Open Archive Toulouse Archive Ouverte (OATAO)

OATAO is an open access repository that collects the work of Toulouse researchers
and makes it freely available over the web where possible.

This is an author-deposited version published in: http://oatao.univ-toulouse.fr/

Eprints ID: 5550

To link to this article: DOI: 10.1016/j.corsci.2011.05.013

URL : : http://dx.doi.org/10.1016/j.corsci.2011.05.013

To cite this version:

Gheno, Thomas and Monceau, Daniel and Zhang, Jianqiang and Young,
David J. Carburisation of ferritic Fe–Cr alloys by low carbon activity
gases. (2011) Corrosion Science, vol. 53 (n° 9). pp. 2767-2777. ISSN
0010-938X

Any correspondence concerning this service should be sent to the repository

administrator: [email protected]
Carburisation of ferritic Fe–Cr alloys by low carbon activity gases
Thomas Gheno a,b, Daniel Monceau b, Jianqiang Zhang a, David J. Young a,⇑
a
School of Materials Science and Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
b
Institute Carnot CIRIMAT, ENSIACET, 31030 Toulouse Cedex 4, France

a b s t r a c t

Model Fe–Cr alloys were exposed to Ar–CO2–H2O gas mixtures at 650 and 800 °C. At equilibrium, these
atmospheres are oxidising to the alloys, but decarburising (aC  10ÿ15 to 10ÿ13). In addition to developing
external oxide scales, however, the alloys also carburised. Carbon supersaturation at the scale/alloy inter-
face relative to the gas reflects local equilibrium: a low oxygen potential corresponds to a high pCO =pCO2
ratio, and hence to a high carbon activity. Interfacial carbon activities calculated on the basis of scale–
Keywords: alloy equilibrium are shown to be in good agreement with measured carburisation rates and precipitate
A. Steel
volume fractions, providing support for the validity of the thermodynamic model.
C. Carburisation
C. High temperature corrosion
C. Oxidation

1. Introduction was formed at the higher temperatures and produced ferrite plus
pearlite, martensite and/or (Fe,Cr) carbides upon cooling to room
Materials to contain and convey hot gases rich in CO2 are temperature, according to the alloy chromium content.
needed in various carbon capture technologies currently being Clearly, the equilibrium carbon activity of the bulk gas phase is
developed. The process gases concerned are expected from their insufficient to form the carburisation products reported. McCoy [3]
composition (principally CO2 plus H2O) to induce oxidation, but suggested that oxide formation from direct reaction of iron with
seem to pose no threat to conventional materials in terms of car- CO2 could generate CO,
burisation. For instance, pure CO2 equilibrated at atmospheric
FeðsÞ þ CO2 ðgÞ ¼ FeOðsÞ þ COðgÞ ð1Þ
pressure and a temperature of 650 °C corresponds to pO2 = 1.5 
10ÿ8 atm and aC = 2.7  10ÿ15. Nonetheless, available information which would then build up in pores within the oxide, causing the
indicates that heat resisting alloys can also be attacked by carbon local carburising potential to increase. Fujii and Meussner [5] pro-
in these gases. posed that carbon was generated on the oxide surface, transported
The oxygen partial pressure of CO2 gas is high enough for iron to through a dense outer oxide layer and produced a CO–CO2 atmo-
oxidise, and dilute Fe–Cr alloys readily form Fe-rich oxide scales sphere within voids in a porous inner oxide layer. However, subse-
upon exposure to CO2 at high temperatures. In addition, carburisa- quent experiments [6] showed that the solubility of carbon in iron
tion of the underlying substrate has been reported by several oxides is extremely small, and its solid-state diffusion through iron
authors. Internal carbides were observed after exposure of 9Cr oxide scales is not feasible. Transport could occur instead by means
(P92) and 12Cr (X20 and VM12) steels to Ar–50%CO2 at 550 °C of molecular CO2, but the scale microstructures are reported [5] to
[1], and after exposure of model Fe–Cr alloys to Ar–30%CO2 at appear impermeable to gas. Furthermore, these scales grow at rates
650 °C [2]. Reaction of pure CO2 with an 18Cr (304) steel at 704, controlled by solid-state diffusion, and continuous pathways for gas
816 and 927 °C [3] and with an Fe–15Cr alloy at 900 °C [4] also phase transport through the scale therefore cannot exist during
led to internal carbide precipitation. Fujii and Meussner [5] re- steady-state oxidation. In more recent studies involving carburisa-
ported simultaneous oxidation and carburisation of Fe–Cr alloys tion of chromia-forming materials [7–9], molecular transport via
containing 1–15 wt.% Cr in pure CO2 at temperatures of 700, 900 grain boundaries has been suggested to account for carbon transfer
and 1100 °C. All alloys remained ferritic at 700 °C and produced through the scale.
(Fe,Cr) carbides, the nature of which varied with alloy chromium Several descriptions [1,2,10,11], which are thermodynamically
content from (Fe,Cr)3C to (Fe,Cr)7C3 and (Fe,Cr)23C6, in accordance equivalent but mechanistically different, have been proposed to
with the Fe–Cr–C phase diagram. Carbon-containing austenite account for the generation of a carburising environment within
the scale. In reference [10], local equilibrium is represented by
the reactions
⇑ Corresponding author. Tel.: +61 2 9385 4322; fax: +61 2 9385 5956.
E-mail address: [email protected] (D.J. Young).

