You are on page 1of 11

Corrosion Science 52 (2010) 33943404

Contents lists available at ScienceDirect

Corrosion Science
journal homepage: www.elsevier.com/locate/corsci

High temperature oxidation of FeAl and FeCrAl alloys: The role of Cr


as a chemically active element
E. Airiskallio a,b, E. Nurmi a,b,c,*, M.H. Heinonen a,b, I.J. Vyrynen a,b, K. Kokko a,b, M. Ropo d,e,
M.P.J. Punkkinen a,b,f, H. Pitknen g, M. Alatalo g, J. Kollr h, B. Johansson f,i, L. Vitos f,h,i
a
Department of Physics and Astronomy, University of Turku, FI-20014 Turku, Finland
b
Turku University Centre for Materials and Surfaces (MatSurf), Turku, Finland
c
Graduate School of Materials Research, Turku, Finland
d
Department of Information Technology, bo Akademi, FI-20500 Turku, Finland
e
Fritz-Haber-Institut der Max-Planck-Gesellschaft, D-14195 Berlin, Germany
f
Applied Materials Physics, Department of Materials Science and Engineering, Royal Institute of Technology, SE-10044 Stockholm, Sweden
g
Department of Mathematics and Physics, Lappeenranta University of Technology, P.O. Box 20, FI-53851 Lappeenranta, Finland
h
Research Institute for Solid State Physics and Optics, P.O. Box 49, Budapest H-1525, Hungary
i
Department of Physics and Materials Science, Uppsala University, SE-75121 Uppsala, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: Good high-temperature corrosion resistance of FeAl alloys in oxidizing environments is due to the a-
Received 20 April 2010 Al2O3 lm which is formed on the surface provided temperature is above 900 C and the Al-content of
Accepted 18 June 2010 the alloy exceeds the critical value. Ab initio calculations combined with experiments on Fe13Al, Fe
Available online 30 June 2010
18Al, Fe23Al and Fe10Cr10Al alloys show that the benecial effect of Cr on the oxidation resistance
is signicantly related to bulk effects. The comparison of experimental and calculated results indicates a
Keywords: clear correlation between the FeCr chemical potential difference and the formation of the protective
A. Alloy
oxide scales.
A. Steel
B. AES
2010 Elsevier Ltd. All rights reserved.
C. Oxidation
C. Passive lms

1. Introduction tom up picture of the physical phenomena responsible for the


excellent high temperature oxidation resistance of FeCrAl alloys.
FeCrAl alloys are often used as high-temperature corrosion The corrosion resistance of FeCrAl in oxidizing atmosphere is
resistant materials [13] due to their ability to form a highly stable due to the formation of a highly protective chromium and alumi-
and protective oxide scale on the open surface when exposed to num oxide layer on the surface, which effectively separates the
oxidizing environment. The physical properties and corrosion oxidizing atmosphere from the pure alloy. Because a-Al2O3 is very
resistance of FeCrAl as a function of the chemical composition stable at high temperatures it would be benecial to maximize its
of the alloy has been studied quite extensively, see e.g. [420]. content at the surface. Unfortunately, for most of the Fe-alloy
Although the overall effects and phenomena of the oxidation and applications the most straightforward procedure of increasing
corrosion of metals has been well characterized [21,22], the atomic the Al-content in bulk is not an acceptable solution. This is due
processes at successive time steps of the oxidation or even the to the fact that increasing Al content makes FeAl alloys brittle,
chemical composition of the surface prior to the oxidation are less which poses a natural upper bound for the Al-content in these al-
well studied. The purpose of our investigation is to extend the loys regarding to most of the applications. [23] Fortunately, there
understanding of the FeCrAl surfaces towards the atomic scale. exist additional alloying elements that can boost the formation of
The knowledge obtained can be used to shed more light on the the Al-oxide scale on the surface up to such a level that the Al-con-
state of the surface prior to oxidation. This will help us build a bot- tent in bulk can be kept within the acceptable limits regarding to
the mechanical properties of the alloy. One of the most common
additional elements for this purpose is Cr. This phenomenon, called
the third element effect (TEE), is still considered a phenomenon
* Corresponding author at: Department of Physics and Astronomy, University of
Turku, FI-20014 Turku, Finland. Tel.: +358 2 33 5661; fax: +358 2 333 5070. without generally accepted explanation [20,2428]. In the present
E-mail address: eknurm@utu. (E. Nurmi). investigation we concentrate on the surface phenomena closely

0010-938X/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2010.06.019
E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404 3395

related to TEE. To give an extensive account on the energetics of


the pure FeCrAl surfaces at the atomic resolution we have calcu-
lated the differences of the atomic chemical potentials of the com-
ponents in the bulk and at the (0 0 1) surface of FeCrAl.
The excellent corrosion resistance of FeCrAl in oxidizing envi-
ronments is based on the rapid formation of the oxide surface scale
of the right type. Iron oxide on the surface is quite vulnerable to
corrosion in most cases. Chromium oxide provides very good cor-
rosion protection, but only at low and intermediate temperatures.
At low temperatures Cr improves the oxidation resistance of
FeCrAl due to the fast formation of Cr2O3 compared to that of
a-Al2O3 [29,30]. Since at high temperature Cr2O3 transforms to
volatile compounds, e.g. CrO3 [31], the corrosion protection at ele-
vated temperatures requires the more stable aluminum oxide
scales on the surface.
Depending on the concentrations of the alloys, details of the oxi-
dizing environment, as well as criteria used for the oxidation resis-
tance, the literature references show marked variance in the upper
limits of the temperature allowing the protective oxide scales of
Fig. 1. The equilibrium WignerSeitz radius (rWS in units of Bohr radius) of FeCr
Cr2O3 and Al2O3 to form and sustain. The most often reported values
Al alloys as a function of Cr concentration with Al-content as a parameter. For bcc
for the upper limits of the feasible temperature range are 900 alloys the lattice parameter (a) and the WS radius are related as a = (8p/3)1/3rWS.
1100 C for Cr2O3 and 12001400 C for Al2O3 [3,7,25,32,33,31,3436].
Depending on the composition of the FeCrAl ternary alloy dif-
ferent types of oxide scales are formed on the open surface when within the scalar-relativistic and soft-core approximations. The
exposed to the oxidizing environments. Because the right compo- EMTO Greens function was calculated self-consistently for 32 com-
sition of the alloy is one of the key prerequisites for a good corro- plex energy points distributed exponentially on a semi-circular
sion protection, we have calculated the chemical potential contour, which included states within 1 Ry (2  1018 J) below
differences of FeCrAl within a wide range of concentrations, the Fermi level. In the one-center expansion of the full charge den-
Cr: 025 at% and Al: 020 at%. Detailed understanding of the equi- sity, we adopted an l-cutoff of 8 and the total energy was calculated
librium surface composition requires knowledge of various proper- using the full charge density technique [43,40]. For each alloy the
ties of a material, e.g. vacancy formation, migration, diffusion, calculated equilibrium volume was used. The optimized Wigner
lattice relaxation, thermodynamics, kinetics, etc. From this diverse Seitz radii are shown in Fig. 1. The convergence of the total energy
eld of phenomena we concentrate here on the energetic driving with respect to the number of k-vectors was tested. It was found
force for the surface segregation and its consequences on the for- that 506 (for bulk) and 231 (for surfaces) uniformly distributed k-
mation of the protective oxide scale on FeCrAl alloys. vectors in the irreducible part of the 3D and 2D Brillouin zones were
To investigate experimentally the high temperature oxidation enough for the present purposes. The alloys were described as sub-
resistance of FeCrAl alloys as a function of Cr- and Al-contents stitutionally disordered ferromagnetic bcc alloys [49,50].
four different alloys were prepared by induction melting. The al- Since in the present investigation we map a large concentration
loys prepared, Fe13Al, Fe18Al, Fe23Al, and Fe10Cr10Al, region of FeCrAl with Cr- and Al-concentrations approaching
(at% used in alloy formulas) cover the interesting concentration zero the conventional supercell method would require enormously
range including the Al threshold corresponding to the onset of large and numerous supercells. Here, we resolve this difculty by
the high temperature oxidation resistance and the effect of Cr to employing the Coherent Potential Approximation (CPA) [51,52].
turn a non-resistant FeAl alloy to a high temperature oxidation Within the CPA, the alloy components are embedded in an effective
resistant alloy by enhancing the Al-oxide scale. The samples were medium, which is constructed in such a way that it represents, on
oxidized at 1000 C and the surfaces were investigated by Auger the average, the scattering properties of the alloy. In this way, the
Electron Spectroscopy (AES), Atomic Force Microscopy (AFM), original alloy problem reduces to the Schrdinger equation for the
and Optical Microscopy. The experimental data are analysed in effective medium plus the real space Dyson equations written for
the light of computational data to deepen the understanding of each single impurity. In the present application, we adopt the
the atomic view of the oxidation of FeCrAl. CPA implemented within the frameworks of the EMTO method
[39,40]. The EMTO approach has been applied successfully in the
theoretical study of various structural and electronic properties
2. Computational methods of alloys and compounds [40], demonstrating the accuracy and
efciency needed for the present investigation.
The calculations are based on the density functional theory The semi-innite bulk surface system is modelled by a slab con-
[37,38] and performed using the Exact Mufn-Tin Orbitals (EMTO) sisting of 8 atomic layers parallel to the surface. To retain the peri-
method [39,40]. The EMTO method is an improved screened Kor- odicity of the model system an innite array of the slabs separated
ringaKohnRostoker method [41], where the one-electron poten- by vacuum layers is constructed. The thickness of the vacuum lay-
tial is represented by large overlapping mufn-tin potential ers is equivalent to 4 atomic layers. The differences of chemical
spheres. By using overlapping spheres, one describes the crystal potentials of a ternary ABC alloy are calculated as
potential more accurately, compared with the conventional non- 
1 dU 
overlapping mufn-tin approach [4245]. lA  lB xC constant; 1
N dxA 
For the exchange-correlation density functional the generalized-
gradient approximation [46] was used. We have tested earlier that where lA, lB, U, N, xA, and xC are the chemical potentials of A and B
this choice leads to the best overall equilibrium volume for the components, the internal energy of the system, the number of
component metals of FeCrAl [47,48]. The EMTO basis set in- atoms in the system, and the atomic fractions of the components
cluded s, p, d, and f orbitals. The one-electron equations were solved A and C, respectively. One should note that since we require the
3396 E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404

