Study of stress effects in the oxidation of phosphated α-iron: in situ measurement by diffraction of synchrotron radiation

Study of stress effects in the oxidation of phosphated α-iron: in situ measurement by diffraction of synchrotron radiation

Applied Surface Science 206 (2003) 149±158 Study of stress effects in the oxidation of phosphated a-iron: in situ measurement by diffraction of synch...

182KB Sizes 0 Downloads 3 Views

Applied Surface Science 206 (2003) 149±158

Study of stress effects in the oxidation of phosphated a-iron: in situ measurement by diffraction of synchrotron radiation B. Panicauda, J.L. Grosseau-Poussarda,*, P.O. Renaultb, J.F. Dinhuta, D. ThiaudieÁrec, M. Gailhanouc a

LEMMA, PoÃle Sciences et Technologie, Universite de La Rochelle, Av. M. CreÂpeau, 17042 La Rochelle CeÂdex 1, France b LMP-CNRS, Universite de Poitiers, SP2MI, 86960 Futuroscope CeÂdex, France c LURE, BaÃt 209D, Universite Paris-Sud, BP 34, 91898 Orsay, France Received 8 July 2002; accepted 14 October 2002

Abstract In the present work the development of strains in the iron-oxide layers growing on phosphated a-Fe at 400 8C in arti®cial air at 1 atm was investigated by X-ray diffraction (XRD) of synchrotron radiation, both in situ during oxide growth and also at room temperature after cooling. This stress state may play an important role during the oxidation process, from the phases as well as from the kinetics point of view. The oxidation kinetics of a-Fe and phosphated a-Fe at 400 8C and the phases evolution during oxidation are ®rst presented. Then a detailed study of the strains in the oxides layers is undertaken. Correlations between the stresses measurements and the successive parabolic oxidation stages for phosphated a-iron are established. It leads to a better understanding of the oxidation behaviour. # 2002 Elsevier Science B.V. All rights reserved. Keywords: High temperature oxidation; Iron; Phosphate layer; Residual stresses; Kinetics; X-ray diffraction

1. Introduction In this paper we describe stress measurements during the oxidation of phosphated a-iron. Indeed it has been shown [1,2] that a phosphating treatment induces a signi®cant improvement of the oxidation resistance of raw iron in the temperature range (350±700 8C). The phosphate layer acts as a protective barrier that slows down the oxidation process. In order to better understand the eventual relationship between


Corresponding author. Tel.: ‡33-5-46-45-86-12; fax: ‡33-5-46-45-72-72. E-mail address: [email protected] (J.L. Grosseau-Poussard).

oxide phases, oxidation kinetics and stresses, an in situ diffraction study has been undertaken. Relatively few published researches dealt with the experimental determination of residual strains in ironoxide layers generated by oxidation of iron [3±7]. In most cases the bent-strip method was used to estimate the sign and the magnitude of the growth strain developing in the surface part of the oxidising specimen. The advantage of using X-ray diffraction (XRD) is that it allows distinction of strains in the different iron-oxide phases. For the ®rst stage of in situ internal stress determination, the classical sin2 C method is not always applicable because of the limited amount of material available for diffraction, giving rise to a very weak signal. Thus the high intensity of the X-ray beam

0169-4332/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 4 3 3 2 ( 0 2 ) 0 1 1 9 3 - 5


