Precipitates in creep resistant austenitic stainless steels



T.Sourmail
Department of Materials Science and Metallurgy
Pembroke Street, CB2 3QZ Cambridge, U.K.

Austenitic stainless steels are, by far, the most widely used stainless steels comprising 70-80% of stainless production [1]. With excellent corrosion and mechanical properties at high temperatures, they are choice materials for powerplant tubes which have to operate at temperatures above 950K, or for aeroengines. The important role of precipitation in the achievement of good creep properties has been understood for long and extensively studied. Although some phases are now well documented, there are still contradictions and missing thermodynamic data, in particular, there is only a limited amount of informations about phases like Z-phase or Cr₃Ni₂SiN which can be predominant precipitates in nitrogen bearing steels. This paper is a review of common precipitates in austenitic stainless steels, in particular wrought heat-resistant steels of the AISI 300 families or 20/25 steels. Precipitates forming in age-hardening austenitic stainless steels are only briefly presented, having been previously reviewed by other authors.

General introduction

Composition and constituents

Austenitic stainless steels are essentially alloys of Fe-Cr-Ni, which owe their name to their room temperature austenitic structure. The addition of chromium has long been known to improve corrosion resistance. Cr is also a `ferrite stabiliser' and Fe-Cr stainless steels have a ferritic structure, possibly martensitic depending on the heat-treatment. The addition of austenite stabilising elements in sufficient quantities can allow an austenitic structure to be stable at all temperatures.

Nickel is the basic substitutional element used for austenite stabilisation. The equilibrium phases depend on the proportion of the three elements, as well illustrated in an isothermal section of the ternary diagram for Fe-Cr-Ni (fig. 1) calculated with MT-DATA [2].

Figure: Isothermal section of the Fe-Cr-Ni diagram at 750 $^\circ$C: a typical 18Cr-12Ni wt% lies in the austenitic field. Calculated using MT-DATA and the SGTE database.

Often, alloying elements, either interstitial such as C or N, or substitutional such as Mo, Mn, Ti, Nb, V, W, Cu, Al,... are used to obtain the required properties. They can be classified as ferrite-stabilisers or austenite-stabilisers and their effect in this respect is often approximated using the notion of chromium and nickel equivalents, calculated by formulae like [3]:

$$\mathrm{Ni_{eq}}=\mathrm{Ni}+\mathrm{Co}+0.5(\mathrm{Mn})+30(\mathrm{C}) +0.3(\mathrm{Cu})+25(\mathrm{N}) {\mathrm{wt\%}}$$

$$\mathrm{Cr_{eq}}=\mathrm{Cr}+2.0(\mathrm{Si})+1.5(\mathrm{Mo... ...athrm{Nb})+1.5(\mathrm{Ti})+0.75(\mathrm{W}) {\mathrm{wt\%}}$$

In this example, the composition has to be given in weight%. The use of such formulae is not always straightforward, as they refer to the austenite content, which is modified by various precipitation reactions involving these elements.

Whether the austenitic structure is retained at room temperature depends on the M_S (martensite-start) temperature. Several empirical formulae have been derived to describe the effect of chemical composition on M_S, an example is [1]:

$$M_{S}(^\circ\mathrm{C})=502-810(\mathrm{C})-1230(\mathrm{N})-... ...\mathrm{Cr})-54(\mathrm{Cu})-6(\mathrm{Mo})\ {\mathrm{wt\%}}$$

A typical type 304 (austenitic stainless steel) has its M_S well below room temperature. However, one should keep in mind that deformation-induced martensite formation can occur well above M_S. M_d30 is the temperature at which 50% of martensite has formed for a true strain of 30%. Again, formulae like:

$$M_{d30}(^\circ\mathrm{C})=497-462(\mathrm{C+N})-9.2 (\mathrm{... ...thrm{Ni})-13.7(\mathrm{Cr})-18.5 (\mathrm{Mo}) {\mathrm{wt\%}}$$

have been derived to describe the effect of alloying elements [1].

At high temperatures, a steel containing 18Cr, 12Ni wt% should be fully austenitic. However, the addition of alloying elements often results in the formation of carbides, nitrides and intermetallics. These phases are not always desirable and a good knowledge of precipitation reactions is required to avoid loss of mechanical or chemical properties. A good example is the sensitisation of non stabilised austenitic stainless steels: sensitisation occurs when the precipitate M₂₃C₆ forms at grain boundaries, depleting the chromium content in the vicinity, which eventually results in intergranular corrosion. This can be avoided by tying up the carbon with strong carbide formers like Ti. The steel is then called stabilised. These second phases will be presented in detail in a next section.

Grades of austenitic stainless steels

From a simple type 304 to the recent NF709, austenitic stainless steel composition covers a large range. The two main alloying elements are chromium and nickel, so the steels will often be referred to by their content of Cr and Ni. For example, 18/10 refers to an austenitic stainless steels with 18Cr, 10Ni wt% .

Table: Composition of two austenitic stainless steels

Wt%	C	Mn	P	Ni	Cr
Type 304 [4]	$\leq$ 0.07	$\leq$2.00	0.04	8-10	17-19
NF-709 [5]	0.06	1.00	0.006	25	20
Wt%	Mo	Nb	Ti	B	N	Si
Type 304	$\sim$ 0	$\sim$ 0	$\sim$ 0	$\sim$ 0	$\sim$ 0	$\sim$ 0
NF-709	1.5	0.26	0.05	0.005	0.167	0.40

The AISI 300 series and other variants

The AISI 300 specifications for the compositions of different austenitic stainless steels (wt%) is shown in table 2 (after [4]).

Table: The AISI 300 series and other examples of heat resistant austenitic stainless steels; E1250 is Esshete 1250.

	C max.	Si max.	Mn max.	Cr	Ni	Mo	Ti	Nb	Al	V
301	0.15	1.00	2.00	16-18	6-8
302	0.15	1.00	2.00	17-19	8-10
304	0.08	1.00	2.00	18-20	8-12
310	0.25	1.50	2.00	24-26	19-22
316	0.08	1.00	2.00	16-18	10-14	2.0-3.0
321	0.08	1.00	2.00	17-19	9-12		5 x %C min.
347	0.08	1.00	2.00	17-19	9-13			10 x %C min.
E 1250	0.1	0.5	6.0	15.0	10.0					0.25
20/25-Nb	0.05	1.0	1.0	20.0	25.0			0.7
A 286	0.05	1.0	1.0	15.0	26.0	1.2	$\sim$1.9		$\sim$0.18	$\sim$0.25

Grades denoted L contain low carbon (0.03 wt%) and N contain nitrogen (eg: 316LN). Most often used as creep-resistant steels are types 316, 321 and 347, or alloys containing all of Mo, Nb and Ti. There are many other variants of these compositions, like the japanese SUS300 series which mirrors the AISI 300 series, but with sometimes addition of both Ti and Nb. For convenience, as is sometimes done in the literature, the AISI 300 series will be used even for steels not strictly belonging to it, like a 316 with Ti addition.

In fact, it is not the intention to describe, in this review, the precipitation sequences in all different grades of creep resistant austenitic stainless steels, but rather to examine the occurrence of the various precipitates in such a way that precipitation behaviour of non documented grades could be inferred from the conclusions reached.

There is a large amount of material on the precipitation phenomena in the 300 series of alloys, which have been used widely as creep resistant steels. The same is true for 20Cr-25Ni steels. However, it appears that the long-term behaviour of Ti, Al alloyed austenitic stainless steels (type A286) is very little documented [6]. This is possibly because production difficulties have restricted the application of such steels to parts requiring relatively small ingot sizes (aeroengine turbine discs), the design life of which is much shorter than the few 100000 h required for a steam plant [7].

Role of alloying elements

• Mn has been introduced in austenitic stainless steels as a substitute for Ni during shortages in the international market or for economical reasons. The nickel content can be halved to 4 wt% by the addition of 2-6Mn wt%. However, although the austenitic structure is achieved, such steels do not exhibit the same corrosion resistance as an 18/8 steel. Mn is also used to increase the solubility of nitrogen in austenite.
• Mo is on the contrary a ferrite stabiliser. It improves the creep properties of stainless steels by solid solution hardening. It also facilitates carbides precipitation. However, it promotes $\sigma$ and Laves phase formation on long term ageing.
• Stabilising elements like niobium, titanium and vanadium greatly improve the creep strength of austenitic stainless steels mainly by precipitating fine carbides intragranularly. On the other hand, they reduce the creep ductility. The ratio in which they are added to carbon is important to maximise the strengthening effect and avoid precipitation of detrimental phases. They can also have a solid solution strengthening effect.
• Carbon acts principally by solid solution strengthening in non-stabilised grades, but mainly by precipitation strengthening when Nb, Ti or V are present.
• Nitrogen is a strong austenite stabiliser. It has also a role in increasing the creep life of austenitic stainless steels: it can act like carbon in stabilised stainless steel by precipitating in the form of titanium or niobium nitrides, but the nitrogen remaining in solid solution has also a much greater strengthening effect than carbon. It has been believed to also lower the diffusivity of chromium and carbon in the matrix, therefore delaying coalescence of precipitates [1],[4]. However, more recent results indicate that nitrogen enhances chromium diffusion, but retards the nucleation of M₂₃C₆ because of its low solubility in this carbide [8].

