Corrosion Science 46 (2004) 12391264 www.elsevier.

com/locate/corsci

Pit to crack transition in stress corrosion cracking of a steam turbine disc steel
A. Turnbull *, S. Zhou
Materials Centre, National Physical Laboratory, Queens Road, Teddington, Middlesex TW11 0LW, UK Received 16 June 2003; accepted 8 September 2003

Abstract Long term exposure tests have been carried out on a 3 NiCrMoV steam turbine disc steel in the form of cylindrical tensile test specimens self-loaded to 90% of r0:2 and exposed to three environmental conditions, viz. deaerated pure water, aerated pure water, and aerated water containing 1.5 ppm of chloride ion. Pitting occurred in all environments but the density and depth of pits in the chloride-containing medium was markedly greater. No cracking was observed in deaerated pure water but cracks initiated in aerated water between 13 and 19 months and in less than 7 months in aerated 1.5 ppm Cl solution. The probability of a crack initiating from a pit of specic depth in aerated solution could be described well by a Weibull function. Proling of pits and cracks in the disc steel tested in aerated 1.5 ppm Cl solution showed that there while there were many cracks with a depth greater than that of the corresponding pit the depth of some cracks was smaller than that of the corresponding pit, suggesting that cracks do not necessarily initiate from the bottom of the pits. The growth rate of short cracks emerging from pits appeared greater than that of long cracks in fracture mechanics specimens. 2003 Published by Elsevier Ltd.
Keywords: A. Steel; C. Stress corrosion; Pitting corrosion

1. Introduction Stress corrosion cracking of steam turbine steels continues to be an area of concern for the power industry. Although failures are rare, compared with that of boiler tubes for example, the potential consequences in terms of system damage are severe. Stress
*

Corresponding author. Tel.: +44-20-8943-7115; fax: +44-20-8614-0436. E-mail address: alan.turnbull@npl.co.uk (A. Turnbull).

0010-938X/$ - see front matter 2003 Published by Elsevier Ltd. doi:10.1016/j.corsci.2003.09.017

1240

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

corrosion cracking of steam turbine disc steel most commonly originates from corrosion pits, except perhaps at high stresses for which the role of pits may be less signicant (because there is strain localisation to disrupt the surface lm and there is no need for stress concentration). When pitting is the precursor to stress corrosion cracking, the fundamental steps in the overall process of crack development include: pit initiation; pit growth; transition from pit to crack; short crack growth; long crack growth. Whilst there are considerable data for the growth rate of long cracks in various simulated environments [1], there is comparatively little information on the growth rate of pits and the early stages of crack growth. Key external factors aecting pit initiation and growth include oxygen concentration, chloride concentration, nature and concentration of other anions, temperature and dynamic strain. Inclusion density, size and distribution would be key material parameters. As part of a larger programme designed to provide a framework for predicting the stress corrosion cracking of turbine disc steels, long-term exposure tests have been conducted to investigate the evolution of pits and cracks on self-loaded tensile specimens under conditions nominally expected to represent service exposure in a conventional coal-red power station. Typically, on-load, a thin liquid lm will form from the steam at temperatures below about 90 C. Recent evaluation [1] suggests that the liquid lm will contain up to about 300 ppb chloride ion under normal operating conditions with well-controlled water chemistry, with perhaps up to 1.5 ppm chloride if the chloride in the inlet steam is not controlled tightly. On-load, the oxygen concentration of the liquid lm will be negligible. O-load, the liquid lm on the discs will usually be aerated, unless there is nitrogen blanketing, and at a temperature of 70 C, but the chloride ion concentration will be very low. During startup, the chemistry is less well understood but a transient combination of oxygen in solution with some chloride is possible.

2. Experimental 2.1. Material The material used in this work was a disc steel (3 NiCrMoV), cut from an exservice steam turbine disc supplied by PowerGen. The chemical composition is listed in Table 1. 2.2. Specimens Self-loaded cylindrical specimens of the disc steel were manufactured according to ASTM E8 [2] with the longitudinal axis perpendicular to the radius of the turbine
Table 1 Chemical composition of the 3 NiCrMoV disc steel (mass %) C 0.30 Si 0.28 Mn 0.45 P 0.017 S 0.013 Cr 0.69 Mo 0.27 Ni 2.89 V 0.091 N 0.21 Fe bal

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1241

Fig. 1. Photograph of the specimen and the loading frame for pitting and SCC tests.

disc. The overall length was 100.0 mm, the shoulder diameter 16.0 mm, the gauge length 25.4 mm and the diameter 6.4 mm. The specimen and the loading frame are shown in Fig. 1. To avoid galvanic corrosion, the loading frames and the nuts were also made of the disc steel. The specimen surface was dry ground using P2400 grinding paper (1200 SiC grit nish). The surface prole [3] was measured using a laser prolometer. Measurements were made at eighteen locations on two specimens. The length of prole for each individual measurement was 2 mm. The arithmetical mean deviation of the assessed prole, Ra , was 0.50 0.28 lm. The corresponding maximum height of the prole, Rz , was 3.13 0.91 lm. The nal grinding direction, and hence any grinding marks, were parallel to the length of the specimen. 2.3. Stress relief To avoid an uncontrolled inuence of residual stress resulting from machining or from grinding, all specimens, after nal grinding, were stress-relieved in vacuum for 2 h at 625 C. The mechanical properties at ambient temperature and at the test temperature are listed in Table 2. The data were the average of three tests under each condition. It can be seen that the stress relief had no eect on the mechanical properties. Since the specimens were cut from an ex-service disc, there was a concern that the
Table 2 Mechanical properties of the disc steel (mass %) Condition As received Stress relieved Stress relieved T (C) 23 23 90 E (GPa) 208 2 206 1 210 4 r0:2 (MPa) 747 9 734 4 705 4 UTS (MPa) 882 5 870 2 827 13 Elongation (%) 23 1 20 0 20 0

1242

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Fig. 2. Microstructure of 3% NiCrMoV disc steel.

