o.

,
IIHI$
ELSEVIER Thin Solid Films 303 (1997) 101-106
J

Chemical reaction path for thin film oxidation of stainless steel

Shigeki ShibagaM a,* Akihiro Koga a, Yoshinori Shirakawa a, Hiroko Onishi b,
Harumi Yokokawa c, Junzo Tanaka d
a Research and De~,elopment DiHsion, Sumitomo Metal Industries Ltd., 1-8 Fuso-cho, Amagasaki 660, Japan
b Technical Research Center, Shinko PANTEC Co., Ltd., 1-4-1 Murotani, Nishi-ku, Kobe 651-22, Japan
c National Institute of Mawrials and Chemical Research, Higashi l-l, Tsukuba, Ibaraki 305, Japan
d National Institute for Research in hzorganic Materials, Namiki 1-1, Tsukuba, lbaraki 305, Japan
Received 14 October 1996; accepted 26 February 1997

Abstract

An oxide thin film formed on a stainless steel is analyzed by microfocused Auger electron spectroscopy (#AES): the oxide film
consists of five sublayers of 2 to 4 nm in thickness. A chemical reaction path of this system is elucidated by comparing a chemical
potential diagram with the elementary distribution observed; thereby the phases in the thin film are specified. The diffusion property of
this system is discussed in relation to the chemical reaction path and it is indicated that an up-hill diffusion occurs in the oxide thin film.
© 1997 Elsevier Science S.A.

Keywords: Auger electron spectroscopy; Oxides

1. Introduction A phase distribution in an oxide thin film formed on the

stainless steel was analyzed by microfocused Auger elec-
In developing elution-free metals, it is important to tron spectroscopy (/zAES). The thickness of tile oxide
elucidate an initial oxidation process on the metal surface. film changed slightly along the surface: the /zAES tech-
For example, an oxide thin film which is formed on the nique was consequently applied for the measurement of
surface of F e - C r alloy suppresses the elution of metal ions the depth profile in the metal-oxide thin film.
from the surface into high temperature extra-pure water or
ozone-additive, extra-pure water [I]. The surface-oxidized
alloy is thereby useful for a production system of extra-pure 2. E x p e r i m e n t a l
water. It is believed that the elution-free property is due to
a Cr-rich phase in the thin film formed in the oxidation
2. l. Sample preparation
process.
In practice, stainless steel is expected to be used as the
Stainless steel 316L was used as the base material for
elution-free material. In the surface oxidation process of
an oxide thin film; this material mainly contains Fe (66%),
this material, thermodynamic phase stability is very com-
Cr (17%), Ni (12%) and Mo ( ~ 2%). After its surface was
plicated since the stainless steel contains many con-
polished mechanically to a mirror finish, damaged layers
stituents; moreover, mass transfer occurs under chemical
on the surface were removed by electrolytic polishing; the
potential gradient due to oxygen activity. The objective of
sample was finally annealed at 700 K in air. Just before
this paper is to elucidate which phases appear in the thin
# A E S measurement, the sample was ultrasonically cleaned
film formed on stainless steel and whether a thermody-
in acetone.
namic approach based on the principle of local equilibrium
[2,3] is valid or not for the oxidation process of stainless
steel, even in a thin film of the order of 10 nm in 2,2. Measurement
thickness.
Surface analysis of the sample was performed by AES
(Microlab 320D: VG Scientific). The accelerating voltage
* Corresponding author. of an incident electron beam was 5 keV; the current was 1

0(140-6090/97/$I7.00 © 1997 Elsevier Science S.A. All rights reserved.

