Materials Sci And E A

Materials Science and Engineering A 415 (2006) 21–32

Relation between the oxidation mechanism of nickel, the microstructure and mechanical resistance of NiO films and the nickel purity I. Oxidation mechanism and microstructure of NiO films A.M. Huntz a,∗ , M. Andrieux a , R. Molins b a

LEMHE, CNRS UMR 8647, Universit´e Paris XI, Centre d’Orsay, F-91405 Orsay, France b ENSMP, Centre des Mat´ eriaux, CNRS UMR 7633, BP 87, F-91003 Evry, France Received in revised form 1 May 2005; accepted 1 August 2005

Abstract The effect of impurities on the oxidation mechanism of nickel and on the mechanical characteristics of NiO films was studied on two industrial nickel grades compared to a pure nickel. In this part, influence of impurities on the oxidation mechanism and on the NiO film microstructure will be detailed. The oxidation mechanism, especially studied at 800 ◦ C in air, was clarified using complementary techniques, SEM and STEM microstructural and analytical investigations, XPS analyses, profilometry, oxygen isotopic exchange and SIMS. Depending on the Ni purity, the morphology of the oxide scale and the porosity amount differs. Duplex oxide layers and internal oxidation are observed on industrial grades, while a simple NiO film grows on pure nickel without internal oxidation. An oxidation mechanism, which differs according to the presence or not of impurities in the nickel substrate, is proposed. The mechanical characteristics will be characterised and discussed in the following paper (part II) in relation with the microstructural modifications induced by changing the nickel purity. © 2005 Elsevier B.V. All rights reserved. Keywords: Oxidation; Nickel; Impurities; NiO; Microstructure; Diffusion

1. Introduction Though nickel oxidation has been largely studied [1–17], the oxidation mechanisms of nickel are not really clarified and discrepancies still exist in the literature data. This is probably due to the differences in purity of studied nickel grades, in the surface preparation and roughness, in furnace atmosphere and in heating procedures . . .. Indeed, all these parameters may have an influence on the oxidation behaviour, but, at this date, the effect of impurities on the oxidation mechanism and on the morphology of the oxide scale is not yet clarified. The characterisation of oxide morphology and of the oxidation mechanisms depending on many parameters such as the oxidation conditions, the porosity of the oxide films, the initial surface roughness and the nickel purity, is of great importance to explain mechanical behaviour and deduce mechanical characteristics of thin films. This part will help us to

∗

Corresponding author. Tel.: +33 1 69 15 63 18; fax: +33 1 69 15 48 19. E-mail address: [email protected] (A.M. Huntz).

0921-5093/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2005.08.225

interpret mechanical results displayed in part II. Moreover, such a study presents an interest not only for a fundamental meaning of environmental aggression phenomena, but also as Ni-based materials can behave similarly to nickel after depletion in more oxidising elements. So, this work was focused on the effect of impurities in a nickel substrate on both the oxidation mechanisms and the mechanical characteristics of NiO scales. The purpose of the first part is related to the determination of the oxidation mechanism, while the second part will be allocated to the determination of the mechanical characteristics depending on the film microstructure and growth mechanism. In order to point out the effect of impurities, the oxidation in laboratory air of three nickel grades was studied, two industrial grades, while the third one, qualified of pure (so called “pure”), was elaborated in “Ecole Nationale Sup´erieure des Mines de Saint Etienne, France” (ENSMSE). The oxidation behaviour was followed by thermogravimetry in isothermal conditions and the oxide films were analysed using many techniques allowing to evidence their microstructure and their chemistry at various scales. A particular attention was paid to the first oxidation stages,

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

22

which can have a great influence on crack initiation in the oxide film.