mechanical properties may vary at the dierent locations of the disc. Therefore, the hardness of the steel was measured on ten randomly selected specimens. There was no signicant variation in the hardness measured; the average value being 28.5 (Rockwell C), with a standard deviation of 1.5. The microstructure of the materials after stress relief is shown in Fig. 2. The mean grain size was 26 8 lm. 2.4. Pre-pitting Some of the cylindrical specimens were pre-pitted to produce single pits, or occasionally a few pits using the procedure developed at NPL [4]. The anticipation was that the single pits would grow preferentially and the cracks might initiate readily at these locations. Should this be the case then crack measurement probes could then be located accordingly in future tests. The use of pre-pits also might be expected to give insight as to the subsequent likely growth rate of pits should their initiation and early development be accelerated by a water chemistry excursion. The pre-pitting procedure established for cylindrical tensile specimens of 3% NiCrMoV disc steel is summarised as follows. A solution of 0.02 M Na2 B4 O7 + 0.01 M NaCl (pH 9.2) was used. The test temperature was 23 1 C. A test cell with a ushed-port seal adapted for cylindrical specimens was used to prevent crevice corrosion at the seals between the cell and the specimen. After the specimen was immersed in the test solution for 1 h, a constant anodic current of 10 lA was applied to the exposed area of approximately 9.6 cm2 . This tends to lead to development of a single dominant pit. To minimise corrosion on the periphery of the pit, the solution was gently stirred but the stirring was applied only after the pit had just initiated as indicated by the decrease in potential. If the applied current were increased to 20 lA, more than one pit (usually less than 5) was produced on the specimen surface.

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1243

2.5. Test environment The environments used for the pitting and SCC tests are listed in Table 3. 2.6. Water supply/circulation systems 2.6.1. Deaerated high purity water at 90 C The quality of the deaerated high purity water was maintained by using a oncethrough system, as shown schematically in Fig. 3. The deaeration was achieved initially by continually passing nitrogen into a 20litre reservoir containing the feed-in deionised water. Further water purication and
Table 3 The environments and test temperatures for pitting and SCC testsa Environment Deaerated high purity water Aerated high purity water Aerated water + 1.5 ppm Cl Conductivity (lS cm1 ) 0.056 pH Assume neutral 5.8 5.9 Test temperature (C) 90 Simulation Liquid lm chemistry with high steam wetness/high purity Liquid lm chemistry during start-up/air leak Liquid lm chemistry during major system upset.

1.1 6.8

90 90

a Prior to the detailed review and analysis of Ref. [1] there was an indication in several papers on stress corrosion cracking that on-load conditions were best represented by laboratory tests in high purity water. This was the basis for this environment when initiating these long-term tests. It is now realised that 300 ppb Cl is likely in the liquid lm whilst on-load.

Fig. 3. Schematic diagram showing the test cell and high purity water supply system.

1244

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

deaeration were achieved and maintained by continually circulating the feed-in deaerated deionised water in a loop containing a second deionising cartridge (purier) and three de-gasser cylinders. The main component of the de-gasser is a membrane that allows only the gas (oxygen) to pass through (Fig. 3). The conductivity and oxygen level of the inlet and outlet water were monitored using on-line probes. The low conductivity and oxygen level were achieved using a once-through water ow system in which the water ows from the above circulation line to the test vessels and then to drainage. The ow rate was approximately 15 ml/min and hence the water in the 1 l test cell was replenished about once an hour. The slow rate of input of solution into the test cell allowed the temperature to be maintained within 1 C without the need for a pre-heater. 2.6.2. Aerated high purity water at 90 C A once-through system, similar to that shown in Fig. 3, was also used for tests in aerated high purity water but in this case the inlet water was aerated using an air pump. The ow rate was also approximately 15 ml/min. 2.6.3. Aerated water containing 1.5 ppm Cl at 90 C The solution circulation loop for aerated water containing 1.5 ppm Cl was slightly dierent from that in Fig. 3. In this case the aerated feed-in solution was recirculated from a 40-l reservoir, again at a ow rate of approximately 15 ml/min. The solution in the 40-l reservoir was refreshed every week and the conductivity and pH were measured before and after the solution change. 2.7. Corrosion potential measurement The corrosion potential was measured using an external 0.01 N Ag/AgCl reference electrode designed specially with an extremely low chloride diusion rate. The reference electrode was connected to the test cell through a glass bridge lled with the test solution. The reference electrode chamber and the test cell were separated with a glass valve, which was closed except when a measurement was taken. The solution conductivity was monitored to ensure that there was no signicant chloride contamination due to the diusion of chloride from the reference electrode chamber to the test cell during the potential measurement. For the potential measurement in deaerated high purity water, the reference electrode chamber was sealed. The reference electrode remained at ambient temperature. All potentials quoted are with reference to the saturated calomel electrode (SCE) at 25 C. 2.8. Test procedure 35 self-loaded cylindrical tensile specimens of disc steel were used; 16 of the specimens were smooth (without pre-pitting), seven specimens had a single pre-pit and 12 specimens had multiple pre-pits. The depth and the surface width of the pit (the mouth of the pit was not always ideally circular) were measured using a travelling microscope (resolution 1 lm). In the latter case, the maximum width in one

