- Email: [email protected]
Microstructure and long-term creep properties of 9–12% Cr steels
ARTICLE IN PRESS
International Journal of Pressure Vessels and Piping 85 (2008) 30–37 www.elsevier.com/locate/ijpvp
Microstructure and long-term creep properties of 9–12% Cr steels$ J. Hald Elsam/Energy E2/IPL-MPT, Technical University of Denmark, Denmark
Abstract Advanced microstructure characterisation and microstructure modelling has demonstrated that long-term microstructure stability in 9–12% Cr steels under technical loading conditions is equivalent to precipitate stability. Mo and W can have a positive influence on longterm creep strength of 9–12% Cr steels by Laves phase precipitation hardening. Unexpected breakdown of long-term creep stability of a number of alloys is caused by precipitation of the complex Z-phase nitride, which may completely dissolve fine V and Nb containing MX nitrides. High Cr contents of 10% and above in the steels accelerate Z-phase precipitation. r 2007 Elsevier Ltd. All rights reserved. Keywords: 9–12% Cr steels; Long-term creep stability; Creep mechanisms; Precipitate characterisation; Precipitate stability modelling; Laves phase; MX nitrides; Z-phase
1. Introduction In the past 40 years, martensitic creep resistant 9–12% Cr steels have found increasing application in thick-section components like piping, large forgings and castings of steam power plants. Beginning with the 12CrMoV steel introduced in power plants in the mid-1960s, alloy development over the past 30 years has led to new steam pipe steels like the modified 9Cr steel P91, introduced in plants in 1988s, to the tungsten-modified 9Cr steels P92 introduced in plants in 2001 and E911 introduced in 2002. Similar steels have been developed and applied for large forgings and castings of steam turbines. The steadily improved creep rupture strength of new martensitic 9–12% Cr steels has been used to construct new advanced fossil-fired steam power plants with higher efficiency. Increase in steam parameters from subcritical 180 bar/530–540 1C to ultra-supercritical values of 300 bar and 600 1C has been realised, and this has led to efficiency increases from 30–35% to 42–47%, equivalent to approximately 30% reduction in specific CO2 emission. With the expected continuous increase in the demand for coal for electricity production, especially in the fast developing $ This article appeared in its original form in Creep & Fracture in High Temperature Components: Design & Life Assessment Issues, 2005. Lancaster, PA: DEStech Publications, Inc. E-mail address: [email protected]
0308-0161/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijpvp.2007.06.010
Asian countries, this technology progress has clearly the potential to become one of the most (if not the most) significant countermeasures against global CO2 emissions in the coming decades. Design strength values for the new W-modified 9Cr steels E911 and P92, published at this conference, are now based on comprehensive long-term creep rupture testing up to 100,000 h. However, the decisions to introduce the steels in plants had to be based on shorter test results leading to less certainty about the long-term stability of the steels. This fact, together with conflicting ideas about long-term strengthening mechanisms of especially tungsten and the quest for even better martensitic steels, has led to intensive research into the microstructure stability of the 9–12% Cr steels under creep exposure. New developments of better microstructure characterisation methods and improved microstructure models have led to significant progress in the understanding of microstructure stability of this class of alloys as presented below. 2. Microstructure and creep mechanisms of 9–12% Cr steels 2.1. Microstructure Chemical composition, heat treatment and creep rupture strength of a number of 9–12% Cr steels are given in Table 1. The microstructure of the 9–12% Cr steels is
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37 Table 1 Chemical composition and heat treatment of a number of 9–12% Cr steels Mass (%)
C Cr Mo W Ni V Nb N B Austenitisation (1C) Pre-tempering (1C) Tempering (1C) sB=105 h=600 C (MPa)
Steam pipe steels
Turbine rotor steels
12CrMoV P91 E911 P92
Steel E Steel F Steel B
0.2 11 0.9 – 0.5 0.3 – – – 1050 – 750 59
0.1 10 1 1 0.6 0.2 0.05 0.05 – 1070 570 690 E95
0.1 9 0.9 – 0.1 0.2 0.05 0.06 – 1050 – 750 94a
0.1 9 1 1 0.3 0.2 0.05 0.07 – 1060 – 770 98
0.1 9 0.5 1.8 0.05 0.2 0.06 0.06 0.001 1065 – 770 113
0.1 10 1.5 – 0.7 0.2 0.05 0.05 – 1120 570 690 E95
0.2 9 1.5 – 0.1 0.2 0.05 0.02 0.01 1120 590 700 E125
The quoted creep rupture strength values for the pipe steels P91, E911 and P92 are from the most recent evaluations by the ECCC. a The most recent ECCC evaluation of steel P91 is from 1995. A new evaluation being made at present.
Fig. 1. Tempered martensite—TEM micrograph.
tempered martensite formed during a final normalising and tempering heat treatment, Fig. 1. After normalising at about 1050 1C, air hardening will lead to a martensitic transformation in sections up to approximately 100 mm thickness due to the chromium content. Tempering in the range 680–780 1C leads to recovery of ductility by annihilation of dislocations and to the formation of ferrite subgrains. Tempering temperatures in the low end of the range are used for components like turbine rotors, where high tensile strength is required. The high end of the range is used for pressurised components like steam pipes, where high toughness is necessary. Carbide and nitride particles precipitate in the steels during tempering, on prior austenite grain boundaries, ferrite subgrain boundaries and on dislocations inside subgrains. When the steels are put into operation in power plants at temperatures below the final tempering temperature, further particles may precipitate, which are thermodyna-
31
mically unstable at the tempering temperature, like e.g. the intermetallic Laves phase. From Table 1, it is clear that minor alloy additions of Nb and N produced the first strength increase from 12CrMoV to P91, and the second increase from P91 to P92 was achieved by partly replacing Mo by W and adding B. The strongest turbine steel B is based on high content of boron. 2.2. Creep and strengthening mechanisms Historically, it has often been assumed that for isothermal creep rupture tests a linear relation should exist between the logarithm of time to rupture and the logarithm of stress if the tested material was metallurgically stable. Inspections of creep rupture data for all technical hightemperature steels show that none of these are metallurgically stable at relevant temperatures. It has been tried with some success to explain the deviations from linearity by a change in fracture mechanism from ductile fracture at shorter testing times to intergranular cavitation at longer testing times. Further explanations include changes in the controlling creep mechanism of the steel from dislocation creep to viscous creep. The technically relevant stresses and temperatures for creep testing and service exposure of the 9–12% Cr steels are in the range 300 MPa/500 1C–50 MPa/650 1C. Relevant loading conditions in this range will lead to rupture times between 100 and 300,000 h and minimum creep rates between 3 106 and 5 1012 s1. In the authors’ opinion, there are strong indications that the creep mechanism in the whole of the technically relevant stress/ temperature range is dislocation creep. Microstructural sources of creep deformation are then the migration of dislocations and subgrain boundaries [1,2]. Subgrain boundaries consist of dislocation networks, which, compared with the subgrain interiors, are hard regions in the microstructure of ferritic steels. A small subgrain size is equivalent to a high-volume fraction of hard regions and could thus be expected to provide high creep strength. This is confirmed by heat treatment experiments with 9Cr steels, where the tempered martensite subgrain microstructure is removed and the creep strength is drastically reduced. However, dislocation densities and subgrain sizes of different 9–12% Cr steels in the normalised and tempered condition do not show systematic variation with creep strength. A 12CrMoV steel may have finer subgrain size than steel P92, but lower creep strength. During creep, the subgrains grow and the dislocation density decreases with strain accumulation in a similar manner for all the 9–12% Cr steels [1]. It is then the ability of a steel to maintain a small subgrain size and a high dislocation density for a long time at stress and temperature, which is significant to high creep strength. This indicates that microstructural explanations for the improved creep strength of the new 9–12% Cr steels should be found in mechanisms that retard the migration of dislocations and subgrain boundaries, and thus delay the