doi:10.1016/j.corsci.2011.05.013
 1 Equilibrium activities of oxygen and carbon calculated from
CO2 ¼ CO þ O2 ð2Þ
2 standard free energies tabulated in Ref. [12] are given in Table 1.
The values are such that all conditions are oxidising and decarbu-
2CO ¼ CO2 þ C ð3Þ rising to the alloys.
At the scale/alloy interface, the oxygen potential is kept to a low Reaction products were examined using X-ray diffraction (XRD),
value, ðpO2 Þeq , by the metal–oxide equilibrium, optical microscopy (OM), scanning electron microscopy combined
with energy-dispersive X-ray analysis (SEM-EDX), and transmis-
y
M þ O2 ¼ MOy ð4Þ sion electron microscopy (TEM) together with EDX and selected
2 area diffraction (SAD) analysis. Metallographic observations were
and might therefore lead to elevated carbon activities, as illustrated carried out on polished and etched cross-sections. Etching with a
in Fig. 1. solution of 1% picric acid and 5% hydrochloric acid in ethanol
Microstructural changes induced in steels by carburisation will was used to reveal alloy grain boundaries and internal carbides,
affect their mechanical properties. This paper is concerned with and to distinguish wustite from the spinel phase in oxide scales.
determining the conditions leading to carburisation of model Fe– Murakami’s reagent (1 g K3Fe(CN)6 and 1 g KOH for 10 mL H2O)
Cr alloys in CO2, and understanding those conditions in terms of was also employed for revealing carbides.
the thermodynamics of reactions (2) and (3). The effect of water
vapour on carburisation is also explored.
3. Results

2. Materials and experiments 3.1. Formation of external oxide scales

Binary alloys were prepared by argon arc melting Fe (99.99% Alloys containing 2.25 and 9 wt.% Cr formed thick, multi-lay-
pure) and Cr (99.995% pure) to yield nominal compositions ered oxide scales in dry and wet CO2 at both 650 and 800 °C. The
(wt.%): Fe–2.25Cr, Fe–9Cr and Fe–20Cr. Ingots were annealed in layers were identified by alternately grinding off some oxide and
Ar–5%H2 at 1150 °C for 48 h, and cut into rectangular samples of performing XRD measurements on the ground surface. The outer
approximate dimensions 14  6  1.5 mm. As-annealed materials layers were iron oxides, while the inner layer consisted of a mix-
had microstructures of coarse-grained (500 lm) ferrite. The com- ture of FeO and (Fe,Cr)3O4 spinel. Analysis by SEM-EDX confirmed
position of an Fe–9Cr ingot was verified by chemical analysis. The the absence of chromium in the outer layers, and its presence in
chromium content was measured as (9.2 ± 0.3) wt.% by ICP-OES, the inner layer. This structure, illustrated in Fig. 2, is typical of di-
and no carbon could be detected after combustion in a LECO lute Fe–Cr alloys and has been observed by others after exposure to
CS230 instrument, with a 4 ppm sensitivity. Samples were CO2 [5,13] and CO2–H2O [1,14] in the temperature range 550–
mechanically ground to a 1200 grit finish, degreased and ultrason- 900 °C.
ically cleaned in ethanol before reaction. Oxidation of the Fe–20Cr alloy was more sensitive to exposure
Isothermal corrosion experiments were conducted at 650 and conditions. When exposed to dry CO2 at 650 °C, the alloy formed a
800 °C in Ar–20%CO2 and Ar–20%CO2–20%H2O mixtures at a total thin Cr2O3 scale interrupted by nodules of iron-rich oxide (Fig. 3a).
pressure of about 1 atm. Linear gas flow rates were set at about The surface area fraction covered by iron-rich oxide increased with
2 cm sÿ1. The wet gas was generated by passing a mixture of Ar reaction time, at least during early stages. In the wet gas, Cr2O3 was
and CO2 through a thermostated water saturator. The deminera- present only in small proportions for short reaction times, and the
lised water in contact with the gas mixture was set at 77 °C. Excess alloy mainly formed a continuous iron-rich scale. The iron-rich
water vapour was subsequently condensed by cooling the wet gas oxide was similar in structure to the oxide formed on dilute alloys
in a distillation column at 60 °C. (Fig. 2). At 800 °C, the alloy formed a continuous Cr2O3 scale in dry
Oxygen is formed by the decomposition of CO2, reaction (2), and CO2, but developed some iron-rich nodules in wet CO2 (Fig. 3b).
of H2O Scale growth kinetics and morphologies will be discussed else-
1 where. Attention is focused here on the alloy phase transforma-
H2 O ¼ H2 þ O2 ð5Þ tions accompanying corrosion.
2
while carbon is produced by the Boudouard reaction, (3), and/or the
syngas reaction 3.2. Internal reactions at 650 °C

CO þ H2 ¼ C þ H2 O ð6Þ Exposure to dry CO2 at 650 °C led to intergranular carburisation

of the Fe–2.25Cr alloy. The precipitates shown in Fig. 4 were found
throughout the whole sample after 120 h of reaction. No other
internal reaction product was observed in this alloy.
In addition to an external scale, the Fe–9Cr alloy formed inter-
nal oxides and carbides (Fig. 5). The carbides were identified as
(Fe,Cr)7C3 by grinding off the oxide scale and performing XRD mea-
surements directly below the scale/alloy interface and deeper in

Table 1
Equilibrium oxygen and carbon potentials in the reacting gas at P = 1 atm.