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1245

direction was measured and the width perpendicular to this was measured also. The average value was used later in calculating the aspect ratio. The location of the prepits was mapped out before testing so that the surface of the tested specimen could subsequently be examined to determine the growth characteristics of the pre-pits. All specimens were loaded to 634.5 MPa, i.e. 90% of the 0.2% proof stress (r0:2 ) at the test temperature (90 C). In order to prevent possible stress relaxation as the specimen and the loading frame were heated to the test temperature, all specimens were loaded at 90 1 C with the specimen and the loading frame placed in a temperature-controlled oven. The loaded specimens were then cooled down and stored in a dessicator for at least 24 h before being placed in the test cell. For each environment, the specimens were distributed in two or three dierent test cells of 1 l capacity so that specimens could be removed and examined at dierent test intervals without aecting the remaining specimens. After placing the specimens in the cells, the test solution was introduced and the test temperature increased gradually to 90 C in about three hours. For deaerated high purity water, the test cells were purged with nitrogen for 4 h before the solution was introduced. The temperature was controlled at 90 1 C and recorded hourly. 2.9. Water chemistry control The conductivity, oxygen level and pH were measured throughout the tests at weekly intervals and the results are summarised in Table 4. In the case of the recirculated chloride solution, the pH and conductivity were measured immediately after refreshment (new solution) and after testing for 7 days, just prior to refreshment (old solution). Table 4 indicates that the conductivity of the inlet deaerated high purity water is almost the same as the calculated value of pure water (0.055 lS/cm) [5]. The conductivity of the outlet deaerated high purity water is much higher but this is mainly due to the fact that it is measured at 50 C (the conductivity of pure water at 50 C is
Table 4 Control of environmental conditions (all data were obtained at ambient temperature 23 1 C, unless stated otherwise) Environments Test temperature (C) 90 Solution into cell Conductivity (lS/cm) 0.056 O2 (ppm) <0.0001 pH nma Solution out of cell Conductivity (lS/cm) 0.34b O2 (ppm) <0.001b pH nma

Deaerated high purity water Aerated high purity water Aerated water + 1.5 ppm Cl

90 90

1.1 6.8c

nma nma

5.8 5.9c

1.2 8.0d

4.4b 2.8b

5.8 6.3d

Note: (a) nm: not measured; (b) measured at 50 C; (c) new solution; (d) old solution.

1246

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

about 0.2 lS/cm) [5]. The oxygen content in the deaerated pure water was controlled at less than 1 ppb, except during one excursion in one corrosion cell. The measured conductivity of the aerated pure water of about 1 lS/cm reects the eect of dissociation of dissolved carbon dioxide. There was only very modest change in the pH and conductivity of the outlet aerated high purity water and chloride-containing solution compared to that of the inlet, highlighting the eectiveness of the water chemistry control. It should be noted that the oxygen concentration of the aerated water and of the 1.5 ppm Cl solution was measured at the outlet at a lower temperature (50 C) than the test temperature. Nevertheless, the oxygen concentration in the test cell at 90 C can be assumed to be the equilibrium value (1.4 ppm at 90 C), since air bubbling was continuous in the reservoir solution. A concern for long-term tests using glass reaction vessels is the inhibiting eect on corrosion of dissolved silica-containing species from the glass [6,7]. To minimise this, all glassware was pre-exposed to water at 90 C for a month, as previous work [7] had shown that using boro-silicate glass treated in this way for corrosion tests gave equivalent results to a quartz glass vessel in terms of dissolved silica and corrosion rate. The impact of dissolved silica is not a problem for tests in deaerated and aerated water, as the water was continuously refreshed at a rate of approximately 1 l/h. To ensure that the accumulated silica level in the test cell was not high with the recirculated aerated 1.5 ppm Cl solution, the silica content in the new solution and the old solution (before being refreshed after one week of testing) was analysed using Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) and Graphite Furnace Atomic Absorption Spectroscopy (GF-AAS). The silicon content was less than 200 ppb for the fresh solution and 700 200 ppb for the old solution. Previous work at NPL has demonstrated that silica species at these levels do not aect the corrosion rate of mild steel in distilled water [7].

3. Results and discussion 3.1. Eect of oxygen and chloride ion on the corrosion potential The corrosion potential of all specimens was measured. There was little variation for the same type of specimens in the same cell, with the spread being less than 0.05 V. Typical corrosion potentials in each cell are shown in Fig. 4. These corrosion potentials reect not only the specimen but also the self-loading frame, which is of the same material. The corrosion potential of the disc steel in deaerated high purity water stabilised relatively quickly, at approximately )0.60 V(SCE). From other observations [8] in which relatively low potentials (<)0.70 V(SCE)) were achieved for a short period it is apparent that the more noble potential reects build-up of corrosion product on the specimen surface. In aerated pure water, the potential rose in a few days to quite noble values, eventually uctuating in the range +0.05 V(SCE) to +0.10 V(SCE). When chloride

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1247

Fig. 4. Variation of the corrosion potential with time for self-loaded disc specimens.

was present in the aerated solution, the corrosion potential increased more slowly and attained a lower potential than the chloride-free environment. It was noted in a similar test that movement of the specimen caused the potential to drop instantly. This transient disturbance of the potential to a lower value suggest that the lm is not a coherent passive lm but at least partially comprises a loosely adherent corrosion product, providing a transport barrier to metal dissolution. The slow gradual increase of the corrosion potential over a long period of time in the chloride solution has implications for testing, as the electrochemical driving force will be changing with time. This may have an inuence on crack initiation, which is likely to be potential dependent, and possibly also on the propagation rate in the short crack regime. As indicated below, all the specimens had pitted, in the case of the chloridecontaining solution, very extensively. In deaerated pure water, the pit development process is likely to be controlled solely by internal reactions, due to the absence of a signicant external cathodic reactant and because of the very low solution conductivity. The pits may then be growing at the free corrosion potential in the pit solution, with an internal cathodic reaction of water reduction or reduction of sulphide species from MnS inclusion dissolution. It would be expected that the potential in the pit would be low, with very large gradients in potential at the pit mouth, but not reected in the measurement of potential remotely. For aerated chloride-free water, the conductivity will be higher due to dissociation products of dissolved CO2 . Hence some inuence of the bulk oxygen on pit growth kinetics is anticipated but with the inuence diminishing as the pit depth increases. 3.2. Stress reduction due to specimen relaxation On completion of the tests, the specimens were removed from the solutions approximately 1 h after the heater was switched o; when the solution temperature had