B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158

available at the French synchrotron (LURE-Beamline H10) has been used to perform stress measurements for oxide thickness as weak as possible. Moreover, in situ and post-mortem stress measurements are compared. This allows to distinguish the growth stresses in the oxides phases from the thermal stresses introduced on cooling, assuming that no relaxation process occur on cooling. 2. Experimental procedure 2.1. Sample preparation The phosphated specimens were prepared from a 1 mm thick sheet of polycrystalline a-Fe produced by Goodfellow, with a chemical purity of 99.5%. Details on the phosphating process are given in [1]. The specimen surfaces were mechanically polished with silicon carbon down to 400 grid and with diamond paste down to 1 mm giving a mean roughness better than 0.05 mm (mean roughness average). 2.2. Characterisation resources 2.2.1. Thermogravimetric analysis (TGA) The oxidation kinetics for both uncoated bulk a-iron and phosphated a-iron were investigated by thermogravimetric analysis in a Setaram TGA 92 thermobalance. The sensitivity of the balance is 1 mg. Samples were heated at 50 8C min 1 up to the desired oxidation temperature under an argon atmosphere. The selected temperature is thought to be stable after 1000 s within 0.1 8C. Then oxidation is performed under arti®cial air (20% O2; 80% N2) at a gas ¯ow in the thermobalance of 0.6 l h 1. 2.2.2. Microstructural evolution and stresses determination The high temperature XRD equipment used in this study consists of a high temperature chamber with an O-setting goniometer equipped with a position sensitive detector. It has been developed in the LURE laboratory (H10 beam line at the French synchrotron) and it allows to follow the evolution of stresses as a function of temperature as well as the nature of the oxide scales that form during the high temperature experiments. The wavelength has been ®xed at

0.190335 nm. For X-ray stress analyses, the sin2 C method was used [8]. Using at least 6C values, it takes about 1 h to determine a stress value. The so-obtained value is thus an average value during the measurement. In order to select the diffraction conditions, several parameters have to be taken into account: the selected (hkl) peaks must be isolated from peaks due to other phases, the corresponding Bragg angles must be as high as possible to obtain good accuracy and must have suf®cient intensity. One dif®culty consists of the very weak diffraction peak of oxides phases due to their small thickness, especially at the beginning of oxidation. It was also veri®ed that the oxide layers were not textured. Depending on the oxide phases, only two planes were selected. The (2 2 6) plane for hematite, for which 2y ˆ 119:47 and the (7 3 1) for magnetite for which 2y ˆ 120:68. The diffraction-line pro®les measured for the determination of lattice strains were corrected for a linear background by a ®t through the extremities of the pro®le. The peak maximum was obtained from the top of a Gaussian or Lorentzian pro®le. The best ®t obtained by a leastsquares procedure to the data in the peak region of the pro®le, was retained. Small errors in peak position due to defocusing of the goniometer were corrected for by calibration of the 2y and the o scales with strain-free well-de®ned powder standards. The maximum average experimental error in the data presented here for the stress in the plane of the oxide layer was estimated to be Ds ˆ 60 MPa. 3. Results and analyses 3.1. Oxidation kinetics The weight change curves (Dm/S) as a function of the exposure time are shown in Fig. 1 for both a-iron in the as polished state and in the phosphated state. From Fig. 1 it clearly appears that at 400 8C the mass gain is less in phosphated than in uncoated iron and even after 48 h of oxidation the difference in mass gain between the phosphated and the untreated specimen is always noticeable as a consequence of the coating ef®ciency. Moreover it appears from Fig. 2, in which the square of the weight gain due to oxygen uptake by oxidation is given as a function of the oxidation time, that the phosphated specimens present two stages

B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158


Fig. 1. Weight gain Dm per unit of surface S (Dm/S) of untreated and phosphated a-iron samples due to oxidation at 400 8C, in arti®cial air, as a function of the oxidation time.

Fig. 2. Square of the weight gain (Dm/S)2 due to oxidation at 400 8C, in arti®cial air, as a function of the oxidation time for phosphated a-iron sample.


