Vous êtes sur la page 1sur 27

Enhanced oxidation of the 9%Cr steel P91

in water vapour containing environments


J. Ehlers
a
, D.J. Young
b
, E.J. Smaardijk
a
, A.K. Tyagi
c
,
H.J. Penkalla
a
, L. Singheiser
a
, W.J. Quadakkers
a,
*
a
Forschungszentrum Ju lich, Institute for Materials and Processes in Energy Systems, 52425 Ju lich, Germany
b
University of New South Wales, Sydney, Australia
c
Indira Gandhi Centre for Atomic Research, Kalpakkam, India
Received 11 February 2005; accepted 20 February 2006
Available online 18 April 2006
Abstract
The short term (~100 h) oxidation behaviour of the 9%Cr steel P91 was studied at 650 C in N
2

O
2
H
2
O gas mixtures containing a relatively low oxygen level of 1%. The oxidation kinetics were
measured thermogravimetrically and the oxide scale growth mechanisms were studied using
H
2
18
O-tracer with subsequent analyses of oxide scale composition and tracer distribution by
MCs
+
-SIMS depth proling. The corrosion products were additionally characterised by light optical
microscopy, SEM-EDX and XRD. It was found that the transition from protective, Cr-rich oxide
formation into non-protective mixed oxide scales is governed by the ratio H
2
O
(g)
/O
2
ratio rather
than the absolute level of H
2
O
(g)
. The results of the tracer studies in combination with the data
obtained from experiments involving in situ gas changes clearly illustrated that under the prevailing
conditions the penetration of water vapour molecules triggers the enhanced oxidation and sustains
the high growth rates of the poorly protective Fe-rich oxide scale formed in atmospheres with high
H
2
O
(g)
/O
2
ratios. The experimental observations can be explained if one assumes the scale growth to
be governed by a competitive adsorption of oxygen and water vapour molecules on external and
internal surfaces of the oxide scales in combination with the formation of a volatile Fe-hydroxide
during transient oxidation. The formation of the non-protective Fe-rich oxide scales is suppressed
in atmospheres with low H
2
O
(g)
/O
2
-ratios, and the healing of any such scale is promoted.
2006 Elsevier Ltd. All rights reserved.
0010-938X/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2006.02.002
*
Corresponding author. Tel.: +49 2461 614668; fax: +49 2461 613687.
E-mail address: j.quadakkers@fz-juelich.de (W.J. Quadakkers).
Corrosion Science 48 (2006) 34283454
www.elsevier.com/locate/corsci
Keywords: Steel; SIMS; TEM; Selective oxidation; Kinetic parameters
1. Introduction
Due to the demand for lower emissions from power generation systems, a number of
projects are being carried out world-wide to improve eciencies of conventional fossil
fuel-red power plants [1,2]. In coal red boilers, eciencies of around 45% can be
achieved if the steam parameters are increased to pressures of 300 bar and temperatures
of 600650 C [3]. At such high temperatures, the commonly used low alloy steels and
the higher corrosion resistant 12%Cr steels can no longer be used as construction materials
for live steam piping or blading materials in steam turbines, because of the lack of creep
resistance of these materials. Therefore, a number of modied 9%Cr steels, such as P91,
P92 and E911 were developed to full the new materials requirements in respect to creep
strength [3]. It has been shown that, in spite of the high temperatures, these steels also
show adequate oxidation resistance during operation in air [4]. However, it was found that
in simulated fossil fuel-red power plant combustion gases which contained oxygen in the
order of 1 vol%, the corrosion rates of these 9%Cr steels can be several orders of magni-
tude higher than in air [46]. Consequently, as thin walled components, they oer no
major benet over 12%Cr steels in spite of their signicantly higher creep strength [6].
The main reason for the high corrosion rates of the 9%Cr steels in the simulated com-
bustion gases was shown [4] to be the presence of water vapour (typically 715 vol%). This
detrimental eect of water vapour on the oxidation resistance of FeCr alloys has in fact
been known for many years, [79] and a number of mechanisms have been proposed by
several authors to explain the eect. A summary of the mechanisms proposed in the earlier
studies is given in Ref. [4].
In recent years, the development and construction of power generation systems with
increased steam parameters has led to a revival of the research on water vapour eects
in steel oxidation. The newer studies relate to exposures in mixtures of water vapour plus
oxygen (see e.g. [1017]) as well as to steam-containing environments to which no oxygen
is intentionally added (see e.g. [1822]). Newer studies on water vapour eects in case of
other types of metallic materials have been described, e.g. in [23,24].
In spite of these extensive investigations, the numerous experimental observations have
to date been only partly explained by the various mechanisms proposed. The diculties in
nding conclusive explanations are probably related to the fact that a number of dierent
steps in the oxidation process may be aected by presence of water vapour. The rate deter-
mining steps in the overall oxidation process may dier depending on the type of water
vapour containing gas, e.g. depending on the content of free oxygen in the environment.
Thus, the dominant mechanisms in oxygen/water vapour mixtures may dier from those
in steam or steam/inert gas mixtures.
In the present paper, the oxidation behaviour of the 9%Cr steel P91 was studied in
model N
2
O
2
H
2
O gas mixtures at 650 C. In most experiments the oxygen content was
1 vol%, i.e. a value similar to that present in the combustion gases mentioned above
[2,4,6]. Some new experimental procedures, including the use of H
2
18
O tracer, were used
to obtain better insight into the mechanisms of the enhanced oxidation of chromium steels
due to the presence of water vapour in frequently encountered service environments.
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3429
2. Experimental
The composition of the studied ferritic steel P91 is shown in Table 1. The two batch
compositions presented were used in various investigations conducted by the present
authors into the behaviour of ferritic 9%Cr steels in water vapour containing gases (see
e.g. [24,6]). Batch A in Table 1 was used in the earlier studies, whereas, for availability
reasons, batch B was used in the more recent studies. The two materials only dier in
respect to minor element concentrations in the steel. This results in slight variations in abso-
lute growth rate in the various environments, however, no fundamental dierences in
respect to conditions under which protective or breakaway type oxidation occurred, were
found. Most of the studies described in the present paper relate to the newer batch B. Only
few results (as will be indicated in the respective gures) relate to the earlier batch A.
Rectangular specimens, 20 10 2 mm in size were machined from the prevailing
pieces of the steel P91, and ground to a 1200 grit surface nish. For a number of short
term experiments, the specimens were subsequently polished with 1 lm diamond paste
to suppress the incubation period frequently encountered under the prevailing conditions,
as will be further explained in the text and the respective gure captions. Studies on
oxidation kinetics up to exposure times of 100 h were carried out at 650 C in
N
2
1 vol%O
2
x vol%H
2
O (x = 27) gas mixtures at a ow rate of 0.15 cm/s, using a
SETARAM thermobalance. To obtain more detailed information on the oxidation kinet-
ics, additional isothermal exposures were performed in an N
2
1 vol%
16
O
2
2 vol%H
2
18
O
gas mixture at a ow rate of 1 cm/s, for oxidation times ranging from 1 to 30 h. In these
tests an N
2
1%O
2
-mixture was bubbled through a glass container containing the H
2
18
O at
controlled temperature. It should be mentioned that actually the water used was not pure
H
2
18
O but contained a 50%H
2
18
O-enrichment. After oxidation, these specimens were ana-
lysed with respect to composition and oxygen isotope distribution in the scales by MCs
+
-
SIMS [25] using a CAMECA IMS 4F secondary ion mass spectrometer (SIMS). The depth
proles were quantied following the procedure described elsewhere [25] and re-calculated
to results which would have been obtained if pure H
2
18
O would have been used. The cor-
rosion products on all specimens were additionally characterised by optical microscopy,
Table 1
Composition of the studied steel batches of P91 in mass%
Element Batch A Batch B
Fe Base Base
C 0.10 0.10
Cr 8.1 8.6
Mo 0.92 0.93
Mn 0.46 0.41
Ni 0.33 0.26
V 0.18 0.21
Al 0.03 n.d.
P 0.02 n.d.
Si 0.38 0.36
S 0.002 n.d.
n.d.: not determined.
3430 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
scanning electron microscopy with energy dispersive X-ray analysis (SEM-EDX) and
X-ray diraction (XRD). For metallographic cross-section preparation, the oxidized speci-
mens were Ni-coated prior to mounting to protect the oxide scales during grinding/polish-
ing and to reveal a clearer contrast between oxide scale and mounting material. One of the
oxidized specimens was studied by transmission electron microscopy (TEM). The cross-
section of the oxide scale was made by the common sandwich preparation method and
subsequent ion milling (PIPS). The analysis was carried out using a Philips CM200-
FTEM.
3. Results
Fig. 1 illustrates, on the basis of earlier [24,6] and more recent studies [26] typical
examples of the scale formation on the steel P91 when exposed in N
2
1 vol%O
2
and
N
2
1 vol%O
2
with water vapour additions in the range 27 vol%. in the temperature range
600700 C for exposure times of approximately 100 h. After exposure in the dry N
2