T (°C) Gas composition (%) pO2 (atm) aC

Ar CO2 H2O
650 80 20 – 5.1  10ÿ9 1.6  10ÿ15
60 20 20 1.1  10ÿ8 7.7  10ÿ16
Fig. 1. Activities in a gas–MOy–M system, where M is a pure metal. Equilibrium 800 80 20 – 1.6  10ÿ7 6.9  10ÿ14
constants, Ki, and boundary values of partial pressures and activities are defined in 60 20 20 2.5  10ÿ7 4.2  10ÿ14
text.
Fig. 2. External oxide scale formed on Fe–9Cr after (a) 120 h exposure to Ar–20CO2 Fig. 3. External oxide scale formed on Fe–20Cr after (a) 120 h exposure to Ar–
at 650 °C; (b) 30 h exposure to Ar–20CO2–20H2O at 800 °C. 20CO2 at 650 °C; (b) 100 h exposure to Ar–20CO2–20H2O at 800 °C.

the alloy (Fig. 6). Intragranular carbide penetration depths, XC, kp = (8 ± 2)  10ÿ11 cm2 sÿ1. Similarly, an average fV = 0.3 was mea-
were measured from the interface between internal oxidation sured in near surface regions after exposure to both gases.
and internal carburisation zones. In both dry and wet CO2, XC in-
creased according to a parabolic law (Fig. 7)
3.3. Internal reactions at 800 °C
X 2C ¼ 2kp t ð7Þ
At 800 °C, both Fe–2.25Cr and Fe–9Cr alloys underwent internal
where kp is the rate constant. Carburisation is seen to be faster in oxidation in addition to external scaling. No carburisation product
the dry gas (kp = 6.5  10ÿ10 cm2 sÿ1) than in the wet gas could be observed in the Fe–2.25Cr alloy, whereas reaction of the
(kp = 2.4  10ÿ10 cm2 sÿ1). Intergranular carbides were observed Fe–9Cr alloy resulted in formation of carbides and martensite be-
throughout the samples within a 40 h exposure to both atmo- neath the Fe-rich oxide scale (Fig. 11).
spheres. The volume fraction of carbides in the ferrite matrix, fV, In the dry gas, intergranular carburisation of Fe–9Cr was fast,
was measured in 50 lm wide zones parallel to the surface using and occurred throughout the alloy within a 5 h exposure. Martens-
the ImageJ software, with a precision of around 2%. Results plotted ite was identified in etched cross-sections, visually (Fig. 11a) and
against the relative depth are shown in Fig. 8. The profiles are sim- by means of hardness testing. Vickers microhardness tests with a
ilar for both gases, but the value at x = 0 is higher in the dry gas, 200 g load yielded HV values of about 100 and 400 for ferrite
with fV = 0.17 compared with 0.14 in the wet gas. and martensite, respectively. The extent of martensite formation
Reaction of Fe–20Cr also led to internal oxidation and carburisa- increased with increasing reaction time. Although the phase trans-
tion, but precipitate morphology and penetration depth varied formation affected entire grains, it did not occur in a uniform man-
with the nature of the overlaying oxide. Underneath the thin ner along the metal/oxide interface: some grains neighbouring the
Cr2O3 scale, internal precipitation was absent or limited to a martensite remained untransformed, while some isolated mar-
semi-connected array of intragranular carbides (Fig. 9a), with a rel- tensite was found deep inside the alloy, surrounded by untrans-
atively low penetration depth (620 lm). Where the alloy formed formed grains.
an iron-rich oxide scale, internal reaction occurred to a larger ex- In addition, internal carbides were occasionally observed in a
tent, producing both small, equiaxed and elongated carbides narrow 10 lm strip between the internal oxidation zone and the
(Fig. 9b). Both external scaling and internal precipitation varied martensite, as seen in Fig. 11. However, in other regions, no car-
considerably. Measurements of the extent of carburisation under- bides were detected between the martensite and internal oxida-
neath the iron-rich scale were subject to substantial error, as seen tion zone. Carbides were never detected in the absence of
from the error bars in the parabolic plots of Fig. 10. Carburisation martensite, i.e. at the internal oxide/ferrite interface. Foil speci-
rates in dry and wet CO2 were similar, and the kinetics are mens were prepared from the sample shown in Fig. 11, by FIB mill-
therefore described with a common, average rate constant ing. The TEM–EDS results in Fig. 12a and b reveal the matrix to be
 Fig. 5. Internal precipitation in Fe–9Cr after 120 h exposure to Ar–20CO2 at 650 °C,
etched with Murakami’s reagent. (a) OM; (b) SEM, backscattered electrons.

Fig. 4. Intergranular carburisation of Fe–2.25Cr after 120 h exposure to Ar–20CO2

at 650 °C, etched with Murakami’s reagent. (a) Scale/alloy interface; (b) deep within
the alloy.

depleted in chromium, and the carbides and oxides chromium-

rich. Selected area diffraction patterns (SADP) were recorded and
interpreted with the help of CaRIne Crystallography software
[15]. Patterns shown in Fig. 12c and d fit well with a-Fe and
(Cr,Fe)23C6, respectively. Fig. 13a and b show a bright field image
and a SADP recorded from the martensite visible in Fig. 11a. The
absence of cubic symmetry in the tetragonal structure allows the
appearance of the (221) and (212) reflections, which are forbidden
in ferrite.
In the wet gas, the Fe–9Cr alloy developed qualitatively similar
carburisation products (martensite and carbides).
The Fe–20Cr alloy was less prone to internal reaction. No oxide
and a very small number of elongated carbides (Fig. 14) were ob-
served beneath the Cr2O3 scale.

4. Discussion

4.1. Activity of carbon at the metal/oxide interface

As noted in the Introduction, several mechanisms have been put

forward to account for the carburisation of Fe–Cr alloys in low aC,
CO2-rich atmospheres. However, a quantitative description of the Fig. 6. XRD pattern recorded at about 40 lm below the metal/oxide interface of Fe–
carbon supersaturation developed in the alloy, relative to the gas, 9Cr reacted 240 h at 650 °C.
has been lacking.
Upon exposure to CO2–H2O mixtures, Fe–Cr alloys form oxide
scales, and the conditions established within the scales might dif- atmosphere, and (ii) thick enough to be considered in local ther-
fer markedly from those prevailing in the gas [1,2,10,11]. Indeed, modynamic equilibrium, the oxygen activity within the scale var-
provided that the oxide scale is (i) sufficiently compact and adher- ies between its ambient value at the oxide/gas interface, and the
ent to constitute a physical barrier separating the metal from the dissociation pressure of the oxide at the metal/oxide interface
Fig. 7. Carburisation kinetics of Fe–9Cr at 650 °C. Values obtained from average of
at least 10 measures, with minimum and maximum indicated by the error bars.