1248

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

decreased to below 50 C. The specimens with their loading frames were washed immediately with high purity water, ethanol and acetone. Unloading was then undertaken on a mechanical test machine. The reduced stress, resulting from specimen relaxation, was estimated from the value of the load at which the nut of the loading frame could be readily unscrewed. For the deaerated and aerated chloride-free water some measurements indicated a decrease to about 80% of the proof stress (from 90%) whilst others showed no reduction, despite longer exposures. This variation could be correlated with the particular experimentalist so it is dicult to be unequivocal. However, it is likely that the decrease is relatively small. For the aerated 1.5 ppm Cl solution, it was less ambiguous; the nal stress decreased with test duration and at the longest exposure (9187 h) the nal stress was about 71% of the proof stress. In this case, the formation of multiple pits and cracks (see below) would be a factor as the change in average cross-section was only 1%. However, the relaxation in the chloride-free environments, with much reduced pitting and cracking, suggests that creep must be a factor. 3.3. Visual examination of the specimen surface After unloading, the specimen surface was carefully examined and photographed before and after chemical cleaning using Super Clarkes solution (10 g/l of 1,3-Din-butyl-2-thiourea (DBT) in 37.5% HCl). It had been established previously that this cleaning solution has negligible eect on the general corrosion of the bare disc steel and that there was no measurable change in the size of pits during the cleaning processes (the measurement resolution of the pit dimension was 1 lm). In the deaerated pure water, a shiny black lm had formed on the specimen surface and pits were readily visible after chemical cleaning. In aerated pure water a brownish lm was apparent on the surface and pits were readily visible. Also, there was evidence of crevice attack in the contact area between the specimen and the loading frame. In aerated 1.5 ppm Cl solution, severe general corrosion was evident with a thick reddish-brown corrosion product covering the surface. Large pits and multiple cracks could be discerned. In some cases, the tips of cracks from adjacent pits had turned towards each other and many cracks had coalesced, as shown in Fig. 5. 3.4. Distribution of the surface widths and depths of pits The surface width (w) and depth (a) of pits in the gauge section were measured using the same procedure as described earlier for pre-pits. Only pits of depth greater than 25 lm were recorded. Measurement of the number of pits of a depth below this value was carried out but was very uncertain and highly subjective because of the diculty of distinguishing very shallow pits from general surface roughening. In view of the possibility of distorting the pit density distribution, such data were neglected. The value of 25 lm was chosen based on condence in the measurement procedure. Since the primary interest was in the evolution of pits to cracks, very small pits depths were less relevant and their measurement added to the measurement burden unnecessarily. Nevertheless, it should be noted that the density of the

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1249

Fig. 5. Surface appearance of a self-loaded tensile specimen tested in aerated 1.5 ppm Cl solution at 90 C for 5107 h.

shallow pits with a depth less than 25 lm was high, usually greater than 25 cm2 . It might be expected that the density of pits with a depth greater than 25 lm would increase signicantly with time if all these pits kept growing. This proved not to be the case (see below), suggesting that the majority of these pits are non-propagating. The measurement of both the width and depth of pits was made only on selected specimens. Examples of the results for the three environmental exposure conditions in terms of the variation of aspect ratio with pit depth are shown in Figs. 68.

Fig. 6. Aspect ratio of pits as a function of pit depth in the self-loaded specimens in deaerated pure water at 90 C for 8322 h (no cracks initiated from pits).

1250

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Fig. 7. Aspect ratio of pits as a function of pit depth in the self-loaded specimens tested in aerated pure water at 90 C for 9138 h (no cracks initiated from pits).

Fig. 8. Aspect ratio of pits as a function of pit depth on the self-loaded specimens tested in aerated 1.5 ppm Cl solution at 90 C for 9187 h.

For comparison of the three exposure conditions, the results for the longest duration only are shown here. The number of pits in these three gures does not reect the pit density, as the number of specimens was not the same in all cases. For example, because of the very high density of pits in the chloride-containing solution, the results for only one specimen are shown. Nevertheless, the data for other specimens are reported elsewhere [9] and these conrm the representative nature of this particular dataset.

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1251

In deaerated pure water the exposure period was 8322 h rather than about 9000 h because of a system failure near the end of the test, so comparison with the other test conditions is not ideal. The comparison of the variation of the aspect ratio (note a value of 0.5 would be expected for a hemispherical pit) with pit depth shows that for the chloride-free environments, the aspect ratio increases with pit depth much more signicantly than for the chloride-containing pits. In the latter case, it is only when the pit depth exceeds the range 200250 lm that the mean aspect ratio exceeds 0.5. In chloride-free environments, the corresponding depth is of the order of 75 lm. In other words, the pits in the chloride-free water tend to be sharper for the same depth, extending in depth relative to opening much more than for the chloride-containing environment. The broadening of the mouth of the chloride-containing pits may reect the greater impact of the cathodic reaction externally and the more noble potential near the mouth, relative to the pit base, allowing enhanced corrosion. Also, the pH at the pit mouth might be expected to be lower for aerated solution as ferrous ions emerging from the pit are oxidised to ferric ions, which hydrolyses more eectively [10]. This would encourage dissolution near the mouth and associated broadening. The extent of that pH fall will depend on the corrosion rate and this may explain the distinction between the chloride-containing and chloride-free aerated water. There is still a basic question as to why a pit once initiated in the deaerated chloride-free water should grow faster at the base relative to the walls. If it were simply a case of corrosion of a lm-free surface in a neutral to alkaline local environment with internal water reduction, in other words simply corroding at a low rate in a lm-free condition with limited connection to the outside conditions, it would be hard to explain. However, one possible factor is dissolution of MnS inclusions. In that context, it may be envisaged that the tighter pit is favoured because it enhances retention of the local chemistry associated with this inclusion dissolution. 3.5. Pit densitydepth relationship The relationship between pit density and depth for the three exposure conditions is shown in Figs. 911. To avoid confusion, not all data sets are included, but details of maximum pit depths and average values are given in Tables 57. In deaerated pure water (Fig. 9), the maximum pit depth after about 22 months was just over 200 lm with a reasonable number of pits between 100 and 200 lm. The increase in the density of pits greater than 25 lm with the increase in exposure time was modest, less than 20%; the pits simply got deeper. Deductions from the data for the aerated water are more challenging because of the unusual degree of variability at dierent exposure times; e.g. the greater pit depths at 5107 h compared with 9138 h and the low density at 15,758 h. There was nothing in the record of the tests in terms of conductivity (monitored weekly) or any other factor that could account for such a variation. It could be related to the material characteristics but the specimens were prepared from the same region of the disc. Despite this variability, there is no doubt that the combination of oxygen and an increased conductivity (due to dissociation of dissolved CO2 ) would seem sucient to give an increased driving

1252

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Fig. 9. Distribution of pit depths on the self-loaded specimens tested in deaerated pure water at 90 C.

