Sei sulla pagina 1di 25

Stainless Steels Aged at Intermediate Temperature: A Review

F. Danoix and P. Auger

Atom Probe Studies of the FeCr System and

Groupe de Mtallurgie Physique, U.M.R. C.N.R.S. 6634, Sonde Atomique et Microstructure, Facult des Sciences, Universit de Rouen, Place E. Blondel, 76 821 Mont Saint Aignan Cedex, France The spinodal decomposition is known to occurr in ironchromium-based model and industrial alloys during long-term aging at intermediate temperatures. This decomposition can be characterized unambiguously by atom probe, and its evolution quantied by means of various microstructural paramaters directly accessible from concentration proles. Comparison with the existing theories of spinodal decomposition will be briey examined. This decomposition has a direct inuence on the material, and can be used to improve magnetic or mechanical properties. Conversely, some mechanical properties, such as impact toughness, are greatly degradated by this evolution, and can lead to serious problems when, for example, the integrity of large industrial components is affected. For this reason, extensive work has been carried out to understand the inuence of the various alloying elements on the phase separation and on the embrittlement process. Atom probe contributions in this area will be presented in this paper. Elsevier Science Inc., 2000. All rights reserved.

INTRODUCTION Stainless steels are one of the most commonly used class of domestic and industrial alloys. In addition to their intrinsic corrosion resistance, those steels can exhibit combinations of high strength, ductility, weldability, castability, etc., according to their structure. These steels can be found with several crystallographic structures, that is, austenitic, ferritic, austenoferritic (duplex), and sometimes martensitic. Among the various applications of stainless steels are boilers in the chemical and nuclear industries, distillation columns, heat exchangers, or pipes. For such applications, they face a severe problem of embrittlement because their service temperature lies in the range of 300500C. After long-term service, loss of impact toughness and duc177
MATERIALS CHARACTERIZATION 44:177201 (2000) Elsevier Science Inc., 2000. All rights reserved. 655 Avenue of the Americas, New York, NY 10010

tility is often observed, which may lead to the ruin of the component. It is, therefore, of technological importance to understand the underlying mechanism of phase separation occurring in these steels. Indeed, predictions of the mechanical evolution, in terms of the component lifetime during holding at the service temperature, are based on a precise knowledge of the embrittlement process and kinetics. The basis of the whole stainless steel family, the ironchromium system, has been known for half a century to be susceptible to embrittlement when aged at intermediate temperatures, i.e., in the range of 300500C. This embrittlement, known for decades as 475C embrittlement [1], is related to the microstructural evolution occurring in the ferrite or martensite of FeCr-based alloys. This is the reason why microstructural evo1044-5803/00/$see front matter PII S1044-5803(99)00048-0

178

F. Danoix and P. Auger

lution in this system has attracted so much interest over the last decades. Many SmallAngle Neutron Scattering (SANS) and Mssbauer Spectroscopy (MS) studies clarified the evolution mechanism, but it has only been for about 20 years that an image in real space could be obtained using Atom Probe Field-Ion Microscopy (APFIM). But, of course, due to the high number of alloying elements present in industrial stainless steels, the details of the FeCr reaction are complicated by the presence of other elements (equilibrium and kinetics modification, other phases present, etc.), and numerous studies have been devoted to the investigation of ternary FeCrX and multicomponent industrial alloys. This paper aims at presenting a survey of the important contribution of atom-probe field ion microscopy in the study of the FeCrrelated alloys aged at an intermediate temperature.

FIG. 1. FeCr binary phase diagram (from [6]).

THE IRONCHROMIUM SYSTEM INTRODUCTION The miscibility gap of the FeCr phase diagram was first identified by Williams and Praxton [2]. It was shown to extend broadly and symmetrically, below a critical temperature of about 560C at 5060% Cr, down to 510 and 9095% Cr at 300C.1 The equilibrium structure consists of two phases, namely (Fe-rich) and (Cr-rich), the latter being considered as the precipitate phase, even though this terminology is not fully accurate in this system. In the upper part of the miscibility gap, the phase is known to precipitate. Its stability domain extends from 700C down to a temperature lying in the range of 430400C, so that the phase may coexist with and . Nevertheless, the phase reaction is so sluggish that it has not been observed even after 35,000 h of aging at 400C, whereas the reaction is very well developed [3]. The shape and limits of the miscibility gap in
1In

the FeCr system has been experimentally confirmed by several workers since, mainly with the use of SANS and MS. In addition, thermodynamic calculations using the regular solution model and including the magnetic transformation occurring during the to transformation also confirmed the location of this miscibility gap. The extension of the miscibility gap at temperatures of 300C was proven experimentally in the late 70s, because the long-term aging treatments necessary for the phase formation (because of too low Cr diffusivity in Fe at 300C) have been undertaken as part of the various nuclear plant survey programs at this period. It is not known whether another FeCr-containing phase would be stable at lower temperatures. An example of the commonly accepted shape for the Fe Cr binary phase diagram, with the miscibility gap clearly seen, is shown in Fig. 1. Within this miscibility gap, two domains exist, reflecting the two different paths possible to reach the equilibrium state, namely nucleation and growth (NG) and spinodal decomposition (SD), of the solid solution into the iron-enriched and the chromiumenriched phases. The classical distinction between those two mechanism can be summarized by saying that, in the first case, nuclei are formed locally with their equilibrium composition since the early stages of phase separation, whereas during spinodal decomposition, domains enriched in iron

the course of this paper, all concentration values will be given in atomic percent.

AP Studies of Fe-Cr-X and Stainless Steels

179

and domains enriched in chromium form all over the parent solid solution, and develop both in size and in composition toward their equilibrium composition. Both domains of the equilibrium phase diagram are separated by the spinodal line, classically described in the linear Cahn-Hilliard (CH) theory as the locus of the points where the second derivative of the free enthalpy vs. composition is zero [4]. The extent of the spinodal domain (where spinodal decomposition is the active phase separation mechanism) is much larger than that of nucleation and growth. This latter is located at the outskirts of the miscibility gap, and extends about 20% at 300C on each side of it (see Fig. 1). EXPERIMENTAL STUDIES Before the atom probe, many techniques were used to study the decomposition process in binary FeCr alloys. The aim of these studies was to give a better description of both the phase diagram and the decomposition paths (i.e., nucleation and growth and spinodal decomposition) as a function of temperature and composition. Among these techniques, MS and SANS have a particular place, as they give access to complementary information related to the microstructural evolution. MS can detect and domains at the early stages of their development, because they induce a modification of the mean hyperfine field [5, 6]. In addition, for a given aging treatment, determining the ferro- to paramagnetic temperature transition gives access to the composition of the forming phase [7]. But with this technique, no indication of the characteristic distances can be obtained; the information is only compositional. Conversely, SANS gives access to such information. That is the reason why this technique has mainly been used to test the validity of the various spinodal decomposition theories. Indeed, each theory predicts a time evolution of the characteristic distances (i.e., wavelengths) that can be compared with the experimental SANS results. Initial work by Vintaykin et al. [8, 9] on Fe-

20% Cr, Fe-28% Cr and Fe-40% Cr, aged up to 1,000 h at temperatures lying in the range 400550C, showed a satisfactory agreement between experimental data and the Cahn-Hilliard theory of spinodal decomposition, particularly for the specimen containing 40% Cr. LaSalle and Schwartz [10] studied a Fe32% Cr alloy homogenized at 1200C, and showed that the observed decomposition kinetic could satisfactorily be described by the CH and by the more refined nonlinear Langer-BarOn-Miller (LBM [11]) theories. By contrast, the same specimen homogenized at 850C exhibited a different behavior [10]. The decomposition kinetics was no longer in good agreement with the CH theory. Nevertheless, the peak time exponent behavior (i.e., the characteristic wavelength evolution) was found to fit well with the LBM theory, and in good agreement with the Monte Carlo simulations by Binder [12]. The same conclusions were drawn from a previous study by the same authors on a Fe-52% Cr alloy homogenized at 850C and aged at 500 and 520C [13]. Therefore, the concentration fluctuations present in the quenched material are supposed to play a significant role on the ensuing phase separation by spinodal decomposition, as previously noticed by Vintaykin et al. [14], although the reasons for this are not clear. Numerous other workers also studied Fe Cr alloys with the use of SANS. Hawick et al. (20 to 40% Cr) [15] and Bley (20 to 50% Cr) [16] both agreed to the fact that the position of the peak in the scattered intensity plots yielded to a time exponent of 0.20.25 as a function of time, in fairly good agreement with the prediction of the LBM theory. ADVANTAGES OF THE ATOM PROBE TECHNIQUE As a conclusion to these studies, the binary ironchromium system is shown to be a good model for studying spinodal decomposition over a wide range of compositions and at accessible temperatures. But the parameters used to compare experimental and theoretical data (mainly the structure

