Historical Review of Alloy 600 Stress Corrosion Cracking: From the “Coriou Effect” to the Quantitative Micro-Nano Approach

Nuclear energy is a carbon-free source of power and a meaningful option in the context of global warming. At the end of 2016, 448 commercial nuclear reactors were connected to the grid, including 289 pressurized water reactors (PWR). Sixty-one other reactors were under construction (including 51 PWRs). The world nuclear electricity supplied in 2016 reached more than 2,400 TW(e) h which accounts for 19.7% of the electric power generation in the world.¹  Current stakes are high for designers and operators: third and fourth generations of nuclear power plants (NPP) are expected to last 60 y and even more, which is twice the initial time for the previous reactor generation. This is possible thanks to operational feedback and to the research efforts in various areas, including the behavior of structural materials in their environments. Since the 1950s, time and money have been spent to select the structural alloys in nuclear environments, develop prediction and mitigation of corrosion phenomena in NPP to prevent failures, and increase safety and to optimize operation time of these reactors. Over time, progress led to better understanding, prevention, modeling, and monitoring of the interactions between alloys and their environments.

Corrosion resistant alloys such as stainless steels and nickel-based alloys have been in use as the first NPP designs. Nevertheless, NPPs suffered corrosion failures, which became one of the major concerns. Stress corrosion cracking (SCC) has been one of the main challenges as the beginning of commercial operation of light water reactors: the first cracks on stainless steels welds occurred during the 1970s in boiling water reactors. For PWRs, Alloy 600, a nickel-based alloy, suffered SCC on tubes of steam generator (SG) since the 1980s and at vessel head penetration in the 1990s.^2-7 

The “Coriou effect” describes the SCC of this nickel-based alloy, Alloy 600 (which contains around 75% nickel and 15% chromium), in pure water at high temperature (or PWSCC for pure-or primary- water stress corrosion cracking). Henri Coriou and his co-workers published in 1959 that Alloy 600 cracks in pure water at 350°C.⁸  The “Coriou effect” was used with some irony and skepticism until the early 1980s, when a growing number of cracks were observed on SG tubes of PWRs under nominal operation conditions. Nearly 70 y after the first publication of these results, it is interesting to understand why Alloy 600 has nonetheless been widely used in light water nuclear reactors up to the 1980s. Some of the last developments and results observed on the mechanisms will follow the historical perspectives that constitute the first part of this paper.

Commercial nuclear power reactors grew out military nuclear reactors developed in the U.S. first for submarines and then for surface ships. Rapidly, water-cooled design and more precisely PWR technology was chosen for nuclear powered ships and at the same time, SCC started to be a major issue: in 1953, SCC occurred in the stainless steel tubing of SG of the prototype for the Nautilus.²  SCC failures occurred on the outside surfaces of the tubes. At that times, “in most cases (of stress corrosion failures), chlorides have been considered as the principal offenders.”⁹  The results in the literature showed that the increase in nickel content of Fe-Cr-Ni alloys decreased their susceptibility to SCC¹⁰ ; the choice for SG tubing was thus an alloy containing more than 72%Ni, 15%Cr to 17%Cr, and 6% Fe to 10% Fe (Alloy 600; Table 1) which was known for its oxidation resistance in hot gases. Most literature data were obtained with U-bend specimens tested in boiling 42% magnesium chloride (154°C) which is a well-known laboratory test for SCC studies of stainless steels. Figure 1 illustrates the “Copson diagram” showing the beneficial effect of an increase in nickel content of Fe-20%Cr-Ni alloys on their resistance to stress corrosion. This work showed clearly that SCC of Fe-Cr-Ni alloys would stop in boiling MgCl₂ when the Ni content is above 40%. So, it was deduced that Alloy 600 was the most suitable industrial alloy for SCC resistance, as its Ni content was very high (more than 72%).

