1. Introduction has been developed [2]. It includes the resulting physical
damage parameters previously identified. Nickel-based superalloys are considered for high tem- perature gas turbines of nuclear gas cooled reactors (GCR). The gas turbine disks, with a diameter about 1.5 m, are 2. Material and experimental methods expected to work at a maximum temperature of 700 ◦ C during around 60 000 h without significant maintenance. The chemical composition of U720 is given in Table 1. Powder metallurgy (PM) processing seems appropriate to Powder particles are obtained by atomization under argon produce segregation-free parts, with a geometry close to the atmosphere. The particle diameter ranges from 10 to 90 m final shape of the component. Since a material with high with an average diameter of about 30 m. Powder billets creep strength is required for this application, PM Udimet used in this study have been consolidated using the hot 720 LI (U720) is considered as a potential superalloy [1]. isostatic pressing (HIP) facility of Aubert&Duval-Holding The creep behavior of U720 was therefore investigated by under 1400 bar at 1100 ◦ C during 3 h. conducting creep tests at 650 ◦ C on two microstructures ob- Both heat treatments applied are the usual ones for tained from two different heat treatments. The deformation cast and wrought U720 [3], known as high strength (HS), and damage mechanisms have been analysed, and a creep improved for tensile strength and low cycle fatigue re- predictive model based on McLean and Dyson approach sistance, and creep resistant (CR), improved for creep strength. HS treatment consists of a 4 h solutioning treat- ment at 1110 ◦ C, followed by two precipitation heat treat- ∗ Corresponding author. Tel.: +33 4 38 78 35 59; ments for 24 h at 650 ◦ C and 16 h at 760 ◦ C. The CR fax: +33 4 38 78 58 91. heat treatment consists of a 4 h solutioning at 1170 ◦ C, E-mail address: raphael.couturier@cea.fr (R. Couturier). then a second 4 h solutioning at 1080 ◦ C, followed by two
doi:10.1016/j.msea.2004.01.094 600 S. Dubiez-Le Goff et al. / Materials Science and Engineering A 387–389 (2004) 599–603
Table 1 Chemical composition (wt.%) of the Udimet 720 LI Ni Bal. Cr 16.2 Co 15.3 Ti 5.18 Mo 3.06 Al 2.47 W 1.33 Fe 0.06 Zr 0.039 C 0.023 Cu 0.02 B 0.018 Si 0.007
precipitation heat treatments for 24 h at 845 ◦ C and 16 h at
760 ◦ C. Creep tensile tests have been performed under air at 650 ◦ C on cylindrical specimens (Ø4 mm) after either HS or CR heat treatments under stresses ranging from 600 to 900 MPa. Stress range applied to the specimens was se- lected to reach fracture between 10 and 5000 h. Creep tests have been conducted under constant load and tensile tests have been performed at 5 × 10−4 s−1 . To analyse the dislocation structure by transmission elec- tron microscopy (TEM) in the strained specimens, thin foils were machined out from specimens strained up to 1% and 3% by creep and tensile tests.
Fig. 1. ␥ distribution in U720 HS (a) and CR (b)—TEM dark-field
3.1. Microstructures than the CR one. Moreover, U720 CR shows the usual three stages of creep whereas U720 HS shows a transition directly HS and CR heat treatments, mainly differing by their so- from primary to tertiary creep, exhibiting a point of inflection lutioning temperature, resulted in two microstructures. Sub- corresponding to the minimum creep rate. solvus solutioning of the U720 HS leads to a ␥-grain size ranging from 1 to 10 m. Supersolvus solutioning of the U720 CR leads to ␥-grain size ranging from 10 to 30 m. In both microstructures, ␥ precipitates constitute about 50% of the volume fraction of the alloy. Fig. 1 shows the distribution of the ␥ in the two microstructures. For U720 HS, there are three different types of ␥ (Fig. 1a) : primary ␥ formed during HIP treatment, with a diameter of 450 nm, then secondary and tertiary ␥ nucleated during cooling after solutioning treatment and coarsened during ageings, with respective average diameters of 40 and 10 nm. For U720 CR, only two types of ␥ are formed during cooling (Fig. 1b) : the secondary ␥ with a diameter of 300 nm and the tertiary ones with a diameter of 30 nm.