2. Experimental Three nickel grades were studied; two of them are industrial grades “Imphy and MSX nickel”, whose composition given by the industry is gathered in Table 1. Note that for Imphy nickel, all the impurities are not mentioned. For instance, for Imphy nickel, Mn is not given as impurity by the industry, but was detected after oxidation in the oxide layer. The third nickel grade “ENSMSE nickel” was provided as a very pure single crystal and contains less than 10 ppm of impurities. The nickel grade provided by Imphy S.A. is as-hot-wired drawn, while nickel provided by MSX was previously hotrolled. Both industrial grades are polycrystalline with a grain size around 30–50 ␮m containing a sub-structure of dislocation cells resulting from the elaboration process. These grades contain various impurities, but they differ according to their type and content, namely silicon, titanium and manganese which are more present in MSX nickel (see Table 1). In order to obtain a pure polycrystalline nickel, the single crystal of ENSMSE Ni was cold-rolled, then heat-treated for 24 h at 920 ◦ C in an argon-hydrogen mixture. Observations by electron microscopies and analyses (EDX, energy dispersive Xray analysis and XPS, X-ray photo-electron spectrometry) did not reveal any impurity and the resulting grain size of this pure nickel was about 100 ␮m. However, a substructure of dislocation cells was still remaining, the annealing treatment having not been long enough. Prior to oxidation tests, all the samples were mechanically polished using SiC paper down to 1200 grade, then ultrasonically cleaned in ethanol. The oxidation kinetics tests were performed in a high sensibility SETARAM thermobalance (sensitivity ±1 ␮g), in air between 700 and 900 ◦ C, but at 800 ◦ C for most experiments. This allowed to choose adequate oxidation treatments for the specimens to be studied by bending-tests, so that oxide scales of the same thickness for the three nickel grades could be obtained. After oxidation tests at 800 ◦ C, all samples were observed using scanning electron microscopy (SEM) on surface and cross section, and EDX analyses were performed to detect impurity rich precipitates. Cross-sectional thin foils were also prepared on the three oxidised nickel grades and characterized using a transmission electron microscope (Tecnai F 20 ST) equipped with analytical spectrometers. The analytical resolution was 1 nm, allowing a better detection of segregated impurities.

The first oxidation stages were studied by in situ XPS analyses and by interferometric profilometry (performed in “laboratoire de physico-chimie de l’´etat solide” in University Paris Sud, France). For in situ XPS experiments, the sample surface was carefully polished. Then, once introduced into the XPS chamber, it was cleaned by an Ar+ sputtering and heated up to 600 ◦ C in ultra-vacuum. At last, oxygen was introduced at 10−5 atm for 1, 5 or 15 min and spectra were recorded during dwell time at 600 ◦ C and after cooling at room temperature. 3. Results 3.1. Kinetics Fig. 1a collects the weight gain versus time obtained for Imphy industrial nickel at three temperatures. In each case, the oxidation kinetics can be represented by a parabolic law. Curves obtained at 800 ◦ C for the three nickel grades are given in Fig. 1b and indicate that pure nickel oxidizes faster than nickel of industrial grades. Similar observations were made at 700 and 900 ◦ C. There is no significant difference between the two industrial nickel grades. Fig. 1c shows that the oxidation kinetics can be fitted to a parabolic law:
2 M = a + kp · t (1) S where a is a small constant term, kp is the parabolic constant whose values are reported in Table 2, and t is the oxidation duration. The oxide thickness xNiO was calculated from: xNiO =

M/S · MNiO MO · ρNiO

(2)

where MNiO and MO are the molar mass of NiO and O respectively, and ρNiO is the volumetric mass of NiO (taken as 6.85 g/cm3 ). Eq. (2) assumes that all oxygen is incorporated in the scale (no internal oxidation) and that the oxide film is compact. 3.2. Morphology of the oxide layers Fig. 2 shows that the oxide scales formed on specimens of industrial grades (Fig. 2a and b) are duplex, consisting in two distinct films, an outer film made up of columnar NiO grains, and an inner one of equiaxed NiO grains. On the contrary, the oxide formed on pure nickel is simple, only composed of one oxide layer. Contrary to references [3,4], the duplex structure for the two industrial grades and the simple structure for the pure nickel were observed whatever the oxidation duration and

Table 1 Composition of nickel provided by Imphy S.A. Ni grade

Imphy Ni MSX Ni

Element (wt%) Ni

Co

C

Ti

Si

S

P

99.700 99.480

0.01

0.058 0.014

0.089

0.01 0.102

0.003 <10 ppm

<10 ppm

Mo

Mn

Cu

Fe

0.266

0.005 0.034

0.01 0.016

0.084

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

23

Fig. 1. Evolution of the weight gain per surface unit by oxidation in air of (a) Imphy nickel at three temperatures, (b) the three nickel grades at 800 ◦ C and (c) same as (b) in quadratic coordinates.

temperature, in the range 700–900 ◦ C. It could be noted that Haugsrud and Pedraza [15,16] worked with pure nickel grades and observed duplex structures for the oxide scale. By comparing their nickel purity to ours, it appears that their nickel is less pure than ENSMSE Ni of this study. It clearly shows that the oxide morphology is very sensible to the degree of nickel purity. On the industrial nickel grades, the thicknesses of the two layers are always near the ratio 1:1, and for a same oxidation time, the oxide scale formed on industrial nickel grades is nearly twice thinner than the oxide formed on pure nickel. On industrial grades, the porosity of the inner layer increased

with time and seemed slightly more important on MSX Ni than on Imphy Ni. On the contrary, as shown by SEM surface observations, the porosity of the outer layer decreased with the oxidation time [5–8]. For industrial grades, under the oxide layers, internal bulk and intergranular oxidation were observed (Fig. 3). Internal oxidation is slightly more pronounced for MSX nickel (Fig. 3b) than for Imphy Ni (Fig. 3a and d), whereas intergranular oxidation seems to develop deeper in Imphy grade. This can be related to a higher carbon content. As shown on Figs. 2c and 3c, there is no internal oxidation for pure nickel.