It is quite interesting to remark that, even in the middle of the 1950s, two cases were known and reported, “in which failure of nickel-based alloys was attributed to SCC under conditions similar to those encountered in NPP”⁹ : in vapor at nearly 400°C (750°F) and in “hot water”, etc. In these two cases, cracking was intergranular in nature and discussions took place regarding the interpretations of these failures.^9,11 

To check the behavior of Alloy 600 under more realistic conditions than magnesium chloride solution, the CEA corrosion laboratory launched tests in the coolant of naval reactors (high-temperature pure deoxygenated water) and accelerating the SCC phenomena by using higher temperatures and plastic deformation. Tests were performed in high-temperature pure water (350°C) in an autoclave with mechanically loaded specimens (bending). The first observations showed obvious intergranular cracking (Figure 2) and intergranular oxidation (Figure 3) after 3 months of exposure. These observations are similar in every aspect to those observed in later years from both operating experience and laboratory studies. The publication of these results⁸  by Henri Coriou in 1959 sharply agitated not only the corrosion community but also designers and manufacturers of water reactors and tube suppliers. Many researchers followed up this study but did not succeed, at least initially, in reproducing the Coriou cracking in pure or simulated PWR water, mainly because the applied stress was much higher in Coriou’s experiments (plastic deformation) than in the other tests (more often at 90% of yield stress).^2,12 

In 1965, H.R. Copson and S.W. Dean published¹³  that “applied stress (…) did not promote SCC or intergranular attack in Alloy 600 in simulated primary and secondary reactor waters at temperatures between 300°F (149°C) and 680°F (360°C).” However, they showed that lead (Pb) contaminant produces SCC of Alloy 600 under these conditions.¹³  Therefore, according to them, it was obvious that the Coriou’s results were due to poor experimental conditions and in particular to a water pollution by lead. The presence of dissolved oxygen and too high stresses were also claimed as parameters leading to nonrepresentative results.

Taking the objections one by one, the CEA corrosion laboratory started a series of tests with not only water chemical analysis, but also determination of the lead content of autoclave steels, of experimental device alloys, and of alumina rods used to isolate the test specimen. Then replacing the alumina rods by zirconia ones when it was claimed that alumina might play a role in SCC of Ni-base alloys, etc. Alloy 600 specimens still cracked intergranulary (but not stainless steels nor Alloy 800 specimens that have been in the same autoclaves).¹⁴  The effect of nickel content was determined by another series of tests with model Fe-Cr-Ni ternary alloys containing a constant chromium content (18% Cr) and a variable nickel content. Tests were performed for 3 months to 6 months in 350°C pure water and in water with 0.1% NaCl. In 1967, Coriou and coworkers¹⁵  published that cracking in pure water at high temperatures occurred only when the nickel content was higher than 60% to 70%, as shown in Figure 4. In particular, Alloy 800 (20% Cr, 35% Ni, balance Fe) was not susceptible to SCC with or without chloride. Alloy 800 was then recommended for the tubing of SG,⁷  and used in some.

With time, after the period of strong the controversy between Coriou and Copson, SCC was broadly observed in Alloy 600. SCC was detected on SG tubes made of mill-annealed (MA) Alloy 600 at Obrigheim reactor in 1972. In about 1978, it was obvious that SCC of MA Alloy 600 was becoming a generic issue. All the SG containing MA Alloy 600 have been replaced today in France and in many countries. SCC mitigation included evolution of the design (hydraulic expansion of the tube on the tube-plate for instance, or procedures of setting mechanical compression in the coldest worked zones like transition of expended zones—various kinds of shot-peening). For the tubing of new SG, thermally treated (TT) Alloy 600, which was less susceptible to SCC, was used at the beginning of the 1980s. Then it was decided to replace Alloy 600 by alloys much more resistant to SCC: Alloy 690 (around 60% Ni, 30% Cr, and 10% Fe) or by Alloy 800 (33.5% Ni, 21.5% Cr, 47% Fe), used from the beginning by German. TT Alloy 690 is the material used today in France for PWR SG tubes. It should be mentioned that the Coriou effect appeared not only on SG tubes, but also on various components containing Alloy 600 (or similar alloys), in particular weld joints (Alloy 82 and 182, vessel head, pressurizer, etc.). This led also to the replacement of many vessel heads. The elimination of Alloy 600 from all the components is the price to be paid to put an end to this saga of the Coriou effect, which in the end has been very expensive for the nuclear industry.