Fig. 9. Non-uniform carburisation pattern of Fe–20Cr after 120 h exposure to Ar–

20CO2 at 650 °C. Samples etched with Murakami’s reagent. (a) Thin Cr2O3 scale and
limited carburisation; (b) thick Fe-rich scale and extensive carburisation.

Fig. 8. Carbide volume fraction in Fe–9Cr exposed to (a) Ar–20CO2 and (b) Ar–
20CO2–20H2O at 650 °C. Values obtained from average of 4 measures at each depth,
with minimum and maximum indicated by the error bars.
Fig. 10. Carburisation kinetics of Fe–20Cr at 650 °C. Values obtained from average
of at least 10 measures, with minimum and maximum indicated by the error bars.

(see Fig. 1). If, in addition, the scale transmits the CO2 and CO spe-
cies, low values for pO2 result in high values for pCO =pCO2 (by virtue
of reaction (2)), and therefore for aC (via reaction (3)). The phase pCO =pCO2 ratio can be expressed as a function of the dissociation
constitution and composition of the growing scale control the pressure of the oxide, ðpO2 Þeq
establishment of chemical potential gradients and resulting driv-
pCO K2
ing forces, while its transport properties limit the availability of ¼ ð8Þ
the molecular species at the metal/oxide interface. pCO2 ðpO Þ1=2
2 eq

It is proposed that all gas species but Ar can reach the scale/al-
loy interface, where local thermodynamic equilibrium is assumed where K2 is the equilibrium constant for reaction (2). This ratio de-
to be achieved. Thus pO2 is controlled by the metal/oxide equilib- pends only on the nature of the oxide, alloy composition and
rium, and aC set by the equilibrium of reactions (2) and (3). The temperature.
 at the interface between such a mixture and the underlying alloy
is the dissociation pressure of FeO on pure iron. In fact, pO2 at the
triple point a-(Fe,Cr)/FeO/(Fe,Cr)3O4 is slightly lower than the dis-
sociation pressure of FeO on pure Fe, because dissolution of chro-
mium in FeO stabilises the latter. However, this effect is
neglected here and the pO2 associated with the equilibrium
1
ð1 ÿ dÞFe þ O2 ¼ Fe1ÿd O ð12Þ
2
is used to calculate the value of b for dilute alloys. Use of the equi-
librium relationship
1
K 12 ¼ ð13Þ
ðpO2 Þ1=2
eq

yields b = K2K12. The value of K12 is calculated for d = 0.053 from

data tabulated in Ref. [12].
High chromium alloys form Cr2O3 according to
2 1 1
Cr þ O2 ¼ Cr2 O3 ð14Þ
3 2 3
where Cr denotes chromium dissolved in the a matrix, with, at
equilibrium,
1
K 14 ¼ 2=3
ð15Þ
aCr ðpO2 Þ1=2
eq

In this case, b ¼ K 2 K 14 a2=3

Cr .
The validity of the proposed model is now tested against
experimental results. Attention is focused on the behaviour of
Fig. 11. Internal reaction products observed in Fe–9Cr after 20 h exposure to Ar– the Fe–9Cr alloy, for which the experimental data are most com-
20CO2 at 800 °C. Sample etched in picral + HCl. (a) Optical microscopy; (b) SEM. plete. Values of aC calculated from Eq. (11) and thermochemical
data tabulated in Ref. [12] are plotted in Fig. 15, along with a
predominance diagram for Fe–9Cr. Phase boundaries in the
The carbon activity can be determined from the Boudouard Fe–9Cr–C system were calculated with the Thermo–Calc software
reaction equilibrium, as and the TCFE4 database [18].
p2CO
aC ¼ K 3 ð9Þ 4.2. Carburisation at 650 °C
pCO2