10 8 Pit density / cm-2 6 4 2 0 0 50 100 150 200 250 Pit depth / m

9138 hours 13528 hours 15758 hours

300

350

400

Fig. 10. Distribution of pit depths on the self-loaded specimens tested in aerated pure water at 90 C.

force for pit growth. Data are shown only up to 9187 h for the aerated chloride solution as the test was terminated because of the high level of damage to the specimens. The pit density was at least of a factor of two higher than for the chloridefree solution. The observation of deep pits in chloride-free water is not conned to this study. Similar observations had been made by Gabetta and Caretta [11]. However, David et al. [12] and Speidel et al. [1315] reported no pitting in high purity water, aerated or deaerated. The temperatures were similar to the present test. Also, David et al. used a 25-year old conventionally smelted steel, suggesting that steel composition does not provide a simple resolution of these dierent results.

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1253

Fig. 11. Distribution of pit depths in the self-loaded specimens tested in aerated 1.5 ppm Cl solution at 90 C.

3.6. Pit growth rates The key parameters from all the test specimens have been analysed and are also summarised in Tables 57. The evolution of pit depth (a) with time is often described by a relationship of the form: a at c 1

where a and c are system specic constants. The variability of the data for aerated water does not allow a detailed assessment of the pit growth rates whilst those for the deaerated solution were limited to two exposure times. Thus analysis is meaningful only for the 1.5 ppm Cl solution and even here the absence of data at short times is limiting. Nevertheless, using this relationship and tting to the data for the average of the 10 deepest pits for each specimen gives values of c 0:38 and a 1:36 106 with t expressed in seconds and a in m. With these parameters, pit growth rates were then estimated at each of the exposure times for the 1.5 ppm chloride solution and are included in Table 7. The values of course are smaller than the average growth rates of the ten deepest pits, which is to be expected as the latter incorporate the faster growth rates at shorter times. The average values were calculated only for comparison with Tables 5 and 6 for which data were limited. On this basis, the chloride-containing environment generated pit growth rates about 46 times faster than the chloride-free water. The pit growth rate in the aerated water was not markedly higher than for the deaerated pure water. The presence of oxygen in pure water might be expected to enhance the initiation of pits, and thus should give a higher pit density, but because of the very large potential drop in such high purity water, the growth rate could proceed independent of conditions of the bulk as the depth increased and should be

1254

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Table 5 Statistics of pits in disc steels in deaerated pure water at 90 C (the specimens were loaded nominally to 90% r0:2 ) Test time (h) 8322 Pit densitya (cm2 ) 14.2 16.7 15.5 16.5 18.8 Maximum pit depth (lm) 163 209 186 211 218 Average depth of 10 deepest pits (lm) 121 139 130 172 163 Average pit growth rateb (1012 ms1 ) 4.0 4.6 4.3 4.8 2.9

9979c 15,402

Results from various specimens are shown with the average value in bold. Data from a compact tension (CT) test is also shown. a Sampling area: 6.0 cm2 . b Average pit growth rate of 10 deepest pits assuming that the pit initiation time is negligible and dividing by the full exposure time. c Compact tension specimen in parallel test set-up.

Table 6 Statistics of pits in disc steels in aerated pure water at 90 C (the specimens were loaded nominally to 90% of YS) Test time (h) 5107 Pit densitya (cm2 ) 12.2 5.2 2.3 6.6 35.3 13.2 32.7 33.3 28.6 24.8 32.2 28.5 14.8 Maximum pit depth (lm) 194 261 209 221 117 155 177 221 168 274 386 330 433 Average depth of 10 deepest pits (lm) 157 180 155 164 105 135 123 133 124 244 323 283.5 249 Average pit growth rateb (1012 ms1 ) 8.5 9.8 8.6 9.0 3.2 4.1 3.7 4.0 3.8 5.0 5.7 5.4 4.4

9138

13,528

15,758

Results from various specimens are shown with the average value in bold. a Sampling area: 6.0 cm2 . b Average pit growth rate of 10 deepest pits assuming that the pit initiation time is negligible and dividing by the full exposure time.

comparable to that for the deaerated solution. However, comparison of such average pit growth rates can be misleading as Fig. 10 clearly indicates pit depths in aerated water signicantly greater than those in deaerated water.

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1255

Table 7 Statistics of pits in disc steels in aerated 1.5 ppm Cl at 90 C (the specimens were loaded nominally to 90% of YS) Test time (h) 5107 Pit densitya (cm2 ) 47.1 46.7 54.0 65.4 53.3 42.8 42.5 42.7 69.3 43.5 56.4 Maximum pit depth (lm) 825 999 792 829 861 925 939 932 1024 1209 1117 Average depth of 10 deepest pits (lm) 668 814 744 761 747 813 813 813 902 970 936 Average pit growth rateb (1011 ms1 ) 3.63 4.43 4.05 4.14 4.06 3.15 3.15 3.15 2.73 2.93 2.83 Pit growth ratec (1011 ms1 )

1.5

7173

1.2

9187

1.0

Results from various specimens are shown with the average value in bold. a Sampling surface area: 6.0 cm2 . b Average pit growth rate of 10 deepest pits assuming that the pit initiation time is negligible and dividing by the full exposure time. c Pit growth rate calculated using Eq. (1) and based on average depth of 10 deepest pits for each specimen.