The SCC data obtained by Coriou and Copson are based on thin specimens and are relevant to components such as tubes in SG that have thin walls. In the reviews of SCC of Alloy 600 made at the beginning of the 2000s, virtually all the data were taken from these types of specimens.^16-17  However, with thicker walls, such as SG divider plate or dissimilar metal welds of main coolant pipes, it is necessary to get data on crack velocity or propagation rates.

In fact, a schematic overview of the progress of SCC includes two main segments: “initiation” and “propagation.”^2,18  Initiation applies to processes that occur on the surface before SCC penetrates; initiation is a property of initially smooth surfaces. Propagation applies to the progressive growth of SCC into the bulk Alloy 600. Initiation involves much longer times than propagation (very often much more than 10 y in PWR primary water). The initiation of SCC is dependent of initial conditions of manufacturing and handling processes where surfaces have been polished, cleaned, scratched, ground, etc. Initiation of Alloy 600 SCC is generally divided into three main steps.

• •

  A precursor step that involves the formation of the oxide layer and changes that may involve preferential dissolution of active species such as Cr or may involve the deposition of species such as Pb.² 

• •

  Incubation is the step where the evolution of the oxide layer continues and that includes the starting of grain boundary oxidation and penetration.^2,18 

• •

  Protocracks are the beginnings of small cracks in the range of a few microns or less²  that are able to sustain growth.

The division between initiation and propagation is dependent of the observer: from an engineering point of view, it is often taken as about 50 μm but this value is not related to fundamental processes. This value is mainly based on an approach developed by Andresen and Ford to define adequate resistance to SCC and which focus on cracks that initiate early and grow actively.^3,19  This led to base the SCC susceptibility mainly on crack growth rates. As very low crack growth rates can be detected with the modern measurements, the concepts of “SCC immunity” and of “thresholds values” are becoming less and less reliable.^20-21  This approach is particularly relevant to structural materials where early cracks and their growth rates must be understood and not the overall distribution of emerging cracks. The situation is quite different for SG tubes where few failures can be tolerated and the evolution of the number of leaker tubes is important.

There is no model that can predict the occurrence and rate of SCC on nickel-based alloys a priori. All SCC modeling is phenomenological: today prediction of SCC is based on what has happened in the past (statistical or semiempirical approach) in the field or in the laboratory. Such information is then interpolated and extrapolated by established approaches of dependences of temperature (1/T), stress (σ⁺ⁿ), environment composition (c⁺ⁿ), pH, electrochemical potential (E), flow, grain size (d⁻ⁿ), and similar phenomenological variables.

The essence of the quantitative micro-nano approach (QMN approach) started by R.W. Staehle in 2010 is “to identify discrete atomic level processes that can participate in SCC, describe them quantitatively, and then synthesize these separate mechanistic parts into coherent models of increasing complexity, which are capable of predicting the occurrence and rate of SCC.”¹⁸  This QMN approach will be illustrated in the following paragraphs with recent developments performed at CEA corrosion laboratory with the use of atomistic simulation and of tracers (deuterium and oxygen 18) to better know the fundamentals of SCC mechanisms.