3.2. Creep behavior at 650 ◦ C
Strain versus time curves of 650 ◦ C creep tests performed
on U720 HS and U720 CR show two different behaviors Fig. 2. Creep curves at 650 ◦ C of the Udimet 720 HS and CR under (Fig. 2): under the same stress, U720 HS creep life is longer stress ranging from 600 to 900 MPa. S. Dubiez-Le Goff et al. / Materials Science and Engineering A 387–389 (2004) 599–603 601
4.1. Early damage by micro-cracking and cavity
nucleation of the specimen
This hypothesis does not seem relevant at 650 ◦ C, as only
few micro-cracks are observed very close to the fracture sur- face on longitudinal cross-section of broken specimens. At 750 ◦ C, a lot of cavities and cracks are observed in the bulk of the specimen after failure. A specific test was therefore con- ducted to show whether the observed damage is responsible for the observed softening. A creep test (750 ◦ C/280 MPa) has been stopped at 2/3 of the creep life. Then, the mi- crostructure has been restored by applying the full HS heat treatment cycle to the specimen (under vacuum), and the creep test was started again until rupture under the same conditions. The behavior of this rejuvenated specimen looks like the one of the virgin material. This suggests that the specimen creep rate is not influenced by cracks or voids af- ter 2/3 of creep life, when softening is already observed.
4.2. Degradation of the microstructure related to γ ageing
during creep deformation
Creep softening of superalloy is often related to ageing
as ␥ coarsen during the test. Nevertheless, at 650 ◦ C, TEM observations have shown that ageing up to 5000 h has no measurable effect on ␥ precipitate size.
Fig. 3. Dislocation structures in the Udimet 720 HS (a) and Udimet 720 4.3. Increase of the true stress during deformation CR (b) (TEM weak-beam) in interrupted specimens crept after 1% strain under 750 MPa at 650 ◦ C. In this study, creep tests were not conducted under con- stant stress, but under constant load. The applied stress thus 3.3. Dislocation structures after creep and tensile increases as the strain increases and the section decreases. tests at 650 ◦ C The true stress is defined as σ = σ 0 exp(ε), with initial stress σ 0 and true strain ε. As proposed by Maldini and Lupinc TEM analysis of crept specimen strained at 1% under [4], the true strain rate of the creep tests shown in Fig. 2 are 750 MPa show large stacking faults inside the ␥ for U720 plotted in Fig. 4 as a function of true stress. The decrease HS (Fig. 3a) and dislocation loops around the ␥ for the U720 of the strain rate corresponds to the primary stage, then the CR (Fig. 3b). This suggests that the deformation mechanisms increase of the strain rate corresponds to the secondary and differ from one microstructure to the other: U720 HS ␥ are tertiary stages. On this representation, creep rate was drawn sheared whereas U720 CR ␥ are bypassed by an Orowan using Norton stress dependence law, as ε̇ ∝ σ n , with Nor- mechanism. ton coefficient n adjusted to the experimental steady-state or Crept specimen of U720 HS strained at 3% under 750 MPa minimum creep rate. For U720 CR, n is 20 for the applied shows even more stacking faults than after 1% strain. By stress range. But a single n-value cannot describe U720 HS contrast, for U720 HS specimen prestrained at 3% in tensile, stress dependence at all stresses. Thus, for 850 and 900 MPa, dislocations form loops around the ␥ . n is 24 and for 600, 750 and 800 MPa, n is 12. The plot of Fig. 4 enables to emphasize the two different kinds of creep behavior. For U720 CR, after primary stage, curves 4. Discussion fall on the Norton’s law plot. This means that the increase of the strain rate observed in Fig. 2 corresponds in fact to a To understand why the creep rate quickly increases after steady-state stage, and that acceleration of creep is only due the minimum creep rate for HS microstructure, four dif- to the increase of the true stress. U720 HS behavior differs ferent hypotheses of softening mechanisms are hereafter from high stresses to low stresses. At 850 and 900 MPa, the investigated: early damage of the specimen, degradation of apparent tertiary stage is linked to the stress increase during the microstructure related to ageing, increase of the true the creep test, as observed for U720 CR. In contrast at 600, stress during creep deformation and interaction between 750 and 800 MPa, strain rate increases more quickly than dislocations and ␥ . Norton’s law prediction. For this stress range, an additional 602 S. Dubiez-Le Goff et al. / Materials Science and Engineering A 387–389 (2004) 599–603
crossing of other dislocations in the same glide plane. In
addition, a higher stacking fault density is observed in the 3% crept U720 HS than in the 1% crept, indicating that there is no localization of the deformation. McLean and Dyson proposed an increase of the mobile dislocation density to explain the softening [2].