Table 2 Kinetics constants deduced from parabolic law Nickel

T (◦ C)

kp (mg2 cm−4 s−1 )

Oxidation time (h)

Oxide thickness (␮m) calculated with ρ = 6.85 g/cm3

Imphy

700 800 900

6.2 × 10−7 4.6 × 10−6 3.2 × 10−5

93 66 22

3.1 7.4 10.7

MSX ENSMSE

800 800

7.3 × 10−6 6.3 × 10−5

66 46

9.3 22.2

24

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

Fig. 2. Transverse sections of NiO scales formed at 800 ◦ C on (a) Imphy Ni, (b) MSX Ni and (c) pure Ni.

Fig. 3. Transverse sections and observation of the underlying substrate with or without internal oxidation. Oxidation at 800 ◦ C: (a) Imphy Ni, (b) MSX Ni, (c) pure Ni and (d) STEM general view of the scale (dark field image, DF, Imphy grade).

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

25

Fig. 4. EDX analysis of precipitates in industrial nickel.

3.3. EDX chemical analyses After oxidation sequences allowing to develop oxide films of 1–10 ␮m, only Ni and O were detected at the surface of the samples by EDX. No outer diffusion of impurities was evidenced. On the contrary, when analysing cross-sections, many internal precipitates were observed (see Fig. 3a, b and d) in the substrate and especially along grain-boundaries of industrial nickel grades. As shown in Fig. 4, these precipitates were mainly composed of Ti and Mn oxides, with some Si. A higher content of precipitates was present in MSX grade than in Imphy grade (compare Fig. 3a and b). 3.4. Transmission electron microscopy (TEM) Cross-sections of specimen of the three studied Ni grades oxidised for 3 h at 800 ◦ C were investigated through the oxide layers and along the interfaces, oxide–metal interfaces but also oxide–oxide interfaces in the case of oxide films formed on industrial grades, and finally in the underlying substrate. Elemental EDX profiles were performed with a probe size of 1 nm and a step size of 1 nm in order to detect any interfacial or intergranular precipitation and segregation. On both oxidised industrial grades, the duplex structure of the oxide scale was confirmed with an outer columnar layer 1–2 ␮m thick, with grains 300 nm large, and an inner layer made of equiaxed grains (Fig. 3a and 5) whose thickness varies (from 0.5 to 1.5 ␮m), due to large convolutions of the internal Ni/NiO interface. For MSX nickel grade, some tiny precipitates enriched in Si and/or Mn were detected in the outer layer, but neither segregation nor precipitation were detected along columnar grain boundaries or at the interface between the two oxide layers (Fig. 6a). Through the inner oxide layer whose grain size is about 100 nm, many pores were observed (Fig. 6b). Large cavities are also present at the oxide–metal interface, associated to metallic nickel areas between the cavities (Fig. 6c, see arrows). No segregation was detected at a nanometric scale at the NiO–Ni interface. Note that the grain size of nickel, under the oxide scale is about 1 ␮m while the grain size of nickel in the bulk is around 50 ␮m.

Fig. 5. Duplex structure of NiO (STEM-DF image, Imphy grade).

Concerning the oxidised Imphy nickel, there is no specific microstructural difference compared to oxidised MSX nickel. The oxide scale is duplex with a planar interface between outer and inner layers, neither segregation nor precipitation occurs at oxide grain boundaries or along oxide/oxide interface and a great amount of porosity is observed through the inner layer. Contrary to the observations made on the MSX grade, silicon and a few Mn were sometimes detected at the Ni/NiO interface (Fig. 7) as segregated elements or as precipitated in oxidized particles. For both industrial grades, the inner NiO/outer NiO interface is rather flat (Fig. 8a) as previously shown by SEM, and without any segregation, while the Ni/NiO interface is very convoluted and irregular with large cavities (Fig. 8b). Below the NiO/Ni interface, in the underlying nickel, transgranular precipitates essentially rich in Mn, Si, Mg and O (Fig. 9a) were observed, while precipitates enriched in Mn, Si, Mg, Ti and O were observed (Fig. 9b) along the grain boundaries, as the result of important internal bulk and intergranular oxidation. Always for both industrial grades, large cavities were observed at the metal/oxide interface after limited oxidation time (Fig. 10). Under the oxide, the substrate displays the same microstructure than in its as-received state. No recrystallisation occurred, perhaps due to a pinning effect by precipitates. The observed microstructure is quite different in pure nickel. Only one NiO layer, 5 ␮m thick, is observed with large equiaxed grains varying between 200 and 500 nm (Fig. 11). The Ni/NiO interface is rather flat, only facetted by the grain shape (Fig. 12a). This facetted interface is possibly related to some crystallographic orientation relationship between Ni and NiO, resulting from initial stages of oxidation. No cavities were detected at this interface. However, there are many pores through the oxide layer (Fig. 12b) often with a geometrical shape. Other chemical elements were detected neither within the oxide layer, nor at the oxide/metal interface and internal oxidation was evidenced

