosion cracking

Documents