A possible role of hydrogen in SCC of Alloy 600 in primary water at 300°C to 350°C is generally discarded for two reasons.²²  On one hand, these temperatures are too high to promote hydrogen embrittlement as they discard low energy trapping. On the other hand, assuming that the hydrogen fugacity is controlled by the concentration of dissolved hydrogen in the primary water (25 cm³Kg⁻¹ to 50 cm³Kg⁻¹ at 320°C) the corresponding hydrogen activity in the alloy beneath the surface is assumed too low to give rise to sufficient trap filling to promote hydrogen embrittlement. However, dissolved hydrogen is not the only source of hydrogen available in primary water: the hydrogen coming from the cathodic reaction (dissociation of water) associated with the oxidation reaction (anodic reaction) on the alloy surface also has to be taken into account. In order to identify the main source of hydrogen, isotopic tracing was used and two experiments were conducted with deuterium for Alloy 600 exposed to primary water.²³  Alloy 600 samples were exposed, under the same conditions, to primary water made with heavy water with dissolved hydrogen (D₂O/H_2,diss) or to natural water with dissolved deuterium (H₂O/D_2,diss). After a rapid cooling of the samples, SIMS analysis (secondary ion mass spectrometry) of ²H and ¹⁶O were done to determine the deuterium concentration profiles together with the oxide film thickness on the alloy surface. Almost no deuterium is observed for the H₂O/D₂ environment and only in the oxide layer (Figure 5[a]), whereas the intensity of the deuterium profile is much larger in D₂O/H₂ (Figure 5[b]) with deuterium observed not only in the oxide layer but also in the alloy showing a deuterium profile representative of a diffusion controlled absorption. Clearly, the main source of hydrogen is the cathodic reaction (water dissociation) and so it is associated to the oxidation of the alloy (anodic reaction).²⁴ 

Two mechanisms may be proposed for modeling the hydrogen transport associated with the oxide growth during alloy passivation: (i) diffusion of hydrogen as an interstitial proton through the oxide lattice, or (ii) diffusion as a hydroxide ion towards the oxide in the anionic sublattice. The latter hypothesis implies the oxygen and hydrogen diffusivities through the oxide layer to be the same. To check which hypothesis is correct, Alloy 600 specimens have been exposed in PWR primary conditions using ²H and ¹⁸O as markers. The values obtained for diffusion coefficient of ²H and ¹⁸O are very close (around 5 × 10⁻¹⁷ cm²/s) which supports the idea of a hydrogen transport mechanism through the oxide layer as hydroxide ions.²⁵ 

The strong correlation between hydrogen absorption and oxidation occurs not only for the formation of the oxide layer on the surface of the alloy, but also during intergranular oxidation of grain boundaries. The question here is to assess whether the oxide grown at the grain boundaries in the case of intergranular corrosion would act as a barrier to hydrogen arrival to the oxide/crack tip or not. After a primary oxidation in nominal primary water (1,340 h) followed by a short period under the same conditions but with isotopic tracers (²H and ¹⁸O), deuterium and oxygen 18 are found at the tip of the intergranular oxidation, even for short exposure times (see Figure 6). This suggests that defects, voids, or cracks could accelerate the hydrogen and oxygen transport within the oxidized grain boundary. The results lead to the conclusion that oxygen and hydrogen transport in the oxidized grain boundary are not the rate-controlling step for SCC initiation in PWR nominal conditions.^24-25 

The oxidation of grain boundaries or slip bands is supposed to be the preliminary step for crack initiation in PWR primary water. Then, crack propagation occurs by successive cracking of the oxide penetration. Nickel oxide is observed in the crack and at the crack tip but nanometric chromium oxide is detected at the end of the crack tip, in association with the formation of a chromium-depleted area aside and ahead of the tip. The formation of Cr₂O₃ coupled with asymmetric Cr-depleted zone adjacent to the oxide characterizes main propagating cracks.²⁶  The chromium-depleted zone extends in only one grain adjacent to the grain boundary and along the grain boundary itself. Based on these observations, the rate-controlling step for SCC crack propagation is probably linked to the chromium diffusion in the alloy. The apparent chromium diffusion coefficient extrapolated form the literature data (either volume or grain boundary diffusion) is far too low to explain the formation of the depleted areas at about 300°C.²⁷  With ⁵²Cr and ⁵⁴Cr, acceleration of chromium diffusion in Ni single-crystals was shown under plastic deformation during creep tests. A linear dependency is observed between measured apparent diffusion coefficients and strain rates during secondary creep. Furthermore, diffusion-induced grain boundary migration (DIGM) was evidenced on model Alloy 600 after exposure to primary water at 360°C. Both mechanisms (DIGM and plasticity-enhanced diffusion) may explain the formation of the Cr-depleted zone.^27-28 