Table 1 Auger line energies used for the concentration determination


Calculated concentration prole of the surface at different temperatures. were: O 510 eV, Cr 530 eV, Fe 705 eV and Al 1395 eV. For alumi-
T (K) Fe5Al Fe5Cr5Al Fe10Cr5Al Fe15Cr5Al num, high energy line was used as there is no interference with
(at%) Fe Al Fe Cr Al Fe Cr Al Fe Cr Al other lines and also the effect of the chemical environment on
the peak shape is smaller. Sensitivity factors for the metallic com-
0 13 87 1 0 99 0 0 100 0 0 100
100 16 84 6 0 94 0 0 100 0 0 100
ponents were measured from pure Cr, Fe and Al samples. Oxygen
200 19 81 9 0 91 0 0 100 1 0 99 sensitivity factor was determined from an Al2O3 layer [29]. Calibra-
300 22 78 12 0 88 1 0 99 3 0 97 tion samples were rst sputter cleaned and then measured using
400 25 75 15 0 85 4 0 96 6 0 94 the same experimental parameters. For aluminum oxide, the sput-
500 27 73 18 0 82 7 0 93 9 0 91
tered surface concentration was supposed to be the same as in bulk
600 30 70 21 0 79 10 0 90 12 0 88
900 37 63 28 0 72 19 0 81 19 0 81 [56].
1200 43 57 34 0 66 27 1 72 26 0 74
1500 49 51 39 0 61 33 1 66 32 0 68
4. Results

4.1. Theoretical surface proles and chemical potentials of alloys


total number of atoms of the system considered to be conserved the
atomic fractions are related as xA + xB + xC = 1. The chemical poten- The calculated surface proles (Table 1) can be compared with
tial differences are calculated for the bulk and for the surface. The the LEED measurements of the top-layer concentration of
surface concentrations for the selected alloys as a function of tem- Fe1xAlx(1 0 0) [5759]. In these measurements the surfaces were
perature T (Table 1) were obtained by minimizing the grand poten- annealed not beyond about 1200 K before quenching. For the Fe
tial (X) of the system. 5Al alloy the surface Al-concentration according to the abovemen-
X tioned LEED measurements is about 57 at%. Our surface prole at
X U  TS  li Ni ; 2
i
1200 K is in good agreement with these experimental results.
The difference of the chemical potentials of Fe and Cr (lFe  lCr)
where S is the entropy of the system, and Ni and li are the total as well as that of Fe and Al (lFe  lAl) were calculated using Eq. (1)
number and the chemical potential of particles of type i. In our cal- and they are shown in Figs. 2 and 3. The bulk data was calculated
culations only the congurational part was included in the entropy. with a dense concentration mesh (xCr, xAl) but since the computa-
Details of the calculational method are published elsewhere [53 tional cost increases considerably for surface calculations the sur-
55]. While the surface proles shown in Table 1 are calculated by face mesh is sparser. Due to the smaller number of surface
varying the surface concentrations the rest of the surface calcula- calculations (16 calculations for (lFe  lCr)surface and 15 calcula-
tions in the present paper refer to homogeneous bulksurface tions for (lFe  lAl)surface) the difference of the surface chemical
systems. potentials was rst tted to a second order two-dimensional poly-
nomial. This polynomial with xed Al-concentrations (1, 5, and
10 at% Al) is shown as surface data in Figs. 2 and 3. Since at this
3. Experimental procedure
point our model system used for FeCrAl alloys is a homogeneous
slab the surface and bulk concentrations in Figs. 2 and 3 are the
The alloys were prepared by induction melting under argon.
same.
Weighted amounts of pure metals (99.99%) were melted in an alu-
Comparison of Figs. 2 and 3 shows that for lFe  lCr the bulk
mina crucible. Nominal compositions were Fe13Al, Fe18Al, Fe
and surface curves intersect, as in the FeCr case [55], whereas
23Al and Fe10Cr10Al in atomic %. Ingots were slowly cooled to
for lFe  l Al the surface curves are far above the bulk curves. Espe-
room temperature. Atomic % checked with EDX are shown in Table
cially the (lFe  lCr)bulk curves have a steep slope within the range
2. Slides of about 3 mm thick were cut from the ingots. Samples
were grounded with abrasive paper down to 1000 grit to atten
and remove the deformed layer of the surface. Final polishing
45
was done with 6 and 1 lm diamond suspensions. Samples were bulk 1% Al
then cleaned with acetone and methanol in ultrasonic bath, and 40
5% Al
dried with air ow. Oxidation at 1000 C was performed under 35 10% Al
oxygen ow in a quartz tube. Heating of the sample was done by
Fe- Cr [mRy]

30 surface
induction and temperature measured with infrared thermometer.
Reaching 1000 C took about 40 s. A PHI 610 spectrometer was 25
used for depth proling. Base pressure during the measurements
was 1  109 Torr. Electron gun was run at 5 keV and 80 nA. Sput- 20
tering was done with 3 keV argon ions at an angle of 40 deg and 15
rastered over 4  4 mm area. The sputtering rate was calibrated
10
using a 100 nm thick Ta2O5 oxide layer as a reference sample.
The estimated etching rate was 4 nm/min. Intensities were deter- 5
mined from the peak-to-peak heights of the differentiated spectra.
0
0 5 10 15 20 25
Table 2 Cr (at. %)
Concentrations of the alloys determined by EDX.
Fig. 2. Bulk and surface chemical potential differences (lFe  lCr) of FeCrAl, (Al
Alloy Fe (at%) Al (at%) Cr (at%) percentages for surface potentials from top to bottom: 1 at%, 5 at% and 10 at%).
Fe13Al 87.4 12.6 Surface data is taken from two-dimensional polynomial t and the calculated bulk
Fe18Al 82.1 17.9 values (shown by symbols) are connected by spline curves. The surface has the
Fe23Al 77.4 22.6 same composition as the bulk. The zero level of lFe  lCr corresponds to
Fe10Cr10Al 79.9 9.8 10.3 443.865 Ry in absolute scale. Atomic chemical potentials are presented, hence
atomic Rydberg units (Ry = 2.17987  1018 J) are used for convenience.
E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404 3397