PII S 0 0 4 0 - 6 0 9 0 ( 9 7 ) 0 0 1 3 0 - 2
102 S, Shibagaki et aL / Thin Solid Films 303 (1997) I01-106

nA and the diameter was 0.1 mm. Depth profiles were D~ D~ D 2 D3 Da

measured by etching with an Ar ion gun; the etching rate i.2
was about 0.2 nm/min.
The contents of constituents were evaluated by using
relative sensitive factors which were determined from stan-
dard samples: Fe20 3, Cr20 3 and NiO (Getler Micro Ana-
lytical Laboratory). FeO
Fe304
F%Os
3. Results o.81 ~, I, , I ~oo ~ ,
0 40 80 120
3.1. Elementary distribution in thin oxide film Etching Time/rain
Fig. 2. Spectrum of LMM Auger transition for Fe (inlet) and relative
Fig. 1 shows a typical example of the depth profiles of
intensity ratio between the L23M45M45 and L23M~3M45 transitions,
the constituents in the oxide thin film. The molar fraction
of the constituents is plotted as a function of distance from
the surface. there exist five sublayers; each sublayer is about 2 to 4 nm
The contents change as follows: (1)The oxygen content thick.
is constant near the surface; it decreases gradually in the
vicinity of 4 nm in depth (point D1) and rapidly below at 3.2. Depth profile of valence of Fe
about 8 nm (point D2). (2) The iron content is also
constant near the surface; it decreases between Dt and D2 The valence of Fe was evaluated from the Auger transi-
and increases between D2 and D4 (D4: point of 16 nm in tion for Fe. A spectrum of LMM Auger transition for Fe is
depth). Below D4, the iron has almost the same content as shown in the inlet of Fig. 2. This spectrum consists of
the base material. (3) The chromium content is zero near three main peaks: L23Ma3M23, L23M23M45 and
the surface; it increases in the vicinity of D1 corresponding L23M45M45 transitions from the lower energy side. The
to the point where the oxygen and the iron decrease, and it relative intensity among these peaks changes correspond-
reaches a maximum near D2. In the region deeper than ing to the valence of iron [4].
point D2, the chromium decreases again and becomes Fig. 2 shows the relative intensity ratio between
almost constant near D4. (4) The nickel content is very L23Mz3M45 and L23M45M45 transitions of the thin film as
low but not zero near the surface; it increases gradually at a function of the depth. The compounds noted on the
about 6 nm (point D2' is slightly more shallow than point right-side ordinate were determined from the intensity
D2) and reaches a maximum followed by a gradual de- ratios for standard samples of Fe, Fe30 4 and Fe20 3 [5]. As
crease below D3. shown in Fig. 2, the valence of Fe changes with the depth:
The reason why the contents of Cr and Ni differ from the valence is + 3 between the surface and D 1, it changes
those of the base material is plausibly the fact that the from + 3 to + 2 between D1 and D2, and proceeds further
sensitivity factors for the oxides were adopted to the to 0 between D2 and D4.
metals.
The total oxide layer is about 16 nm thick, in which
4. Discussion

D1 D~ D2 D~ D4 4.1. Chemical potential diagram of Fe-Cr-O system

The metal makes contact with the air at the surface

0.6
from where the oxygen is transferred into the growing
oxide. The activity of oxygen is thereby largest at the
E
h 0.4 surface and becomes lower in a deeper region of the oxide.
The oxidation process on the metal consequently occurs
g
under chemical potential gradient, resulting in the content
0.2
gradient shown in Fig. 1.
There are normally two approaches to elucidate such
0.0 content gradient or phase stability under the chemical
0 40 80 120
potential gradient. A straight approach is the diffusion
Etching Time/rnin process analysis in which Gibbs free energy is minimized
Fig, 1. Depth profile of constituents in the oxide thin film of stainless under a certain temperature and pressure. This method is
steel. quite complicated since the material consists of many
 5. Shibagaki er aL / T h i n Solid Films 303 (1997) I 0 ] - 1 0 6 103