Carbides and nitrides

MX precipitates

The formation of MX precipitates in austenitic stainless steels occurs when strong carbides/nitrides formers (Ti, Nb, V, Zr, Ta...) are added to the alloy.

These additions have two purposes:

• stabilising the alloy against intergranular corrosion. When stabilisation is the first aim, a stabilising heat-treatment is conducted at 840-900 $^\circ$C for several hours before service. Most of the carbon is then `tied-up' by precipitation of MC.
• providing good creep resistance: when this is the first aim, a solution heat-treatment is given in order to dissolve as much MX as possible. Subsequent precipitation occurs during creep and increases strength. A typical solution treatment is 30-60 min at temperatures between 1100 and 1250 $^\circ$C.

The second use is the one we will focus on. Many studies have concentrated on MX precipitation, but also on the appropriate content necessary to obtain the best properties. The solubility of the MX precipitates and the stoichiometry are essential in determining the behaviour of the steel with regard to precipitation.

Structure and features of precipitation

MX carbonitrides have a NaCl fcc (face-centered cubic) structure, with lattice parameters as listed in table 3.

Table: Lattice parameter of some MX precipitates.

MX	a, Å	ref
NbN	4.39	JCPDS 38-1155
NbC	4.47	JCPDS 38-1364
TiN	4.24	JCPDS 38-1420
TiC	4.33	JCPDS 32-1383

Often, measured lattice parameters have intermediate values, reflecting the existence of a solid solution between the different carbonitrides.

MX precipitates usually form on dislocations within the matrix, on stacking-faults (most often with TiC), on twin and grain boundaries. They have a characteristic cuboidal shape after sufficient ageing.

Solubility of MX carbides in austenite

Stabilisers such as Ti or Nb have long been known to reduce the solubility of carbon in austenite. Early studies of the problem consisted in the determination of two parameters: H, the heat of solution, and A, a constant as in:

$$\log[{\mathrm M}][{\mathrm X}]={\mathrm A}-\frac{\mathrm H}{\mathrm T}$$

where [M] is the weight percent of stabilising element in solution in the matrix, and similarly for [C]. The product [M][X] is often referred to as the solubility product. Strictly speaking, H is expressed in K^-1.

The two following relationships can be found in many publications (e.g. [3],[4]):

$$\displaylines{ \log[\mathrm{Ti}][\mathrm{C}]={2.97}-\frac{678... ...\log[\mathrm{Nb}][\mathrm{C}]={4.55}-\frac{9350}{\mathrm T(K)}}$$

They are essentially valid for a typical 18/12 steel, the concentrations are in weight percent. For a 20/25 steel, Kikuchi et al. [9] used for TiC H=10475, A=3.42 and for NbC, H=7900 and A=4.92, but it is not clear whether these values have been measured for 20/25. The solubility is, as shown in their work, an important factor in the achievement of good creep properties: they showed that adding M and X in excess of their solubility limits resulted in coarse MX in the matrix and induced faster coarsening of MX later precipitated. However, until this limit is reached, the more M and X added the better because more MX particles will be formed.

It is therefore clear that knowing the solubility limits of MX carbides is important. However, modern high-temperature austenitic stainless steels often contain both carbon and nitrogen, and more than one strong carbide former (Ti+Nb, Nb+V...), and relationships as above are of limited use when it comes to estimate the solubility of multicomponent carbonitrides (e.g. (Ti,Nb)(C,N)).

Recently, some studies proposed different approaches to the problem of the solubility of multicomponent carbonitrides in austenite (P.Rios [10], [11], Zou et al. [12]...). For example, Rios [10] proposed for Nb(C,N):

$$\displaylines{ \left(\%{\mathrm {Nb}}\right) {\left[ \left( \... ...}}}={\Delta G}^{\circ}_{{\mathrm{NbN}}_{0.87}}/{\mathrm {RT}} }$$

However, one must notice that all these studies deal with MX in austenite for micro-alloyed steels. It seems clear, from the literature, that the presence of Ni and Cr introduces a further difficulty in austenitic stainless steels, the main one being the formation of Z-Phase.

There is a general agreement that Z-Phase (CrNbN) forms in Nb stabilised austenitic stainless steels, with a sufficient amount of nitrogen. In fact, it seems that as soon as 0.06N wt% is present in a typical 347 steel, Z-Phase can be expected (Hughes, [13]).

Moreover, in Cr/Ni steels, MX precipitates have been reported to start growing largely under-stoichiometric (Andrén et al. [14], [15]). No approaches have been found that dealt with the solubility of multicomponent carbonitrides in austenitic stainless steels susceptible of Z-Phase formation. Indeed, Z-Phase is even absent from the SGTE (Scientific Group Thermodata Europe) thermodynamic databases accessed by programs such as MT-DATA or Thermocalc.

Formation and stoichiometry of MX precipitates

A number of studies have been devoted to the determination of the correct quantities of M and X to use in order to have the best mechanical properties. As stated above, one factor is the solubility of MX. This has been clearly outlined by the work of Kikuchi et al. [9].

The second factor is the proportion of each element that has to be added, in such a way as to maximise the amount of precipitation for a given product [M][X]. It is also important to know, if stoichiometry is not respected, which of the elements is in excess, as this may influence the precipitation sequence.

Wadsworth et al. [16] proposed a quantitative approach to stoichiometry and showed that the amount of precipitate that can form drops quite sharply when M and X are not added in proportions corresponding to the composition of the expected carbide (figure 2). Using the data of Keown and Pickering [17], they showed that best creep lives were obtained when the Nb/C ratio was matching the stoichiometry Nb₄ C₃ (for 18/12 or 18/10).

Figure: The amount of niobium carbide available for precipitation at 923K (after solution treatment at 1373K) as function of the degree of deviation from stoichiometry r, defined by r=pB_T-qA_T, where p and q are the mass fractions of A and B in the compound AB_n, and A_T and B_T the mass fraction of these elements in the alloy. After [16]

Other studies have shown that best creep life was obtained for a Nb/C atomic ratio of 1/1 (Adamson and Martin [18], Kikuchi et al. [9], all for 20/25). The work done by Kikuchi et al. does not highlight any serious contradiction since the first composition they tried around the atomic ratio 1:1 was 2:1, which means that any variation within this interval was not investigated. However, Adamson and Martin have results for the weight ratios 8:1 and 10:1 (respectively atomic ratios 1:1 and 4:3), but find better creep properties for the 1:1 atomic ratio. A possible explanation is that NbC actually forms with the stoichiometry 1:1 in 20/25 steels. It seems important to note that all these studies rely on relatively short term creep tests (as an average, less than 1000 h). No study has been found which confirms the importance of stoichiometric addition in long term creep properties. Moreover, this explanation of the good creep properties of steels with stoichiometric additions of M and X assumes that MX precipitates form with their final stoichiometry, which has been proved wrong by different authors.

The non-stoichiometry of MX precipitates in 18/12 austenitic stainless steels has been confirmed by microprobe analysis (Andrén et al., [14], [15]). The authors noticed that MX was largely sub-stoichiometric for short ageing (typically 3 h at 750 $^\circ$C), and contained a large amount of Cr substituting for M (Ti,Nb,V...). They proposed the following explanations: as MX has a lattice parameter larger than that of the austenite, a flow of vacancies is needed for the growth of MX and stress is generated. The substitution of M by Cr and the low carbon content both reduce the lattice parameter. Moreover, Cr is readily available while M has to diffuse over long distances. The formation of sub-stoichiometric, carbon depleted MX precipitates is therefore kinetically advantageous. In more theoretical terms, this corresponds to two effects. Capillarity is likely to modify substantially the local equilibrium as the precipitates are very small. In fact, the composition changes reported by these authors correlate with size changes. The second effect is the modification of local equilibrium to satisfy simultaneously the flux balance for different solutes. This effect could affect strongly the composition of MX, since the diffusivities of the elements involved are very different. It would similarly correlate with a size change unless the precipitates are only coarsening.