B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158

(stages I and II) of parabolic oxidation. The transition from stages I to II seems to occur between 600 and 1100 min. The slopes for stage I is much smaller as compared with the slope in stage II indicating that the kinetics for stage I is reduced. It appears that stage I could be considered as slowing down the oxidation process. The parabolic-rate constants kp are determined from the slopes of the curves in Fig. 2. In stage I, kp ˆ 8  10 14 g2 cm 4 s 1 and in stage II, kp ˆ 2:5  10 13 g2 cm 4 s 1. It is to be mentioned that during the oxidation at 400 8C of a polycrystalline etched a-iron, Jutte et al. [9] also observe two successive stages in their oxidation kinetics. Moreover, the parabolic constant corresponding to their second stage is between 1.5 and 2.10 13 g2 cm 4 s 1 in agreement with the present kp value obtained in the stage II for the phosphated specimen. On the other hand, a kp value has not been deduced in [9] for the ®rst stage of oxidation. Taking into account the results from the ®rst 100 min of oxidation at 400 8C in that reference, we have deduced a kinetic constant kp ˆ 4:710 14 g2 cm 4 s 1, which also compares favourably with the present value in stage I. Jutte et al. explain that the etching pretreatment of the metallic surface results in the development of an hydroxide ®lm prior to the oxidation. It appears that such a super®cial hydroxide layer could play the same role that the phosphate coating in the present study. 3.2. Microstructural evolution of the oxide layers XRD patterns recorded on untreated and phosphated iron samples oxidised at 400 8C for 2880 min are presented in Fig. 3. It has been obtained after cooling at the end of the oxidation process (Cu Ka radiation). It appears that both oxides magnetite (M; Fe3O4) and hematite (H; a-Fe2O3) are present in the two spectra. From the observation of the peak intensities for each specimen, a signi®cant distinction can be outlined: after 2880 min, the presence of Fe3O4 on untreated iron is dominant while on the phosphated iron, a-Fe2O3 is the main phase. As already observed in the oxidation kinetic study the oxide layer grown on the phosphated sample is thinner than the oxide layer grown on the untreated specimen. Indeed from the examination of the a-Fe lines in the scans, its obvious that the peak intensities are smaller on the untreated

specimen compared with the phosphated one. See for example, the 1 1 0 peak of a-Fe at 2y  458. It corresponds to an increasing absorption of the X-rays through the thickest oxide layer for the untreated specimen. From the in situ synchrotron diffraction study, a more detailed analysis of the spectra has been done: taking into account the theoretical non-normalised intensity for a-Fe2O3 and Fe3O4 and as the absorption coef®cient and the density are about the same for both oxides, the proportion's evolution of hematite and magnetite can be deduced from the ratio of the integrated intensity of the diffraction peaks. It has been done for two diffraction lines: (1 0 4) for hematite and (2 2 2) for magnetite. It appears that for the untreated specimen (Fig. 4a), the magnetite Fe3O4 is the dominant phase whatever the oxidation time and its proportion increases continuously during oxidation. On the other hand, for phosphated iron (Fig. 4b) hematite is the dominant phase. But from 150 min upon oxidation until about 700 min its proportion increases slowly and then an abrupt inversion is clearly observed which indicates that the proportion of magnetite starts to increase. The transition observed around 700 min of oxidation has to be compared with the kinetic transition already mentioned: so on the untreated sample mainly grains of Fe3O4 are present while on the samples which were phosphated prior oxidation, aFe2O3 is the dominant phase. This observation is in agreement with the study on bulk a-iron [9] in which it has been established that a-Fe2O3 ®rst develops before Fe3O4 occurs. Moreover, the development of Fe3O4 corresponds to an increasing oxidation kinetic. In that case [9] the initial etching of the metallic surfaces leads to a complex oxide ®lm constituted of iron hydroxides that will transform during oxidation into a-Fe2O3. So in the present study the phosphate layer, in the ®rst part of the oxidation may also transform preferably to a-Fe2O3 or lead to the preferential growth of a-Fe2O3 as is suggested by the present X-ray diffractogram. This process should correspond to the stage I observed in the oxidation kinetic. Then when the stage II appears, it could be associated with the preferential development of magnetite with a corresponding increasing kinetic. This assumption agrees with the fact that magnetite support a larger ¯ux of rate controlling species and thus growths faster than hematite [10]. The very good agreement between the

B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158


Fig. 3. Diffractogram (Cu Ka radiation) showing the magnetite (M; Fe3O4) and the hematite (H; a-Fe2O3) line pro®les for (a) untreated a-iron sample and (b) phosphated a-iron sample, oxidised during 48 h at 400 8C in arti®cial air. Note that only the main diffraction lines have been indexed.