constituents, whose diffusion coefficients are all required diagonally the axes of the corresponding elements. For
but unknown. example, the plane P~p shown in Fig. 3 represents the
The other approach is to show stable phases on a region where a spinel-type oxide FeCrzO4 can stably exist.
chemical potential diagram, which is determined on the Presently, solid solution data have not been included in
basis of local equilibrium. By this principle, an equilibrium the database MALT2; thus, no solid solution is shown on
condition is assumed for a limited small area even if the the chemical potential diagram of Fig. 3. In fact, the Fe
system is completely in a non-equilibrium state. This and Cr metals form a solid solution and their oxides also
assumption combines both diffusion and thermodynamic form solid solutions, e.g., spinel structure between FeCr, Q
approaches for a system under the chemical potential and Fe304, and between Cr203 and Fe203. The regions
gradient. We have used the chemical potential diagram to for such solid solutions are schematically shown by dotted
analyze the phases in the oxide film. lines in Fig. 3.
According to the /zAES measurement, the oxide film is
of the order of 10 nm thick and consists of five sublayers 4.2. Chemical reaction path in oxide thin fihn
as shown in Fig. 1. The thickness of each sublayer,
corresponding to only several to a few tens of unit cells, is
too thin for a normal thermodynamic analysis, i.e., statisti- 4.2.1. Simplified phase relation
cal fluctuations cannot be neglected. On the other hand, in As the first step, we consider the phase changes in the
a two-dimensional chemical potential diagram parallel to thin film based on three phase diagrams of F e - C r - O ,
the surface, there exists a large number of atoms, so that N i - F e - O and N i - C r - O systems [9]. Since the oxygen
the statistical condition is sufficiently satisfied in this activity decreases from the surface to the inside, the com-
direction. In the following sections, it will be shown that bination of the above three-phase diagrams leads us to the
the local equilibrium approach is approximately valid for following possibility about the phases which may appear
the depth direction [6-8]. in the oxidation process at T = 700 K.
Fig. 3 shows the schematic three-dimensional chemical At the surface with the highest oxygen activity, four
potential diagram for a F e - C r - O system at 700 K. This phases--FeaO 3, Cr203, NiCr204, NiFe204--and their
diagram was calculated by using a thermodynamic database solid solutions are all possible, i.e., all or a few of them
(MALT2 [9]) published by The Society of Calorimetry and might appear. The phases which appear depend on the
Thermal Analysis, Japan. composition of the base material and on diffusion of the
As shown in Fig. 3, a polyhedron is formed in the components, as will be discussed in Section 4.2.2 to
chemical potential diagram, with faces representing stable Section 4.3. In the next region, spinel-type compounds (or
compounds. For example, pure Fe metal is stable on the solid solutions) in a N i - F e - C r - O system are possible, in
plane PFe vertical to the /x(Fe) axis and pure metal Cr is which Ni and Fe ions occupy the so-called A sites in the
stable on the plane Pc, vertical to the /z(Cr) axis. The spinel-type structure while Cr and the rest of the Fe ions
stable phases naturally change depending on the chemical occupy the so-called B sites. As the oxygen activity is
potentials, /z(Fe), /z(Cr) and /z(O). In the chemical poten- further diminished, the spinel-type compounds are decom-
tial diagram, the compound which consists of two (or posed into metals a n d / o r simple oxides in order of
three) elements is thus defined by the plane which cuts NiFe204, NiCr204, Fe304 and FeCr, O4. Corresponding to
such decompositions, Ni metal, Fe metal, Cr20 3 and Cr
metal appear one by one in the internal region; finally, the
~(Fe) base material appears (at T - - 7 0 0 K, l e O is unstable at
any oxygen pressure).
This phase sequence is, however, ambiguous; in particu-
lar, the phase which exists near the surface cannot be
specified. We thus apply a two-dimensional chemical po-
tential diagram to specify the phases in the thin film.