3.7. Extension of pre-pits Since multiple pits developed readily from the smooth specimens in all environments pre-pitting to achieve reasonable depths turned out not to be a necessary requirement. Nevertheless, the nal depth of the pits that had evolved from pre-pits was measured, and the maximum depths are shown in Table 8. Comparing Table 8 with Tables 57, it would appear that the growth rate of the pre-pits is lower than those of the pits developed from a smooth surface, especially in aerated 1.5 ppm Cl solution. However, the comparison is valid only with respect to the same range of pit depths (i.e. between the initial pre-pit depth and the nal depth from the pre-pit), as

Table 8 Depths of pre-pits on the disc steel before and after testing at 90 C (initial stress: 90% r0:2 ) Environment Test duration (h) 8322 9138 7173 9187 Specimen no. 13 41 47 1 Pit depth (lm) Before 89 103 72 102 After 182 140 133 193 Pit growth rate (1012 ms1 ) 3.1 1.1 2.4 2.8

Deaerated high purity water Aerated high purity water Aerated 1.5 ppm Cl solution

1256

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

the pit growth rate is known to decrease with time. In that context, the dierence in growth rate for the deaerated solution does not seem signicant. Also, when account is taken for the distribution of pit depths in Fig. 11, rather than considering just the maximum value, it really is not possible to suggest any special features associated with the pre-pit. In pursuing this pre-pitting approach, the hope was that the pre-pits would grow preferentially and cracks might initiate more readily at these locations because of the greater depth. Crack measurement probes could then be located accordingly in future tests. However, not only did the pre-pits have a modest growth rate in aerated 1.5 ppm Cl solutions but cracks did not develop from these sites. For this environment, about 40% of pits had cracks at the depth corresponding to the extended pre-pit depths. It is feasible that with less than one or two pre-pits on these specimens, the statistics were just not favourable enough. 3.8. Pit-to-crack transition In aerated water, cracks were observed on specimens removed at 13,528 h (about 19 months) and at 15,758 h (22 months) but not on specimens removed at 9138 h (about 13 months). By contrast, no cracking was observed in deaerated pure water even after 22 months, although long crack growth is reported in such an environment [1,13,14]. Cracking was most extensive in the aerated chloride-containing water. The generally accepted phenomenological criteria for a pit to crack transition are that the threshold stress intensity factor has to be exceeded (the pit has to exceed a minimum depth) and the growth rate of the crack must exceed that of the pit [16]. For the tests with chloride, analysis of the data from all the tests is summarised in Table 9 and for aerated water, the results are shown in Table 10. In the case of very long cracks, the curvature of the specimen made reliable estimation of the crack length problematic using the normal optical methods of focusing on the specimen and adjusting the position laterally to measure the length. Thus, in Table 9, the true value will be greater than the quoted value. The extent of cracking in the aerated water and in the 1.5 ppm Cl solution is represented in Fig. 12. A key feature of Fig. 12 is the remarkable similarity in the percentage of pits with cracks at specic pit depths despite changes in exposure time and the presence or absence of chloride. Of course, the density of deep pits is greater in the 1.5 ppm solution. Thus, overall, cracking will be more extensive and the cracks larger in the 1.5 ppm solution. Crack coalescence will also be more signicant. The likelihood of cracking from a pit was described reasonably by a Weibull distribution function: F a 1 expfa a0 =g a0 b g 2

where a0 is the threshold pit size, g is the scale parameter (equal to the value of a at which F a is 0.632) and b is the shape parameter, the value of which reects the extent to which the probability of crack initiation increases or decreases with time. In the case of pits, the probability of the transition to a crack increases with pit depth

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1257

Table 9 Crack density in disc steels in aerated 1.5 ppm Cl at 90 C (the specimens were loaded to 90% of YS) Test time (h) Crack densitya (cm2 ) Average crack length (lm) Average surface crack growth rate (1010 ms1 ) Maximum crack length (lm) 2397 2230 2591 3421 2660 >3662 >4528 >4363 >5240 Maximum surface crack growth rate (1010 ms1 )

5107

11.5 21.3 28.7 21.0 20.6 27.7 22.7 25.2 38.3 28.2 33.3

1157 1085 985 1075 1076 >1479 >1792 >1684 >2110

0.6 >0.5 >0.7 >0.5 >0.6

1.4 >1.4 >1.7 >1.3 >1.6

7173

9187

Results from various specimens are shown with the average value in bold. a Sampling surface area: 6.0 cm2 .

Table 10 Statistics for cracks in disc steels in aerated pure water at 90 C (the specimens were loaded to 90% of YS) Test time (h) Crack densitya (cm2 ) Average crack length (lm) Average surface crack growth rate (1011 ms1 ) N/A N/A N/A 4.99b 2.33b Maximum crack length (lm) N/A N/A N/A 1619 1169 Maximum surface crack growth rate (1010 ms1 ) N/A N/A N/A 1.02b 0.49b

5107 9138 13,528 15,758

No crack No crack No crack 5.33 1.33

N/A N/A N/A 789 556

Sampling surface area: 6.0 cm2 . Assuming crack initiation time is 91,388 h.

and this is consistent with the calculated value of 1.5 for b. For g, a value of 295 lm was determined. The value of g would be expected to be stress-dependent and conceptually would be expected to depend on the nature of the environment. In that context, the overlap of the data for the chloride-free and chloride containing solution was surprising. Nevertheless, recent measurements [8] of the potential at the tip of a crack indicated no distinction between an aerated solution containing chloride and a chloride-free solution, although there was a distinct eect of oxygen concentration. This may be a possible explanation for the similarity of the results of Fig. 12, although some caution is necessary in extrapolating data from deep (6 mm) cracks to relatively shallow pits. However, if the Weibull relationship is combined with the pit density distribution to predict the likely number of cracks on the two specimens in aerated water removed at 9138 h, a value of 17 is obtained. In practice, no supercial

1258

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Fig. 12. Likelihood of cracks developing from pits as a function of depth.