1 vol%O
2
mixture the surface scale is extremely thin. Its morphology and composition
are very similar to those found after air oxidation [4]. The scales formed in wet gas consist
of four regions (Fig. 1b and c). In the outer part, a thin Fe
2
O
3
and a thicker Fe
3
O
4
layer
had been formed. The inner scale consists of an Fe
3
O
4
matrix with (Fe,Cr)
3
O
4
stringers.
These (Fe,Cr)
3
O
4
stringers mainly resulted from oxidation of the chromium-rich carbide
precipitates in the alloy [21]. Consequently, they show a morphology and distribution sim-
ilar to that of the alloy carbides. Near the oxide/metal interface an internal oxidation zone
exists which contained Cr-rich phases such as Cr
2
O
3
or Cr-rich (Fe,Cr)
3
O
4
, frequently in
combination with FeO. The latter phase was identied by XRD analysis of sequentially
ground samples, as will be shown later, and the Cr
2
O
3
by Raman spectroscopy. The scale
formed in the wet gas shows substantial porosity and locally, a degree of separation
between inner and outer layer is seen to have developed.
Fig. 2 shows the eect of water vapour on the isothermal oxidation kinetics of P91 in
N
2
1 vol%O
2
with and without additions of water vapour at 650 C. In the dry gas, the
alloy exhibits extremely low weight changes. It is evident that addition of water vapour
to the test gas results in an enhanced reaction rate. The latter is preceded by a short, appar-
ent incubation period in which the weight change rate is relatively small. The switch-over
to enhanced (breakaway) oxidation does, especially in case of ground specimen sur-
faces, not start at the same time over the whole specimen surface. It was found [2,26] that
nodules of rapidly growing oxides nucleated and spread over the surface after extended
exposure times, as has frequently been observed during long term exposures of similar
materials in wet gases (see e.g. [19]). The rates measured in wet gas during these relatively
short term exposures after the onset of breakaway (Fig. 2) therefore do not necessarily
reect the exact, quantitative eect of water vapour on the oxidation rates of the post-
breakaway scales, but rather the extent of nodule formation. The incubation period
can be substantially suppressed by diamond polishing the steel to a mirror-nish prior
to exposure [26] so that the change from protective to non-protective oxidation in wet
gases can be studied already in experiments with exposure times of only a few hours, as
already mentioned in Section 2.
The TEM cross-section in Fig. 3 shows that the slowly growing scale which forms dur-
ing exposure in the dry gas, consists of two layers. EDX analyses in combination with
results from glancing angle X-ray diraction strongly indicate that at the outer side
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3431
Fig. 1. Typical examples of metallographic cross-sections of P91 when exposed in the temperature range 600
700 C in N
2
1 vol%O
2
with and without water vapour additions of 27 vol%H
2
O for exposure times of
approximately 100 h: (a) dry gas, (b) wet gas and (c) schematic of typical scale formation in wet gas illustrating
nomenclature used in text.
3432 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
Fe
2
O
3
is present, whereas the inner scale consists of Cr-rich (Fe,Cr,Mn)
3
O
4
spinel. Small
voids are seen to have formed at the oxidemetal interface. These could have resulted from
condensation of inwardly diusing cation vacancies at the scale/alloy interface. The alloy
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0 5 10 15 20 25 30
Exposure time [h]
W
e
i
g
h
t

g
a
i
n

[
m
g

.

c
m
-
2
]
650 C;
N
2
- 1%O
2
- H
2
O
dry gas;
ground surface
2% H
2
O;
ground surface
4% H
2
O;
ground surface
4% H
2
O;
polished surface
Fig. 2. Weight changes during isothermal oxidation of the 9%Cr steel P91 in N
2
1 vol%O
2
x vol%H
2
O mixtures
at 650 C.
Fig. 3. TEM cross-section of P91 after 5 h oxidation in N
2
1 vol%O
2
at 650 C, (a,b) overview picture plus Cr-
distribution, (cf) larger magnication with corresponding element distributions of Cr, Mn and Fe, showing two-
layered structure and re-oxidation of metal surface in voids.
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3433
side of these voids formed new, protective, Cr-rich oxide. The element mapping in Fig. 3b
clearly shows Cr depletion in the sub-scale zone of the steel beneath the protective scale.
Fig. 4 shows tapered metallographic cross-sections of the surface scales formed in wet
gas (N
2
1%O
2
2% water vapour) after various oxidation times. In the early stages of oxi-
dation (1 h) the scale mainly consists of a relatively thick, outer Fe
2
O
3
layer and an inner
layer of Cr
2
O
3
-stringers embedded in FeO. The inner and outer layer are almost com-
pletely separated by a gap and, unlike the scale commonly found after long exposure times
[16], hardly any magnetite is present. The outer hematite exhibits a whisker type mor-
phology and is extremely poorly adherent to the substrate. Extension of the exposure time
(2 h) at rst mainly leads to a thickening of the inner sub-scale, in which hardly any mag-
netite is found. Relative to the overall scale thickness, the gap seems to be shifted outward.
Upon further exposure, the relative amount of magnetite strongly increases and the gap
becomes gradually lled with oxide. After 16 h, the overall scale morphology is very
Fig. 3 (continued)
3434 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
similar to that frequently described for specimens after long term exposures ([16] and
Fig. 1b).
Fig. 5 shows weight change data for the P91 steel (batch A) during isothermal oxidation
at 650 C in wet (N
2
1 vol%O
2
4 vol%H
2
O) and dry gas (N
2
1 vol%O
2
), with in situ
Fig. 4. Metallographic tapered cross-sections of oxide scales on P91 after oxidation in N
2
1 vol%O
2
2 vol%H
2
O
at 650 C: (a) 1 h, (b) 2 h, (c)7 h and (d) 16 h. Specimens were mirror-polished prior to oxidation.
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3435
switching from wet to dry gas and vice versa every 24 h, i.e. without intermediate cooling.
The gravimetric data were presented earlier in Ref. [11]. In the rst stage (wet gas) fast
Fig. 4 (continued)
80 90
0
1
2
3
4
5
6
7
8
0 10 20 30 40 50 60 70
Time [h]
wet wet dry dry
M
a
s
s

c
h
a
n
g
e

[
m
g
.
c
m
-
2
]