180

F. Danoix and P. Auger

and the pair correlation functions) give no information on the detailed morphology in real space, and direct comparison between experimental results and computer simulation in particular is not possible. Because of the nanometric scale at which the microstructural evolution takes place, transmission electron microscopy would be the best-adapted investigation technique. Unfortunately, in the ironchromium system, the combination of both the similarity of the X-ray scattering factors of these two elements, and the high coherency of the forming phases (namely and ) prevents its use to characterize the fine scale microsctructural evolution in the ferrite (or in the martensite). Conversely, the atom probe is a very effective technique to study this ultrafine scale phase separation because it provides both atomic-scale microstructure (because it is always combined with a Field Ion Microscope, FIM) and atomic-scale microchemical analysis of the observed phases. Indeed, the lateral resolution of the instrument is smaller than the extend of the emerging phases, and the mass resolution of this time-of-flight technique is more than sufficient to separate iron and chromium. It is, thus, possible to characterize the nature, the spatial extend, and the composition of the forming Cr-rich domains. EXPERIMENTAL PROCEDURE FOR ATOM PROBE STUDIES Both 1D and 3D atom probes have been used to study binary, ternary, or industrial FeCr-based alloys, but for all the studies concerned in that paper, only 1D atom probes were equipped with energy-compensating devices. As none of the 3D investigations presented here was carried out with energy-compensation systems, the mass resolution for 1D analyses is always better that for 3D analyses. Corresponding mass spectra are shown Fig. 2. That is the reason why 3D instruments were mainly used to obtain topological information from the emerging phases, whereas 1D instruments were preferred for quantitative

analyses and accurate phase composition measurements. In contrast to TEM, and as shown in the mass spectra, the separation of Cr and Fe is not a major issue for the atom-probe analysis. The only minor problem is the existence of a common isobar for chromium and iron at 54 atomic mass units. But, knowing the chromium and iron isotope natural abundance, this overlap can be accounted for, and the measured composition, accordingly, easily corrected. Although the atom probe has the best spatial resolution among the microstructural investigation tools, it may, in the case of a very fine-scale microstructure, face the problem of spatial convolution when analyzing nanometer scale domains, like and domains at the early stages of their formation [1719, 24]. This problem is enhanced because the lateral (and, therefore, the spatial) resolution of the instrument is not constant during analysis (in the case of 1DAP) [20]. As a consequence, the spatial convolution due to the detector lateral extent are not stationary, and therefore, cannot be accounted for and corrected. The only method to eliminate this problem is to position a diaphragm between the specimen and the detector. Reducing the diaphragm aperture when the specimen voltage is increased balances the increase of the lateral extent of the analyzed surface. Thus, the analyzed surface is kept constant, and so are the spatial convolutions that can, therefore, be taken into account during the derivation of the actual amplitude of the concentration fluctuation amplitude [20]. Such diaphragms can be implemented either by setting a variable size aperture at a constant distance from the detector [21], or by moving away from the specimen a constant size aperture [18]. Unless otherwise stated, the results presented in this paper were obtained with a constant lateral resolution, and the given values of fluctuation amplitude corrected to account for spatial convolutions. Another aspect to consider during analysis of FeCr alloys is the lower stability of chromium at the specimen surface, i.e., its lower evaporation field. As a consequence,

AP Studies of Fe-Cr-X and Stainless Steels

181

FIG. 2. Atom probe mass spectra with: (a) 3D Tomographic Atom Probe [52], (b) 1D energy-compensated instrument [21].

chromium may evaporate between the evaporation pulses, and thus not be detected, resulting in underestimated measured chromium concentration. Another consequence of the chromium lower evap-

oration field is the so-called local magnification effect, which can result in an overestimation of iron content in the phase [22]. These effects are enhanced when the decomposition proceeds (i.e., increases

182

F. Danoix and P. Auger

its Cr content and lowers it) in such a way that underestimation of the chromium content in the phase is getting worse, and thus not constant, when aging proceeds. There is a consensus for the evaporation conditions that should be used to minimize these consequences when analyzing FeCr alloys. A pulse fraction (pulse voltage over standing voltage) of 1920%, and a specimen temperature lower that 60K (40K) are commonly accepted values [23]. INITIAL STUDIES OF BINARY IRONCHROMIUM ALLOYS First analyses were conducted in the early 80s. Pioneer work by Brenner et al. [24] was dedicated to an Fe-32% Cr alloy aged at 470C up to 11,000 h. Their concern was to show metallographic evidence of the early stages of phase transformation occurring within the miscibility gap. Field ion microscopy investigations revealed the development, with aging time, of darkly imaging regions, as shown Fig. 3. Those regions showed a rather unusual morphology: they

consisted of veins threading through the imaged surface. In addition, those veins built a continuous network. Field evaporation of the specimen also proved the continuity of the network within the depth of the material. Thus, the morphology of the darkly imaging phase is a fully interconnected structure. In addition, no preferential orientation of the veins was observed with respect to the crystallography of the specimen, in such a way that the microstructure can be described an an isotropic sponge-like network, leading to a fully isotropic interconnected structure. Selected area atom probe analyses on those darkly imaging regions confirmed that these regions were Cr rich, containing up to 80 90% Cr. Random-area analyses of these specimens gave concentration profiles within the material with a spatial resolution lower the 1nm, as shown Fig. 4. The evolution of the concentration profiles with aging time can be described as follows. In the as-quenched conditions, the only observable fluctuations can be attributed to a statistical counting effect, giving no evi-

FIG. 3. FIM of decomposition in the Fe-32% Cr aged at 470C (from [24]).

AP Studies of Fe-Cr-X and Stainless Steels

183

FIG. 4. Concentration profiles in the Fe-32% Cr aged at 470C (from [24]).

dence of concentration variations (i.e., no phase separation has occurred). After aging 50 h, concentration fluctuations become clear, and may no longer be interpreted as statistical fluctuations: phase separation has begun. For prolonged aging times (193 and 669 h), the amplitude of the fluctuations increases, with the mean distance between Cr-rich peaks. As a conclusion, this work was the first reported atom-probe experimental evidence of spinodal decomposition within the miscibility gap of the Fe Cr system: the development of both the amplitude and wavelength of Cr concentration fluctuations is consistent with the continuous transformation mechanism, the classical description of the spinodal mechanism. If spinodal decomposition was known to take place in the miscibility gap of the FeCr system, one of the important aspects of this pioneer study is the description of the fine-scale microstructure, i.e., an isotropic fully interconnected network of veins in the solid solution. The isotropic nature of

the morphology is attributed to the small strain developed between the emerging and domains. For the most heavily aged specimens (11,000 h), the vein-like structure still persists, indicating that the coarsening of is very sluggish, presumably because of the small interfacial energy between and . After this initial work, some others were conducted on FeCr alloys, the most important ones by Miller et al. and Hyde et al. [2527] in the framework of a larger project involving Monte Carlo simulations for direct comparison (with the use of the reconstruction capabilities of the 3DAP). Alloys containing from 17 to 45% Cr were investigated during aging at 500C. One of their results is that only the 24% specimen exhibits isolated particles, whereas the 32% specimen showed the vein-like structure. This difference is a result of the different mechanisms involved during phase separation, and indicates that the spinodal line at 500C lies between 24 and 32% Cr. More gener-

184

F. Danoix and P. Auger

ally, this work focused on the determination of the evolution of the and domain size and composition, and of their interface. Another interesting point of this study is the characterization of the network morphology, as extensive analysis of 3D atom-probe reconstruction and Monte Carlo simulations lead to the description of the sponge-like structure in terms of fractal. More recently, Okano et al. [28] studied a Fe-75% Cr alloy aged at 500C. They showed that 2,600 h of aging produced, presumably by NG, 4nm large particles containing 80% iron embedded in a chromium-rich matrix containing 10% iron, in good agreement with the existing phase diagrams. APFIM is thus shown to be a very wellsuited technique to study phase separation in the binary FeCr system, and can be used to elucidate the effects of various parameters (such as composition, temperature, etc.) on the underlying thermodynamic process, and this is the reason why, since the early 80s, the atom probe has been widely used to study phase separation in the miscibility gap of various ironchromium-based systems.

miscibility gap, and element partitioning between the two phases [29, 30]. The addition of cobalt to the binary FeCr system produces an asymmetry in the miscibility gap, and raises the critical temperature. It also keeps the atomic misfit small, similar to that of the FeCr binary system, but produces a slightly higher internal stress, making observation of this system easier by TEM [31]. On a technical point of view, this ternary system is the basis of the so-called ductile permanent magnet alloys. The magnetic properties of this system depend upon both the morphology and the composition of the and phases, which form during aging in the miscibility gap. The decomposition in the Fe30.1Cr9.9Co ternary alloy aged at 525C up to 485 h is similar to that observed in the binary FeCr system. Coarsening of the microstructure follows a t0.2 law. In addition, it was found that in both and domains, the Fe:Co ratio is about 3:1, indicating that Co strongly partitions in the domains. Uemori et al. [32] also studied an FeCrCo with an addition of 2% Si. They showed that Co partitions into the , whereas Si partitions into the phase. FeCrAlX The FeCrAl system has been studied as they form the basis for the Fecralloy family. A Fe-20.2% Cr-8.8% Al-0.55% Ti have been investigated during aging at 475C [33]. FIM analysis of the aged samples showed that the interconnected structure is also present, indicating that phase transformation through a spinodal mechanism has occurred. Atom-probe concentration profiles confirmed the FeCr solid solution decomposition, as the concentration modulation is compatible with the one expected from a spinodally decomposed material, even though in that case only specimens aged 588 h were analyzed, which is not sufficient to prove the spinodal nature of the decomposition. Most interesting is aluminium behavior during the FeCr phase separation. Indeed, despite Thermocalc predictions that indicate a very strong partitioning of