26

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

Fig. 6. STEM transverse sections, MSX grade (a) outer/inner NiO interface, (b) grain size and porosity through the inner layer, (c) interfacial cavities (marked by arrows) and Ni extrusion (EDX analysis performed between the two cavities).

neither in the bulk nor along grain boundaries of the underlying substrate. 3.5. First oxidation stages and in situ XPS chemical analyses It was observed that the spalling at the oxide/oxide interface was easier for Imphy nickel than for MSX and segregation of impurities at the initial surface of nickel was suspected to have an influence. Thus, “in situ” XPS analyses were performed before oxidation, after heating at 600 ◦ C in ultra-vacuum and after a limited oxidation duration. Note that the XPS device

cannot heat over 600 ◦ C, and analyses at 600 ◦ C only indicates what occurs at the very beginning of oxidation and what can be the consequence of the first oxidation stages. Contrary to “accelerated tests” which are generally performed in most studies in order to simulate the life-time of systems, our XPS tests, that can be considered as “delayed tests”, give information on first oxidation stages. In the case of commercial materials, no segregation was detected before oxidation. After oxidation at 600 ◦ C, for instance, up to 15 min or even one hour at an oxygen pressure of 1 Pa, it was surprising to observe that, for industrial grades, the experimental peak of nickel was composed of Ni2+ (peak at 854.6 eV) and Ni0 (peak at 852.6 eV) [11], while

Fig. 7. STEM view of NiO/Ni interface and associated EDX silicon profile (Imphy grade).

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

27

Fig. 8. STEM images (Imphy grade, bright field, BF), (a) outer/inner NiO interface and (b) inner NiO/Ni interface.

on pure nickel, the Ni2+ peak only was present. It was verified, by thermogravimetry performed in pure argon at 600 ◦ C, that the oxide thickness, developed in these in situ oxidation tests, was greater than the XPS analysis depth. Indeed, after oxidation for 1 h in pure argon, the NiO scale thickness was equal to 290 nm whereas the allowed depth analysed by XPS was about 5 nm. Then, the appearance of the oxide surface was observed by interferometric profilometry and 3D images were obtained on industrial and pure nickel after 15 min oxidation at 600 ◦ C in 10−5 atm O2 in the XPS chamber. It clearly appeared that

walls of about 100 nm thick were present along grain boundaries of the underlying metal in the case of industrial nickel (Fig. 13a) as also along some grain sub-boundaries (Fig. 13b), while the surface of pure nickel remained flat [11] (Fig. 13c). EDX analyses allowed to detect that these walls were only made of metallic nickel, while nickel oxide was present on the bulk of Ni grains [11]. Further sputterings by Ar+ ions in the XPS chamber were successively repeated. Observations by interferometric profilometry after each sputtering showed that the height of nickel walls along Ni grain boundaries progressively decreased,

Fig. 9. STEM views, Imphy grade: (a) internal precipitate (arrows) and EDX spectrum showing the presence of Si, Mn and O; (b) intergranular precipitate along grain boundaries of Ni and associated EDX profile showing Mn, Si, Mg and O.

28

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

Fig. 10. BF-STEM view (Imphy grade) of an interfacial cavity and EDX analysis of precipitate hung to the cavity (rich in Mg, Si, Mn and O).

the observations made at 600 ◦ C. Moreover, the Ni walls along grain and sub-grain boundaries were also observed after a short oxidation time at 800 ◦ C in air [7]. 4. Discussion on the oxidation mechanism Concerning the order of magnitude of the parabolic constants (Table 2), our values are in the range of values given in the literature [1,3–6,15–17]. Haugsrud and Pedraza [15,16], who worked with pure nickel (clearly purer than our industrial grades), found parabolic constants of the same order of magnitude than the kp value of our pure nickel. So, it is important to note that the pure nickel oxidation constant is greater than that of industrial grades and, simultaneously, the oxide scale thickness is about twice greater. 4.1. Initial stages