ARTICLE IN PRESS 32
J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
accumulation of creep strain with time. Such mechanisms include (i) solid solution strengthening by formation of clouds of solute atoms around dislocations and (ii) interactions with precipitate particles. 2.2.1. Solid solution strengthening vs. precipitation hardening Solid solution strengthening has often been referred to in discussions of the effect of Mo and W on creep strength of 9–12% Cr steels. It has since long been clear that during creep exposure at temperatures around 600–650 1C, most of the Mo and W in the steels will precipitate as intermetallic Laves phase ((Fe,Cr)2(Mo,W)). The dominating opinion was that this would cause creep instability in the steels because the solid solution strengthening effect from Mo and W would be lost, and the precipitation strengthening effect from Laves phase was believed to be insignificant. This opinion seemed to be supported by the breakdown of long-term creep strength of some W alloyed 9–11Cr steels, and by the excellent long-term stability of the Mo alloyed 9CrB steel B (Mo causes less Laves phase precipitation than W). However, the excellent long-term stability of the W-alloyed 9Cr steel P92 seemed to contradict the opinion, and experimental evidence to demonstrate and quantify the solid solution strengthening mechanism is sparse. A rough estimation of the solid solution strengthening effect from Mo and W can be obtained by considerations based on recently obtained experimental in-situ observations of dislocation movement in ferrite [4]. The back stress s* acting on a dislocation moving with velocity v is given by s ¼ v=B,
(1)
where B is the dislocation mobility (m/Pa s). The mean dislocation velocity is a function of the creep rate _ and the mobile dislocation density r: _ ¼ rbv=M,
(2)
where b is the Burgers vector and M is a Taylor factor. Eqs. (1) and (2) give the back stress as a function of the creep rate: M _ . (3) s ¼ rbB Measurements of dislocation velocities in [4] give a value of B at 600 1C in ferrite of 5 1018 m/Pa s for a W content of 2 atm%. Loss of W from solid solution will increase B and lower the back stress. In [3], B is estimated to be 2.5 1019 m/Pa s. In [5], the dislocation density is measured to be higher than 1 1014 m2 in steel P92. For a long-term creep test at 410,000 h at 600 1C/o150 MPa the minimum creep rate in steel P92 is below 0.0001%/ h ¼ 2.8 1010 s1. With M ¼ 3, b ¼ 0.254 nm and B ¼ 2.5 1019 m/Pa s, Eq. (3) gives s o0:13 MPa ðlongterm test 410; 000 h=600 C=o150 MPaÞ.
This back stress should be considered as a maximum value with all tungsten in solid solution. The very low value appears because the solute tungsten atoms diffuse at a rate similar to the dislocation velocity. It can be concluded that solid solution strengthening from W and Mo has no significant effect on long-term microstructure stability of the 9–12% Cr steels. Precipitation hardening by pinning of dislocations and subgrain boundaries should be regarded as the most significant strengthening mechanism in 9–12% Cr steels, and microstructure stability of the 9–12% Cr steels under creep load is equivalent to precipitate stability. This is consistent with two findings: (a) the compositional changes, which have improved the creep strength of the 9–12% Cr steels, have also resulted in clear changes in the precipitate populations and (b) breakdowns in creep stability presented below can be explained by unexpected precipitate reactions. 3. Precipitate stability The interaction between precipitate particles and subgrain boundaries or dislocations under creep conditions can be complicated, and a detailed discussion of interaction mechanisms and back stress formulations will not be attempted here. However, equations to describe the contribution of precipitate particles to creep strength have similarities to the Orowan stress pffiffiffiffiffi fp sOrowan ¼ 3:32 G b , (4) dp where G is the shear modulus, b is the Burgers vector, fp is the precipitate volume fraction and dp is the mean precipitate diameter. This formula demonstrates that the basic information needed to describe the effect of precipitate particles on creep strength is their volume fraction and their mean size. Predictions of long-term stability of precipitate particles and their influence on creep stability of 9–12% Cr steels have to rely on measurements of volume fractions and mean particle sizes at certain points in time and on models describing the evolution of these parameters as functions of temperature, time and stress. 3.1. Precipitate characterisation The new 9–12% Cr steels contain several precipitate types, which form either during the final normalising and tempering heat treatment or during subsequent creep, see Table 2. For the validation of models of long-term precipitate stability it is necessary to know volume fractions and mean particle sizes of the individual precipitate types. With the newly developed energy-filtered transmission electron microscope (EFTEM), it is now possible to make accurate measurements of particle sizes down to a few nm, and on-line discrimination between the individual precipitate types [6]. Due to technical constraints there is a lower limit to the usable magnification in EFTEM microscopes, which means
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37 Table 2 Main precipitates in some 9–12% Cr steels Steel
Precipitate Formula
12CrMoV M23C6 P91 M23C6 MX
P92
(Cr,Fe,Mo)23C6 (Cr,Fe,Mo)23C6 (Nb,V)(N,C)
Remark Precipitate during tempering Precipitate during tempering Undissolved during austenitisation Precipitate during tempering Precipitate during creep (To650 1C)
MX Laves phase
(V,Nb)(N,C) (Fe,Cr)2Mo
M23C6 MX
(Cr,Fe,Mo,W)23C6 Precipitate during tempering (Nb,V)(N,C) Undissolved during austenitisation (V,Nb)(N,C) Precipitate during tempering (Fe,Cr)2(Mo,W) Precipitate during creep (To720 1C)
MX Laves phase
that statistically sound size determination for a population of particles with mean diameters larger than approximately 200 nm become unpractical with the EFTEM method. However, in recent years, the introduction of new highintensity electron sources in the scanning electron microscope (SEM) has improved the resolution power of this instrument. This means that mean particle sizes down to approximately 100-nm diameter can be measured with SEM. Particle discrimination in the SEM can be obtained from the atomic number contrast in backscattered electron images [7]. 3.2. Modelling of precipitate stability Modelling of long-term precipitate stability should include predictions of phase stability, nucleation rates, growth rates and coarsening rates for precipitate phases as functions of chemical composition, temperature and stress. 3.2.1. Phase stability Recent developments of databases and software, like the Thermocalc [8], to calculate thermodynamic equilibria in multicomponent systems allow predictions of phase stability for 9–12% Cr steels based only on chemical composition. Accuracy of the predictions depends critically on the available phase descriptions in the thermodynamic databases. A few critical gaps and inaccuracies in these descriptions have been identified, e.g. the predictions of Laves phase stability in 9–12% Cr steels alloyed with molybdenum but no tungsten has been proven wrong by experiments. In such cases, new thermodynamic phase descriptions are needed [9]. Otherwise, the thermodynamic calculations have demonstrated accurate predictions of stability range, amount and chemical composition of all phases in the 9–12% Cr steels [10]. These predictions are fundamental to the predictions of long-term stability of precipitate particles as demonstrated below.