70

60
Fe- Al [mRy]

50 bulk 1% Al
5% Al
40 10% Al
surface
30

20

10

0 5 10 15 20 25
Cr (at. %)

Fig. 3. Bulk and surface chemical potential differences (lFe  lAl) of FeCrAl, (Al
percentages for surface potentials from top to bottom: 1 at%, 5 at% and 10 at%). Fig. 4. AES surface depth proles of Fe18Al oxidized at 1000 C and 1 atm for
Surface data is taken from two-dimensional polynomial t and the calculated bulk 5 min.
values (shown by symbols) are connected by spline curves. The surface has the
same composition as the bulk. The zero level of lFe  lAl corresponds to
2059.943 Ry in absolute scale (Ry = 2.17987  1018 J).

of 010 at% Cr. This can be related to the transition from the
Cr-miscible to the Cr-immiscible region in the FeCr bulk phase
diagram [55]. It is also interesting to note that adding Al to the
alloy shifts the chemical potential difference curves down in Figs. 2
and 3. This can be related to the effect of increasing the volume of
the alloy with increasing Al-content (Fig. 1). Similar effect has been
obtained also in FeCrV alloys. [60] The trends of the bulk chem-
ical potentials in Figs. 2 and 3 are also consistent with the experi-
mentally observed Al partitioning in FeCrAl alloys which shows
the depletion of Al from the Cr-rich phases [6163].

4.2. Experimental Auger proles of oxidized surfaces

The prepared ingots of Fe13Al, Fe18Al, Fe23Al, and Fe


10Cr10Al were exposed to oxidizing atmosphere and the result-
Fig. 5. AES surface depth proles of Fe23Al oxidized at 1000 C and 1 atm for
ing oxide scales were investigated by repeated Ar sputtering and
5 min.
Auger measurements. Our Auger measurements of the surface
depth proles of oxidized Fe13Al, Fe18Al, Fe23Al, and Fe
10Cr10Al show that Fe13Al does not form the protective oxide
scale. Instead, the surface was almost entirely covered with a
rough and thick Fe-oxide layer without any traces of Al. In striking
contrast to this the other three alloys form thin protective oxide
layers consisting mainly of Al-oxides. Our experimental AES-re-
sults are shown in Figs. 46.
As shown in the gures the amount of Al (Fe) increases (de-
creases) in the oxide scale in the sequence of Fe18Al, Fe23Al,
and Fe10Cr10Al. This is also the sequence of the improved oxi-
dation resistance, since the Al-oxide gives better protection against
oxidation at high temperature compared with the Cr- and Fe-
oxides.

5. Discussion

5.1. Scaling losses

To take a full advantage of our electronic structure calculations


Fig. 6. AES surface depth proles of Fe10Cr10Al oxidized at 1000 C and 1 atm
in the quest for veiled atomic phenomena related to the oxidation for 5 min.
of FeCrAl we rst survey the published data focusing on the for-
mation of oxide scales on the FeCrAl surfaces. The rates of the
scaling loss (weight lost by oxidation) of FeCrAl alloys at alloying components (Cr and Al) affect the formation of the oxide
1200 C, reviewed in Ref. [4], are rst converted into a two-dimen- lms. One should note that the experimental data that we have
sional plot (Fig. 7). This representation shows in parallel how both used in our tting procedure, is limited to the region where the
3398 E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404

contours at levels: 0.5, 1, 2, 5, 10, 50 40


40 200 threshold
35
180
35 30
160
140 25

Cr (at. %)
30 120 20
100 15
25 80
10
Cr (at. %)

60
20 5
40
20 0
0 5 10 15 20 25
15
0 Al (at. %)

Fig. 8. FeCrAl alloys rated as good (lled symbols) or poor (open symbols) with
10
respect to the high temperature (P1000 C) oxidation resistance. The data is from
Ref. [4] (triangles pointing downward) and from Ref. [20] (triangles pointing
upward). The dividing threshold curve between the poor and good oxidation
5 resistance domains is the 5 g/(m2h) scaling loss curve adapted from Fig. 7.

0 centrations of the alloys. For instance, if the 5 g/(m2h) contour from


0 5 10 15 20
Fig. 7 is replotted in Fig. 8 it approximately splits the alloys in two
Al (at. %) groups possessing good or poor high temperature oxidation resis-
Fig. 7. Experimental scaling losses (g/m2h) of FeCrAl alloys at 1200 C after 240 h
tance. According to the review article of Thomaszewicz and Wall-
oxidation. The colour coding is from 0 to 200 g/m2h. To illustrate the groups of work [4] the Al-concentration in bulk FeAl alloy should be at least
alloys having similar oxidation rates ve equiscaling contours are shown in the plot. 1015 at% in order to the oxidation resistance to be effective at
The plot is obtained by tting to the data shown in Ref. [4]. high temperatures. They also report that adding 10 at% Cr into
FeAl allows one to reduce the Al-content down to 3 at% without
oxidation rate is 10 g/(m2h) or less. Therefore, the data shown in weakening the corrosion resistance of the alloy. The abovemen-
the lower left corner of Fig. 7 is a result of extrapolation. However, tioned 5 g/(m2h) contour starts from the Fe15Al and goes to about
one should note that at high temperatures (1000 C) J 10 at% Al 4 at% Al level when there is 10 at% Cr in FeCrAl. Therefore, the
(p. 84 in [4]) or J 14 at% Cr (p. 126 in [2]) are reported to give suf- 5 g/(m2h) oxidation rate curve seems to give an approximate lower
cient protection against corrosion in FeAl and FeCr, respec- limit for the combined Al and Cr concentrations in high-tempera-
tively. It is interesting to note that for these two threshold values ture corrosion resistant FeCrAl alloys. Correspondingly, we sug-
our tting function in Fig. 7 gives scaling losses approximately of gest that for FeCrAl alloys to have good high temperature
the same order in magnitudes. This clearly suggests that the tting oxidation resistance their concentrations should be chosen accord-
procedure works well and the obtained function can also be used ing to the following relation, corresponding to the dividing thresh-
within extrapolated regions. old curve in Fig. 8,
As Fig. 7 shows, at 1200 C in most cases Al is about 1.52 times
as effective oxidation retardant as Cr if scaling losses in FeAl and 150
cAl > 4 ; 4
FeCr are compared. Nevertheless, considering FeCrAl one can 8 cCr
dene a line in Fig. 7 where Al and Cr are equally effective, i.e.
the reduction in scaling loss is the same for the same amount of where concentrations are given in at%.
added solute Al or Cr. This dividing line is obtained by joining Additionally one should note that the oxidation resistance of
the points where the normal of an equiscaling contour is at FeAl and FeCrAl alloys seems to correlate with the Al-concen-
45 deg to the horizontal direction. This happens approximately tration in the surface layer. According to Fig. 8 the Fe10Cr5Al
for the concentrations (cCr, cAl) satisfying and Fe15Al alloys have approximately similar oxidation resis-
tances. Table 1 shows that for the Fe10Cr5Al the calculated Al
cCr  0:7cAl  2; 3 concentration in the surface at 1200 K is 72 at% and the LEED mea-
surements [59] give about 74 at% Al for the top layer of Fe15Al,
where at% are assumed for the concentrations. For alloys above this
annealed not beyond 1200 K and quenched.
line in Fig. 7 adding Al while keeping Cr content xed reduces scal-
ing losses more than doing vice versa (adding Cr with Al-content
xed). Below this line the effectivity of Al and Cr is reversed. 5.3. Oxide type and chemical potentials

5.2. Protective oxide scales We are now at the point when we can compare the calculated
chemical potentials (Figs. 2 and 3) with experiments. For that
The reported high temperature (P1000 C) data concerning the purpose we present the difference of the chemical potentials
rating of the oxidation resistance of FeCrAl as good or poor [4] is (lFe  lAl)bulk  (lFe  l Al)surf and (lFe  lCr)bulk  (lFe  lCr)surf
shown in Fig. 8. Although the results scatter due to dissimilar in two-dimensional plots (Figs. 9 and 10). In these gures we also
experimental conditions and rating criteria used in different inves- show the results of the experimental characterization of the main
tigations one can divide the diagram approximately into two parts, component of the surface oxides of FeCrAl alloys above 1000 C
showing good or poor oxidation resistance depending on the con- [4] and at 800 C [64].
E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404 3399

-70 AlO dominance at about 5 at% Al, close to the saddle point of the
-68
-66 20 FeCr chemical potential difference plot. The reason for the sug-
-64
-62
gested importance of Cr on the oxidation of FeCrAl in general
-60
15 could be the fast formation of Cr2O3 on the surface, a phenomenon
-58

Cr (at. %)
-56 that would give more time for a-Al2O3 to form up, provided there
-54
-52
is enough Cr at the surface at the initial stage of the oxidation. This
CrO
10 is in line with the grazing angle X-ray diffraction (XRD) measure-
AlO
(Cr,Al)O ments of FeCrAl. In contrast to the commonly held view that
FeO 5 corundum-type oxides only form above 900 C [25], there is a clear
evidence in the XRD results of FeCrAl that a-Al2O3 forms even at
0 700 C [29]. This was proposed to be due to the fact that the corun-
0 5 10 15 20 25 dum nucleation is greatly facilitated on a surface consisting of the
Al (at. %) isostructural Cr2O3.