cracking was observed but more detailed evaluation is being conducted to check for sub-surface cracks perhaps initiating from the base of pits but yet to break surface. However, since some deep pits were observed at 5000 h with no cracking also, it perhaps suggests that there is some other factor that is inuencing the transition that can be related to exposure time. The same predictive approach for the likelihood of cracking can be used for the results in the deaerated solution, for which no cracks were observed, since it could be argued that the pits were simply not deep enough in this case to make the transition to a crack. In this case, the probable number of cracks to be expected in the specimens tested in the deaerated solution was 22, assuming the same Weibull function parameters. It could be argued that the behaviour for the deaerated solution is distinct from that of the aerated solution, particularly in the light of the crack-tip potential measurements [8]. However, in view of the absence of cracks in the shortterm aerated chloride-free tests, too strong a conclusion would not be merited. Longer-term tests in the deaerated pure water would be required to resolve this but would be dicult to justify since this environmental combination of pure water and deaeration is now recognised to be not so practically relevant. 3.9. Crack growth rates As indicated, the pit to crack transition beyond the threshold value has been associated with the condition that the growth rate of the crack should be greater than that of the pit [16]. The crack growth rate can be estimated approximately from measurement of the crack depth by sectioning of the specimen and by assuming initiation on initial exposure. Work on this aspect is in progress. Table 11 shows preliminary estimates for the maximum crack depth derived from four longitudinal slices on specimens tested in aerated 1.5 ppm Cl solutions for 5107, 7173 and 9187 h. The maximum crack growth rate was calculated by dividing the maximum depth by

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1259

Table 11 The depth and growth rate of cracks initiated from pits in the disc steel tested in aerated 1.5 ppm Cl solution at 90 C Test time (h) 5107 7173 9187 Maximum crack depth (lm) 820 960 1030 Maximum growth rate (1011 ms1 ) 4.5 3.7 3.1

the total exposure time, ignoring any initiation period, and is also shown. (Of course, from a practical viewpoint, the concern over initiation time may be academic since it is the growth rate of the aw (pit plus crack) that is important). The estimated crack growth rates are about a factor of 3 less than the surface crack growth rates (Table 9). They are consistent with the average pit growth rates for the deepest pits (Table 7). For this steel, the stress corrosion growth rates of long cracks were determined previously [17] using FM specimens with a 6 mm pre-fatigue crack immersed in the same aerated 1.5 ppm solution. The test period was 7176 h. The extension in crack depth was measured by fracturing the specimen at the end of the exposure period rather than by in-situ crack depth monitoring. The crack grows only about 250 lm over this extended period and the stability in the crack depth measuring system was not adequate, although this has now been overcome with a new system capable of stable 1 lm resolution. Assuming the cracks initiated on initial exposure, the crack growth rates were estimated to be 8.9 1012 and 1.0 1011 ms1 for static loads of 40 and 50 MPa m1=2 respectively. The uncertainty in the time to initiate a crack and some dierences in the rate of change of corrosion potential with time for the two types of specimen means that comparison of growth rates from the tests on plain specimens and from the fracture mechanics specimens should be treated with caution. With that caveat, the indications are that the growth rate of the short crack initiated from pits is about four times faster than that of the long crack. 3.10. Crack morphology A typical etched cross-section of the specimens tested in aerated 1.5 ppm Cl solution is shown in Fig. 15. Examination of a number of such sections indicated that the crack was predominantly transgranular, although there was some evidence of intergranular cracking (Fig. 13). 3.11. Pit and crack proles A detailed study has been conducted on the proling of pits and cracks using a layer removal technique and an image analyser. For the self-loaded specimen exposed to 1.5 ppm solution for 5107 h, longitudinal slices were removed along the gauge length in controlled intervals of approximately 50 lm.

1260

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Fig. 13. The etched cross-section surface of specimens tested in aerated 1.5 ppm Cl solution at 90 C, showing predominantly transgranular cracking.

The individual images of each exposed new section are shown in Fig. 14. For a tensile specimen, the width of the slice will increase progressively from the top surface. As described above, the transition from a pit to a crack is a probabilistic process with the likelihood increasing with pit depth. Hence, there will be some deep pits from which no crack develops (e.g. Pit A). It is also interesting to note that the depth of some cracks is smaller than that of the corresponding pit (e.g. Pit B), suggesting that cracks do not necessarily initiate from the bottom of the pits. The alternative explanation that the pit subsequently grew faster seems less likely because corrosion blunting might have been anticipated. On the other hand, there were many cracks with a depth greater than that of the corresponding pit (e.g. Pit C), but it is dicult intrinsically to ascertain whether the cracks initiated at the bottom of the pits or the initiated crack progressed beyond the depth of the pits. The possibility that the crack does not initiate from the base of the pit has implications for the prediction of the pit to crack transition using fracture mechanics concepts, as the stress intensity factor is usually based on treating the pit as an equivalent crack of the same depth. An added complication is the probabilistic nature of the transition as evidenced by Fig. 12. There is also clear evidence that some pits and cracks coalesced (e.g. Area D). 3.12. MnS inclusions It is evident that anionic species in solution are not required to cause pitting attack in this steel. Inclusions are likely to be involved, of which the most aggressive and plentiful is MnS. The MnS inclusion distribution in the disc steel was analysed by sectioning a cylindrical specimen longitudinally. This was then mounted in phe-

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1261

Fig. 14. Microphotograph of the cross-section surface using the layer removal technique, constructed with an image analyser.

nolic resin such that the curved surface of the gauge length was visible. The surface was then ground and polished such that a plane surface was obtained in which the MnS inclusions could be clearly seen using optical microscopy. For analysis of the MnS size and size distribution, 56 digital images were obtained using a camera attached to an optical microscope. Each image covered an area of 0.021 mm2 . The area of each inclusion was measured and the diameter of the

1262

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

Fig. 15. The distribution of the equivalent circle diameter of MnS inclusions in 3% NiCrMnV disc steel. Table 12 Statistics of MnS inclusions in the 3% NiCrMnV disc steel Diameter of the equivalent circle (lm) Maximum 15.3 Minimum 0.37 Average 1.72 1.94 Density (cm2 ) 2.24 104

inclusion was estimated by tting it to a circle (the diameter of the equivalent circle). The size distribution obtained from these images is shown in Fig. 15 and Table 12. The density for this rather old steel is very high indeed and would provide plentiful sites for initiation of pitting and indeed for perhaps continuing to inuence the propagation rate and shape for the reasons articulated above.