33
3.2.2. Nucleation Modelling of nucleation rates is challenging because it depends critically on the interfacial energy, which is very difficult to predict or determine experimentally. It is therefore most often necessary to make simplifying assumptions about nucleation. However, with thermodynamic equilibrium calculations, the driving force for nucleation can be calculated, and based on the normal ranges for the interfacial energy gE0.1–1 J/m2 critical nucleus sizes as well as nucleation rates can be estimated. More ambitious modelling of nucleation has to include strain energy and nucleus shape [11]. 3.2.3. Growth The author has previously modelled Laves phase growth in steel P92 using an Avrami-type equation together with thermodynamic calculation [12]. However, recent developments of multicomponent diffusion databases like the DICTRA [13] allow more advanced modelling. The model needs experimental input to determine the nucleus density or the final size of fully grown particles [9,14,15]. 3.2.4. Coarsening Particle coarsening is one of the most important degradation processes for creep-resistant steels, and a detailed understanding of the influence of chemical composition on coarsening rate is essential to the understanding of long-term microstructure stability. Coarsening can be modelled by DICTRA in a similar way as for growth by incorporation of a contribution from the interfacial energy to the Gibbs energy of the particle. This leaves the interfacial energy as the only fit parameter in the coarsening model [16]. However, the DICTRA calculations can be time consuming, and a simpler approach allows a quicker overview of coarsening rates for different steels. Particle coarsening is observed to follow the well-known Ostwald ripening equation: r3 r30 ¼ K p t,
(6)
where r is the precipitate radius at time t and r0 is the precipitate radius at time 0. Liftshitz and Slyozov [17] and Wagner [18] calculated the coarsening rate Kp for binary alloy systems, but only recently have extensions to multicomponent systems been obtained by Umantsev and Olson [19] and by A˚gren et al. [20]. The latter resulted in the following formula for the coarsening rate constant Kp in a C-component system of b precipitate particles in an a matrix: 8 gV bm i, Kp ¼ P h 9 C ðxb xa=b Þ2 =ðxa=b Di =RTÞ i i i i¼1
(7)
where g is the interfacial energy and V bm is the molar volume of the precipitate phase. Di is the diffusion coefficient of element iin the matrix, xbi is the mole fraction a=b
of element iin the precipitate and xi
is the mole fraction
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
34
of element iat the precipitate/matrix interface. The diffusion coefficients are extracted from the DICTRA database, and include effects of chemical composition. Volume diffusion is assumed. The effect of interface curvature on the equilibrium solubilities is relatively small in this case, a=b and the values of xbi and xi obtained by thermodynamic equilibrium calculations assuming a flat precipitate/matrix interface can be used as a good approximation. V bm is calculated from lattice parameter data and precipitate composition. The only unknown is then the interfacial energy g, which is used as a fit parameter when compared to experimental observations.
rupture strength between the steels. The Laves phases in steel P92 produce a significant precipitate strengthening effect, which together with the observed high stability of the M23C6 carbides provide the high creep strength of this steel. It is interesting to note that the loss of tungsten from solid solution in steel P92 actually contributes to the increased coarsening stability of the M23C6 carbides, which is mainly controlled by the ferrite matrix content of the substitutional elements, which have a high concentration in the carbide. The observed differences in Laves phase behaviour in steels P91 and P92 is explained by the lower solubility of the Mo Laves phases compared with W Laves phases, see Table 2. When the creep temperature is close to the solution temperature for Laves phase, nucleation of the phase is very difficult, and only few particles nucleate. This results in the extended growth phase and large mean particle size. Similar observations have been made for the 9CrMoB steel B [24]. This is also seen in the very strong effect of temperature on Laves phase particle size in steel P92, Fig. 3. At 650 1C, the Laves phase particles are approx. 60% larger than at 600 1C. This could explain an observed tendency to reduced rupture ductility after more than 10,000 h at temperatures 625 and 650 1C in steel P92 since large Laves phase particles can act at nucleation sites for creep cavities. It is clear from the discussion above that alloying with either W or B has a strong positive effect on the microstructure stability of 9–12% Cr steels. W produces fine stable Laves phase particles in steel P92—even finer than the M23C6 carbides in steel P91—when the creep temperature is sufficiently below the solubility temperature for this phase. The Laves phase particles contribute significantly to particle strengthening and indirectly to M23C6 carbide stability through removal of Mo and W from solid solution. B has a strong stabilising effect on the M23C6 carbides, even though the mechanism for this remains unclear. The MX precipitates appear to be extremely stable against coarsening and could probably be characterised as the backbone of the long-term stability of the new 9–12% Cr steels.
4. Evaluation of precipitate stability In recent years, experimental measurements of particle sizes of individual precipitate types have been obtained with the new microstructure characterisation methods, and the models described above have been used to evaluate the precipitate stability of a number of 9–12% Cr steels. Fig. 2 shows evaluations by measurements and modelling of the evolution in mean particle sizes of precipitates in steels P91 and P92 at 600 1C [21–23]. In both steels, the MX carbonitrides show a similar very high stability against coarsening. The M23C6 carbides show considerable coarsening in steel P91, whereas in steel P92 this carbide is highly stable, this is attributed to the presence of B in steel P91, which seems to improve the carbide coarsening stability. This probably explains the high stability of the rotor material steel B, but the exact mechanism for boron stabilisation is at present unknown. Large differences are found between the Laves phase particle sizes in the two steels. After an initial growth phase of approx. 10,000 h, the Laves phase particles in steel P92 are more stable against coarsening than the M23C6 carbides in steel P91. In steel P91, the Laves phase particles grow to very large sizes during an extended growth phase, which lasts approx. 30,000 h. The observed differences in particle stability can explain the observed differences in creep
P92 - MX 600°C
500
Mean particle radius, nm
Mean particle radius, nm
P91 and P92 Laves phase and M23C6 size - 600°C
400 300 200 100 0 0
10000
20000 30000 Time, h
40000
50000
30 25 20 15 10 5 0 0
10000
20000
30000 40000 Time, h
50000
60000
70000
Fig. 2. Left: Particle sizes of M23C6 (thin lines) and Laves phase (thick lines) in steels P91 and P92. Right: Size evolution of MX particles in steel P92. A similar MX evolution is observed in P91.
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
300x10-9
Diameter [m]
250 200 150 100 50 0 0
20
40
60
80x10 3
30
40x10 3
Time [h] 500x10-9
Diameter [m]
400 300 200 100 0 0
10
20 Time [h]
35
58 MPa/approx. 610 1C and after 12,000 h at 45 MPa/ 660 1C. The rupture times were only approx. 10% of the expected minimum lifetimes according to published ASME allowable stresses, Fig. 4. Inspection of the tubes showed no signs of extensive corrosion or oxidation attack, and temperature estimates based on oxide thickness measurements did not indicate any overheating. Detailed microstructure investigations showed extensive precipitation of the complex nitride Z-phase (Cr(V,Nb)N) in the steels, and this had led to an almost complete dissolution of MX nitrides. The T122 tubes contained 12% Cr. Parallel investigations of a steel P122 pipe with 11% Cr after isothermal exposure 650 1C/10,000 h showed only a very little sign of Z-phase formation, and a large population of MX was still present in the steel. It was concluded that the Z-phase transformation was responsible for the breakdown in creep stability of the T122 tubes. Since the only significant difference between the T122 and P122 steels was the Cr content it was further concluded that a high Cr content accelerates Z-phase formation. Severe creep instabilities have been observed in a number of experimental 10–12% Cr steels, e.g. the 11CrWCo steels TAF 650 and NF12. Is was for long believed that this was caused by Laves phase precipitation, but measurements of Laves phase particles in a steel similar to TAF650 have demonstrated an even more favourable particle size distribution of the Laves phase than in steel P92 [9]. Similarly, a large number of test alloys with 11–12% Cr, and with or without W, intended for operation at 650 1C have been manufactured and tested. All of these, which have been tested to long times, show a severe breakdown in creep stability between 5000 and 30,000 h at 650 1C. Recent investigations have identified precipitation of the complex
Fig. 3. Model of growth and coarsening of Laves phase in steel P92. The diameter of Laves phase vs. time. (a) 600 1C and (b) 650 1C. Full squares denote measured values.