B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158

Fig. 4. Phases evolution percentage in hematite (1 0 4) and magnetite (2 2 2) during oxidation at 400 8C in arti®cial air for (a) untreated sample and (b) phosphated sample.

B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158

kp values obtained in the stages I and II in the present study with the values obtained by Jutte et al. [9] indicates that the phosphate layer likely acts as the hydroxide layer. Summarising, it has been established that during the stage I an hematite layer growths preferentially on the phosphated a-iron. Then during stage II a magnetite layer develops preferentially. In order to better understand the kinetic transition, the evolution of the stresses during the growth of the oxide layer for phosphated a-iron is now investigated. 3.3. In situ stress measurement during oxidation First of all, it is assumed that linear elastic stress/ strain conditions occur. Moreover, assuming an inplane isotropic stress in the plane of the oxide layers (s11 ˆ s22 ˆ s and s33 ˆ 0), the sin2 C relation becomes  hkl    d sin y0 1 ejc ˆ ln hkl ˆ ln ˆ S2 s sin2 c ‡ 2S1 s 2 sin y d0 where dhkl is the hkl lattice spacing determined in a direction characterised by the angle C with respect to the surface normal, d0 the strain-free lattice spacing, e the strain and S1 and (1/2)S2 are the X-ray elastic constants. Values for S1 and (1/2)S2 of the oxides scales under consideration were calculated from the corresponding single crystal elastic constants, using the Reuss approximation [11±13] yielding for the analysis using the Fe3O4(7 3 1) and the a-Fe2O3(2 2 6) re¯ections: Fe3O4 a-Fe2O3

S1(7 3 1) ˆ 1.21  10 6 MPa 1 S1(2 2 6) ˆ 0.756  10 6 MPa 1

(1/2)S2(7 3 1) ˆ 5.69  10 6 MPa 1 (1/2)S2(2 2 6) ˆ 5.58  10 6 MPa 1

The value for the stress parallel to the surface, s can be obtained from the slope of the straight line obtained by least-squares ®tting through the experimental data in a plot of ln(1/sin y) versus sin2 C. The strain at the oxidation temperature can be obtained by subtracting the thermal expansion of the oxide phase from the total strain. e ˆ etotal ˆ

a… 13

ethermal expansion DT†


ethermal expansion (1)


where a is the temperature-dependent linear expansion coef®cients of oxide layers and DT is the temperature change. Fig. 5 shows ln(1/sin y) versus sin2 C for the phosphated-iron oxidised for 1200 min in (a) hematite and (b) magnetite. The linear behaviour is well observed and the plot exhibit negative slope which is characteristic of a compressive strain/stress state. Growth stresses were also determined during isothermal oxidation at 400 8C both in magnetite and hematite. The results obtained are collected in Fig. 6. Thermal expansion effects on the strain measurements have been removed according to Eq. (1) in order to verify that the strain-free lattice spacing d0 is in agreement with the JCPDS data for a-Fe2O3 and Fe3O4. It has to be noticed that for shorter oxidation time than 700 min it was not possible to determine growth stresses/strains in Fe3O4 owing to very faint and broad diffraction peaks. The following conclusions can be drawn on the basis of the results of Fig. 6.  A large compressive growth stress/strain occurs in a-Fe2O3. Its absolute value increases from 480 to 720 MPa on continued oxidation (between 160 and 300 min). Then it remains in the range 700±760 MPa until about 700 min, and finally it decreases from 700 to 335 MPa after 1100 min of oxidation. It has to be noticed that the subsequent cooling induces an additional compressive stress of about 300 MPa.  A small compressive growth stress/strain occurs in Fe3O4 on phosphated a-Fe after 700 min of oxidation and it remains about constant until the end of the oxidation process. A thermal stress arises also on cooling.  The observation of a compressive strain/stress state in the oxide layer is in agreement with the already mentioned work of [9,14] for which hematite was also the dominant phase at the beginning of the oxidation. They suggest that the oxide growth is rate controlled by short circuit diffusion of iron cation and therefore the formation of oxide along the short circuit (e.g. grain boundaries, microcracks) could be partly responsible for the compressive nature of the growth strains in hematite.  The observed increasing value of the compressive stress in hematite during the first 300 min of oxidation and its subsequent stabilisation at, about