ATOM PROBE STUDIES OF IRONCHROMIUM-BASED ALLOYS ATOM PROBE STUDIES OF TERNARY IRONCHROMIUM-X ALLOYS In parallel to the study of the binary FeCr system and industrial stainless steels, some ternary FeCr-based alloys have also been studied with the atom probe, in particular, FeCrCo, FeCrAl, and FeCrNi. A short review of these atom-probe studies is presented, with particular emphasis on the aging aspects and alloying element partitioning. FeCrCo The FeCrCo ternary system has been studied by several groups to study the kinetics of phase separation, morphological evolution of the microstructure within the

AP Studies of Fe-Cr-X and Stainless Steels

185

Al to the phase (virtually no Al is expected in the equilibrium phase), Al concentration profiles remain homogeneous: no partitioning is observed. To the authors knowledge, it is the only substitutional specie that does not show partitioning, despite the thermodynamic predictions. It is not clear yet whether the reason for that is an inadequate thermodynamic description of this system, or if it is due to insufficient Al mobility in this system. The first hypothesis is not supported by an observation made by Zhu et al. [34] in the FeCrCoAl system, where Al is shown to partition in the Fe-rich phase after aging 2 h at 525C. It has, therefore, been suspected that Al mobility is reduced to values that do not allow partitioning, leading to the so-called paraequilibrium transformation. If correct, it would be the first observation of paraequilibrium in a pure substitutional solid solution. One of the possible explanations would be related to the nanostructure, preventing stress relief, and thus lower Al diffusion to the domains. FeCrNi Major work on this system was conducted by Brown [18] on a series of Fe-26% Cr-(0-35-8)% Ni alloys, in parallel to the characterization of decomposition in industrial duplex stainless steels. The effect of nickel, and to a lesser extend Cr, on the aging behavior of FeCr alloys with a basic composition close to the one of commercial-cast duplex stainless steels, was analyzed, as well as the influence of austenite when present. Brown showed that the addition of nickel increases the rate of spinodal decomposition upon aging in the temperature range of 300450C. This result is in good agreement with thermodynamic calculations by Hayes et al. [35]. This increase is not perceptible on the activation energy of the aging process, which remains consistent within the 95% confidence limit with the value of 230kJ/mol of Cr diffusion in Fe-10 to 20% Cr [36]. It is more likely to be the preexponential term that is affected. Brown also

showed that, although his model alloys has a composition similar to that of the ferrite of his industrial steel, the aging kinetics were very different for the two materials [37]. To study the influence of the presence of the second phase, the austenite , the Fe26% Cr-5% Ni alloy was homogenized at 1060C (temperature used for the industrial alloys) to produce this duplex structure. Both inter- and intragranular austenite could then be produced, and the microstructure was closer to that of duplex steels, making this alloy a more reliable ternary model for commercial steels. This alloy was shown to age drastically and more rapidly than the purely ferritic one, consistently with the aging kinetics of the ferrite phase of the industrial alloy, the spinodal reaction being about 50 times faster. Alloying element partitioning rate, residual elements effects, dislocation pipe diffusion or vacancy supersaturation have been shown not to be responsible for such an enhancement. If the reasons are not yet clear, the presence of austenite is the key point for this enhancement, even when its volume fraction is as small as 1%. Conversely, increasing it up to 90% does not significantly change the rate of spinodal decomposition. The effect of austenite on the rate of spinodal decomposition has also been observed in the FeCr Co system [31]. More work is needed to understand the mechanism involved. In the 300350C temperature range, the activation energy remains unaffected, but in the range 400450C, it is reduced by a factor of 2, in good agreement with results obtained from impact toughness tests on duplex stainless steels [38]. Among the other results of this study, Ni is shown to strongly partition into the Fe-rich domains.

ATOM PROBE STUDIES OF DUPLEX STAINLESS STEELS INTRODUCTION Cast duplex stainless steels are used in some components of primary cooling circuits in pressure water reactor (PWR) nu-

186

F. Danoix and P. Auger

clear plants. These materials show good mechanical properties, corrosion resistance, weldability, and cast ability. As expected from their chemical composition, these duplex stainless steels are susceptible to embrittlement [39, 40]. At temperatures above 500C, various intermetallic phases, such as , R, , may precipitate in the ferrite, in addition to nitride or carbide precipitation [41]. At intermediate temperatures (i.e., 300500C), the embrittlement of cast duplex stainless steels is due to the phase transformation process of the ferritic phase, as the austenitic phase remains unaffected by such aging conditions [42]. Although relatively slow, these evolutions are still observed at temperatures as low as 280C, that is to say, the service temperature of duplex stainless steels of the cooling circuits. During holding at service temperature, the macroscopic mechanical properties (hardness, toughness, etc.) may evolve significantly differently according to the chemical composition, the homogenization treatment, and the service temperature [38, 43]. The presence of a number of alloying elements such as Ni, Si, Mo, Mn, etc., in addition to chromium, increases the number of possible decomposition paths leading to the equilibrium state of the solid solution. The material, quenched from the domain 10501150C, is highly unstable at a service temperature where the atomic mobility can be sufficient to achieve atomic redistribution and phase separation. The study of decomposition in these industrial materials is, therefore, very complex, and of a clear technological importance, as demonstrated by the number of studies initiated in the late 70s. As expected from the chemical composition of the steels, the phase separation mechanism in the ferritic phase of duplex stainless steels is very similar to the one observed in binary ironchromium specimens, as these two elements represent 90% of the atoms [44]. According to the temperature and composition of the steel, nucleation and growth or spinodal decomposition may be responsible for the ironchromium

phase transformation occurring in the ferrite. But at service temperature, the investigated steels are situated well within the miscibility gap, where the involved mechanism in clearly spinodal decomposition [19, 37, 4550]. One of the first aims of the survey programs was to identify the temperature extent of the spinodal domain for such steels, to ensure that a temperature increase will only affect kinetic aspects (simply accelerated aging treatments), but not the nature itself of the phase-separation mechanism. This point is a major issue for validation, in terms of reliability and representativity, of accelerated aging treatments [51]. If the main structural evolution in the ferrite is the transformation of the initially homogeneous FeCr-based solid solution, some Mo-bearing duplex stainless steel grades may exhibit an extensive precipitation of small Ni- and Si-enriched intermetallic G-phase particles [19, 44, 48]. The estimate of the component lifetime, in particular through the validation of extrapolation to long-term behavior from accelerated aging treatments, requires a precise understanding of all the underlying microstructural phenomena involved. The experimental approach of this problem faces two main difficulties: the very fine scale of the microstructural evolution (a few nanometers), and the complexity induced by the chemical complexity of the material, and therefore, of the phases to be characterized. This second difficulty has been partly overcome through the study of more simple binary or ternary alloys, such as FeCrNi. But, as shown previously, ferritic alloys are not fully representative of austenoferritic duplex ones, at least when kinetic aspects are concerned. Therefore, as far as estimation of the components lifetime is involved, the ferritic phase of duplex steels must be analyzed to obtain proper estimates of phase separation kinetics. In addition, although indispensable, the study of the model alloys cannot account for the complexity of the multielements industrial duplex stainless steels. With the number of elements, the number of phases

AP Studies of Fe-Cr-X and Stainless Steels Table 1 Composition of Duplex Stainless Steels Analyzed by Atom Probe (at.%)
Ref. Cr Ni Si Mo Mn C P S

187

CF3 CF8 CF8 CF8 CF8 CF8M CF8M CF8M

[47, 37] 19.521.5 8.79.1 1.36 0.15 0.4 0.12 0.03 0.02 0.05 [45, 37, 49] 21.6 8.0 2.74 0.19 0.50 0.13 0.03 0.013 0.06 [46] 20.2 8.3 2.0 0.08 0.28 0.18 0.08 0.04 0.06 [48] 20.4 8.4 1.88 0.05 0.83 0.18 0.027 0.08 [45, 49] 20.822.0 8.09.6 2.22.8 0.030.19 0.30.5 0.140.29 0.020.06 0.040.12 0.060.1 [44] 20.8 10.6 1.62 1.6 0.79 0.23 0.04 0.08 [48] 20.0 10.0 2.44 1.7 0.60 0.23 0.06 0.02 0.06 [19] 20.121.4 10.1 2.4 1.8 0.7 0.14 0.04 0.02 0.14

present in the steel may increase drastically. Experimental MS and SANS spectra become rapidly intractable, and the models used to interpret them are much too simplistic. This problem is even more difficult for industrial duplex stainless steels, where the ferrite-to-austenite ratio is only about 25%. APFIM, as it provides a direct view in real space of the microstructure of the ferrite, has became a key instrument for such studies. Indeed, the nanometric scale quantitative concentration profiles of all the elements can be obtained individually for all the present phases. In addition, no microstructural model is required for data processing. As a result, major fundamental and technologically important results were obtained with the atom probe on those complex and multiphase industrial alloys. MATERIALS Cast duplex stainless steels investigated by APFIM belong to the three different

gradesCF3 (low C), CF8 (high C), and CF8M (high CMo-bearing grade). A list of the various grades analyzed by the atom probe and their compositions is given in Table 1. They all contain 15 to 25% ferrite, and have been homogenized in the temperature range of 10501150C, in the duplex / domain. SPINODAL DECOMPOSITION IN DUPLEX STAINLESS STEELS Morphology and Concentration Profiles As illustrated in the FeCr system, iron and chromium concentration fluctuations emerge and develop progressively during aging, giving rise to the fully interconnected structure. A 3D reconstruction of such a structure obtained with the Tomographic Atom Probe [52] is shown Fig. 5. Each dot is a single atom, and only chromium atoms are represented for clarity.