P/T122 1000
The presented experimental observations and modelling produce improved understanding of the excellent stability of the strongest of the new 9–12% Cr steels, i.e. the P92 and rotor steel B, but the models fail to explain observed microstructure instabilities in a number of steels. These instabilities can however be attributed to precipitate reactions. Imbalances in composition may lead to formation of unwanted phases like AlN in steel P91 at the expense of MX nitrides [25], and unexpected phase transformations like the Z-phase formation may occur with a similar result as described below. 5.1. Z-phase precipitation In 2001, two tubes of steel T122 (12CrWMoCuVNbN) installed in a test superheater facility at a Danish power plant ruptured prematurely after 30,000 h exposure at
Rupture stress, MPa
5. Microstructure instabilities
100
10 28000
30000
32000
34000
36000
38000
40000
PLM = T(C+logt), K and h
Fig. 4. Mean rupture strength of steel P/T122 and observed lifetimes of T122 superheater test tubes.
ARTICLE IN PRESS 36
J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
nitride Z-phase ((Cr(V,Nb)N)) as the root cause of this instability in the 10–12% Cr steels. The Z-phase can completely dissolve the MX carbonitrides, which are major contributors to creep strength. The Z-phase precipitates as very large particles, which will not contribute to precipitation hardening, and which may each consume more than 1000 MX particles, Fig. 5. The Z-phase nitride in the form CrNbN has been known from Nb-alloyed austenitic stainless steels since 1950, and here it is most often regarded as a beneficial strengthening phase, since it precipitates as very fine particles. In martensitic 9–12% Cr steels alloyed with V and Nb, the Z-phase in the form Cr(V,Nb)N was first observed in 1986 in the 12% Cr bolting steel X19CrMoVNbN 11 1 [26]. In 1996, detailed investigations of creep specimens of older 12CrMoVNb steels showed that the Z-phase precipitated after 10,000–30,000 h of creep at 600 1C, and that MX (V,Nb)-carbonitrides dissolved. It was concluded that the Z-phase precipitation reaction was responsible for observed creep instabilities in the steels [27]. However, the older 12% CrVNb steels including the X19CrMoVNbN 11 1 have Nb contents, which are approx. 5 times higher than the new generation of high-strength 9–12% Cr steels, and it was not considered that Z-phase precipitation would occur in the new steels. Thus, the observed Z-phase precipitation could represent a serious problem to the long-term microstructure stability of the new generation of 9–12%Cr steels, in which a major part of the high creep strength is obtained by particle strengthening with MX (V,Nb)-carbonitrides. Work at TU Denmark has resulted in a first version of a thermodynamic model of the Z-phase [28]. Model predictions indicate that Z-phase is the thermodynamically stable
nitride at temperatures below 650 1C in all of the 9–12% Cr steels, including P91, E911, P92 and even 12CrMoV, which do not contain niobium. This means that it is the kinetics of Z-phase precipitation (nucleation and growth), which is the key to understanding the differences in creep stability of the steels. Estimates of nucleation rates of the Z-phase based on driving force calculations show that Cr is a very significant element to accelerate Z-phase precipitation. Steels with 11–12% Cr can show Z-phase precipitation with complete MX dissolution within a few thousand hours at 650 1C. The 9% Cr steels like P91 and P92 have only shown beginning Z-phase precipitation after 30,000–40,000 h at 650 1C, with large populations of MX particles still present in the microstructure. At 600 1C and lower temperatures, Z-phase precipitation in the 9% Cr steels seems to be even slower, and may in fact be insignificant to the long-term stability up to 300,000 h. 6. Summary New microstructure characterisation methods combined with thermodynamic and kinetic modelling have provided improved understanding of strengthening mechanisms and long-term microstructure stability of the 9–12% Cr steels. Precipitate hardening controls the long-term microstructure stability, and solid solution strengthening from Mo and W plays no significant role in the long-term microstructure stability of 9–12% Cr steels. Significant particle strengthening can be obtained by intermetallic Laves phases provided that a large number of particles nucleate during creep exposure. Phase stability to high temperature, like for W Laves phases, promotes the nucleation of Laves phase. M23C6 carbide stability against coarsening is improved by boron, but also significantly by loss of Mo and W from solid solution through Laves phase precipitation. MX carbonitrides rich in V and Nb are extremely stable against coarsening, but they may be dissolved by precipitation of the complex nitride Z-phase (Cr(V,Nb)N). This is mainly a problem in Nb-containing steels with Cr contents of 10% and above. References
Fig. 5. Z-phase particle in a 11CrWCoVNbNB steel.
[1] Straub S. Verformungsverhalten und Mikrostruktur warmfester martensitischer 12%-Chromsta¨hle, VDI Forschungsberichte, vol. 5. Du¨sseldorf: VDI Verlag; 1995. p. 405. [2] Polcik P. Modellierung des Verformungsverhalten der warmfesten 9–12% Chromsta¨hle in Temperaturbereich von 550–650 1C. Aachen: Shaker Verlag; 1999. p. 92. [3] Hald J. Creep strength and ductility of 9 to 12% chromium steels. In: Holdsworth S, editor. Procedings of advanced creep data for plant design & life extension. SVUM, Prague, September 2003. UK: ETD; 2003. p. 58. [4] Terada D, Yoshida F, Nakashima H, Abe H, Kadoya Y. ISIJ Int 2002;42:1546. [5] Zielinska-Lipec A, Czyrska-Filemonowicz A, Ennis PJ, Wachter O. In: Lecompte-Beckers J, et al., editors. Procedings of the conference
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
[6] [7] [8] [9]
[10]
[11] [12] [13]
[14] [15] [16] [17]
on materials for advanced power engineering 1998, Lie`ge, October 1998, Part I. Ju¨lich, 1998. p. 559–66. Ha¨ttestrand M, Andre´n H-O. Micron 2001;32:789. Korcakova L, Hald J, Somers MAJ. Mater Character 2002;47:211. Sundman B, Jansson B, Andersson J-O. Calphad 1985;9:153. Korcakova L. Microstructure evolution in high strength steel for power plant application: microscopy and modelling. PhD dissertation, TU, Denmark, 2002. Hald J. In: Strang A, McLean M, editors. Modelling of microstructural evolution in creep resistant materials. London: The Institute of Materials; 1999. p. 1. Gustafson A˚, A˚gren J. Acta Mater 1998;45:81. Hald J. Steel Res 1996;67:369. Andersson J-O, Ho¨glund L, Jo¨nsson B, A˚gren in J. In: Purdy CR, editor. Fundamentals and application of ternary diffusion. New York: Pergamon Press; 1990. p. 153. Bja¨rbo A, Ha¨ttestrand M. Metall Mater Trans A 2001;32A:19. Bja¨rbo A. Scand J Metall 2003;32(2):94. Gustafson A˚, Ha¨ttestrand M. Mater Sci Eng A 2002;333:279. Liftshitz IM, Slyozov VV. Chem Solids 1961;19:35.
[18] [19] [20] [21] [22] [23] [24]
[25]
[26] [27] [28]
37
Wagner C. Z Elektrochem 1961;65:581. Umantsev A, Olson GB. Scr Metall 1993;30:235. A˚gren J, Clavaguera-Mora MT, Golcheski J, Inden G, Kumar H, Sigli C. Calphad 2000;24(1):42. Hald J, Korcakova L. ISIJ Int 2003;43(3):420. Ha¨ttestrand M, Schwind M, Andre´n H-O. Mater Sci Eng A 1998;250:27. Hald J. VGB PowerTech 2004;12:74. Stocker Ch, Spiradek K, Zeiler G. In: Lecompte-Beckers J, et al., editors. Proceedings of the conference on materials for advanced power engineering, Lie`ge, October 2002, Part III. Ju¨lich, 2002. p. 1459. Brett S. In: Proceedings of the fourth international conference on advances in materials technology for fossil power plants, Hilton Head Island, SC, October 2004. Schnabel E, Schwaab P, Weber H. Stahl u Eisen 1987;107(II):691. Strang A, Vodarek V. Mater Sci Technol 1996;12:552. Danielsen H, HaldJ. In: Proceedings of the fourth international conference on advances in materials technology for fossil power plants, Hilton Head Island, SC, October 2004.