Denoting by pT the total pressure at the interface, and omitting 4.2.1. Fe–9Cr
the small pO2 value, one obtains According to the proposed model, Eq. (11) predicts carbon
activities at the scale/alloy interface of oxidised alloys. However,
pT ¼ pCO2 þ pCO þ pH2 O þ pH2 ð10Þ
the Fe–9Cr alloy produces both internal oxides and internal car-
where again argon is assumed not to enter the scale. Eq. (9) can bides, and the carburisation zone is not in contact with the scale/
then be rewritten as alloy interface (Fig. 5). Thus, predicted and experimental results
apply to two distinct locations. In addition, the width of the car-
2
b pT burisation zone is, a priori, affected by the inward movement of
aC ¼ K 3 ð11Þ the oxidation front. Under certain conditions, the formalism devel-
ð1 þ bÞð1 þ abÞ
oped by Meijering [10,19] to describe the simultaneous internal at-
where b ¼ pCO =pCO2 , a ¼ pH2 þ pH2 O and b ¼ pCO þ pCO2 . tack of dilute alloys by two oxidants may be used to overcome
Describing the formation of carbon via the syngas reaction, (6), these difficulties.
yields the same expression. According to Eq. (11), aC is dependant The reaction morphology and schematic concentration profiles
upon two parameters: (i) pT, which is a function of both the exter- are depicted in Fig. 16. The two internal reaction zones are defined
nal pressure and the scale transport properties; and (ii) the by the thermodynamic stability of the precipitates and the concen-
ðpH2 þ pH2 O Þ=ðpCO þ pCO2 Þ ratio. The latter is determined by its value tration of the oxidants. Essentially, as carbon diffusion in the metal
in the reaction gas and the relative degree of uptake and transport is rapid, and carbon cannot react with the more stable chromium-
of H-bearing and C-bearing molecules by the scale. The value of pT rich oxide, the carbon concentration NC does not decrease signifi-
is set at 0.2 atm for Ar–20CO2 and 0.4 atm for Ar–20CO2–20H2O. In cantly across the oxide precipitation zone. Thus, the value of aC
the absence of any data, hydrogen and carbon penetration of the calculated for the scale/alloy interface can to a good approximation
scale might be considered equally probable. In this case, a/b = 1 be applied at the internal oxide/internal carbide interface. Further-
and the same aC value is arrived at for both gases. The wet gas case more, as the internal oxidation front advances, carbides are con-
is considered further when discussing carburisation of Fe–9Cr. verted into oxides, and the carbon thus released diffuses inward
Dilute Fe–Cr alloys reacted with CO2–H2O mixtures form multi- to react with chromium at the carburisation front. Consequently,
layered, Fe-rich oxide scales. The layer in contact with the metal if the conversion is complete, the carburisation depth XC can be
consists of an FeO + (Fe,Cr)3O4 mixture. Calculated ‘‘log pO2 -compo- defined from the interface between the internal oxidation and
sition’’ diagrams in the Fe–Cr–O system have been published for carburisation zones. In Fig. 12b, the TEM–EDS profile shows that
627 °C [16] and 900 °C [17], and indicate that the equilibrium po2 the internal oxidation zone contains no carbon, and that the two
Fig. 12. TEM analyses at the interface between internal oxidation and carburisation areas in Fe–9Cr after 20 h exposure to Ar–20CO2 at 800 °C. (a) Bright field image and
corresponding EDS elemental maps; (b) EDS linescan; (c) and (d) selected area diffraction pattern of the a matrix ([011] zone axis) and of a (Cr,Fe)23C6 carbide ([0 0 1] zone
axis), respectively. The arrow and star in (a) indicate the location of the linescan in (b) and of the pattern in (d), respectively.

precipitation zones are indeed distinct. That result was obtained Provided that local thermodynamic equilibrium is achieved
after reaction at 800 °C, but is assumed also to apply at 650 °C. throughout the carburisation zone, a diffusion path on the phase
The carbon activity predicted from equilibrium at the Fe–9Cr diagram can be used to define the compositions of coexisting
alloy/scale interface is seen in Fig. 15 to exceed the value required phases and their mass fractions, fm. Mass and volume fractions
to form M7C3 at 650 °C, thus qualitatively accounting for the are related through the densities of the phases (Table 2). Fig. 8
observed carburisation. The quantitative success of the model is shows that in both gases, the carbide volume fraction decreases
now tested by examining the amount of carbide formed. with increasing relative depth, according to a profile which is
Fig. 13. TEM analyses at the interface between internal carburisation area and
martensite in Fe–9Cr after 20 h exposure to Ar–20CO2 at 800 °C. (a) Bright field
image; (b) selected area diffraction pattern of martensite, showing characteristic
Fig. 16. Diffusion profiles during simultaneous reactions of oxygen and carbon with
(221) and (212) reflexions in the [322] zone axis.
a dilute Fe–Cr alloy [10].

Table 2
Density of a-Fe and Cr carbides.

q (g cmÿ3) References
a-Fe 7.87 [20]
Cr7C3 6.92 [21]
Cr23C6 6.97 [21]

independent of reaction time. The steady-state distribution of

phases within the precipitation zone supports the hypothesis of lo-
cal thermodynamic equilibrium. Accordingly, the Fe–Cr–C phase
diagram shown in Fig. 17 was constructed using Thermo–Calc. In
this isothermal section, the points M, A and B represent the compo-
sition of the global system, the carbide precipitates and the metal
matrix, respectively.
At 650 °C, the diffusion coefficients DC [22,23] and DCr [24] in
a-Fe are estimated as 2.4  10ÿ7 and 5.6  10ÿ14 cm2 sÿ1, respec-
tively. Since DC  DCr , chromium is assumed not to diffuse on a
macroscopic scale, and the diffusion path lies along the dotted line
Fig. 14. Internal precipitation in Fe–20Cr after 100 h exposure to Ar–20CO2 at in Fig. 17a. This path cuts tie-lines at points corresponding to dif-
800 °C, etched with Murakami’s reagent. ferent fm values. Expressed in terms of volume fraction, fV, these