FIG. 5. 3D reconstruction of CF8M specimen aged 30,000 h at 400C: network. The represented volume is 8 8 80nm3. Each dot is a single Cr atom.

188

F. Danoix and P. Auger

The interconnected nature of the structure is clearly seen. The Fe-rich network would be the complement, as Fe-rich and Cr-rich domains are exclusive. This 3D reconstruction provides a good representation of the morphology of the microstructure. Salient characteristics are the nanometric size of the various domains, and the progressive evolution of the concentration from to domains, even in this heavily aging treatment of 30,000 h at 400C. But most of the atom probe, work on duplex stainless steels has been realized with 1D instruments. The concentration profiles obtained are simply a projection of the 3D microstructure into the analysis direction, and the reduction of the information must be taken into account when concentration profiles are analyzed [47]. It is only relatively late in the study of duplex stainless steels that 3D instruments became available, and they have been used only for specific studies, or for illustration purposes, which explains that only minor reference will be made to this technique in this paper. The analysis of decomposition kinetics have been undertaken on a variety of duplex stainless steels, and all showed the same general trends. As an illustration, concentration profiles and related concentration frequency distributions obtained on a CF8M steel for various aging conditions are shown Fig. 6. In the unaged conditions, no phase separation has occurred; the concentration fluctuations on this profile have a simple counting statistics origin. This is proven by the corresponding concentration frequency distribution, which is identical to the binomial distribution expected from a purely random homogeneous solid solution [53]. During isothermal aging treatment, significant concentration fluctuations appear and develop as aging proceeds. The first important information provided by the concentration profiles is the characteristic length of the decomposition, i.e., the size and distance between the growing and domains. Statistical tools, such as Fourrier transforms and autocorrelation functions, have been used to derive characteristic dis-

tances from concentration profiles [23]. In parallel, the experimental concentration frequency distributions broaden and differ more and more from the random distribution observed in the as-quenched sample. Frequency distribution analysis is used to derive the significant amplitude of the concentration fluctuations, which directly accounts for the evolution of the unmixing process. With amplitude and wavelength information, the kinetics of phase transformation can be followed accurately, and then compared with mechanical data to derive relationships between the fine-scale microstructure and the mechanical properties. Wavelength Concentration profiles give a good idea of the size of Cr-rich domains, and of the correlation distances between them. In addition, their evolution with aging conditions may also be derived from concentration profiles. Three different techniques have so far been used to calculate this mean distance. The first one, and the most simple, consists of determining the mean distance between successive chromium peaks. The disadvantage of this methods lies in the fact that statistical fluctuations can be misinterpreted as domains, and therefore, lead to an underestimated mean wavelength. For that reason, statistical methods are usually preferred. In addition, they give the possibility of processing concentration profiles in a standard way. The most commonly used technique is the autocorrelation function [5456], although power spectra methods may also be applied [57, 58]. Results presented in Table 2 are obtained with the autocorrelation technique, where the mean wavelength is taken as the average value of the first five maxima of the autocorrelation function [59]. For the first stages of aging, concentration profiles show fluctuations with wavelengths smaller than 5nm. These first stages correspond to aging at 300C, up to 2,500 h at 350C, and up to 25 h at 400C. When aging is continued, these wavelengths increase progressively, but

AP Studies of Fe-Cr-X and Stainless Steels

189

longer wavelength fluctuations appear to modulate the short wavelengths [60]. Their presence modifies the local chromium concentration, and probably the local thermodynamics. These long-range fluctuations are also affected in wavelength and amplitude by the aging treatment. Such longrange order fluctuations have already been detected in stainless steels by Epperson et al. [61], by means of SANS at very small angles. They have been associated, because of similar size, to the existence of the magnetic domains of the ferrite, even though no physical argument has been brought to support it. Their existence may also be interpreted in terms of self-similarity of the microstructure, at least at the observed scale (1 to 1,000nm). Within this framework, it can be suspected that this self-similarity extends up to the size of the ferritic grains themselves, that is, several tens of microns in duplex stainless steels. This hypothesis was initially made by Langer [62] from a purely theoretical point of view, because of the formal similarity between the diffusion equation and the Schrdinger equation, leading to similar mathematical treatments of both equations. The modulation of the driving force, G, could play the role of a periodic potential, and thus lead to the existence of Bloch-type waves. Experimental evidence for them is, nevertheless, difficult to give, as the expected amplitude of these fluctuations should be very small. Another description of the same self-similarity can be given in terms of fractal dimension. First suggested by Camus et al. [63], this approach has been formalized by Hetherington and Miller [64], Cerezo and Hetherington [65], and Miller and Hetherington [66]. Numerical resolution of the Cahn Hilliard equation indicates that the microstructure fractal dimension evolves from 3 (random solid solution) down to 2 during aging. They have also shown that the fractal dimension of the 1D concentration profiles (more precisely, the atom data chain) increases monotonically with aging time from 0.5 (Brownian motion) to less than 1, indicating chromium clustering. This evolution

means that the microstructure scales with aging. These authors suggested that this fractal dimension could be used to parameterize the material-aging kinetics, as it is shown to be independent on the sampling volume used to calculate concentration data, but no extensive use of such a parameter has been reported so far. Amplitude The general shape of the concentration frequency distributions, and its evolution with aging, has been used to define statistical parameters that account for the increase of fluctuation amplitudes, and more generally, for the spinodal decomposition advance. As discussed elsewhere [67], two main regimes have to be considered when the amplitude of a spinodal is to be estimated. In a first approach, according to the CH linear model, concentration fluctuations can be considered as sine-like waves expanding all over the decomposed material. This simplistic model has a limited validity domain, as it is based on a linear resolution of the diffusion equation, which can only apply for the first stages of decomposition. Indeed, according to this same model, fluctuation amplitudes should grow exponentially, and lead to fluctuations that are no longer symmetrical with respect to the mean composition. In the particular case of stainless steels, the decomposing ferrite only contains about 25% chromium, and fluctuations become nonsymmetrical as soon as the equilibrium composition of the Fe-rich domains is reached. This model must, therefore, be considered with care when analyzing long-term aged stainless steels, and also the statistical models derived from it [47, 68]. For this problem, the nonlinear LBM model should be preferred. Most pertinent of the approximations used in this model is the assumption that the concentration probability distribution is written as the sum of two Gaussians, both having the same width and centered respectively at C and C, assumed to be the concentration of and domains [69]. If sampling statistical effects, as they appear

190

F. Danoix and P. Auger

FIG. 6. Concentration profiles and related frequency distributions: (a) CF8M steel aged at 350C, (b) CF8M steel aged at 400C.

in the atom probe analysis, are incorporated in this model, , C and C can be adjusted to fit to the best experimental distributions, through a maximum likelihood method [70]. C and C are considered as measured chromium concentrations of the and phases. As Gaussians may expend to be negative, or larger than 100%, concentration values, Auger et al. [71] proposed to replace them by binomial distributions. If convolution effects are taken into account and corrected [17], the amplitude of the concentration fluctuations, C, can be simply obtained [60]. The values reported Table 2 were obtained this way. It can be seen that no estimate of C is given for aging conditions longer than 2,500 h at 400C. The reason is that concentration frequency dis-

tributions are no longer bimodal; the importance of domains containing more than 60% Cr (up to 80%) increases, whereas the importance of domains containing about 40% Cr remains stable. At the mean time, the chromium content of Fe-rich domains has reached a stable value of 810%, which can be considered as the equilibrium concentration in the range 350400C. This complexity precludes the description of the solid solution with simple models, at least after aging more than 2,500 h at 400C. The existence of retained (possibly metastable) domains containing 40% Cr can possibly be associated with the ferro- to paramagnetic transition occurring at that concentration in binary FeCr alloys. This observation confirms the influence of magnetic

AP Studies of Fe-Cr-X and Stainless Steels

191

FIG. 6. Continued.