Fig. 9. Calculated chemical potential difference (lFe  lAl)bulk  (lFe  lAl)surf (in 5.4. Tuning the chemical potentials by Cr substitution
mRy) as a function of Cr- and Al-concentrations. The results of the experimental
characterization of the main component of the oxide scale grown on FeCrAl
Next we discuss how the substitution of Al and/or Fe by Cr in
above 1000 C [4] are shown by symbols. Gray triangle upward: Cr2O3, blue triangle
downward: Al2O3, green square: mixed Cr2O3 and Al2O3, black circle: Fe2O3. (For bulk FeCrAl affects the chemical potentials of the components
interpretation of the references to colour in this gure legend, the reader is referred and how this is expected to be reected in the state of the Fe
to the web version of this article.) CrAl surfaces. The Cr substitution in bulk has direct effects on
both bulk and surface quantities. Starting with bulk properties
we note that lFe  lAl and lFe  lCr both decrease steeply with
15
10 increasing Cr content, within the region from 0 up to 1015 at%
9 20 Cr (Figs. 2 and 3). In contrast to the situation in bulk the chemical
8 II IV
7 potential of the surfaces changes rather slowly within the same
6
5 15 concentration region. This implies increasing chemical potential
Cr (at. %)

4
3
induced driving force for Al atoms to diffuse from the bulk to the
2
10 surface, with increasing Cr content in bulk. Fig. 2 also shows that
1
0 the driving force of the Cr diffusion reverses from the surface to
1
bulk direction to the bulk to surface direction at about 10 at% Cr
2 5
3 in bulk FeCrAl. Summarising from above, the Cr addition to
4
5 I III FeAl has a double-edged effect: the equilibrium surfaces will con-
CrO 0 tain more Al than they would have otherwise and within a certain
AlO 0 5 10 15 20 25
(Cr,Al)O concentration region also Cr appears in the surface. This effect can
FeO Al (at. %) be clearly seen in Table 1 where the 10 at% increase in the bulk Cr
content increases the Al-concentration in the surface layer by
Fig. 10. Calculated chemical potential difference (lFe  lCr)bulk  (lFe  lCr)surf (in
mRy) as a function of Cr- and Al-concentrations. The results of the experimental about 20 at%. Table 1 also shows that traces of Cr are expected to
characterization of the main component of the oxide scale grown on FeCrAl be found in the surface at high temperature when the Cr concen-
above 1000 C [4] are shown by symbols. Gray triangle upward: Cr2O3, blue triangle tration in bulk reaches the 10 at% level. Increasing the Cr content
downward: Al2O3, green square: mixed Cr2O3 and Al2O3, black circle: Fe2O3. The in FeCrAl up to 15 at% shows that the Cr-driven surface enrich-
major types of scaling behaviours fall into four regions IIV, separated by solid black
lines [64]. The latter data refer to 800 C and the major components of the oxide
ment with Al is saturated. This is due to the attening of the
scales are, I: Fe2O3, II: Cr2O3, III: Al2O3 with Fe2O3 nodules, IV: Al2O3. (For lFe  lAl (Fig. 3) when the Cr content in bulk is increased beyond
interpretation of the references to colour in this gure legend, the reader is referred 10 at%.
to the web version of this article.) Because the changes in the relative chemical potentials in the
bulk seem to be the key quantities in determining the surface con-
centrations of FeCrAl alloys at equilibrium conditions it is inter-
Somewhat surprisingly, it is observed that although Al-oxide is esting to analyse the mechanisms through which the substi
the protective component in the oxide scale, there is no correlation tutional Cr affects the chemical potentials in more detail. For that
between the calculated FeAl chemical potentials and the measured purpose we calculated for FeCrAl the density of states (DOS) at
composition of the oxide scale (Fig. 9). The reason for this result Fe, Cr, and Al sites as a function of Cr content. Comparing the site pro-
could be that the chemical potential difference in Fig. 9 is within jected DOSs, relative to the common Fermi energy, reveals that the
the range from 70 to 50 mRy (1.5  1019 to 1.1  1019 J). peaks in the occupied part of the Fe DOS shift down with the increas-
This means that the driving force for the Al diffusion between the ing Cr concentration. Therefore, the projected band energy in Fe sites
bulk and the surface regions due to chemical potentials is always di- is reduced and since Fe, Cr, and Al sites have the same Fermi energy
rected from the bulk to the surface and the relative variations in the this lowering of the Fe bands is consequently transferred to
Al driving force are small within the concentration region consid- (lFe  lAl)bulk and (lFe  lCr)bulk which decrease similarly with
ered. On the contrary, as Fig. 10 shows, the FeCr chemical potential increasing Cr content (Figs. 2 and 3).
changes sign within the investigated region implying signicant To some extent it is also possible to relate the chemical poten-
changes to the chemical potential induced driving force of the Cr dif- tials to atomic pair energies. To compare our results with these
fusion. The FeO dominated scales appear in the lower left corner of quantities we refer to investigations on FeAl with Cr impurities.
the plot (Fig. 10), where the driving force for Cr is from the surface to A study on the effect of Cr on the antiphase boundaries of FeAl al-
the bulk and there is practically no Cr at the equilibrium surfaces of loy [65] reports the interatomic energies. The interatomic energies
these alloys. The experimental borderline of the changing of the Fe of AlAl, CrAl, and FeAl pairs are of the same order of magnitude,
O dominance to CrO or AlO dominance lies approximately at the approximately from 0.4 to 0.6 eV. The FeFe interaction energy
line where the chemical potential difference, i.e. the driving force is about from 1.2 to 1.3 eV and the CrFe interaction energy is
of Cr diffusion changes its sign. The CrO dominance is changed to about 1.6 eV. This suggests that adding Cr into FeAl increases
3400 E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404