Fig. 15. Interfacial carbon activity from Eq. (11) (solid line) and predominance diagram from data in Thermo–Calc for Fe–9Cr (dashed lines).
 ðsÞ
where N C is the surface concentration of carbon in the metal ma-
ð0Þ
trix, m the C/Cr ratio in the carbides, and N Cr the initial Cr mole frac-
tion in the alloy. Diffusional blocking by the small equiaxed
precipitates is ignored. In the presence of an external oxide scale,
ðsÞ
the surface concentration of carbon, N C , is replaced by the interfa-
ðiÞ ðiÞ ð0Þ
cial value, N C . For Eq. (16) to hold, (i) the condition NC DC  N Cr DCr
must be met, (ii) carbon must react with chromium but not iron,
and (iii) the carbides must be sufficiently stable for both chromium
and carbon concentrations in the metal matrix to be negligible
throughout the precipitation zone. These conditions are now exam-
ined in light of the volume fraction measurements presented in
Fig. 8.
Using an average value of 0.15 for fV at x = 0 and the phase dia-
ðiÞ
gram of Fig. 17 yields N C = 8.6  10ÿ5. The carbon permeability is
ðiÞ
then calculated from DC values in a-Fe [22,23] as N C DC = 4.0 
ð0Þ
10ÿ11 cm2 sÿ1. Standard data for DCr [24] lead to N Cr DCr = 5.4 
ðiÞ ð0Þ
10ÿ15 cm2 sÿ1, and the condition N C DC  N Cr DCr is satisfied.
Volume fraction profiles together with the phase diagram indi-
cate that both M7C3 carbides in the near-surface precipitation zone,
and M23C6 carbides deeper in the alloy, contain large amounts of
iron. Values of NC range from 1.3  10ÿ4 at x = 0 to 9.4  10ÿ6 at
x = XC, while NCr ranges from 6.6  10ÿ3 at x = 0 to 8.3  10ÿ2 at
x = XC. The latter values are not negligible. Thus the conditions of
pure chromium carbide formation and complete chromium precip-
itation are not met. Consequently Wagner’s simplified Eq. (16),
whilst providing good order of magnitude prediction of kp, lacks
the accuracy required for present purposes.
The present situation of partial precipitation is described by
Ohriner and Morral [26], applying the lever rule to the matrix-pre-
cipitate two-phase field to more accurately reflect the mass bal-
ance underlying Eq. (16). This procedure has been combined here
with a numerical treatment of varying precipitate composition (de-
ðiÞ
fined in the phase diagram from fV profiles), allowing N C DC to be
evaluated from kp. Independently determined values of DC
ðiÞ
[22,23] then allow calculation of N C . The corresponding carbon
activities are shown in Table 3, together with the values obtained
from volume fraction measurements and those calculated from
Eq. (11). Order of magnitude agreement between values obtained
from the carburisation rates and those calculated for scale-alloy
Fig. 17. Isothermal section of the Fe–Cr–C phase diagram at 650 °C. (a) Overall
view; (b) Fe-rich corner. Points A, B and M are defined in text.
equilibrium further supports the validity of the thermodynamic
model.
range from 0 to 0.07 in the a + M23C6 field and from 0.06 to 0.2 in Estimates of aC arrived at from measurements of carburisation
the a + M7C3 field. Experimental fV values, shown in Fig. 8, thus extent and rate are consistently lower than predicted from the lo-
indicate that the near interface region corresponds to M7C3 precip- cal equilibrium model (Table 3). This may represent a systematic
itation in ferrite, and the deeper region to a + M23C6. Measured vol- failure to image all carbide precipitates metallographically. The
ume fractions and XRD data (Fig. 6) are consistent and in discrepancy is even larger in the case of reaction with wet gas. This
agreement with the predicted reaction products. can be understood in terms of preferential uptake and/or transport
Using the lever rule, a tie-line in the a + M7C3 field correspond- of H-bearing molecules over C-bearing species. The extent of
ing to the volume fraction measured at the interface between hydrogen enrichment can be estimated from Eq. (11), assuming
internal oxidation and carburisation zones is selected. This defines again that pT = 0.4 atm. Comparison of the aC values determined
ðiÞ
the carbon content of the metal matrix N C . Thermodynamic for dry and wet gases (Table 3) yields ðpH2 þ pH2 O Þ=ðpCO þ pCO2 Þ
data for the dissolution of carbon in a-(Fe,Cr), obtained from = 6.2 and 3.1, for the kp and fV calculations, respectively. The con-
Thermo-Calc, are then used to calculate the local equilibrium clusion that selective uptake of H-bearing species decreases carbon
carbon activity. The amount of carbide formed during reaction penetration of an oxide is in agreement with earlier suggestions
with Ar–20CO2 is found in this way to correspond to aC = 0.43. In [1,9,27].
the case of reaction with Ar–20CO2–20H2O, the value aC = 0.21 is
obtained. These values are to be compared with that calculated 4.2.2. Fe–2.25Cr
from scale-alloy equilibrium (Eq. (11)) as 0.47. This good agree- Exposure of the Fe–2.25Cr alloy to dry CO2 at 650 °C led to
ment provides support for the validity of the proposed model. intergranular carburisation, but no intragranular carbides were ob-
Further confirmation is available from the carburisation rates. served (Fig. 4). Eq. (11) yields aC = 0.47, which exceeds the value of
The observed parabolic kinetics (Fig. 7) indicate diffusion control, 5.5  10ÿ2 required to form M7C3 carbides in a Fe–2.25Cr alloy. The
and the rate constant kp is related to carbon permeability according driving force for carbide precipitation in this alloy is less than in
to Wagner’s equation [25] Fe–9Cr, where a value of 1.2  10ÿ2 is sufficient to stabilise M7C3.
ðsÞ The relatively low carbon supersaturation with respect to carbide
N C DC
kp ¼ ð16Þ formation is evident in preferential precipitate nucleation at
mNð0Þ
Cr grain boundaries. The qualitative implication of intergranular
 Table 3
Carbon activity at the scale/alloy interface for the Fe–9Cr and Fe–20Cr alloys reacted at 650 °C.