aspects in the spinodal decomposition in the ferritic phase of duplex stainless steels. A more simple statistical parameter has also been used in the course of this work. It is called the variation, V, and is the integral of the difference between the Cr concentration frequency distribution of the aged sample and the corresponding (same average) binomial distribution (random solution) [53]. No concentration modulation model is required to estimate V; thus, this parameter can be used whatever the experimental distribution shape is, and in particular, for the longer aging times at 400C, where other models fail. Although simplistic, this parameter has rapidly shown to be excellent criterion to follow aging, as will be illustrated later. V values obtained for different aging treatments are reported Table 2. Kinetic of Phase Separation and Activation Energy The evolution of the microstructural parameters obtained by atom probe analysis

can be compared with the prediction of spinodal decomposition models. As shown in Fig. 7, the main characteristic of both the mean wavelength and the amplitude is a slow but continuous evolution, even after heavy aging conditions. The evolution of the amplitude of concentration fluctuations does not match the CH theory predictions of an exponential increase with time for the early stages of phase separation [4]. Conversely, the evolution of the mean wavelength, both at 350 and 450C, is in good agreement with the LBM theory, which predicts with a time exponent of about 0.18 [11]. As has been reported in the case of binary FeCr alloys, the LBM model is shown to be better adapted than the CH for the study of long-term aging of stainless steels. This evolution clearly appears thermally activated, as the evolution is faster when the temperature is increased (see Table 2). It is, therefore, reasonable to write this evolution as an Arrhenius law over the studied range of decomposition. As similar decomposition states are observed at different aging temper-

192

F. Danoix and P. Auger Table 2 Evolution of the FeCr Phase Separation in a CF8M Duplex Stainless Steel

Aging condition

C (%)

(nm)

HV0.05 ( )

Unaged 10,000 h at 300C 30,000 h at 300C 1,000 h at 350C 2,500 h at 350C 10,000 h at 350C 30,000 h at 350C 100,000 h at 350C 25 h at 400C 250 h at 400C 2,500 h at 400C 10,000 h at 400C 30,000 h at 400C

0.02 0.18 0.32 0.31 (0.01) 0.51 (0.03) 0.67 (0.05) 0.87 (0.08) 0.85 (0.11) 0.33 (0.05) 0.68 (0.05) 0.97 (0.09) 1.1 (0.1) 1.3 (0.1)

5.9 (0.5) 8.4 (0.6) 8.7 (0.6) 12.6 (0.8) 15.6 (0.8) 17.4 (1.0) 18 (1) 9.3 (0.5) 15.3 (0.8) * * *

3.4 (0.5) 5.3 (0.5) 4 (1) 5.5 (0.5) 7.0 (0.5) 8 (1) 9 (2) 5.4 (0.5) 6.5 (0.5) 17 (3) 15 (2) 20 (3)

304 (30) 348 (38) 450 (45) 435 (55) 553 (73) 640 (86) 716 (104) 730 (80) 510 (71) 660 (55) 763 (52) 890 (60) 989 (38)

Standard deviation on microhardness determined over 50 measurements. Impossible to determine because of a too weak concentration fluctuation signal. *Impossible to fit experimental frequency distributions with bimodal ones.

atures, it is possible to estimate the corresponding activation energy. A value of Q 230kJ/mol is found, which is in very good agreement with the activation energy of chromium diffusion in -iron [36]. It should be noted that this value is very similar to the one obtained by Brown [18] and Leax et al. [19]. It is also quite interesting to notice, as shown Fig. 8, that V accounts nicely for the evolution of the ferrite microhardness. Brown [72] found that the same type of linear fit (using Pa as an estimate of the fluctuation amplitude) is observed for single- and two-phase model alloys. This confirms that the spinodal decomposition in the ferrite is the main source of embrittlement of duplex stainless steels, and gives additional information on the hardening process. Indeed, this result may be interpreted in the framework of Park et al.s [73] model, who attribute the hardening to the increase of the elastic strain, although small, resulting of the crystallographic misfit between the forming of the and domains. G-PHASE PRECIPITATION In some duplex stainless steel grades, an intermetallic phase has been reported to form

in the ferrite after aging at intermediate temperature. This phase has been crystallographically identified as the G-phase [74], a fcc structure (space group Fm3m, lattice parameter 1.09nm), having a cube-on-cube orientation with the ferrite matrix. On a chemical point of view, it has been shown to be an alloyed silicide, incorporating most of the solute elements present in the steel, in particular, a large amount of nickel. Its stoechiometry is based on Ni16Si7Ti6 (55% Ni-24% Si-21%Ti), but with many variations reported [75]. In particular, Cr and Mn may substitute to Ti, and Fe and Mo to Ni [48, 76, 77]. The extent of the G-phase precipitation has been shown to be very dependent on the chemical composition of the steel (more precisely of the ferrite), in particular, to the C and Mo contents [78]. Indeed, Mo-bearing grades are most susceptible and sensitive to this precipitation, which, combined with the high Mo content of the G-phase particles, emphasized a supposedly important role of molybdenum in the precipitation process. Figure 9 represents the position of nickel, silicon, molybdenum, and manganese in the same volume as the one represented Fig. 5. Conversely to what is observed for

AP Studies of Fe-Cr-X and Stainless Steels

193

FIG. 8. Evolution of the ferrite microhardness with V. Data from specimens aged at 300, 350, and 400C.

FIG. 7. Time evolution of concentration amplitude and wavelength during aging: (a) concentration amplitude at 350C, (b) wavelength at 350C, (c) wavelength at 400C. Last point in (a) and (b) due to Hdin [87].

chromium atoms, these elements partition to roughly spherical precipitates, the G-phase particles, which are less than 10nm in diameter for this heavily aged specimen. Those precipitates are, thus, very small, and again, the atom probe is a very well-suited instrument to analyze them chemically. Because of its crystallography, different from that of the ferritic matrix, this phase can be observed by TEM, which gives a larger view of the precipitation of G-phase than a simple atom-probe analysis. G-phase particles are shown to grow preferentially along dislocations, which is

the only sites where these particles are observed in Mo-free grades. In Mo-bearing grades, the precipitation is homogeneous over the all-ferritic solid solution, although larger particles are observed on dislocations. Atom probe studies give a fairly accurate description of the size, number density, and composition evolution of G-phase precipitates during aging at 300400C. In addition to nickel and silicon, the particles are also molybdenum, manganese, and to a lower extent, carbon and phosphorus enriched. As a general feature, it appears that all the alloying elements of the steel, except chromium, partition to the G-phase. As a consequence, all these elements can be considered as G-forming elements, and the G-forming element content is often taken as a compositional characteristic of the phase, and the sum of all these individual contributions taken as a tracer of this phase. Tables 3 and 4 provides the evolution of these various parameters during aging. Neither the evolution of the composition of the precipitates at a given temperature nor kinetic arguments (which will be discussed later) are in agreement with this classical nucleation and growth model, which can definitely not apply to a G-phase precipitation. Internal Complexity Dark-field TEM micrographs, as well as atom probe analysis of G-phase particles reveal a complex internal structure within each individual precipitate, although these particles are less than 10nm in diameter. TEM reveals a pseudo-fivefold symmetry in the diffraction pattern, accompanied by

194

F. Danoix and P. Auger

FIG. 9. 3D reconstruction of CF8M specimen aged 30,000 h at 400C: G-phase precipitates. The represented volume is 8 8 80 nm3. Ni, Si, and Mo atoms are represented. The volume is the same as in Fig. 5.

a large number of internal stacking faults [48]. Compositional analysis by the atom probe also reveals some sort of internal microsegregation, leading to the idea that G-phase particles may, in fact, constitute an aggregate of parent nanodomainsor phaseswith particular orientation relationships. Nucleation Mechanism As shown on the 3D reconstruction (Fig. 10), G-phase particles are located at interfaces. No particle have been observed within or domains. This indicates that the spinodal decomposition and the G-phase precipitation are highly dependent. As it is clear that spinodal decomposition may take place without G-phase particles, unmixing must be regarded as the governing process, and G-phase precipitation as a secondary process. The confirmation is given by the comparison of the

relative of these two mechanisms. With data from Table 3, it is possible to estimate the time evolution of G-phase particles after aging at 350C. A ta, with a 0.15 fits nicely with the experimental data. This time law is significantly different from the one expected for a classical nucleation and growth (a 0.5), but very similar to the one measured for the evolution of the mean distance (a 0.150.18). From this kinetic argument, it is clear that G-phase precipitation is controlled by the concomitant iron chromium unmixing process [79]. The formation mechanism of the G-phase has been studied in details by comparison between the thermal evolution of a CF8 and a CF8M steels. The compositions of the and phases in these two steels have been measured after aging 30,000 h at 350C. They are given in Table 5, in addition to the composition of the ferrite of these two steels. It appears that both steels have more nickel and silicon that both the

Table 3 Evolution of G-Phase Precipitation in a CF8M Duplex Stainless Steel


Aging conditions Particle diameters (nm) Number density (m3) Volume fraction (%) G-Forming elements content (%)

30,000 h at 300C 1,000 h at 350C 2,500 h at 350C 30,000 h at 350C 30,000 h at 400C

4 4 5 6 10

8.5 1023 4.5 1023 6.5 1023 5 1023 2.3 1023

4 1.5 4 5.5 12

37 40 49 60 80

Standard deviation is about 10%. G-Forming elements are Ni Si Mo Mn Al C P.