4.2.2. Two-dimensional chemical potential diagram

Fig. 4 indicates the two-dimensional chemical potential
diagram for the F e - C r - O system, in which the polyhedron
shown in Fig. 3 is projected onto a plane involving the
/~(O)-axis by means of a horizontal projection. The ab-
scissa of Fig. 4 thus represents the logarithm of oxygen
activity, i.e., log a(O2); the ordinate represents the loga-
/z(Cr) rithm of activity ratio of Fe to Cr, i.e., log[a(Fe)/a(Cr)],
Fig. 3. Three-dimensionaIchemical potential diagam for a Fe-Cr-O which is proportional to the difference of chemical poten-
system at 700 K. tials of Fe and Cr.
104 S. Shibagaki et aL /Thin Solid Fibns 303 (1997) 101-106

30 , ' t ' P '/,, A reaction path for the F e - N i - O system is similarly

given in Fig. 5. In this case also, four phases appear along
20 Fe / F%044~ ~ the reaction path, i.e., from the surface: (1) co-existing
J- _.,~ Fe203 phases of NiFe204 and Fe203 above log a(O2)= - 2 1 ;
"~
~ 10 "sF%*×Cr2-x04~ (2) spinel phases: Fe304, NiF%O 4 and their solid solution
at - 21 > log a(O 2) > - 2 8 ; (3) co-existed phases of Fe304
~ FeCr204 and Ni metal at - 2 8 > 1 o g a ( O 2 ) > - 3 1 ; (4) Fe, Ni
metals and their solid solution below tog a(O 2) = - 3 1 .
.= 0 Comparing Fig. 1 with Fig. 4 and Fig. 5, the point D1
._~ [~side Cr20.,' corresponds to the point where the oxygen activity is
nearly equal to log a(O 2) = - 2 1 ; similarly, the points D2',
cr D2, D3 and D4 correspond to the oxygen activities of log
DI D3 D, a(O2) = - 28, - 31, - 37, - 48, respectively.
-20 i ~ 1 i I
-60 -40 -20 0 4.2.3. Phase sequence in thin fihn
IogPo2 Based on the reaction paths of Fig. 4 and Fig. 5, the
Fig. 4. Two-dimensionalchemicaldiagramfbr a Fe-Cr-O systemplotted phase relation and the texture in the thin film are summa-
as log P(O2) vs. log[a(Fe)/a(Cr)] at 700 K. rized as follows: (1) Between the surface and DI: there
exist a corundum-type Fe203 and a spinel-type NiFe204
form a mosaic pattern. (2) Between D1 and D2': spinel-type
compounds, F%O 4, FeCr204, NiFe2Q and NiCr204, form
Yokokawa et al. [2] and Yokokawa [3] studied the a solid solution. This subtayer is probably almost homoge-
usage of the two-dimensional chemical potential diagram. neous. (Precisely speaking, the point D2' should split into
When a chemical reaction occurs at the interface of two two points since two spinel compounds, NiFe204 and
materials, its reaction path is given by a continuous, nearly NiCr204, are formed at different oxygen activities; how-
straight line, in the two-dimensional chemical potential ever the difference between the activities is very small and
diagram. The chemical reaction path is ordinarily mapped neglected here.) (3) Between D2' and D2: spinel-type
to be consistent with experimental data such as composi- compounds and Ni metal coexist; the spinel-type com-
tion, etc. pounds consist of Fe304, FeCr204 and their solid solution.
On the other hand, Clark [10], van Leo [11] and This sublayer can take a penetration-type form. (4) Be-
Yokokawa et al. [2] have discussed the phase sequence and tween D2 and D3: the spinel-type compound FeCr204,
texture for an A-B-C system in which an element A and a coexists with a solid solution of Ni and Fe. This sublayer
compound BC exist. Two possible cases were indicated: also can take the penetration-type form. (5) Between D3
(a) When a reaction path goes forward across a phase and D4: the solid solution metal exists with only one
boundary between A and BC in the phase diagram, two oxide, i.e., the corundum-type compound Cr203. (6) Be-
phases appear sequentially and their interface is flat; (b)
when a reaction path goes forward along the phase bound-
ary, two phases coexist and form a mosaic pattern in which
one phase penetrates into the other phase. 50 ' ' I '
For the interface between the Fe-Cr alloy and oxygen,
Fe F%04{ Fe20 3
the reaction path can be represented by a bold line in Fig.
20
4, which takes account of Fig. 1.
p •