However, stoichiometric carbides still have a lower free energy and, during further ageing, they grow at the expense of sub-stoichiometric ones. This is because, according to Andrén et al., the diffusivity of metallic elements within MX precipitates is so small that these precipitates can not change composition [14].

Z-Phase

Z-Phase is a complex carbonitride which forms in Nb stabilised austenitic stainless steels containing a relatively high level of nitrogen. It has only recently been discovered, with its structure being fully determined by Jack and Jack in 1972 [19], and it appears from the most recent reviews (e.g. [3]) that the conditions of its formation are not very clear, and even less is its relative stability when compared to other carbonitrides.

Structure and composition

The structure of Z-Phase is commonly accepted to be the one described by Jack and Jack [19]: it has a tetragonal unit cell (space group P4/nmm), with a=3.037 Å and c=7.391 Å. The unit cell is obtained by ordering of Cr, Nb and N atoms, the formula being Cr₂Nb₂ N₂ per unit cell. Vodárek [20] reports the following orientation relationship for Z-phase in a type 316LN:

$$\displaylines{ {\left( 001 \right)}_{Z} \parallel {\left( 001... ...ine{1}0 \right]}_{ Z} \parallel {\left[ 100 \right]}_{\gamma} }$$

In contrast to MX precipitates, it seems that Z-Phase dissolves little (if at all) substitutional elements. Raghavan et al. [21] suggest that it can dissolve carbon and write CrNb(CN), without giving any evidence for this. The absence of a CrNbC phase suggests that if carbon can dissolve in Z-Phase, it can only do so within limits. Also, when Nb is in sufficient quantity, both Z phase and NbC are found.

Occurrence

Z-Phase is seldom reported, even in alloys liable to form it on ageing. This could be due to its composition and its general features of formation which are not very different from common MX precipitates.

It generally forms on grain-boundaries, very rapidly, but also on twin boundaries and within the matrix, where it is associated with dislocations [22]. When it forms, it is usually as a fine dispersion of particles (see [13],[21], [22]), which makes it an interesting phase when good creep properties are sought. The morphology has been reported to be either cuboidal [22] or rod-like [14].

There is a good agreement that it forms at high temperature: Raghavan et al. [21] report its formation during annealing (1 h at 1300 K) of a 18/12 containing 0.3Nb wt% and 0.09N wt%, with an orientation relationship indicating that these are not residual particles. After ageing 8000 h at 866 K, it is still the predominant precipitate. Few M₂₃C₆ and $\sigma$-phase particles are present. These authors indicate that a 10 s heat-treatment at 1573 K is sufficient to dissolve all the Z-phase particles, which is not inconsistent with the work of Robinson et al. (quoted in [21]) which locates the solvus of Z-phase to be between 1573K and 1623K in a steel containing larger amounts of niobium and nitrogen.

Robinson and Jack [22] report the formation of Z-Phase in a 20Cr/9Ni steel containing 0.38N wt% and 0.27Nb wt% between 700 and 1000 $^\circ$C. At 1000 $^\circ$C Z-Phase is the first and only phase formed. It is found uniformly distributed after 30 min but coarsens rapidly. At 700 $^\circ$C it starts to precipitate in the matrix after 16 h. At lower temperatures, Vodárek et al. [23] report a considerable dimensional stability of Z-phase in a type 316LN, with a mean size of 6nm after 82 h at 650 $^\circ$C and 12nm after 37890 h at the same temperature.

Thorvaldsson and Dunlop [24], studying the effects of stabilising elements in austenitic stainless steels, use a 18/12 steel with 0.4V, 0.13Nb and 0.43N at%. After 5000 h at 750 $^\circ$C, no Z-Phase has formed, but a (Nb,V)(C,N) fcc carbonitride forms instead. Andrén et al. [14], with the same composition, in very similar conditions, found, on the contrary, that the MX precipitate had always a stoichiometry not far from M X_0.5, and contained sensibly as much chromium as Nb+V. However, the characteristic tetragonality was not observed and the authors suggested that they found a precursor state of the fully ordered Z-Phase.

Considerations on the stability of Z-Phase

It seems quite reasonable to propose that Z-Phase is the stable Nb containing phase in austenitic stainless steels with high nitrogen: Hughes [13] suggests that it can be expected in a typical 347 steel as soon as the weight percentage of N is greater than 0.06 wt%. Knowles [25] find Z-phase in a 20/25 0.41Nb wt% steel with 0.028N wt%, which could indicate that chromium and/or nickel content play an important role in the solubility of this phase.

The problem arises, in many steels, to know whether NbC or Z-Phase is the more stable. Very few studies have dealt with additions of Nb, C and N together in a way that could determine which phase is more stable: often Nb is in excess and both NbC and CrNbN form. Uno et al. [26] found that only Z-phase formed in a 18/12 Nb steel with carbon and nitrogen, the niobium content being lower than that required to combine either all C or all N. This would indicate that Z-phase is fast enough to form first when competition between NbC and CrNbN is likely to occur. On the other hand, Knowles [25] reported Nb(C,N) after 2 h at 850 $^\circ$C in a 20/25 steel, transforming to Z-phase with further ageing. From the precipitation behaviour of NF709 [5], where Z is reported after ageing times of 10⁴ h at 750 $^\circ$C, and C containing phases are M₂₃C₆ and M₆C one can infer that Z-Phase is more stable than NbC and that the latter will dissolve for the former if Nb is in too small quantities. However, Raghavan et al. [21] deduce from a comparison between 347 (18/12 with 0.8Nb and 0.07C wt%) and 347AP (18/12 with 0.3Nb, 0.09N, 0.009C wt%) a greater stability of NbC, since Z is found to precipitate from the solid state but the NbC is found as residual particles. The conclusions of such a comparison have to be examined carefully as the compositions were different.

Conclusions on Z-phase

Generally, one can conclude that Z-phase precipitation is not fully understood. Points of agreement are its stability at high temperature and the features of its precipitation (fine dispersion), proposed solvus temperatures are different between the studies, but these differences are consistent with different compositions. It is difficult as yet to build a coherent view of the different kinetics observed. It is not clear whether Z-phase actually forms from MX precipitates as suggested by the results of Knowles [25] and Andrén et al. [14], or directly as suggested by Robinson et al. [22]. These observations could be reconciled if one could prove that the driving force in a 20Cr 10Ni (Robinson et al.) is significantly different than that of a 18/12. The importance of Z-phase in both ferritic and austenitic power plant steels, renders particularly necessary the assessment of the ternary system Cr,Nb,N to provide the thermodynamic parameters required for any prediction.

M₂₃C₆

Structure and composition

M₂₃C₆ is a more general notation for Cr₂₃C₆, as often, Ni, Mo and Fe are found to substitute partially for chromium. It is the main carbide in non-stabilised C containing austenitic stainless steels; nitrogen bearing steels form Cr₂N. It has a fcc structure (and space group Fm3m) of lattice parameter varying between 10.57 and 10.68 Å, which is about three times that of the austenite. It is often reported to grow with a cube to cube orientation relationship:

$$\displaylines{ {\lbrace 100 \rbrace}_{\gamma } \parallel {\lb... ...rallel {\langle 010 \rangle }_{\mathrm{M}_{23}\mathrm{C}_{6}} }$$

It is most of the time the main carbide found in austenitic stainless steels. Although it can be only metastable, it is always found in the early stage of precipitation because it nucleates very easily.

Morphologies and locations

M₂₃C₆ is most of the time reported to precipitate on the following nucleation sites, and in this order: grain boundaries, incoherent and coherent grain boundaries, intragranular sites. In Nb stabilised grades, it is sometimes reported to nucleate on undissolved (that is, during the solution treatment) niobium carbonitrides. Generally, M₂₃C₆ only shows $\lbrace 111 \rbrace$ and $\lbrace 110 \rbrace$ interface planes. This has been explained by Beckitt and Clarck (1967, [27]), who showed these planes have the best atomic correspondence with the austenite.