Fig. 11. Pure Ni, DF-STEM view of the simplex layer.

and, simultaneously, precipitates identified by SEM–EDX as Mn or Ti rich oxides appeared along the grain boundaries, as seen in Fig. 14. These precipitates were also viewed by TEM. Post-mortem examinations by XPS of the surface of the oxidised samples after treatment at temperatures of 800 ◦ C always agree

At first, let us consider the initial oxidation stages, as previously mentioned and studied by “in situ” XPS experiments associated to interferometric profilometry and EDX analyses. In the case of industrial grades, impurities such as Mn, Mg, Ti, Si, which have a strong affinity for oxygen were present mostly as intergranular precipitates in the underlying metal. These impurities promoted inward oxygen diffusion, and, during the initial stage of oxidation, internal oxidation of manganese

Fig. 12. Pure Ni: (a) DF-STEM view of NiO/Ni interface and (b) pores through the NiO layer.

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

29

Fig. 13. Surface of the oxidised samples observed by interferometric profilometry after 15 min oxidation at 600 ◦ C: (a and b) Imphy nickel, grain boundaries and sub-boundaries, respectively, and (b) pure nickel.

and titanium occurred along the grain boundaries of nickel while NiO crystallites formed at the nickel grain surface. The increase in volume due to internal oxidation induced an outward flow of metal and extrusion of metallic nickel on the outer surface (Fig. 13), extrusion which appeared as walls on grain boundaries of the underlying nickel, while the surface of grains were

covered of NiO islands. Thermo-elasto-plastic calculations were performed using finite element methods to verify the possibility to form nickel extruded walls. Taking into account the volume of each phase and the mechanical properties of NiO, Ni and precipitates, calculations confirm such extrusion of nickel which were indeed observed using TEM (cf. Fig. 6c). Such a particular reactivity induced an additional driving force for oxygen. 4.2. Comparison of oxide thickness calculated from thermogravimetric results or measured in SEM When comparing the oxide thickness calculated according to Eq. (2) from thermogravimetric data (Fig. 1b) and the oxide thickness determined by SEM observations (see all the related micrographs), a slight scattering can be remarked. This is not surprising as values deduced from thermogravimetric data do not take into account several phenomena discussed in the following:

Fig. 14. SEM observation of intergranular oxide precipitates found after sputtering of the previous metallic nickel walls formed along grain boundaries during the first oxidation stages.

(i) The calculations were made considering the volumetric mass given for the massive oxide (i.e. 6.85 g/cm3 deduced from crystallographic data). As a great density of pores were detected through the oxide films, the real volumetric mass is smaller and the oxide thickness greater than the calculated value. This is sometimes the case, but not always.

30

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

(ii) As shown in Fig. 3, internal oxidation occurred in the bulk (particularly for MSX nickel, Fig. 3b) and especially along grain boundaries of both industrial nickel grades. Thus, the oxygen content of the film was overestimated in the calculation from thermogravimetric data and consequently the oxide film thickness also. This parameter induces an opposite variation when compared to the previous one. (iii) In fact, calculation from thermogravimetric data corresponds to an oxide free growth in the direction perpendicular to the substrate surface (the surface of the oxide film is taken as S, the surface of the substrate). Thus, it corresponds to an anisotropic formation of the oxide film by nickel ions piling up at the outer surface of the scale [6]. In such a theoretical case, a nickel layer of 1 ␮m would give an oxide layer of 1.65 ␮m (taking 6.85 and 8.9 g/cm3 for the volumetric mass of NiO and Ni, respectively). In this case also, the oxide film would be stress-free. Such events could correspond to the growth of NiO on pure nickel, but cannot satisfy the case of industrial nickel for which, as it will be detailed further on, a non negligible part of the oxide growth is associated to inward growth by oxygen diffusion. It means that, at least for industrial substrates, the thickness calculated from thermogravimetric data is overestimated. Consequently, it appears that, according to the importance of each of these three phenomena, differences can appear in the oxide thickness measurement. 4.3. Oxide growth mechanism After these considerations, it is important to discuss about the oxidation mechanism according to the nickel purity. Pure nickel developed one simple film, while the industrial grades developed, at the same time, two oxide layers, one outer columnar layer and an inner one with equiaxed grains (see Fig. 2). The size of equiaxed grains for pure nickel was higher than those of inner scales for industrial grades. It is expected that the oxide growth on industrial grades is controlled by countercurrent diffusion, the outer columnar layer being due to outward cationic diffusion, and the inner one to inward anionic diffusion. Such an assumption agrees with the shape of the interfaces: the oxide/oxide interface, from which the outer layer is growing by cationic diffusion, has a flat surface corresponding to the initial surface of nickel prior to oxidation. As observed for other systems (alumina or zirconia forming materials [18,19]), the oxide/metal interface is convoluted due to the oxygen reactivity at the inner interface. Such convoluted shapes correspond to a minimisation of the energy of the system in relation with the stresses generated by the inner growth of an oxide layer. In the case of pure nickel, the oxide layer made of equiaxed grains seems to be related to a cationic growth mechanism. Moreover, the facetted morphology of the interface seems to be due to a recession of this later, thus avoiding cavity formation. In order to verify the assumption concerning the growth mechanism of NiO on industrial grades and to determine the growth mechanism on pure nickel, isotopic exchange tests were made by successive 16 O2 and 18 O2 treatments (Roberval Labo-