International Journal of Pressure Vessels and Piping 85 (2008) 30–37 www.elsevier.com/locate/ijpvp
Microstructure and long-term creep properties of 9–12% Cr steels$ J. Hald Elsam/Energy E2/IPL-MPT, Technical University of Denmark, Denmark
Abstract Advanced microstructure characterisation and microstructure modelling has demonstrated that long-term microstructure stability in 9–12% Cr steels under technical loading conditions is equivalent to precipitate stability. Mo and W can have a positive influence on longterm creep strength of 9–12% Cr steels by Laves phase precipitation hardening. Unexpected breakdown of long-term creep stability of a number of alloys is caused by precipitation of the complex Z-phase nitride, which may completely dissolve fine V and Nb containing MX nitrides. High Cr contents of 10% and above in the steels accelerate Z-phase precipitation. r 2007 Elsevier Ltd. All rights reserved. Keywords: 9–12% Cr steels; Long-term creep stability; Creep mechanisms; Precipitate characterisation; Precipitate stability modelling; Laves phase; MX nitrides; Z-phase
1. Introduction In the past 40 years, martensitic creep resistant 9–12% Cr steels have found increasing application in thick-section components like piping, large forgings and castings of steam power plants. Beginning with the 12CrMoV steel introduced in power plants in the mid-1960s, alloy development over the past 30 years has led to new steam pipe steels like the modified 9Cr steel P91, introduced in plants in 1988s, to the tungsten-modified 9Cr steels P92 introduced in plants in 2001 and E911 introduced in 2002. Similar steels have been developed and applied for large forgings and castings of steam turbines. The steadily improved creep rupture strength of new martensitic 9–12% Cr steels has been used to construct new advanced fossil-fired steam power plants with higher efficiency. Increase in steam parameters from subcritical 180 bar/530–540 1C to ultra-supercritical values of 300 bar and 600 1C has been realised, and this has led to efficiency increases from 30–35% to 42–47%, equivalent to approximately 30% reduction in specific CO2 emission. With the expected continuous increase in the demand for coal for electricity production, especially in the fast developing $ This article appeared in its original form in Creep & Fracture in High Temperature Components: Design & Life Assessment Issues, 2005. Lancaster, PA: DEStech Publications, Inc. E-mail address: [email protected]
0308-0161/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijpvp.2007.06.010
Asian countries, this technology progress has clearly the potential to become one of the most (if not the most) significant countermeasures against global CO2 emissions in the coming decades. Design strength values for the new W-modified 9Cr steels E911 and P92, published at this conference, are now based on comprehensive long-term creep rupture testing up to 100,000 h. However, the decisions to introduce the steels in plants had to be based on shorter test results leading to less certainty about the long-term stability of the steels. This fact, together with conflicting ideas about long-term strengthening mechanisms of especially tungsten and the quest for even better martensitic steels, has led to intensive research into the microstructure stability of the 9–12% Cr steels under creep exposure. New developments of better microstructure characterisation methods and improved microstructure models have led to significant progress in the understanding of microstructure stability of this class of alloys as presented below. 2. Microstructure and creep mechanisms of 9–12% Cr steels 2.1. Microstructure Chemical composition, heat treatment and creep rupture strength of a number of 9–12% Cr steels are given in Table 1. The microstructure of the 9–12% Cr steels is
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37 Table 1 Chemical composition and heat treatment of a number of 9–12% Cr steels Mass (%)
C Cr Mo W Ni V Nb N B Austenitisation (1C) Pre-tempering (1C) Tempering (1C) sB=105 h=600 C (MPa)
Steam pipe steels
Turbine rotor steels
12CrMoV P91 E911 P92
Steel E Steel F Steel B
0.2 11 0.9 – 0.5 0.3 – – – 1050 – 750 59
0.1 10 1 1 0.6 0.2 0.05 0.05 – 1070 570 690 E95
0.1 9 0.9 – 0.1 0.2 0.05 0.06 – 1050 – 750 94a
0.1 9 1 1 0.3 0.2 0.05 0.07 – 1060 – 770 98
0.1 9 0.5 1.8 0.05 0.2 0.06 0.06 0.001 1065 – 770 113
0.1 10 1.5 – 0.7 0.2 0.05 0.05 – 1120 570 690 E95
0.2 9 1.5 – 0.1 0.2 0.05 0.02 0.01 1120 590 700 E125
The quoted creep rupture strength values for the pipe steels P91, E911 and P92 are from the most recent evaluations by the ECCC. a The most recent ECCC evaluation of steel P91 is from 1995. A new evaluation being made at present.
Fig. 1. Tempered martensite—TEM micrograph.
tempered martensite formed during a final normalising and tempering heat treatment, Fig. 1. After normalising at about 1050 1C, air hardening will lead to a martensitic transformation in sections up to approximately 100 mm thickness due to the chromium content. Tempering in the range 680–780 1C leads to recovery of ductility by annihilation of dislocations and to the formation of ferrite subgrains. Tempering temperatures in the low end of the range are used for components like turbine rotors, where high tensile strength is required. The high end of the range is used for pressurised components like steam pipes, where high toughness is necessary. Carbide and nitride particles precipitate in the steels during tempering, on prior austenite grain boundaries, ferrite subgrain boundaries and on dislocations inside subgrains. When the steels are put into operation in power plants at temperatures below the final tempering temperature, further particles may precipitate, which are thermodyna-
31
mically unstable at the tempering temperature, like e.g. the intermetallic Laves phase. From Table 1, it is clear that minor alloy additions of Nb and N produced the first strength increase from 12CrMoV to P91, and the second increase from P91 to P92 was achieved by partly replacing Mo by W and adding B. The strongest turbine steel B is based on high content of boron. 2.2. Creep and strengthening mechanisms Historically, it has often been assumed that for isothermal creep rupture tests a linear relation should exist between the logarithm of time to rupture and the logarithm of stress if the tested material was metallurgically stable. Inspections of creep rupture data for all technical hightemperature steels show that none of these are metallurgically stable at relevant temperatures. It has been tried with some success to explain the deviations from linearity by a change in fracture mechanism from ductile fracture at shorter testing times to intergranular cavitation at longer testing times. Further explanations include changes in the controlling creep mechanism of the steel from dislocation creep to viscous creep. The technically relevant stresses and temperatures for creep testing and service exposure of the 9–12% Cr steels are in the range 300 MPa/500 1C–50 MPa/650 1C. Relevant loading conditions in this range will lead to rupture times between 100 and 300,000 h and minimum creep rates between 3 106 and 5 1012 s1. In the authors’ opinion, there are strong indications that the creep mechanism in the whole of the technically relevant stress/ temperature range is dislocation creep. Microstructural sources of creep deformation are then the migration of dislocations and subgrain boundaries [1,2]. Subgrain boundaries consist of dislocation networks, which, compared with the subgrain interiors, are hard regions in the microstructure of ferritic steels. A small subgrain size is equivalent to a high-volume fraction of hard regions and could thus be expected to provide high creep strength. This is confirmed by heat treatment experiments with 9Cr steels, where the tempered martensite subgrain microstructure is removed and the creep strength is drastically reduced. However, dislocation densities and subgrain sizes of different 9–12% Cr steels in the normalised and tempered condition do not show systematic variation with creep strength. A 12CrMoV steel may have finer subgrain size than steel P92, but lower creep strength. During creep, the subgrains grow and the dislocation density decreases with strain accumulation in a similar manner for all the 9–12% Cr steels [1]. It is then the ability of a steel to maintain a small subgrain size and a high dislocation density for a long time at stress and temperature, which is significant to high creep strength. This indicates that microstructural explanations for the improved creep strength of the new 9–12% Cr steels should be found in mechanisms that retard the migration of dislocations and subgrain boundaries, and thus delay the