Precipitation on grain boundaries

The precipitation of M₂₃C₆ has been the focus of many investigations, motivated by its importance in terms of corrosion resistance. M₂₃C₆ on grain boundaries is often associated with intergranular corrosion. It is found after very short ageing times (30 mn at 750 $^\circ$C,Lewis et al.,[28]), even in stabilised steels. It causes the grain boundary to move, and therefore shows on one side, the new grain boundary, while the other has a serrated aspect composed by the $\lbrace 111 \rbrace$ interfaces. These precipitates are usually large. When boron is added, the number density of M₂₃C₆ along the grain boundaries increases (Tanaka et al.,1997, [29]), with beneficial effect for the creep rupture strength, as the grain boundary sliding and surface cracking are reduced. Other effects of boron in M₂₃C₆ will be dealt with later.

Figure: M₂₃C₆ on grain boundary.

Precipitation on incoherent and coherent twin boundaries

On incoherent and coherent twin boundaries, M₂₃C₆ forms long plates parallel to the twin boundaries. The plates are, as for any form of M₂₃C₆ in austenitic stainless steels, bounded by $\lbrace 110 \rbrace$ and $\lbrace 111 \rbrace$ planes only. The large faces are $\lbrace 111 \rbrace$ planes parallel to the twin boundaries. Plates are initially growing from the incoherent twin boundary, but some are found later apparently detached from the boundary. Precipitation on the coherent boundary occurs more slowly than on incoherent ones.

Figure: Plates of M₂₃C₆ growing from an incoherent twin boundary (left), and around a Nb(C,N) precipitate (right), work done for this project.

The mechanism of their formation has been subject to controversy: Beckitt and Clarck (1967, [27]) suggested that carbides growing on the incoherent twin boundary repels Shockley partial dislocations in the twin, and these cause nucleation of new plates in the twin. But detailed observations have lead Singhal and Martin (1967, [30]) to conclude that stacking fault were first growing in the twin. M₂₃C₆ would then nucleate on the Shockley bounding dislocation and grow in the stacking fault, possibly helped by an enrichment of this stacking fault in chromium. Both fail to explain why these plates most of the time stand separated from the boundary and from each other (according to these mechanisms, they should be connected to the incoherent twin boundary by a stacking fault). Another mechanism has been proposed more recently by Sasmal (1984, [31]). Noticing that plates do not form away from the boundary or on coherent twin boundaries if the samples are directly put at ageing temperature after the solution treatment, Sasmal suggested that the formation of plates of M₂₃C₆ close but detached from the incoherent twin boundaries, and on coherent ones, was an effect of residual stress developed in these regions. One could here argue that stress only influences the formation: if it favors the formation of M₂₃C₆ plates parallel to the twin boundaries, there should still be a small percentage of plates growing in other equivalent directions.

Intragranular precipitation

In the austenite grains, M₂₃C₆ precipitates as cuboids or plates. The latter have the same characteristics as the plates growing from twin boundaries. The cuboids grow on dislocations until they form a continuous chain sharing their corner. These chains can form branches, forming secondary stringers of M₂₃C₆. Beckitt and Clarck (1967, [27]) have proposed satisfactory explanation for the formation of these secondary stringers: the growth of primary M₂₃C₆ generates dislocations on which other precipitation can occur. The direction of these secondary stringers is that expected by the mechanism, that is the direction of usual edge dislocations in austenite.

Another kind of intragranular precipitation has been reported by Sasmal (1997, [32] ). Plates of M₂₃C₆ can form around undissolved Nb carbonitrides in Nb stabilised steels. In this case, the large faces are $\lbrace 110 \rbrace$ and the edges are $\lbrace 111 \rbrace$. The reasons for this change are not clear, although a contribution of the strain around the undissolved particle is invoked.

Role of other alloying elements

M₂₃C₆ is, initially, a chromium-iron carbide. However, its composition is susceptible to large variations. For example, in a typical 2.25Cr1Mo wt% steel, M₂₃C₆ has a composition of about 0.3Fe, 0.38Cr, 0.1Mo, 0.22C mole fraction. While in a typical 304, it is closer to 0.04Fe, 0.65Cr, 0.11Mo, 0.22C. There are also evidences that, for a given steel composition, M₂₃C₆ composition undergoes large changes in the early stages of the precipitates. Lewis and Hattersley [28] quote a study by Philibert et al. in which the authors propose that M₇C₃ forms in the early stages of ageing in a 18/8 steel, as the Fe/Cr ratio is more consistent with the known composition of this phase. Similarly, Marshall [1] quotes a study in which the authors show that up to 40 wt% Fe can be found in M₂₃C₆ in the early stages of precipitation, but also that the Mo content increases with time. It is proposed here that these composition changes can be explained by the growth mechanism involved. In a multicomponent alloy, the interface has to move at a rate which is consistent with the flux of each element, in order to maintain local equilibrium. Therefore, the interfacial compositions are not in general given by a tie-line passing through the bulk composition, but by an alternative tie-line which permits the flux-balance (details are out of the scope of this review, see for example [33] or [34]).

In the same way, on the interstitial lattice, B can substitute for C; however both in very small quantities. Boron is of particular interest as it promotes the formation of intragranular M₂₃(C,B)₆. There is some controversy about the mechanism involved, but it is possible that B increases the lattice parameter of M₂₃C₆, therefore reducing the mismatch with the austenite. The presence of B in M₂₃C₆ has been shown using Atom Probe Field Ion Microscopy (APFIM, e.g. [35]). The same is not true for nitrogen. It has long been believed to have a small solubility in M₂₃C₆ (e.g. [36]). However, recent experimental results supported by ab initio calculations indicate zero solubility of nitrogen in M₂₃C₆, the structure of which is destabilised if a small amount of carbon is replaced by nitrogen [8].

Except in the fine intragranular form, M₂₃C₆ precipitation is not desirable for good creep properties. It is often associated with intergranular corrosion, as its formation along the grain boundaries causes a local depletion in chromium and possibly local loss of the stainless property (the steel is then said to be sensitised, i.e. susceptible of intergranular corrosion).

Solubility and kinetics of precipitation

The solubility of carbon in austenite, with regard to M₂₃C₆, has been described by empirical relationships such as [37]:

$$\log {\left[ \mathrm{C} \right]}_{\mathrm{ppm}}= 7.71-\frac{6272}{\mathrm{T(K)}}$$

Figure: Solubility of C in austenite with regard to M₂₃C₆ precipitation. Calculated with MT-DATA (allowing for austenite, M₂₃C₆, M₇C₃ and liquid, with elements Fe, Cr, Ni, 0.3Si, 1.0Mn wt%). Solid lines for 18/12 and dashed lines for 20/25. The squares correspond to the empirical relationship from [37]

As can be seen in figure 5, this is valid for a typical 18/12 steel, but one should be careful as it is sensibly modified by the chromium and nickel content. Increasing the chromium content reduces the solubility of carbon in austenite. It also increases the kinetics of M₂₃C₆ precipitation, as does an increase in carbon content: from 0.02 to 0.08C wt%, the nose of the C-curve is shifted from 100 h to 0.1 h.

M₂₃C₆ first precipitates on grain boundaries, then, with increasing time, on incoherent twin boundaries, coherent twin boundaries and finally in the matrix on dislocations. In the matrix, it forms as evenly spaced angular blocks. At long ageing times, grain boundary carbides can form a completely interlocked structure.

The kinetics of precipitation is affected by Mo, which stabilises the carbide and accelerates its formation. On the other hand, nitrogen is well-known to retard both formation and coarsening rates of M₂₃C₆, and an often proposed explanation is that nitrogen reduces the diffusivity of Cr and C in the austenite [4], [9]. However, Degalaix and Foct [38] found that if the carbon content was higher than 0.08 wt%, increasing the nitrogen content could have the opposite effect. More recent investigations indicate that N actually enhances the diffusion of substitutional elements by increasing the formation of vacancies, but delays the nucleation of M₂₃C₆, the structure of which is destabilised when carbon is partially substituted by nitrogen [8].

Relative stability of M₂₃C₆

In a simple type 304 steel , M₂₃C₆ is often the only carbide found at any ageing time. Although sometimes M₆C is found (Biss and Sikka, [39]), it is as a minor phase and no mention is made of an instability of M₂₃C₆ with regard to M₆C. When molybdenum is added (type 316), M₂₃C₆ can partially transform to M₆C (see further); it is however always reported that M₂₃C₆ remains the main carbide even after long ageing. In fact, M₂₃C₆ and M₆C also coexists in the X5CrNiMo1713 (17Cr, 13Cr, 5Mo wt%) studied by Thier et al. [40]. This is not the case when niobium is involved in the formation of M₆C, which seems then more stable and becomes the dominant carbide at long ageing times [41].