Fig. 15. 18 O2 penetration profiles in NiO scales formed at 800 ◦ C on the three nickel grades (16 O2 , 1 atm, 7 h then 18 O2 , 10−2 atm, 2 h).

ratory, UTC, France), at 800 ◦ C, in the same conditions for the two industrial grades and the pure nickel [11]. The depth profiles were established by secondary ions mass spectrometry (SIMS), CNRS Bellevue, France, with a Cs+ primary beam. The sputtering rate of the nickel oxide was determined from the depth of the crater measured by profilometry. The obtained profiles are gathered in Fig. 15 and show that the oxidation mechanism is different according to the nickel purity. Indeed, it clearly appears from these depth profiles of 18 O, that in case of pure nickel, 18oxygen is just localised at the outer part of the scale due to its reaction with Ni2+ ions coming from the nickel sample, while, for the industrial samples, its penetration in the oxide is very deep, down to the two interfaces (oxide/oxide then metal/oxide) where it laterally extends. This verifies that, for industrial grades, the NiO scale grows by simultaneous counter-current diffusion of Ni2+ cations and O2− anions or gaseous oxygen, while for pure nickel, the NiO layer grows only by cationic diffusion. Due to this growth mechanism, the Ni/NiO interface on pure nickel is fairly flat, only facetted by the oxide grains. Several questions arise from this result. The first one concerns the correlation between the growth rate and the diffusion data of NiO. According to Ref. [1], at 800 ◦ C, the nickel diffusion in the bulk of NiO is somewhat 109 times higher than the oxygen bulk Ni ∼ 104 –106 DO . diffusion, and along grain boundaries, DNiO NiO Moreover, the comparison of parabolic oxidation constants values and diffusion coefficients of nickel in NiO indicates that the oxide scale growth cannot be controlled by bulk diffusion only, but is ensured by an effective diffusion taking into account bulk and grain boundary diffusions. The effective diffusion coefficients can be calculated according to Hart equation [20]: Deff = (1 − f )Db + fDgb where f is the fraction of atoms diffusing along grain boundaries, calculated as f = 3δ/Φ (δ being the grain boundary width conventionally taken as 10−7 cm) and Φ the oxide grain size, varying between 100 and 300 nm in our case. Considering the pure nickel with a scale growth ensured only by cation diffusion, it leads to an effective diffusion coefficient around 3 × 10−11 cm2 s−1 , i.e.