650C;
Isothermal
Fig. 5. Weight gain during isothermal oxidation of P91 (batch A in Table 1) at 650 C whereby the gas was in situ
changed from wet (N
2
1 vol%O
2
4 vol%H
2
O) to dry gas (N
2
1 vol%O
2
) and vice versa every 24 h.
3436 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
oxidation kinetics were observed, similar to those shown in Fig. 1. Switching to dry gas
after 24 h almost immediately decreased the oxidation rate. Switching back to the wet
gas after 48 h again led to an increase of the oxidation rate after a short incubation period.
Similar observations were made by Narita et al. [16] during in situ gas changes of an FeAl
model alloy at 800 C. In Fig. 5, the oxidation rate at the beginning of the second wet
Fig. 6. Metallographic cross-sections of oxide scales on P91 (batch A) after the various oxidation stages including
in situ gas changes between wet and dry gas indicated in Fig. 5: (a) 24 h, (b) 48 h (c) 72 h and (d) 96 h.
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3437
oxidation stage was lower than that found at the end of the rst oxidation stage. Switching
back to dry gas after 72 h resulted in a very low oxidation rate.
Fig. 6 shows cross-sections of the oxide scales formed after the four oxidation stages of
Fig. 5. After the rst wet gas stage, the oxide scale consists of three layers plus an inner
oxidation zone and is comparable to that shown in Figs. 1b and 4d. The Fe
3
O
4
in the outer
part of the scale exhibits substantial porosity. After subsequent oxidation in dry gas, this
porosity nearly completely vanished (Fig. 6b) and the Fe
2
O
3
layer increased in thickness.
Fig. 6 (continued)
3438 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
0.00
0.01
0.02
0.03
0.04
0 20 40 60 80 100
Time [h]
M
a
s
s

c
h
a
n
g
e

[
m
g
/
c
m
2
]

M
a
s
s

c
h
a
n
g
e

[
m
g
/
c
m
2
]

wet dry
0.00
0.01
0.02
0.04
10 20 30 40 50 60
Time [h]
Isothermal
Change to Wet Gas
Dry Gas
at 650 C
Cooling to
Room Temperature
in Wet Gas
Heating to 650 C
in Wet Gas
0.03
0
b
a
Fig. 7. Isothermal oxidation of P91 at 650 C whereby, after 24 h, the gas was switched from dry (N
2
1 vol%O
2
)
to wet gas (N
2
1 vol%O
2
4 vol%H
2
O): (a) without intermediate temperature change, (b) cooling to room
temperature during exposure in wet gas and (c,d) SEM pictures of oxide morphology after oxidation according to
conditions in (b).
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3439
In the third stage (wet gas), the Fe
2
O
3
was largely transformed into an Fe
3
O
4
layer, and a
large amount of porosity again appeared (Fig. 6c). Some remnants of Fe
2
O
3
are apparent,
both near the outer Fe
2
O
3
layer and near the inner Fe
3
O
4
/(Fe,Cr)
3
O
4
interface. After the
last stage in dry gas, practically the whole outer layer consisted of hematite and only small
fragments of Fe
3
O
4
are seen in the outer scale layer (Fig. 6d).
Fig. 7 shows weight change data measured during an in situ gas change, after the expo-
sure was started in dry gas. The data show that the protective oxide is up to the maximum
exposure time not destroyed if the dry gas is in situ switched to wet gas. However, if after
the change to wet gas, an intermediate cooling to room temperature is introduced, a tran-
sition to rapid oxidation occurs after an apparent, short incubation period. The external
appearance of the scale grown in this rapid reaction is shown in Fig. 7c and d.
To gain more insight into the roles of oxygen and water vapour species in the scale
growth processes of P91, a number of experiments were carried out for dierent oxidation
times in a N
2

16
O
2
H
2
18
O gas mixture (1 vol%
16
O
2
and 2 vol%H
2
18
O). Isotope distribu-
tions in the resulting surface scales were analysed by MCs
+
-SIMS (see Section 2 for details
on the test procedure and quantication method).
After a very short oxidation time of 1 h, the scale seems to consist of two regions
(Fig. 8a). Based on the results in Fig. 4a and XRD data, the inner Cr-containing part con-
sists of a Cr-rich scale. The outer scale consists of pure Fe-oxide in which no clear change
in oxygen/iron ratio is visible as a function of penetration depth. This can be explained if
one assumes the iron oxide to nearly exclusively consist of hematite (compare Fig. 4a). The
18
O/
16
O-ratio diers only slightly as a function of penetration depth, although the ratio is
slightly higher in the outer than in the inner part of the scale. Similar proles were found
after 2 and 4 h oxidation.
The SIMS depth proles after 7 h and 30 h oxidation in N
2

16
O
2
H
2
18
O are shown in
Fig. 8b and c. It is obvious that three layers of dierent compositions exist in the oxide
scale. Comparison of the SIMS depth proles with the metallographic cross-sections
(Fig. 4) and XRD data (Fig. 9) reveal that the outer scale consists of Fe
2
O
3
and Fe
3
O
4
,
the inner of Fe
3
O
4
+ (Fe,Cr)
3
O
4
whereas substantial amounts of FeO were found in the
zone near the scale/alloy interface. The Cr/Fe-ratio in the inner layer, consisting of
Fe
3
O
4
+ (Fe,Cr)
3
O
4
, equals approximately 1:4.
In all measured SIMS depth proles, the ratio
18
O/
16
O in the inner part of the scale is
smaller than in the outer scale. After 1 h and 7 h oxidation, the region in which the
18
O
concentration becomes higher than the
16
O concentration is located very near the interface
between outer Fe
3
O
4
- and inner Fe
3
O
4
+ (Fe,Cr)
3
O
4
layer (compare Figs. 4 and 8). After
30 h oxidation, the cross-over point occurs approximately in the middle of the outer
Fe
3
O
4
-layer. Comparison of these SIMS-data with the metallographic cross-sections in
Fig. 4 strongly indicate, that the area in which the concentration of
18
O becomes higher
than that of
16
O, coincides with the gap in the scale.
Fig. 10 shows the eect of water vapour and oxygen content on the oxidation behaviour
of the P91 steel at 650 C in N
2
O
2
H
2
O gas atmosphere. It is seen that in a gas with a low
oxygen content, very small amounts of water vapour are sucient to initiate the rapid
non-protective oxidation. The growth rate of the non-protective oxide scale appears to
be practically independent of the water vapour content. If the oxygen is increased to
20 vol%, non-protective oxidation does not occur even if the concentration of water
vapour is as high as 10 vol%. Only if the water vapour content is increased to very high
levels, is the protective oxide destroyed. This result explains why 9%Cr steels can exhibit
3440 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
0
20
40
60
80
100
0 0.05 0.1 0.15 0.2
Depth [m]
C
o
n
c
e
n
t
r
a
t
i
o
n

[
a
t
.
-
%
]
O
total
Fe
Cr
18
O
16
O
Fe
2
O
3
Alloy Cr-rich oxide
.1
0
20
40
60
80
100
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Depth [m]
C
o
n
c
e
n
t
r
a
t
i
o
n

[
a
t
.
-
%
]
O
total
Fe
Cr
18
O
16
O
Fe
2
O
3 Fe
3
O
4
+
(Fe, Cr)
3
O
4
Alloy Fe
3
O
4
)
0
20
40
60
80
100
0 2 4 6 8 10 12 14 16 18 20
Depth [m]
C
o
n
c
e
n
t
r
a
t
i
o
n