Four phases lie on the line according to the decrease in

the oxygen activity: (1) a corundum-type phase: Fe203 Z
IO
above log a(O z) = - 2 i ; (2) spinel-type phases: Fe304,
FeCr204 and their solid solution at - 2 1 > tog a(O a) > Is.
.-" S,S.NiI.XFe2÷xO
4)
- 37; (3) coexisting phases of Fe metal and corundum-type 0
Cr203 at - 37 > log a(O 2) > - 4 8 ; (4) (Fe, Cr) metals and inside
_o
=
their solid solution below log a(O~)= - 4 8 .
From the simplified thermodynamic consideration (Sec- -I0
tion 4.2.1), corundum-type compounds Fe203, Cr203 and
their solid solution may be produced at the surface; how-
-20
ever, Cr did not appear at the surface in Fig. 1. Further- -60 -40 - 20 0
more, the order of the phases that appeared is qualitatively logPo2
in agreement with the depth profile of the composition Fig. 5. Two-dimensionalchemicaldiagramfor a Fe-Ni-O systemplotted
shown in Fig. 1. as log P(O2) vs. log[a(Fe)/a(Ni)]at 700 K.
 S. Shibagaki et aI./Thin Solid Films 303 (1997) 101-106 105

low D4: no oxide exists; thus, the structure should be Based on this assumption, the following diffusion pro-
homogeneous. Inside the alloy, both the structure and cesses are conjectured: (1) Fe can diffuse to the surface
composition approach that of the base material. through all of the four inner sublayers, i.e. (Cr, Fe)203-
Fe-Ni, FeCrzO4-Fe-Ni , Fe304-FeCrzO4-Ni ~ Fe304-
4.2.4. Comparison with experimental results FeCr204-NiF%O4-NiCr204 layers. Fe reacts with the
The Cr and Ni ions in oxides at high temperature other elements existing in each sublayer; at the surface, it
always have charges of + 3 and +2, respectively, while forms the corundum type of Fe:O 3 by reacting with oxy-
Fe ions change their charge from + 3 to + 2 depending on gen. ( 2 ) N i diffuses in a similar way to Fe. Although Ni
the oxygen activity. On the basis of the phases predicted can reach the surface, the amount is much smaller than that
above, the charge of the Fe ion can be estimated as a of Fe because of its smaller diffusion coefficient. At the
function of depth, as indicated by a broken line in Fig. 2. surface, Ni reacts with Fe and O, forming the spinel type
The predicted result is qualitatively in accordance with the of NiFe204. (3) Cr can diffuse to the Fe304-FeCr204-
experimental data. NiF%O4-NiCr204 (spinel) sublayer through the three in-
Similarly, taking the composition and the ionic charges ner sublayers. Because of the very small diffusion coeffi-
of the constituents into account, the depth profile of the cient of Cr, the diffusion stops at the spinel sublayer, so
oxygen content can be calculated from those obtained the phases near the surface cannot involve Cr. The sub-
experimentally for Fe, Cr and Ni. The result is shown by a layers through which Cr diffuses all involve the metal
broken line in Fig. 1. The calculated oxygen content is phases: Ni a n d / o r Fe, suggesting that Cr diffuses probably
almost in agreement with that obtained experimentally. through the metal phases to the spinel phase.
These agreements about the valence of Fe and the In this way, the chemical reaction paths in Fig. 4 and
oxygen content thus indicate that the reaction path is Fig. 5 result from the diffusion property of the thin film; in
reasonably determined in the two-dimensional chemical other words, the point corresponding to the surface in Fig.
potential diagram and the principle of local equilibrium is 4 and Fig. 5 is decided by the diffusion coefficients of the
valid in the thin film of a few nanometers thick. metals which are in order of Fe > Ni > Cr.
Strictly speaking, the oxygen content should jump As noted in Fig. 1, the Fe content increases in the
steeply at the points where the phase-changes occur; how- direction of the surface between D2 and D1, and the Cr
ever, the experimental data are somewhat diffuse as shown content similarly increases between D3 and D2. The mass
in Fig. 1. This might be partially due to experimental transfer thus occurs in the direction for the content to
limitations arising from the escape length of Auger elec- {ncrease, though the direction of the mass transfer accords
trons from the surface and from the ion-etching technique with that for the chemical potential to decrease. This
with an Ar ion gun. phenomenon is known as "up-hill diffusion": it is thus
considered that the up-hill diffusion for Fe and Cr occur in
4.3. Diffitsion proper O, in thin fihn the present thin film.
The present results of the analysis on diffusion in the
The reaction path is normally decided by a diffusion oxide thin films are consistent with recent results of simi-
property of the reactant materials and the reaction prod- lar analysis by Yokokawa et al. [13] on reactions of alloys
ucts. Here, we discuss the above-mentioned chemical reac- with alkali carbonates.
tion path based on the diffusion property.
As shown in Fig. 3, the chemical potential for oxygen
/~(O) gradually decreases from the surface toward the
inside along the reaction path; in contrast, those for metals 5. Summary
/z(Fe) and /x(Cr) gradually decrease from the inside to-
ward the surface. Mass transfer for the oxygen thus occurs An oxide thin film formed on a stainless steel was
from the surface toward the inside while the mass transfer analyzed by ttAES to evaluate the depth profiles of com-
for the metals oppositely occurs from the inside toward the position and valence of Fe. From thermodynamic consider-
surface. ation, it was indicated that a chemical potential dia~am
In general, a diffusion coefficient depends on the con- could be well applied to elucidate the depth profiles of
tent of defects, microstructure, etc. As described in Section both composition and valence; this result suggested that
4.2, the microstructure in the present system is so compli- the thermodynamic approach based on a principle of local
cated that the diffusion coefficients cannot be simply equilibrium was valid for the thin film of 10 nm thick. A
stated. It is, however, well known that a diffusion coeffi- chemical reaction path in the thin film was discussed in
cient for Fe is much higher than that for Cr in the relation to the diffusion property.
spinel-type structure [12]. Thus, we assume hereafter that, The oxide film consisted of five sub-layers of 2 to 4 nm
in the oxides, the intrinsic diffusion coefficient for Fe is in thickness, and the phases which appeared in them were
much higher than that for Cr; in addition, Ni takes an specified. It was indicated that an up-hill diffusion oc-
intermediate value. curred in the oxide thin film.
106 S, Shibagaki et aL / Thin Solid Films 303 (1997) 101-106