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

to an oxide thickness, approximately calculated by x2 = 4Deff t, of the order of 20 ␮m after 48 h oxidation. Such approximating considerations agree with the microstructural observations and confirm that on pure nickel, at 800 ◦ C, the oxide scale grows by cationic diffusion in the bulk and along grain boundaries of the NiO film. Now, considering the NiO scale on industrial nickel, two questions have to be answered: how can oxygen reach the inner interface of the scale when the diffusion coefficient of O2− ions is smaller than the diffusion coefficient of Ni2+ ions by several order of magnitudes and, secondly, why the oxide film is thinner when two counter-current diffusions have to be taken into account. The first question can be solved if gaseous diffusion is considered instead of, or simultaneously to, ionic diffusion for oxygen. This is given possible due to the great porosity in the oxide films and it is confirmed by the presence (Fig. 15) of oxygen peaks along the interfaces (oxide/oxide and oxide/nickel interfaces). Thus, while Ni2+ ions ensure the outward growth of the oxide scale, gaseous oxygen goes through the porous oxide film and allows the inward growth of the scale. This gaseous oxygen diffusion was already suggested by Graham et al. [17]. Moreover, it is important to note that, in the case of industrial nickel grades, impurities in the substrate which have a great affinity for oxygen (Si, Ti, Mn) act as a driving force promoting inward diffusion of oxygen during the first stage of oxidation, and initiating anionic growth of equiaxed NiO film. This oxide film progressively formed into the metallic substrate is subjected to compressive stresses [21] which must induce a decrease of the diffusion coefficients, enhancing the barrier effect when compared to that of the oxide film formed by cationic diffusion (on pure nickel). Ni cations have now to go through this barrier and consequently the oxide growth is reduced. Thus, the whole thickness of the duplex scale is smaller than the thickness of the simple scale. Such a difference of behaviour of the oxide layers formed on the two types of nickel grade will be encountered also when considering the mechanical characteristics. Nevertheless, an effect of silica precipitated along the nickel/oxide interface [22] cannot be completely excluded though TEM observations did not evidence a continuous silica film. Another question subsists: why the oxide layer growing by cationic diffusion on pure nickel is equiaxed while it is columnar on industrial grades? In the present case, it can be noticed that the diameter of the columnar grains is of the order of magnitude of the diameter of the sub-grains initially observed in the substrate. In the case of industrial nickel grades, the grain sub-boundaries are marked with tiny metallic precipitates (some nanometers in size), as are the grain boundaries, and it can be suggested that the oxidation mechanism along sub-boundaries is the same as for grain boundaries: in the initial oxidation stages, oxygen penetrates in sub-boundaries leading to impurity oxidation and nickel outward flow and extrusion. This was observed in some cases along dislocation alignment by profilometry (see Fig. 13b). Such phenomena correspond to anisotropic oxidation and induce a columnar growth of the oxide grains. In the case of pure nickel, as the sub-boundaries or grain boundaries are not marked by precipitates, the oxide nucleation is homogeneous which leads to an equiaxed oxide film structure.

31

In all cases, due to cationic diffusion, cation vacancies are created and can coalesce to form the pores observed through the oxide layers (see Figs. 6b and 12b). Vacancies can also arrive at the oxide/metal interface. For the pure Ni, creep of the substrate is allowed and the interface recession seems to compensate the vacancies, as no interfacial pore was detected. This justifies the facetted interface. In the case of industrial grades, most of the vacancies resulting from cationic growth (diffusion in the outer oxide layer) are injected in the inner oxide layer (see Figs. 5, 6b, 8a) and creep of the substrate is avoided by grain boundary pinning by precipitates. In this case, vacancies can condense at the metal/oxide interface to form voids (Figs. 6c and 10). This formation of interfacial cavities can then enhance the anionic contribution. Moreover, some pores can result from the oxidation of carbon and gaseous release (CO, CO2 ). Indeed, the effect of carbon, even in small amount, was already mentioned by other authors [13,14], and carbon can also be responsible for the important voids, i.e. cavitation, which are observed in the industrial nickel substrates (Figs. 3d and 6c). This was not observed for pure nickel, may be because our pure nickel contains less than 10 ppm impurities. Concerning the internal oxidation into the industrial nickel substrates, it is clearly related to the presence of oxidisable impurities as Si, Mn and Ti. Indeed, no internal oxidation was observed in pure nickel. The internal oxide deepness slightly differs between the two industrial grades as also the localisation of the oxide precipitates (Fig. 3). In case of MSX nickel, internal oxide particles are present in the bulk and along grain boundaries while they are mainly localised along grain boundaries (but at a deeper extent) in the Imphy nickel. This can be due to the carbon content higher for Imphy grade whereas metallic impurity content is higher for MSX grade. According to the order of magnitude of the internal intergranular oxidation depth, it leads to a grain boundary diffusion coefficient of oxygen in nickel of the order of magnitude of 10−11 cm2 s−1 which agrees with literature data [23] concerning oxygen diffusion in nickel at 800 ◦ C. 5. Conclusion The oxidation behaviour of nickel is well-known to depend on many parameters such as surface preparation, atmosphere, heating procedure, oxidation temperature, presence of impurity, etc. But, at this date, the effect of impurities on the oxidation mechanism of nickel is not yet clarified. It was the objective of this first part to clarify the influence of impurities on the morphology of the NiO scales and on the oxidation mechanisms. It was possible by coupling microstructural observations at various scales, analytical techniques and kinetics studies. Three nickel grades were selected, two industrial ones and a very pure nickel for which no impurity was detected, even on thin foils, neither at the initial state, nor after thermomechanical treatments or oxidations performed mainly at 800 ◦ C in air. Both the scale morphology and the oxidation mechanism are clearly modified according to the presence or not of impurities. For pure nickel, without any impurity detected, a NiO film grows by outward diffusion of nickel, leading to the presence of a sim-