[
a
t
.
-
%
]
O
t ot al
Fe
Cr
16
O
18
O
Fe
2
O
3
Fe
3
O
4
+
(Fe, Cr)
3
O
4
Alloy Fe
3
O
4
+
)
a
b
c
Fig. 8. MCs
+
-SIMS depth proles of P91 after oxidation in N
2
1 vol%
16
O
2
2 vol%H
2
18
O at 650 C: (a) after 1 h,
(b) 7 h and (c) 30 h. Specimens were mirror-polished prior to oxidation.
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3441
very low oxidation rates, during exposure in laboratory air up to 10,000 h [4,10]. This very
protective behaviour occurs in spite of the fact that during such a long term exposure, the
20 25 30 35 40 45 50 55 60 65 70
2 theta [degrees]
I
n
t
e
n
s
i
t
y

[
a
r
b
i
t
r
a
r
y

u
n
i
t
s
]
Oxide surface
After 3
rd
grinding step
After 4
th
grinding step
Fe
2
O
3
FeO
Fe
3
O
4
Alloy
Fig. 9. XRD patterns of the oxide scale formed on P91 during oxidation for 30 h in N
2
1 vol%O
2
2 vol%H
2
O at
650 C. The specimen was mirror-polished prior to oxidation. The rst XRD spectrum was taken of the as-
oxidized specimen (in the gure indicated as oxide surface). Subsequently the specimen was ground in four
steps before reaching the metal surface. XRD spectra were take after each grinding step. Presented are the XRD
spectra after grinding steps 3 and 4. The phases which could be detected after steps 1 and 2 were similar to those
after step 3.
0
0.5
1
1.5
2
2.5
3
3.5
4
0 10 20 30 40 50 60
H
2
O concentration [%]
M
a
s
s

c
h
a
n
g
e

[
m
g
/
c
m
2
]
20% O
2
2% O
2
1% O
2
Fig. 10. Eect of O
2
and H
2
O content on the weight gain after 24 h oxidation of P91 at 650 C in N
2
O
2
H
2
O gas
mixtures. Specimens were mirror-polished prior to oxidation.
3442 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
water vapour content in the gas can, for certain time periods, be as high as around 2 vol%,
a value which was shown (see e.g. Figs. 1 and 10) to cause non-protective oxidation in low-
oxygen environments after only very short exposure times.
4. Discussion
4.1. General remarks
With its chromium level of 9%, the P91 steel has marginal ability to develop a protec-
tive, chromium-rich scale at the intermediate temperature of 650 C. When such a scale
does develop, it has been shown [4] to be capable of providing very long term protection.
If, on the other hand, an iron-rich scale forms, it grows rapidly, consuming the steel. It is
obviously of interest to identify and understand the processes determining which of these
outcomes is arrived at.
4.2. Growth rates and morphologies of protective and non-protective scales
As seen in Fig. 2, reaction of P91 with dry N
2
1 vol%O
2
was extremely slow. However,
addition of even small amounts of water vapour to the gas led to breakaway reaction,
i.e. to a rapid acceleration in rate after a period of slower reaction. As the value of p(H
2
O)
was increased, breakaway occurred at shorter times. Higher p(H
2
O) values also led to
some acceleration of the initial pre-breakaway rate.
The detrimental eects of water vapour on FeCr alloy oxidation have long been
known [79] but not adequately explained. The protective scale grown in N
2
1 vol%O
2
is seen in Fig. 3 to be very thin, and is similar [4] to those produced by oxidation in air.
This scale consists of an outer layer of Fe
2
O
3
and an inner layer of mixed chromium spinel,
(Fe,Mn,Cr)
3
O
4
. If the presence of manganese is ignored, and if phase equilibrium in the
FeCrO system can be approximated by the high temperature ternary isotherm in
Fig. 11 [27], then the phase assemblage present in the scale can be mapped onto the dia-
gram as the diusion path shown. This is seen to be consistent with local equilibrium, and
to imply substantial chromium depletion at the alloy surface.
Selective oxidation of chromium must, of course, lead to its preferential removal from
the alloy. However, the localisation of this eect to the alloy sub-surface region is a con-
sequence of the relatively low alloy diusion coecient, D
A
. This is conrmed by a value
for an average diusion coecient D
A
= 10
13
cm
2
s
1
which can be extrapolated from
high temperature data for FeCr alloys from Ref. [28]. A similar value can be derived from
more recent data (see Ref. [29] and compilation in Ref. [12]). It is clear that any disruption
of the scale would expose an alloy surface with a low chromium concentration [10]. Re-
growth of continuous, chromium-rich spinel would then be dicult, and iron-rich oxide
formation favoured.
Scale development during reaction in wet gas is illustrated by the series of cross-sections
in Fig. 4 (N
2
1 vol%O
2
2 vol%H
2
O) and the longer term oxidation product in Fig. 2. The
scales shown in Fig. 4a and b are seen by reference to Fig. 2 to correspond to pre-break-
away reaction, where the rate is only a little faster than in dry gas. The outer layers of these
scales are Fe
2
O
3
, just as in the dry gas, but now with an irregular, whisker-like surface.
Unlike the continuous chromium spinel formed in dry gas, the inner layers of these scales
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3443
consist of various phases such as FeO, Cr
2
O
3
, (Fe,Cr)
2
O
4
and possibly more stable oxides
of alloying elements such as Si [18].
The multi-phase mixtures e.g. of FeO + Cr
2
O
3
are, according to the phase diagram
thermodynamically unstable both at high temperature (Fig. 11), and at 650 C, because
the reaction
FeO(s) +Cr
2
O
3
(s) =FeCr
2
O
4
(s) (1)
is thermodynamically favoured. The formation of the two-phase reaction product reects
the low value of D
A
relative to the scaling rate, enabling the formation of FeO by limiting
the chromium availability. Its continued metastability reects the slow rate of reaction (1).
Scales shown in Fig. 4c and d reect various times of reaction after breakaway. In all
cases, large amounts of Fe
3
O
4
are present, along with chromium-rich spinel. The magne-
tite phase was not present in the protective scale grown in dry gas, or in the scales grown
in wet gas prior to breakaway. Its presence is characteristic of breakaway reaction, and
reects the failure of the slow diusing alloy to supply chromium to the scalealloy inter-
face. This interface advances rapidly into the alloy, oxidising the prior microstructue of
ferrite plus chromium-rich carbides, and reproducing it as the Fe
3
O
4
plus chromium-rich
spinel scale layer [21]. Thus the interface between Fe
3
O
4
and Fe
3
O
4
+ (Fe,Cr)
3
O
4
seen in
Figs. 1 and 4 represents the prior alloy surface location after oxidation. Additional oxide
is formed outside this interface as result of outward iron diusion through Fe
3
O
4
[21].
Fe 10 20 30 40 50 60 70 80 90 Cr
Cr (at %)
O
40
90
80
70
Cr O
2 3
50
20
10
30
O (at %)
+ Cr O
2 3
-Fe + Cr O
2 3
+
C
r
O
2
3
+
S
+
C
r
O
2
2
3
+
S
1
-
F
e
+
F
e
O
+
S
1
-
x
1
-Fe
+
Fe O
1-x
Fe O
2 3
Fe O
3 4
S = Fe Cr O
S = FeCr O
1 1.5 1.5 4
2 2 4
+
S
+
C
r
O
2
2
3
+
S
1
Fe O
1-x
Fe O + S
1-x 1