AP Studies of Fe-Cr-X and Stainless Steels Table 4 Composition of G-Phase Particles in Various Mo-Bearing Stainless Steels (%)
Aging conditions Concentration (%) Fe Cr Ni Si Mo Mn

195

Steel

Ti

CF8M [48] CF8M [76] CF8M [74] 308 CRE [77]

30,000 h at 400C 10,000 h at 450C 7,500 h at 400C 24 h at 550C

515 16 21 (incl.Mn) 4

1018 25 12 47

3035 26 24

2025 14 28 28

515 16 13

510 3 with Fe

21

and phases can accommodate. It is also clear that the ferrite of the CF8M grade contains 1.5% Ni and 0.8% Si in excess, compared to the total acceptance of both and domains (resp., 4.5% Ni and 2.0% Si). Furthermore, the distribution of Ni and Si in and domains is not symmetrical: Ni is better accepted in Fe-rich domains than in Cr-rich domains, from which it is rejected. The reverse situation (with respect to and ) is observed for Si. The growth of and domains during the spinodal

decomposition will result in the rejection of Si from domains and of Ni from domains. The opposite fluxes of both species at the interfaces is supposed to facilitate the nucleation of the NiSi-rich G-phase at this location, as experimentally observed. Furthermore, Ni and Si supersaturations are much lower in the CF8 steel, which can explain why G-phase precipitation is lower in this grade than in CF8M steels. Partition ratios of the major alloying elements between the and G-phase are seven for

FIG. 10. Respective positions of network and G-phase precipitates. The gray isoconcentration surface corresponds to the limit of domains (threshold 30% Cr) and the dots to Ni atoms. Represented volume is 8 8 25nm3.

196

F. Danoix and P. Auger

Table 5 Phase Composition in the Ferrite of CF8 and CF8M Duplex Stainless Steels Aged 30,000 h at 350C
Concentration (%) Ni Si Mo Mn

CF8ferrite CF8 phase CF8 phase CF8Mferrite CF8M phase CF8M phase CF8MG-phase

4.7 0.4 5.7 0.3 3.6 0.2 6.1 1.0 5.0 0.3 3.8 0.2 25 1

2.3 0.3 1.8 0.1 2.3 0.1 2.8 0.4 1.7 0.1 2.4 0.1 14 1

0.16 0.03 0.1 0.1 0.2 0.1 2.2 0.3 1.6 0.1 2.0 0.1 41

0.6 0.05 0.53 0.10 0.7 0.10 0.4 0.1 0.17 0.05 0.26 0.05 51

Si, six for Ni, and two for Mo. These arguments lead to the conclusion that Ni and Si, rather than Mo, are the decisive elements in the precipitation of the G-phase [80]. The reason for which Mo was supposed to play a major role is, in fact, indirect. Indeed, when Mo (an -like element) is added, the Ni (a -like element) level is also increased to keep the / ratio stable. As Ni is shown to be a driving element for G-phase precipitation, Mo addition indirectly favors G-phase precipitation. As shown in Table 6, for a given V, the hardness of CF8M steel is larger than that of CF8. Two mechanisms may account for this observation. First, as the major cause for hardening is spinodal decomposition of the FeCr solid solution, it may have a more pronounced effect in Mo-bearing grades. As Brown [18] did not noticed any influence of nickel on hardening, Mo can be suspected to have one. Indeed, as this element partitions in the domains, and be-

cause of his relatively large atomic radius (0.2725nm compared to 0.2490nm for chromium), it probably enhances the atomic misfit between and domains, and thus, the elastic strain primarily responsible for hardness [73]. The second possible mechanism is the excess precipitation of the G-phase in the Mo-bearing grade; as such, an intense precipitation of small particles must have an effect on the mechanical properties of the material. But the role of G-phase precipitation is not clear, and is still a matter of debate. Indeed, as G-phase precipitation is hardly dependent on the spinodal decomposition advance, it is very difficult to deconvolute individual effects of these two mechanisms on hardness. In addition, because decomposition is shown to be the main reason responsible for the hardness increase, the effect of G-phase precipitation will be included in it, and possibly concealed. As a matter of fact, experimental results are contradictory. Vitek et al. [81]

Table 6 Evolution of the Microstructural Parameters and of the Hardness for Given V
Aging treatment V

p(nm)

Np (part/m3)

Fv (%)

HV0.05

CF8M CF8M CF8M CF8M CF8M CF8

30,000 h at 300C 1,000 h at 350C 25 h at 300C 10,000 h at 350C 250 h at 400C 30,000 h at 350C

0.32 0.31 0.33 0.67 0.68 0.67

4.5 4.0 none detected 6.0 3.3 3.0

8.4 1023 4.5 1023 none detected 5.4 1023 12.0 1023 0.1 1023

4.0 1.5 none detected 5.0 2.3 0.05

450 435 510 660 648 440

AP Studies of Fe-Cr-X and Stainless Steels

197

and Leax et al. [19] showed a hardening effect of G-phase particles, whereas Chung and Leax [49] and Miller and Bentley [50] did not. Data in Table 6 seem to indicate that there is no effect. Indeed, for specimens showing the same amount of spinodal decomposition (same V value close to 0.3) obtained for different aging temperatures, the intensity of G-phase precipitation decreases with increasing temperature, whereas the hardness increases. It should be noted that the evolution of G-phase precipitation is in good agreement with Leax et al.s work [19], who calculated an activation energy for precipitation of 140 60kJ/ mol, almost half of that measured for decomposition. The situation is not clear at all, and the only sensible conclusion is that the influence of G-phase precipitates on hardness, if any, must be weak, and most probably hidden by more subtile or indirect effects. In particular, the precipitation of G-phase will deplete the matrix in molybdenum. Elastic strain between and domains may, thus, be reduced, leading to a decrease of the material hardness associated with G-phase precipitation.

on the early-forming Cu clusters [84]. A 17-4 PH (17.5% Cr-4% Ni-3% Cu-1% Si0.3% C) martensitic stainless steel has also been investigated. It is strengthened by -Cu particles produced during a tempering treatment at 580C for about 4 h. A subsequent increase in hardness is observed after long-term aging is in the range of 300400C. Although Miller and Burke [85] found direct evidence of precipitation during aging at 482C, Murayama et al. [86] found that the phase separation occurs through a spinodal mechanism, with G-phase heterogeneous nucleation on Cu precipitates after 1,000 h at 400C. In addition, these authors indicate that the increase in hardness and yield strength after aging at 400C is mainly due to spinodal decomposition, whereas G-phase precipitation does not appear to have a significant contribution to aging, as previously observed in duplex stainless steels.

CONCLUSION The presented atom probe studies of longterm aging at intermediate temperatures on ironchromium-based model and industrial alloys have confirmed that the major microstructural evolution is the spinodal decomposition. This decomposition process have been characterized from atom probe concentration profiles, and its evolution quantified by mean of various microstructural parameters directly accessible, such as the mean wavelength and the amplitude of the concentration fluctuations. If the linear Cahn-Hilliard theory does not adequately account for the observed evolution, the nonlinear Langer-BarOn-Miller theory is much more adaptable. But for heavily decomposed specimens, the observed complexity, both in terms of wavelength and concentration, cannot be described by this theory. More sophisticated models, including magnetic interactions, should be developed and compared with experimental data obtained by the atom probe.

ATOM PROBE STUDIES OF MARTENSITIC STAINLESS STEELS Several investigations have been conducted on martensitic stainless steels aged at an intermediate temperature. Because of their good corrosion resistance and mechanical properties, these steels are used as a structural material for chemical and power plants, at a temperature in the range of 300400C, a region where decomposition occurs. Two different steels have been analyzed by the atom probe, to clarify the phase formed in the martensite, and responsible for the good mechanical properties [82]. In the case of IRK91 maraging stainless steel (13% Cr-9% Ni-2% Mo-2% Cu), no transformation at 475C (even after 1,000 h) is observed [83]. The good mechanical properties are attributed to the heterogeneous precipitation of Ni-rich/(AlTi) precipitates