Alloy Gas Experimental Thermodynamic

From kp From fV Equilibrium
Fe–9Cr Ar–20CO2 0.25 0.43 0.47
Ar–20CO2–20H2O 0.07 0.21
Fe–20Cr Both 0.01 0.1

carburisation is nonetheless clear. Carbon enrichment at the metal/ alloy equilibrium (Eq. (11)). This apparent conflict could be ration-
oxide interface relative to the gas has occurred, in agreement with alised by proposing that preferential oxidation of chromium had
the model. occurred locally. Not only would this produce chromium-depleted
metal capable of forming single-phase austenite when carburised,
4.2.3. Fe–20Cr but also a chromium-rich oxide at the alloy surface, with a conse-
The Fe–20Cr alloy produced a non-uniform reaction pattern at quently lower oxygen activity and elevated aC value.
650 °C, as illustrated by Fig. 3a. The cases of Cr2O3 scaling and of The very irregular pattern of subsurface carburisation evident in
iron-rich oxide growth are discussed separately. Fig. 11 is certainly consistent with local variations in the scale/alloy
Since the dissociation pressure of Cr2O3 is extremely low, a very interface conditions. However, the additional observation of iso-
high carbon activity is expected from thermodynamic equilibrium lated martensite within the alloy interior cannot be accounted
at the Cr2O3/alloy interface. Values calculated from Eq. (11) exceed for in this way. What is needed is a determination of chromium
104 in both gases. However, no graphite was observed, and carburi- and carbon concentration profiles throughout the reacted zones.
sation occurred at a relatively low rate (Fig. 9a). This indicates a Nonetheless, formation of martensite and carbides beneath the
low interfacial carbon activity. Thus while carburisation did indeed oxide scale clearly demonstrates that carbon supersaturation was
occur beneath the Cr2O3 scale, the predicted carbon activity was established at the metal/oxide interface, relative to the gas.
not achieved. The assumption that CO2 and CO access the metal/
oxide interface at a sufficient rate for reaction (2) to reach equilib- 4.3.2. Fe–2.25Cr
rium, and for Eq. (8) to be satisfied, is incorrect in this case. It is No carburisation product was observed in the Fe–2.25Cr alloy
therefore concluded that transport of carbonaceous species after reaction at 800 °C. This is in agreement with the proposed
through Cr2O3 is much slower than in iron oxide, as has already model, since Eq. (11) predicts a carbon activity of 3.4  10ÿ2, below
been suggested [4]. that required to form austenite (0.15) or M7C3 carbides (0.34).
In regions beneath an iron-rich oxide scale, carburisation rates
were similar in reactions in dry and wet CO2 (Fig. 10). Average val- 4.3.3. Fe–20Cr
ues of kp = 8  10ÿ11 cm2 sÿ1 and fV = 0.3 are used to calculate the As discussed for reactions at 650 °C, Eq. (11) results in very high
carbon activities shown in Table 3, using the procedure described carbon activities below Cr2O3 scales, provided that they transmit
for the Fe–9Cr alloy. The values are lower than those predicted the CO2 and CO species. The very occasional presence of internal
by Eq. (11). The low aC values corresponding to the amount and carbides after reaction at 800 °C suggests that the oxide was, in
rate of carburisation indicate that some degree of protection was general, impervious to CO2 and CO, but that microscopic failure oc-
obtained, even in locations where an iron-rich oxide scale had curred locally.
developed.
The value estimated via Eq. (16) from carburisation rates is 5. Summary and conclusions
likely to be in error. Non-uniform carburisation depths indicate
that the onset of carburisation, which would correspond to break- Binary Fe–Cr alloys exposed to dry and wet CO2 at 650 and
down of the protective scale, varied from place to place. Thus the 800 °C form external oxide scales. Despite the very low gas phase
times associated with measured precipitation depths would be carbon potentials, the alloys generally sustain carburisation attack
shorter than the exposure time, and carburisation rates underesti- beneath their oxide scales.
mated. Protection against carburisation of the Fe–20Cr alloy at Carbon activities required to form the observed reaction prod-
650 °C was partial, and only transient. ucts are calculated from thermodynamic data, and shown to ex-
ceed gas phase equilibrium values by factors of about 1012 to 1013.
4.3. Carburisation at 800 °C A local equilibrium model is used to relate carbon and oxygen
activities via the CO2 and H2O dissociation and Boudouard reac-
4.3.1. Fe–9Cr tions. Application of this description to the scale/alloy interface,
Carburisation of Fe–9Cr at 800 °C produced chromium-rich where the oxygen potential is low, leads to the prediction of carbon
M23C6 precipitates and martensite. Fig. 15 shows that the aC value activity values high enough to form chromium-rich carbides.
predicted from scale-alloy equilibrium is sufficient to form M23C6, Quantitative testing of the model was performed in the case of
but too low to reach the c + M7C3 phase field. Moreover, no single- Fe–9Cr reaction at 650 °C. Measured amounts of carbide precipi-
phase austenite field is available at this temperature, and the tated at the scale/alloy interface are shown to agree well with
observed formation of martensite cannot be accounted for on this the thermodynamically predicted value of aC = 0.47.
basis. Parabolic carbide precipitation kinetics indicate diffusion con-
At 800 °C, chromium diffusion is significant, and subsurface trol. Rate constants measured for Fe–9Cr at 650 °C provide an addi-