B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158

Fig. 5. Evolution of the sin2 C curves for (a) the hematite phase and (b) the magnetite phase obtained during oxidation of phosphated a-iron. The straight full lines represent the linear regressions on the experimental data.

750 MPa during 400 min, may be explained as follows: as hematite is the dominant phase at the beginning of oxidation, its continual growth upon oxidation is associated with an increasing compressive strain. Then after 300 min a steady-state with a maximum compressive stress is attained which is maintained for about 400 min. Nevertheless after 1100 min, the compressive stress in hematite has been significantly reduced.

 Compared with the evolution of the oxidation kinetics and oxide phases at 400 8C, it obviously appears that the transition region between stages I and II from 600 to 1100 min, overlaps both the steady-state and sharp decrease regime of stress evolution. Moreover the abrupt increasing proportion of magnetite in the second oxidation stage is clearly correlated with the sharp decrease regime of stress evolution in hematite. The majority growth of

B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158


Fig. 6. Evolution of the stresses in the oxide layer vs. oxidation time for (a) the hematite phase and (b) the magnetite phase. Note the additional compressive stress induced by cooling at the end of the oxidation process.

magnetite in the oxide layer from 700 min upon oxidation should be able to relax the stress in the initial hematite layer. 3.4. Calculation of thermal stresses As in the case of phosphated a-iron an hematite layer ®rst preferentially develops and then a subsequent magnetite phase appears, thus after 1200 min

upon oxidation the scale is mainly constituted of two partial layers of respectively hematite and magnetite which upon cooling induce an additional stress component sthermal, owing to the oxide/metal thermal expansion mismatch. When a scale consists of several partial layers and if the interface roughness is ¯at, the stress si in each partial layer can be calculated from si ˆ … Ei …am ai †=…1 ni †‰…1 ai †DTŠ†DT [15], where index i stands for the parameters of the


B. Panicaud et al. / Applied Surface Science 206 (2003) 149±158

Table 1 Experimental and calculated thermal stresses in hematite and magnetite phases Thermal stresses/oxide phases a-Fe2O3 Fe3O4

sexp (MPa) 279 150

scal (MPa) 319 46

individual partial layer and am is the mean coef®cient of thermal expansion of the scale that results from Pn …di Ei ai =…1 ni †‰…1 ai †DTŠ† am ˆ Pi n ni †‰…1 ai †DTŠ† i …di Ei =…1 In our case the appropriate thermal expansivities and elastic constants for a-Fe, a-Fe2O3 and Fe3O4 have been deduced from [14,16]. Thus the thermal stress in each layer can be calculated and compared with the experimental value. In this aim the thickness of each partial oxide layer di is necessary. The kinetic evolution of Fig. 1 allows to estimate the total oxide thickness after 1200 min upon oxidation, and the proportion of each oxide phase at this time can be deduced from Fig. 4. It leads to an oxide scale which is constituted of 0.4 mm of hematite and 0.15 mm of magnetite. Experimentally, the thermal stress is obtained from the difference of the stresses at 400 8C and at room temperature, assuming that the relaxation process plays a small role on cooling. The calculated scal and measured sexp stresses values are reported in Table 1. It appears that a good agreement is observed between experimental and calculated values for hematite. On the other hand, a signi®cant difference is obtained for magnetite. Nevertheless, for both oxide phases the calculated thermal stress range order is good compared with the experimental values. 4. Conclusions In this work the in¯uence of a phosphating treatment on the oxidation of a-iron was studied. The main conclusions can be summarised as follows.  At 400 8C, the mass gain is less in phosphated a-iron than in untreated iron showing the protective character of the phosphate treatment. Two parabolic stages occur during the oxidation of phosphated a-iron.  Both phases magnetite and hematite are present in the oxide layer. Hematite is the dominant phase