4. Conclusions Pitting was observed under all exposure conditions with the pit density and growth rates increasing generally in the order: deaerated pure water; aerated pure water, and aerated 1.5 ppm chloride-containing water. Pits formed in the pure water tended to be tighter with a higher depth to mouth opening ratio than for the chloride-containing water. This is rationalised on the basis of retention of local pit solution chemistry inuenced by MnS inclusions for the pure water bulk solution and a combination of mildly acidic pit environments and potential drop considerations in the chloride-containing solution.

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

1263

Extensive stress corrosion cracking from pits had initiated in the chloride-containing solution at exposure time less than 7 months. In aerated water, cracking was observed at 19 months but not at 13 months. A minimum pit size of 50 lm for initiation of cracks was identied. However, no cracking was observed in deaerated pure water after 22 months despite pit depths of up to 200 lm. There was no evidence of cracking in the absence of pits. The probability of a crack initiating from a pit of specic depth in aerated solution was surprisingly similar for the 1.5 ppm chloride solution and the chloridefree water at long exposure times. The transition from a pit to a crack was well described by a Weibull cumulative distribution function. Of course, the density of deep pits is greater in the 1.5 ppm solution. Thus, overall, cracking will be more extensive and the cracks deeper in the 1.5 ppm solution. The growth rate of short cracks initiated from pits in aerated 1.5 ppm Cl solution was about four times greater than that of long cracks determined using FM specimens with a 6 mm pre-crack but in both cases there was uncertainty in the time to initiate the crack. The cracking mode in aerated 1.5 ppm Cl solution was identied as predominantly transgranular. Proling of the pits and cracks by progressive layer removal and imaging indicated that whilst the crack depth usually exceeded the pit depth there were examples where the pit depth extended signicantly beyond the crack; i.e. the crack may not have initiated from the base of the pit. In that circumstance, simple application of fracture mechanics principles based on the pit acting as an eective crack of the same depth would not be pertinent. It should be emphasised that these measurements of pit and crack growth rates are based on continuous exposure and in the case of the 1.5 ppm chloride under very severe conditions. More work is required to translate these measurements into likely growth rates in service, for which the exposure conditions will be transient and in the case of the severe conditions, short-lived.

Acknowledgements This work was conducted as part of the Life Performance of Materials programme, a joint venture between the United Kingdom Department of Trade and Industry and an Industrial Group comprising Alstom, BNFL Magnox, British Energy, Innogy, Powergen, and Siemens. The authors are grateful to Angela Mensah, Linda Orkney and Gareth Hinds for assistance with the experimental work.

References
[1] S. Zhou, A. Turnbull, Steam Turbine Operating Conditions, Chemistry of Condensates and Environment Assisted CrackingAn Overview, NPL Report MATC (A) 45 (2002). [2] ASTM, E 8-94a, Standard Test Methods for Tension Testing of Metallic Materials, 1994.

1264

A. Turnbull, S. Zhou / Corrosion Science 46 (2004) 12391264

[3] British Standard, BS ISO 4287, Geometric Product Specication (GPS)Surface TextureTerms, denitions and surface texture parameters, 1997. [4] S. Zhou, A. Turnbull, Development of a pre-pitting procedure for turbine disc steel, British Corrosion J. 35 (2) (2000) 120. [5] W.L. Marshall, J. Chem. Eng. Data 32 (2) (1987) 221226. [6] P.E. Francis, A.D. Mercer, Corrosion of a mild steel in distilled water and chloride solutions: Development of a test method, in: G.S. Hayne, R. Baboian (Eds.), Laboratory Corrosion Tests and Standards, ASTM STP 866, American Society fro Testing and Materials, Philadelphia, 1985, pp. 184 196. [7] A.D. Mercer, G.M. Brook, La Tribune du, CEBEDEAU 31 (417418) (1978) 299306. [8] A. Turnbull, S. Zhou, G. Hinds, Corros. Sci. 46 (1) (2004) 193211. [9] A. Turnbull, S. Zhou, Eect of Oxygen and Chloride on Pitting and Stress Corrosion Cracking of a Steam Turbine Disc Steel, NPL Report MATC (A) 133, 2003. [10] C.F. Baes, R.E. Mesmer, The Hydrolysis of Cations, John Wiley and Sons, New York, 1976. [11] G. Gabetta, E. Caretta, Investigation on pit susceptibility in low pressure rotor steels, in: Proceedings of Materials for Advanced Power Engineering, 1994, 36 October, Kluwer Academic Press, Dordrecht, 1994, pp. 209218. [12] W. David, G. Rottger, K. Schleitho, H. Hamel, H. Termuehlen, Disk-type LP turbine rotor experience, PWR-Vol 21, The Steam Turbine Generator Today: Material, Flow Path Design, Repair and Refurbishment, ASME 1993, pp. 8391, 1993. [13] M.O. Speidel, R. Magdowski, in: 2nd International Symposium Environmental Eects on Degradation in Nuclear Power Plants, Monterey, 912 September 1985, pp. 267275. [14] M.O. Speidel, J. Denk, B. Scarlin, Stress corrosion cracking and corrosion fatigue of steam-turbine rotor and blade materials, Cost Project, EUR 13186 EN, 1991. [15] M.O. Speidel, B. Scarlin, Stress corrosion cracking and corrosion fatigue of steam-turbine rotor and blade materials, in: Proceedings of High Temperature Materials for Power Engineering, Liege, Belgium, 2427 September 1990, p. 343. [16] Y. Kondo, Corrosion 45 (1989) 711. [17] S. Zhou, A. Turnbull, Environment assisted cracking of steam turbine disc and blade steels description of test procedure and preliminary results, NPL Report, MATC(A)41, 2001.