In stabilised grades, the situation is much more complex: from the literature found, it is not possible to state clearly whether M₂₃C₆ is sometimes first formed or always follows MX precipitation, nor which phase is the more stable on long term ageing.

Thorvaldsson and Dunlop ([24],[42],[43]), studying the effect of different stabilising elements and their combinations, found that M₂₃C₆ was more stable than TiC, but less than NbC. This is consistent with Grot and Spruiell [44], who found formation of M₂₃C₆ in a type 321 after long term ageing, and with Kikuchi et al. [9] for a 20/25. These authors concluded that TiC retards the precipitation of M₂₃C₆ but does not suppress it.

On the other hand, Bentley and Leitnaker [45], studying a type 321 steel having been in service for 17 years at 600 $^\circ$C did not find any M₂₃C₆ and concluded that TiC was more stable. Lai [46] observed no M₂₃C₆ in a type 321 used up to rupture (16000-29000 h at 600 $^\circ$C).

Obviously, considerations on the relative stability of the two phases only make sense if the titanium content is enough to combine all carbon present, as in the opposite case, M₂₃C₆ forms with the excess carbon. The studies in which M₂₃C₆ formation appears as an anomaly deal with steels in which Ti content is higher or equal than that required for full stabilisation.

The agreement is better for NbC, which is more stable than M₂₃C₆. However, M₂₃C₆ can form as a transient phase.

M₆C

M₆C is also known as $\eta$-carbide, but will be referred to as M₆C to avoid confusion with $\eta$-phase (Laves phase). It generally refers to a phase of very variable composition. In most cases, it is formed only after long ageing times, and therefore is quite little documented. Moreover, its appearance and disappearance are strongly linked to that of other constituents and it is difficult to give it any absolute position in a TTP (Time Temperature Precipitation) diagram.

Structure and composition

M₆C is a diamond-type fcc carbide whose lattice parameter varies between 10.95 and 11.28 Å. Its space group is Fd3m, which distinguishes it from M₂₃C₆ and G-phase which both are fcc of similar lattice parameters, but have a Fm3m space group. The $\eta$-carbide structure is described by Stadelmaier in [47] for example. This structure encompasses a wide range of compositions and only the one which are commonly found in creep resistant austenitic stainless steels will be reviewed here.

M₆C composition can be molybdenum-rich ( (FeCr)₂₁Mo₃C₆) or niobium-rich ( Fe₃Nb₃C). The molybdenum rich Fe₃Mo₃C (a=11.11 Å) was reported in high Mo steels, but never in austenitic steels [48]. Instead, a fraction of molybdenum is replaced by iron or chromium, and the lattice parameter is reduced (a=10.95 Å) [48]. The composition reported by Brun et al. [49] in a 316 steel containing Ti shows substantial amounts of Ni also dissolving in M₆C (see table 4).

Table: Composition of M₆C as measured by Brun et al. in a 316 steel containing titanium.

Si at%	Mo at%	Ti at%	Cr at%	Mn at%	Fe at%	Ni at%
4.5	6.5	0.8	30	0.5	26.7	31

Silicon has been reported to dissolve in this phase to form M₅SiC, but such a phase is very seldom found. Other elements which can be included in the general notation M are Ni, Ti, Co. In NF-709 (a 20/25 Nb stabilised, with 0.17N wt%), long-term ageing allows formation of Cr₃Ni₂SiC; such a composition has been reported by Williams et al. in a study of irradiated type 316 (1984,[50]) and by Titchmarsh et al. in a similar steel (1981,[51]), at rather low temperatures (466 $^\circ$C) but not above 670 $^\circ$C. Its formation in such steels is linked with the segregation effects caused by irradiation, in particular the Si segregation to point defect sinks. This particular composition has a lattice parameter of 10.62Å (JCPDS 17-330), which makes it extremely similar to M₂₃C₆ from a structural point of view. Although always referred to as Cr₃Ni₂SiC, its actual composition includes substantial amounts of Mo and Fe (see table 5), the concentrations of which increase with temperature (Williams,[50]). Williams therefore proposed the more general formula (Cr,Mo)₃(Ni,Fe)₂SiC.

Table: Some compositions of Cr₃Ni₂SiC in irradiated 316, from [50]

Temperature ($^\circ$C)	Cr at%	Mo at%	Si at%	Ni at%	Fe at%
500 $^\circ$C	35	11	20	24	10
600 $^\circ$C	32	15	20	24	10

Jargelius-Peterson [52] found a nitrogen rich similar phase ( Cr₃Ni₂SiN) after furnace ageing of a 20Cr25Ni5Mo0.2N steel. It is reported after 5h and 3000 h at 850 $^\circ$C, therefore being probably an equilibrium phase. In agreement with the observation that the Mo content increases with temperature, the composition found here includes as much as 25 wt% Mo (this is also due to the larger Mo content of the alloy).

Occurrence and stability

Mo rich compositions

In a type 316 steel, the composition of M₆C is close to (FeCr)₂₁Mo₃C₆. Weiss and Stickler [48] proposed that it formed as follows

$${\mathrm{M}}_{23}{\mathrm{C}}_6 \longrightarrow {\mathrm{(Fe... ...}}_{3}{\mathrm{C}}_{6} \longrightarrow {\mathrm{M}}_6\mathrm{C}$$

It is not reported by Minami et al. [41] after 10000 h at 750 $^\circ$C (instead, $\chi$-phase is found), neither by Lai and Meshkat [53] after 10000 h at 600 and 650 $^\circ$C, nor by Barcik [54] for the same time, up to 900 $^\circ$C. However, it was found by Stoter [55] in specimen of 316 having been in service at 650 $^\circ$C for 28000 and 60000 h, in very small amounts and always associated with M₂₃C₆.

Nitrogen seems to have a large influence on M₆C formation: Thier et al. [40] did not find this carbide after 1000 h in a type 316 with 0.037N wt%, but found it after only 1 h ageing at 900 $^\circ$C when the nitrogen content was 0.069 wt%. Gavriljuk and Berns [8] suggest that the calculations performed by Jargelius-Petterson [56] raise controversy, as she shows that an increase of nitrogen in a 20Cr, 25Ni, 4.5Mo wt% steel reduces the driving force for M₆C. However, one must notice that M₆C is quite poorly described in the SGTE databases, as it only contains information about the Mo rich carbide, that is to say it describes only one of the possible compositions of the $\eta$-structure. In this regard, it is possible to say that the $\eta$-structure is in both cases stabilised but with different compositions. In particular, Jargelius-Petterson [56] reports that M₅SiN is always present and favoured by high nitrogen contents. It seems therefore that nitrogen generally stabilises the $\eta$-structure, but the composition of this phase varies with the alloy.

In type 321 (Ti-stabilised), none of the studies found report M₆C.

Nb rich compositions

In type 347, on the contrary, the presence of Nb seems to promote the formation of a niobium rich M₆C. Care should be taken as Powell et al. [57] have given sensible arguments showing that G-phase and M₆C had often been mistaken. However, if the structures are very similar, the composition should differentiate clearly Fe₃Nb₃C or Cr₃Ni₂SiC (as in NF709) from Ni₁₆Nb₆Si₇. The experimental methods must be linked with the results of the different studies. In general, Nb-rich M₆C seems to form only on long-term ageing. Kikuchi et al. [9] do not report it after 1000 h at 700 $^\circ$C in a 20/25. In a 18/8 steel, it is reported between 600 and 800 $^\circ$C by Minami et al. [41] at very long time ($\sim$50000 h) for 600 $^\circ$C but faster ($\sim$2000 h) for 800 $^\circ$C. They proposed a sequence illustrated in fig 6.

Figure: Precipitation sequence leading to M₆C when Nb is in excess, as proposed by Minami et al. in [58]

This mechanism implies excess of Nb is present. If Nb is less than that required for combination of all carbon, no such phase is observed (Steel Tempaloy A1 in [41], similar to a 304 with 0.13Nb, 0.06Ti, 0.07C wt%)

In 20/25 alloys, as written above, M₆C and G-phase have probably been often mistaken (e.g. [59]) in early studies where identification relied on X-ray or electron diffraction only. It is difficult to conclude which phase forms preferentially: Ramaswamy et al. [60] report M₆C in a 20/25 with low Si content (0.03 wt%), but other studies report G-phase (Si content 0.4-0.7 wt%).

When nitrogen is present (in 347 or in 20/25) in sufficient quantities, Z-phase forms and it is difficult to have an idea of its stability with regard to M₆C.