References [6] C.S. Giggins, F.S. Petit, J. Electrochem. Soc. 118 (1971) 1782.
[7] E.J. Felton, F.S. Petit, Oxid. Met. 10 (1976) 189.
[1] T. Isagawa, M. Ogure, T. Imaoka and T. Omi, Proc. 15th Syrup. on [8] H.M. Hindam, W.W. Smeltzer, J. Electrochem. Soc. 127 (1980)
ULSI Ultra Clean Technology, Advanced Wet Cherm Proc. III 1622.
(1992) 196-215. [9] Thermodynamic data base MALT2, published by Soc. Calorimetry
[2] H. Yokokawa, N. Sakai, T. Kawada and M, Dokiya, Proc. Internat. and Thermal Analysis JPN., Kagakugijutsusha, 1992.
Symp. on Mater. Chem. Nucl. Envir. (1992) 547-554. [10] J.B. Clark, Trans. Metall. Soc. AIME 27 (1963) 1250.
[3] H. Yokokawa, Progress of new thermodynamic measurement, 1 [11] F.J. van Loo, Prog. Solid. Sta. Chem. 20 (1990)47-99,
(t990) 9-22. [12] J.D. Hodge, J. Electrochem. Soc. 125 (1978) 55C-57C.
[4] C.N.R. Rao, D.D. Sarma, M.S. Hedge, Proc. R. Soc. London A370 [13] H. Yokokawa, N. Sakai, T. Kawada, M. Dokiya, J. Electrochem.
(1980) 269-280. Soc. 140 (1993) 3565-3577.
[5] Experimental results and their analysis will be published in detail
elsewhere in relation to the texture and thickness of each sublayer.