+

Fig. 11. Phase diagram FeCrO at 1200 C [27]; dotted lines showing diusion path.
3444 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
The remainder of the post-breakaway scale consists of an outer Fe
2
O
3
layer and an
innermost multi-phase sub-layer. The sequence of oxides from top to bottom of the scale
is consistent with their increasing thermodynamic stability and the expected decrease in
oxygen activity from the outside to the interior of the scale. Whilst the requirements of
local equilibrium between oxides and oxygen activity are satised in this sense, true local
equilibrium is not achieved. As already discussed phase mixtures such as e.g. FeO +
Cr
2
O
3
, or Fe
3
O
4
+ (Fe,Cr)
3
O
4
are metastable; at equilibrium the former would form a
spinel and the latter a single phase.
An important additional feature present in breakaway scales is the extensive porosity
evident in Fig. 4. Whilst ne pores are present in the inner Fe
3
O
4
+ (Fe,Cr)
3
O
4
layer
[21], large cavities exist in the Fe
3
O
4
layer, and these form a more or less continuous
gap at intermediate times. This gap is present also in pre-breakaway scales grown in
wet gas (Fig. 4a and b). The existence of the gap probably explains the formation of
the unusually thick Fe
2
O
3
layers developed in these experiments. Because solid-state dif-
fusion in Fe
2
O
3
is much slower than in the lower iron oxides [8] it usually develops as only
a very thin layer. However, once the gap develops and separates the outer scale from the
underlying material, the supply of iron by outward diusion ceases. Inward oxygen diu-
sion then leads to Fe
3
O
4
oxidation, resulting in Fe
2
O
3
layer thickening. Although the gap
blocks solid-state diusion, scale growth nonetheless continues, both above and below the
gap, as seen in Fig. 4. A similar phenomenon is seen on a much smaller scale in the pro-
tective oxide formed in dry gas (Fig. 3) where voids at the oxidealloy interface form oxide
on the metal surface.
It is clear that water vapour can prevent the formation of protective chromium-rich
oxide scale layers on the Fe9Cr steel. The eect increases in severity with increased
p(H
2
O), and is associated with the development of voids and/or a gap within the scale,
as well as the appearance of large amounts of Fe
3
O
4
. Continued reaction despite the pres-
ence of this gap must be supported by gaseous mass transfer, which is now considered.
4.3. Gas phase mass transfer within the scale
In the dry gas reaction, the only relevant vapour species within the oxide is O
2
(g). If we
assume that pO
2
at the interface between hematite and magnetite is set by the equilibrium
2Fe
3
O
4
+1/2O
2
=3Fe
2
O
3
(2)
and neglect the presence of chromium, then it is estimated from thermodynamic data [30]
that pO
2
equals approximately 10
13
atm at 650 C. At such a low pressure, the rate of
oxygen vapourisation through dissociation of Fe
2
O
3
can be estimated from the Hertz
Langmuir equation as [8]
k
i
=
a
i
p
i
(2pm
i
kT)
1=2
(3)
where p
i
is the vapour pressure and m
i
the mass of the evaporating molecules; a
i
is termed
the evaporation coecient. When p
i
is expressed in atmospheres and the evaporation rate
k
i
in g cm
2
s
1
, Eq. (3) takes the form
k
i
= 44 3a
i
p
i
(M=T)
1=2
(3a)
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3445
where M is the molecular weight of the evaporating molecules. If a
i
is set at unity, then a
ux of 3 10
14
mol cm
2
s
1
is calculated. Arrival of this ux at the underlying metal
surface can support oxide re-growth at a rate of 0.01 nm h
1
. This dissociation mecha-
nism [7,31,32] is thus seen to provide insucient mass transfer to account for the
re-grown oxide, observed in Fig. 3 to be approximately 5 nm thick in a protective scale.
It is therefore concluded that the oxygen activity must have risen to higher values as
the detached Fe
2
O
3
approached equilibrium with the ambient atmosphere.
Such an explanation is not available for mass transport across the gap formed in break-
away scales. As seen in Fig. 4, the gap in these scales is located within the Fe
3
O
4
layer. The
value of pO
2
within the gap must therefore be below the equilibrium value for Eq. (2), and
the oxygen ux due to the O
2
(g) species is much too low to account for the rapid growth of
post-breakaway scale beneath the gap. Furthermore, it obviously provides no mechanism
for iron transport to support continued growth of oxide outside the gap. It is therefore
concluded that additional transport mechanisms must be facilitated by the presence of
H
2
O [8].
If H
2
O molecules can enter the scale, they can provide a means of oxygen transport [7,9]
through the reaction
H
2
O=H
2
+1/2O
2
(4)
as illustrated in Fig. 12. If inward H
2
O transport is relatively fast, then the partial pressure
of H
2
O in the cavity will approach that of the external gas, in the present case approxi-
mately 10
2
atm. As discussed earlier, local equilibrium between gas and solid oxide
appears to be closely approached. For local oxygen potentials of 10
22
10
13
atm, corre-
sponding to the Fe
3
O
4
existence range at 650 C, it is calculated from the thermodynamics
of Eq. (4) [4,30] that p(H
2
) values lie in the range 5 10
7
to 10
2
atm. According to Eq.
(3), these hydrogen pressures could support oxygen transfer rates of 5 10
7
to
10
2
mol cm
2
s
1
. The breakaway oxidation rate in N
2
- 1 vol%O
2
2 vol%H
2
O shown
in Fig. 1 corresponds to 1 10
9
mol cm
2
s
1
of oxygen atoms. If approximately half
of this uptake occurs below the gap in the scale (Figs. 2 and 4), then the available gas
phase oxygen transport rate is more than enough to support it.
An alternative possibility in the presence of water vapour is the formation of volatile
(oxy) hydroxides:
FeO H
2
O = Fe(OH)
(g)
2
(5)
Fe
3
O
4
3H
2
O = 3Fe(OH)
(g)
2
1=2O
2
(6)
Fe
2
O
3
2H
2
O = 2Fe(OH)
(g)
2
1=2O
2
(7)
Cr
2
O
3
2H
2
O 3=2O
2
= 2CrO
2
(OH)
(g)
2
(8)
The formation of Fe(OH)
(g)
2
during pure iron oxidation was proposed by Surman and
Castle [33]. Volatile Fe-species were observed by Viefhaus [34] during in situ AES studies
on steam oxidation of 9Cr steel and by Jaron et al. [35] during high ow experiments with
Fe in steam. The eects of steam on formation of volatile Cr-oxy-hydroxides Eq. (8) are
well documented [3638] but they seem not to be directly relevant to the transport of iron
across the scale gap considered here. Astemann et al. [14,15] clearly showed, that in high-
oxygen/high-water vapour mixtures formation of volatile Cr-species can trigger break-
away type oxidation. However, comparing the dependence of the Cr-oxy-hydroxide
3446 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
partial pressure on pO
2
and pH
2
O
(g)
implied by Eq. (8) with the experimental data in
Fig. 10, leads to the conclusion that chromium volatilisation was not a critical factor in
the overall oxidation process studied here. As the process under examination involved
the growth of iron oxides, this is not surprising.
Assuming that H
2
O can enter the scale and that local oxide-gas equilibrium is achieved,
it is seen from Eqs. (6) and (7) that the lower oxygen potentials in the inner part of the
scale will produce higher Fe(OH)
(g)
2
partial pressures than in the outer scale regions.
The resulting gradient in pFe(OH)
(g)
2
leads to outward transport of the hydroxide. The