Cr₃Ni₂SiX and G-Phase

G-phase, which is considered further, is an alternative Si rich phase to Cr₃Ni₂SiC. As mentioned above, the composition Cr₃Ni₂SiC is very seldom reported in the literature. Titchmarsh and Williams have reported its formation in irradiated steel of composition close to that of 316 with addition of 1.8wt% Nb. They noted [51] and provided evidence [61] that G-phase formed preferentially only when carbon was not available.

In 20/25-Nb-C steels though, Powell et al. (1984 [62], 1987 [57]) and Ecob et al. (1987 [63]) have found G-phase and observed that NbC partially transforms to G-phase with time. It seems reasonable to propose that in 20/25, G-phase is stabilised with regard to Cr₃Ni₂SiC, probably because of the larger Ni content.

However, in NF709 [5] (composition in table 1), Cr₃Ni₂SiC is reported. This is inconsistent with the studies quoted above which seem to indicate G-phase as a more stable phase. It is not clear however whether the presence of carbon or nitrogen was investigated or the composition Cr₃Ni₂SiC assumed. Investigations of the exact nature of this precipitate are required to determine whether it is a nitride or a carbide. A nitride would be expected as it would not be incoherent with the former observations that G-phase is more stable than Cr₃Ni₂SiC and that the $\eta$-structure is stabilised by nitrogen.

Intermetallic phases

Sigma phase

Sigma phase is a well-known intermetallic phase which forms in the Fe-Cr system, with a composition FeCr. In highly alloyed steels, its composition is variable. Its precipitation is associated with embrittlement. In creep resistant steels, it has a detrimental effect on creep properties when precipitated on grain boundaries, but little effect when it precipitates intragranularly.

Structure and Composition

Sigma has a tetragonal unit cell (space group P4₂/mnm) with a=8.80 Å and c=4.54 Å (FeCr). This unit cell contains 30 atoms. Reported orientation relationships are [4]:

$$\displaylines{ {\left( 111 \right)}_{\gamma } \parallel {\lef... ...amma } \parallel {\left( {\overline 1}10 \right) }_{\sigma } }$$

The composition varies quite widely and it is difficult to give a formula. For example, Jargelius-Petterson [56] reports the following range of composition for sigma in a 20/25 with Mo content varying between 4.5 and 6.0 wt%:

	Si	Cr	Mn	Fe	Ni	Mo
wt%	0-1	27-32	1-9	35-43	8-15	10-16

for a type 316 [36], a typical composition is (other elements not given):

	Fe	Cr	Mo
wt%	44	29.2	8.3

Occurrence

An in-depth review of $\sigma$-phase precipitation in Cr-Ni austenitic steels has been done by Barcik [64]. Sigma phase precipitates first on triple points then on grain faces. After long term ageing at high temperature, it also forms on incoherent twin boundaries and intragranular inclusions. The mechanism of nucleation is still a matter of controversy, particularly on the role of $\delta$-ferrite and M₂₃C₆ in the nucleation process. Studies have reported its formation associated with the dissolution of M₂₃C₆, but also independently [4]. It is found in most of the grades of austenitic stainless steels: type 304, 316 (Mo), 321 (Ti stabilised) and 347 (Nb stabilised) (see for example [41]). However, it forms after different times. The 25Cr-20Ni steels can exhibit as much as 20 wt% of $\sigma$-phase.

Figure: Precipitation of sigma in different grades of austenitic stainless steels. Tempaloy-A1 is an 18/10 with a Nb/C ratio of 1.86, whereas the 347 has a ratio of 17.40. The percentage is the area etched by KOH. After [41]

In stabilised grades, its formation is faster than in other grades: Minami et al. report precipitation of $\sigma$ after 1000 h at 700 $^\circ$C in 347 and 321 (347 precipitating $\sigma$ slightly faster than 321). In 304, 316 and Tempaloy-A1, $\sigma$ phase is found in significant quantities only after 10000 h (fig 7). It is worth noting that Tempaloy-A1 is Nb stabilised, but with a Nb/C ratio of 1.86, while this ratio is 17.40 in the 347. This correlates well with the fact that $\sigma$ forms when the carbon content falls below a critical value when the chromium equivalent is higher than 18wt%. In the 347, almost all the carbon is rapidly precipitated as NbC, while the low Nb content of Tempaloy-A1 leaves some carbon in solution. The different trend for 321 could be linked to the instability of TiC with regard to M₂₃C₆. The precipitation of M₂₃C₆ lowers both the carbon and the chromium content. The results of Grot and Spruiell [44] show on the contrary no sigma phase forming up to 2000 h in a type 321. $\sigma$-phase is also found in 20/25 (fig 8). Different factors affect the formation of sigma phase. Elements like Cr, Nb, Ti or Mo are known to promote $\sigma$ formation. Silicon promotes and accelerates its formation. In general, the formation of sigma in austenite is about 100 times slower than in ferrite. Consequently, the presence of $\delta$-ferrite accelerates sigma precipitation. [64]

A method has been developed by Woodyatt et al. [4] to estimate the $\sigma$-forming tendency of an alloy, based on the electron vacancy number ${\overline N}_{v}$:

$$\begin{eqnarray*} {\overline N}_{v} & = & 0.66{\mathrm{Ni}}+1.71{\mathrm{Co}}+2.... ... + 6.66{\mathrm{Zr}} + 10.66{\mathrm{Nb}} { {\mathrm{at\%}}} \end{eqnarray*}$$

If ${\overline N}_{v}$ is higher than 2.52, the alloy should form $\sigma$.

Laves phase

Laves phase is found in various grades of austenitic stainless steels. It is often a minor constituent. It precipitates intragranularly in the form of equiaxed particles, occasionally on grain boundaries [48]. Whether Laves phase has a detrimental effect or not on creep properties is still discussed. In niobium stabilised steels however, it is likely that Laves phase are detrimental because they allow formation of M₆C (see below).

Structure and composition

Laves phase is a hexagonal phase of space group P6₃/mmc, with a=4.73 Å and c=7.72 Å. The main factor determining its formation being the relative atomic size of the constituent atoms, the ranges of composition are quite small. Jargelius-Petterson [56] measured the following composition for Laves phase in a 20/25 with 5Mo wt%:

After 10 min at 850 $^\circ$C
	Fe	Mo	Cr	Ni	Mn	Si
wt%	$29\pm2$	$36\pm2$	$19\pm2$	$13\pm1$	$1.5\pm0.8$	$1.8\pm0.5$
After 3000 h at 850 $^\circ$C
wt%	$37\pm1$	$35\pm1$	$12\pm1$	$14\pm1$	$\sim0$	$0.7\pm0.2$

This shows that Laves phase probably does not form at its exact equilibrium composition. In stabilised grades, Fe₂Nb or Fe₂Ti (more seldom) can form.

Denham et al. (1969, [65] proposed for Fe₂Nb the following orientation relationships:

$$\displaylines{ {\left( 0001 \right)}_{{\mathrm{Fe}}_{2}{\math... ...{\mathrm{Nb}}} \parallel {\left( \bar{1}10 \right)}_{\gamma} }$$

Occurrence

In type 304, Laves phase is not found, due to the absence of Mo, Nb or Ti. In type 316, which contains typically 2-2.5Mo wt%, it is found after relatively long ageing times: Minami et al. [41] report its formation after a minimum of 1000 h, between 625 and 800 $^\circ$C. The experimental results of Lai [66] are in agreement with the dissolution of Fe₂Mo above 800 $^\circ$C. Coherently, White and Le May [67] do not report any Laves phase precipitation between 640 and 800 $^\circ$C for ageing times up to 900 h. Weiss and Stickler [48] report a slightly higher maximum temperature for Laves precipitation in type 316 as they found it up to 815 $^\circ$C. Its formation interacts in a competitive way with that of $\sigma$ and $\chi$ phases. For example, the presence of $\delta$-ferrite delays its formation because it enhances that of $\sigma$ and $\chi$ phases.

In titanium stabilised grades, the formation of Fe₂Ti is never reported in compositions similar to that of a 321 steel. Minami et al. do not report it for ageing times up to 50000 h between 600 and 750 $^\circ$C, in a type 321. However, it is found by Beattie and Hagel [68] in a A286 type alloy, containing 16Cr, 26Ni and 1.8Ti wt%, after 1000 h at 815 $^\circ$C. This is directly related to the large amount of Ti used in such steels compared to a typical 321.