198

F. Danoix and P. Auger


5. T. De Nys and P. M. Gielen: Spinodal decomposition in the FeCr system. Met. Trans. 2:14231428 (1971). 6. D. Chandra and L. H. Schwartz: Mossbauer effect study of the 475C decomposition of FeCr. Met. Trans. 2:511519 (1971). 7. H. D. Solomon and L. M. Levinson: Mossbauer effect study of the 475C embrittlement of duplex and ferritic stainless steels. Acta Metall. 26:429442 (1978). 8. Ye. Z. Vintaykin and A. A. Loshmanov: The character of embrittlement in FeCr chromium alloy at 475C. Fiz. Metallov. Metalloved. 22:473476 (1966). 9. Ye. Z. Vintaykin, V. N. Dmitriyev, and V. Yu. Kolontsov: Kinetic study of phase separation of FeCr solid solutions by neutron diffraction analysis. Fiz. Metallov. Metalloved. 29:12571267 (1970). 10. J. C. LaSalle and L. H. Schwartz: Further studies of spinodal decomposition in FeCr. Acta Metall. 24: 9891000 (1986). 11. J. S. Langer, M. Bar-On, and H. D. Miller: New computational method in the theory of spinodal decomposition. Phys. Rev. A 11:14171429 (1975). 12. K. Binder: Theory for the dynamics of clusters. II. Critical diffusion in binary systems and the kinetics of phase separation. Phys. Rev. B 15:44254447 (1977). 13. J. C. LaSalle and L. H. Schwartz: Spinodal decomposition in Fe52Cr. In Decomposition of Alloys: The Early Stages, P. Haasen, V. Gerod, R. Wagner, and M. Ashby, eds., Pergamon Press, Oxford, UK, pp. 104109 (1984). 14. Ye. Z. Vintaykin, V. N. Dmitriyev, and V. Yu. Kolontsov: Effect of prior heat treatment on the rate of separation of ironchromium solid solutions. Fiz. Metal. Metalloved. 27:11311133 (1969). 15. K. A. Hawick, J. E. Epperson, C. G. Windsor, and V. S. Rainey: Chemical phase separation in binary ironchromium alloys. Mater. Res. Symp. Proc. Materi. Res. Soc. 205:107112 (1992). 16. F. Bley: Neutron small angle scattering study of unmixing in FeCr alloys. Acta Metall. Mater. 40: 15051517 (1992). 17. F. Danoix, P. Auger, A. Bostel, and D. Blavette: Atom-probe characterization of isotropic spinodal decomposition: Spatial convolution and related bias. Surface Sci. 246:260265 (1991). 18. J. E. Brown: An atom probe study of the spinodal decomposition in FeCrNi. M.Sc. Thesis, Department of Materials, University of Oxford, Oxford, UK (1990). 19. T. R. Leax, S. S. Brenner, J. A. Spitznagel: Atomprobe examination of thermally aged CF8M cast stainless steel. Met. Trans. 23A:27252736 (1992). 20. F. Danoix, M. Bouet, P. Pareige, and A. Menand: Controlled aperture atom-probe: Principle and applications. Appl. Surface Sci. 67:451458 (1993). 21. J. M. Sarrau, F. Danoix, M. Bouet, B. Deconihout,

The hardness of the ferrite, in both single and duplex models and duplex industrial alloys, is directly proportional to the amplitude of concentration fluctuations. The lattice misfit between the emerging and domains, and the resulting elastic strain can, therefore, be regarded as the main cause of hardening, and by extension, of embrittlement of this alloy family. In Mo-bearing industrial duplex stainless steels, the precipitation of the intermetallic G-phase is observed. The precipitation process of this phase is controlled by the concomitant reaction, which make any deconvolution of the influence of each process on the mechanical properties difficult. This issue has not yet found any clear solution. Regarding the role of alloying elements, there is clear evidence that nickel enhances the spinodal decomposition mechanism, but it has also proven to be a key element, with silicon, in the precipitation process of the G-phase. The situation with molybdenum is not as clear. It probably amplifies the hardening effect of the spinodal decomposition, at least in the early stages before partitioning in the G-phase, but does not seem to be a major element for its precipitation. These studies show that, even in very complex industrial alloys, the atom probe can give an original view of the very finescale structural evolution, and to some extent, identify the processes responsible for the evolution of the mechanical properties.

References
1. H. D. Newell: Properties and characteristics of 27% chromium-iron. Metals Progress 49:9771006 (1949). 2. R. O. Williams and H. W. Praxton: The nature of aging of binary iron chromium alloys around 500C. J. Iron Steel Inst. 185:358374 (1957). 3. S. M. Dubiel and G. Inden: On the miscibility gap in the FeCr system: A Mossbauer study on long term annealed alloys. Z. Metallkde 78:554559 (1987). 4. J. W. Cahn: Spinodal decomposition. Trans. AIME 242:166180 (1966).

AP Studies of Fe-Cr-X and Stainless Steels


D. Blavette, and A. Menand: The Rouen energy compensated atom probe. Appl. Surface Sci. 76/77: 367373 (1994). 22. M. K. Miller: The effects of local magnification and trajectory aberrations on atom probe analysis. J. Phys. 48C6:565570 (1987). 23. M. K. Miller and G. D. W. Smith: Atom-Probe Microanalysis: Principles and Applications in Materials Science, Materials Research Society, Pittsburg, PA (1989). 24. S. S. Brenner, M. K. Miller, W. A. Soffa: Spinodal decomposition of Fe-32 at % Cr at 470C. Scripta Metall. 16:831836 (1982). 25. M. K. Miller, J. M. Hyde, M. G. Hetherington, A. Cerezo, G. D. W. Smith, and C. M. Elliott: Spinodal decomposition in FeCr alloys: Experimental study at the atomic level and comparison with computer models I. Introduction and methodology. Acta Metall. Mater. 43:33853401 (1995). 26. J. M. Hyde, M. K. Miller, M. G. Hetherington, A. Cerezo, G. D. W. Smith, and C. M. Elliott: Spinodal decomposition in FeCr alloys: Experimental study at the atomic level and comparison with computer models II. Development of domain size and composition amplitude. Acta Metall. Mater. 43: 34033413 (1995). 27. J. M. Hyde, M. K. Miller, M. G. Hetherington, A. Cerezo, G. D. W. Smith, and C. M. Elliott: Spinodal decomposition in FeCr alloys: Experimental study at the atomic level and comparison with computer models. III. Development of morphology. Acta Metall. Mater. 43:34153426 (1995). 28. R. Okano, K. Hono, K. Tukanashi, H. Fujimori, and T. Sakurai: Magnetoresistance and phase decomposition in CrFe bulk alloys. J. Appl. Phys. 77: 58435849 (1995). 29. S. S. Brenner, P. P. Camus, M. K. Miller, and W. A. Soffa: Phase separation and coarsening in FeCr Co alloys. Acta Metall. 32:12171227 (1984). 30. F. Zhu, P. Haasen, and R. Wagner: An atom probe study of the decomposition of FeCrCo permanent magnet alloys. Acta Metall. 34:457463 (1986). 31. L. L Horton, M. K. Miller, and S. Spooner: Characterization of Spinodally Decomposed Fe-30.1% Cr9.9% Co: Parts 1 and 2. Proc. of the 44th Annual Meeting of the Electron Microscopy of America, G. W. Bailey, ed. San Francisco Press, CA, pp. 580 583 (1986). 32. R. Uemori, T. Mukai, and M. Tanino: APFIM Study of the Partitioning of Alloying Elements During Spinodal Decomposition in FeCrCoSi alloys. Proc. of Phase Transformations 87, Cambridge, UK, G. W. Lorimer, ed., Institute of Metals, London, UK, pp. 4446 (1988). 33. H. G. Read and H. Murakami: Microstructural influences on the decomposition of an Al-containing ferritic stainless steel. Appl. Surface Sci. 94/95:334 342 (1996).

199
34. F. Zhu, H. Wendt, and P. Haasen: Atom probe field ion microscopy of FeCrCo permanent magnet alloy. Scripta Metall. 16:11751180 (1982). 35. F. H. Hayes, M. G. Hetherington, and R. D. Longbottom: Thermodynamics of duplex stainless steels. Mater. Sci. Technol. 6:263272 (1990). 36. T. Heumann and H. Bohmer: Determination of coefficient of diffusion in FeCr alloys. Arch Eisenhttenwesen 31:749754 (1960). 37. J. E. Brown, A. Cerezo, T. J. Godfrey, M. G. Hetherington, and G. D. W. Smith: Quantitative atom probe analysis of spinodal reaction in ferrite phase of duplex stainless steel. Mater. Sci. Technol. 6:293 300 (1990). 38. S. Bonnet, J. Bourgoin, J. Champredonde, D. Guttmann, and M. Guttmann: Relationship between evolution of mechanical properties of various cast duplex stainless steel and metallurgical and aging parameters: Outline of current EDF programmes. Mater. Sci. Technol. 6:221229 (1990). 39. H. D. Solomon and T. M. Devine: Duplex stainless steels: A tale of two phases. General Electric Report No. 82CRD276, Nov. 1982, also in Duplex Stainless Steels, ASM, Metals Park, OH, pp. 698 756 (1983). 40. Proc. of Intermediate temperature embrittlement processes in duplex stainless steels, Oxford, UK, 1989, special issue of Mater. Sci. Technol. 6 (1990). 41. B. Josefsson, J. O. Nilsson, and A. Wilson: Phase Transformations in Duplex Stainless Steels and the Relation Between Continuous Cooling and Isothermal Heat Treatment. Proc. of Duplex Stainless Steels 91, Beaune (Fr), J. Charles, and S. Benhardsson, eds., Les Editions de Physique, Paris, France, pp. 6778 (1991). 42. M. Guttmann: Intermediate Temperature Aging of Duplex Stainless Steels. A review. Proc. of Duplex Stainless Steels 91, Beaune (Fr.), J. Charles, and S. Benhardsson, eds., Les Editions de Physique, Paris, France, pp. 7993 (1991). 43. P. H. Pumphrey and K. N. Akhurst: Aging kinetics of CF3 cast stainless steel in temperature range 300400C. Mater. Sci. Technol. 6:211220 (1990). 44. M. K. Miller, J. Bentley, S. S. Brenner, and J. A. Spitznagel: Long term thermal aging of type CF8 stainless steel. J. Phys. 45-C9:385390 (1984). 45. T. J. Godfrey and G. D. W. Smith: The atom probe analysis of cast duplex stainless steels. J. Phys. 47C7:217222 (1986). 46. M. K. Miller and J. Bentley: Microstructural characterization of primary coolant pipe steel. J. Phys. 47-C7:239244 (1986). 47. J. M. Sassen, M. G. Hetherington, T. J. Godfrey, G. D. W. Smith, P. H. Pumphrey, and K. H. Akhurst: Kinetics of Spinodal Reaction in the Ferrite Phase of a Duplex Stainless Steel. Proc. of properties of stainless steels in elevated temperature service,