process is shown schematically in Fig. 13. The Fe(OH)
(g)
2
vapour species is unstable at
high-oxygen pressures and deposits as solid oxide, which is believed to be mainly
Fe
2
O
3
, but Fe
3
O
4
deposition should also be possible in the outer part of the magnetite
layers.
Thiele et al. [4] have extrapolated thermodynamic data from much higher temperatures
[30] to calculate Fe(OH)
(g)
2
partial pressures at 650 C. They found pFe(OH)
(g)
2
to be
approximately 10
11
atm at oxygen potentials in the Fe
3
O
4
stability range. These values
are too low to provide signicant mass transfer, but reliable thermodynamic data are
not available for this low temperature. Given that the oxide continues to thicken above
Fig. 12. Schematic illustration showing transport of water vapour molecules through the scale and oxygen
transfer across in-scale void via H
2
OH
2
bridge (based on Ref. [7]).
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3447
the gap in the scale (Fig. 4), despite the impossibility of solid-state diusion, it must be
concluded that vapour phase transport of iron is occurring. The hydroxide species pro-
vides the vehicle for this transport.
Two mechanisms for gas phase transport within the scale have been identied. Both
involve H
2
O
(g)
, and can only operate if that species can enter the scale. Relevant informa-
tion on the interaction between water vapour and the scale is provided by the cyclic expo-
sure experiments.
4.4. Alternating wet and dry oxidation
As seen in Fig. 5, switching from wet to dry gas after 24 h of breakaway oxidation led to
a rapid decrease in scaling rate. Comparison of scale cross-sections after these two stages
(Fig. 6a and b) reveals that the dry gas caused an increase in the amount of Fe
2
O
3
at the
expense of Fe
3
O
4
, densication of the oxide and elimination of the gap. It is clear that dur-
ing the second stage of this experiment, oxygen entered the scale interior where it con-
verted Fe
3
O
4
to Fe
2
O
3
. The volume expansion accompanying this transformation,
together with perhaps some additional oxide growth led to elimination of much of the pore
space. The weight of oxygen uptake measured during the second stage was about
0.6 mg cm
2
, corresponding to an Fe
2
O
3
thickness of about 4 lm, in reasonable agreement
with the metallographic evidence of Fig. 6.
For this to occur, the scale originally grown in wet gas (Fig. 6a) must have been perme-
able to gas species. It is therefore concluded that the outer Fe
2
O
3
layer, despite its compact
appearance, allowed inward gas species diusion. Since, nonetheless, a large gradient in
oxygen activity was sustained (as shown by the distribution of oxide phases), this diusion
process must have been much slower than gas phase transport. It is suggested that molec-
ular diusion along internal surfaces provided the transport mechanism.
Fig. 13. Schematic illustration showing proposed mechanism for transport of Fe from inner to outer part of the
scale via volatile specie Fe(OH)
(g)
2
.
3448 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
Elimination of the pore spaces indicates that H
2
O
(g)
was no longer present within the
scale. Either H
2
O
(g)
diused out into the dry atmosphere, or it was reduced by reaction
with the scale, and H
2
then diused out and/or partly into the metal as shown to occur
during oxidation of Fe in wet environments at high temperatures [7].
Subsequent re-introduction of wet gas led to some acceleration in rate (Fig. 5), but
much less than was observed in the rst stage. The acceleration seems to be preceded
by a short incubation period. During this third reaction stage, the scale (Fig. 6c) re-devel-
oped porosity and a gap in the interior. These changes were accompanied by reduction to
Fe
3
O
4
of part of the thick Fe
2
O
3
layer remaining from stage 2. These eects can be under-
stood if H
2
O
(g)
gained access to the scale interior, causing volatilisation via reactions (6)
and (7), and oxygen transfer via the process shown in Fig. 12. The observation of irregu-
larly distributed Fe
2
O
3
remnants in the outer Fe
3
O
4
layer is consistent with molecular gas
transport within the void space, rather than solid-state oxide ion diusion. Evidently, the
oxidation step in stage 2 did not completely densify the outer Fe
2
O
3
, as some gas access
was still possible. The nal oxidation (stage 4) in dry gas led again to conversion of
Fe
3
O
4
in the scale interior, densication and elimination of the gap, by the same processes
as occurred in stage 2.
Commencing the experiment in dry instead of wet gas led to very dierent results, as
shown in Fig. 7a. The protective scale grown in dry gas was up to the test times used,
not aected by subsequent exposure to H
2
O
(g)
, and retained its protective character. If,
however, the scale was cooled and reheated, the coecient of thermal expansion dierence
between scale and metal led to scale damage and subsequent rapid reaction in wet gas
(Fig. 7b and c). Clearly the Fe
2
O
3
grown during isothermal exposure in dry gas is not sub-
sequently permeable to H
2
O
(g)
, at least up to the maximum exposure times employed.
Unlike the scale grown in wet gas, the oxide grown in dry gas appears to be fully dense
as long as no scale damage, e.g. by thermal cycling, is introduced. A similar conclusion
was drawn by Schu tze et al. [10] who found breakaway of the initially formed protective
layer on P91 to occur during prolonged exposure in air with large amounts of water
vapour. From their acoustic emission analyses, the authors concluded the occurrence of
scale damage to be related to growth stresses or thermally induced stresses caused by ther-
mal cycling.
A nal illustration of the gas permeability of breakaway scales is provided by Fig. 14,
taken from Ref. [26]. It is known that long term oxidation in Ar50 vol%H
2
O leads to a
thick, porous scale, similar to that grown in N
2
1 vol%O
2
4 vol%H
2
O (Fig. 1) except that,
as long as the scale is suciently dense and no barrier layers (e.g. of Cr- and/or Si-rich
oxides) are being formed, an outer Fe
2
O
3
layer is not present [21,26]. A two-stage reaction
involving exposure rst to Ar50 vol%H
2
O and subsequently to air (Fig. 14) led in the sec-
ond stage to oxidation of most of the outer Fe
3
O
4
layer but not to a change of the inner
Fe
3
O
4
+ (Fe,Mn,Cr)
3
O
4
layer. It is clear that molecular oxygen penetrated the outer layer,
but not the inner layer. This was presumably a result of their dierent porosities: large and
connected in the outer layer; small and isolated in the inner layer [21,22], apart from some
larger voids near the scale/alloy interface. The observation of isolated Fe
3
O
4
islands
remaining in the oxidized outer layer is also consistent with molecular transport rather
than solid-state diusion.
Because no H
2
O
(g)
was present in the second stage of this reaction, no mechanism for
volatilisation was available, and the overall scale growth was low. Because the rate of
scalealloy interface movement was therefore low, diusion in the alloy had time in which
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3449
to deliver chromium to this interface, where a chromium-rich oxide layer formed
(Fig. 14b), i.e. a zone with internal oxidation of chromia was no longer present.