the average bonding of Fe to the alloy matrix which is in line with side is expected to be due to the different sensitivity factors in
our result of lowering the band energy of Fe. The site resolved mix- the oxide and metal matrices as already discussed in the exper-
ing energies in FeAl show slight preference for Cr to substitute Al, imental procedure section.
i.e. Cr prefers Fe neighbors [66] suggesting stronger bonding in Fe
Cr than in FeAl pairs. Comparing the Cr and Ni additions in FeAl, it The four zones described above are clearly discernible also in
has been found that contrary to Cr, Ni prefers to substitute atoms the Auger proles of Fe23Al (Fig. 5). In this case the rst two
in the Fe sublattice [67] which is in line with the fact that Ni tends zones are almost identical to those found in the Fe18Al case. A
to destroy the oxidation resistance of FeAl [4], quite an opposite slight difference is found at the top of the surface zone: it suggests
effect compared to that of Cr. that the Fe content decreases and correspondingly the Al-content
increases when the Al-concentration in the base alloy is raised
5.5. The role of Cr in the oxidation of FeCrAl from 18 to 23 at%. The zones of Fe23Al are slightly thinner than
the zones of Fe18Al.
The Cr induced changes in the state of the FeCrAl surfaces Compared to the Fe23Al case, the zone structure of Fe10Cr
suggest important consequences on the formation of the protective 10Al (Fig. 6) differs more from the zone structure of Fe18Al. From
oxide scale on the surfaces of FeCrAl. The initial oxidation de- top to bottom the thickness of the rst three zones in Fe10Cr
pends of course on the state of the alloy surface prior to oxidation. 10Al are about 85, 50, and 80 nm. The thickness of the deection
The Cr addition has both direct and indirect effects. The predicted zone cannot be determined reliably in this case because our mea-
increased Al-content in the surface suggests direct improvement in surement does not extend to a depth which would show the Al-
the formation of protective Al-oxide scales. On the other hand, Cr content to have reached the bulk value. However, we estimated
itself in the surface may induce transient oxidation containing this zone to be about 60 nm thick. In addition to the fact that the
Cr-oxide patches which act as oxidation retardants and nucleation thicknesses of the zones are somewhat larger in Fe10Cr10Al
centers for a-Al2O3 [29]. Since the formation rate of Cr-oxide is compared to those in Fe18Al the heights of the concentration pro-
higher than that of Al-oxide [68] and Cr2O3 has the same crystal les within the zones differ from those of Fe18Al as well. The con-
structure as a-Al2O3 the above facts are suggested to be benecial centration variations between the surface and saturation zones in
to the growth of the protective a-Al2O3 scale on the FeCrAl Fe10Cr10Al are smaller than in Fe18Al. Another clear difference
surfaces. is that in Fe10Cr10Al the Al-concentration does not have a dip in
the deection zone as it has in the Fe18Al and Fe23Al cases. This
5.6. Oxidation experiments is in line with our earlier prediction of the Cr-induced Al-pump
[70]. Furthermore, in Fe10Cr10Al there is a Cr contribution in
Now we are ready to analyse our experimental observations in the inversion and deection zones. No Cr in the surface and satura-
the light of our theoretical results and the published data. We start tion zones were detected which is in agreement with the differ-
with the characterization of the oxide scales of FeAl and FeCrAl ences between the diffusion constants of the components. The
alloys. Analysing Fig. 4 we can distinguish four different zones integrated interdiffusion constant of Cr is much lower than that
within the Auger prole of Fe18Al. Although the division of the of Fe and Al. The approximate ratios are 1:4:5 for Cr, Fe, and Al
Auger proles into different zones is to some extent arbitrary it in (Fe, Cr)3Al2 [71]. The observed weak maximum in the Cr concen-
is, however, a very useful concept in investigating the oxidation tration could be a trace of the initial surface since Cr-rich oxide has
mechanism in general and in comparing the oxidation of different been related to the position of the initial surface before oxidation
alloys. Here we omit the approximately 5 nm thick topmost sur- [28].
face layer which we expect to have been affected by atmospheric Comparing the Auger proles of Fe10Cr10Al with those of
contaminants. Based on the observed atomic fractions of Fe, Al Fe18Al and Fe23Al reveals that if the Fe and Cr proles are added
and O we dene the different zones in sequence from the surface up the obtained sum prole within the inversion zone is very sim-
to the base alloy as: ilar to the Fe proles of Fe18Al and Fe23Al. Also the Al proles of
Fe10Cr10Al, Fe18Al, and Fe23Al are quite similar. One should
1. Surface zone. Within this about 75 nm thick zone the oxygen note that it is the Al-oxide that is considered to block the atomic
content is practically constant while the Fe content decreases diffusion at high temperatures in these alloys. Therefore, substitut-
from 5 at% to zero and the Al-content increases from 35 to ing 1/9 of Fe by Cr in the Fe10Al alloy seems to raise the Al-con-
40 at% correspondingly. tent in the oxide scale to the same level as that found in the Fe
2. Saturation zone. Here the Al-content reaches its maximum and 18Al and Fe23Al alloys.
there is practically no Fe in this zone. The oxygen content is
at the same level as in the previous zone. The thickness of the 5.7. Initial oxidation
saturation zone is about 40 nm.
3. Inversion zone. In the next, about 60 nm wide section the Fe con- So far we have not discussed much the important initial stage of
tent increases and the Al-content drops down, both approach- the oxidation [2,3] of FeAl alloys just at the oxidation resistance
ing their bulk values. The right hand side end of the inversion borderline, i.e. alloys containing 1015 at% Al [4]. Therefore, we
zone is dened here to be at the inexion point of the Al-con- now try to get better understanding of the high temperature oxida-
centration curve. tion of Fe13Al alloys. For that purpose we re-examined our Fe
4. Deection zone. This about 50 nm thick zone is found below the 13Al sample at the initial stage of oxidation. Instead of 5 min oxi-
inversion zone. Within this zone the atomic concentrations are dation the sample was oxidized only 1 min at normal pressure and
still deected from their bulk values to some extent. One should 1000 C temperature. After oxidation the surface was scanned by
note that different sputtering yields of the chemical elements AFM and a surface with nodules randomly scattered on an other-
may lead to somewhat biased values for the atomic fractions wise smooth base was observed (Fig. 11). Our observation is in
in the proles. Especially the result for Al fraction might be agreement with earlier results, see e.g. [64,72], which show ran-
slightly underestimated. However, in our investigations we domly positioned Fe-oxide and Fe-sulde nodules on a smooth
used such Auger lines which are not very surface sensitive Al-oxide surface of FeAl and FeCrAl alloys. Analysing the com-
[69]. Additional uncertainty to the atomic fractions in the alloy position of the nodules reveals that they consist almost exclusively
E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404 3401

Fig. 11. AFM image of an FeO nodule on the smooth surface oxide scale of Fe13Al
oxidized at 1000 C and 1 atm for 1 min.

of iron and oxygen. The nodular structure has been observed also
earlier for FeAl and FeCrAl alloy surfaces [20,27,73,64].
Between the nodules we found a smooth thin oxide scale con-
taining both Fe- and Al-oxides. A typical Auger prole measured Fig. 13. Optical microscope image of FeO nodules at the surface of Fe13Al
in this smooth surface region between the nodules is shown in oxidized at 1000 C and 1 atm for 1 min.
Fig. 12. Comparing Fig. 12 with Fig. 4 one realizes that the thin
smooth oxide scale in Fe13Al shows similar features as the pro-
tective scale in Fe18Al. The signicant differences between Fe oxide scale, eventually breaking it. There can be also hidden de-
13Al and Fe18Al are found in the surface zone, i.e. in the topmost fects in the protective scale through which Fe atoms can gradually
part of the oxide scale. In Fe13Al the surface zone contains more diffuse from the bulk alloy to the surface and O atoms can diffuse
Fe than Al and also the oxygen content, compared to the metal from the surface to the bulk alloy. This kind of leak process would,
atoms, is higher than the ratio 3/2 which was found in Fe18Al. however, trigger new Fe-rich nodules on the smooth oxide scale.
Actually, the total iron content within the surface zone of Fe To obtain information on the large scale properties of the nod-
13Al is several magnitudes higher than in the Fe18Al case. This ular structure of the oxidized Fe13Al surface optical microscope
suggests that at the initial stage the Al-oxide scale in Fe13Al leaks was used. As Fig. 13 shows the Fe-oxide nodules are randomly
and there exists a substantial Fe ux from the base alloy to the sur- scattered throughout the otherwise smooth surface. Since after
face through the Al-oxide scale during the rst moments of the 1 min of oxidation all the nodules have the same symmetrical cir-
oxidation. cular shape and they all are of the same size, we conclude that they
Our investigations suggest that the surface of Fe13Al is par- all have emerged at the very rst moments of the oxidation from
tially covered by the protective Al-oxide scale. This Al-oxide scale point defects scattered randomly on the initial alloy surface. Dur-
shows two blocking fronts (Fig. 12). The blocking front for the Fe ing the oxidation the nodules grow and after 5 min they practically
diffusion from the bulk alloy to the surface is located about cover the whole surface. To improve the oxidation resistance of
125 nm below the surface and the blocking front for the O diffusion low-Al Fe-alloys it is important to investigate what are those point
from the surface to the bulk alloy is found about 50 nm below the defects that nucleate the nodular growth of Fe-oxide and what are
surface. However, the protective Al-oxide scale in Fe13Al is only a the possibilities to inactivate them.
transient feature. We already know from our earlier oxidation The main features of our Auger depth proles agree with those
experiments that after 5 min of oxidation most of the surface is of other similar measurements on FeAl and FeCrAl alloys [73
covered with a non-protective thick Fe-oxide scale. The protective 78]. Summarizing from above, the following conception of the oxi-
Al-oxide scale of Fe13Al can be destroyed principally in two ways. dation process of the investigated alloys is proposed.
The destructive Fe-oxidation may start from the Fe-oxide rich nod-
ules and proceed sideways along the surface into the protective (i) For FeAl alloys under oxygen ow at 1000 C and 1 atm the
threshold value of the Al-content for the protective Al-oxide
scale to form up and hold for a longer time is between 13
and 18 at% in the base alloy.
(ii) In the main part of the oxide scale the oxygen and metal
atom concentrations are 60 and 40 at%, respectively, cor-
responding to the atomic fractions in the stoichiometric oxi-
des Al2O3 and Fe2O3. This suggests that in the surface and
saturation zones practically all oxygen and metal atoms
are combined into oxides.
(iii) The building up of the Fe-free saturation zone can be under-
stood in the following way. At the early stages of oxidation
there presumably exist both Al- and Fe-oxides in the satura-
tion zone region. However, the oxygen ow from the gas
phase is decreased step by step as the oxide layer grows. If
there is no free oxygen available and the Al ux from the
base alloy brings more Al to this region the higher oxygen
afnity of Al [79], compared to that of Fe, leads to the pro-
cess 2Al + Fe2O3 ? 2Fe + Al2O3 which combined with the
Fig. 12. AES surface depth proles of the smooth oxide scale of Fe13Al oxidized at subsequent Fe escape due to diffusion saturates this zone
1000 C and 1 atm for 1 min. with Al-oxide.
3402 E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404