on phosphated a-iron while magnetite develops preferentially on untreated iron.  For phosphated a-iron a large compressive growth stress occurs in a-Fe2O3 and its value changes during oxidation. A small steady compressive growth stress is measured in Fe3O4 from the beginning of the stage II until the end of the oxidation process.  For phosphated a-iron an important correlation has been established between oxidation and growth stresses evolution. A transition has been observed at about the same time in the oxidation kinetics, the oxide phase evolution and the growth stresses evolution.  A multilayer model has been used to calculate thermal stresses and is quite correct. Errors are likely due to restrictive hypothesis on the model, such as adherent or flat layers, which is not necessary the case here. Moreover the relaxation process which may occur on cooling should also be considered. References [1] S. Rebeyrat, J.L. Grosseau-Poussard, J.F. Dinhut, P.O. Renault, Thin Solid Films 379 (2000) 139±146. [2] S. Rebeyrat, J.L. Grosseau-Poussard, J.F. Silvain, B. Panicaud, J.F. Dinhut, Appl. Surf. Sci. 7932 (2002) 1±11. [3] S. Taniguchi, D.L. Carpenter, Trans. ISIJ 18 (1978) 523. [4] S. Taniguchi, D.L. Carpenter, Trans. ISIJ 18 (1978) 530. [5] P.B. Abel, A.H. Heuer, R.W. Hoffman, J. Vac. Sci. Tech. A 1 (1983) 260. [6] P.B. Abel, R.W. Hoffman, J. Vac. Sci. Tech. A 4 (1986) 2938. [7] S. Taniguchi, T. Shibata, M. Murahoshi, Corr. Eng. 36 (1987) 295. [8] G. Maeder, J.L. Lebrun, in: D. David, Caplain (Eds.), MeÂthodes Usuelles de CaracteÂrisation des Surfaces, Eyrolles, 1988, p. 248. [9] R.H. Jutte, B.J. Kooi, M.A.J. Somers, E.J. Mittemeijer, Oxide Metals 48 (1/2) (1997) 87. [10] P. Sarrazin, A. Galerie, J. Fouletier, Les MeÂcanismes de la Corrosion SeÁche, EDP Sciences, 2000. [11] C. Sariogliu, J.R. Blachere, F.S. Pettit, G.H. Meir, in: S.B. Newcomb, J.A. Little (Eds.), Proceedings of the Third International Conference on Microscopy of Oxidation 3, Trinity Hall, The Institute of Materials, 1997. [12] K. Messaoudi, A.M. Huntz, L. Di Menza, Oxide Metals 53 (1/2) (2000) 49. [13] M. van Leeuwen, J.D. Kamminga, E.J. Mittemeijer, J. Appl. Phys. 86 (4) (1999) 1904. [14] B.J. Kooi, M.A.J. Somers, R.H. Jutte, E.J. Mittemeijer, Oxide Metals 48 (1/2) (1997) 111. [15] M. Schulte, M. SchuÈtze, Oxide Metals 51 (1/2) (1999) 55. [16] J. Robertson, M.I. Manning, Mater. Sci. Technol. 6 (1990) 81.