32

A.M. Huntz et al. / Materials Science and Engineering A 415 (2006) 21–32

ple oxide layer with equiaxed grains. The growth rate of this oxide layer is controlled by effective diffusion of Ni2+ , i.e. contribution of bulk and grain-boundary diffusion. When impurities are present, i.e. case of commercial nickel grades, two oxide layers are formed: an outer one with columnar grains due to cationic diffusion, and an inner one with equiaxed small grains formed by anionic diffusion. The growth rate is controlled by diffusion of Ni2+ through the equiaxed inner oxide layer which acts as a good barrier, so that the growth rate of NiO is slower on industrial nickel than on pure nickel. Simultaneously gaseous diffusion of oxygen occurs and ensures the formation of the inner layer. As a consequence, a duplex scale is formed. Under the oxide scale, internal oxidation, especially along grain-boundaries, extends on distances greater than the oxide scale thickness. The duplex structure of the NiO films is related to the great affinity for oxygen of elements present in the nickel substrate as precipitates (Mn, Mg, Ti, Si). These elements react with oxygen inducing, in the first oxidation stages, an outward flow and extrusion of nickel along grain boundaries and subboundaries on account of the stresses created by the volume increase. Acknowledgements Thanks to C. S´ev´erac for XPS experiments and to LPCES (laboratoire de Physico-chimie de l’Etat Solide), Orsay, for optical interferometry. References [1] A. Atkinson, Rev. Modern Phys. 57 (2) (1985) 437–470. [2] P. Kofstad, High Temperature Corrosion, Elsevier Applied Science, London, 1988.

[3] R. P´eraldi, D. Monceau, B. Pieraggi, Oxid. Met. 58 (3/4) (2002), 249–273 and 275–295. [4] R. P´eraldi, Ph.D. Thesis, INP Toulouse, France, 2000. [5] L. Gaillet, Ph.D. Thesis, Universit´e de Technologie de Compi`egne, France, 2000. [6] A.M. Huntz, J. Mater. Sci. Lett. 18 (1999) 1981–1984. [7] A.M. Huntz, B. Lefevre, F. Cassino, Mater. Sci. Eng. A 290 (2000) 190–197. [8] O. Bernard, G. Amiri, C. Haut, B. Feltz, A.M. Huntz, M. Andrieux, Mater. Sci. Eng. A 335 (2002) 32–42. [9] J. Balmain, C. Savall, R. Molins, C. S´ev´erac, C. Haut, A.M. Huntz, Mater. Sci. Forum 369–372, Trans. Tech. Publications, 2001, pp. 125–132. [10] B. Chirat, C. Haut, A.M. Huntz, M. Andrieux, B. Feltz, J. Phys. IV France 106 (2003) 151–160. [11] A.M. Huntz, A. Lefeuvre, M. Andrieux, C. S´ev´erac, G. Moulin, R. Molins, F. Jomard, Mater. High Temperatures 20 (4) (2003). [12] W. Przybilla, M. Sch¨utze, Oxid. Met. 58 (1/2) (2002) 103–145. [13] S. Perusin, B. Viguier, J.C. Salabura, D. Oquab, E. Andrieu, Mater. Sci. Eng. A 387–389 (2004) 763–767. [14] S. Perusin, B. Viguier, D. Monceau, E. Andrieu, Mater. Sci. Forum 461–464, Trans. Tech. Publications, 2004, pp. 123–130. [15] R. Haugsrud, Corros. Sci. 45 (2003) 1289–1311. [16] F. Pedraza, M. Reffass, J. Balmain, G. Bonnet, J.F. Dinhut, Mater. Sci. Eng. A357 (2003) 355–364. [17] M.G. Graham, G.I. Sproule, D. Caplan, M. Cohen, J. Electrochem. Soc. 119 (1972) 883–1205. [18] R. Molins, A.M. Huntz, Mater. Sci. Forum. 461–464 (2004) 29–36. [19] M. Parise, O. Sicardy, G. Cailletaud, J. Nucl. Mater. 256 (1998) 35–46. [20] E.W. Hart, Acta Metall. 5 (1957) 597–598. [21] A.M. Huntz, G. Calvarin Amiri, H.E. Evans, G. Cailletaud, Oxid. Met. 57 (5/6) (2002) 499–521. [22] A.M. Huntz, V. Bague, G. Beaupl´e, C. Haut, C. S´ev´erac, P. Lecour, X. Longaygue, F. Ropital, Appl. Surf. Sci. 207 (2003) 255–275. [23] Landoklt-B¨ornstein, in: O. Madelung, H. Mehrer (eds.), Diffusion in Solid Metals and Alloys, New Series III/26, Springer-Verlag, Berlin, 1990.