(iv) When the Al-concentration of an FeAl alloy is close to the


threshold value of the onset of the high temperature oxida- Ds
tion resistance both protective and non-protective surface
oxide scales coexist at the surface. The non-protective oxide
scale grows by the rapid generation of the FeO nodules dur-
Eb Es
ing the rst moments of the oxidation. These nodules are

Energy
nucleated by the surface point defects. While the oxidation
continues the nodules grow at the expense of the protective
scale.
(v) Replacing Fe by Cr in the base alloy pushes more Al to the
surface region leading to the formation of the protective Db
Al-oxide scale with considerably reduced Al-content in the
base alloy. In the oxide scale of Fe10Cr10Al the concentra-
tion of Al is even higher than in the oxide scales of Fe18Al
and Fe23Al alloys. 0 1 2 3
n (Distance form the surface)
5.8. Atomic scale mechanisms Fig. 14. Schematic gure of the energetics of the exchange processes of AlFe pairs.
The red wavy curve represents the energy of the system at different stages of the
To shed some light on the underlying mechanisms of the sur- pair exchange of an FeAl pair. The conguration Al at n = 1 and Fe at n = 2 is shown
face oxidation of FeAl alloys we reconsider our experimental data by the green sphere on the left and the conguration Fe at n = 1 and Al at n = 2 is
shown by the green sphere on the right. On average Al atoms in FeCr matrix will
in the light of our computational results. The computational results
diffuse from right to left, i.e. from the high Cr region (bulk side, high potential) to
refer to the (0 0 1) surface. However, the oxidation experiments for the low Cr region (surface side, low potential). (For interpretation of the references
FeAl above 500 C lead to the formation of well-ordered Al2O3 lms to colour in this gure legend, the reader is referred to the web version of this
on all low-index surfaces (0 0 1), (0 1 1), and (1 1 1) [34]. Moreover article.)
Al segregation is observed for all low-index surfaces [69]. We
therefore expect that the theoretical results for the (0 0 1) surface
with successively increasing n. From Eqs. (5)(8) one obtains the ra-
can be justiably used in analysing our experimental data.
tio of the diffusion coefcients
From the theoretical point of view we can describe two features
that affect the oxidation resistance of FeAl alloys at high temper- n n
Ds elFe lAl =kB T
atures. The rst one is a straightforward mechanism to improve n1 n1 : 9
Db elFe lAl =kB T
the oxidation resistance simply by increasing the Al-concentration
in the base alloy. This accelerates, in a natural way, the formation
According to our calculations (Table 1) the Cr concentration at the
of the protective Al-oxide scale which then reduces both internal
surface is lower than that in the bulk. Therefore, it is expected that
oxidation as well as the growth of the outer oxide scale.
in the region between the bulk and the surface ln1 Fe  lAl
n1
is less
However, there exists another, a more complex, mechanism
than lnFe  lnAl (in line with Fig. 3) leading to Ds/Db > 1, i.e. the above
that is also benecial to the oxidation resistance of FeAl alloys.
model predicts the driving force of the Al diffusion to be towards
This can be explained by considering the calculated chemical
the surface. The energetics of the diffusion is illustrated in Fig. 14.
potentials shown in Fig. 3. As seen in Fig. 3 there exists a strong
Due to the anisotropy of the Al diffusion, the Al-concentration in
chemical potential induced driving force for Al to diffuse from bulk
the surface region is expected to exceed that in the bulk. This
to the surface which is in line with the surface energies of the com-
enrichment of the surface with Al continues as long as the mini-
ponents [80,81] and the results of the surface segregation investi-
mum of the free energy has been achieved or the surface has be-
gations [82].
come saturated with Al. When this stage is reached there is no
It is also useful to relate the data shown in Fig. 3 to the diffusion
net driving force for Al diffusion to any direction.
coefcients in FeCrAl. Considering either direct or concerted
In Table 1 we show the calculated equilibrium surface concen-
AlFe pair exchange process [83] the corresponding diffusion
tration of Fe5Al alloy as a function of temperature. We see that at
coefcient for the Al moving towards the surface (Ds) or Al moving
0 K the surface in the equilibrium state contains 87 at% Al. Using
towards the bulk (Db) are approximated as
the data in Fig. 3 we estimate that in the case of Fe13Al, Fe
18Al, and Fe23Al the equilibrium surface at 0 K would contain
Ds D0 eDEs =kB T ; 5
about 94, 98 and 100 at% Al, respectively. However, at high tem-
Db D0 eDEb =kB T ; 6 peratures the entropy contribution in the free energy becomes sig-
nicant and consequently at about 1000 C the corresponding
where T is the temperature and kB is the Boltzmanns constant, the equilibrium Al surface concentrations for Fe13Al, Fe18Al, and
activation energies are Fe23Al are 67, 72 and 77, respectively. It is interesting to note that
the Al surface concentrations at 1000 C of Fe13Al and Fe5Cr5Al
DEs Esp  EAln1;Fen ; 7 are approximately the same and also according to Fig. 7 the exper-
sp Aln;Fen1
DEb E  E : 8 imental scaling losses of these alloys are of the same order. An-
other similar pair of alloys is Fe18Al and Fe10Cr5Al suggesting
Here D0 and Esp are the pre-exponential factor and the energy of the for a connection between the Al-concentration at the surface layer
system when the AlFe pair is in its saddle-point conguration and the scaling loss in oxidation atmosphere.
along its migration path, both assumed to be independent of the What is even more interesting here is the role of Cr in enhancing
direction of the exchange process. EAl(n + 1),Fe(n) and EAl(n),Fe(n+1) are the oxidation resistance of FeAl alloys. The effect of Cr on the driv-
the energies of the system in the initial states corresponding to ing force of Al diffusion is clearly seen in Fig. 3. The chemical poten-
AlFe pair having the Al end in the bulk side or in the surface side, tial difference (lFe  lAl)bulk decreases substantially when Cr
respectively. The index n refers to the atomic layer, n = 0 refers to content in the base alloy is increased from 0 to 10 at%. This decrease
the surface layer and deeper layers towards the bulk are indexed means that in the bulk the chemical potential of Al is raised com-
E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404 3403