200
M. Prager, ed., PVP-ASME, New York, p. 132, pp. MPC 26, pp. 6578 (1978). 48. P. Auger, F. Danoix, A. Menand, S. Bonnet, J. Bourgoin, and M. Guttmann: Atom probe and transmission electron microscopy study of aging of cast duplex stainless steels. Mater. Sci. Technol. 6:301313 (1990). 49. H. M. Chung and T. R. Leax: Embrittlement of laboratory and reactor aged CF3, CF8 and CF8M duplex stainless steels. Mater. Sci. Technol. 6:249262 (1990). 50. M. K. Miller and J. Bentley: APFIM and AEM investigation of CF8 and CF8M primary coolant pipe steels. Mater. Sci. Technol. 6:285292 (1990). 51. G. Slama, P. Petrequin, and T. Mager: In Structural mechanics in reactor technologyPost conference seminar 6Assuring structural integrity of steel reactor pressure boundary components, Monterey, CA (August 1983). 52. D. Blavette, B. Deconihout, A. Bostel, J. M. Sarrau, M. Bouet, and A. Menand: The tomographic atom probe: A quantitative three dimensional nanoanalytical instrument on an atomic scale. Rev. Sci. Instrum. 64:29112919 (1993). 53. D. Blavette, G. Grancher, and A. Bostel: Statistical analysis of atom-probe data (I): Derivation of some fine scale features from frequency distributions for finely dispersed systems. J. Phys. 49-C6: 433438 (1988). 54. J. Piller and H. Wendt: Autocorrelation Analysis of Atom-Probe Concentration Profiles. Proc. of 29th IFES, H.-O. Andren and H. Norden, eds., Almquist and Wiksell, Stockholm, pp. 265274 (1982). 55. M. G. Hetherington and M. K. Miller: On the statistical analysis of atom-probe data. J. Phys. 48-C6: 559564 (1987). 56. C. J. Dalzell: Statistical analysis of an atom-probe study of iron chromium. J. Phys. 49C6:411414 (1988). 57. S. Hill and B. Ralph: Continuous phase separation in a nickel aluminium alloy. Acta Metall. 30:2219 2225 (1982). 58. D. Blavette, S. Chambreland, A. Loiseau, J. Planes, and F. Ducastelle: FIM atom probe investigation of long period superstructures in Cu3xPd. J. Phys. 50-C8:365370 (1989). 59. M. Oehring and L. V. Alvensleben: Evaluation of atom-probe concentration profiles by autocorrelation analysis. J. Phys. 49-C6:415421 (1988). 60. F. Danoix, B. Deconihout, A. Bostel, and P. Auger: Some new aspects on microstructural and morphological evolution of thermally aged duplex stainless steels. Surface Sci. 266:409415 (1992). 61. J. E. Epperson, J. S. Lin, and S. S. Cooper: The fine scale microstructure in cast and aged duplex stainless steels investigated by S.A.N.S. In Radiation Induced Changes in Microstructures: 13th Int. Symp. (part I), F. A. Garner, N. H. Packan, and A. S. Ku-

F. Danoix and P. Auger


mar, eds., A.S.T.M., Philadelphia, PA, pp. 595614 (1987). 62. J. S. Langer: Theory of Spinodal Decomposition in Alloys. Ann. Phys. 65:5386 (1971). 63. P. P. Camus, W. A. Soffa, S. S. Brenner, and M. K. Miller: Quantification of interconnected microstructures by FIM. J. Phys. 45-C9:265268 (1984). 64. M. G. Hetherington and M. K. Miller: Some aspects of the measurement of composition in the atom-probe. J. Phys. 50-C8:535540 (1989). 65. A. Cerezo and M. G. Hetherington: Visualisation and analysis of 3-dimensional atom probe data. J. Phys. 50-C8:523528 (1989). 66. M. K. Miller and M. G. Hetherington: Morphological and scaling behaviour of ultrafine isotropic microstructures in FeCr alloys from atom-probe field ion microscopy data. Scripta Metall. 24:1375 1380 (1990). 67. M. K. Miller: The development of Atom Probe Field-Ion Microscopy. Mater. Charact. 44:1127 (2000). 68. P. Auger, A. Menand, and D. Blavette: Statistical analysis of atom-probe data (II): Theoretical frequency distributions for periodic fluctuations and some applications. J. Phys. 49-C6:439444 (1988). 69. F. Danoix: Phenomne de dcomposition de la ferrite des aciers austeno-ferritiques: Une tude par microscopie ionique et microanalyse la sonde atomique. Thse de doctorat, Universit de Rouen, Rouen, France (1991). 70. M. G. Hetherington, J. M. Hyde, M. K. Miller, and G. D. W. Smith: Measurement of the amplitude of a spinodal. Surface Sci. 246:304314 (1991). 71. P. Auger, F. Danoix, M. Guttmann, and D. Blavette: 300C400C Aging of the Ferrite of a Duplex Mo-Bearing Steel: An Atom-Probe Investigation. Proc. Duplex Stainles Steels 91, Beaune (Fr), J. Charles, and S. Benhardsson, eds., Les Editions de Physique, Paris, pp. 101109 (1991). 72. J. E. Brown and G. D. W. Smith: Atom-probe study of spinodal processes in duplex stainless steels and in single- and dual-phase FeCrNi alloys. Surface Sci. 246:285291 (1991). 73. K. H. Park, J. C. LaSalle, L. H. Schwartz, and M. Kato: Mechanical properties of spinodally decomposed Fe-30wt% Cr: Yield stength and embrittlement. Acta Metall. 34:18531865 (1986). 74. J. Bentley, M. K. Miller, S. S. Brenner, and J. A. Spitznagel: Proc. of 43rd Annual Meeting of the Electron Microscopy Society of America, G. W. Bailey, ed., San Francisco Press Inc., San Francisco, CA, pp. 328329 (1985). 75. F. X. Spiegel, D. Bardos, and P. A. Beck: Ternary G and E silicides and germanides of transition elements. Trans. Met. Soc. AIME 227:575 (1963). 76. M. Vrinat, R. Cozar, and Y. Meyzaud: Precipitated phases in the ferrite of aged cast duplex stainless steels. Scripta Metall. 20:11011106 (1986).

AP Studies of Fe-Cr-X and Stainless Steels


77. J. M. Vitek: G-phase formation in aged type 308 stainless steel. Met. Trans. A 18:154156 (1987). 78. H. M. Chung and O. K. Chopra. Proc. of Environmental Degradation of Materials in Nuclear Power SystemsWater Reactors, G. J. Theus and J. R. Weeks, eds., The Metallurgical Society, Pittsburgh, PA, pp. 359370 (1988). 79. F. Danoix, S. Chambreland, J. P. Massoud, and P. Auger: Influence of Ni, Mo and Other Elements on the Microstructural Evolution of Duplex Stainless Steels Aged at 300400C: An Atom-Probe Investigation. Proc of Duplex Stainles Steels 91, Beaune (Fr.), J. Charles, and S. Benhardsson, eds., Les Editions de Physique, Paris, pp. 111119 (1991). 80. F. Danoix, P. Auger, and D. Blavette: An atomprobe investigation of some correlated phase transformations in Cr, Ni, Mo containing supersaturated ferrites. Surface Sci. 266:364369 (1992). 81. J. M. Vitek, S. A. David, D. J. Alexander, J. R. Keiser, and R. K. Nanstad: Low temperature aging behavior of type 308 stainless steel weld metal. Acta. Metall. Mater. 39:503516 (1991). 82. K. Stiller, F. Danoix, and A. Bostel: Investigation of

201
precipitation in a new maraging stainless steel. Appl. Surface Sci. 94/95:326333 (1996). 83. K. Stiller, M. Hattestrand, and F. Danoix: Precipitation in 9Ni-12Cr-2Cu maraging steels. Acta Mater. 46:60636073 (1998). 84. K. Stiller, F. Danoix, and M. Httestrand: Mo Precipitation in a 12Cr-9Ni-4Mo-2Cu maraging steel. Mater. Sci. Eng. A. 250:2226 (1998). 85. M. K. Miller and M. G. Burke: Proc. of 5th Int. Symp. on the Environmental Degradation of Materials in Nuclear Power SystemsWater Reactors, Sept. 1991 Monterey, E.P. Simonen, ed., American Nuclear Society, La Grange Park, pp. 689695 (1992). 86. M. Murayama, Y. Katayama, and K. Hono: Microstructural evolution in a 17-4 ph stainless steel after aging at 400C. Met. Mater. Trans. A. 30A:345353. 87. M. Hdin: Sensibilit aux conditions initiales de lvolution structurale de la ferrite daciers austno-ferritiques vieillis dans le domaine 300 400C. Thse de doctorat, Universit de Rouen, Rouen, France (1998). Received November 1998; accepted December 1998.

Potrebbero piacerti anche