In niobium stabilised steels, Fe₂Nb is frequently reported after long ageing times, but as a transient phase which disappear for Fe₃Nb₃C. It is reported to form in a type 347 with 0.87Nb and 0.05C wt%, after 1000 h between 650 and 800 $^\circ$C, and disappear after 5000-10000 h [41]. However, its formation is dependent on the availability of niobium. In the same study, the Tempaloy-A1 (18/10 with 0.13Nb for 0.07C wt%) do not precipitate Laves phase after ageing treatments up to 25000 h. Instead, only NbC is found (see figure 6). The same results are reported by Raghavan et al. [21], who found NbC and Laves phase in a type 347 with 0.8Nb and 0.07C wt% (8000 h at $\sim$600 $^\circ$C), but only Z-phase in a modified 347 with 0.3Nb and 0.09N wt%. It is therefore probable that both NbC and Z-phase are more stable than Fe₂Nb.

$\chi$ phase

$\chi$ phase is most of the time a minor intermetallic phase. It is mainly found in type 316 above 750 $^\circ$C

Structure and composition

$\chi$ phase is a bcc (body-centered cubic) phase of space group $1{\overline 4}3m$. The unit cell contains 58 atoms and has a lattice parameter varying between 8.807 and 8.878 Å. A typical composition is Fe₃₆Cr₁₂Mo₁₀ but the phase has a high tolerance for metal interchanges. The structure has large holes which allow presence of carbon, sometimes making this phase better described as a M₁₈C carbide. It nucleates on grain boundaries, incoherent twin boundaries but also intragranularly on dislocations.

Occurrence

Although it is sometimes (but seldom) present in very small quantities in 321 [44], $\chi$ phase is only a significant phase in type 316, when the ageing temperature is high enough. For example, it is reported by Minami et al. for times up to 25000 h and temperatures between 700 and 850 $^\circ$C, and by Lai et al., as being the main phase with $\sigma$ between 715-850 $^\circ$C for times up to 5000 h, but not by Stoter [55] after 28000 and 60000 h at 650 $^\circ$C.

G-Phase

G-Phase is a silicide forming in austenitic stainless steels stabilised with Ti or Nb. It has been first reported by Beattie and Versnyder in 1956, in a A286 type steel, that is a precipitation-hardening 26Ni-15Cr with variable (0 to 2.3wt %) Ti and Al contents. It has long been believed to form only under irradiation in steels of the 300 series and in 20/25 steels. However, Powell et al. showed, in 1985 [57] that it formed under normal ageing in 20/25 Nb stabilised alloys, in the temperature range 500-850 $^\circ$C, and suggested that it had been identified as M₆C in earlier studies of the same steel.

Structure and Composition.

G-Phase has a general formula A₁₆D₆C₇ where A and D are transition elements, and C a group IV element [63]. A is usually nickel, D niobium (ref [57],[62],[63]) or titanium [46]. Ecob et al. [63] mentioned the possibility of substitution, in small quantities, of D (ie Nb or Ti) by Fe or Cr, and Powell et al. [57] measured the following composition:

Table: Composition of G-Phase measured in [57]

	Ni	Nb	Si	Fe	Cr
Wt%	51	29	14	4.5	1.5

Lai found a titanium containing G-phase in a type 321 steel, and gives a composition sensibly richer in nickel (table 7).

Table: Composition of G-Phase measured in [46]

	Ni	Ti	Si	Fe	Cr
Wt%	63.3	20.9	12.2	3.47	0.13

Note that none of these compositions clearly shows substitution of D when trying to match the stoichiometric formula.

G-Phase has a fcc structure with a lattice parameter of 11.2 Å, this corresponds to a content of 116 atoms per unit cell. The space group for this structure is Fm3m. It is remarkable that the lattice parameter seems to be the same for Ni₁₆Nb₆Si₇ (see [57],[62]) and Ni₁₆Ti₆Si₇ (see [46],[68]). In a earlier study, Sumerling et al. [59], studying a 20/25 Nb stabilised steel, found a lattice parameter of $11.25\pm0.05$ Å for a phase they identified as M₆C .

As mentioned before, Powell et al. have suggested that G-Phase was identified as M₆C in early works on 20/25 Nb steels, because of close compositions and structures. A detailed investigation of the structure is sufficient to solve the problem (Ecob et al.,[63]), but additional evidence has been sought by the use of EELS (Electron Energy Loss Spectroscopy) (as in [57]) or a wavelength-dispersive crystal spectrometer in SEM (as in [46]), both techniques making possible quantitative measurements of light elements (C,N). These studies have confirmed the absence of interstitial elements in G-Phase.

Occurrence of G-Phase in Austenitic Stainless Steels.

The general features of G-Phase formation are:

1. It forms very predominantly on grain-boundaries (from where its name)
2. It is now clear that it forms in 20/25 Nb stabilised steels and to an extent which depends on the Si content.
3. It has also been reported in a type 321 steel, but other results are contradictory.

Ti rich G-phase

In one of the first studies on G-Phase (Beattie and Hagel,[68]) the Ti rich G-phase is reported in a A286 type steel. It is not clear in what conditions of temperature and time it forms, as it was often found to resist the solution-treatment. However, G-Phase was not found in the alloy containing only 0.01Si wt%, but formed at higher contents, and the volume increased with the Si content. Adding 2 wt% of Al suppressed the formation of G-Phase. This work also included a steel in which the Ti content is more similar to that used in the 300 series, and no G-phase was found.

The only report found of Ti rich G-phase in type 321 has been quoted above (Lai, [46]). Three out of seven type 321 steels show G-phase after 16000h and 50000h at 600 $^\circ$C. The common factors between the three type 321 steels which are found to form Ti rich G-phase in this study are their small grain sizes and an excess of Ti with regard to the amount of carbon present. However, in similar conditions, other studies (Minami et al., [41]; Bentley et al. [45]) do not report G-phase.

Nb rich G-phase

In 20/25-Nb-C stabilised steels, it is now clear that G-Phase can form (see [57], [63], [62]) and, following the suggestion of Powell et al., the works of Sumerling et al. and Dewey et al. may be regarded as other evidences for G-Phase formation if one accepts that M₆C must be read as G-Phase. In [59], in particular, the authors measured a lattice parameter of $11.25\pm0.05$ Å which is much closer to the G-Phase parameter.

Figure: TTP diagram from Powell et al. [57]

Powell et al. summarised their work in a TTP (Time Temperature Precipitation) diagram shown in fig 8. In their alloy (0.68Nb, 0.037C, 0.61Si wt%), G-Phase forms between 500 and 850 $^\circ$C, first on residual grain boundary particles of NbC, and only after very long ageing on NbC particles in the matrix. The authors propose that this is due to the easier diffusion of required elements in the grain boundaries.

Ecob et al. [63], comparing the occurrence of G-Phase in similar 20/25-Nb stabilised steels, noticed that alloys apparently similar in composition exhibited different relative stabilities of NbC with regard to G-Phase. They found that an increase in the oxygen content led to a greater instability of NbC with regard to G-Phase, and proposed that oxygen and silicon are involved in a co-segregation process, a greater amount of oxygen segregating around NbC leading to a greater segregation of Si in the same way. The silicon rich region is more favorable to G-Phase formation.

In 20/25-Nb-N steels however, as it has been discussed in the section regarding M₆C, it seems that Cr₃Ni₂SiN forms preferentially to G-phase. No results could be found that support this hypothesis.

In common 18/12-Nb steels, G-phase seems to be found only under irradiation (Titchmarsh and Williams [51], Williams [50]). Increasing the Si content to 6 wt% resulted in the formation during furnace ageing of Nb G-phase, although provided that carbon was not available to form Cr₃Ni₂SiC. However, no study has been found which report niobium G-phase in ordinary 18/12 steels.

Ni₃Ti and related precipitates

Precipitates such as Ni₃Ti, Ni₃(Ti,Al), are only found in a particular class of austenitic stainless steels: the precipitation hardening ones. A typical steel in this category is A286, a 15Cr-25Ni-Al-Ti steel. They contain titanium and aluminium in quantities substantially larger than the 300 series and related compositions, to form $\gamma^{\prime}$ or other Ni, Ti, Al precipitates, as summarised in figure 9.

Figure: Effect of aluminium and titanium on the equilibrium phases present at 800 $^\circ$C in 15Cr-25Ni austenitic steels, from [69]

However, there are very few studies in which the creep properties of these steels are given attention, and long term informations are not available (see for example [70]). For this reason and because the precipitation in these steels has been reviewed in [71], [69], it has been decided not to cover in depth these precipitates. These are [71]:

• $\gamma^{\prime}$ is a fcc Ni₃Al or Ni₃(Ti,Al) intermetallic compound of lattice parameter 3.60 Å. It is also the transition structure of Ni₃Ti.
• $\eta$ is hexagonal (a=5.09 Å c=8.32 Å) and is the equilibrium structure of Ni₃Ti.
• $\beta$ is NiAl, formed at lower temperatures (up to 700 $^\circ$C) and short ageing times (10-100h).