ARTICLE IN PRESS 32
J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
accumulation of creep strain with time. Such mechanisms include (i) solid solution strengthening by formation of clouds of solute atoms around dislocations and (ii) interactions with precipitate particles. 2.2.1. Solid solution strengthening vs. precipitation hardening Solid solution strengthening has often been referred to in discussions of the effect of Mo and W on creep strength of 9–12% Cr steels. It has since long been clear that during creep exposure at temperatures around 600–650 1C, most of the Mo and W in the steels will precipitate as intermetallic Laves phase ((Fe,Cr)2(Mo,W)). The dominating opinion was that this would cause creep instability in the steels because the solid solution strengthening effect from Mo and W would be lost, and the precipitation strengthening effect from Laves phase was believed to be insignificant. This opinion seemed to be supported by the breakdown of long-term creep strength of some W alloyed 9–11Cr steels, and by the excellent long-term stability of the Mo alloyed 9CrB steel B (Mo causes less Laves phase precipitation than W). However, the excellent long-term stability of the W-alloyed 9Cr steel P92 seemed to contradict the opinion, and experimental evidence to demonstrate and quantify the solid solution strengthening mechanism is sparse. A rough estimation of the solid solution strengthening effect from Mo and W can be obtained by considerations based on recently obtained experimental in-situ observations of dislocation movement in ferrite [4]. The back stress s* acting on a dislocation moving with velocity v is given by s ¼ v=B,
(1)
where B is the dislocation mobility (m/Pa s). The mean dislocation velocity is a function of the creep rate _ and the mobile dislocation density r: _ ¼ rbv=M,
(2)
where b is the Burgers vector and M is a Taylor factor. Eqs. (1) and (2) give the back stress as a function of the creep rate: M _ . (3) s ¼ rbB Measurements of dislocation velocities in [4] give a value of B at 600 1C in ferrite of 5 1018 m/Pa s for a W content of 2 atm%. Loss of W from solid solution will increase B and lower the back stress. In [3], B is estimated to be 2.5 1019 m/Pa s. In [5], the dislocation density is measured to be higher than 1 1014 m2 in steel P92. For a long-term creep test at 410,000 h at 600 1C/o150 MPa the minimum creep rate in steel P92 is below 0.0001%/ h ¼ 2.8 1010 s1. With M ¼ 3, b ¼ 0.254 nm and B ¼ 2.5 1019 m/Pa s, Eq. (3) gives s o0:13 MPa ðlongterm test 410; 000 h=600 C=o150 MPaÞ.
This back stress should be considered as a maximum value with all tungsten in solid solution. The very low value appears because the solute tungsten atoms diffuse at a rate similar to the dislocation velocity. It can be concluded that solid solution strengthening from W and Mo has no significant effect on long-term microstructure stability of the 9–12% Cr steels. Precipitation hardening by pinning of dislocations and subgrain boundaries should be regarded as the most significant strengthening mechanism in 9–12% Cr steels, and microstructure stability of the 9–12% Cr steels under creep load is equivalent to precipitate stability. This is consistent with two findings: (a) the compositional changes, which have improved the creep strength of the 9–12% Cr steels, have also resulted in clear changes in the precipitate populations and (b) breakdowns in creep stability presented below can be explained by unexpected precipitate reactions. 3. Precipitate stability The interaction between precipitate particles and subgrain boundaries or dislocations under creep conditions can be complicated, and a detailed discussion of interaction mechanisms and back stress formulations will not be attempted here. However, equations to describe the contribution of precipitate particles to creep strength have similarities to the Orowan stress pffiffiffiffiffi fp sOrowan ¼ 3:32 G b , (4) dp where G is the shear modulus, b is the Burgers vector, fp is the precipitate volume fraction and dp is the mean precipitate diameter. This formula demonstrates that the basic information needed to describe the effect of precipitate particles on creep strength is their volume fraction and their mean size. Predictions of long-term stability of precipitate particles and their influence on creep stability of 9–12% Cr steels have to rely on measurements of volume fractions and mean particle sizes at certain points in time and on models describing the evolution of these parameters as functions of temperature, time and stress. 3.1. Precipitate characterisation The new 9–12% Cr steels contain several precipitate types, which form either during the final normalising and tempering heat treatment or during subsequent creep, see Table 2. For the validation of models of long-term precipitate stability it is necessary to know volume fractions and mean particle sizes of the individual precipitate types. With the newly developed energy-filtered transmission electron microscope (EFTEM), it is now possible to make accurate measurements of particle sizes down to a few nm, and on-line discrimination between the individual precipitate types [6]. Due to technical constraints there is a lower limit to the usable magnification in EFTEM microscopes, which means
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37 Table 2 Main precipitates in some 9–12% Cr steels Steel
Precipitate Formula
12CrMoV M23C6 P91 M23C6 MX
P92
(Cr,Fe,Mo)23C6 (Cr,Fe,Mo)23C6 (Nb,V)(N,C)
Remark Precipitate during tempering Precipitate during tempering Undissolved during austenitisation Precipitate during tempering Precipitate during creep (To650 1C)
MX Laves phase
(V,Nb)(N,C) (Fe,Cr)2Mo
M23C6 MX
(Cr,Fe,Mo,W)23C6 Precipitate during tempering (Nb,V)(N,C) Undissolved during austenitisation (V,Nb)(N,C) Precipitate during tempering (Fe,Cr)2(Mo,W) Precipitate during creep (To720 1C)
MX Laves phase
that statistically sound size determination for a population of particles with mean diameters larger than approximately 200 nm become unpractical with the EFTEM method. However, in recent years, the introduction of new highintensity electron sources in the scanning electron microscope (SEM) has improved the resolution power of this instrument. This means that mean particle sizes down to approximately 100-nm diameter can be measured with SEM. Particle discrimination in the SEM can be obtained from the atomic number contrast in backscattered electron images [7]. 3.2. Modelling of precipitate stability Modelling of long-term precipitate stability should include predictions of phase stability, nucleation rates, growth rates and coarsening rates for precipitate phases as functions of chemical composition, temperature and stress. 3.2.1. Phase stability Recent developments of databases and software, like the Thermocalc [8], to calculate thermodynamic equilibria in multicomponent systems allow predictions of phase stability for 9–12% Cr steels based only on chemical composition. Accuracy of the predictions depends critically on the available phase descriptions in the thermodynamic databases. A few critical gaps and inaccuracies in these descriptions have been identified, e.g. the predictions of Laves phase stability in 9–12% Cr steels alloyed with molybdenum but no tungsten has been proven wrong by experiments. In such cases, new thermodynamic phase descriptions are needed [9]. Otherwise, the thermodynamic calculations have demonstrated accurate predictions of stability range, amount and chemical composition of all phases in the 9–12% Cr steels [10]. These predictions are fundamental to the predictions of long-term stability of precipitate particles as demonstrated below.