alloy depletion could result from selective oxidation. Reference to tional means of estimating interface carbon activities. The results
the Fe–Cr–C phase diagram at 800 °C (not shown) indicates that reinforce the conclusion that carbon activities are controlled by
single-phase austenite can form by carbon diffusion into chro- scale–alloy equilibrium.
mium-depleted alloy. However, the minimum value of aC required Carburisation of Fe–9Cr also occurs at 800 °C, producing mar-
to form austenite in chromium-depleted alloy is of order 0.1, which tensite and chromium-rich carbides. However, the reaction did
exceeds the maximum value of aC = 3.4  10ÿ2 calculated for FeO– not achieve a steady state condition in the times examined.
 Carbon activities in Fe–Cr alloys beneath external oxide scales [9] D.J. Young, Simultaneous oxidation and carburisation of chromia forming
alloys, Int. J. Hydrogen Energy 32 (2007) 3763–3769.
are elevated far above ambient gas phase values, provided that car-
[10] D.J. Young, High temperature oxidation and corrosion of metals, Elsevier,
bon can penetrate the scale. The mobile species has not been iden- Amsterdam, 2008.
tified, but is modelled successfully as CO–CO2. The presence of [11] N. Birks, in: Z.A. Foroulis, F.S. Pettit (Eds.), Corrosion of High Temperature
water vapour lowers the carbon activity, probably by partially Alloys in Multicomponent Oxidative Environments. Electrochem. Soc., Las
Vegas, 1976, pp. 215–260.
excluding the carbonaceous species from the scale. A chromia scale [12] I. Barin, Thermochemical Data of Pure Substances, second ed., VCH, Weinheim,
provides a more effective barrier to carbon than does an iron-rich 1993.
oxide. [13] M.R. Taylor, J.M. Calvert, D.G. Lees, D.B. Meadowcroft, The mechanism of
corrosion of Fe–9%Cr alloys in carbon dioxide, Oxid. Met. 14 (1980) 499–516.
[14] A.M. Pritchard, N.E.W. Hartley, J.F. Singleton, A.E. Truswell, Oxygen-18 and
deuterium profiling in thick films on Fe–9%Cr alloys by 3 MeV nuclear
Acknowledgements
microprobe, Corros. Sci. 20 (1980) 1–17.
[15] C. Boudias, D. Monceau, 1989–2010, <http://carine.crystallography.pagespro-
The authors acknowledge access to the UNSW node of the Aus- orange.fr/>, date accessed (2010).
tralian Microscopy and Microanalysis Research Facility (AMMRF), [16] H. Davies, A. Dinsdale, Theoretical study of steam grown oxides as a function of
temperature, pressure and p(O2), Mater. High Temp. 22 (2005) 15–25.
and thank Electron Microscope Unit staff for help with TEM analy- [17] A.D. Pelton, H. Schmalzried, J. Sticher, Computer-assisted analysis and
ses. Financial support from the Australian Research Council Discov- calculation of phase diagrams of the Fe–Cr–O, Fe–Ni–O and Cr–Ni–O
ery program is gratefully acknowledged. systems, J. Phys. Chem. Solids 40 (1979) 1103–1122.
[18] B. Sundman, B. Jansson, J.O. Andersson, The Thermo–Calc databank system,
CALPHAD: Comput. Coupling Phase Diagrams Thermochem. 9 (1985) 153–
190.
References
[19] J.L. Meijering, in: H. Herman (Ed.), Advances in Materials Research, Wiley-
Interscience, New York, 1971, pp. 1–81.
[1] J. Pirón Abellán, T. Olszewski, H.J. Penkalla, G.H. Meier, L. Singheiser, W.J. [20] Physical Constants of Inorganic Compounds, in: W.M. Haynes (ed.), CRC
Quadakkers, Scale formation mechanisms of martensitic steels in high CO2/ Handbook of Chemistry and Physics, 91st Edition, CRC Press/Taylor and
H2O-containing gases simulating oxyfuel environments, Mater. High Temp. 26 Francis, Boca Raton, 2010, pp. 4/43–4/101.
(2009) 63–72. [21] H.J. Goldschmidt, Interstitial Alloys, Butterworths, London, 1967.
[2] G.H. Meier, K. Jung, N. Mu, N. Yanar, F. Pettit, J. Pirón Abellán, T. Olszewski, L. [22] E. Budke, C. Herzig, H. Wever, Volume and grain-boundary diffusion of 14C in
Nieto Hierro, W.J. Quadakkers, G.R. Holcomb, Effect of alloy composition and a-Fe, Phys. Status Solidi A 127 (1991) 87–101.
exposure conditions on the selective oxidation behavior of ferritic Fe–Cr and [23] K. Tapasa, A.V. Barashev, D.J. Bacon, Y.N. Osetsky, Computer simulation of
Fe–Cr–X alloys, Oxid. Met. 74 (2010) 319–340. carbon diffusion and vacancy-carbon interaction in a-iron, Acta Mater. 55
[3] H.E. McCoy, Type 304 stainless steel vs flowing CO2 at atmospheric pressure (2007) 1–11.
and 1100–1800F, Corrosion 21 (1965) 84–94. [24] A.W. Bowen, G.M. Leak, Solute diffusion in alpha- and gamma-iron, Metal.
[4] C.S. Giggins, F.S. Pettit, Corrosion of metals and alloys in mixed gas Trans. 1 (1970) 1695–1700.
environments at elevated temperatures, Oxid. Met. 14 (1980) 363–413. [25] C. Wagner, Reaktionstypen bei der Oxydation von Legierungen, Zeit.
[5] C.T. Fujii, R.A. Meussner, Carburization of Fe–Cr alloys during oxidation in dry Elektrochem. 63 (1959) 772–790.
carbon dioxide, J. Electrochem. Soc. 114 (1967) 435–442. [26] E.K. Ohriner, J.E. Morral, Precipitate distribution in subscales, Scr. Metal. 13
[6] I. Wolf, H.J. Grabke, A study on the solubility and distribution of carbon in (1979) 7–10.
oxides, Solid State Commun. 54 (1985) 5–10. [27] D.J. Young, Effects of water vapour on the oxidation of chromia formers, in: P.
[7] X.G. Zheng, D.J. Young, High-temperature corrosion of Cr2O3-forming alloys in Steinmetz, I. Wright, A. Galerie, D. Monceau, S. Mathieu (Eds.), 7th
CO–CO2–N2 atmospheres, Oxid. Met. 42 (1994) 163–190. International Symposium on High Temperature Corrosion and Protection of
[8] X.G. Zheng, D.J. Young, High temperature corrosion of pure chromium in CO– Materials, vol. 595-598, Les Embiez, France, pp. 1189–1197.
CO2–SO2–N2 atmospheres, Corros. Sci. 36 (1994) 1999–2015.