Other precipitates

Attention has been concentrated on precipitates most often reported and studied in creep resistant austenitic stainless steels. However, other phases are sometimes reported depending on the alloying elements.

Cr₂N

It forms in non-stabilised austenitic stainless steels with high nitrogen (0.2 wt% at 900 $^\circ$C with 20Ni wt%) [38]. It has a hcp (hexagonal close-packed) unit cell of lattice parameters a=4.78 Å and c=4.44 Å [4]. Although an major precipitate in high nitrogen non stabilised steels, it is not in either nitrogen-bearing non-stabilised creep resistant austenitic stainless steels because of the lower N content used, or in stabilised nitrogen bearing grades, because Z-phase seems to form preferentially. A detailed review of this precipitate can be found in [8].

Pi-nitride

It is reported by Jargelius-Petterson [56], [52] to form in a non-stabilised 20/25 steel with 0.21N wt%. The composition after 10 min at 850 $^\circ$C is:

	Si	Mn	Cr	Ni	Mo	Fe
wt%	1.0 $\pm$ 0.2	1.1 $\pm$ 0.3	47 $\pm$ 1	20$\pm$ 1	13 $\pm$ 1	19 $\pm$ 1

It has a cubic structure of lattice parameter a=6.3 Å [52].

Titanium carbosulphides

The titanium carbosulphide Ti₄C₂S₂ is reported by Lai [46] in different type 321 steels aged at $\sim$600 $^\circ$C for times between 16000 and 53000 h, and by Minami et al. [41] in a similar steel, they precise that this phase is already present in the solution treated condition. It has a hcp structure of lattice parameter a=3.21 Å and .

Copper precipitates

Few data are available concerning the effect of copper in creep resistant austenitic stainless steels. Tohyama et al. [72] used 3 wt% copper in Tempaloy-A1, a steel similar to 347, with addition of a small amount of titanium. This results in precipitation of a copper rich phase, independently of the precipitation of other precipitates. The creep rupture strength is significantly increased in comparison to the original composition.

Chromium phosphides

The chromium phospide Cr₃P has been reported by Rowcliffe et al.[73], in a 18/10 steel with 0.3 wt% P. It has a tetragonal unit cell with a=9.186Å and c=4.558Å.

Concluding Remarks

An attempt has been here made to review the characteristics of the main precipitates found in austenitic stainless steels. Whereas the early stages of precipitation are quite well understood in the usual grades of austenitic steels, this is far from the case for the new generations of nitrogen-bearing stabilised austenitic steels. There is a dirth of knowledge about the phases which precipitate during long term ageing, particularly with respect to thermodynamic data.

Z-phase, which plays an essential role in recent creep-resistant austenitic or ferritic stainless steels, is not present in the SGTE databases commonly used with thermodynamic calculation packages such as MT-DATA or ThermoCalc. The kinetics of its formation are also rather obscure.

In discussing the stability of M₆C, there is some confusion which arises from the fact that this phase is most often referred to as one particular composition of the eta structure rather than the eta-structure itself. Similarly, thermodynamic data are only available for the Mo-rich pole of the eta structure. Other important compositions which are not present in the SGTE databases include Fe₃Nb₃C and Cr₃Ni₂SiX.

The formation of MX precipitates is well documented, but again, thermodynamic data are missing to model the solubility of Cr, which is likely to be important in the kinetics of precipitation. An assessment of the Cr-Nb-N system could help improving the description of NbX precipitates, and would also provide thermodynamic data for Z-phase.

Acknowledgements

I am grateful to David Gooch of National Power, who funded the project this literature review is part of. I am also thankful to Professor Alan Windle of the university of Cambridge for provision of laboratory facilities, and to my supervisor, Professor H.K.D.H. Bhadeshia, for his support, enthusiasm and fruitful discussions.

Table 8: Crystal structure and composition of the main precipitates in austenitic stainless steels

Precipitate	Structure	Parameter (Å)	Composition
NbC	fcc	a=4.47	NbC
NbN	fcc	a=4.40	NbN
TiC	fcc	a=4.33	TiC
TiN	fcc	a=4.24	TiN
Z-phase	tetragonal	a=3.037 c=7.391	CrNbN
M23C6	fcc	a=10.57-10.68	Cr16Fe5 Mo2C (e.g.)
M6C	diamond cubic	a=10.62-11.28	(FeCr)21Mo3 C; Fe3Nb3C; M5SiC
Sigma	tetragonal	a=8.80 c=4.54	Fe,Ni,Cr,Mo
Laves phase	hexagonal	a=4.73 c=7.72	Fe2Mo, Fe2Nb
$\chi$-phase	bcc	a=8.807-8.878	Fe36Cr12 Mo10
G-Phase	fcc	a=11.2	Ni16Nb6 Si7, Ni16Ti6 Si7

Bibliography

1

Pickering F. B.
Physical Metallurgical Development of Stainless Steels.
In Proc. Conf. Stainless Steels 84, Gothenburg, Sept 1984, pages 2-28, 1984.

2

MT-DATA.
National Physical Laboratory, Teddington, Middlesex, U.K., 1989.

3

Lacombe P., Baroux B., and Beranger G., editors.
Stainless Steels.
Les Editions de Physiques, Les Ulis, 1993.

4

Marshall P.
Austenitic Stainless Steels, Microstructure and Mechanical Properties.
Elsevier, London, 1984.

5

Nippon Steel Corporation.
Quality and Properties of NF709 Austenitic Stainless Steel for boiler tubing applications.
Technical report, Nippon Steel Corporation, 1996.

6

ASM International.
Irons, steels and high performance alloys.
In Davis J.R., editor, Metals Handbook, volume 1. 1990.

7

Naylor D.J. and Cook W.T.
Materials Science and Technology, volume 7.
VCH Publishers Inc., 1992.

8

Gavrilnjuk V.G. and Berns H.
High Nitrogen Steels.
Springer, 1999.

9

Kikuchi M., Sakabibara M., Otoguro Y., Mimura M., Takahashi T., and Fujita T.
An Austenitic Heat Resisiting Steel Tube developed for Advanced Fossil Steam Plant.
In International Conference on Creep, Tokyo, April 14-18, 1986, pages 215-220, 1986.

10

Rios P.
Expression for Solubility Product of Niobium Carbonitride in Austenite.
Mater. Sci. Techn., 4:324-327, 1988.

11

Rios P.
Method for Determination of Mole Fraction and Composition of a Multicomponent f.c.c. Carbonitride.
Mater. Sci. Eng., 142:87-94, 1991.

12

Heilong Z. and Kirkaldy J. S.
Thermodynamic Calculation and Experimental Verification of the Carbonitride-Austenite Equilibrium in Ti-Nb Microalloyed Steels.
Metal. Trans. A, 23:651-657, 1992.

13

Hughes H.
Complex Nitride in Cr-Ni-Nb Steels.
J.I.S.I., pages 775-778, 1967.

14

Andrén H. O., Henjered A., and Karlsson L.
MX Precipitates in Stabilized Austenitic Stainless Steels.
In Proc. Conf. Stainless Steels 84, Gothenburg, Sept 1984, pages 91-96, 1984.

15

Andrén H. O., Henjered A., and Norden H.
Composition of MC Precipitates in a Titanium Stabilised Stainless Steel.
J. Mater. Sci., 15:2365-2368, 1980.

16

Wadsworth J., Woodhead J. H., and Keowm S. R.
The Influence of Stoichiometry upon Carbide Precipitation.
Metal Sci., pages 342-348, 1976.

17

Keown S. R. and Pickering F. B.
Effect of Niobium Carbide on the Creep Rupture Properties of Austenitic Stainless Steels.
In Creep strength in steel and high-temperature alloys, pages 229-234, 1974.

18

Adamson J. M. and Martin J. W.
Tertiary Creep Processes in 20%Cr, 25%Ni Austenitic Stainless Steels of differing Nb/C Ratios.
J.I.S.I., pages 271-275, 1972.

19

Jack D. H. and Jack K. H.
Structure of Z-Phase.
J.I.S.I., pages 790-792, 1972.

20

Vodárek V.
Morphology and Orientation Relationship of Z-phase in Austenite.
Scripta Metall. Mater., 25:549-552, 199