33
3.2.2. Nucleation Modelling of nucleation rates is challenging because it depends critically on the interfacial energy, which is very difficult to predict or determine experimentally. It is therefore most often necessary to make simplifying assumptions about nucleation. However, with thermodynamic equilibrium calculations, the driving force for nucleation can be calculated, and based on the normal ranges for the interfacial energy gE0.1–1 J/m2 critical nucleus sizes as well as nucleation rates can be estimated. More ambitious modelling of nucleation has to include strain energy and nucleus shape [11]. 3.2.3. Growth The author has previously modelled Laves phase growth in steel P92 using an Avrami-type equation together with thermodynamic calculation [12]. However, recent developments of multicomponent diffusion databases like the DICTRA [13] allow more advanced modelling. The model needs experimental input to determine the nucleus density or the final size of fully grown particles [9,14,15]. 3.2.4. Coarsening Particle coarsening is one of the most important degradation processes for creep-resistant steels, and a detailed understanding of the influence of chemical composition on coarsening rate is essential to the understanding of long-term microstructure stability. Coarsening can be modelled by DICTRA in a similar way as for growth by incorporation of a contribution from the interfacial energy to the Gibbs energy of the particle. This leaves the interfacial energy as the only fit parameter in the coarsening model [16]. However, the DICTRA calculations can be time consuming, and a simpler approach allows a quicker overview of coarsening rates for different steels. Particle coarsening is observed to follow the well-known Ostwald ripening equation: r3 r30 ¼ K p t,
(6)
where r is the precipitate radius at time t and r0 is the precipitate radius at time 0. Liftshitz and Slyozov [17] and Wagner [18] calculated the coarsening rate Kp for binary alloy systems, but only recently have extensions to multicomponent systems been obtained by Umantsev and Olson [19] and by A˚gren et al. [20]. The latter resulted in the following formula for the coarsening rate constant Kp in a C-component system of b precipitate particles in an a matrix: 8 gV bm i, Kp ¼ P h 9 C ðxb xa=b Þ2 =ðxa=b Di =RTÞ i i i i¼1
(7)
where g is the interfacial energy and V bm is the molar volume of the precipitate phase. Di is the diffusion coefficient of element iin the matrix, xbi is the mole fraction a=b
of element iin the precipitate and xi
is the mole fraction
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
34
of element iat the precipitate/matrix interface. The diffusion coefficients are extracted from the DICTRA database, and include effects of chemical composition. Volume diffusion is assumed. The effect of interface curvature on the equilibrium solubilities is relatively small in this case, a=b and the values of xbi and xi obtained by thermodynamic equilibrium calculations assuming a flat precipitate/matrix interface can be used as a good approximation. V bm is calculated from lattice parameter data and precipitate composition. The only unknown is then the interfacial energy g, which is used as a fit parameter when compared to experimental observations.
rupture strength between the steels. The Laves phases in steel P92 produce a significant precipitate strengthening effect, which together with the observed high stability of the M23C6 carbides provide the high creep strength of this steel. It is interesting to note that the loss of tungsten from solid solution in steel P92 actually contributes to the increased coarsening stability of the M23C6 carbides, which is mainly controlled by the ferrite matrix content of the substitutional elements, which have a high concentration in the carbide. The observed differences in Laves phase behaviour in steels P91 and P92 is explained by the lower solubility of the Mo Laves phases compared with W Laves phases, see Table 2. When the creep temperature is close to the solution temperature for Laves phase, nucleation of the phase is very difficult, and only few particles nucleate. This results in the extended growth phase and large mean particle size. Similar observations have been made for the 9CrMoB steel B [24]. This is also seen in the very strong effect of temperature on Laves phase particle size in steel P92, Fig. 3. At 650 1C, the Laves phase particles are approx. 60% larger than at 600 1C. This could explain an observed tendency to reduced rupture ductility after more than 10,000 h at temperatures 625 and 650 1C in steel P92 since large Laves phase particles can act at nucleation sites for creep cavities. It is clear from the discussion above that alloying with either W or B has a strong positive effect on the microstructure stability of 9–12% Cr steels. W produces fine stable Laves phase particles in steel P92—even finer than the M23C6 carbides in steel P91—when the creep temperature is sufficiently below the solubility temperature for this phase. The Laves phase particles contribute significantly to particle strengthening and indirectly to M23C6 carbide stability through removal of Mo and W from solid solution. B has a strong stabilising effect on the M23C6 carbides, even though the mechanism for this remains unclear. The MX precipitates appear to be extremely stable against coarsening and could probably be characterised as the backbone of the long-term stability of the new 9–12% Cr steels.
4. Evaluation of precipitate stability In recent years, experimental measurements of particle sizes of individual precipitate types have been obtained with the new microstructure characterisation methods, and the models described above have been used to evaluate the precipitate stability of a number of 9–12% Cr steels. Fig. 2 shows evaluations by measurements and modelling of the evolution in mean particle sizes of precipitates in steels P91 and P92 at 600 1C [21–23]. In both steels, the MX carbonitrides show a similar very high stability against coarsening. The M23C6 carbides show considerable coarsening in steel P91, whereas in steel P92 this carbide is highly stable, this is attributed to the presence of B in steel P91, which seems to improve the carbide coarsening stability. This probably explains the high stability of the rotor material steel B, but the exact mechanism for boron stabilisation is at present unknown. Large differences are found between the Laves phase particle sizes in the two steels. After an initial growth phase of approx. 10,000 h, the Laves phase particles in steel P92 are more stable against coarsening than the M23C6 carbides in steel P91. In steel P91, the Laves phase particles grow to very large sizes during an extended growth phase, which lasts approx. 30,000 h. The observed differences in particle stability can explain the observed differences in creep
P92 - MX 600°C
500
Mean particle radius, nm
Mean particle radius, nm
P91 and P92 Laves phase and M23C6 size - 600°C
400 300 200 100 0 0
10000
20000 30000 Time, h
40000
50000
30 25 20 15 10 5 0 0
10000
20000
30000 40000 Time, h
50000
60000
70000
Fig. 2. Left: Particle sizes of M23C6 (thin lines) and Laves phase (thick lines) in steels P91 and P92. Right: Size evolution of MX particles in steel P92. A similar MX evolution is observed in P91.
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
300x10-9
Diameter [m]
250 200 150 100 50 0 0
20
40
60
80x10 3
30
40x10 3
Time [h] 500x10-9
Diameter [m]
400 300 200 100 0 0
10
20 Time [h]
35
58 MPa/approx. 610 1C and after 12,000 h at 45 MPa/ 660 1C. The rupture times were only approx. 10% of the expected minimum lifetimes according to published ASME allowable stresses, Fig. 4. Inspection of the tubes showed no signs of extensive corrosion or oxidation attack, and temperature estimates based on oxide thickness measurements did not indicate any overheating. Detailed microstructure investigations showed extensive precipitation of the complex nitride Z-phase (Cr(V,Nb)N) in the steels, and this had led to an almost complete dissolution of MX nitrides. The T122 tubes contained 12% Cr. Parallel investigations of a steel P122 pipe with 11% Cr after isothermal exposure 650 1C/10,000 h showed only a very little sign of Z-phase formation, and a large population of MX was still present in the steel. It was concluded that the Z-phase transformation was responsible for the breakdown in creep stability of the T122 tubes. Since the only significant difference between the T122 and P122 steels was the Cr content it was further concluded that a high Cr content accelerates Z-phase formation. Severe creep instabilities have been observed in a number of experimental 10–12% Cr steels, e.g. the 11CrWCo steels TAF 650 and NF12. Is was for long believed that this was caused by Laves phase precipitation, but measurements of Laves phase particles in a steel similar to TAF650 have demonstrated an even more favourable particle size distribution of the Laves phase than in steel P92 [9]. Similarly, a large number of test alloys with 11–12% Cr, and with or without W, intended for operation at 650 1C have been manufactured and tested. All of these, which have been tested to long times, show a severe breakdown in creep stability between 5000 and 30,000 h at 650 1C. Recent investigations have identified precipitation of the complex
Fig. 3. Model of growth and coarsening of Laves phase in steel P92. The diameter of Laves phase vs. time. (a) 600 1C and (b) 650 1C. Full squares denote measured values.
P/T122 1000
The presented experimental observations and modelling produce improved understanding of the excellent stability of the strongest of the new 9–12% Cr steels, i.e. the P92 and rotor steel B, but the models fail to explain observed microstructure instabilities in a number of steels. These instabilities can however be attributed to precipitate reactions. Imbalances in composition may lead to formation of unwanted phases like AlN in steel P91 at the expense of MX nitrides [25], and unexpected phase transformations like the Z-phase formation may occur with a similar result as described below. 5.1. Z-phase precipitation In 2001, two tubes of steel T122 (12CrWMoCuVNbN) installed in a test superheater facility at a Danish power plant ruptured prematurely after 30,000 h exposure at
Rupture stress, MPa
5. Microstructure instabilities
100
10 28000
30000
32000
34000
36000
38000
40000
PLM = T(C+logt), K and h
Fig. 4. Mean rupture strength of steel P/T122 and observed lifetimes of T122 superheater test tubes.