pared to that of Fe. Because the same Cr addition does not change the AES surface depth proles are compared with the calculated atom-
chemical potential difference at the surface appreciably, the driving ic chemical potentials and surface concentration proles. To con-
force of Al from bulk to the surface is therefore considerably en- clude we list the main results found in our investigation.
hanced due to the Cr addition as veried computationally and shown
in Table 1. The Cr-enhanced oxidation resistance discussed above is 1. Increasing the Al-content of FeAl from 13 to 18 at% makes the
effective mainly at the initial stage of oxidation or during the healing alloy resistant to oxidation at 1000 C. Further increase of the Al
of a broken oxide scale, i.e. in cases where an open metal surface is content to 23 at% slightly improves the protecting Al-oxide
exposed to oxidizing environment. However, this is not the whole scale on the surface. At the borderline of the onset of the oxida-
story of the benecial effect of Cr in FeAl alloys. Cr improves the oxi- tion resistance in FeAl alloys the nodular growth of Fe-oxide is
dation resistance of FeAl alloys also during the steady state of the nucleated at randomly scattered points along the surface. Dur-
oxide scale growth. Comparison of Figs. 5 and 6 shows that an Fe ing the oxidation these nodules grow in size but not in number.
10Cr10Al alloy has a better supply of Al in the oxide scale than an Further investigation of the nucleation and the early kinetics of
Fe23Al alloy even though the Al-content in the latter is more than these nodules is expected to be useful in improving the quality
twice of that of the rst one. In the oxide scale of Fe10Cr10Al of FeAl surfaces.
the Al-content is higher and correspondingly the Fe content lower 2. Substituting Fe by Cr in FeAl alloys clearly increases the driv-
than in the oxide scale of the Fe23Al alloy. This can be understood ing force of Al to diffuse from the bulk to the surface. Because
on the basis of the calculated chemical potentials shown in Fig. 3. the increase of the Al driving force is mainly due to bulk prop-
During the initial oxidation also Cr oxidizes leading to a Cr-depleted erties, this increase is not very sensitive to the type of the sur-
zone [84,15] under the oxide scale. Due to the slow mobility of Cr face, whether it is metal or oxide, or the type of the oxide. A
[71] this Cr-depleted zone is stable. If the Al-concentration in bulk driving force for the Al diffusion based on the Cr concentration
is below the critical level, then due to the high oxygen afnity of Al gradient in FeCrAl is predicted.
[79] the exposure of the surface to oxygen results in a formation of 3. Comparing the measured chemical composition of the oxide
an Al-depleted zone under the oxide scale, which could eventually scales with the calculated chemical potentials of the atoms in
lead to the breakdown of the protective oxide scale [4]. The impor- the bulk and at the surface of the alloy a correlation between
tant issue within the maintaining of the protective oxide scale is the composition of the oxide and the energetic balance between
the recovery of the Al level in the depleted zone. At this point Cr plays Fe and Cr in the bulk and at the surface (namely (lFe  lCr)-
bulk
an important role. As seen in Fig. 3 the Cr depletion in FeCrAl raises  (lFe  lCr)surf) can be seen. In the case of Fe and Al a sim-
the chemical potential of Fe relative to that of Al and consequently ilar kind of correlation is not found.
helps in restoring the Al to Fe ratio in the depleted zone to the level 4. With increasing Cr content the Fe bands shift down, relative to
that is needed to maintain the protective oxide scale. the Cr and Al bands, which can be related to the calculated
Using the chemical potentials shown in Fig. 3 we can give a the- changes in the chemical potentials in FeCrAl. Increasing the
oretical estimate, based on our calculations, on the effectivity of Al Cr concentration in the bulk results in two effects on the surface
and Cr on the improvement of the high temperature oxidation of the FeAl alloys: the Al-content at the surface increases and
resistance of FeCrAl alloys. To get the same improvement in Cr is also found at the surface.
the oxidation resistance the Al addition should be about twice of 5. Corrosion resistant surfaces of FeCrAl are quite robust
the amount of the Cr addition. This result is in line with our Auger because their formation and self-healing properties are to a
measurements since comparing Figs. 46 one observes that the large extent determined by the bulk part of the alloy.
oxide scale of Fe10Cr10Al contains less Fe than that of Fe
23Al. Therefore, adding 10 at% Cr is effectively more than adding Acknowledgements
13 at% Al. Also the experimental results of the scaling losses
(Fig. 7) show that if one considers, for instance, Fe10Al which The computer resources of the Finnish IT Center for Science
has the scaling loss of about 50 g/(m2h), there are two extreme (CSC) and Mgrid project are acknowledged. Financial support from
ways to improve its oxidation resistance. One can either substitute the Academy of Finland (Grant No. 116317), Outokumpu Founda-
Fe by Cr or Al. If the Cr concentration is increased from 0 to 5 at% tion (E.A.), National Graduate School in Materials Physics (E.N.),
we arrive at the Fe5Cr10Al alloy which has the scaling loss of the Swedish Research Council, the Swedish Iron Ofce (B.J., L.V.),
about 2 g/(m2h). In order to achieve the same scaling loss level and the Carl Tryggers Foundation (M.P.) are acknowledged.
by Al addition we have to add Al twice of the Cr amount, which
leads to the Fe20Al alloy having approximately the 2 g/(m2h) scal- References
ing loss.
As Fig. 2 shows, adding Cr into FeAl alloys affects also lFe  lCr. [1] S.L. Case, K.R. van Horn, in: F.T. Sisco (Ed.), Aluminium in Iron and Steel, Alloys
As above, also in this case the drop of lFe  lCr in the bulk is much of Iron Research Monograph Series, John Wiley and Sons Inc., New York, 1953.
[2] A.S. Khanna, Introduction to High Temperature Oxidation and Corrosion, AMS
larger than at the surface. Therefore, considering FexCr10Al alloys International, Materials Park, OH, 2002.
one observes that for small x the driving force for Cr (relative to that [3] A.S. Khanna, High temperature oxidation, in: M. Kutz (Ed.), Handbook of
of Fe) is from the surface to the bulk. Increasing the Cr concentration Environmental Degradation of Materials, William Andrew Publishing,
Norwich, NY, 2005, pp. 105152.
this tendency gradually weakens and at about 10 at% Cr in the base [4] P. Tomaszewicz, G.R. Wallwork, Rev. High Temp. Mater. 4 (1978) 75105.
alloy there is no relative driving force anymore and Cr and Fe have [5] B.A. Gordon, W. Worrell, V. Nagarajan, Oxid. Met. 13 (1979) 1323.
equal probabilities to occupy bulk and surface sites. This phenome- [6] G.B. Abderrazik, G. Moulin, A.M. Huntz, E.W.A. Young, J.H.W. de Wit, Solid State
Ionics 22 (1987) 285294.
non of the increasing of the probability of Cr, relative to that of Fe, to [7] F.H. Stott, Rep. Prog. Phys. 50 (1987) 861913.
occupy surface sites has also a benecial effect on the formation of [8] F.H. Stott, F.I. Wei, Oxid. Met. 31 (1989) 369391.
the protective oxide scale on the surface. [9] R. Prescott, M.J. Graham, Oxid. Met. 38 (1992) 7387.
[10] J.H. DeVan, P.F. Tortorelli, Corros. Sci. 35 (1993) 10651071.
[11] I. Gurrappa, S. Weinbruch, D. Naumenko, W.J. Quadakkers, Mater. Corros. 51
(2000) 224235.
6. Conclusions [12] C. Schwalm, M. Schtze, Mater. Corros. 51 (2000) 161172.
[13] N. Babu, R. Balasubramaniam, A. Ghosh, Corros. Sci. 43 (2001) 22392254.
[14] D.B. Lee, G.Y. Kim, J.G. Kim, Mater. Sci. Eng. A339 (2003) 109114.
Ab initio electronic structure calculations have been performed [15] I.G. Wright, R. Peraldi, B.A. Pint, Mater. Sci. Forum 461464 (2004) 579590.
for Fe1xyCrxAly alloys, 0 6 x 6 0.25, 0 6 y 6 0.25. The measured [16] J.A. Nychka, D.R. Clarke, Oxid. Met. 63 (2005) 325352.
3404 E. Airiskallio et al. / Corrosion Science 52 (2010) 33943404