5. Modeling
McLean and Dyson proposed a specific model to pre-
dict creep behavior of complex alloy: a physically based continuum creep damage model with internal state vari- ables which relates the strain rate response to the applied stress [2]. The equation of the chosen model results in a set of a hardening parameter H to describe the primary creep stage and a parameter Dd to describe the multiplication of the mobile dislocation density, only for U720 HS at low stress: Fig. 4. True strain rate as a function of the true stress for 650 ◦ C creep
tests. The rates obtained from Norton law is also drawn as dashed lines. ε̇0 σ(1 − H) ε̇ = sinh (1) 1 − Dd σ0 damage parameter is required to describe the softening of the alloy during creep. h H Ḣ = 1 − ∗ ε̇, Ḋd = C(1 − Dd )2 ε̇ (2) σ H 4.4. Dislocation/γ interactions
TEM analyses indicate that the difference of behavior at
750 MPa is related to different deformation mechanisms: U720 CR ␥ are by-passed by an Orowan mechanism whereas U720 HS ␥ are sheared. These two mechanisms, already observed in other recent PM nickel-based super- alloys [5,6], seem to be strongly related to the presence of tertiary ␥ . The large ␥ corridors observed in U720 CR promote an Orowan mechanism whereas the presence of tertiary ␥ in U720 HS decreases the corridor width and promotes a shearing mechanism. According to the previ- ous investigation, Orowan mechanism observed in U720 CR could be connected to a classical behavior with a steady state stage, whereas shearing observed in U720 HS could be connected to a softening mechanism. However, stress-assisted tertiary stage also appears at high stress for U720 HS and could thus be related to Orowan mechanism as for U720 CR. This is confirmed by TEM observations of interrupted tensile U720 HS specimen showing Orowan loops around ␥ . The deformation mechanisms hence de- pend on the stress level. The transition between the two mechanisms should be the Orowan stress.
4.5. Possible origins of the softening observed at low
strain rate
Particle shearing during deformation is usually associated
with softening of alloys, because particle size decreases in the glide plane. However, shearing in superalloys implies Fig. 5. Comparison of the model predictions with experimental data of complex stacking faults [7] which may not facilitate the U720 HS and CR. S. Dubiez-Le Goff et al. / Materials Science and Engineering A 387–389 (2004) 599–603 603
Table 2 6. Conclusions Model parameter set for U720 HS and CR used to generate the curves in Fig. 5 Creep behavior of PM U720 depends on its microstruc- ε̇0 (s−1 ) σ 0 (MPa) h (MPa) H∗ C ture and stress range. U720 CR and U720 HS crept at high CR 1.7 × 10−13 34.0 58835 0.08 0 stress are deformed by an Orowan mechanism and present a HS 1.8 × 10−13 33.7 58357 0.1 1600 steady state stage. By contrast, U720 HS behavior presents a tertiary stage at low stress that can be related to a ␥ shear- ing mechanism. These behaviors can be fairly well predicted by McLean and Dyson creep model by selected physically based parameters. with characteristic rate ε̇0 , true stress σ, kinematic back stress σ 0 , normalized saturation stress H∗ , effective Young modulus h , and a material constant C. Best parameter sets References identified are given in Table 2. For U720 CR and U720 HS creep at high stress, both [1] R. Couturier, C. Escaravage, in: Proceedings of the IAEA Technical having a classical behavior, no damage parameter is required Committee Meeting, IAEA, Palo Alto, CA, 2000. to describe the full creep curve. For U720 HS at low stress, [2] M. McLean, B.F. Dyson, Eng. Mater. Technol. 122 (2000) 273–278. softening is modeled by an increase of mobile dislocation [3] M.P. Jackson, R.C. Reed, Mater. Sci. Eng. A 259 (1999) 85–97. density. [4] M. Maldini, V. Lupinc, H. Watt, University Edinburgh, UK, 1999, pp. 109–118. Fig. 5 shows experimental creep curves compared with the [5] P.R. Bhowal, E.F. Wright, E.L. Raymond, Metal. Trans. A 21A results of the simulation, using the proposed model: creep (1990) 1709–1717. behavior of U720 CR and U720 HS at 850 and 900 MPa is [6] D. Locq, M. Marty, P. Caron, in: Proceedings of the Superalloys fairly well predicted, but creep behavior of U720 HS at 750 2000, TMS, Metals Park, OH, 2000, pp. 395–403. and 800 MPa has to be further analysed. [7] M. Condat, B. Décamps, Scripta Metal. 21 (1987) 607–612.