ARTICLE IN PRESS 36
J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
nitride Z-phase ((Cr(V,Nb)N)) as the root cause of this instability in the 10–12% Cr steels. The Z-phase can completely dissolve the MX carbonitrides, which are major contributors to creep strength. The Z-phase precipitates as very large particles, which will not contribute to precipitation hardening, and which may each consume more than 1000 MX particles, Fig. 5. The Z-phase nitride in the form CrNbN has been known from Nb-alloyed austenitic stainless steels since 1950, and here it is most often regarded as a beneficial strengthening phase, since it precipitates as very fine particles. In martensitic 9–12% Cr steels alloyed with V and Nb, the Z-phase in the form Cr(V,Nb)N was first observed in 1986 in the 12% Cr bolting steel X19CrMoVNbN 11 1 [26]. In 1996, detailed investigations of creep specimens of older 12CrMoVNb steels showed that the Z-phase precipitated after 10,000–30,000 h of creep at 600 1C, and that MX (V,Nb)-carbonitrides dissolved. It was concluded that the Z-phase precipitation reaction was responsible for observed creep instabilities in the steels [27]. However, the older 12% CrVNb steels including the X19CrMoVNbN 11 1 have Nb contents, which are approx. 5 times higher than the new generation of high-strength 9–12% Cr steels, and it was not considered that Z-phase precipitation would occur in the new steels. Thus, the observed Z-phase precipitation could represent a serious problem to the long-term microstructure stability of the new generation of 9–12%Cr steels, in which a major part of the high creep strength is obtained by particle strengthening with MX (V,Nb)-carbonitrides. Work at TU Denmark has resulted in a first version of a thermodynamic model of the Z-phase [28]. Model predictions indicate that Z-phase is the thermodynamically stable
nitride at temperatures below 650 1C in all of the 9–12% Cr steels, including P91, E911, P92 and even 12CrMoV, which do not contain niobium. This means that it is the kinetics of Z-phase precipitation (nucleation and growth), which is the key to understanding the differences in creep stability of the steels. Estimates of nucleation rates of the Z-phase based on driving force calculations show that Cr is a very significant element to accelerate Z-phase precipitation. Steels with 11–12% Cr can show Z-phase precipitation with complete MX dissolution within a few thousand hours at 650 1C. The 9% Cr steels like P91 and P92 have only shown beginning Z-phase precipitation after 30,000–40,000 h at 650 1C, with large populations of MX particles still present in the microstructure. At 600 1C and lower temperatures, Z-phase precipitation in the 9% Cr steels seems to be even slower, and may in fact be insignificant to the long-term stability up to 300,000 h. 6. Summary New microstructure characterisation methods combined with thermodynamic and kinetic modelling have provided improved understanding of strengthening mechanisms and long-term microstructure stability of the 9–12% Cr steels. Precipitate hardening controls the long-term microstructure stability, and solid solution strengthening from Mo and W plays no significant role in the long-term microstructure stability of 9–12% Cr steels. Significant particle strengthening can be obtained by intermetallic Laves phases provided that a large number of particles nucleate during creep exposure. Phase stability to high temperature, like for W Laves phases, promotes the nucleation of Laves phase. M23C6 carbide stability against coarsening is improved by boron, but also significantly by loss of Mo and W from solid solution through Laves phase precipitation. MX carbonitrides rich in V and Nb are extremely stable against coarsening, but they may be dissolved by precipitation of the complex nitride Z-phase (Cr(V,Nb)N). This is mainly a problem in Nb-containing steels with Cr contents of 10% and above. References
Fig. 5. Z-phase particle in a 11CrWCoVNbNB steel.
[1] Straub S. Verformungsverhalten und Mikrostruktur warmfester martensitischer 12%-Chromsta¨hle, VDI Forschungsberichte, vol. 5. Du¨sseldorf: VDI Verlag; 1995. p. 405. [2] Polcik P. Modellierung des Verformungsverhalten der warmfesten 9–12% Chromsta¨hle in Temperaturbereich von 550–650 1C. Aachen: Shaker Verlag; 1999. p. 92. [3] Hald J. Creep strength and ductility of 9 to 12% chromium steels. In: Holdsworth S, editor. Procedings of advanced creep data for plant design & life extension. SVUM, Prague, September 2003. UK: ETD; 2003. p. 58. [4] Terada D, Yoshida F, Nakashima H, Abe H, Kadoya Y. ISIJ Int 2002;42:1546. [5] Zielinska-Lipec A, Czyrska-Filemonowicz A, Ennis PJ, Wachter O. In: Lecompte-Beckers J, et al., editors. Procedings of the conference
ARTICLE IN PRESS J. Hald / International Journal of Pressure Vessels and Piping 85 (2008) 30–37
[6] [7] [8] [9]
[10]
[11] [12] [13]
[14] [15] [16] [17]
on materials for advanced power engineering 1998, Lie`ge, October 1998, Part I. Ju¨lich, 1998. p. 559–66. Ha¨ttestrand M, Andre´n H-O. Micron 2001;32:789. Korcakova L, Hald J, Somers MAJ. Mater Character 2002;47:211. Sundman B, Jansson B, Andersson J-O. Calphad 1985;9:153. Korcakova L. Microstructure evolution in high strength steel for power plant application: microscopy and modelling. PhD dissertation, TU, Denmark, 2002. Hald J. In: Strang A, McLean M, editors. Modelling of microstructural evolution in creep resistant materials. London: The Institute of Materials; 1999. p. 1. Gustafson A˚, A˚gren J. Acta Mater 1998;45:81. Hald J. Steel Res 1996;67:369. Andersson J-O, Ho¨glund L, Jo¨nsson B, A˚gren in J. In: Purdy CR, editor. Fundamentals and application of ternary diffusion. New York: Pergamon Press; 1990. p. 153. Bja¨rbo A, Ha¨ttestrand M. Metall Mater Trans A 2001;32A:19. Bja¨rbo A. Scand J Metall 2003;32(2):94. Gustafson A˚, Ha¨ttestrand M. Mater Sci Eng A 2002;333:279. Liftshitz IM, Slyozov VV. Chem Solids 1961;19:35.
[18] [19] [20] [21] [22] [23] [24]
[25]
[26] [27] [28]
37
Wagner C. Z Elektrochem 1961;65:581. Umantsev A, Olson GB. Scr Metall 1993;30:235. A˚gren J, Clavaguera-Mora MT, Golcheski J, Inden G, Kumar H, Sigli C. Calphad 2000;24(1):42. Hald J, Korcakova L. ISIJ Int 2003;43(3):420. Ha¨ttestrand M, Schwind M, Andre´n H-O. Mater Sci Eng A 1998;250:27. Hald J. VGB PowerTech 2004;12:74. Stocker Ch, Spiradek K, Zeiler G. In: Lecompte-Beckers J, et al., editors. Proceedings of the conference on materials for advanced power engineering, Lie`ge, October 2002, Part III. Ju¨lich, 2002. p. 1459. Brett S. In: Proceedings of the fourth international conference on advances in materials technology for fossil power plants, Hilton Head Island, SC, October 2004. Schnabel E, Schwaab P, Weber H. Stahl u Eisen 1987;107(II):691. Strang A, Vodarek V. Mater Sci Technol 1996;12:552. Danielsen H, HaldJ. In: Proceedings of the fourth international conference on advances in materials technology for fossil power plants, Hilton Head Island, SC, October 2004.