Atomistic simulations have been performed to describe the sequence of phenomena occurring at the crack tip of an SCC fracture. Molecular statistics simulations with embedded atom method potentials have been performed on a nickel bicrystal with and without hydrogen at the grain boundary. Ni atoms located on the crack flanks are subjected to a force from the oxide growth. This mechanical loading tends to spread the crack flanks which is the focus of the ongoing simulation. The oxide growth at the crack flank is modeled with a numerical force impressed on the crack flanks. Without H in the grain boundary, dislocation emission is observed, as shown in Figure 7(a). When H is present, the crack propagation is brittle which implies that no dislocation emission occurs, neither defect nor amorphization (Figure 7[b]). These atomistic simulations have shown the detrimental consequence of H in the grain boundaries.²⁹ 

Interactions of a nickel substrate with water molecules have been investigated using reactive force field molecular dynamics simulations.³⁰  After the hydration of nickel, the dissociation of water molecules is the first stage observed (H₂0 → OH⁻ + H⁺). The OH⁻ is adsorbed forming a thin hydroxide layer and the H⁺ is bonded with another water molecule to form H₃O⁺. The last stage includes the penetration of oxygen and hydrogen coming from the OH⁻ into the nickel substrate, as shown in Figure 8. These simulations show also that the crystal surface orientation is an important factor in the oxidation: the oxidation kinetics (and the penetration of hydrogen in the metal) is higher on (110) orientation than on (111) or on (100). All these data are consistent with the results obtained with tracers.³¹ 

Other atomistic simulations were performed to predict the behavior of the passive layer on nickel-based alloys, as this passive layer play a major role in the passivity of the alloy and thus on SCC initiation. Ni_1 – xFe_xCr₂O₄ spinels have been identified as part of the corrosion layer on nickel-based alloys in PWR.³²  The investigation of the thermodynamic properties and grain boundaries properties of the NiCr₂O₄-FeCr₂O₄ system has been done in the framework of molecular dynamics using fixed charged empirical potential. Detailed analysis of the grain boundaries structure shows that the cation coordination number (x in Ni_1 – xFe_xCr₂O₄) is a key parameter in the stability of the grain boundaries. The structural and energetic differences are caused only by nickel and iron cations. The grain boundary energy is shown to be also a function of the cation coordination number (and of the grain boundary distance).^32-33 

• •

  Nickel-based alloys are still widely used for PWR components. “Coriou cracking” has been detected in nearly all the components including dissimilar metal weld joints made of Alloy 600 and its welding alloys. However, after the development of new alloys like Alloy 690 or new heat treatments for alloy X-750, “Coriou cracking” seems to have been solved practically. Nevertheless, Alloy 600 and other similar susceptible materials are still in service in PWRS, and the immunity of new materials has to be based on the understanding of SCC mechanisms. In that sense, the QMN to predicting SCC of Fe-Cr-Ni alloys, which Roger W. Staehle has proposed is the way to follow: the QMN approach emphasizes qualitative and quantitative descriptions of mechanistic elements within SCC initiation and propagation, at various scales, including atomistic dimensions in the environments of water-cooled reactors. Of course, lot of works is still needed to develop a comprehensive understanding of SCC in PWR conditions.

• •

  The cost to nuclear energy not to have taken into account the Coriou’s results has been very high. These results were contrary to the dominant ideas regarding corrosion of materials in the 1950s: nickel alloys with high nickel content were supposed to be very resistant to corrosion and pure water was supposed to be a noncorrosive environment (only solutes or impurities should have a detrimental effect). Even today, when laboratory results are in disagreement with the global consensus, bad experimental conditions are claimed to exclude these results. Those who have to face unexpected results are encouraged to check them carefully and rigorously, and to follow Henri Coriou’s example.

The authors thank the European Federation of Corrosion (EFC) for permission to use the figures of the EFC publications number 67.