It is clear from these experiments that the gas permeability of an iron-rich oxide scale
depends on the gas in which it is grown. When oxygen is the sole oxidant, the scale is dense
and virtually impermeable, and any prior porous oxide becomes densied. When oxygen is
in the presence of water vapour, this desirable result is not arrived at. Conversely, water
vapour, either alone or in the presence of oxygen produces a gas-permeable scale. It is
therefore concluded that the volatilisation processes possible in the presence of H
2
O
(g)
are responsible for creating the scale defects which permit molecular gas transport. How-
ever, this description does not explain why the oxygen species present in a mixed gas does
not penetrate the scale and lead to oxidation and densication of the scale interior. The
competition between oxygen uptake from O
2
and H
2
O
(g)
is discussed later.
4.5. Distribution of oxygen in scales
It has been deduced that during breakaway oxidation, much of the reaction is due to
penetration of the scale by H
2
O
(g)
which facilitates vapourisation processes. On this basis,
it would be expected that oxygen deriving from H
2
O
(g)
would be distributed dierently
from that coming from O
2
. Fig. 8 shows the distributions of oxygen within the scale, where
18
O derives from H
2
O
(g)
and
16
O from molecular oxygen. The phases marked on these pro-
les were identied from the total concentrations of Fe, Cr and O. It is seen that
16
O was
always more abundant than
18
O in the inner part of the scale. In the outer part of the scale,
the two species were present at approximately equal concentrations in a pre-breakaway
scale (Fig. 8a) but
18
O was enriched in this region after breakaway (Fig. 8b and c). This
Fig. 14. Metallographic cross-sections of 9%Cr steel showing oxide scales after oxidation at 650 C: (a) 1000 h
oxidation in Ar50 vol%H
2
O and (b) 250 h oxidation in Ar50 vol%H
2
O and subsequent oxidation in air for
750 h.
3450 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
nding supports the earlier conclusion that the H
2
O
(g)
species participates in breakaway
oxidation.
The oxygen distributions developed in breakaway scaling indicate that the O
2
species
penetrates the outer scale region without being completely consumed, and reacts preferen-
tially with the inner layer. They also indicate that the H
2
O
(g)
species does react in the outer
region, relatively little of it reaching the inner scale.
A redistribution process could also occur via Fe(OH)
(g)
2
volatilisation in the scale inte-
rior and re-deposition at higher pO
2
values toward the scale surface. If the initial transient
scale is assumed, for the sake of argument, to consist of
16
O and
18
O in a 1:1 ratio, then the
volatilisation reaction (6) with H
2
18
O
(g)
produces Fe(OH)
(g)
2
and O
2
each with an
16
O to
18
O ratio of 2:5 if mixing is random. Re-deposition through reaction (7) with
16
O
2
pro-
duces Fe
2
O
3
with an
16
O to
18
O ratio of 3:4. Thus enrichment of
18
O in the outer layer
is achieved. This isotopic transfer would not continue indenitely, because the H
2
O
(g)
pro-
duced in reaction (7) is enriched in
16
O. To the extent that H
2
O
(g)
is recycled within the
scale, rather than being replaced by H
2
18
O
(g)
from the external gas, the isotopic distribu-
tion would approach a steady state.
The oxygen distribution experiments conrm that when water vapour is present in suf-
cient quantity, oxygen is incorporated into the scale interior, not merely at the scale sur-
face, consistent with the inward diusion of molecular species. They also conrm that
oxygen in the presence of water vapour does not react with (and thereby densify) the outer
scale. Instead, reaction with H
2
O
(g)
is favoured in this region. The relative contributions of
reaction with the two oxidant species is now considered further.
4.6. Eect of oxygen partial pressure on water vapour eects
Fig. 10 shows weight uptake after 24 h oxidation at 650 C in gases with dierent oxy-
gen and water vapour partial pressures. It is seen that breakaway rates are not very sen-
sitive to pH
2
O
(g)
, but the value of pH
2
O
(g)
at which breakaway initiates increases with
increasing pO
2
. To a rst approximation, the data in Fig. 10 can be summarised by the
condition for breakaway oxidation for the exposure times used in the present study:
pH
2
O
pO
2
P1 (9)
Hayashi and Narita [17] also proposed that the change from slow to rapid oxidation of
Fe5%Al alloys at 800 C depended on the H
2
O
(g)
/O
2
-ratio. Explanations for this nding
are sought on the basis that the condition for breakaway is that H
2
O enters the scale and
facilitates gaseous mass transfer.
Formation of volatile Fe(OH)
(g)
2
is dependant on both pH
2
O
(g)
and pO
2
. In the scale
interior, where reaction (6) is in eect, local equilibrium leads to
pFe(OH)
2
= K
1=3
6
pH
2
O (pO
2
)
1=6
(10)
Alternatively, reaction with Fe
2
O
3
through Eq. (7) leads to
pFe(OH)
2
= K
1=2
7
pH
2
O (pO
2
)
1=4
(11)
Obviously neither of Eqs. (10) and (11) explains the condition of Eq. (9).
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3451
Considering now the entry of the molecular species into the scale, we write the surface
adsorption equilibria
H
2
O(g) S = H
2
O=S (12)
O
2
(g) S = O
2
=S (13)
where S represents a vacant surface site, and H
2
O/S, O
2
/S represent adsorbed species. Dis-
sociative equilibria are ignored in light of the fact that isotopic mixing between H
2
18
O
(g)
and
16
O
2
is extremely slow at these low temperatures. Assuming that at any instant during
scale development, surface sites are conserved, then
M = [S[ [O
2
=S[ [H
2
O=S[ (14)
Here, M is a constant and square brackets denote area concentrations. Eliminating [S]
between Eqs. (12)(14), one nds
[H
2
O=S[ =
MK
12
pH
2
O
1 K
12
pH
2
O K
13
pO
2
(15)
[O
2
=S[ =
MK
13
pO
2
1 K
12
pH
2
O K
13
pO
2
(16)
and it follows immediately that
[H
2
O=S[
[O
2
=S[
=
K
12
pH
2
O
K
13
pO
2
(17)
This competitive adsorption process provides an explanation for the observation that
higher pO
2
values require higher pH
2
O
(g)
values to initiate breakaway oxidation and the
condition of Eq. (9).
When Eq. (9) is satised, it is likely that K
12
pH
2
O > K
13
pO
2
, reecting the preferred
adsorption of the polar H
2
O molecule. Then Eq. (15) can be approximated as
[H
2
O=S[ ~
MK
12
pH
2
O
1 K
12
pH
2
O
(18)
If, in addition, K
12
pH
2
O > 1, then the surface would saturate with adsorbed water and
the rate of its inward diusion and participation in internal mass transfer processes would
be independent of pH
2
O
(g)
. This would explain the relative insensitivity of breakaway rates
to pH
2
O
(g)
(Fig. 10).
The competitive adsorption process is also consistent with the isotope distribution
experiments (Fig. 8), which showed that in the breakaway regime, oxygen from water
vapour was the major species incorporated into the outer scale and molecular oxygen
the major species taken up by the inner scale. The preferential adsorption of H
2
O
(g)
in
the outer part of the scale largely excludes the O
2
species from the surface and thereby
reduces its uptake. Only deep within the scale, beyond the part at which most of the
H
2
O
(g)
has been consumed, is O
2
an eective reactant. Finally, the competitive adsorption
process explains the ability of scales formed in breakaway-inducing atmospheres to resist