[17] V. Rohr, Development of novel protective high temperature coatings on heat [51] P. Soven, Phys. Rev. 156 (1967) 809813.
exchanger steels and their corrosion resistance in simulated coal ring [52] B.L. Gyrffy, Phys. Rev. B 5 (1972) 23822384.
environment, Ph.D. dissertation, Centre Interuniversitaire de Recherche et [53] M. Ropo, K. Kokko, L. Vitos, J. Kollr, Phys. Rev. B 71 (2005) 045411-1045411-
dIngnierie des Matriaux (CIRIMAT), 2005. 6.
[18] H.J. Grabke, Mater. Technol. 40 (2006) 3947. [54] M. Ropo, K. Kokko, L. Vitos, J. Kollr, B. Johansson, Surf. Sci. 600 (2006) 904
[19] H. Asteman, M. Spiegel, Corros. Sci. 50 (2008) 17341743. 913.
[20] Y. Niu, S. Wang, F. Gao, Z.G. Zhang, F. Gesmundo, Corros. Sci. 50 (2008) 345 [55] M. Ropo, K. Kokko, M.P.J. Punkkinen, S. Hogmark, J. Kollr, B. Johansson, L.
356. Vitos, Phys. Rev. B 76 (2007). 220401(R)-14.
[21] N. Birks, G.H. Meier, Introduction to High Temperature Oxidation of Metals, [56] K.D. Childs, B.A. Carlson, L.A. LaVanier, J.F. Moulder, D.F. Paul, W.F. Stickle, D.G.
Edward Arnold Ltd., London, 1983. Watson, Handbook of Auger Electron Spectroscopy, third ed., Physical
[22] G. Wrangln, An Introduction to Corrosion and Protection of Metals, Chapman Electronics Inc., Minnesota, 1995.
and Hall, New York, 1985. [57] V. Blum, L. Hammer, W. Meier, K. Heinz, M. Schmid, E. Lundgren, P. Varga, Surf.
[23] M. Palm, Intermetallics 13 (2005) 12861295. Sci. 474 (2001) 8197.
[24] C. Wagner, Corros. Sci. 5 (1965) 751764. [58] W. Meier, V. Blum, L. Hammer, K. Heinz, J. Phys. Condens. Matter 13 (2001)
[25] F.H. Stott, G.C. Wood, J. Stringer, Oxid. Met. 44 (1995) 113145. 17811791.
[26] C. Badini, F. Laurella, Surf. Coat. Technol. 135 (2001) 291298. [59] L. Hammer, W. Meier, V. Blum, K. Heinz, J. Phys. Condens. Matter 14 (2002)
[27] Z.G. Zhang, F. Gesmundo, P.Y. Hou, Y. Niu, Corros. Sci. 48 (2006) 741765. 41454164.
[28] H. Gtlind, F. Liu, J.-E. Svensson, M. Halvarsson, L.-G. Johansson, Oxid. Met. 67 [60] E. Airiskallio, E. Nurmi, I.J. Vyrynen, K. Kokko, M. Ropo, M.P.J. Punkkinen, B.
(2007) 251266. Johansson, L. Vitos, Phys. Rev. B 80 (2009) 153403-1153403-4.
[29] H. Josefsson, F. Liu, J.-E. Svensson, M. Halvarsson, L.-G. Johansson, Mater. [61] H.G. Read, H. Murakami, K. Hono, Scripta Mater. 36 (1997) 355361.
Corros. 56 (2005) 801805. [62] C. Capdevila, M.K. Miller, K.F. Russell, J. Mater. Sci. 43 (2008) 38893893.
[30] G. Berthom, E. NDah, Y. Wouters, A. Galerie, Mater. Corros. 56 (2005) 389 [63] C. Capdevila, M.K. Miller, K.F. Russell, J. Chao, J.L. Gonzlez-Carrasco, Mater. Sci.
392. Eng. A490 (2008) 277288.
[31] F.J. Prez-Trujillo, S.I. Castaeda, Oxid. Met. 66 (2006) 231251. [64] P. Tomaszewicz, G.R. Wallwork, Oxid. Met. 20 (1983) 75109.
[32] B.B. Ebbinghaus, Combust. Flame 93 (1993) 119137. [65] L. Senying, J.A. Leiro, Jpn. J. Appl. Phys. 38 (1999) 28062811.
[33] M.P. Brady, B. Gleeson, I.G. Wright, JOM J. Miner. Met. Mater. Soc. 52 (2000) [66] A. Strutz, D. Fuks, Comput. Modell. New Technol. 9 (2005) 714.
1621. [67] C.L. Fu, J. Zou, Acta Mater. 44 (1996) 14711478.
[34] H. Graupner, L. Hammer, K. Heinz, D.M. Zehner, Surf. Sci. 380 (1997) 335351. [68] J. Klwer, Mater. Corros. 49 (1998) 758763.
[35] D. Pilone, Recent Patents Mater. Sci. 2 (2009) 2731. [69] L. Hammer, H. Graupner, V. Blum, K. Heinz, G.W. Ownby, D.M. Zehner, Surf. Sci.
[36] P.F. Tortorelli, K. Natesan, Mater. Sci. Eng. A 258 (12) (1998) 115125. 412/413 (1998) 6981.
[37] P. Hohenberg, W. Kohn, Phys. Rev. 136 (1964) B864B871. [70] E. Airiskallio, E. Nurmi, M.H. Heinonen, I.J. Vyrynen, K. Kokko, M. Ropo, M.P.J.
[38] W. Kohn, L. Sham, Phys. Rev. 140 (1965) A1133A1138. Punkkinen, H. Pitknen, M. Alatalo, J. Kollr, B. Johansson, L. Vitos, Phys. Rev. B
[39] L. Vitos, I.A. Abrikosov, B. Johansson, Phys. Rev. Lett. 87 (2001) 156401-1 81 (2010) 033105-1033105-4.
156401-4. [71] P.C. Tortorici, M.A. Dayananda, Mater. Sci. Eng. A244 (1998) 207215.
[40] L. Vitos, Computational Quantum Mechanics for Materials Engineers: The [72] S.W. Banovic, J.N. DuPont, A.R. Marder, Metall. Mater. Trans. A 31A (2000)
EMTO Method and Applications, Engineering Materials and Processes Series, 18051817.
Springer-Verlag, London, 2007. [73] C. Houngniou, S. Chevalier, J.P. Larpin, Oxid. Met. 65 (2006) 409439.
[41] J.N. Andersen, D. Hennig, E. Lundgren, M. Methfessel, R. Nyholm, M. Schefer, [74] P.Y. Hou, G. Ackerman, Appl. Surf. Sci. 178 (2001) 156164.
Phys. Rev. B 50 (1994) 1752517533. [75] P.Y. Hou, J. Am. Ceram. Soc. 86 (2003) 660668.
[42] O.K. Andersen, O. Jepsen, G. Krier, Exact mufn-tin orbital theory, in: V. Kumar, [76] P.Y. Hou, Annu. Rev. Mater. Res. 38 (2008) 275298.
O.K. Andersen, A. Mookerjee (Eds.), Lectures on Methods of Electronic [77] J. Engkvist, U. Bexell, M. Grehk, M. Olsson, Mater. Corros. 60 (2009) 876881.
Structure Calculations, World Scientic Publishing Co., Singapore, 1994, pp. [78] J. Engkvist, T.M. Grehk, U. Bexell, M. Olsson, Surf. Coat. Technol. 203 (2009)
63124. 28452850.
[43] L. Vitos, Phys. Rev. B 64 (2001) 014107-1014107-11. [79] J. Camra, E. Bielanska, A. Bernasik, K. Kowalski, M. Zimowska, A. Biaas, M.
[44] L. Vitos, H.L. Skriver, B. Johansson, J. Kollr, Comp. Mat. Sci. 18 (2000) 2438. Najbar, Catal. Today 105 (2005) 629633.
[45] M. Zwierzycki, O.K. Andersen, Acta Phys. Pol. A 115 (2009) 6468. [80] M. Aldn, H.L. Skriver, S. Mirbt, B. Johansson, Surf. Sci. 315 (1994) 157172.
[46] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 38653868. [81] L. Vitos, A.V. Ruban, H.L. Skriver, J. Kollr, Surf. Sci. 411 (1998) 186202.
[47] K. Kokko, M.P. Das, J. Phys. Condens. Matter 10 (1998) 12851291. [82] B. Eltester, C. Uebing, H. Viefhaus, H.J. Grabke, Fresenius J. Anal. Chem. 358
[48] M. Ropo, K. Kokko, L. Vitos, Phys. Rev. B 77 (2008) 195445-1195445-6. (1997) 196199.
[49] F. Lechermann, M. Fhnle, B. Meyer, C. Elssser, Phys. Rev. B 69 (2004) 165116- [83] A. Antonelli, S. Ismail-Beigi, E. Kaxiras, K.C. Pandey, Phys. Rev. B 53 (1996)
1165116-7. 13101314.
[50] P.G. Gonzales-Ormeo, H.M. Petrilli, C.G. Schn, Scripta Mater. 54 (2006) [84] R.M. Deacon, J.N. DuPont, C.J. Kiely, A.R. Marder, P.F. Tortorelli, Oxid. Met. 72
12711276. (2009) 87107.

You might also like