densication and retain their gas-permeability. Adsorbed H
2
O excludes O
2
from the inter-
nal surfaces of the outer scale region, whilst itself reacting only relatively slowly. Only
when H
2
O
(g)
is removed from the gas phase, can O
2
gain access to these surfaces. Finally,
the adsorption model is consistent with the nding that dense, protective scales grown in
3452 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454
dry oxygen are not subsequently permeated by H
2
O
(g)
. In the absence of internal surfaces,
adsorption and penetration of molecular H
2
O
(g)
is clearly impossible.
5. Summary and conclusions
The presence of water vapour in oxygen-bearing gas mixtures at 650 C has been shown
to provide conditions for breakaway oxidation of P91 steel. Breakaway was found to be
associated with the formation of large amounts of porous Fe
3
O
4
, and the development
of an essentially continuous gap in the scale. Under there conditions, rapid scale growth
combined with slow chromium diusion in the alloy led to loss of local equilibrium
between the solid phases of the reacting system. However, gassolid local equilibrium
was probably still closely approached.
It is concluded that entry of molecular H
2
O
(g)
into the scale interior was the critical pro-
cess leading to breakaway. The H
2
O can create porosity by vapourising the Fe(OH)
(g)
2
spe-
cies and re-depositing in parts of the scale where higher pO
(g)
2
values exist. The H
2
O
(g)
species also facilitates oxygen transfer within the scale through operation of the H
2
O/
H
2
couple.
The conclusion that H
2
O
(g)
entry into the scale interior occurs is supported by the
ndings that scales grown in wet gas develop and maintain gas permeability, that this per-
meability is not developed in dry gas, and that the permeability of a wet gas grown (break-
away) scale can be sealed by subsequent reaction in dry oxygen. The conclusion is
conrmed by the nding that oxidation in a mixture of
16
O
2
and H
2
18
O
(g)
leads to a
non-random distribution of isotopes within the scale, consistent with gas entry.
A competitive adsorption process in which strongly adsorbed H
2
O
(g)
largely excludes
O
2
from internal surfaces is shown to account for the development in wet gas of gas-per-
meable scales and its resistance to densication by reaction with O
2
. This process also
accounts for the observations that, in the time-frame examined, a critical condition for
breakaway is pH
2
O/pO
2
> 1 and that breakaway rates are relatively insensitive to pH
2
O
(g)
.
Acknowledgements
The authors are grateful to their colleagues Mr. Olefs, Mr. Lersch, Mr. Gutzeit and Mr.
Wessel for their assistance in carrying out the oxidation studies, XRD analyses, optical
metallography and SEM/EDX. Prof. B. Gleeson is gratefully acknowledged for his stim-
ulating discussions in the interpretation of the experimental data. They also acknowledge
the European Commission and Siemens Power Generation who partially funded this
project.
References
[1] K.H. Mayer, W. Bendick, R.U. Husemann, T. Kern, R.B. Scarlin, VGB Power Tech 1 (1998) 22.
[2] M. Thiele, W.J. Quadakkers, F. Schubert, H. Nickel, Report Forschungszentrum Ju lich, Ju l-3712, Ju lich,
FRG, 1999, ISSN 0944-2952.
[3] P.J. Ennis, in: R. Viswanathan, W.T. Bakker, L.D. Parker (Eds.), Advances in Material Technology for
Fossil Power Plants, Institute of Materials, London, Book no. 0770, 2001, pp. 187196, ISBN 1-86125-145-9.
[4] M. Thiele, H. Teichmann, W. Schwarz, W.J. Quadakkers, VGB Kraftwerkstechnik 77 (1997) 135;
M. Thiele, H. Teichmann, W. Schwarz, W.J. Quadakkers, VGB Kraftwerkstechnik 2/97 (1997) 129.
J. Ehlers et al. / Corrosion Science 48 (2006) 34283454 3453
[5] K. Zabelt, B. Melzer, A. Reuter, in: Conference Korrosion in Kraftwerken, Wu rzburg, FRG, 2930
September 1999, 11. VDI-Jahrestagung Schadensanalyse, VDI Verlag, Du sseldorf, 1999, pp. 99111.
[6] W.J. Quadakkers, M. Thiele, P.J. Ennis, H. Teichmann, W. Schwarz, in: EUROCORR 97, Trondheim,
Norway, 2225 September 1997, Proceedings, European Federation of Corrosion, vol. II, pp. 3540.
[7] A. Rahmel, J. Tobolski, Corrosion Science 5 (1965) 333.
[8] P. Kofstad, High Temperature Corrosion, Elsevier Applied Science, London, 1988.
[9] C.T. Fujii, R.A. Meussner, Journal of the Electrochemical Society 110 (12) (1963) 11951204.
[10] M. Schu tze, D. Renusch, M. Schorr, Corrosion Engineering, Science and Technology 39 (2004) 157166.
[11] C. Williams, M. Thiele, W.J. Quadakkers, in: EUROCORR 96, 2426 September 1996, Nice, F.,
Proceedings, vol. III, pp. 10/110/4.
[12] R. Peraldi, B.A. Pint, Oxidation of Metals 61 (56) (2004) 463483.
[13] K. Nakagawa, Y. Matsunaga, T. Yanagisawa, Materials at High Temperature 18 (1) (2001) 5156.
[14] H. Asteman, J.-E. Svensson, L.-G. Johansson, M. Norell, Oxidation of Metals 52 (12) (1999) 95111.
[15] H. Asteman, K. Segerdahl, J.-E. Svensson, L.-G. Johansson, Material Science Forum 369372 (2001) 277
286.
[16] M. Fukumoto, S. Maeda, S. Hayashi, T. Narita, Oxidation of Metals 55 (56) (2001) 401421.
[17] S. Hayashi, T. Narita, Oxidation of Metals 56 (34) (2001) 251270.
[18] M. Ueda, Y. Oyama, K. Kawamura, T. Maruyama, Materials at High Temperature 22 (12) (2005) 7985.
[19] I.G. Wright, B.A. Pint, in: Proceedings, NACE CORROSION/2002, Denver, CO, USA, April 811, 2002;
paper no. 02377.
[20] A. Agu ero, R. Muelas, Materials Science Forum 461464 (2004) 957964.
[21] R.J. Ehlers, P.J. Ennis, L. Singheiser, W.J. Quadakkers, T. Link, in: M. Schu tze, W.J. Quadakkers, J.
Nicholls (Eds.), European Federation of Corrosion Monograph, no. 34, The Institute of Materials, London,
2001, pp. 178193, ISSN 1354-5116.
[22] J. Zurek, E. Wessel, L. Niewolak, F. Schmitz, T.K. Kern, L. Singheiser, W.J. Quadakkers, Corrosion Science
46 (2004) 23012317.
[23] M. Hansel, W.J. Quadakkers, D.J. Young, Oxidation of Metals 59 (34) (2003) 285301.
[24] A. Galerie, Y. Wouters, M. Caillet, Materials Science Forum 369372 (2001) 231238.
[25] J.-P. Pfeifer, H. Holzbrecher, W.J. Quadakkers, W. Speier, Fresenius Journal for Analytical Chemistry 346
(1993) 186191.
[26] J. Ehlers, W.J. Quadakkers, Report Forschungszentrum Ju lich, Ju lich, FRG, Ju l-3883, June 2001, ISSN
0944-2952.
[27] N. Birks, G.H. Meier, Introduction to High Temperature Oxidation of Metals, Edward Arnold, London,
UK, 1983.
[28] W. Bowen, G.M. Leak, Metallurgical Transactions 1 (1970) 2767.
[29] C. Piehl, Zs. To kei, H. Grabke, Materials at High Temperature 17 (2) (2000) 243246.
[30] Program Factsage, Data bank SGTE, GTT Technologies, Herzogenrath, FRG.
[31] A. Rahmel, W. Schwenk, Korrosion und Korrosionsschutz von Stahlen, Verlag Chemie, Weinheim, New
York, 1977.
[32] S. Mrowec, in: Proc. 3rd JIM Int. Sympos. High Temperature Mater. Chem. 99-38, 2000, p. 115.
[33] P.L. Surman, J.E. Castle, Corrosion Science 9 (1969) 77.
[34] H. Viefhaus, Max Planck Institute, Du sseldorf, FRG, 2001, unpublished results.
[35] A. Jaron, Z. Zurek, A. Stawiarski, J. Zurek, M. Homa, Ochrona przed Korozja 11 s/A/, 2003, pp. 97101.
[36] D.L. Douglass, P. Kofstad, A. Rahmel, G.C. Wood, Oxidation of Metals 45 (1996) 529.
[37] K. Hilpert, D. Das, M. Miller, D.H. Peck, R. Wei, Journal of Electrochemical Society 143 (1996) 3642.
[38] B.B. Ebbinghaus, Combustion and Flame 93 (1993) 119.
3454 J. Ehlers et al. / Corrosion Science 48 (2006) 34283454

Vous aimerez peut-être aussi